#Arcademuseum Seligenstadt Repair Logs

Google Translate still sucks big time - I tried putting my arcadefix.txt
through it, you couldn't understand it.
If you put "Let's see if Google translate still sucks" through it and 
translate it back to English, you get "Let's see if Google still vacuums
translating".

So, let's just manually translate this text, since I'm definitely better at it
than Google Translate.

Let's start with the logic graveyards (machines without a CPU).

Breakout (Cocktail table)
First I got a replacement for the exploded capacitor in the monitor and put it
in. It takes half a minute to get a picture and another 15 minutes until it's
sharp.
Symptom: No bricks. Player 2 can't control the paddle.
If you start a game, the ball will deflect off the invisible bricks of player
2!
Suspicion: SRAM. Output is fixed high, R/W is stuck low (permanently writing
data, must not be), data is near low (1/3 high) - invalid signal.
According to the schematics, the signal originate from a 7427 triple NOR.
The gate gets low, high, low, output should be low, instead, it adds the
inputs in an analog fashion. If you pull down the output with a resistor, the
bricks magically appear. This chip is probably dead (but it could be the SRAM
is trying to pull up the signal). I happened to just so have a spare IC, bent
up the outputs and capped it on the other one. The signals were correct and
the original IC had another gate with invalid output.
I just desoldered the IC and replaced it (socketed, of course). The bricks
appeared and the game was perfectly playable. At least for player 1.
Player 2 still couldn't move the paddle. If you turned off cocktail mode, you
could control player 2 with player 1's potentiometer. Something's wrong with
the pot, pot's power supply or the pot switchover. Let's ring through the
wires (using a continuity tester) where pot 2 goes. Hm, doesn't help.
According to manual there's a 4066 used to switch between player 1 and 2.
Measured it, it's broken. But only 2 of the 4 gates are used, so I just
bridged the gate over to another one and the machine was fully restored.
Oh wait - not fully: The score of the player up should blink. It doesn't, not
even if you turn it to Upright mode. (but the game is perfectly playable and
I don't give a damn).
The extra credit circuitry also works. Once I got 400 points, it beeped and
Player 1 Start lit up. When I was Game Over, only 6 bricks were left.
I'm gooood.
After about a year, the paddles started wobbling up to unplayable. Bending and
pushing on every connection and IC showed that it got better when I pushed on
the known defective 4066. Well, I thought, this one bit the bucket completely.
Swapped it, barely any difference. Damn. But now it seems to lie only within
the contacts. Cleaned'em, no dice. Removed that funny RF shield card and
directly hooked up the edge connector to the game board fixed it.
But still it likes to wobble, by that I mean the AC going into the board is
missing every second half wave sometimes. Jiggling the connector brings it
back, but it uses to immediately go bad as soon as you close the machine.
So I soldered in a Molex connector to bypass the edge connector and that
works and you still can remove the board if you want.
After like 2 years I got bored and wanted to know why the Player Up score
doesn't flash. Only 3 ICs could be responsible for that and they are a 555,
an LS04 and an LS27. Of course, it had to be the last one. And the blink
output wasn't the only one that didn't work.
After turning on, the picture is wobbling pretty badly for a while.
There's probably another capacitor on its way out (the other machine's filter
cap died long ago)
And yeah... after the machine stood unused for a few months due to Corona, the
picture is dim and wobbly. And the capacitor doesn't look healthy, but up to
10V it works almost passable (with some tens of Ohms ESR...).
Well I put a new one in and the pic's back to normal.

Oh cool - in the meantime, the other Breakout board surfaced...
(we have a Breakout upright with a very poor disfunctional MAME conversion).
With the cocktail table as a testing station I thought - Let's test!
Turned it on - everything there. Bricks, ball, attract mode ran fine, great!
I put in a credit, nothing happens. Doesn't coin up.
Oookay schematic diagram....yesss the inputs go into these hex inverters
someone ham-fistedly soldered over the old sawed-off ones. Away with it, new
ones in (socketed...). Well, it coins up now. But doesn't start a game if you
push any start button. What's the scope say? Jup, the signals arrive. And go
into some flipflop for some reason. The flipflop swallows the signals instead
of buffering them and creating an edge. Hm. Invalid signal on the clock pins.
Bent up the output pin of the chip that sends the signal (which I suspected
already because the picture was always upside down no matter cocktail mode or
player up). Signal's OK now. Removed flipflops, Signal OK. Replaced flipflops
with factory new, signal's crap. Weird. Oh wait now he doesn't coin up again?
What's the problem? Looking for a reason for 3 hours, I found that a 74193,
which stores the credits, immediately gets a reset signal as soon as you
push the coin button, where does that come from? From the antenna circuit of
the coin checking mechanism (so you can't get credits with a piezo lighter).
OK I unplug this pin and leave it floating (easy because I socketed this chip
earlier on). What, the machine's on Free Play now? (unplugged pin 6 of the
right 7404 when the edge connector points away from you)
So why can't I start a game? I don't understand, I'm fed up. The other machine
didn't took half as long to fix!
So what happens when you pull down the flipflop's outputs? The game starts up.
And stays running. So - removed the flipflops and directly wired the Start
button inputs to the flipflop's outputs. Not quite kosher, but it works.
When I put the board back into the original machine I could see that it used
double throw (I think - I mean ON-ON) switches for coins instead of that
antenna thingy.
Well at least I don't need a credit button.
After 5 minutes, the exact same capacitor blew up in the monitor.
(same model monitor)
I got bored (a few weeks before fixing the score flash on the other Breakout)
and wanted to see what happens if you swap the suspicious 7408 (the one that
drives the Cocktail mode as well as the Ingame Flipflops I worked around).
Well, it was bad and swapping it fixed the Game circuit.
But for some reason, the machine is not on Free Play anymore. OK put in a
credit button, that had to happen, saw that coming.
And a month later the machine was eating credits again. Back to square one!
Sunk 4 more hours into the machine, checking lots of ICs with my fancy new
In Circuit IC tester and Debugger with no results. There is a discrete
transistor Circuit named Q (great name) that keeps its state after giving a
credit in the working machine and falling back in the bad one. Socketing the
credit counter IC (that LS193) so I could prevent it from resetting didn't
work at all, the machine seems to have caught me cheating and had
countermeasures installed. Since I need to push the credit button quite often
and the fact that there's a live CRT in the way (I could turn it off) doesn't
help and there's no one around who can push the button for me, so I put it in
the Cocktail table. And left it in there since here it works somewhat.
If you push the credit button, the game resets. But you'll have two credits.
So it doesm't count to 15 but at least you can play both machines again.
Huh now the paddles are wobbling horribly, almost unplayable.
That big capacitor does look like it's blown up before... let's put a new one
in. Still some wobble left, but way more playable. And hey I can throw in
credits like mad without the game resetting.
But after half a year there's no more sound. Not even crackling when I turn
the speaker potentiometer. Reason being a via flew away, so the 12V rail
didn't reach the audio amplifier.

Road Champions (Model Racing version, although it's pretty certainly identical
to the Taito version - both are rare):
Symptom: Screen turns on and immediately off again. If it stays on, you get a
solid dark gray screen.
When the screen turns off, it sounds more like a kind of standby mode.
The machine then doesn't put out a video signal, just a solid 1V...
Gotta mention the moron who put the machine together swapped signal and ground
(gotta protect the ground level from interference). Duuuh.
Optical inspection shows: Almost all ICs are extremely corroded. Plus side:
Almost all ICs are socketed (although the sockets are corroded, too...
spoiler alert: Reseating the IC usually wards off whatever problem the machine
comes up with now). The only chips that aren't socketed are on the mainboard
and the audio card (main board huh, it's just a bus card, there's 4 ICs on
this).
Voltage regulator delivers clean 5V. Each of the 4 circuit boards draws around
1 to 2 amps, so nothing's completely busted. A real problem is the 2 boards in
the middle aren't accessibe to oscilloscope investigation.
Where do I start? When just nothing shows up on screen, on logic graveyards
you should check the address generator first. The address generator generates
the signals which determine the current position of the cathode ray.
Using logical equations, the machine determines whether a pixel should be
black or white.
In the manual (so it exists, thank god yes) you should keep the eyes open for
HSYNC, VSYNC or signals like 1H 2H 4H 8H 16H and so on until 128H, then 1V 2V
...128V. These determine the position of the cathode ray and the ENTIRE rest
of the game depends on these signals.
So if you randomly probe such a machine and nothing appears to happen, the
address generator is most likely at fault.
If you don't have a manual or schematics, look out for a quartz.
It's most of the time hooked up to a 7404 or 74F04. Check if there's a clean
oscillation going on (doesn't mean to look perfect, but amplitude and 
frequency should match somewhat. The address generator should get a somewhat
clean signal). Well that was the case. Lots of ringing, but stable and good
amplitude. If you follow the signal through the address generator ICs
(Atari likes to use 9316=74161, Road Champions uses 7493), you see a signal
of each half the frequency on each pin (most the time, these chips have 4 bit
and are preloadable so if you see less than half the pins toggling, that's ok)
The signal disappeared into the first vertical IC. No output. NOTHING.
Swapping the 2nd vertical IC with the first yielded 4 more signals with each
half the frequency than the one before. This chip's DEAD! Great. Checked 4 ICs
and already one is broken - this machine's got great chances of making lots of
trouble (spoiler alert: yes.)
Ordered replacement (74xx93 are hard to get - had to get HC) put it in and -
dark gray screen. But sync! And on the lower screen border: "ROAD CHAMPIONS"
nicely pixelated black on white (bright white) together with some random
symbols for time and points. The caption disappears after a few seconds only
to slowly be replaced letter by letter.
Well at least something. I leave it running and go over to Steeplechase for
which I also have bought spare parts.
After an hour I return and all the sudden Swoosh - the road appears. And 4
"computer" cars running on it. And not moving. Except very slowly back and
forth a few pixels. At least the road appears to move even though it just goes
straight and doesn't have curves. Enemy car movement matches a Youtube video
i've seen.
I give a credit, nothing happens. I swap a couple of chips in the RACE circuit
(behind the COIN Circuit which is on the "main"board) which determines if the
game is running or attract mode. No dice.
I fetch a colleague who gives credits on the far side of the machine while I
measure the signals on this side of the machine. Result: The coin pulses enter
the first chip, but garbage out. Either this chip is broken or one of the
others pulls it down. As mentioned, the 4 COIN chips are with the only chips
in the machine that aren't socketed - you assholes!
So... removing all the modules, all the plugs, torture myself through 7 of
these plastic holders, socket all the chips and bend up the output pin.
Result: High with garbage on top. Close the coin switch, the signal goes to a
clean low. The next chip OUGHTA work with that. Put the pin back in and the
machine starts up on the next credit. I don't understand it. Fixed by 
socketing. Well it's working now. I push the accelerator, turn the wheel,
where's my car? I switch to second gear and it appears from past the bottom of
the screen and crashes. After succeeding to get the car on the road I could
play. Meh. So so. Not bad for 1975, but when Night Driver came out in '76,
this game was ripe for the scrapyard. (the road wobbles normally when you
play).
With the colleague I adjusted the car's Y position so you spawn directly
above the scorebox instead of far below the screen, also adjusted the timer so
a game takes more than 30 seconds.
What more's to do: the game never returns to attract mode after playing,
Beginner's course never resets (and easy mode is booooring) and the score has
a garbage character in it. Swapped chips and bent up a pin, pulling it low,
hoping no one manages more than 799 points until I get a replacement.
But when you get more than 400 points, the game goes to FINAL LAP at the end
of which you get the finish line. Sometimes it resets easy mode but it also
turns it back on when you give another credit...
The open saturday the inverted output of the car chip died, leaving the
player's car dark gray instead of white. Also the monitor is on the edge.
Replaced the score chip and the car chip and tweaked the 12V rail. Had only
9V and ripple. Swapped a resistor in the voltage regulator with the value 
that's in the manual. Result: 12V with LOTS of ripple. Changed capacitor. 
Less ripple, but not much. Put in a compromise resistor, no more ripple, but
it returned after a couple of hours (might be the rectifier).
Shortly after that the collision detection for offroad kicked the bucket.
Studying the schematics showed there is a chip that gets car1, !street and
cars2-5. Blindly swapping that chip fixed the problem. Had to clean the relais
(Gas/gear shift also quit working some time). This machine makes a lot of
trouble.
(What's missing now: clean 12V, that I find the broken trace that makes the
road vanish from time to time, Beginner's course, attract mode).
Managed to come in first. The machine plays a little tune!
Every 3 to 20 operating hours, another chip bites the dust.
I turned the power supply down to 220V. The ripple on the monitor and the 12V
supply are gone now.
What now: Tire squeal sound in sharp curves works intermittently. Doesn't coin
up all the time. Says I've won although I came in third. Beginner's race never
resets (unless you get to FINAL LAP). The squeal sound now works again and for
a long time, no chips have died. But the enemy cars are so far up the screen 
that you have trouble even reaching the last one. And I have trouble finding
someone to play the game so I can adjust it. My colleagues suck so bad at the
game that they crash before the enemies fall back.
So. After a couple of months of peace'n'quiet the following defect appeared:
The delay mechanism that draws the road and memorizes the curves as the road
scrolls down as you drive broke. At the same time the road markings flicker
funnily. The game now plays a little like Monaco GP, only harder.
I must say, it's a lot more fun in this broken mode! The track changes you!
After lots of hours of oscilloscope fun: No dice. The delay line doesn't get
!RECI signal (whatever that is). It should theoretically be generated every
VBLANK, but nothing comes out. The input signals don't seem to overlap.
For that I need a REAL 2channel scope. Swapping the chip didn't do nothing.
Weird. That the circuit boards are sandwiched doesn't help either.
At least the enemy cars move slowly up and down the screen. They should do
that only in attract mode. For the book: At 200 points, the leading car falls
back so it can ram you from behind at about 300 points (happens EVERY TIME!).
(enemy car Y position should be set so that you're first with 1 car length
after 200 points).
Well great now it doesn't work anymore at all. First the street fidgeted
around the street (as a whole), now it's gone for good. Swapping chips back
and forth didn't do anything. Gotta get a logic analyzer...
...or just buy every chip on the board (except for the 4 delay line chips,
it's illusory to get those) (the schematics mention a variant with RAM, but
that also uses unobtainium chips and I'm not speaking of the RAM itself).
!RECI means recirculate. The delays are 256*2 bits per chip, so 256*8 in 
total. Recirculate means the bits flowing out the bottom are getting stuffed
back into the top. The way I would have that wired up is that the thing stores
the X coordinate of the left edge of the street and gets read out once per
frame (every scan line is one coordinate). When the car moves 1 pixel, the
!RECI goes low for one clock cycle and a new coordinate gets pushed in.
Makes sense. The other coordinates slip down one scan line and the road
appears to move. Sadly, the machine seems to be implemetned completely
differently? No, I'm right.
Task: Bend up !RECI and pull it high. For that, I'm gonna need a road.
So I bought a bag of chips from Reichelt, basically the entire board.
I started with the LS123. I noticed a difference in behaviour when swapping
those so at least one of them is knackered. Turned machine back on, there's
the road (or something). It was fuzzy and looked... broken. But the track
markings aren't flashing as a whole. But the road still moves as a whole, but
now not soft anymore, position changes are instantly.
There's a ROM which stores the road and some weird electronics softening it
so curves look natural.
The !RECI-signal doesn't look half funny. High is at 3V, low at 2V. So I swap
the LS73 which generates that signal and the machine's fixed again.
The street doesn't disappear randomly when you flex the circuit boards,
wouldn't have thought the chip is to blame.
As a finishing touch I adjusted the enemy car's position so that with a lot of
skill you can finish first sometimes.
And one month later the monitor died. Worked-fine-yesterday-symptom.
No horizontal deflection (and thus no high voltage). Didn't have my tools, so
had to wait for next week. It appears to go into short-circuit protection.
Hm no, all voltages are present (except 130V which comes out the flyback).
Now the fuse blew. Put in new one, ugly noises, thanks. Let's see how far I
can get without the HOT. Well, there's a TCA511 monitor controller. Signals go
in, nothing comes out, chip gets awfully hot (I don't think the chip got very
warm when the monitor worked). Got no spares, so I bought one off eBay (the
seller only sold them one at a time at pricy shipping costs).
Well, the monitor works with the new chip, but takes a few seconds before it
even starts giving that typical high-pitched hum you know and love).
Might be that it bites the dust again soon. (spoiler: That didn't happen yet,
and it's many months past)
After some months the road width has the same problem as once the street.
You flex the board and the street has correct width again, let it go and it's
so narrow that it's unplayable. The pot is broken, but changing it didn't do
anything anyway. Well at the moment it works, let's see how long...
Another thing that broke is the Winner's Tune. It just plays the first note of
each sequence. And the collision detection isn't properly processed.
It screeches, but you can drive on merrily. I reseated the according chips and
that fixed it. I suspected a bad 7493 or 7442 in the music circuitry since the
music stalls on one note. But it was neither. It was the 7407 bus driver which
selects which resistor goes into the 555 timer. Weird. The last 2 notes of the
melody are different, but at least it plays.
And the machine still sometimes doesn't coin up. Swapped the four mainboard
chips. Didn't change much. Sometimes, sometimes not.
Well, most of the time it works now, but the finish line looks weird for quite
a while. Since instead of chequered there were just two lines, it looked like
a bad XOR gate and wouldn't you think there is an XOR in the finish line logic
which I swapped and the problem went away.
After quite a while something went bad and it took me a LONG time to figure it
out: The entire screen is light gray, cars are still visible, but no road and
no scorebox. If you look closely, the edges of the road are just barely
visible. The video mixer is three ICs, one transistor and a couple of 
resistors that are fed directly into the mix.
Fun fact: Remove any or ALL of these three ICs changes NOTHING.
After some desparation I built a video probe (an Idea I came up with around
middle 2017: Feed composite sync into a TV with SCART, hook fast switching to
5V over a 150-220 Ohms resistor and connect green or whatever color you like
to the IC pin you want to probe) and it turned out that a signal that was
toggling intensely on the scope would actually fill the entire screen so
instead of SCOREBOX it was !CSYNC. That lead me to the scorebox generator.
Also, removing that pin did nothing, but connecting it to +5V brought the
picture back without the scorebox, so I knew it was the scorebox generator.
Only 4 or 5 hours wasted here so far... the signal originates from a 9300
which the manual states can also be an LS195. Which I should have. 
There's two of them. Swapping both also brought the picture back with a
heavily distorted scorebox that only shows pixel garbage characters.
I need to organize my logic chips, there's so many different... LS193...192...
194...getting warmer...finally 195. Swapped the suspicious chip out and the
game works fine again.
Aww drat, the enemy cars are too far up again. The enemy car position pot
doesn't work anymore as planned, it shifts the entire road with it.
The pot on the enemy car generator board does nothing unless you touch the
board near it. No clue what the pot is for (and the manual is Italian),
but the relais switches it. Knowing these relais don't work relaisably (ha ha
chirp) I swapped it with the Game On relais and the error changed in that the
enemy cars still couldn't be properly adjusted but instead of them being
constantly near the top of the screen, they now come dowen far enough so you
can get first place, although that's really hard now and lining it up with
the finish line is near impossible. But not entirely.
We just got a huge batch of New Old Stock relais, maybe I should put two of
these in...
Aww crap dead again. After running for about half a year! Picture looks a lot
like back when the !RECI signal was missing/corrupted, but instead of the road
being one pixel it now repeats every 16 pixels. On the surface it looks like
one of the bucket brigades hit the bucket, but the road moves smoothly across
the screen and the distortions are dependant of your speed, so it might be
Clock or Recirculate.
Recirculate gets Clear Sync from A5,A6 / B5,B6.
Let's try the standard method first - jiggle everything around.
Do I even need to say that worked. A couple hours later the road was gone.
Well, you guessed it. The machine has been moved, that's why now it has all
these problems.
Now after 3 years the monitor's capacitors appear to have started going bad
(the ripple starts to reappear)
And one open saturday the enemy cars disappeared. Shook the machine to no
avail, at least I got more than 800 points (finally it paid off swapping that
defective chip), but still got PLACED BRAVO so the enemies must be above the
screen.
Opening up the machine brought the cars back... and yes it's possible to win.
We've been having the problem for a while now that collision detection only
works a bit - the squealing sound turns on, but nothing else happens, no
CRAAAAASH! and the car isn't set back to the bottom of the screen, also the
bash sound is missing. So I just went over every chip on the board with the
chip tester and the first chip I tested (a 7400) said one of the inputs was
shorted to 5V, I checked with the scope, yeah it was possible for the tester
to drive the line low, but it looked weird, so I swapped that chip.
Now it works. (I think it was S7, next to a relais)

Super Road Champions
I got the monitor almost up and running. RGB module still seems defective
(it was damaged because it was salvaged from a monitor that got vandalized and
the original RGB module as well as the MQ module were missing).
But not bad for something that wasn't in operation since the '80s.
Sadly, the game didn't survive the testing, doesn't even output a sync signal.
That's gonna be fun. I'm gonna poke around with the scope probe and if nothing
happens, it's the address generator and if lots happen, it's the video buffer.
I discovered more broken traces on the RGB board and now I have a picture on
red and green, but blue is still permanently on and shows retraces. The image
is inverted, but sharp and clear. It also barely synchs. My homemade MQ board
doesn't do anything, but at least something resembling a picture. Not bad for
a Hantarex MTC90, I've been told these are rare in a working state.
Well the RGB module isn't entirely compatible and creates a short circuit.
When I remove it, all the colors work, but the picture is still inverted.
Checked the schematics again and I put the switches on the MQ module together
wrong. Fixed that didn't put the picture back right, but at least it now has
no more trouble synching.
Mario Kart Koopa Beach is a challenge! But the Mushroom cup was doable.
Well today is Circuit Board Meeting. There, people are running around which
you can give an unknown circuit board and 3 hours later you get the complete
schematics. "Sadly", it was also my mother's birthday and she had clear
priorities... Also my left foot is damaged and impairs my maneuverability.
So... the 5V voltage regulator is broken. If I apply 5V (4.5Amps!) with an
external power supply, I can see the address generator isn't running. If I
remove the circuit board with the address generator and operate it on 5V,
it runs fine. Without the schematics for the logic board, it's gonna take me
weeks at 8 hours a day to find out what's the problem.
The schematics surfaced about half a year ago. Sadly, they're not annotated.
Well, barely, just the edge connectors. Which I can't measure since the probes
don't reach. Great. But the circuitry doesn't differ much from Road Champions,
so we have a machine to compare some of the waveforms to.
After disassembling the shot power supply, I couldn't find the problem so I
reassembled it and sure enough, the wires weren't soldered on properly and now
it works fine. The screen just shows a solid blue screen with a black line at
the bottom showing in yellow text "GAME OVER INSERT COINS".
I tried measuring the circuit that creates the road video signal from the raw
road signal and found a weird waveform, but it went away as I ordered the
replacement chip. So now I concentrate on the raw road generation itself.
The delay lines are suspect. There is RECI, the signal looks very weird but as
far as TTL levels are concerned is a clean signal, there is CLK, but way way
too slow (should happen once per scanline, it appears it happens once per
frame) and all the outputs are stuck and some don't match the inputs.
Spoiler alert: Yes the bucket brigade chips were bad but they weren't the
cause.
After some lengthy thinking and reading the schematics I figured, the graphics
encoder consists of two priority encoder chips. So these prioritize the
signals, meaning one stuck signal with high priority would cause the entire
screen to turn a color. The Player 2 Car is blue and sprites should have
priority over the road background so maybe...
So I studied both the schematics for Regular Road Champions and Super Road
Champions (since the regular schematics ARE annotated) to set me some
comparison points since the game has two identical Player Car generators, so
I could compare the two. Just finding the generators wasn't easy. The only
annotated things were the edge connector so I looked for the steering wheel
pots of both players. They go into the sound board (which uses them to
deduct if a tire squeal sound should be played) and into the graphics board.
There they were and almost symmetrically in the schematics.
And after some measuring and comparing I found an LS00 at C1 to have both
inputs low and the output low too. Should be high. Swapped the chip and the
game came back including colorful enemy cars. Also turned out blue was too low
so there was barely any blue on the screen. Turned it up, still barely any but
at least white and purple. The bottom text is white, too.
So the road doesn't move. It goes straight and then wobbles a little like only
the least 3 significant bits were working and yes, I thought so, the bucket
brigade delay line chips are all bad but one. And one has one working line
(with a few bad bits for good measure) and another line that does something
but not what it should. Ordered "new" ones on eBay and the road now curves,
even in attract mode. Computer cars slowly move down the road.
If the road goes to extreme left, it glitches and partially disappears.
I couldn't have set the road to appear more right, but setting the horizontal
VCO to a different value got the road to move more evenly over the screen.
It still goes partially offscreen to the left AND right, but it's at least
balanced. And I haven't mentioned, the game is playable. It plays a lot more
like Monaco GP (poor man's version), but it's two player.
I also noticed that each player has some different sounds missing (so having
two players yields a complete soundscape for a single player), but I found the
right speaker to not work or be VERY quiet only having some treble, but no
middle, no bass, just slightly touching the volume pot crackled the volume up
where it belongs. Probably all the volume pots (and there are LOTS) are
garbage.
Another thing not working is the VCOs. The road looks normal and matches the
flyer screenshot perfectly, but the cars stay the same size all over the
screen, which they shouldn't. Also if a player car explodes, it doubles in
width. So the chip that applies X VCO to all the cars must be defective.
Maybe the Road board VCO output is defective (but it uses it with no problems)
Tobi also fixed the stuck gas pedal of player 2 and we were the first people
in this millenium who played Super Road Champions. It's not easy, but you can
enter extended and the game ends when you score 600 points, then it plays a
lengthy melody. There is no finish line, but the game has a start man with a
flag and instead of the finish line there is a cup that appears.
The flaky 5V rail was due to a corroded fuse holder.
After two entire days of tinkering where did I get? Nowhere.
VCO appears to work fine. I'll have to study RC and SRC's schematics some more
and do side-by-side comparisons.
At least, the High-Score circuit doesn't work because first the LS174s that
store the current high score are defective and the 74C85 circuitry that checks
if the last game's high score (since it's a two player game) is higher than
the current high score is also busted.
First I swapped the 174s and put a resistor into the high score comparison
circuit so that thing at least shows something different than 000 for once
(if it did. Some segments are defective and the LS48 are to blame).
Sadly, the 74C85 has a totally different pinout to ANY 4 bit magnitude
comparator IC that ever existed, even the CMOS 4063 has the same exact pinout
as the 74x85 with x being anything including nothing but C).
And everywhere else in the entire machine they used 7485 or LS85. WTF?
The machine worked fine through the entire open saturday. Not bad, since I
expected the remaining original bucket brigade chip to fail as well as the
Hantarex MTC90 but neither wanted to give up. Also, many people have actually
played the game, definitely more than N-SUB, even though that one is better.
So I built myself three adaptor boards. At least it works better than the
resistor. Actually, it works just as it should.
So now the high score circuit works, let's fix the high score display.
I removed the most badly malfunctionning chip as well as one that has the
bottom left segment missing. But, damn. You can't get LS48 anymore and LS47 do
exactly the opposite as well as needing a pullup resistor for each segment.
Now the middle digit has the segments dark that should be lit and vice versa.
You can at least read it if you know how. The other chip's missing segment I
fixed by adding an external pullup resistor. It's a little erratic and doesn't
turn on or off all the way on every number, just on some, but it shows some
signs of life. The remaining unsocketed LS48 has some erratic segments on some
numbers as if the internal ROM that contains the matrix had bad bits.
Well, *facepalm*, the good old CMOS 4511 does exactly what I want. Oh wait, it
doesn't work at all. Hmm. Built a test circuit. Hmm I can't get it to work
reliably. Let's test these three enable/lamp test/latch pins. Turns out, you
have to pull latch low instead of high as is on the LS48, but besides that,
they're identical. Phew.
After four more weeks of unemployment giving me vast time to troubleshoot the
faulty VCOs (I just don't get why the cars get smaller as they near the top of
the screen on the older Road Champions but keep their size on the newer Super
Road Champions, ruining the entire 3D effect), I've rewatched the Nazareth
music video Holiday, which features probably the only remaining footage of a
working Super Road Champion. And much of it, too. And... their cars don't get
smaller towards the top of the screen either. And the video was recorded in
1980, so their machine couldn't've been bad. Broken by design. WTF.
It took a day or two to fix the sounds. The left motor and the right ice sound
are missing. They took CMOS 4069, sandwiching each gate inside a volume pot
(again, WTF?) and both 4069 had one bad gate each. But I didn't notice that
before pulling at least two other chips.
Note: Don't use HC04 as replacements for 4069. They get really hot.
After a few months, the road was looking ugly again. One of the replacement
bucket brigades bit the dust, I should have ordered more than four, so that
when the next one dies, I won't have to wait. The one remaining original chip
still works.
There's a weird game glitch, sometimes the second player gets awarded massive
amounts of points (going from 199 to 800 at one point), but it cannot
reliably be reproduced and appears to happen at any score in any increment.
Lately the start flag guy appeared mid-game and the road stopped moving and
had to be removed by jiggling around the circuit boards. You could also
"shake off" one of the "Computer"-cars, making the game easier.
It also appears as if it was supposed to be played by either player 1 or 2 in
a one-player-mode.
And after a little more than a year, the picture drowned in red.
Swapping inputs to the monitor and outputs to the RGB module proves the
monitor is at fault. After taking some measurements: TDA 2350 color IC three
signals go in, two come out. The IC is socketed so I can just order a
replacement and stick it in on open saturday. Lifting the red leg makes the
screen go normal, albeit with no red. Injecting a signal with my Testofon made
the red channel react so the output stage transistors are good.
Just... the chip arrived and it's completely dead. So let's order some more
and see if one's working. Nope. This can't be...
Checking through my pile of ancient TVs for one that uses this chip I only
found a METZ from '82 that uses a 2532, only difference being that this chip
uses pin 7 as another blanking input as opposed to red gain.
So, inserting the defective NOS TDA 2350 into that TV makes... no difference,
it just works without even discolorations. So either this is a mislabeled 2532
or that monitor uses it in an illegal configuration.
So it *should* work in that monitor. Blanking input is a little low, changing
a resistor pulled the signal to specification. But still no image.
Lifting pin 8 brought back a picture, but without blanking, the reference
voltages shift away from the dominant color (green).
But wait it gets even more crazy: The floating pin stabilizes at 4V.
Connecting a pot between 0 and 12V to that pin results in no picture at any
setting. OK let's put pin 8 back in and lift pin 7. Now everything looks
normal! It IS a mislabeled 2352.
I have some trouble adjusting green so that the orange car doesn't look red,
But you can clearly spot the difference when player 1 car is on screen.
Now the guy from zzzaccaria.com has actually sent me an MD and MQ module.
He's crazy! He doesn't even want my money! Shipping accross the ocean must've
cost a small fortune. Thanks, man!
So the MQ module has all chips socketed and is known bad.
My tester can't do 4016 (4066) or 74LS221, the LS00 passed as did the LS32.
Putting the 4066 into my homemade MQ module (thanks again to the same guy who
made the schematics useful by adding chip types and pinouts) got me a black
screen. Bad, throw it out. Same with the LS221, I just got a ghosting of the
outlines of graphics objects. Cool, now I have a working official Hantarex MQ
module.
So what's the RGB (MD) module doing? Shorting out. Expected that. My monitor
isn't compatible. Doing a Testofon checksum with the output stage transistors
passed, the TDA 2350 is the genuine article and works fine, I'll for now just
keep the old MD module.
And guess what was in those bags from Reiner's father's radio/tv repair shop?
74C85 galore!
In the meantime I've written a program for the In-Circuit-Chip-Tester which
can dump the 4 bit PROMs. And used it on the Model Racing machines. Just to
find out that undumpedwiki is being down for three months and I can't upload
them anywhere.
And then someday I try replacing capacitors on the car-board hoping the VCO
would come back, but not only didn't that get me anywhere, now the pink car is
stuck at the top of the screen and won't come down. No fun.
The schematics aren't annotated, making it all guesswork. After some time I
gave up and removed every single graphics ROM one at a time. Eventually, I got
the pink car to turn into a pink rectangle by removing the one that is on the
bottom right in the schematics (the bottom left one is for orange/green cars).
So now let's check the clear signal input on the counter driving the ROM.
The orange/green ROM has one static pulse and one that moves down the line on
the oscilloscope (it's always static pulse - moving pulse, so adjust the scope
to trigger on the next static pulse). And the moving pulse syncs to the car.
The clear pulse on the pink car is on turbofire.
I was searching for this signal forever. It comes from another board (CS219)!
Unfortunately, the chiptester only fails 3 ICs, one of them looks perfectly
fine on the scope (good old false alarm you get in-circuit), the second one is
a 7400 where two gates aren't responding to inputs, but these aren't even
connected to anyting (no traces connected to them) and the other two gates
work fine, so bite me. The third is a 7404 where the input to one gate is
shorted to ground and the output is nagged at to be shorted to +5V, so maybe
the gate is bad, but it's doing exactly what the machine demands from it so
bite me again.
Doing all these measurements I see that there's a transistor bent over and
shorts to a trace (no soldermask anywhere on this board). Maybe that was it.
Nah. All this crap is timered by an NE556 on the first board. The second half
is constantly firing away because the first half isn't running. Swapping a new
one in did nothing. Beepchecking capacitors, swapped a 4.7æF tantalum and a
.1æ disc cap (wonder if that did anything - the tantalum measured OK ad the
disc cap measured only .07æ - wait a minute, these can't break, but this one's
way out of spec) and now the first half works again and the pink car drives
somewhat peculiarly, but it's going down the screen. Sometimes when it reaches
the middle, it goes up a little before going down and respawns on top of the
screen.

Steeplechase
Symptom: No ponies. Wait, only unsynched white lines. Wait, no more picture.
As with Road Champions, I start at the address generator. Schematics are found
on The Web (the Road Champions manual I have to scan in, this one isn't on the
internet).
The horizontal portion is fine until HSYNC, which doesn't appear. NULL.
The vertical portion is driven by HSYNC. What's going on? A 7474 drives a 7402
which outputs HBLANK on pin 10, inputs 8 and 9, having signals of different
frequency, but no output. A 2-input gate with different frequency inputs MUST
have an output. So I swapped the 7402. That didn't do anything. Can it be that
LOW is above 1V? What else is on that line? A 7411. I remove that one and...
nothing. I socket the 7474. Even if you bend out its output pins, it looks
like that. Clear case. Sadly I don't have a spare. So I build a voltage 
divider to get low somewhere around 0.8V. That resulted in 6 HSYNCs per HBLANK.
The result is a laser light show on the monitor. Looks promising.
Got a replacement LS74. The game now shows... TADAA! The normal picture incl.
horses. But only 3 and one moves backwards at ludicrous speed. If you coin up,
the game is over before you can start it. Push the button fast enough unlocks
all 7 horses, 3 of which move backwards and the ones standing still are
flashing.
The 7411 I socketed has only garbage on the outputs and controls the horses'
backwards movement. Removing the chip does nothing. If you measure the pins
while touching them you might as well just measure your finger. Stuck open.
Sadly, I don't have a 7411. Gotta wait another week. And Zimmermann doesn't
stock them. Now I gotta go to EBG again just for ONE chip (and now EBG closed
down forever...).
I put the broken 7474 in my Atari 2600 multicart. It just BARELY works.
So I put in the replacement 7411 and the machine doesn't eat the credits.
The game runs normal except you can't control any of the 6 horses.
Checking the inputs (without schematics), I found a 7414 (which I have).
The scope says the outputs are either stuck low or high and don't react on
input the slightest. Swapped it and the machine is fully restored.
Well the crowd noises sound like a ripped trace ripping one, but the whistles
sound OK and the hoofbeat sounds great. And the fanfare works too.
I thought the game sucked (it does, if you play alone that is), but with 4 or
6 people it is a LOT of fun.
But the buttons don't work all that great after over 40 years.
The continuity tester says the same as the game board:The buttons barely work.
I spare you the details of how every button doesn't work unless you push it in
a VERY specific way which is different for every button. Long story short,
I just bought new buttons.
But I should mention that between September 1981 and February 2017 only about
700 credits were added. For the first time after 35 years, the machine is
getting some love again. Well I have to help, because most people just run
past that game (like me, they think it sucks). So I'm standing around asking
everyone in sight who's not currently playing a game to play this one and if
I do get 6 people to play, they like it and often enough they play another
game or two. Or sometimes 6. Or more.
After more than a year, the horses started jumping by themselves. Even in
attract mode. If you started a game, it looked normal apart from the jumping,
but as soon as you hit an obstacle, alle the horses started moving slowly
back and forth and then the game stopped. I couldn't find out what it was and
after an hour - miracle cure.
Next day it worked fine. A week later on an open saturday the error reappeared
after 2 hours and went away after 2 more hours. Some weeks later (one week
before the next open saturday) I tried again. First time it worked fine, an
hour later I turned it on again and the error was back. The jump gate gets
half track signal plus... an invalid logic signal. I traced them back to a
7404 directly under the voltage regulator, which had garbage on all outputs.
Swapped it, game runs.
The machine gets about 1000 credits a year, but barely any games. Probably
because the credit button I put in is big and bright yellow. Small kids see it
and think if they push it 900 times something might happen.
And just a few months later, the game developed a nwe problem - it doesn't get
faster anymore. The game is still playable, but it's a challenge not to win.
Very boring. The schematics don't say right away which is the acceleration
circuit, but two parallel to serial encoder chips which output the picket
and ground line signals with a clock controlled by a preloadable LS161 were a
clear indication where to look. All those counters worked fine. The last chip
I checked was a 04 inverter. 6 signals went in, 5 went out. Swapped the chip
and the fun returned to the game.
Ah and by the way... the machine was stationed at GamesCom Cologne for a week
in 2016 or so and they told me it was a big hit there, lots and lots of
people kept playing it, so the fact that it had 16k credits before and 22k
credits in 2017 is credible. I still need to make a circuit that turns off the
credit button light when player 3/6 lights are on (simple. Use a relay)
OK now the machine has some weird problem: No horses except for Number 7.
Number 7 doesn't stop at the finish line, it just keeps going out of the
screen (although the horse only gets rendered until the end of the finish line,
then it just disappears even though the track continues for a while)
Tried turning it of and on again, now it plays two notes from the melody in an
endless loop, as if it would watchdog (if it had a CPU).
Here's the cause: The fastening ring of the credit button has loosened and
blocked the credit button's mechanism, locking it in closed position.
Why I write this down? So you know the symptoms for a shorted credit input!
Now I've installed that credit button lighting hack so a relay will kill the
credit button illumination as well as the button itself so once the machine
is maxed out on credits it will stop counting up. I will now have more
accurate readings concerning credits. But still 120 new credits per month so
it appears that people are indeed playing it.
And after like 3 years (2023) you can only play a four player game. If you put
another credit in, Horsie 3 swooshes to the finish line with in a quarter
frame and keeps wrapping. According to the schematics, that must be signal
FWD3, that comes from a 7404 of which already a couple have died in this
machine. Or is it the 9301 mux that generates the signals? But my money is on
the 04 (at K3, if my memory serves me well. It doesn't). So I remove it and
it's the culprit. I put it in the tester and one of the outputs is floating.
I have a little fun and replace all the replaced ICs with period correct ones
(because I can and because I have lots of them) and eventually I take a look
at the chip tester and it says PASS. According to the scope, the chip fixed
itself! Not putting it back though.

PONG (Mirco Games Challenge - a 1-4 player PONG game)
What do you want, it works!
Oh wait if you coin it up twice, it doesn't go to 4 player mode.
Doesn't help that there are no schematics (nowhere in the machine and nowhere
on the web). Well using the old ringer again (continuity tester)...
There are a handfull of 7439 (same as 7400 but open collector, but as opposed
to the 7439 I have them as spares). Coin signal goes into a memory as well as
into that 7439. When the second coin pulse comes, it should go out the second
7439, but doesn. Piggybacking a 7400 fixed the machine. 4 players, fixed.
The machine seems to be a clone of SEGA's ELEPONG 2 (or was that Taito?)
which has the same one player mode.
This machine seems to be pretty popular, getting around 100 credits per month
(and I did the minimal invasive hack that makes the coin reject button the
credit button).
Sadly, the 5V power regulator is on the fritz. First, the picture started
wobbling after a couple of hours (and works fine when cold).
It took me some time to find a faulty rectifier diode by bridging them one by
one after the machine started wobbling.
But now the fuse blows after some hours and it appears to have glowed for some
time before blowing. Putting in my bench power supply limiting it to 3 amps
hoping the defect shows up so that I can put my finger on it shied it away
like cockroaches when you turn on the light. So I put fuses between many
components. The one after the rectifier blew. It's not the game, it's the
voltage regulator itself. Not very helpful. I decided screw that, now the game
operates on a 5V 5A power brick.
The problem doesn't go away if I replace the regulator board, so there's not
too many parts that could be the reason.
Hmm in the basement we have another B/W monitor with a 5V regulator circuit.
Maybe I should get that out to circle the fault further...
After 1 1/2 years the clamping power of my hat chip failed requiring me to
really remove the old chip and put a socket in there. Also, the ducktape
holding the Coin Reject Is Now The Credit Button hack has weakened slightly.
This results in the game not coining up or - very strange - the game only
coining up once per game, disallowing the four player mode!
But still by march 2020, the machine finally has more than 10000 credits.
I also tried pinpointing what the fault is, so I changed the big capacitor
(it got really hot - anything in there got really hot). Let's see if that
fixes it.
By the way: March 2020 the machine broke the 10000 credits on the counter
(had 4300something when I initialy fixed it in March 2016)
And by now the original power supply survived more than one open Saturday, so
it must've been the capacitor.

Monaco GP
The game runs fine, only sometimes the offscreen scoretable works fine.
Most the time it doesn't and just overwrites all the entries with the last one
no matter how bad it was.
I remove the score board and take a look at it. A single chip has been
socketed. I remove it. Boy, someone rammed it into the socket so that 3 pins
have been bent away. Put it back in correctly. I look at the circuit board
from below. What a nice collection of dry joints in various states!
2 hours later and high on solder smoke I put the board back in and hope for
the best. At the end of the next open saturday my record from the beginning of
the day was still there (ok it slipped to 4th because someone was better than
me and I had to get my lead back:-) ). Looks good. Although the machine 
creates dummy entries for every score that have 10 or 70 points. That explains
"Ranking: 032 of 028". Not so great is that someone did 3700something points
which should have been a second place, but the machine entirely ignored the
score and said 2nd is still at 2400something points.
It stays like that: Scores with 3xxx points are ignored unless there are no
higher scores in the table (although sometimes, it works fine).
Well better than nothing.
Uhh I should stop using this machine as a test platform for the other
machine's boards. My car was gone and the enemy cars went through the scenery.
Got rid of those problems by jiggling the boards.
Hey I finally made it to 9999 points. It gets really boring afterwards as the
game is locked in normal road mode and I basically died of boredom a minute
after getting to 9999.
After listening to the sound modules from the other two machines, I now know
that there are two enemy car sounds and not just one and the ice sound (slip
zone) is missing.
The ice sound is there, but way too quiet and with too high a DC offset.
Since everything seems OK, putting a piggyback on IC54 which is next to a
4016 next to the ice sound potentiometer (it's an LM324) seemed the most
reasonable thing. This made the outputs change dramatically so I swapped this
IC and the ice sound came back.
The Other Car circuit I went back to forth. The car sounds arrive at the mixer.
There are 4 ICs in a row mirrored vertically (each half of the ICs is
responsible for one of the two sounds) and they're absolutely identical apart
from the resistors which makes it really easy to check because the pin to
check against is literally the same pin on the other side of the IC.
First I got blown off track by the fact that there is one tantalum that has
the same voltage potential on both legs and the other side of the circuit had
two different voltages. But the cap was OK. So I checked against a working
sound board and this has two frequencies on these two caps.
So this time I started at the beginning of the chain. The IC immediately in
from of the mixer had great differences from side to side, especially one
output was completely off when compared to the inputs. Put a piggyback LM324
on and all of a sudden everything started oscillating and the signal now
looks like on the working machine. So another dead LM324. That finishes (off)
the sound board.
Now the scoreboard is the only thing having issues. (the aforementioned dummy
entries)
So I checked all the chips with the chiptester. The board not having a clock
making this really easy (there's just one LS393 somewhere that shows all
outputs stuck LOW which is a false alarm).
However, an LS00 at 13H was not a false alarm, one output was stuck high.
The chip failed out of circuit, too. Let's see what the scoreboard does.
OK instead of SOMETIMES getting dummy entries, now the entire board is filled
with them. Turns out it must be one of the RAM chips and the bad IC just hid
the fact how broken this RAM must be. Since all the other chips pass and all
I can't test is the 8216 BUS transceivers (the machine can test 8212 however).
Must be the RAM. I socketed them all, they're 2111. Three made by NEC, one by
AMD. The AMD one failed in the tester (btw. the 8-bit was error free, making
this a 256 bit IC, it should be a 256 nybble IC).
Now I gotta make an adapter for 2114. Here, both chip selects are tied to
ground, making it even easier than the Atari Middle Earth pinball.
The scoring is working fine again.
After a year or so, it creates phantasy entries in the score table. Someone
played the machine, got around 1200 points and it said rank 3 of 1 with rank 1
being 8700something, rank 2 being over 3000, rank 3 was his and rank 4 was 17
or something.
And all the chips pass in the tester! Maybe a bad connection. I played it
afterwards and it didn't create weird extra entries...

Monaco GP Boardset 2
We have 3 Monaco GP machines, 2 of which are in the basement. Since I'm
running out of fixable machines (only 6 left and some require unobtainium
chips and won't run properly on MAME) I decided to give one of these a try.
Well... the score board works even better, no dummy entries. The big logic
board #1 works great, no difference to the working one, the big logic board
#2 seemed to work fine at first, but doesn't take inputs. It doesn't coin up
and in attract, when driving through a tunnel, the headlights move right
rapidly then reappearing left and repeating so either there are more chips at
fault or one deeply buried within the system. Sadly, these are specimen of 74
series logic chips that are as unobtainable as some customs but could be
replaced by other 74 series chips.
My chiptester fails almost all the ICs on the board with Mid Level signals.
Scoping the signals showed nothing out of the ordinary, the chiptester is just
overly sensitive to a board that sucks amps.
Only few chips are left untested (LS123) and those are made by Hitachi, so
they last next to forever. Even Toshiba chips die more often and these are
known to be very reliable.
So. After finishing Phoenix, mid 2020 and no more parties every weekend due to
Corona means I can finally get some serious work done. So let's get at it!
Let's do the stupid "check every IC with the In Circuit Chip Tester" again.
This time maybe hooking the power supply lines directly next to the chip under
test, possibly eliminating these stupid Mid Level alarms. But that did next to
nothing, they got less, but not by much. And some chips show stuck outputs
that definitely aren't stuck which made me desolder another LS193 and an LS192.
So, OK, let's put the board back in and see if I can revive the machine
(it hasn't seen the juice of Powerline in probably decades).
The monitor is DONE FOR! But with lots of turning pots (and even changing one
that fell apart) it produces a somewhat usable picture.
Hm - all the graphic errors, especially the headlight cone in the tunnel
being rendered as just flickering garbage - GONE! But I can't coin up or steer
the light cone. The steering board I already tested in the other machine.
(side note - the steering board is brilliant. Instead of just relaying the
light switch outputs like every other board on the planet that doesn't use
pots, this has two lines that tell the main board, in which direction you're
turning and how fast - with just 3 simple logic ICs - this is one case of work
smarter, not harder)
And coinup didn't work when I tested the board in the other machine.
(btw. we're talking about the middle board in the 3-stack to the right of the
machine when you're behind it).
So I remove the known working left board so I can work (measure) on the middle
board. And after tracing the signal through lotsa ICs, I get to one that takes
the three coin inputs at once. It's a triple NOR and one input is stuck low.
Traced it back to one of the coin switches that isn't connected. And dead as
well. Replaced it and soldered in place. And I can start a game. And throttle
the accelerator. And steer. Let's just skip that I wasted one or two hours on
troubleshooting a perfectly working steering.
Playing it I notice the timer doesn't count down. It's stuck at 90.
After checking the 555 timer circuit in and out of the machine (and even
forcing the  clear signal High), I put the third machine's board in. Which
also doesn't count down. So I checked with the working one which produces a
ca. 1Hz square wave. So we got ourselves two broken timer circuits.
First I wanted to remove the 555. Then I saw the fat tantalum cap next to it.
Out with it, Testofon says it's fried, swapped a new one in (another NOS
tantalum ;-) ) and yes, I can make it toggle out of circuit.
Put it in, now it counts down and the gameplay is as normal with Extended and
Game Overs and everything, hooray!
Onwards. Even in Game Over you can hear a deep wobble repeating roughly once a
second from the speaker (uhh I just remembered the speaker didn't work, one of
the leads to the cone was apparently frayed close to the speaker terminal so
applying a blob of solder fixed that, the speaker works great now)
It sounds like a Super-Lo-Fi sample being played. If you start a game, it gets
much louder. And then I noticed that my own car engine sound never changes.
The other sounds seem to be there. (notice that was before I tested the third
machine where the ice sound was... existent)
The circuit for the frequency of the Own Car engine sound shares an IC with
the circuit for turning the engine sound off and on. It's a CD4016. By Fujitsu.
Now Fujitsus pre-1985 are known to be unreliable and CD4016s are also 
notorious for going bad so listening to the internet this time I immediately
pulled this chip. And nothing happened, so back in with it. There's a big
tantalum in the circuit for the engine frequency (the board is littered with
tantalums and LM324s which also often go bad, that's gonna be fun).
I checked that tantalum and it leaks big time. Replaced it. I still got the
wobble, but I also got my engine back, proving the chip I pulled has nothing
to do with the wobble and works just fine.
OK, breaking out the second power supply so I can operate the sound board on
the bench. Since all the oscillators are running all the time and just getting
switched on or off by the game I can now easily get at it with the 
oscilloscope. And after a while I found an LM324 where some grumble is going
in and huge grumble comes out. Checking in Chop-mode that it's the same signal
going in. Hm. Don't think it is the bad one, but removing it allowed me to
unplug pins and see what happens. OK this chip's good. If I remove the output,
the sound is normal, but my car explosion sound is missing. But it worked
before! Really weird. Now I gotta check against the working board.
Explosion comes from a SEGA custom. Great. The other machine has no output
there however! So let's compare all the inputs. And there's an input at 7V
where the broken machine has 1.5V. Pulled it up through a resistor and the
grumble fades. Looked at the schematics, there are 2 identical circuits, maybe
for two different explosion sounds so I can compare them. And finally I found
an LM324 there one output doesn't react to its inputs (by resistor shorting
again). Swapped it and that fixed the problem. Spent roughly half a day on
fixing this sound board.
Another thing: While testing the steering board I broke the gear switch switch.
It has a little roller on its end and I don't have a replacement for that.
So a normal microswitch with a reed (or how do you call that little metal
strip that actuates the pin that pokes out of the switch). With lots of
bending I could get it so you can shift gears, but shifting the gears makes
weird stuff happen, eg. you can go from 0 to 100 in 5 frames.
But the machine shows that the switch is generally speaking doing what it
should. And we have now two fully functional Monaco GPs!
Had.
Next week I came back, two of the cars just moved right through the scenery.
Swapped board with machine 3, error went away so I know it's on this board.
So ran the chiptester down the board again and noticed 2 ICs working worse
than before. Removed them and an LS193 had an output stuck high and an LS157
was completely garbage, doesn't switch inputs. Replaced'em both and no
difference.
But I fixed the slip zone sound (as I now had revived the third machine).
Was a little tricky since the sound is there but looks weird and doesn't react
to the potentiometer. With op-amps you notice best that the output is 
somewhere in between the inputs (unless you got negative feedback) but here I
had an input with a fixed voltage and one with a variable input of similar
voltage offset and the output was fixed (stuck). Piggybacked a new LM324 and
got almost the same waveform as the working board so another chip bites the
dust and the soundboard is fully functional now.
I swapped the soundboard with the third machine because that one's fully fixed
except for the displays and the machine is in way better condition 
cosmetically and monitor-wise.
So I got back to the Cars-move-through-the-landscape-Bug and put a working
board next to it as a reference. And there I noticed that I replaced an LS193
with a 192 *facepalm*. What effect did it have? Right again. None.
So kept on searching. This time in the schematics where I pinguessed the area
where the enemy-car-counters could be (the signal for the Bridge gets in there
too and most of it is written in Japanese which even if I could read it the
scan quality is too bad for it to be deciphered). And IC89 (LS157) had an
output stuck high but not on the other board.

Monaco GP Boardset 3
Got a few hours left, let's see how far I get.
I had the inner big board in the second machine and can confirm that two of
the enemy cars are moving smooth in one direction and jerk back quickly in the
other.
If you observe Monaco GP for a while you will notice the sides of the road and
the other cars moving in a simple triangle wave, they just bounce back and
forth courtesy of simple up/down counter ICs. The cars move at three different
speeds (the blue one being the most drunk). The cars using the middle speed
move left to right normally, but right to left they jerk by 16 pixels each
animation clock pulse.
And the timer doesn't count down here either. Removed the tantalum. It's good!
Put it back in and pulled the 555. Guess what, here the 555 was bad.
Now I got the ca. 1Hz clock there.
So I put the board back and slowly revived the machine. I couldn't revive the
spiders though, the machine was once a Metropolis (of spiders).
The monitor needed the filter cap to be swapped, the Testofon said it's
shorted. I could shock it back to life on my bench supply, but Testofon still
said there is too much leakage so it would have blown up sometime anyway
(the monitor would very likely just have worked with the cap given the fuse
is very time-laggy). Anyway it's got a much nicer picture than the other one.
But the scoreboard is completely dead. Dark. Nothing.
And the credit counter immediately locks on after powerup.
But I already know, that's normal without a coin switch. So I soldered the
credit switch back in and would you guess it, the machine works immediately!
Throttle, Steering and gear shift works just fine.
Except for the erratic car movement (clearly a bad counter chip).
And the sound has even worse problems.
The fanfare... someone detuned big time. The notes are way out of tune and one
seems to be missing. And if you drive over ice, you just get a tinnitus beep.
Weird, I don't think there's a sound for ice, just the puddle splash sound
which works. I probably need an entire day to work it out. I can probably fix
the cars within an hour.
No, it's the other way round. The frequency potentiometers were all turned up
to eleven. I turned them back according to what was left of the paint marks.
One or two I couldn't quite make out where they were but two out of six is
easier to debug so I turned them until the fanfare wasn' *completely* out of
tune.
And there's a potentiometer for Slip Zone and yes, this is the ice, and yes,
it's supposed to make sound. With the pot, you can set the chaos feedback of
the noise generator. If it's set too high, the noise generator will find its
resonance frequency which is a high-pitched beeeep.
So I turned the pot until it sounded more like slipping on ice.
The cars took longer. According to the In-Circuit-Chiptester, almost all the
LS161s are bad with completely nonsensical outputs and if I turn on the
oscilloscope I even believe it. Five of them have really broken outputs so I
desolder them. I tested them out of circuit and they just pass like fine!
Put them back in, turned it on and it works. Fixed it. Wot?
So the displays are missing. First I thought there's a fuse missing, maybe 
it's for the displays. But the fuse holder wasn't connected to anything.
If I buzz the edge connector with the Testofon, I can get segments to light up
so the wiring is not the problem.
If I put the display board into machine #2, everything is working except that
place 1 and 2 are shown as place 1, making the place 1 score double as bright
and the place 2 score is completely dark.
At least what's on screen is identical to the other Monaco GPs.
But once more the sound module... ambulance doesn't make a sound.
Checked the part of the circuit generating the signal, works.
Checked the part that switches the signal in and out of circuit, went 
backwards from the Enable-IC (pulled the control input Low), works, and then
I got to an LM324 that only has one input hooked up and the output is stuck
low. Put a piggyback on, no difference. Bent up the piggyback's output and it
goes High, so that's the one. (and it was)
Back to the score. They had their own fuse indeed and it was blown. I still
had to take the machine apart just to see where the wire went.
What's really weird is that the LS161s still work a week after I took'em out
and put'em back (my experience shows that sometimes a thermal shock 
temporarily fixes a bad IC, but temporarily usually spans from a few seconds
to half a week).
Back to the score. Yet again.
The scoreboard is relatively simple to check. Testing the board gave the
following results: Place 1 shows up at Place 5. Place 2 doesn't show up at all.
(after a while, Place 1 also failed). I played 4 games but still places 3 and
4 stayed 0000. Or 0020 or something.
First I checked the ICs generating the line strobes.
The IC generating the strobes is OK. The IC generating the blink signals is
OK (both LS259). The strobe IC is driven by an LS390. That is two LS90 for the
price of one. And the half controlling the strobe generator is broken.
I don't have 390s, but I might make one out of an LS90 and the working half of
the 390.
Am I too stupid to look_ I got 390s, even factory new.
So I checked all the ICs in the chip tester, it just has a bank of LS174s,
LS151, LS48, LS85 and four 2111 SRAMs.
One of the LS174 was completely shot with five outputs stuck low and one stuck
high. The others were error free.
So I swapped it out and at least the problem with the random numbers in the
score vanished, now every score shows as 0000 (except for the two dead ones).
So. Put in a new LS390 and no difference. After checking all the ICs with the
chip tester again I had to sit down next to the machine with the oscilloscope.
And here I could finally see that the 7416 (or 17?) doesn't pull two of its
outputs low. Piggybacked it, worked, swapped the chip, displays work and the
broken one passed the test anyway. Huh?
OK spent a day calibrating the fanfare with a frequency counter against a
known working board. I had to replace a 680 Ohm resistor with a 1k one since I
couldn't set the frequency right. And it still sounds completely off.
One of the notes is completely off. One thing foiling the calibration routine
is that the ROM defaults on pulling the 2-switch LOW, so that this switch
doesn't make a difference and you can't calibrate it or rather everything you
calibrate will be calibrated against the #2 switch being on.
Well - I would have to remove the ROM from both boards (isn't socketed).
And it still doesn't put your score into the table (this machine has the
score board from the first machine)

Assortment of Monaco GP boards
Years ago, Ari received a package. It sat in the corner of the arcade workshop
for years - she recently opened it - it contained some Monaco GP boards:
3* Clock / Scenery
2* Car video
1* Score
1* Sound
And nothing works except the scoreboard.
One of the clock/scenery boards is almost working (the leftmost big board in
the cage) - it's playable. There are some weird rectangles on the screen which
don't move. Looks like a character RAM addressing error - if the machine had
anything like that! The road color is also slightly off. These errors
disappear when you reach Extended, except for the ice part.
Another fault - the board completely screws up the clocking of the scoreboard.
It stays on some displays for much too long and only drives the other ones for
super short pulses or not at all.
The second clock/scenery board makes cool copper bar looking lines scroll
across the screen. Scoping around for a while revealed a 74161 with floating
outputs (not a tri-state device). Swapped the chip and the board didn't have
anything else wrong with it, it works perfectly fine now.
The next board shows random garbage on the right part of the screen and on the
left side it draws a perfect roadside scenery (that moves left to right like
it should) but never draws a road. (scoping for an hour I found nothing too
far off - yet)
One of the car boards only draws unsynchronized pixels going all over the
screen (like h-hold is misadjusted, but only for the cars) (with collision
detection - now try dodging TV static)
On this board, some 74157 get really hot, but still pass the test. Two which
don't, fail. Also there's a 161 with two dead outputs.
Pulled them, they all fail pretty horribly out of circuit. Still, it didn't
make the tiniest difference.
The other car board repeats the cars horizontally (Y-direction technically).
You can't overtake them, so the only way to stay alive is to match their speed.
Here I found two LS157 with stuck outputs shown on my tester on this board,
but not on the other board. They're just as dead out of circuit and curiously
this completely fixed that board. So we have one completely working spare set
of Monaco GP Boards (except sound, see below).
The sound board worked a lot better than expected. One of the "Other car"
sounds constantly plays in the background and sounds weird. There is no
fanfare (maybe something barely audible in the background?).
All the other sounds are working though (even ice and ambulance).

TANK:
Here comes a beast from 1974! The machine had its original monitor lost and a
Grundig TV (The German #1 brand until 2000) as a replacement, wired to a
Pentium 3 PC with an emulator running (for reasons unknown) Ultra Tank.
First, restored the power supply as good as I can. Unplug the MAME machine and
replug the original game boards after ringing through the capacitors
(NOTE: my continuity tester isn't just that: It's called a Testofon and the
frequency it produces is dependant of the current (<2mA). It can determine
short circuits, resistances (ballpark, but good enough to see if the printed
value matches), capacitances (ballpark, ditto, also you can measure leakage
current - if ANY can be measured, throw the cap away, it's bad), 
semiconductors and inductors - and you don't have to look at a display, it
doesn't have one. It also doesn't have an on-off switch. Your brain writes the
software to even measure caps and semicons in-circuit and determine their
value and if that transistor is shot and since it measures in extreme 
real-time, you'll be able to find really quick where a signal is going in a
huge PCB you don't have a manual for.

OK back to the translation.
Found a shorted cap in the 12V rail. Replaced it and applied power.
Both boards have 3V on their 5V supply (both boards have individual 5V
regulators). Pushing 5V into each circuit board with my lab power supply
yields 2 amps per board. A bit power hungry, but it is from 1974, nothing 74LS
to be found. Will be hard to supply both boards at once but the right board
has the address generator. The schematics sadly were used as a mattress by a
rat, had to puzzle it back together. Horizontal looks good, vertically it
doesn't but I was looking in the wrong place. Also there are 9313, not 9316.
Hard to read the schematics, not because of the rat but the copy is barely
legible, the manuals on the internet have the same problems, parts are
illegible. Maybe that was some kind of copy protection...
That cap I replaced must have shorted out in operation. The 12V rectifier had
2 diodes badly charred, replaced them. Still no 12V. Removed the boardset.
After the cap has shorted, the diodes must have shorted too, because the
12V traces were burnt through. BOTH. Bridged them with wires and got 4.87V on
the 5V rail. Good. But lots of ripple on it (50Hz). Problem is, I measured the
input to the rectifier at 6.7V and it wants 7V. This was a very narrow margin
for a 5V voltage regulator even in 1990, but the 1974 regulator doesn't like
it and now I have 0.3V of ripple that I will probably never gonna get rid of.
(unless I buy a new transformer with VERY custom oh wait 7.5V should be
obtainable).
Well, the machine produces a picture! But almost everything is broken!
The background is gray (correct), the minefield is in the middle of the
screen with black mines (correct).
The minefield has a dark gray background (wrong) and is arranged in a clean
matrix (wrong). Scorebox shows solid blocks in the color of the player.
There are no obstacles and only 2 different gray level squares flicker accross
the screen. Well the machine's 43 years old, there ought to be lots of broken
ICs in there...
Swapping the chips will be the greatest fun of all. TANK consists of 2 boards
held together by a "flexible" bus bar that looks like its wires will rip just
from looking at it. (EDIT: The bus bar is remarkably robust)
Since according to the schematics all the logic blocks are driven by the
single ROM, I thought, let's pull it out and see what happens. Well, nothing
happened. The ROM's dead. We have 2 more TANK boardsets (one of it accessible),
let's try that one's ROM. No difference. Whatever drives the ROM looks OK on
the scope, but that part of the schematics is illegible. Might be the socket,
but I couldn't see floating or illegal signals. They looked clean.
Oh wait. The fat rom needs -12V to operate and I can't remember measuring
anything negative at that chip. So the -12V is to blame. Should be easy.
Jup, can only be the 741 Op Amp or the transistor. I cautiously socketed both
because you can never know how much mileage you can get from that bus bar and
should remove the boardset as seldom as possible. Turns out the 741 was to
blame. I had spares (3 years younger but looked identical). That restored the
-12V. Also laid a bypass for the 5V traces which were pretty charred as well,
but the image now looks worse (without the ROM), just a gray screen with a
flickering giant pixel. Putting the ROM back in brought the game back 
entirely. Everything looks like it should. Tanks, Mines, Background, Scores.
Let's see if it plays.
Almost. The black tank won't stop flashing, even in Attract mode. If one tank
hits the other, it starts flashing for a short time and the attacking tank
can't shoot for that time, giving the shot tank a chance for revenge (or
escape). Sadly, it doesn't stop. Traced the error to a NE556 timer. Inputs OK,
support circuitry OK (tantalum caps, you can never know but these were ok) so
it's gotta be the chip. I don't have a replacement right now. Also one of the
amplifier chips doesn't work. Gotta buy. So it worked fine for a few minutes,
but I got curious investigating a color upgrade and finding out I could see
for every pin of every IC where in the picture it fires (NOTE: Put composite
sync into a TV and around 1V into SCART fast switching signal and you can use
the TV as a two-dimensional oscilloscope which for video circuitry of an
arcade machine is a lot better than a normal oscilloscope). Problem: I must
have hit the -12V and shorted it to some chip's output. Now the backgrounds
are rubbish. Traced the problem to a bank of 74153 that control the fat ROM.
I shot two of them. More to buy. Week later all my chips have arrived, put 
them in, machine's fixed, hooray. Hooked it to the Grundig TV (used a small
surveillance monitor before that, which is useful if you want to have a
picture while working in the back of the machine) and there's one problem:
It doesn't like the sync timings. The picture is shifted upwards so far you
can barely see the score. Hacked into the TV's hidden geometry setup, but I
couldn't get the picture down completely. Even the black&white TV had trouble
synching the top 20 scanlines, showing they wouldn't normally be on screen.
The scores aren't on screen entirely, but around two thirds are, enough so you
can see the score at any given time (as long as you remember the one has a
base and the seven doesn't).
Playing the game showed a weird bug: The white player can't aim in the upper
left quadrant. He always shoots at 10 o'clock no matter where you aim.
The schematic wasn't the greatest in legibility, but structured well enough to
guess which part is the aiming "mechanism". There's a 9316 per player per
direction that gets preloaded with values buffered in a bank of 7474 which get
loaded directly from the tank image address of the fat ROM (so the graphics in
the ROM coincide with the direction in which they shoot - smart!).
One of the Flipflops (E2) had High for Q and !Q so this chip was clearly dead.
Now the machine is completely restored except for the ripple.
Swapping the diodes didn't get me anywhere as did bridging the 0.22Ohm
resistors after the power caps (who knew Computamite still exists?)
BTW the CSYNC signal looks OK. Back then they probably figured "it runs with
this monitor, who cares if it runs with any regular TV".
And after like 3 years the blue tank can't shoot left. If you aim somewhere
left, the shots will fly right as if mirrored, but not for all directions.
But I know where I have to look. And when the shot explodes (if you hit
something or run out of range), the explosion animation flies horizontally to
the right. Looks more like a feature than a bug, but it's certainly a bug.
And when I had time to fix it, I couldn't reproduce the problem for the life
of me. (maybe it's doing the Steeplechase - you remember where the IC faded in
and out of working whenever it felt like it)
And after like a year or two the fault finally came back and stayed until I
scoped it. And I went straight to all the 7474s and one had a weird floating
output signal on pin 9, but I forgot which one. One to the right and two or
three up from E2. Funny, the chip tester says that pin 9 is the only broken
output, so both flipflops work fine, only the noninverted output of flipflop 2
doesn't work.
Another thing I noticed that's kinda new but looks a lot more like a feature,
if you hit your opponent, not only does your tank turn slightly randomly to
prevent you from shooting again right away, but even if you line back up and
shoot again, if you hit your opponent, it doesn't count. The explosion
animation appears, but your opponent doesn't flash and you don't get points.
This is until either you or your opponent move. As in move. Turning doesn't
count. And since the behaviour is the same for both players and kinda makes
sense, I guess it's a feature.
And I  do have to fix the left player fire button once and for all.
It keeps breaking apart, I probably have to re-engineer something.

As I said, we have another boardset of TANK (the third boardset is 
inaccessible and also said to be working). This one does NOTHING. No sign of
life. The transformer center tap contact has completely disintegrated.
Used a strand of wire to bend over the edge connector to fix that.
Afterwards it appeared to be working better than the first boardset (before
fixing the -12V). But the 5V regulator doesn't regulate at all (showing the
same exact signal as the input to the regulator, only 2V less).
The backgrounds and mines are completely faulty, mines are big black pixels
instead of X's. Backgrounds look like 64V or 32V is missing.
The tanks move right at blazing speed. Knowing what I broke on the other
boardset I'm gonna keep fixing that one.
And after lots of years I decided to revisit this board. So out goes the
quartz and in comes the In-Circuit-Chiptester.
This one marks a 7486, almost all 9316, all 74153 and one 7420 as bad.
The last chip turned out to have 7420 as a date code... nothing wrong with it.
The 7486 had one gate bad, I shorted the inputs but the output didn't change
so in the bin it went.
The 9316s look healthy, just that the chiptester doesn't get how they're wired
up, that they can reset themselves or one will drive another that controls the
first, so they fail because another chip interferes.
The 153s are wired so that they should pass in circuit with the quartz
removed. So I removed the '86, one 9316 and one 153 as a sample. The 9316
passes. The 153 is just as dead outside as it was inside. So I removed them
all. None of them did anything. That explains the total lack of graphics with
a ROM that tested good.
So juiced it up and... still totally garbaged. A plus: The scoreboard works.
The minefield is semi-working (mines look glitchy) but the arena is only 32
pixels in size and repeats itself over and over. And no traces of any tanks.
Trying to get reference values for the still dead voltage regulators I of
course killed the one from the working TANK. I found out pretty quickly that I
just bashed the pre-driver. Datasheet says it's identical to the BC547 except
it can do 60V instead of the 547's only 30V. But since the circuit deals with
nothing more than 12V so it should work just fine and guess what, it does.
Also just half of the boardset failed as killing the regulator doesn't cascade
to the other one.
Well now I have voltages. None of which matches in the defective regulator.
The resistors are OK. The Zener is OK. The caps are all ceramic, they only go
bad when physically damaged (or hit with a couple hundred volts in which case
there's probably a little more bad than just one cap).
So I shotgun EVERY transistor. None of which fails the Testofon test (holding
my finger to the collector to see if it switches on). Stupid situation.
Well all the parts are working but the circuit isn't. And yes, we have the 12V.
For fun I replaced one of the beefy power transistors, and that made it work.
The old one still switches voltage, but no current.
And when you replace the entire battery of bad 153s with 157s and ram a chip
into the socket so hard you bend a pin out so it shorts something, you don't
need to wonder why it doesn't work. Because that's all that was left to make
it work (even tried shooting every direction with every tank).
Although the right tank has a motor sound that sounds more like a UFO.

TANK color upgrade
For people who have a color monitor in their TANK for reasons, this upgrade
makes their machine much more attractive.

We have pictures of our TANK machine with the original monitor, but that was
before my time. No clue what happened to it. Instead, as mentioned, there's
a standard color TV in there, so you can watch TV, too. In color. Even DVDs.
The original mainboard is a logic graveyard consisting of logic blocks that
contribute a part of the video signal. So there's a block with just the video
signal of tank 1, another with tank 2, the backgrounds (sadly including the
scores, so the original machine has to do some trickery to get one score
black), the mines, the shots.
Since I want tank 1 to be red and tank 2 to be blue, I have to do even more
work to do get the scores correct. For this I need to separate the scores from
the background.
For this (as well as other things) we need a 7404, a 7408 and a bag full of
diodes. Or a lot of OR gate chips (probably 2 or 3). I'm going with DTL here.
If you happen to have historically correct chips at hand (I have, but only
00,01,03,10), you should preferrably use these. If color monitors were
affordable and/or reliable in 1974, TANK would definitely have been a color
game. But a color TV that size would have costed more than the entire machine
including its monochrome monitor back in the day. And it would have broke
after a few weeks of 24/7.
You should also run the generated signals through some kind of buffer if 
you're afraid to expose the inner signals of the original machine to your
monitor (you should.). You can use a 7407. I used a 7404 with three outputs
connected to three inputs to invert the signal back to normal, because 7407
are open collector which means I need pull-ups and have no clue how to
dimension them.

OK take note the following plan might have errors. After putting it together,
tank 1 was green instead of blue (put a diode to green instead of blue) and
the scores were mismatched (red score for blue player and vice versa).
So you better triple-check the schematics before heating up your soldering gun.
A small refresher on Boolean Logic:
! = NOT
& = AND
| = OR
So where are the image signals coming from:
Tank 1		= B10 pin 2
Tank 2		= D7 pin 12
Mines		= D7 pin 11
Playfield'	= F14 pin 13 (I think, but no matter)
Playfield' also contains the scores, so we have to derive them from existing
signals (math without the math and the headaches).
The shots also have to be derived, but they already contain the explosions,
too.

Now let's derive the signals. We need some more signals to do so, like
Scorebox (F14 pin 1 I think) and left half of screen (=H7, also on F14).
Score 2 is Playfield' & Scorebox & !H7.
Accordingly, Playfield = Playfield' & !Scorebox &CSYNC.
I don't rely on blanking, so I use CSYNC to blank. Otherwise your TV might do
weird things. Some monitors don't care so if you feel brave, you can omit
CSYNC.
Score 1 and the two shots exist as inverted signals, so we'll just need to
invert them again.

From theory to practical, with chip positions and pin numbers, like painting
by numbers (almost).
Shot 1	= !L7 pin 13
Shot 2	= !N7 pin 10
Score 1	= !F14 pin 12
These just need 1 gate of a 7404 each.
The next signals are hairy:
Score 2	= F14 pin 13 & F14 pin 1 & !F14 pin 2
(again the disclaimer: I might have score 1 and 2 mixed up)
Playfield = F14 pin 13 & !F14 pin 1 & B11 pin 4
Looks like it would be more comfortable to use a 3-input AND gate.

So all we need is 2 7404 (or a 7404 and a 7407) and a 7408, a couple of 1n4148,
3 pulldown resistors for the diodes and 3 limiting resistors for RGB (since
TVs and monitors don't like TTL signals). I used 3k9 for the pulldowns since
I'm running out of 4k7, for the SCART RGB I used 510 Ohm (because I have tons
of them). 390-820 Ohms should be OK as should be 2k-10k for the pulldowns.
Also you need a 150 Ohm resistor for SCART fast switching to tell the TV to
use RGB instead of composite. This one you make switchable and put the switch
in front of the machine so you can switch between the original black&white
signal and the color signal any time you like or someone comes and says
"that's not original".

So now that we have all the signals we need, it's time to turn them into
colors.
The combination I used leaves the walls white, backgrounds black but makes
the shots yellow, mines blue, tank 1 and score 1 red as well as tank 2 and
score 2 blue. (again, it is possible that I switched the score signals, so
make sure you can swap them easily)

R = tank 1 | score 1 | shot 1 | shot 2 | playfield
G = mines | shot 1 | shot 2 | playfield
B = tank 2 | score 2 | playfield

An alternative for G/B would be
G = mines | shot 1 | playfield
B = tank 2 | score 2  | shot 2 | playfield
this makes player 2's shots pink, using yet another color.

Who likes, can remove playfield from the red, rendering it cyan.

If you like a gray background like the original has, you just need to turn up
the brightness. You'll have to, since the original video signal has black
level even below that so when you switch from color to b&w, you'll just barely
see the black objects.

Sanity check: The 'real' 7404 has 5 gates of 6 used and the 7408 all 4 gates.
Update: In mid 2020 I did the "blue tank shoots cyan red tank shoots yellow"
upgrade and the level select upgrade (the hidden arena looks like someone took
more time to design it than the standard one)

Another TANK
This one's gonna be fun - I don't have access to this machine. I'm helping
another arcade repair technician fixing his.
Symptom: The mine graphics is the only thing that's correct.
The mine placement however looks like the ROM is missing except for a missing
column in the middle. And the arena and the scores look the same.
It appears that there is only one byte of data getting recalled from the ROM
(except for mine placement data), hinting at stuck address lines to the ROM.
I told him he should check what comes out the 153s and what goes in, at least
the select signals. IIRC most of the signals come from the address generator.
I sent the guy screenshots of using the video probe from the ROM's pins.
He sent me a link to two PDFs, one which was a legible schematics for TANK and
another which was like a theory of operation, repair and upgrade guide
(which also had the "switchable arena mod" explained) which also had
screenshots of the video probe for various signals.
So he says he has no VINFO while my machine has a few horizontal bars across
the screen. This signal uses signals exclusively from the address generator.
Adding to that he says there's nothing out of the second horizontal address
generator. That's gotta be our suspect here. Probably the carry still works,
but not the other outputs.
Damn stupid, my remote colleague says this chip is OK. As the only chip in a
swamp of defective ICs he found. Swapping these didn't do anything either.
But he did talk about rusty parts and where there's rust, there's short
circuits. Or it's just the ROM is bad.
Turns out the ROM is bad, but thats's not the only problem.
He ordered another "tested working" Tank board on eBay. It arrived without a
quartz, so how anyone could have tested this board and find it to be working
is beyond me. Maybe it's supposed to make a black screen with no sync, who
knows.
So with the ROM from the "working" machine, his Tank board does a lot more.
He says the picture now looks like it should, but there are no tanks.
They also don't appear if you start a game.
That means one of three things: Both tank circuits broken (very unlikely),
video circuitry bad (unlikely), or tanks offscreen, maybe the counters that
store the tank's position never get latched, set or reset.
In fact the Tank-Reset-Circuit was bad. Now the black tank appears in the
right spot and looks good. There are fragments of the left tank scattered all
across the screen (1 pixel high but 1 tank wide each)
He writes that both white tank counter ICs were bad but still had all outputs
pulsing - finding something is quite an accomplishment.
After swapping one of the ICs, the tank's fragments were only scattered over
one dimension, so swapping the other brought the white tank back as well.
He says that shooting doesn't quite work, but the machine is running again for
the most part - he had to swap every 5th IC so far...
(no I'm not translating the e-mail correspondence)
Trying to power up the eBay board violently blew one of the big caps and took
the rectifier diodes with it.
And last thing I heard of him was he found many more bad chips and now
everything works except tank 2.

Quadrapong
Sadly we don't own it, but came just in for fixing.
The machine had 00120 games and a 40 year thick dust sheet on the credit 
counter and no burn ins or high-voltage attracted dust on the monitor, so it
was neither tampered with nor rolled over. Wow. This cocktail table really was
used only as a table. (also, I confirmed the credit counter works. It didn't
all the time, but back then it surely did)

Symptom: Monitor doesn't sync. That a machine with so few hours on and
inoperational for so long doesn't comprise the picture out of differently
colored magic smokes is a miracle. At least you can kinda sync.
What do you see? horizontally 4 playfields, vertically 4. And the VSYNC in the
middle of the screen. MIDDLE of the screen? And again at the bottom?
Since the monitor gives off a 15 kHz whine that sounds just fine one can
assume it runs at standard frequency and not some weird high-def (ever slipped
adjusting your computer monitor and got it tuned to a much much higher freq?)
Also, there is no trace of HSYNC. Looks more like a logic board problem to me.
The scope shows the same: The video signal has no HSYNC. I wonder where the
VSYNC comes from... so pulling out the logic board. Hm
?SERVICE MANUAL UNOBTAINABLE ERROR IN 10120.
So let's see what counts. There are lots of 9316 which do. Two of them have
signals that look like fight. Put on a 74LS161 as a cap, bent up the outputs
and it looked nice. Replaced it, nothing. Went through all 9316 on the board,
shorting their outputs and checking if the picture changes. No dice.
In the middle of the board I see 4 7493. Shorting their outputs changed the
picture. Finally. After an hour of searching I found the address generator.
And all the outputs look... somewhat healthy. So time for a piggyback.
First 2 chips make no difference, but chip 3 twitched and for a second, the
picture looked normal. So I replaced it and the machine was fixed.
Just 1 chip bad. Not bad for a machine from mostly 1973 (some chips had date
codes from 74 but many had 73).
Oh wait: One potentiometer doesn't work, you can influence but not control one
paddle. Measured the pot out of circuit showed the same. Removed it showed
that at some point liquid was spilled into it. Took the pot apart, cleaned it
and it works fine. The game is more fun than PONG, even the 4 player PONG.

Missile-X
Our Missile-X usually works just fine, but if you look closely it has Problems
with the IC sockets - sometimes there are stripes through the sprites and the
last time I checked the explosion graphics were missing, but all comes and
goes so that's due to bad contacts.
This is the story about the spare board we have of this rareish machine.
Kinda interesting, years ago 2 people have asked me the pinout without knowing
of each other, individually. One used my old mail address which I rarely ever
check so I saw that just now. I reverse-engineered the pinout for the first
dude back then and so I could just send a copy to that other guy, he responded
that someone else slipped him the pinout already. Lol.
Plugged it in and it's just X with no Missile. Like Frogger without the frog.
Or Donkey Kong with no Mario (or Jumpman). And there's only one tank in the
top row and it's going too fast, it should only go that fast in Extended.
Everything else seems to work. But it's not as if the missile was invisible,
it doesn't get rendered. You can get the chaff to pop up and even the
explosion graphics, you can start a game, but you can't set which of the three
depths you want to go to (or rows).
So, looking for invalid logic signals got me the first chip in the bottom left
corner above the ROMs (or whatever these big 24 pin ICs are) (bottom left with
the edge connector on top). It's a 7474 that has half of its outputs floating.
And all it does is fix the top tank's speed. And after looking for bad ICs for
more than an hour I couldn't find any, but instead of following differences to
the working board, I just kept going on until I finished half of the board and
everybody else went home so I had to go too.
So next weekend I went looking for differences and traced them back.
On the fourth chip from the bottom left (2 above the bad 7474) I found a
signal that's missing on the broken one and it even looked like the missile
video. Following that signal through what felt like 20 ICs (in reality it was
3 or 4), but luckily all simple gates that also only had one input missing
each, I arrived at a 7410 where all inputs were stuck (or was it a 7420?) but
were toggling on the working one. They came from a 74107 that was configured
as a divide-by-two and didn't have an input signal either and that came from a
74161 which actually got a clock input and was set up so it should count but
it didn't. If you piggyback it, it still doesn't count and some sporadic 1Hz
pulse shouldn't hinder it either.
But if you bend up the outputs of the piggyback, it counts. And since there's
another identical 161 next to it, I remove that too (together with the 7474)
and... well the second 161 could have stayed, it works fine. Well now it's in
a socket. Put a replacement 161 for the first one and the rocket is back.
And for some reason, the game now works perfectly fine apart from the fact
that there's only one tank in the top row. But it's fully playable.
---END OF CPU-FREE ZONE---
------------------------------------------------------------------------------
Q*Bert
Problem: No sound, no knocker, sometimes no red, most the time completely dead.
Resoldered the soundboard first, wasn't really necessary, but well-soldered
looks different. Still no sound. How does the voltage regulator board look
like? It's one single dry joint. You can just pull out the components.
Sunk lots of solder into it, still no sound. Oh, the 30V fuse is out.
Great, sound now works fine. Knocker still doesn't, but the game plays great
again. Red still fails sometimes. After a few hours Q*Bert called it quits,
crashed and stayed dead afterwards. Only the next weekend. Over the next
months, Q*Bert just worked when he felt like it. Meanwhile the Pinball People
replaced the Knocker coil which was completely dead and showed traces of
charring, also the 30V fuse blew again a couple of times, then we noticed,
it's a 2A fuse (or 1A) and someone put a 7A fuse for the knocker, so when the
game crashes, the knocker sometimes permanently turns on until the 30V fuse
blows because the knocker fuse should blow at 1A...
But now he's completely functional IF he feels like working. Although 
sometimes the sounds are weird and you have to wiggle the sound board cable to
fix that. And red fails more and more often.
After many months I noticed that the fuse holder for the 5V fuse gets hot.
And is horribly corroded. Replaced the fuse holder, resoldered the RGB
connector and - no more crashes or picture problems.
But the sound got even weirder. Most sounds work fine but some are wrong
(plays the wrong sound effect or repeats the first note of a melody instead
of playing the entire melody) and Q*Bert doesn't stop talking and does so an
octave too high. The sound control signals are slightly weak, but not too
much. Swapped the 74LS377 that produces the signals (the only socketed IC in
the entire machine besides ROM, RAM and CPU), worked for half a minute, then
reset to aforementioned behavior. At least the game is playable.
(in the German version I mention -5V spikes. I probably forgot to ground the
scope). Tried swapping ALL the ICs that are responsible for sound board
communication, nothing.
When I resoldered the monitor, I missed the vertical deflection which failed
half a year later due to very dry joint (?OUT OF SOLDER ERROR).
Meanwhile, the sound malfunctions got way worse. Substituting the receiver
chip with a couple of switches I could determine the fault lies within the
sound board entirely. Can't be the 6502 or the RIOT or the speech chip.
When I was fixing the Video Pinball a couple of weeks later (see below), I
wanted to test its very corroded CPU, so I put it into Q*Bert's Sound board.
And left it there. Firstly, the pins are so corroded, they will likely stay
in the socket when I remove the CPU. Secondly ... it works fine again!
All the sounds are back to normal (ok he had that loose-wire-problem I
mentioned above but jiggling that fixed it).
Very weird. The soundboard CPU is dead but the soundboard still works for the
most part. Even says "Hello I'm turned on" on powerup. Put the CPU in my CBM
8032 and was greeted by a garbled screen. Case closed.
(btw. back when I was little and had no clue and a tiny underpowered soldering
gun, I tried to remove the CPU of a dead 1541-II. It lost one third of its
pins in the process and I kept it flying around for over a decade. Since the
8032's CPU is now in Video Pinball, I need a 6502 CPU... I soldered the 
missing pins onto the ripped 6502 (guess who was the pin donor) and it still
worked!)
One open saturday the red failed again, but differently, there was still some
tiny traces left and a dark red bar on top of the screen, where the graphics
area ended. Mugg turned it off, waited 10 seconds, turned it on again and
fixed it forever.
I also replaced the score saver battery (a miracle it hasn't leaked the
slightest, now it only leaks electrons - a full charge lasts an hour).
It now keeps scores, but for some reasons re-set to factory defaults in the
middle of operation (I now leave it on factory defaults since after adjusting
the values it crashed and wouldn't leave the service menu ever until I removed
and put back the battery - at least it keeps the scores for months).
Now there's one more thing I want to tackle: The 30V rail. Why 30V? The 
knocker coil is perfectly happy with 24V and the sound amplifier chip is rated
27V maximum. I got a ten-pack of 24V Zener diodes at a bargain (no, that is
not a translation for the bad pun I made in German and which can't be
translated). Was good that it's a 10-pack, since they're rated 1.3W while the
original is rated 5W. The board below the original Zener is baked crispy.
Sooo if I take 4 of them in parallel, I get 5.2W, sufficient... only... if one
of then has a lower voltage due to tolerances, the entire load goes through it
and kills it. What to do? I solder the new diodes (5 of them to be on the safe
side) simply over the old diode. If the new diodes fail short, the fuse pops.
If the new diodes fail open, the old one is gonna take over and we're back at
30V, which didn't kill our machine in over 155000 games!

Donkey Kong Jr.
The monitor trolled me. I thought it needed recapping, the picture rolls
vertically (if the monitor was mounted landscape). After months of using a
10-turn-pot for VSYNC I looked at the signals with a scope. No VSYNC.
The 74LS123 which generated it was dead. Thankfully, Nintendo put tags on some
testpoints (as well as ... testpoints!) (not a bootleg). New LS123 fixed it.
After some years without problems all of a sudden the sound turned up all the
way. The amplifier module which is part of the monitor, is bad. If you remove
the audio in, it gets less noisy, but the weird howling sounds which you can
change if you turn the volume knob, stay.
Damn. It needs 100V and 120V power supply and they go straight to the power
transistors. I'm not gonna try to debug this thing in-circuit.
Let's salvage one of these cheap (free!) green speakers, that worked.
Yay got 93600 points yesterday.

Another DK Jr
There's no sprites. Oh wait, you can see eyes and teeth of the Klaptraps, but
only if they climb up/down the vines. Since some more parts of some more
sprites appear if you jiggle the connections and DK hardware is known to
produce graphical glitching below 4.9V, I guess my test rig isn't adjusted
right. And yes, only 4.5V at the mainboard. And if you crank it up to 5.25V,
you get 4,97V at the board and... it's still glitchy, but every sprite is
almost completely there. It's definitely playable, not a single sprite missing
but most sprites do still have single lines or couple pixels missing/
flickering. It's pretty clear it'll come solid if I give it a few more mV.
What doesn't come back though is sound. Analog effects (walk, jump, climb,
fall) work. They sound wonky though.
But nothing else. No music, no other sound effects. Gotta look after the sound
CPU another time. I ask myself how many boards I've written OK on even though
sound doesn't work just because I couldn't test it...

Even more DK Jrs
Here we got one that only displays sprites. No backgrounds. Instead just solid
colors. Swapping EPROMs back and forth between a working one brought the
backgrounds back when swapping 3J or something.
Reading the EPROM reveals it's almost blank. The last 100 bytes contain some
zeroes. The EPROM right next to it looks the same.
Now I noticed some sprites have glitches which are constant, weird and hinting
towards another semi-erased EPROM. And another fault - when the Klaptraps go
down, they paint the screen red to their right until the right screen border.
And while the EPROMs were busy programming their contents back on (none of
which were defective, they just got greeted by the Sun for decades and had no
stickers on) I noticed a ripped off pin on an IC, but that was a red herring
as that wasn't connected to anything. I did see a scratched trace, beeped it
and it had no continuity, all the errors were gone.
Then I had another board where all the colors were dim except blue which was
exceptionally bright. Must've been a bad transistor, but replacing a 2SA1015
with a BC558 didn't fix it, now the blue channel is inverted!
But since the transistor's input was pretty strong already, I just soldered in
a 330 Ohm resistor and that got the blue to the same level as the other colors
(the colors are weird on this machine anyway, still dim and only one shade of
brown. Well...)
And then I had another one missing the blue - but also the PROM for the blue
channel and some of the green. I happened to find that PROM and put it back.
And then there was another one where the digital sound wouldn't work, but this
time it was just a ripped out resistor.
And then we had another one where the digital sound wouldn't work. This time
it turned out to be a bad DAC08. Too bad, that was my last one, so I had to
order new ones.
And then there was a collector who owned a DK Jr which also had no digital
sound. There was a ROM socket on the board that was suspiciously empty.
And yes, the regular DK Jr has a ROM there. Is that the sound ROM? I make a
duplicate from one of the few original Nintendo boards we have and yes, that
was the sound ROM and nothing else was bad. Btw. I searched the machine for
the original sound ROM but couldn't find it anywhere.

Xevious
Symptom: Bluescreen.
Optical inspection shows corroded pins. Can't do much here now. Let's take a
look at the fuses. One looked very definitely blown. Measured the rectifier,
one diode was shorted. Well I thought that can't be the reason for the
bluescreen, this machine shouldn't have had any picture at all.
But it was. Replaced the rectifier, game fixed. I then was in an urge to find
the volume pot and turn it down. You could hear the machine in the far corner
of the hall when all the other machines were running and the music of the game
is optimized for maximum annoyability.
And after a few years the picture starts getting brighter, maybe the monitor
needs some fresh caps.
Nah, it's the same model monitor as Kangaroo with the same scratchy pots that
just feel like randomly creeping up or down. Just not as bad as on Kangaroo.
I did do a recap of that monitor. Turns out only 2 caps were marginal and the
others I checked were fine.
Oops I totally forgot that one, after like 5 years one event evening it
crashed. I didn't think much of it and turned it off and on again and it ran.
But no such luck next weekend, open Saturday. But now I know what they meant
by "blue screen"... that was because of these kn0rky color potentiometers in
the monitor. Well anyway the 5V was measuring 1.4V. This time it was the
2N3055 though.

N-Sub
Boyohboy I'm really gonna put me up against a machine THIS rare?
Measured voltages. 5V was at 3.7V. A little low. Video output looks weird,
monitor shows a solid amber picture. Scope shows the same, but you can see a
clean sync and the video level is zero during the blanks.
And... no short circuit in the 5V. With an external power supply I get 3A on
4.7V, if I add the original power supply it sinks to 2.5A at 5V.
Suspected the big power transistor, 5V is now at 4.2V. uuhh.
Checked the data and address bus. Ringing, but ok, the machine's running.
Gotta get a circuit diagram for the power supply. 12V is at 0.something. Hm.
Ah short circuit. What's on the 12V? Just the DRAM. Please, not the DRAM.
Checked the caps, one of the tantalum caps was short. Replaced it, 12V.
5V is now at 4.5V. But what happened to the video output? It looks like a
video signal! Turned on the monitor and the game works OK.
In between I had to replace all the jumper wires to the daughterboard. They
used very thin unflexible wire which broke after bending twice.
(I forgot to mention I accidentally swapped 2 wires and got the gradient for
player 2 cocktail table configuration and the water sparkle was missing -
after swapping the wires back, everything looked like the flyer which was the
only image source for an N-Sub on the entire web!)
OK game runs, but no sound. Volume control is set to 0. Why? Turn it up, 
instant tinniiiiiiiiiiitus. In the background you can faintly hear the game
noises. Ah that's why it's on zero. Tried randomly putting caps at different
places in the sound module. Got rid of the beeeeep by putting a cap to the
+-15V power supply. But there is still some hum on the sound, but at least
it's quieter than the effects.
And on some days there's only 4.3V on the 5V rail and then he crashes every
half hour, I gotta take a look at that.
OK we've got ringing. A buttload of ringing in the power supply. That gets
smoothed and the mainboard sees almost clean 4.5V. But the ground rail as
well as 5V have ringing relative to the power supply. hmmm what did I read
about blocking caps around a voltage regulator? Maybe they're dead.
There's a 7805 driving a beefy power transistor. Put a blocking cap at the
output of the 7805. Ringing halved, stabilized at the middle, which means 
still 4.5V. Also, the speaker started buzzing louder. Let's try the 7805's
input. Instantly, ringing and buzzing vanished. 5V is now at 4.8V, shouldn't
crash anymore, the sounds (analog soundboard!) sound marvellous.
Only no one ever plays it. 200 credits in one year (ok counting the 5DM clock
in makes for about 400 credits). Still too few. And around 100 are from me
playing the thing. My record is at 44800. I love this game!
Even Super Shot gets more games. And this game's horrible.
Ari has the daughterboard partially reverse-engineered and added the
background gradient as well as sound to MAME. The water sparkles (AKA 
starfield) are still missing in MAME.
Well when Ari took the board apart to reverse engineer it, she knocked off a
heatsink to some custom chip, making the game freeze after a few minutes of
operation. Putting the heatsink back on fixed the problem. Luckily there
appeared to not be any side effects.
Hmm March 2018: Turned it on, Garbage Screen. Yeah great, the Museum people
just added the machine to the list of machines to take to VCFE Munich. DAMN!
By the way the N-Sub version of Garbage Screen is a checkerboard-like pattern
made of mostly yellow  random characters in front of the gradient/starfield)
Turned it off and on again, different garbage.
Opened the machine and it just came back on like so. Weird. Hm there are
weird graphics glitches at the left hand of the navigatable area. They are
deadly to the touch. They get erased if you die, but as soon as anything moves
or detonates there, weird glitches stay back that even eat your torpedo and
won't go away. Weird.
OK they want to take the machine to VCFE, so let's clean it up again.
I remember the last time cleaning that contrast filter screen in front of the
CRT. I had to clean it again, had to do it thrice from both sides with the 
Ammoniak (Ammonium? anyway this stuff bites!), since the Kleenex were still
yellow after the second wiping. I even cleaned the window and the bezel
(for the bezel I just used water). The machine looks like new (except for the
tons of spider poop on the bezel), but - the Red is gone! OK the game has
barely any red, but anyway, PANIC! Hm the starfield still has pink pixels in
it, so scoping around the daughterboard, connecting the mainboard directly to
the monitor shows that there is no red from the mainboard. It comes from an
LS157 that switches between Sync and Video signal or something. Well there is
no video signal input to that chip, that comes from an LS175 (no typo) that
has an illegal input (floating), which is connected to a custom IC, probably
a PROM, that doesn't do anything. Or invalid signal? Wait, no, there is a
lovely digital signal going on, only the LS175 doesn't get it. Re-socketed the
custom chip and there it was again. 2 hours of searching for a corroded socket
pin. At least now I know why the machine crashes after moving it.
All the IC sockets are corroded. By the way, at VCFE the machine crashed
seconds after turning it on, then minutes after reseating the chips the first
time, then again the next morning and of course again after moving it back to
Seligenstadt. At the VCFE, people were given little choice (only Asteroids,
Steeplechase, Space Invaders and Bomb Jack) and played the hell out of it,
you could hear the machine across the entire building. I bet it got more games
than in an entire year. Someone managed 17000 points. Even DaddlerTL struggles
getting such a score.
Ari did some more reverse engineering and found that the machine has only 5
levels, if you beat level 5 (I got there. I thought level 4 was a kill screen
already), the game just crashes. Which is great since there's no watchdog.
My current record on this game is exactly 69000. And here's how you can get
that many points:
This game has a bug you can exploit.
If you surface, sometimes one of these Space Invaders UFO gunboats comes on 
the same level as you and just rams you unless you shoot it first.
On the third wave per level, these black ships that circle around, these
gunboats respawn instantly as soon as you shoot them. So surface and keep
shooting left and right (and watch enemy subs). If you get lucky, you can
make it so the next one appears just slightly before your shot exits the
screen. This will give you 450 points (rare!). Also every time you hit one,
the game pauses, so once you're in that rhythm, the enemy subs have barely any
time to move. You'll rack up points in no time.

Sega Vic Dual (Head On+Space Combat)
(11/2020, almost the same hardware as N-Sub)
Well this machine does nothing, not even a picture.
Since all the ICs are socketed except for a single video RAM and the 555 that
resets the CPU, I can put them all in my chip tester. Except for the PROMs
(and the video RAM which is 4027 and I can test it elsewhere).
And there wasn't much broken! I remember an LS253 where an input was shorted
to ground and an LS374 with a bad output, nothing that should have kept it
from working.
The only thing obviously bad is the chip sockets. On some of them you can just
pull out the ICs with your fingers.
Interestingly, I got the machine working though. I was scoping the signals to
the RAMs and one doesn't get an input. But if you just push on it, the picture
changes. And eventually I got it running with no errors and clearly identified
as Head On. The scope says only one row of ROMs is being addressed, so it's
really gotta be a DUAL machine, let's swap the banks over and look at what the
other game is. Only it doesn't work. Maybe it's compiled to a specific
address in ROM, so I need the pinout.
You have to connect a toggle switch (like the old nightstand lamp switches)
and now I can change between this and... a 1:1 clone of Space Invaders that
uses 8k of ROM instead of 4. Well whatever, both games work, machine works,
kinda rare, so I'm done.

Night Driver
Graphics. Where we're going, we don't need graphics.
No, I'm not gonna waste any words about fixing the monitor, the company who
made it does exist although it took a couple of hours of intense googling,
but this model of monitor has never existed and the closest schematics don't
nearly match and it has a trolling transistor whose sole purpose it is to
jam the SYNC signal whenever it feels like it (after 6 hours of diagnosing
with a monitor expert we just removed the transistor and the monitor worked,
then 2 more hours to get it working properly)
Symptom: No steering. Ari thinks it's the optocoupler. It was its amplifier
transistor. Replaced it fixed the steering and since that is the game's only
problem (apart from that monitor which is also upside-down so you can't just
replace it with any other monitor without removing the sticker on front of the
CRT), the game's fixed.
I hooked up a VCR to the monitor and tried to watch an episode of Knight
Rider (what else on this game?) but the picture is horrible. It only shows the
brightest parts of the signal even if you turn brightness all the way up and
contrast down and it really has massive problems with the geometry.
You know these horizontal shifts really cheap portable black&white TVs give
when some very bright object appears? This monitor makes them look tolerable.
Every couple of seconds it completely lost sync and then the horizontal
deflection collapses (that's probably what that transistor was supposed to
do is keep that running, now it's doing the opposite) and once it even made an
audible bang, so after 10 minutes of this I decided to watch the rest of the
episode on Space Invaders. That one has only the cheap B&W TV's problem of
taking contrast quite literal (as in when the picture's dark, black is gray
and when the picture's bright, gray is black).
That monitor is done for. It probably has to be built up from scratch.
And now in late 2017 the track selector switch selects only 2 of 3 tracks.
Traced that to a LS151, 8 to 1 mux, put a cap on which fights so this is now
a 7 to 1 mux. Sad. After 41 years the first IC in this machine broke.
Also broken wires on the gearbox are a common recurring problem.
And then in March 2023 it was out of order and did weird things, angry noises,
but no picture. I jiggled the edge connector and it came back. Turned out to
be a burnt off power contact. I used a strand of wire and pushed it flat and
tinned it, so it made a somewhat decent connection again. (dunno if strand is
the right word here)

Afterburner, 720degrees, Super Hang on all just had some bad contacts.
Tapped against the motherboard and they were fixed.
2 months later Afterburner started to keep crashing every few minutes.
There was some voltage drop in the wiring so that it ran on 4.5V. It's a
little more energy hungry than N-Sub or Donkey Kong (which was glitchy at 4.5V
but ran stable). So cleaned it, bent contacts and set the PSU to 5.2V.
Even 4.9 arrive at the mainboard. Donkey Kong only gets 4.7V (still some
glitches but barely noticeable -  I don't wanna set the PSU higher because
maybe the voltage drop goes away and then I'm frying the chips)
I went and eliminated one of the intermediate power adapters, now there's only
.3V drop left.
Amd then after years of the Afterburner monitor having whiteout every now and
then, it finally stayed in that fault state. The Lock-On lamp blinks
occasionally so the game runs, gotta check when it's quiet. A week later - it
works. A day later - white screen. So I opened it, jiggled everything to no
avail, jiggled everything harder and the picture came back. Hm.
So I took the monitor out, which was a Hantarex 900, one of the last ones that
already had the "gimme what you got I can take everything" video input board
and that funky little geometry daughterboard.
And it was clear to see that one of the big white concrete resistors could be
bent around a little too easily. The resistor is parked close enough to the
power supply to be likely responsible for a power rail going to the neck board
and that makes me check the solder - of which nothing was left, so I soldered
it back on and now it works.
While I was at it, I also checked a few toasty looking capacitors and my
Testofon said they were also a decade past their best before date.
Also the machine has some video problems (ringing on the video signals) which
also indicate bad capacitors (I'm still talking Monitor here).
And after another few years (~5 after Winnis repair), the monitor went up in
smoke and what looks like flames. Ugh and I wasn't even there.
Now there's a hole burnt into the PCB and another half a hole that didn't burn
long or hot enough, an exploded foil capacitor (WTF?), a shorted TDA2953 and
a bunch of scorched resistors. And a solder joint that had anything conductive
burnt away around it.
So what the copulation happened? And in which order? So I asked Winni and he
says the linearity coil was the cause. There are insane currents going through
this coil, so much that if you remove R109 it can start to glow red hot.
So I suppose the other way round also holds true, causing the resistor to go
nuclear when the coil is missing from the circuit, even arcing a lot which is
basically the same as on all these Valvo VCC91 chassis where the solder goes
bad and arcs away the solder joint, but on these chassis, that's all that
happens. Here, half the components are dead. So this is what happened:
Bad joint on the linearity coil causes arcing. The arcing destroys the solder
and the copper until it hits the soldermask, then it extinguishes.
The coil is now out of the circuit, the resistor R109 goes nuclear until parts
of the plastic from the coil and the entire capacitor C200 start melting.
The melting C200 shorts out, sending these huge currents to pin 3 of the combi
IC (TDA2953?) which shorts to VCC, toasting all the resistors along the path.
Now I gotta check all components on the board including the LOPT.
Er well so much can't be bad - the transistors in the path seem to be good and
the 12V regulator regulates, so I just replace the bad components and retry...
except for... well... R179 has also been part of the disaster and completely
shed its paint coating so you can't read its value anymore. I measure 4 Ohms.
And here's the deal - it's not in the manual. It's not in any of the other
Hantarex 900 or 900E manuals. The internet doesn't know it exists.
Winni doesn't know its value. All the other Hantarex 900s in the museum either
don't have a spot for R179 or if they do, it isn't populated.
Well it works with the spare chassis. After some work.
The TDA2953 had to be replaced and now it starts, but the picture keeps dying.
Looks like bad LOPT, so I put the one from the wreck in and that didn't change
much, if you jiggle the board, it arcs and dies, then comes back, and the
arcing appears to not come from the LOPT. So I checked the board, especially
under the rails, and I found the solder joint going to the HOT's case to be in
really poor condition, it has arced and charred the board and corroded the pin
of the HOT's "socket". So I scratched off the corrosion, put a piece of wire
in and to the next okay solder joint and now that's a problem of the past, but
it has another problem, if you jiggle the voltage regulator, the +B voltage
fluctuates. I didn't see any bad joints, but resoldered the big ones anyway
and now it's solid. Let's see how long it lasts.

Red Donkey Kong (Radarscope Hardware):
Sadly the machine is irreversibly wired for JAMMA so we have to make a JAMMA
to Nintendo adaptor. But I can easily test the game in Donkey Kong Jr.
It does produce a picture, but just utter garbage. It appears it tries to show
the High Scores screen, the colors even match, but the letters are pure 
garbage and the machine crashes. Jiggling everything just made it worse, but
at some occasion, the game came back. It looked like crap (well actually just
an address line in the video section hanging), but it resembled Donkey Kong
(he did more damage to the building than usual). You could even play, but you
ran against invisible walls, dying from nothing, then I fell through the floor
and came back out of the ceiling and the game thought I beat the level and
loaded the next one. So I thought there sure must be a connection problem.
I plucked the 4 board stack apart and wanted to look at the chip sockets.
The sockets looked back. And blinked. Changing the chip sockets and cleaning
the pins fixed the game apart from the sound, which still doesn't work.
The sound CPU appears to be running, I see enough oscillators oscillating, so
the problem must be relatively far along the audio path. Maybe a defective
transistor in the output stage. Huh I don't know what I saw. The sound section
has its own connection to the power supply, which is missing. It shouldn't be
doing anything.
Damn, some vandal ripped out the joystick. And the pinball people built a new
control board from scratch (the wood panel was extensively damaged).
And they put the original 4 way joystick back in.
After finding the wiring harness (but still no trace of the original PSU even
though I am sure it was sitting in the corner for a long time) I put it back
into DK Jr and was greeted with the garbage screen. Pushing on all ROMs got
the game back to life, reminding me to swap these sockets as well.
The sound was kinda odd volume wise but everything was there and sounded
normal. We also have a DK whose walk and jump sounds are like three times as
loud as the rest of the sound.

Missile Command had some weird colorful retraces one day, something with
blanking. The chip receiving RGB+Blanking is OK. But the VBLANK is way too
short. !VBlank goes in there which is produced out of VSYNC by a 7404.
Uh the signal's crap, 3V is low. Replaced the chip, picture is ok.
2 months later the picture turned solid dark red after 5 hours.
The machine shouldn't be capable of dark colors. Turned it off and on again,
dark red picture with rolling blanking. Player start buttons flickered and
stayed off then. They should be blinking as the machine is on Free Play.
Doesn't coin up or start a game. Half an hour later the game was working
again. Problem with Power supply? Nope, voltages are OK, but after a cold
start, picture was dark red again, but after a minute it just started.
Went to service menu. The monitor can't display solid backgrounds, the
color then just gets darker. Useless for video games (unless most the picture
is black), it looks like it's been designed that way.
And now I can't reproduce the problem anymore. Maybe a bad contact on the
edge connector.

Another Missile Command
Says "REPLACE 10MHZ QUARTZ" on a sticker on the board.
What do you want, the quartz is OK, it's just the legs aren't physically
connected to the case anymore - anyway whatever I have these so let's put in a
new one. Since this machine needs 3 voltage rails and the Debug-Tabs aren't
populated, I put this board in our machine. And all I get is broken with funny
colorful pixel snow (at least some pixels are changing).
In test it does the same garbage.
The ROMs are of that ultra reliable and trustworthy kind ("what, you looked at
my pin? Now that must fall off."), but they still beep out OK (huh...)
The CPU however stops beeping at pin 9. Out, back in, now I get a few pins
further before it stops making contact. That socket goes in the garbage and
now test says doot doot doot doot deet doot doot doot. According to the manual
that's 4K. Now it beeps 4J. Couldn't it have beeped ...doot deet deet doot...?
This way I could have swapped both bad RAMs at once.
With the two new RAMs in it, it makes a weird new beep code. So first I
thought that new RAM must screw up the BUS, but it turns out that third
frequency is the OK beep and if you switch it back from Test to Game, it just
works perfectly.

Enduro Racer one evening developed the problem that you could only steer right.
Well I thought a wire snapped loose from the pot, but nope. From what I
measure and from the pot-values the self test shows, the ADC gotta be bad, or
the BUS transceiver connecting the ADC. Databit 80h is missing (and a few
others, too). The BUS transceiver appears to be working, my In Circuit Tester
says the same. So I put in a new ADC (you can still buy them new) and that
worked.
Our other Enduro Racer had the problem of randomly crashing. Sometimes it
works for days, sometimes it crashes every five minutes. Swapping all the
ribbon cables almost fixed that, now when I play, I can sometimes see the
entire screen flicker down so the road kinda disappears and you can see the
horizon through where the road should be, but at least it doesn't crash
any more. Except for that one day when it said KOSGST COKOS instead of
INSERT COINS.
But now After Burner won't stop with the white screen. The Lock lamp blinks
and it reacts to start (but no sound - found out later someone turned the
volume all the way down), so the circuit board gotta be working.
The screen did work after a few hours of turning it off and on again.
A week later: Turned on, works. Next day: White screen. Opened it up, jiggled
all the wires and sockets and stuff, nothing. Jiggled some more, picture came
all by itself, no need to turn it off and on again. Hmm.
Couple weeks later it died for good. No amount of physical dislocation would
have any impact on the (white) picture. So I removed the monitor. It's one of
those very late Hantarex 900s, the one with the "give me what you got, just
set my jumpers correctly" and that weird image geometry daughterboard.
Easy to spot is a big white resistor near the power supply section that you
could bend very easily. Way too easily. That resistor is next to a bunch of
wires going to the neck board. White screen with retraces means there's a
voltage rail missing. Checking the solder joints showed there weren't any.
Dumped a load of solder on the connections, I checked capacitors adjacent to
hot resistors which also looked baked, Testofon confirms they're well past
their best-before date. The monitor works reliably now. The picture has some
fuzz artifacts (over-ringing on the video signal) that looks like more bad
caps in here, but I'll do that on next occasion.
Speaking of occasion... Degauss isn't hooked up, but this machine needs it...
Well that didn't happen becasue the monitor went up in flames.
R109 got so hot that the circuit board turned into coal and C200 right next to
R109 melted and half exploded. Also the resistor behind C200 going to pin 3 of
the combination IC (I think TDA 2953 or something) got pretty crispy and also
fried the board below, and the resistor hooking the IC to the power rail also
looks toasty.  Another big resistor next to R109 jettisoned its paint coating.
By the way, the charred spot under R109 disintegrated when I tried to scratch
off the remains of C200. Then I notice a solder joint to the linearity coil
has ceased to exist - including the copper, gone to the soldermask.
Now the question is who dunnit. Which order did it happen in? Winni says the
linearity coil is the likely starting point. There's some insane current
running through that and he says if that resistor R109 is missing, that coil
can become red hot. So I assume, the same holds true the other way round, if
the coil is missing, the resistor goes nuclear. He also says there can be
quite some arcing, like on all those VCC91, only on these the joint is the
only thing that's bad... here, half the components are probably fried!
So - dry joint on the linearity coil leads to arcing. The arcs eat away the
copper until the separation is big enough for the arc to extinguish (the
solder mask helped too). Then R109 turns into a small inferno, melting C200
which shorts out, which in turn shorts the combination IC to the supply rail.
The supply rail can now flow through the shorted blob of capacitor through the
small resistor on pin 3 which lasts long enough for the board to also burn
there (Glass Inferno, anyone?) which of course exceeds the current the power
rail resistor is rated for. I probably have to check every single component
now including the LOPT.
Well it isn't as bad as I thought. The transistors in the path aren't shorted
and the 12V regulator regulates, so I just replace the broken components and
retry... except... well R179 also suffered from the meltdown and completely
shed its layer of paint so now you can't make out what value it should have
had. I measure 4 Ohms. And here's the deal: It doesn't exist! It's not in the
schematics. The internet doesn't know its value (nor is this variants' manual)
None of the other Hantarex 900 monitors has it. It's either not populated or
missing entirely.

Bubble Bobble has the same weird undervolt glitches.
By now, it reliably crashes just before level 2. Just before it finished
scrolling. I suspected the chip sockets, they used the cheapest variant
possible (only one contact tongue instead of two).
Swapping the sockets did nothing. We have two Bubble Bobbles.
The other one is japanese and pretends to not be a bootleg (but where's the
TAITO logo? Maybe it's a japanese bootleg, too bad I can't find out since its
security chip is bad and the only way to find out if it's genuine is if the
letters spell EXTEND instead of EEEEE and the random number generator is in
that chip which on our machine is bad!)
The japanese Bubble Bobble has weird colors that are different on each boot.
The video board is not the problem, I swapped it and now our real bootleg now
crashes randomly instead of always at level 2.
Sadly a week later it now crashes just after level 1 (it scrolls 8 pixels and
crashes). I guess there are timing problems somewhere. A hell to troubleshoot.
So let's fix the japanese Bubble Bobble (a dip switch sets it to English, but
I prefer Japanese - looks way cooler)
As I expected, the color RAM never gets written. The color RAM is attached to
a BUS that's attached to a BUS which and so on. Until I found that in the
schematics, Christmas will be over (I think I did that in April).
The Testofone is the quicker way here. It led me to a PAL which should 
generate the Color RAM Write signal but doesn't. Explains why the error 
doesn't wander with the video board. Waaaiit a minute is there an input to the
PAL floating!? According to the schematics they're all hooked up to the CPU,
but I can't measure a contact! So I put a jumper wire to where it's supposed
to go. The colors should work now, I wonder if that fixed anything else
(the game crashed immediately if you tried to start one - just Title and
Attract worked) err yeah. The game is fully functional and playable now.
Except for sound. That was even worse, I take you on the entire oddyssey.
Sound CPU NMI's stuck high. There's an OR gate which gets LOW for Sound 
Request on one input and the other one's stuck high. Went on a wild goose
chase from an LS74 to an LS155 to an LS138 back to the Sound CPU and could
only guess that one of the address lines has no contact. Scoped them all and
A8 is dead. Weird, there is already a jumper wire on A8 and when I measure it,
it checks out good, but if I measure the Component Side, no contact there.
So fixed the fix and it works. And there is no CGA snow when it builds up the
screen like the other one does, so there's probably a BUS driver out of whack.
Am happy with the fact that this machine works great and the bootleg now
crashes randomly again instead of after level 1.
Cool, we have a Zaccaria machine in our storage that has two boardsets in it,
one looking suspiciously like Bomb Jack, the other (single board) has a
Sticker saying S. Bobble (not sure about the S but it doesn't look like a B)
on it. After fixing the monitor (early Hantarex 900 with corroded fuse holder)
I could test the boards. It indeed is Bomb Jack (albeit with lots of sounds
missing). The other board also indeed is, oh wait, no, it's Bobble Bobble.
But this time, it's the Official Bootleg. It has EXTEND!
And it produces CGA snowing artifacts when the next level loads, just before
scrolling.

Another Bubble Bobble from a big joblot of 3 crates had "kicked the bucket"
written on it. The CSYNC felt kinda weird and went completely wonky (no H part
and just some occasional low pulses long enough to be the V part) and the
machine just boots to a black screen with the B CPU stuck in IRQ.
Swapping the video board from a known working boardset isolated the problem to
be on the video board. I tried the counter ICs but couldn't find anything
obvious. I then conducted a finger test which found a warm LS08. It wasn't
warmer than the ICs around it, but those were counters. A simple AND gate
shouldn't get that warm, so I scoped it and garbage in, garbage out, just like
a day earlier when I fixed that Double Dragon board, the chip had floaty
outputs and corrupted some of its inputs. Replaced the chip and the scope
signal looks like the attract mode now.
Hooking it up to a monitor shows that we're not finished yet... there's
graphics corruption going on, quite obviously every other two characters are
the next two characters repeated, so we're clearly missing 16H, easy, is it?
IS IT?
Weeeelll it's a Taito and the horizontal address counter isn't even used
according to the schematics (apart from 1H and... 256H!?). WTF? That's hard to
believe! And now it's getting wild...
After scoping and piggybacking for 3 hours (pretty much every IC), I got just
as far as no I'm not going for a political joke, maybe cite the movie All
quiet on the western front? So I'm still on square one. And just when I wanted
to call it quits (it was quarter to 1 at night and I had to go to work
-tomorrow- today), another fault manifested, the right half of the characters
is missing every second scan line.
So I thought how about removing the CPU board so whatever magic Taito employs
as a video counter is left alone and the 16H or the lack thereof can be traced
but nope. The most interesting thing is that if I scope one of the outputs of
my new LS08, every second column gets pixel snow. With a different 08 that
disappears - until the chip gets warm.
As an aside, with a working video board, the game crashes when trying to start
the attract mode, corrupt graphics (Title screen is perfect though) and I get
an error message WORK RAM BAD which I gotta investigate but it's pretty
reproducable and looks as if it's really the work RAM.
Another aside, with the bad video board, the game doesn't crash in the same
place every time, sometimes the attract mode starts to run and you can see
that the sprites are not affected by the missing 16H.
And under-/overvolting like with Scramble Formation back then didn't change
anything either.
One of my strategies was to randomly short pins to ground which didn't help
but also didn't harm, until I shorted pin 9 and 10 of a 163 counter together
and suddenly the board's power consumption spiked and all the video RAMs got
cooking, and even though I turned it off quickly, the consumption is like
double the expected and the RAMs get really hot). The entire thing smells like
bad video RAM, but their inputs are bogus. Stuff doesn't work, half the
signals on the board now look like gobbledigook (illegal levels...), there's
6 MHz everywhere (where there was data before), the 12 MHz are almost floating,
CSYNC is utter garbage (looks like 16V). Parts of that (especially CSYNC)
christined themselves, but the RAMs still get hot and my 12 MHz looks like
crap.
So I socketed all the RAMs and an LS244 that *sorta* was hooked into the clock
generation. That didn't bring back the 12MHz, but the 74S74 under the clock
generation IC gets suspiciously hot and has only floaty garbage on pins 8 and
9. A piggyback gives a crappy signal, but hey, signal! So gotta remove that
one too. As luck will have it, I happen to have the 7474 as LS, HC, F and S.
I tried them all, they all seem to fit, but the S variant seems best for the
job. The F doesn't make a difference and the HC produces a signal that goes
from -2V to +7V and doesn't even get warm while the others do.
And even then the machine doesn't work because the RAMs are toast. Except for
two which somehow survived. Weeell at least my address busses look how they
should now.
And yes, with new RAMs it works a lot better, all the problems I had before
are gone now - except for the random crashes. They're also not so much random
anymore - basically it shows the title screen, it shows a black INSERT COIN
screen, then it wants to go to attract mode and it either crashes right away,
or it starts drawing the level and crashes or it completely draws the level
and then it either crashes or it even brings Bub in in a bubble, draws the
first enemy, corrupts Bub and crashes. No more WORK RAM ERROR message though.
(and guess what - I swapped the work RAM, no change).
So what do I notice? It crashes whenever it wants to switch to the second Z80.
I swapped the ROMs and the PALs out, but the CPUs and RAMs are soldered in.
The second Z80 also doesn't seem to answer its IRQ when it happens.
Btw. I didn't want to steal all the RAMs from the arcade museum (I need 7,
they have 6), so I used some of the 386/486 Cache SRAM I had laying around
from machines I threw away. At least the rightmost four RAMs are set up in a
4k x 16 bit configuration so I could get away with replacing them with two.
Note: In this case, you'll need to generate the ANDed chip selects. There is
conveniently an LS08 right next to the RAM where one gate isn't used.
Unfortunately, we also need to AND the chip selects of the other two and there
is no other LS08 with an unused gate... but there's an LS32 even closer to the
RAM that has an unused gate and also right next to it, the Quartz 74S04 has
three gates unused. And while negating the inputs and the outputs sound like
a timing problem (propagation delay of THREE gates), it appears to work fine,
I tested it with a working CPU board and the only thing I noticed is the
colors seem to be different (my test monitor is monochrome... but NOS!), which
could be because the last two non-dead RAM chips are in the palette RAM now.
Damn, the museum's JAMMA adapter isn't compatible with this bootleg, so I can
forget about the freeze spray. What I can't forget about though is the chip
tester. Which flags the LS08 and the LS125 as bad. The 125 drives a BUS and
looks good. The LS08 is said to only output High on pin 3 which it isn't, but
pin 1 is stuck high, but sometimes there is something resembling a signal.
It has 4.5V for Low though. After some searching I came upon 1) one of the
many vias of that trace is exactly under the big end of a resistor, scratched
the soldermask and the paint off and shorted and 2) the PAL generating the
signal didn't like that at all and the signal still looks like that with the
short gone. Why has that ever worked? This turns out to be a red herring.
And now it stopped workingm only graphical garbage. After some looking I noted
that the two 4K cache SRAMs have an address pin ripped out, both the same.
I had to bend that way up so I could fit another wire under it that mustn't
touch and connect something different to the pin in the socket. Damn, now I
have to throw them away and make new ones.
Now we're back where we were (good luck finding someone who spells that
right nowadays). Since I brought some replacement RAM from the museum, let's
try one of the 8K RAMs as work RAM. With that, it's... worse with graphical
garbage and crashes on the title screen. So I put it back out and, following
an instinct, back in. And now it's running.
OK now let's see how our remaining Toshiba 2K RAMs are... one of them is
pretty screwy, the other one is actually still OK.
So I put the broken 2K and the weird 8K work RAM in an Euro League Soccer
board and of course the colors are wrong and every 8 pixels there is a
shimmering line of faintly flickering pixels. Now that's a super subtle fault
if you use something like that as RAM.
And then he tells me that starting in level sixtysomething, the machine starts
doing weird stuff with the enemies, like they are the wrong kind or don't move
or don't do what they're supposed to. OK I told him to swap the one remaining
original Toshiba RAM with one of the palette-RAMs. He couldn't do that. So I
told him which chips he would have to take out which sockets (by numbers) and
even that was too complicated for him. So we send the boardset across Germany
a few times to cross-swap two ICs. Did it do something? I don't know. I had to
get some idiot to play through the entire game with me just to find out
because some levels are impossible for one player only, so we're standing in
front of the machine for like an hour or two and we don't notice anything odd.
Until, it might have been level sixtysomething, we started being invulnerable
all the time. It's not like collision detection stopped working, enemies die
when they touch us, even the invulnerable white Hurry Up! monsters.
No idea but I really don't want to try and fix that. Forget it! It can't be
the ROM because he burned a Redux version, but didn't burn a new sprite-CPU
ROM so that the game still runs at double speed.

Scramble Formation
OK I bought me this machine. Wait no I didn't. I bought an Atari Pinball and
two weird rare horse race machines (think Quarterhorse with computer generated
horses instead of Laserdisc and real money payout) and got this machine for
free. But it was terribly broken (Line filter smoked, Power supply smoked,
Picture tube smoked (vacuum was gone), with new picture tube the monitor
smoked...)
Back to Scramble Formation. No one knows this game, but I don't care. There is
one other game that Taito released on that hardware. It is Bubble Bobble.
This board has no encryption CPU, so I can apply the Bubble Bobble Redux hack
that brings the RNG back and spells EXTEND instead of EEEEE.
But before that it'll have to work. Turned it on and - Garbage screen with
random garbage that changes on each powerup.
My fancy Chip Tester nags a few chips, but that's normal when testing In
Circuit and my scope says OK. The PAL gets a little hot though.
So the BUS is completely dead, none of the main CPUs do anything.
It's constantly watchdogging, but even when RESET is high, there's literally
NOTHING going on So I checked WAIT and HALT. Both high. WRITE, which is an
active low output on the CPU, is LOW however on one CPU. Piggybacking another
Z80 brought the BUS back to life as soon as RESET goes high.
(I took that CPU from an Apple ][ Z80 card, it's a Z80A that only goes to 4
MHz while BB/SF has Z80B clocked at 6 MHz, so I had to buy some new ones from
Week 11, 2017), but I'm not gonna desolder the CPUs with my cheap solder
sucker.
At the museum, I'm gonna try the PAL and the video board in our Bubble Bobble,
turns out they're incompatible, but that's no problem since replacing the 2
CPUs brought the game back to life.
And just as Tobi (DaddlerTL) wanted to film it, it died with garbage screen.
My first guess was a side effect from cleaning the edge connector contacts,
but the garbage was solid even though the machine was watchdogging and
appearing to try to run. So I checked the chip select lines to RAMs and ROMs
and found one to be stuck entirely low with no high at all, but other chips
getting low sometimes. Traced the signal to an LS157 where this output was the
only bad one. And the machine's back. That was quick.
Just one and a half hour later, no the monitor dind't die even though it's on
the verge of it since like forever, no, the sprite layer circuitry now
produces garbage. The entire foreground layer is complete garbage, the
backgrounds are OK and the game still plays.
So I thought - maybe same problem, chips fighting over bad chip select lines.
But found nothing, the BUS looks super clean as well, but only two ROMs are
ever selected. One of the lines going to the LS138 doing the bankswitching was
dead, it came from an LS374 which got no input on that line, which came from
an LS174 which had an input. Put a cap on. No fighting, even with output bent
up. So after randomly checking chips with the chiptester and not finding
anything suspicious, I thought let's short some of the RAMs data lines.
My trust in the indestructability of Toshiba SRAMs was shattered after Hang On
(see below, I put this machine here since it's basically Bubble Bobble).
So by shorting pins I could find two chips responsible for the foreground
layer. Putting a cap on one changed next to nothing, but next to nothing is a
difference, so mark that chip. Putting a cap on the second one made everything
look a whole lot better. Swapped the latter RAM and everything was back to
normal. The other one is OK.
There's just one problem with the sound. After about an hour, the machine 
starts playing sound effects erratically, like not at all or entirely the
wrong effect, or not playing music, playing the wrong music or keeping the
music playing after Game Over and it degrades the longer you let it run.
(sounds like a job for cold spray)
To me it sounds like a communication problem. There are two LS155 and two
LS374 which I have to check. But that's one of the differences between
Bubble Bobble and Scramble Formation, this one's different.
Turned out that there's more than one cause to the sound problem:
If the 12V line drops (because of bad contacts), the sound CPU crashes and
doesn't get watchdogged and has a rather crappy power on reset that doesn't
appear to work right - you have to turn on diagnostic mode and fire a few
sound commands for the sound section to begin working again (I think it's
code 0 followed by code 6).
So not only do the graphics glitch when you wiggle the edge connector, the
sound comes and goes, too. So since there's a Cabinet to Konami Scramble
connector that has a tongue board plugged in that is soldered directly to the
Scramble Formation motherboard, I also soldered the Konami Scramble connector
to the machine's interface. Now the graphics are solid and the sound only
starts failing after a few hours. But the picture starts glitching again and
I can't get it to go by jiggling this time. But it got worse and worse and
persisted even after cooling off.
So: The entire screen is shifted by almost half of it and there is no more
scrolling going on (and you can see where the new graphics are being loaded
in) and sprites are missing or corrupted. The fault is pretty certainly on the
video board. Trying to diagnose it however turned into a major headache.
The chiptester nags, but I can't stop the clock, so too many signals fighting
the chiptester. Scoping didn't reveal anything amiss, either.
Trying to short circuit certain signals didn't make the screen scroll, either.
Hmmm. if you just MEASURE some of the score layer RAMs, the signals changing.
Looks like a weak BUS transceiver. But: I piggybacked every chip I had. Nope.
I tried freeze spraying EACH chip (except the ROMs). Nope.
Then I noticed that if I tweaked the power supply to 5.6V, the entire screen
flipped and instead of having the left part of the screen black with broken
sprites, now the right part of the screen is black. So I scope *all* the
signals while wiggling the power supply. After many hour I find an LS169 that
appears to be part of a scrolling circuit where only one signal coming in
changes (I see 2 distinct patterns), but the signals going out change more
radically and with more than 2 patterns, and the big jump in how the signal
looked like happens when the output dies completely and does not coincide
with the change in the input! Sadly, I don't have 169s there (or 669).
By the way - the chip passes the test out of circuit.
I guess replacing this chip won't fix the machine, but I will have an entire
picture instead of half of it. Btw. forcing cocktail table mode rotates the
screen without changing the brokenness of the picture at all.
Out of desparation I removed the remaining 169 as well as two 374s that I had
a grudge on even though they're innocent. But if you just leave the two LS169
out of circuit, the picture is stable (won't change over voltage) and
complete-ish.
The four LS669s (schematics calls for six 169s) get a slowly increasing
binary number for LOAD input. No clue why the picture doesn't scroll.
Well whatever. A slight cold kept me home, so I made an adapter LS169 to
LS191. I have ONE LS191. So I can test. Thankfully the LS169 has both enable
pins shorted since the LS191 has only one but a MIN/MAX output instead of that.
And... with the adapter in place of the more important of the two LS169s,
the machine is almost back to normal. The screen looks normal until it starts
to scroll (it SCROLLS again!), every second scanline goes black instead of
scrolling, but I already know that one of the LS169s is for odd, the other for
even scanlines, so replacing this will certainly fix the game.
And the sprites are in the right place.
At least I know that if I get LS169s, the machine will be fixed. Oooof!
With the two new chips everything now works OK, but I need 5.2V for the 
machine to not glitch up. And even then I measure 4.66V at the video board.
The machine probably needs the same fix as Bubble Bobble. Oh wait, I'm
measuring 4.66V at the edge connector already. These alligator clips have
quite the voltage drop.
And just two weeks later we're back at the beginning. No scrolling, broken
sprites and tweaking the 5V I can decide which half of the screen to look at.
Great. And after trying different things for two days (including befuddling
different signals with a 100pF capacitor which turned out that some pins will
make the screen solidly broken) I finally piggybacked the only LS175 and all
the sudden it works. And if I measure the outputs with the oscilloscope while
removing the piggyback, there is *ALMOST* no difference. Diabolic!
And interestingly the chip worked after desoldering. And after a week it's bad
again and I just replaced the chip and it works since.
One weird thing stays: After power on the machine takes much longer to boot
than before. Before, it just started right away. Now it doesn't even generate
a clean sync, the picture will not sync, it'll restart a couple of times
courtesy of the watchdog, then the picture syncs and every reset you see
different garbage on the screen until eventually the machine boots up.
It does that since I socketed the chip behind the quartz, that's a 74S04
I think. And the machine doesn't work with LS, HC or F.
Also... 24 MHz. No joke, most 7400 ICs can't go that fast and the parasitic
capacitance of a chip socket can be enough.

Scramble Formation Monitor
This machine had something out of everything broken.
Logic Board: Two Z80s. Power supply: Line filter went up in smoke. Also one of
the two smoothing/filtering caps was shorted which caused two diodes to de-
solder themselves.
Now the monitor. Turning it on produced a lovely satisfactory 15kHz whine, but
also sparks in the neck. That CRT got aired. Great, I know someone who stored
away a Philips TV from '93 which didn't even have any kind of AV, so is
completely useless nowadays, not just a little, as if you bought a TV in 2019
that didn't have HDMI or DVI or even VGA (or Display Port or any of these
new-fangled connections. Not even Wi-Fi or Bluetooth).
OK salvaged the tube (after watching Wreck-It-Ralph on it). Circuit board
trace went up in smoke. Picture comes on bright white though (with retraces).
Patched up the trace, replaced a shorted-out diode, went up in smoke again,
this time with a dark raster, Diode still shorted again.
So why does that diode blow again and again?  Winni's got the schematics.
The monitor is a Cabel. Looks like a Valvo knockoff.
The schematics show that the heater circuit is a small independant circuit
that also feeds the -20V. So how can that diode blow? It goes to ground.
If the other side of that diode shorts to ground as well, the diode will blow.
Measured it and indeed there was a short where there shouldn't be.
Found out the short is in the CRT itself! Winni says that mustn't be, the tube
has got to be defective. So I went to the storage to measure other tubes, they
all have the same short however! So I re-engineered the circuit so it can work
with a short in this place. But now I have no picture no matter what. At least
it stopped blowing diodes. Took the monitor home where I found out that the
CRT socket is shot completely and barely makes a connection. Fidgeting it
around a lot finally produced a picture. After about half a year!
(I should have listened to everyone else and just thrown the thing away
initially). So I put the monitor back into the machine and of course nothing.
Jiggling the CRT connector finally got me something resembling a picture.
Getting a decent picture is harder than playing the game itself (boy Scramble
Formation is HARD, at least if you try to beat the factory High Score!), as
soon as you start playing, the picture starts fading in and out.
For some insanity reasons I forgot to remove the socket from the original
Philips TV when I threw it away after removing its CRT, so I need another one.
Couple months later the E-Waste Angel was nice to me and got me a few TVs next
door which I'd rather use whole, but when I turned one on, I got a picture,
but also High Voltage Rock'n'Leakage, making this TV highly unsafe to use) -
bad LOPT. So I got its CRT socket. Had to drill some holes into the neck board
for the new socket - no picture, just dark raster. Duuuuuurrrh...
After a couple hours of searching: No -20V. Hmmm. If I put the old socket back
in, it works. The new socket has less pins connected. Hmmm.  Including the one
shorted to ground... So now the short I reengineered the circuitry to is gone.
So... had I swapped the tube AND the socket in one go, I could have had a
working monitor about 3/4 years earlier saving me LOTS of brain damage.
So I created a new short circuit and the monitor got back to work.
Joy didn't last however, I put the machine into the museum, and just a week
later it was dead, they turned it on when I wasn't even there!
Now all I get is a vertical (horizontal) line with some deflection left to it.
So the vertical deflection is bad. The sync generator IC should produce a
needle pulse, but the pulse is more 30/70 duty cycle. Winni doesn't know more,
I'll just buy both the sync control IC and the deflection IC.
The 2578 I bought was DOA, the picture didn't look any different so I put in
the new 3651, but then there was no vertical deflection at all. So I tried the
old 2578 again and then there was a picture.
But I can still adjust vertical size with the vertical hold pot and not just a
little! Also the picture starts jittering after a few minutes and stops doing
that after a few more minutes. I recapped almost the entire board with a net
total effect of zero point zero.
The monitor is really touchy. When cold, the picture gets 50Hz hum on +B when
you REDUCE +B but becomes stable if you turn up +B. When warm, it's the other
way round, it starts wobbling but decreases if you reduce +B.
The main filter capacitor gets really hot, but on the other hand everything in
that monitor gets really hot since this is a cabaret machine with a 19 inch
CRT and NO ventilation whatsoever. But, whatever, it's working.
And I closed the only one hole with a power socket with a line filter.
Out of order. Again. At least it was running for 3 or 4 months.
Switched on, picture collapses 50 times a second. Filter capacitor. Gotta be.
Removed it, it's dead. Put the capacitor from the GDR Poly Play monitor in,
switched it on for a test, perfect picture. Put the machine back together,
Line Fault Switch trips. Huh!? Took a look, the degauss wires had come off and
lean against the capacitor's metal can. Plugged them back in and taped the
cap (it's too big and hangs around on two wires). Power up, picture is
wobbly and the 15kHz whine sounds painful. Weeelll if you have a short to
ground, it often kills one of the rectifier diodes. Measured it, all OK.
Turned it on, BRRRRZZZZZ! (but barely any magic smoke). Now the voltage
regulator is fried, but I doubt it was the cause. So let's hook up the monitor
to a "low" voltage power supply (up to 32) and see what gets hot or how the
power consumption looks like, but as long as the monitor fries itself without
a clear reason, that sucks andn I can't measure it in operation!
This monitor is more often broken than working!
So if I turn it on without the voltage regulator in, it doesn't really
consume any power, which is weird since the schematics look like this tranny
only burns off excessive voltage, so without it, the monitor should get MORE
power. But in fact it does nothing. So I hold the transistor to its contacts
and the current draw goes to ONE AMP at 12V? So there's gotta be something
wrong with the pre-driver, if the pre-driver voltage is whacked, the regulator
may pull the entire B+ line to ground (and fry itself in the process).
The pre-driver seems to be OK though. Gotta ask Winni for the schematics
(again telling the story out of order).
Took it home, measured, tested, can't reproduce. My only guess is that the
big resistor (carbon rod clamped in by two metal cups) got jammed under the
heatsink and shorted to it. So carefully switched the machine on and off.
It twitched in a very promising way, so I turned it back on.
The picture looked ugly. But not power supply ugly, more vertical deflection
ugly (which we already know doesn't work just right). Now the picture is
awfully dim and very very red.
After adjusting V-Hold and tweaked B+ as well as the colors, I get some weak
blue, no green, red is OK. No green, not even if I turn up the screen grid.
Swapped red and green on the neckboard, still no green, so the problem is on
the neckboard. Swapped red and green on the CRT socket, no red, so the tube's
OK, so gotta be the neckboard.
New problem: Every time I switch on the machine, V-Hold randomizes.
That's annoying as hell. I'm gonna need another TDA2578.
Let's go back to the neckboard. Testofon says the RGB driver transistors are
problematic. At least the green one. The blue one works but is out of spec.
The green one is blown. Electrically non-existant. Let's check 
alltransistors.com for replacement. Lots of choice, Reiner's father's 
collection of spares should have something. After 2 hours: Nothing.
Every time I came up with something promising, it was either one number too
high or too low. Another day later I check what other monitors use...
a Hantarex 9000 just lying around uses BF459. The ones I need are NSE459.
459!? But the BF has less bandwith and less current switching capabilities.
But the Hantarex is for a 26 inch CRT. Let's try that, because we have BF459
with X, without X and even some with a green case!
I soldered the remaining working NSE into the green channel. Switched it on
and... BLUE ONLY? Didn't expect that! Looked at the neck board, the RGB wires
are soldered directly to the board because the manufacturer didn't think of
putting holes for the wires in and my soldering was shoddy and the wire
crosses a ground wire and shorted. So I resoldered it in a way that should not
create a short and now we have blue and green, but no red. Red gain doens"t
do anything, so I removed the "new" (old stock) and put the marginal one back
in and we have all three colors back. My testofone finds no difference
between a working BF459 and the bad one I just removed. Weird.
The marginal transistor (whatever happened, the green one died and the blue
one got damaged) doesn't have steep edges anymore, red text looks fuzzy and
red is still somewhat low, even with gain all the way up.
By now the TDA2578 is so wrecked it doesn't start shaking. Or it starts after
an hour. Gotta buy an LOPT at D”nberg, so I might as well buy one, they have
them. And it does the same as the eBay one: No vertical deflection, even
though it is made by Philips in the late 80s. So I guess the monitor doesn't
need a 2578, but a 2579 and is running a mislabeled chip or whatever.
Ah right - I bought a bag full of SF359 transistors from Pollin (probably GDR
surplus) which according to the datasheet have similar specs as the BF459 so I
thought, let's try one of these for the marginal red output stage, because
looking at DaddlerTLs videos the red text on Mappy is illegible and the grid
on the Nibbler title screen doesn't have vertical lines.
So the SF359 appears to have very little gain, I only had to crank the pot
down a little compared to the bad transistor (and I remember having had to
crank it a lot) but at least it's got the bandwidth. Red text is sharp as it
should be.
Also interesting - a colleague bought the machine and put a 60-in-1 board in
it and with that, the picture is rock solid. Looks like the TDA2578 doesn't
like real arcade boards.

Space Invaders Cocktail Table
The manual from the internet was perfectly right with defective op-amps when
sounds are missing. Swapped two and the sounds came back.
Sadly, the UFO hit sound died again. The chip is OK though. I damaged a trace
when desoldering. Fixed it. The sounds work, the machine is dead. But only on
open saturdays. On smaller events it can run for 8 hours with no problems, but
on open saturdays it rarely runs longer than 3 hours. One day it started dead.
Rolled it into the lab, works great. Scope says power supply is fine.
Rolled it back, dead. After the event was over, miracle cure. I just bought a
portable scope just to find out what's going on. There's ripple on the +12V
and -5V lines, which gets slightly worse when the machine crashes, then data
and address bus halt. The CPU socket sucks. There's just noise on one pin.
I sightly push against it and I get a noisy signal. I push even harder until
there is just the signal and the crashed machine resets. I reseat the CPU and
now the machine still keeps crashing, but at least now it gets resetted by the
watchdog so people can still play (for a couple of seconds). On one saturday
the main fuse blew and only the foyer still had power (that's where the
cocktail tables stand). Then, Space Invaders worked great.
Power supply might need new caps. No, that doesn't help. Maybe the internet is
right and the diodes are crap. The ROM chipsockets are bad, too, had some
graphics glitches which went away after reseating.
Swapping the sockets is out of the question. The traces come off but the
solder doesn't.
If the pinout on the mains transformer is correct, a blind person has put it
together. The power supply diodes have finally blown up now, put in new ones,
did nothing at all (well it still works).
I decided to give up and give it a switched mode power supply. No more crashes
since, but the picture is still wobbly as soon as all the machines are running.
Video Pinball's monitor is an identical model and it too wobbles as soon as
there are many machines running. I got both to stop that by slightly reducing
the +B voltage of the monitor.
OK the crashes get worse and worse, so I finally replaced the ROM chip sockets.
It runs more stable now, but has been reported to still crash, so I'll need to
replace the CPU socket as well. Did I mention they used the cheapest sockets
available? With the desoldering station I managed to only damage one pad.
OK since several people told me that the machine still crashes every now and
then, I decided to swap the CPU socket as well (see above). Getting it out was
simple because you could just peel off the plastic and remove the pins one by
one without even slightly damaging the vias.
Put in a new socket, works, but no Base Explosion sound.
Does it get triggered? Jup. Does the buffer LS17 work? Nope. In Circuit Tester
says PASS. Put in an LS07 instead, the trigger now reaches the Op-Amp.
Does it work? Nope. Swapped the Op-Amp. Work? Nah. Are all the traces still
going where they should? Damaged vias? No and yes. Bridged the damaged traces.
Does it work? Nope. Wassup!? Checked the caps, checked the resistors. WTF?
Even worse: Thump-thump is now 4 nearly identical tones instead of the
descending order. Huh... there's 8V on the sound BUS. These come from the
Thump-Thump! Put a cap in between that sound and the sound BUS. Thump-thump is
now back to normal, still no base explosion. If I bend the output pin of the
op-amp up and hold it in a weird way I get it to work, but not reproducable.
I wasted 11 (ELEVEN) hours on that thing and the result is:
All components are working. All the signals are there. All traces have contact.
Still not working! That contradicts my view of the world. Problems must have
a cause! All I can do now is give up. I'm gonna recreate the circuit in my
spare time (as if I had any...), all it needs is ground, +12V, Noise, Trigger.
And then outputs audio.
Well, the recreated circuit doesn't work either, but the background noise is
a lot louder than before.
So. Here's the Mega-Facepalm the world was waiting for: If you put in the
correct IC, it's working. Don't know what made me put a 4066 in there!
I kinda remember having even tried an LM339.
The original LM3900 was defective, though.
And even though swapping chip sockets made it a lot more stable, there still
are sometimes lines through the invaders, but that's a known issue with the
cart interconnect to the daughterboard.

Space Invaders Upright
It's a Deluxe Cab with 3D projection and rainbow color overlay and MAME.
The original mainboard still works...kinda.
The invaders have a line on top and bottom of them. An invader only counts as
hit if you hit the pixel that would have been white in that line anyway.
Then the invaders go right. Then drop down one line, then they go left, then
they drop down... all the way. GAME OVER.
ROM, CPU and RAM are OK (7 of the 8k of RAM are just for the framebuffer).
It is interesting to note that the invader graphics on the title screen are
correct. The thing must have some kind of blitter and that one's bad.
Short googling revealed I was right. There's a barrel shifter consisting of
9 LS151 on the daughterboard (which other purpose is sound - btw. many of the
sounds are missing here, too). One of the LS151 had garbage on the output.
Piggybacked my only spare and the graphics errors went away even though the
scope still showed garbage. So I socketed it and replaced it and the game is
now 100% playable (the invaders don't drop down all the way anymore).
Only most the sounds are still missing and the few that work are very quiet.
I only have one LM3900 left. Put it on top of the one next to the volume pot.
The volume control now works and is capable of LOUD. Replaced the chip.
Which sound effects work? Thump-thump, Base explosion, UFO Sound, 1up. 1up
doesn't beep rhythmically, it's a steady beep. Maybe a 556. Maybe some other
time. Which ones are missing? Shot sound, UFO hit, invader hit. I can swap up
to 2 LM3900 with the other Space Invaders machine. Got me nowhere. Well not
much anywhere. The shot sound ICs are both shot so I could reanimate that
effect. Couldn't get the invader hit to work again, there's just some faint
click in the speaker when it should turn on. Gotta buy some more...
I could get UFO hit working again by socketing the chip above. By piggybacking
the chip, it didn't work. We learn from that, LM3900s can fail short, 
preventing to find the culprit by piggybacking.
So I just fixed the sound effect I didn't want to fix yet. Checked the
schematics and... invader hit and saucer hit are swapped.
So I socketed the remaining two chips and now all the sound effects work
(except 1up which I don't care enough since it's audible).
So only thing left to do is burn a Deluxe ROM set. Did that. Well, it's not
compatible with a standard space invaders board. Weird, why should someone put
a regular SI in a DX cabinet? So it stays running on MAME until Ari makes it
compatible (she's GREAT at reverse engineering).
After the cocktail table gave me hours of non-fun trying to fix my own fault
which I didn't notice, I tried using this machine's sound board.
It made the machine fail completely. I forgot that the reset pulse comes from
the power supply and the cocktail has a switcher which doesn't generate reset
and you have to run a jumper wire across two pins on the sound board in order
to get the machine working.
And yeah right, hacking SI doesn't work, SI deluxe's daughterboard has extra
hardware that's missing on our regular SI board.

Space Invaders Upright (Hi-Score Hanover)
The machine crashes from time to time - sometimes lines through the graphics,
sometimes with frozen "snow" garbage over the (frozen) picture. Most the time
it's running fine though...
...I play it and immediately notice neither the shot sound nor the UFO hit
sound work. According to the schematics they both go through M4 (UFO and
invader hit are swapped), so that one's gotta be it. I completely borked
localizing the chip and without further looking, I can only see 4 rows of
chips so I determine that must be the one and removed M5 instead and wonder
why it still doesn't work. Instead of keep looking, I just piggyback'em all.
And the sounds are back. So I pull one after the other until the sounds
disappear. After pulling the real M4, the sounds are gone. There's still one
more piggybacked chip, I reinstall M4 and pull N?4 and the shot sounds
disappear too. So I swap the real M4 and N4. Now I want to know which of the
three chips I pulled work - turns out none of them.
According to the schematics they're all involved in the shot sound and all 3
were bad. And one of them is half involved in the UFO hit.
So to the glitches and crashes... The board set has a black interconnect which
is older and more reliable - and most the time the machine crashes by drawing
other garbage, like text where it shouldn't be, eg. the PRESS <1> PLAYER START
message in the wrong location or following it up with random letters, or
filling the UFO or Base line with random flickering pixel garbage.
What I noticed is that the daughterboard only gets 4.7V. Jiggling a power
connector brought that up to 4.9V so I turned up the PSU just a little.
The ROM sockets are by AUGAT and the ROMs itself from GI, those with the nice
black and brittle legs. So I cleaned them and replaced the sockets.
The mechanical construction of the sockets is interesting as it's a deviation
of how I remember those crappy sockets, these are dual wipes and grip the legs
the right way, from the thick side...
The CPU has thick shiny legs so I shouldn't need to swap its socket I guess.
With the new sockets, the machine is way more stable but still crashes into
short endless loops that either play the Thump-Thump sound or draw garbage in
the UFO or Invaders line while also making the watchdog happy.
At least it now crashes approximately once in 5 hours which can be checked by
making a high score (>4000, most people don't manage more than 2000) and it's
still there in the evening. But some time after, it's gone.
Also, if you leave one of the lower rows of invaders for last, once it gets to
the bunkers, it'll drag garbage behind, like ours does too, but once it
reaches the right side of the screen, it instantly goes down (instant GAME
OVER) whereas ours just goes left and clears the garbage. That's gotta be down
to an early revision ROM with a bug. I guess.
Or one of the ROMs is flaky. I can't read them because they're incompatible
with my reader and I don't wanna bend up a leg, they fall off if you look at
them weird.

Space Attack cocktail table (Lich Space Invaders bootleg)
Except for replacing the word "INVADERS" with "ATTACK" they changed nothing.
The CRT is done for. Easily visible through the massive burn-in that's even
clear through all the color film that they stuck on the tube.
First I thought the monitor wouldn't work, then I noticed if I turn it off and
on again you could see a line expanding but there was nothing hinting of a
video signal there. But then I turned it on at night with no lighting, you
could see a picture. If you try to turn it up, it immediately starts bleeding
and tearing without even remotely getting brighter. That tube's spent! Baked.
Well apart from the monitor, the power supply had to be checked, they were
nice and put a red, green and yellow LED for the three power rails and they
all light up and the multimeter says they're within spec.
On the bench, the game board clears the screen and starts drawing the score
table and reboots. It reboots at the exact same time and does so abruptly so
that I think it's the watchdog resetting it. So I traced where Reset comes
from (remember: Lich bootleg, has some resemblence to the original).
It comes from a 161 counter which counts. Its reset input doesn't go to LOW
all the way. That comes from a 7442 and some of its outputs look weird, stuck
high or low, floating or garbage. Put a new one in and now the game runs.
Next problem: Player shot sounds weird and base explosion is missing.
If the noise source is missing, that's exactly what it sounds. It's comprised
of a 4006 and a 4030. And even though I rummaged through *all* the museum's
chip storage (where I also got the not-quite-period-correct 7442 that was a
few years older than the machine), we don't have either. The one in my bag was
immediately flagged beyond broke by my chip tester.
Had to buy new ones, now it works.
Next problem: The sound has mains hum and the amplifier chip gets crispy hot.
It's a stereo amp abused as a differential pair, maybe one of the inputs is
stuck because the amp basically works and we all know how awesome the LM3900's
reliability record is.
But, just for fun, I rummaged through the museum's bags of old chips and
wouldn't you believe they have one. And it doesn't even get warm and is much
louder and doesn't hum. It's so loud I had to replace the volume pot because
you could only set it to "no sound" or "instant hearing damage".
We have a bunch of more SI bootlegs by Lich flying around, now I have a
platform to test these.
I stumbled upon two bad CPUs (and even though we have a big box of big ICs and
there's everything and anything, even an RCA1802, we don't have any spare
8080s).
Two boards don't do nothing (there's the shortest of blips on the data bus,
like it reads two instructions and the second is HALT). If you put ROMs from
another machine in, they try to run but on the screen it looks just the same,
garbage. Maybe they have an external interrupt source and/or they're just
incompatible. Or all the ROMs are bad, but Lich uses 2708 which I can neither
read nor write.
Another board starts working, but crashes and then every 16 pixels you can see
dead pixels starting to appear, a clear sign of a bad RAM. Through shorting
the outputs I could circle the problem to two ICs and through piggybacking I
found the bad guy. We have a RAM board from an ancient supercomputer in
storage (just the RAM board, but it is what you think - an enormous circuit
board with hundreds of ICs and probably 64k of RAM) and that RAM happens to
fit in Space Invaders hardware. This is my first time installing one of these.
And the chip works. The game now runs up until the ingame-attract, draws most
part of a single enemy and crashes every time.
But since there are still black pixels visible in some of the bases, I'd say
there's another bad RAM with a different, sneakier failure mode.
Err well yes or something like that. Shorting told me the last row to the edge
connector was the one with the black pixels, so I piggybacked one with another
supercomputer RAM and there were some weird white lines in the picture.
Piggybacking the other RAM in the row didn't do nothing, so I swapped the
first and... now the machine draws 5 invaders, a 6th one on the left of the
screen and then it crashes. Progress! And there's still one black pixel in one
of the bases. And it's still in the same row... was the RAM that I just
replaced even bad in the first place? Let's swap it with the first RAM I
replaced - and I get a gltiching garbage screen, weird, since the RAM seems to
be mostly OK. Well whatever, lemme just swap the last RAM in the row too,
let's see what happens. And the game runs! No more problems! Let's see what
happens if I swap that last chip with the first one - and the enemy shots
don't get erased, tehy stay on screen.
Cool - all the sounds are working - the Thump-Thump sounds a bit sick, but
there's a broken ceramic cap on one of the Op-Amps, that might already be it.

Geiger Electronics Space Invaders Bootleg
We got two of these, both dead, one worse than the other, no hardware shifters.
That held me up a little, because the ROMs aren't compatible, but I'm getting
ahead of myself.
But we have a fully functional Jeutel Space Invaders where the missing Quartz
was the only thing that kept it from running!
One of the boards has fancy purple ceramic RAMs, the other cheap plastic
National Semiconductor plastic RAMs and the crappiest crumbliest 0.1æF ceramic
caps more than half of which had to be replaced.
Both boards have the same exact symptom: The CPU executes one instruction and
HALTs. And I can't find a problem! Swapped the CPU with a working board makes
no difference for neither board, so at least we have a good CPU. Swapped ROMs,
now I get broken chaos. I installed LICH ROMs though, these have hardware
shifters, maybe they're incompatible. I have here a board that looks pretty
much identical (Jeutel as announced), if I swap just ROM #0 over (which is the
leftmost ROM for some reason), the machine erases the screen and draws
PLAYER<1> HIGH SCORE PLAYER<2>
 0000        0000      0000
And further down the screen
   *** C
                    CREDITS 00
Now that's a start - it boots. There's just nothing else happening.
If you replace more ROMs, eventually the *** C stops appearing, on ROM#4 the
HIGH SCORE changes into JEUTEL (so that's how I knew where we got it from),
but it never does more than this. I almost suppose it requires a 60Hz AC
signal to proceed.
Board 2 doesn't even boot if you change ROM#0 - it doesn't immediately go to
HALT now, but nothing happens on the screen except after power up you can see
a bunch of pixels "fading" away like bad RAM loves to do.
Once upon a time I have replaced an LS157 on this board, probably because it
oscilloscoped bad, but that didn't bring the machine back.
Let's investigate the fading pixels. Shorting DO gets me the two lines of RAM
where that happens. Scoping these, you can see some bits move up the area from
LOW to HIGH (the LOW voltage slowly increases). That's not right! Clearly bad
RAM. So we have that mainframe RAM board, piggybacking the first defective
column bottom - still the same behaviour. Moving to the upper chip, same.
Piggyback both - now it's stable. On the second column I only had to piggyback
the bottom chip to get the outputs stable. But it still doesn't run, but at
least I can see it trying to draw stuff where it's supposed to when drawing
the intro screen. So I guess that's either more bad RAM or the original RAMs
are too strong for my piggybacks.
No matter for now as we have run out of 8080 CPUs.

Raiden and PANG have similar problems with the power supply/edge connectors.*
*EDIT that was when there were no further Space Invaders below the Cocktail
table.
Raiden doesn't work anymore at all (sometimes it shows the title screen but
immediately crashes) and PANG is ultra glitchy. Both machines are known for
bad edge connectors, but some bonehead lost the keys to the machines!
At least they both work well in the winter.
After the key for Raiden reappeared I cleaned the contacts and it appears to
work fine now. And then someone bought it without warning.
I did try at some point to find out, why the most rarely used character layers
on PANG has graphics errors every 16 pixels but even though I fixed shorted
IC leads and scratched traces, there is zero indication on the screen that
anything has changed at all...
NOTE TO SELF: Legend has it there's a test mode where you can test all the
graphics layers.
Well we do have the Raiden board in another machine now (where PANG used to
be) and testing something else I noticed the 5V were only 4.2V.

Galaxian
Symptom: No picture, no nothing.
Pushing the Test button yields very silent beeping (as opposed to the 
deafening sound effects I get when I slip with the probe and short lines).
Data and Address bus look good, no conflicts, barely any ringing (after
cleaning the corroded IC pins).
The machine has a quite good self diagnosis menu. If you see flickering
monochrome garbage on screen in an endless loop followed by a message that
stays for just 1 frame, flick the test switch and it stays on, then you can
read BAD ROM or BAD RAM x.
Checking the video signals: Nothing, not even sync. Where would that come 
from?  After going on safari with my Testofone, I found a 7408 which gets
VSYNC and HSYNC. But gets nothing. Where are these coming from? The schematics
leave out such details. Another safari later (not even knowing which is H and
which is V) I had a 74161. We know this kind already - a drop-in replacement
for the 9316 which is used as address generator in all these old ATARI games.
Galaxian has 4 of these.,, my god this is a CPU controlled logic graveyard.
Well '79 they didn't do things much differently than '75...
So clock comes in and 4 clocks with half frequency each come out. Next chip,
nothing. But nothing comes in. Clock input of chip 2 should be ripple carry
out from chip 1. Nothing comes out of there, stuck low. One of the two is bad.
I piggyback one of my LS161 and bend up ripple carry out and it had output.
I still don't know which of the 2 chips is bad. Desoldering with Ari's half-
broken cheap desoldering pump is out of the question since Namco/Midway has
bent ALL the chips' legs around. To be continued...
Settled on clipping off the chip. The first chip was defective. The machine
now has a picture, but it is double and every 4 columns, the next 4 are
missing. And it plays at double speed. The last of the 4 LS161 had garbage on
a signal but still counts normally. Piggybacking brought no difference, I
replaced it which didn't make a difference. Piggybacking the third completed
the picture. Damn, now I killed the working fourth one. Well, replaced the
third. The image is complete and it plays great, better than MAME which is set
on such a high difficulty I have trouble getting to level 2. Here I got to
level 4 barely even trying (the enemies don't start swooping the entire width
of the screen until level 3, in MAME they do that right away).
So one problem persists: The colors are utter crap. Everything's red! The
enemies' wings are dark blue at least. The starfield looks normal, so do the
shells and the ship's shot (which all contain green).
The scope shows that there's signals on R,G and B that aren't just spurious.
But the green signal is a little weak. The resistors are OK. The color ROM
just outputs 4V max. If I bridge the G/B resistors, the enemies' faces show
up. Weirdly, all the enemies are cyan except for the flagships which are
red/green/blue. The player's ship is red/green/blue. All the text is red even
in lines that shouldn't contain a red component.
Two of the color ROM address lines are stuck. If you pull one of them high,
almost everything dissappears. Pull up the other one and all the text becomes
cyan. The tables in that color ROM are bullshit (or are they?).
Whatever I gotta continue on here some other time, after fixing the voltage
regulator which just crapped out.
Naah curiousity got the better of me. Using a bench supply I press on finding
out where the color ROM's address signals come from. There are 5 color RAMs in
line 1 which get warm and have output signals which look ugly. One of the
chips has an output signal which looks really ugly and is pretty much invalid.
Since these RAMs are rare and none to spare, I replace the 74174 which buffers
the RAMs' signals. And it happened to be that 74174, the colors now look
normal, except way too dark. Now I need the remote control for that TV which
replaced the original monitor. Damn, it's in hotel mode and only reacts to on/
off, volume and program. So I turned up the Screen pot at the HV transformer
and now you won't know the difference. The machine was working fine on the
next open saturday and people were hogging it (maybe long-time players found
out it's way easier now - the game gets harder gradually instead of starting
up in hardcore).
Oh. I still have some words to say about the voltage regulator. It uses a
78S05 or something like that which is a controllable voltage regulator used
to drive a high power transistor from a reference Zener diode. It also
requires 12V to work. So on the bench, I replaced the transistor and found the
new one would still not regulate anything and wouldn't even stop at 5V.
So when I added the 12V rail, it started regulating (both 5 and 12V came from
one PSU set to 9V, the 5V were at 5.12V, the farther I upped the voltage, the
closer the 5V got to real 5V. Back in the machine we had clean 5.01V)
So two months later the machine was submitted to Cologne GamesCom. And died
after 1 and a half days with a static garbage screen. At least... the video
portion I fixed still works. CPU still kinda runs, but address bits 8,10 and
one more were stuck low (no short) and the watchdog kept resetting.
And I sunk hours on hours into finding the fault. CPU support chips SEEM to
work. Putting the MAME computer back in was no fun as it was way too hard and
kept dropping you into DOS. As far as I got there's a 74139 which has a signal
on only one output when it should have one on two. Since all the input signals
vary fast and there's a lot of input signals, this might well be OK.
And since the inputs depend on the outputs of another IC whose inputs depend
on the outputs of another IC and so on and so on and the inputs of the last
IC depend on the outputs of the FIRST IC, it could be any. Still the 139 is
the most suspicious. But if I can't fix it, I'll have to dump the ROMs or get
a Galaxian Debugger. (NOTE I didn't have a spare 139).
In our basement we have a lot of Galaxian hardware, but only a UniWarS I got
as far as to know what's wrong (we also have a working UniWarS): If you just
put it in and turn it on, it beeps and you see a static garbage screen.
If I put swap the daughterboard with the working UniWarS, I see flickering
monochrome garbage, Raster and then for a brief instant BAD ROM #3. So... only
the ROMs are bad.
So I swapped the 139. I damaged a lot of traces in the process since this time
I wanted to desolder instead of clip the chip. Continuity test showed all OK
except 5V, I had to solder a bypass. With the new 139, it beeps. And play a
weird broken tune. Then he crashes. If you remove the leftmost chip on the
daughterboard, he does more. The graphics never changes, but he sometimes 
flips the picture (cocktail mode) turns starfield on and off, pulses the
credit counter, and beeps like R2D2 on bad acid. I then checked the chip 
select pulses: It only ever selects the leftmost chip which I just removed,
even if I put it back in. Where are the CS signals coming from? The 139.
Oh, an input is floating. Should go to A13, but doesn't.  I damaged the trace.
Laid a bypass, machine now shows monochrome flickering garbage, then raster,
then something with BAD ROM, but no number. Earlier I already found out the
sockets are wonky and our favourite doesn't even ring the Testofone.
So I wiggled the chips until they had contact and the game was working again.
At least for 5 months (wow...).
Now it's stuck in boot loop, showing flickering monochrome garbage.
Flicking the test switch settled the machine on BAD RAM 2. Since the machine's
behaviour changes and is similar to the one I got with the bad socket contacts,
I suspected the leftmost and again it had no contact on the CS line.
Now the flickering monochrome garbage is solid (as in: stuck in a loop) and
keeps saying BAD RAM 2. In the middle of that static there's a single
character that doesn't change much. Maybe the test is right this time.
Uhh I put the daughterboard in the wrong way after another measurement orgy.
Next week the graphic garbage looked different and it said BAD ROM, and one of
the daughterboard ROMs isn't getting warm anymore. Reading it out reveals it's
dead (well done moron - that is, me). Burnt a replacement, no dice. Burnt a
Galaxian Test ROM, guess what: All tests pass! Not a single error. Video image
identical to Youtube. Good thing we have another Galaxian in the external
storage. This one's dead too showing nothing but a screen of white zeroes.
It didn't have a daughterboard, but a 2k and 8k Mask ROM, but these are dead
as well. With the daughterboard of our Galaxian it behaved similar, but the
flickering garbage stayed longer before BAD RAM 2 showed up. And pressing on
one of the ICs brought the game to boot! So I swapped the daughterboard
sockets and at least the spare board now works fine. Putting the daughterboard
back in ours, the problem was still there and identical. What's your stupid
problem???
The spare Galaxian developed the following tiny problem at the end of the day:
Flagposts were missing and the player's ship had weird colors in its center.
The problem went away after the machine cooled down.
OK back to ours. Since I suspected support logic and not the RAM itself, I
started searching high and low, but couldn't find anything. Swapped all the
sockets but nothing. I gave up and since Ari bought a brand new desoldering
station (a REAL ONE!) I thought it would be better in the long run if the RAM
was socketed. So removed the RAM (no problem with bent chip legs now).
The spare board had the RAM already socketed so I put in the chips from ours
and see - BAD RAM 2. So I replaced the RAM on our machine and the error went
away. The self-test was right and the extra test we burned was wrong.
BTW. BOTH RAMs were defective, but you can't see it! Maybe just a few nybbles.
I put them into my Emerson Arcadia 2001 (the by far best feature of this game
console is that it uses 2114 RAMs which are very sought after and the worst
and also main feature of the console is that it doesn't have a single good
game for it) and it works fine. Maybe they can't cope with the Galaxian's
strict timing. Although my CBM8050 double floppy's self test rejects both RAMs.
I finally burned a Midway Set 1 Z-ROM, so both Galaxian boards now work.
The TV that replaces the monitor in our Galaxian gets senile and often forgets
to turn on or turns on at channel 1, showing static instead of Galaxians.
It has a serial EEPROM which is a 24C08, I only have a 24C32, let's see if
setting Write Protect to 5V after writing sensible data in there and it seems
to have worked.
And what's even cooler, now that I have access to our storage, I could finally
try and get the other Galaxian machine working. It still has the original
delta shadowmask CRT from '79, our only machine to have a delta mask!
And it just turns on fine as if it was turned off just yesterday and not 20+
years ago. Few days later they took it to an expo without knowing that it's
not supposed to be working. And it ran all day. And got played a lot!
And the color purity is way off whack. But I noticed that the purity magnet
assembly has come loose and is "flapping around in the breeze" (not really
loose, just not tight) so I shoved it this way and that until I got a
satisfactory picture. Dynamic convergence is still off, but static convergence
is decent and the purity is good so it'll look nice enough.
Hm if you compare Kirby's fun pak (aka Kirby Superstar) on this monitor to the
DECO cassette system monitor, the latter wins by far.
Now back to the first Galaxian (Namco). Some of the guys got really bored and
developed a new hobby: Shuffling machines around. So Galaxian ended up rotated
by 180ø and the colors were all magnetized again. I told them about it and
they were like "We're not gonna put the machine somewhere convenient for you,
you must fix the degauss!". So let's now do that. Or die trying. Wait I'll
better take that back. The TV controls degauss through a relay. I connect a
switch to pull the control transistor to 5V (through a resistor), nothing.
I try ground, nothing. I also note that the relay does appear to click on
startup, but no degauss! I measure the relay, 0.16 Ohms. OK I give up, I short
the relay, Klonnnnng! And the colors? Look as shitty as before. So the CRT
must have fallen down before and the shadowmask is bent. So much about
"fix the degauss".

UniWar S (Board #2)
So. In 2016 I was struggling with the Galaxian and noted that I found a UniWar
S board I managed to get to display ROM ERROR 3 or something. It was in the
fourth to last box in storage and now we have 2021 and I finally got to it.
Also we finally swapped the UniWar S cocktail table board with Moon Cresta
(about time!) so I have a compatible board to test against.
And I get... the same. Garbage screen. The fault moves with the daughterboard.
Swapped all the ROMs, no difference. Swapped the ROMs, they're all good. Hmm.
Is that LS42 on the board knackered? That's a Hitachi, so it can't!
No difference with a piggyback. Hmmm. Looked at the underside of the board...
there's a blocking cap shorting A7 to +5V. Duuuuuuuh!
Now I'm back to ROM ERROR 3. (spoiler: must've been RAM ERROR 3)
So with the scope and knowledge of the self test I can see it hangs at the
sprite test. So I swap the sprite RAM (2x 5101 or similar) and now it's
working. I stole the RAM from a very faulty Moon Cresta (see below)

UniWar S (Board #3)
This is not the broken board from the story above. This one isn't a cocktail
table version, just a normal Galaxian minus the voltage regulators.
And I get greeted with garbage and a barking watchdog. Completely dead
otherwise. NOTHING on the Z80. Except clock. HALT is high. Reset works as well.
Tried another Z80, game runs perfectly for there being three of five color
RAMs missing. Now I gotta make an adapter for D2125 of which I got two dozen
for a song recently. They do fit on Scramble though.
Interesting... I stumbled across a dead Kangaroo lately, both CPUs seemed dead
and tested dead, but one of them worked just fine - at least in this UniWar S.
But my crappy adapter didn't. I didn't get anything but Low out of the
replacement ICs.
I did manage to write a working test program for the In Circuit Chip tester,
it takes forever, but the ICs pass and without a chip the program fails, so it
appears I did it right. Checking two color channels with two alligator clips I
can see all 3 outputs are identical. Testofon confirms short to ground.
And there is a sync signal present on the video output, even with a standard
conform amplitude (monochrome video for old TVs). So that explains it -
someone did a monochrome hack. The missing color RAMs had their output shorted
to ground intentionally. Now finally the outputs on my homebrew adapter show
something, but it looks ugly and doesn't match what the two others do.
There's even Hsync on the signal! Stupidly, the output is fed back into the
input through some unnecessarily complicated circuitry (OK I take that back
whenever I read a Taito schematic) and the output doesn't look too clean to
begin with. I bet it oscillates and the D2125 aren't usable for this purpose.
Well at least I have more than 4 colors now. And starfield also was disabled
as per hack.
And then I noticed that 27LS00 have the output inverted. So I hacked an HTC04
onto the daughterboard I made and now the colors make sense (certainly a lot
more than four. Every one of my bodge RAMs has a signal on the output)

Ladybug (very subtle hack of the Galaxian hardware, only one extra wire, you
can just put the Galaxian Daughterboard in and off you go)
I get the good old garbage screen with barking watchdog.
When I swap the CPU, I get flickering garbage. In between you can see signs of
intelligent life like 1 COIN 1 PLAY and other short strings that aren't random
and if I put the old CPU back I get nothing of that. But if I bend the
daughterboard around, a little something happens. Enough to suspect this CPU
is good too. But I can't get it to run, so I put in the ROMs from Galaxian,
the machine is just next door.
And somehow I managed to put one ROM in backwards. And on a PC power supply
this poor chip didn't stand a chance... gotta burn another one...
And it still doesn't work. The other machine says BAD ROM now.
Well then let's take a look at the work RAM now that I have the daughterboard
off. Uhh these pins are corroded and the sockets are weird, but both chips and
sockets are made by TI and the sockets look less trustworthy than rotten dual
wipes.
Removing the RAMs left a few pins behind, one fell off from looking at it
weird, so put new RAMs in, put the daughterboard back and the machine just
works. I wonder if these ICs work when I clean the pins and replace the
missing ones. Weird. The video background RAM are the same ICs and that works
perfectly.
Well. Both RAMs are so corroded that my chip tester didn't like either, even
after cleaning and re-tinning the legs. One just doesn't work whatever, the
other one passes every now and then if you insert it just right. The other one
must've been broken before.
And somehow I must've forgot to say the game now works.

Moon Cresta (Board #5)
The board already says from afar "Oh please just let me die!", the hacky power
supply with a way too tiny transistor without heatsink and no direct
connection from the edge connector to the 5V say "Boy leave that thing alone."
Well, by now that's a challenge I can't reject.
And... it's very dead. Nothing stirs. Only the quartz itself is running.
Looking at the clocking chain, there's an LS20 that gets lots of signals that
can stop the clock, one being... VSYNC!? And it's the only thing that's low.
It comes from the LS161 vertical counter. Forcing VSYNC high doesn't make
anything work. I think I'm wrong here. Ah right, there's a branch from the 6
MHz and there's something going into the LS368 buffer and nothing going out.
So I pulled the chip and it was innocent but probably already knackered
because it didn't survive the unsoldering. So no difference with a new chip.
Even if you lift the output. Hmm. The input from the LS107 looks crappy, hangs
around 3V which at such a high frequency can be ok, but the new LS368 doesn't
like it. Piggybacking it makes the chip work. But so does a 680 Ohm pulldown
resistor. And I already noticed the quality of this bootleg board is pretty
crappy, the pads go out before the solder does. At least they didn't bend the
legs around. OK so the LS368's output is shorted to ground. So what do we do?
Yeah - jumper an alligator clip to +5V. And my power supply goes into over
current protection. Let's only connect Ground and hook the 5V straight up to
the TTL output. 1.5 Amps! Healthy! Now I can do the finger test to see what
gets hot. And there's a 7474 at 5C or thereabouts getting nice and crisp.
Removed it, still not working (after patching two traces ripped when
desoldering) (but a tiny bit of 6 MHz visible). So connected the jumper again,
goes into protection again, still 1 Amp, what's getting hot? There's four
LS194 under the character ROM daughterboard. Two of which have been circled in
red. These aren't getting hot. But the two that aren't marked got so hot I
burnt my finger. So I pulled all four. Unfortunately I only have three sockets
and three LS194... at least the 6MHz goes where it should and the machine
produces something akin to a Garbage Screen... after 3 hours of tinkering.
Nah, not really. Just a white raster with borders.
And I forgot my solder sucker again, I'm gonna buy myself another one just for
storage!
First I checked the Reset circuit - the machine doesn't watchdog even though
the CPU is guaranteed not to process any valid data. And here's the deal -
the watchdog disable is shorted to ground from the factory. No solder pad you
can scratch through. But they did populate the 1k pullup.
*facepalm* (or *TILT* what old folks used to say in situations like these)
So I scratched the trace and the LS393 which has been completely useless
before starts doing its job. Interestingly the machine doesn't always watchdog,
sometimes it lands in an endless loop that resets the watchdog (and does some
more, see below).
So I beeped all the traces and checked the chip selects, nothing suspicious,
nothing where a piggyback would disagree, but there's four ROM sockets under
the PRG ROM daughterboard. With no ROMs! The daugherboard ROMs only have 2k
each, making 8k, still 2k less than the original Galaxians. So I gotta burn 4
more ROMs. Unfortunately I'm out of 2716s except for TMS which need 5, 12 and
-5V.
Despite... with the video probe you can see it goes into selftest from time to
time! It tests screen RAM and fails at sprite RAM (which I put into UniWar S,
see above) and we already know it's bad, but still nothing happening on the
screen.
Btw... funny thing: There's an LS245 next to the defective 2101s. Video
probing one side shows the contents of the screen RAM and probing the other
side looks like someone tuning an ancient TV from static into a TV station.
I think it looks totally cool.
Continuing on the video board there's an LS273 that does nothing and is
disabled. The schematics show the enable signal coming via a pullup from the
5V! ANOTHER rail shorted to ground by bad ICs. So the jumper comes back out
and only 80mA? That took a while for one of the three LS157 next to the 194s
to get warm: Without the solder sucker I can only clip the offending pins.
That chip was clearly bad, but the LS273 still does nothing. It gets its data
from the broken 2101 though...
After quite a while I noticed an LS86 next to the five line RAMs get hot.
My fingers over"looked" it because the Line-RAMs usually get hot, especially
when they don't have heatsinks which is the case here. So I expected a hot
chip here and didn't notice there was one more hot chip than expected.
The outputs are dead. The manual states that with this chip bad, all you get
is a white screen as all backgrounds and sprites are rendered as blocks
(remember the self test fills the background).
Replaced the LS86 and the LS157 to no avail except for getting sensible
signals out of the LS86. But one of the output pins is still low (Pin 8), so I
gave it a taste of 5V and whatever was shorted burned through before it got
hot. But according to the schematics the output only goes to an LS157 next to
the one I already removed. Swapped that one and the signals look OK now, but
still no difference on the video.
I went over the board with the In Circuit Chip tester and couldn't find
anything bad.
Now there's only one known problem remaining: One of the line RAM chip doesn't
do anything. And the LS174 in the corner has an output stuck low.
Swapped it and got an invalid signal on this output going up to 2V. Comes from
2P. Swapped it, nothing. The signal from the 174 looks good now and 2P now has
a signal on all outputs, but the LS08 next to it doesn't, and it's the pin
going to the line RAM. But the input looks different than the others, a
piggyback agrees and... well I gotta do the Big ROM hack.
Naah I finally got my order and can now populate the missing LS194.
And finally - a normal looking garbage screen! Let's see if I get into self
test again... yes it does and then I can see a flash of BAD 86 every now and
then. Whatever that means (Edit: I took a photo, it really said BAD 86).
I mean OK there's half the ROMs missing and the sprite RAM is known bad, so
Big ROM hack up next.
Nope. Well OK firstly I dumped the other Moon Cresta's ROMs and the assumed
identical UniWarS. Two of the four ROMs don't match MAME, but they match each
other. So only 2 ROMs to burn (2732 replacing two 2716 each).
So let's see how the chip selects are hooked up. Exactly the other way round
as in UniWarS, but normal again on the main board. So instead of going 1A-4A
on the motherboard and 5A-8A on the daughterboard, it's 1A-4A then 8A-5A.
And what happens? Watchdogging with no signs of intelligent life.
Well what did I search for the cause... While paying a visit to the toilet,
the radio reception suddenly cleared up (these boards give off interference).
I come back, the 6 MHz has gone again. Now the LS107 finally died. Well it
soft-died, every time you piggyback it shortly, the signal comes back and then
you can see the signal getting thinner, getting erratic and then rots at 2V.
Well. After successlessly looking for dead chips I noticed the address decoder
glue logic not doing much at all (and changing when you bend the board).
I stumbled across an LS365 that does nothing and has a floating enable.
There's a wire to ground on the circuit board, but the Testofon says there is
nothing. So I wired it. The patterns still change when bending the board.
The CPU socket's gotta go.
The higher CPU addresses go to a LS139 that constructs the chip selects for
the (almost) entire rest of the address decoder logic and its global enable
almost never goes low. And eventually I noticed the decoders not doing
anything even when it does. Out of desparation I beeped all the CPU pins and
those from that LS139 (8E or 8D or thereabouts - bootlegs don't always have it
in the same spot). And the global enable is pin 6, not pin 5.
Now I gotta beep the working cocktail table. And there it's the same.
So I can forget the Big ROM hack, I'm going semiBig with two 2732 instead of
four 2716.
Which also doesn't work. Well now it reliably doesn't go into self test.
What it does is still random. After changing the CPU socket I noticed that the
ROM daughterboard doesn't get 5V so I changed these sockets too. No dice.
I even put the known good ROMs from the other machine in - nothing.
When you turn it on you see a garbage screen with flickering characters that
get less (damn that looks like a broken trace) until it eventually stops.
It's still arbitrary if it watchdogs or not.
Eventually I went over it with the chip tester again.
The LS164 from the starfield completely fail - I've never seen a starfield on
this machine, just a single color background on two channels... that could be
the starfield being stuck on.
But for that, other chips that I had marked bad now pass perfectly.
Even weirder - I got a 50Hz signal on some ICs even though I removed the clock
IC. Took a while to find out it comes from the bad sprite RAMs.
But - whatever I do I can't move an inch. Both Moon Crestas are incompatible
to each other.
My next idea would be to check everything with the chip tester again and
compare it to the working board, see if I can find inputs FLOATING that aren't
on the working board.
Slowly I lose the will to work on that corpse.
Or I could use the logic generator to generate some read accesses and see what
pops up on the data bus. Or try to write to spaces.
Or I just beep the entire addressing logic, don't find any errors, check the
chip selects for the mainboard ROMs again and notice the order being 4321 8765
and not 1234 8765.
And the machine runs! At least if you replace the sprite RAMs.
I found more than 8 in a box in the basement so I helped myself.
Now there's the problem that instead of a starfield, we got a background
consisting of colorful columns. And some characters are wrong, an address line
being missing. But they're all there. Took forever to notice the address pin
of the character ROM missing the pin on the socket. D'oh!
Where do these colorful columns come from? Not the line RAM, not the sprite
RAM not even the character RAM. They come from the fixed LS194s which get
their data from the CHR ROMs, two of which are floating and if you scope them,
you see an inverted starfield.
Turned out, two paralleled CHR ROMs on the CHR ROM daughterboard were bad, one
had the magic smoke trapped between the die and the quartz window.
If you pull the data lines low or insert working ROMs, you can see the
starfield. So next week I bring my EPROM burner.
Well at least the ROMs on the CHR ROM daughterboard are in the same order,
unbelievable! So I duplicated the ROMs and at least the picture is now error
free (well wait until I put it inside an arcade machine).

Moon Cresta (Board #7)
Originally I put it down as a non-working Galaxian Bootleg, on order of it
being completely dead. I recently ran across it again just after fixing the
above Moon Cresta and now it snapped - the boards look identical, I mistook it
for the recently fixed one - only some of the color RAMs still have their
heatsink and the Model Racing logo is on the mainboard and not just on the
daughterboards. And all 8 PRG ROMs are populated.
Oh - and Watchdog Disable is a jumper wire here.
Removed the wire and still no watchdogging. No video signal either. Wait.
Wasn't there a little blip? So the CHR RAM test must already fail!
So let's test the PRG RAM - OK. Testing the first CHR RAM shows it's OK.
Testing the second CHR RAM shows it's a 3-bit chip now. I put that aside, they
like to fail that way, so now I can make a good one from two bad ones.
And with a new RAM installed, the game just runs flawlessly.

Knock Out (Board #6)
Would you look at that board? In this sea of crappy bootlegs, here's a real
authentic NAMCO Galaxian! You don't get these outside of Japan!
And it's been hacked up to play Crazy Kong. Watched it on Youtube - graphics
wise you wouldn't be able to tell the difference (unless a "real" Crazy Kong
was next to it), but the sound is crap. Galaxian Sound hardware was made
specially for Galaxian and super inflexible (and Namco starkly overcorrected
with the next hardware revision and thus created the first wave table synth).
And what's it doing? Lots of garbage - literally, if you jiggle the wires to
the daughterboard, you can get up to 60 garbage screens per second. So these
connections are garbage, they're so bad, you can't get them all to work at the
same time.
And now I gotta correct - the game's Knock Out, not Crazy Kong. Someone wrote
KONG ??? on the board with a sharpie (a blunt one) and now I can decipher the
??? to say OUT. Ouch. (still definitely KONG, not KNOCK)
So I tried to take care of the daughterboard connection and can't really get
it to work, but I got all the wires to eventually make contact all the time,
and it's Knock Out (an Amidar clone). First time trying, the Attract Mode
commited suicide like 5 seconds into the game, after jiggling the wires some
more, it now doesn't do that and instead gives a nice demonstration on how to
play the game.

Crazy Kong (for real this time)
IIRC there was a garbage screen. The frightening amount of hacky hacks is also
frightening (Tautology club had special offers today). When I was there 3000
years ago I made the note "looks like work RAM".
There's a wire soldered to one of these, making the socket it's in essentially
useless. So I found a via that went to that pin and soldered the wire there
and just swapped both RAMs (and put the daughterboard back on top).
Now I get a black screen with a line of mini Marios and P1 and L=00 flashing
while it watchdogs.
For shits&giggles I replaced a character RAM with a work RAM. The first one
worked just fine. The second one had a bit of garbage stuck in the background,
enough for the program to act up. So that suspicion was correct.
The daughterboard (which looks like stock Galaxian) has funny constructions of
towers of ICs that have sockets soldered onto them containing more ICs some of
which have sockets soldered onto them as well containing ROMs or RAM.
One of which had a pin bent over. Straightened it, put it in, game runs.
Until Mario arrives at the middle of the screen. Then he wraps around to the
left border and starts doing unholy things.
And also there's a wayward sprite in the middle of the screen that looks like
a debug symbol, never changes, never moves.
So I cross-swapped the sprite RAMs and Mario now traverses the path he
normally takes, only he moves very jerkily. So I borrowed sprite RAM from
another Galaxian to find out which of the sprite RAMs is bad (or both, but I'm
pretty sure one was alright).
Now the remaining problems are kinda trivial - all 32 lines there's a line of
color information missing - and the board is missing one line RAM.
(but then it should be missing one part of the entire color information, but
that hints to one of the remaining line RAMs being bad)
I still find it impressive they got DK running on a Galaxian.
But can it run Crysis?
Here I would write about another Crazy Kong on Galaxian (in this case Model
Racing Moon Cresta) Hardware where the Character ROM daughterboard including
ROMs was missing, but I just put the two ROMs per bank on one 4k ROM each
(the one closer to the socket is the lower 2k) and bent up A11 and soldered it
right to the bank select line and that's it, nothing wrong with the board.

Pac Man
I kinda overlooked that one when I took stock.
Kinda awesome, this is a clean Galaxian bootleg board (actually a Model Racing
licensed Moon Cresta board) and it works almost without hacks.
Also interesting is that it isn't watchdogging even though there's nothing but
a garbage screen. If you remove the ROMs it starts watchdogging so here the
Watchdog Disable isn't hardwired in.
RAMs and CPUs are all socketed so I fetch another Galaxian for testing.
Sprite RAMs were fine although they had no output in the board.
Of the 2114 character and work RAMs one each was bad. With those replaced, the
machine is back up and running.
The colors are a bit weird and it's also not a clean bootleg program code wise
(the ghosts are named differently and the colors follow the Ms Pac Man scheme
more closely with 2-color maze walls) and some sprites have stripes through
them. The line RAMs are socketed and swapping them around reveals the bad guy.
The scope says the pulses don't go all the way to 0V, so it's a crap shoot if
the latch detects the signal as a 0 or a 1.
And... that's really strange. A week later, I couldn't find that Galaxian
board I used for testing. I thought I must've put it in the wrong box...
turns out it's not in any of the boxes.

Galaga, Pole Position:
It sufficed cleaning the legs of the Namco custom chips. These were so
corroded, the games wouldn't run or crash constantly.
Uh cool we have 3 more Pole Positions in the basement. PREPARE TO REPAIR.
Put a highscoresaver battery in. Managed to beat the game. Twice!
Now I'm trying to fix our upright Pole Position which runs with MAME.
OK standard procedure swap all transistors and resistors in the battery area
and check all the traces for continuity. I didn't swap the resistors since
they read fine, one trace was bad but the game still boots to the garbage blue
screen. Battery RAM WE is kept down. This is generated by some of the battery
transistors. But that circuit gets a solid LOW as an input so probably that
custom chip above the battery is dead.
It's a PAL or GAL so getting a replacement isn't impossible. Hell we even have
blank PALs in our replacement bin and the GALEP should take care of these.
Out of boredom (ran out of fixable broken machines) I tried to really get it
working and socketed the PAL on all the Pole Positions we had except for the
working one. After replacing the PAL from another board the machine came to
life. It wouldn't turn though since the steering wheel was connected to an old
ball mouse for the MAME PC. After hooking up the original steering board, the
machine was dead, no 5V. There's a 10 Ohms resistor that looks like it has
burned! I didn't smell anything burning though.
If I leave power idling (boards not connected) I can see the voltage slowly
creeping up on the main filter cap, so the problem is even before the voltage
regulator, so the Atari curse of bad wires. And the big rectifier was bad.
It didn't kick the fuse since the wires were already gone. Generously cut away
the wires, but they were still green after 5 cm. So I changed the rectifier
and the wires, the game now works (that charred resistor I swapped before that
of course) and I can steer. Nothing is smoking, so keep fingers crossed.
On the open saturday it was only working for a couple of hours and since then
it's back to Garbage screen, albeit with a black background which is atypical
since they usually have blue background. Swapping chips again brought no
results, even the known bad PAL gives a black background, so I suspected one
or two of the acid soaked ICs like in Qix, but nope.
Ah damn it, I almost got as far as to get it working with a Pole Position 2.
I took the board set out of the basement to help me tracing the traces.
The PP1 had another broken trace. Or two. But fixing these did nothing.
The PP2 gives me the standard blue garbage screen. I tried to swap-test the
ROMs when I found that the PP2's Big ROM ground pin was bent off. Straightened
it, put it in, blue screen full of white "W"'s and the message "RAM 6". Cool!
Looked it up, marked it, put the PP1's other ROMs in and whoopsie, it's 
working now. I can even play! The graphics look weird, but it works.
Putting back the ROMs got me a fully working Pole Position 2 CPU board with a
working Pole Position 1 Graphics board. I can even play all four tracks, but
since it's a PP1 GFX board, it's missing those extra graphics ROMs which
results in glitchy tracks. The roads even are textured, one track has like the
Tetris S-Piece and another something that almost looks like a swastika.
It does look very interesting how the graphics CPU handles this.
So I thought, cool, let's add in the PP2's video board so we have a fully
working PP2 machine, but nope again, it's smoking. And after that the PP2 CPU
board was dead again as well.
OK back to PP1. First I tried fixing the watchdog which didn't work either.
It's frustrating. Nothing works and no clue why.
And since I don't know why the watchdog's output signal won't go low enough to
trigger, I fumbled in a resistor so the capacitor will discharge quicker and
try different resistor values until eventually the watchdog started to work.
Great, I fixed the watchdog and don't even know how. And why.
Surprise: To the beat of the watchdog, the screen now changes from black
garbage screen to blue garbage screen.
Another hot clue is that address line A11 never toggles. This line is used for
communication between sound CPU and graphics (looks like the Z80 does at least
one tile layer and controls which graphics bank to use).
To find out what smoked off this time I put the boards in the known working
PP1 which isn't too easy since the power supply has been soldered directly to
the board (not a bad thing, since normally it tries to push 4 Amps per board
through two small contact pins)
So, both CPU boards are broken, the PP1's video board still works though.
After swapping a couple of chips I noticed that the PP2 does the same exact
crap each time you turn it on. Measured the BUS and found out that DB0 doesn't
hit LOW. Swapped everything that was connected to DB0. And in the end, it was
the CPU. That a CPU breaks down from just now to now doesn't happen very often.
OK, now that the PP2 CPU board is back to life, let's try the PP2 GFX board
again. LED is dimly lit, hum, quickly turn it off!
OK no, I have no clue why the machine doesn't like that GFX board, since it
works just fine on the bench (or appears, since there's no CPU to play with it,
it outputs a nice garbage screen)
OK back to PP1's CPU board. How often does it help looking at the problem from
another perspective? Like, how does the watchdog get reset?
Erm... not at all. Another trace eaten through by the remains of the battery.
Yeah, now I got rid of the black garbage screen, it now stays blue!
So what signals does the chip that resets the watchdog need?
Nothing that I haven't found working before, so now I got the board on the
bench, let's test the LS138(@M6 I think, right next to the Z80) with the in
circuit tester/debugger. Big Great Garbage. Half the outputs is stuck low, the
other is stuck high. A multiplexer mustn't have more than one output low at a
time! Weird, my oscilloscope says the output looks OK. I believe it though,
since most of the other chips on the board fail with less severe problems
(again: a chip failing in-circuit doesn't necessarily mean it's bad).
So, I remove the chip and tested it out of circuit. The same thing on the
screen. Chucked it right away, swapped it and - GAME ON! It's working now!
The music sounds weird though, the Attack phase of the organ (ADSR) sounds
like it uses the wrong sample and it plays a few sour notes, also explosions,
start/finish line and some billboards have lots of bad pixels (which is a VERY
common problem with Pole Position) but FU, it's working now.
But there's one problem I need to fix: The video signal is way too hot and the
monitor doesn't even have a contrast pot. Trying serial resistors in the video
signal cable didn't help at all, so I soldered them to ground and now the
picture is still pretty bright, but stays really sharp.
(I guess in a few years this Monitor will say "GAME OVER" when turned off just
like the other Pole Position Monitors we have)
Back to Galaga... I found a bootleg in a box where the starfield is missing...
and another one with no sprites... and testing another board in this Galaga
machine I found out it runs at only 4.25V. Jiggling everything got no more
than 4.5V so I turned that up a little.
After a few years (like two of them) I got to check the boards in the boxes in
storage. And when I marry the known good CPU board from above with its video
board, I get the road test screen (with the circus tent road) with a few lines
of a complete garbage road and RAM 0 written in the sky.
According to the manual that's a 2114 at 7F or thereabouts. A piggyback shows
subtle differences (also on 6F) (or was it 7F and 8F and 8F was the bad guy?).
So I pulled them both, only the one in position 0 was bad. Now it says RAM 3.
At least it likes the completely corroded TI RAM I put in position 0
(from that Ladybug - I couldn't get the other one to work, maybe it was bad).
So I remove RAM 3 along with some traces. Damn. After fixing those, I got RAM
0 back along with a few signals missing on RAM 0. Beeping through showed more
bad traces, fixing them got me RAM 3 back. Swapping with RAM 1 gets RAM 1
error on the screen. I find it hard to believe that a brand new 5114 from '91
is bad, so I blame timings. Trying a 2114 from Hitachi got me to RAM 8..
That one is on the CPU board - it's the battery SRAM. Flipping the boardset
over got me RAM 31. I flipped it back and forth and so were the error messages
flipping back and forth. Short: The socket of RAM 8 is CRAP!
RAM 31 is one of the 6116s. Chiptester says it's a 7 bit chip now, with a new
one I get RAM 46 (uhh was that 8K?). That one doesn't exist according to the
manual, but we got two RAM 45. So let's check the second one, the scope says
it's dead as a doorknob. Pulled it (pulled only two traces this time) and the
game is up and running. With major graphics glitches. Sprites are missing each
other scan line and have addressing errors, making it hard to guess what
they're supposed to look like if you didn't knew already.
And on top of all that, I leave the room and when I came back, the red is
missing on some overlays. White text, GAME OVER and the own car are cyan now.
TOP SCORE is dark gray on blue and barely readable. The high score table
screen has PINK background now. Red isn't missing, it's just... WEIRD.
All the signals are there, all input bits react to being grounded or pulled up
and all the output bits are toggling. As if the tables in the color PROM got
scrambled (spoiler: they did).
About the sprites... there are two 2114 RAM on 9K and 10K (or was it F?).
These can't get tested by the self test. One seems slightly unhealthy, the
other one is definitely fried. Also there's two LS283 in the same column, top
row, one has a piggyback fighting it and the other has a piggyback changing
the picture on the screen, so I pulled all 4 chips (and 4 traces), each type
had one bad IC. Now the sprites look passable, but glitch sometimes, eg. in
curves. The Roman II from the Pole Position logo glitches too and looks more
like a Roman III.
After searching the color circuit for forever and a day because of the weird
colors, I swapped one of the customs with another identical one and that fixed
these glitches (bad socket). After swapping a regulatable power supply in for
the video board and cranking it to 5.1V, the picture is rock solid.
Not a single pixel where it shouldn't be.
Finally arriving in storage, I consult the salvaged PP2 board and find it's
only missing all the EPROMs but none of the PROMs or custom ICs.
So let's try swapping the red color PROM and the wacky colors are gone, the
colors are now perfect as they should be.
Now all I gotta do is convince Ari to allow me to convert our Pole Position II
deluxe sitdown back into a PP2 and not a Buggy Boy running in MAME on a PC
monitor with a Thin Client PC that barely has enough horsepower for Scramble
and overheats so the game runs in slo-mo after five minutes.

Speaking of it - there's another Pole Position (II, if you can trust the
stickers), that looks different though, maybe original Namco or a bootleg
(but where do the customs come from then?)
It only gives a blue screen and doesn't boot, but seems to try.
Something tells me I should check the Z8002s. One of them looks mechanically
hit by something. Swapped both CPUs and the error changes. We got a PP
salvage board (all the EPROMs, some RAMs and some customs are missing as well
as the PAL which I salvaged to get that other PP going), so I stole a Z8002
off of that and... the one that looked perfectly fine was the bad one.
Weird now - it passes its self test and then it just resets. Again and again.
Hm Joe's Classic Video Arcade has one of them with about the same problem and
here one of the Z8002 sockets was bad. But these beep fine. Also weird, in the
z8002 area 3 ICs are missing - and if you shove all the chips over into a
working board, that one just works fine. Must be a different software version
on the same exact hardware.
And... one of the pins of the left PAL (near the lower Z8002) has no output,
but on the working board it does. The custom ICs are also completely different.
I can only put the chips from the working Atari into the Namco or whatever it
is and see if that brings me any further.

And another Pole Position 2
With the same problem: Eaten by batteries.
Instead of a blue garbage screen I get a broken (but straight) road in circus
tent colors. And it's the same every time. The Z80 doesn't reset so the Z8002s
must start running without being bootstrapped properly.
And the Z80? Dead. With another one it doesn't work either but the signals
look iffy as if the power rails have trouble connecting. How does the socket
look? Green corroded single wipes.
Wouldn't be surprised if that's it.
Well no it wasn't. Now it's stuck in a loop without even reaching the blue
screen of death.
The CMOS RAM chip select never goes high.
That comes out of the PAL above the battery that likes to die.
Whose output has an internal short to ground and that's it. Trying to tie the
signal High with an alligator clip I could get it to show a blue screen after
a while of fidgeting around, along with a RAM 6 error. (O Rly?)
Now I have to burn another PAL because we ran out of original ones.
And yes - with a new socket (which I can change again because I ripped a pin
out...) the machine runs with no glitches or graphics errors. Only the
watchdog doesn't work. That's the LS161 left of the bad PAL and it works -
except for RCO which does the Reset, that's the only thing not working on this
chip. I can fix it, but I don't have to now the system works.
I burned a GAL and put it in there, works fine.

Galaga Bootleg
Most of these work. Here we got one that watchdogs.
The board is very different to another bootleg, but only in size.
FOUR Z80s. One watchdogging, three held in RESET. One of the ROMs has lost two
bits (not lines, bits). Unfortunately, my GALEP stopped being able to burn,
because I replaced a bad transistor with the wrong one (edit: Got the right
one in and now it works - it needs a BC337, a BC547 is too weak).
Putting a good ROM in doesn't make a difference. Beeping the sockets show they
sould be replaced, but should work well enough (for now).
Scoping around the main CPU I found an LS74 with an output stuck high while
the other one toggles. Put a piggyback on, no difference. Bent the stuck
output up, NO DIFFERENCE! Bent both outputs up (to make sure it gets the
input signals right), nO dIfFeReNcE! Two other LS74 (one factory new) behave
exactly the same. WTF?
The "defective" ROM had two bits flipped from a good one. One makes it say
"COPYRIGHT 1985" instead of '81 and then the last byte doesn't match which I
bet is the checksum.
I socketed the LS74 to make sure the signal isn't externally pulled high
(after bending up like everything...) and no it isn't and it's low without a
chip in. And the other output doesn't do anything without a chip either.
With whichever chip I put in. W. T. F.
So. Blind Me didn't see we got more than one of this exact kind of bootleg
board and these have a custom IC replicated in 7400 logic in the empty socket
on the board I was having trouble with.
Out of curiousity I checked the LS74 on the working bootleg and /Q behaves the
same and Q has a completely different signal on it. It's a valid logic signal,
but it's not //Q (don't make me not use no double negation)
So the task is: Scrape the boxes for daughterboards and see if it's in any one
of them (wouldn't surprise me) and if it isn't, make a replica. It's only 6 ICs.
Indeed there wasn't any daughterboard to be found so I duplicated one from an
identical board and now the machine works just fine - except it had a JAMMA
adapter and I wanted to test right away and it didn't take any credits.

Another Galaga bootleg:
This one belongs to a pinball arcade in the bottom left corner of Germany,
forgot the name (but Retrogames Karlsruhe knows who I mean)
It's dead simple - on the sprites, every 2nd scan line is missing.
Moving sprites "interlace" so you can see the missing graphics every other
pixel when they move.
And I went on a crazy hunt even though it's dead simple - there's two 2148s as
line buffers and an LS298 as demuxer. And it had to be these chips I don't
have on hand and my 6514s and 5114s are too slow.
The 298 seems to have scan lines on pin 12 only. On one input it doesn't react
if I short it to 5V, only if I short it to ground. Signal looks good though.
That comes from an LS365 of which I have two, piggyback says OK, so I gotta
wait until my 298 comes.
And guess what, that wasn't it! Even better, if I jiggle the IC, the stripes
go away except for pin 12. Oh wait now they're gone! How? OK let's wait until
they reappear and beep the traces out to compare them with the working one.
And everything has connection. With the video probe you could see the inputs
to the 298 that did not have stripes, the inputs had weird moving blocks
moving with the sprites. The one input where there's stripes doesn't have that.
It's hooked to one of the line RAMs. If you swap the line RAMs, the fault
moves with it. So we have a flaky 2148 RAM that sometimes works.
I could get it to work by shorting the pin to ground through a 75 Ohm resistor
but never got it to last. Maybe I should just try heating up the IC.
Or just replace the RAM with one from the Pole Position spares board, that
fixed it. The original line RAM's pin 12 also flunked in the chip tester.

Stargate:
Machine one day decided to stop working entirely. Burnt connector pins and
dry joints in the power supply.
The game still doesn't make any sound (except on power up you hear the blub of
the amplifier). The sound test switch does nothing and the chips stay cold.
The rectifier for the 5V is shot (why there even is a rectifier I don't know
since the sound board gets DC right away). A diode across it on the rear of
the board tells that another part of the rectifier has already failed earlier.
I accidentally zapped something next to the CPU when testing something.
The machine now randomly resets every couple minutes, but with no warning or
crash (although from the schematics and theory of operation, the watchdog
needs refreshing every couple scanlines). But still playable.
Checking reset signals didn't show anything too suspicious though.
And all of a sudden it went dark. 7V fuse blown, let's see if the rectifier is
rekt and yes it is. Swapped it, party on (let's see if it keeps crashing -
sudden short voltage dips are often seen days or weeks before a rectifier
blows, at least that's what I'm thinking)
And a couple months later it says RAM ERROR 35. According to the manual that's
Row 3 chip 5, swapped it and the machine passes the self test again.
For some weird reason there are a few glitchy pixels almost off screen with
this RAM in, but nothing else looking wrong. Can't be too much broken in this
chip.
And after a long time it failed. Someone came to me and said the game was
unplayable because it had no sound. My idea of Unplayable differs, but without
sound, Defender isn't half as much fun.
When I finally had time to care about the problem, the machine had failed
completely. Stuck in rug test pattern with what appeared to be lots of bad RAM
so I opened the back and was greeted with a missing 12V LED. Err, correction,
the LED was still there, it was just lit VERY dimly.
So I got the board out and on the bench, that's a little more complicated,
there's an LM723 in there, so first we check all the passives and the
transistors and they were all fine, so it could only be the 723.
Let's see what we got... there we got a mystery chip from the 70s in the same
compartment as a few other related analog ICs as well as a real 723.
The mystery chip has something with 723 in its model number, so I'll try my
luck, on a lab supply, what can go wrong, especially now that it's socketed?
I cranked it up and output voltage stopped increasing at 12V. Clear win!
Back into the machine and game on! The mystery chip also barely gets warm.
Just... oh wait, the original problem is still there - no sound!
And I did everything to the board. Tested all the ICs and most components,
swapped the CPU, dumped the ROM until I finally saw that it in fact did
produce sound, it was just getting lost somewhere along the way to the amp.
For whatever reason there are no schematics for the Williams sound board even
though it's used in their pinball machines and arcades from 1980 to whenever
(the pins started using it in '78!). So now let's start going backwards from
the amp. There's a single transistor. And there's signal going in but nothing
coming out. And we have a bigger problem, no schematics and no idea what was
once written on this transistor, it has faded away leaving barely any trace.
So I got another sound board just to take a look at the transistor and it's
something slightly exotic, but the specs scream insanely close to 2N3904, so
I just thought, for the heck of it, let's try a BC187 (or what it was).
We have tons of them, it has a metal can, it's ancient, the application it's
in doesn't have much current flowing (certainly less than the 500mA the 3904
can put out), let's just try that. And lo and behold, it just worked!
And after another couple of months it stopped working. It does a rug pattern,
stays there for a while, clears the screen but the screen won't fully clear,
there's fuzzy lines of random flickering pixels like several bad RAMs, so I
thought the -5 or +12 has taken a dump again, but nope.
Inside the machine, I'm getting all 3 lights on, the voltages check out, error
1-2-1. Changing the chip has no effect (not unexpectedly). A little Google
says to remove chip 1 in bank 3 to see if it shows 1-3-1 now which it did.
But no clear solution, just to check the chips in the bank select circuitry.
So out came the oscilloscope and I went on Safari. Couldn't find anything out
of the ordinary - all the RAMs get all the same signals except Write Enable
and the data outs are individual. They look the same too, so I thought maybe
the LS374 putting the RAM's output on the BUS was bad. Of course I also
checked the one LS32 and the two 7410s and the LS139. There was little
activity on all the outputs, but they all eventually had logic LOWs on them.
So... time for the Chip Tester. And that stated that one of the 7410s was bad.
An output was stuck low. Which is weird, the scope showed output on that pin,
I'm pretty sure it did. But then again if one of its inputs internally
disconnects, it will assume it high and it'll still react to the other inputs
and if they're all high, the output goes low, so the scope shows activity.
I replaced the chip with one from '77 and the machine is working again.
While I was at it, I also replaced the decoder and CPU sockets.

Joust:
Same platform different(?) problem: Doesn't boot.
It stays on a cyan colored screen with about 100 pixel rug pattern on the left
side of the screen. Turning it off and on again gets you the same.
Well the cause was the same as Stargate. Turned it on, worked!
I didn't trust in a miracle healing so I measured the voltage on the ICs.
Most boards had 4.9V, but the one big one read 4.2V. Had to look again to see
the game still works fine. 4.2V! Running stable! So I cleaned the connector
pins, found out that there are two +5V lines only one of which is used so I
put a wire in the second one and lo and behold, it's 4.8V now.
The joysticks have even worse problems with the pin connectors. It's rare to
get one joystick and flap button to function completely, let alone two.
After years and years (first time visit player 2 left didn't work) I finally
gave up and soldered the wires directly to the connector... that should solve
the problem once and for all.

Defender:
Yay we got one without MAME. It's in storage, but I have the key, so I can fix
it. "Why must I fix everything I touch?"
Started by checking power supply rails. They're all there and OK.
Soundboard is running and playing the standard boot sound.
CPU board only blinks the four ROM board LEDs twice. Oh wait, not anymore.
I had the wrong schematics (spoiler: even the correct schematics didn't match
the circuit board more than 90%), but there are definitively frequencies
missing. There is no blanking and from scanline 200 (+-20) on it stops
counting up vertically, so the rug pattern goes to vertical stripes.
And now that I got the correct schematics, that doesn't make me smarter.
Especially now the machine does random garbage. The rug pattern doesn't always
come on, just some garbage with CGA snowing behind it, the ROM board diodes
blink weird patterns and the sound board produces sounds like an AM radio
trying to tune into something).
After fixing Stargate's bad rectifier I could try swapping the socketed ICs.
The upper decoder ROM is defective, making our Stargate's rug pattern go to
lines at around scanline 200 as well (and making the game display look really
weird - but the game runs...). So I checked all the 7400s with my Chip tester
and most of them pass. Two didn't pass but passed once I removed them and two
more are marginal, as soon as I measure the bad output, the chip starts
passing.
Let's take a look at the ROM and CPU sockets. They're single wipes, so let's
replace them. Didn't do a thing. Let's try comparing the ROMs to MAME...
uhh there are so many different versions of Defender and ours doesn't even
seem to be in there... gotta put them through ROMIDENT someday.
Err whaaa? I just put Defender with its ROM board into our Stargate and...
it came right up. Rug pattern, Initial tests determine Unit OK and then goes
to audits (no joystick board connected). It passes the self test even with the
bad decoder ROM, but then it looks funny.
Back in the original machine it's back to crap, not even rug pattern here.
What's different!? Oh wait, the PIA joystick board is hooked up. If I remove
it, the machine boots (stays a little longer in rug test than it should, but
still... it boots and passes nevertheless). The three 4049s all pass the test
so the PIA 6821 must be bad. But someone cleaned the desoldering station and
disposed off the metal plate and filter as well... but whatever it appears we
have a nearly working Defender now.
Ah dammit, our new storage ran out of juice so I gotta test the board here.
I burnt myself a decoder ROM (or both since I don't know which is which and
I can put a couple of these on a 27C16 at 512 bytes each)
So... there's no blanking (btw. I checked the shorted trace that is left open
on some production runs of Defender that are known for not having blanking),
but that doesn't even matter since all garbage colors are in the overscan 
area. Well the game runs. But it's not playable. The controls from Stargate
aren't compatible. And it turned out the PIA on the joystick board was really
toasted. So I tried another one (going for Williams PIAs) which is nearly
perfect, but Player 1 Start appears shorted and up and down are swapped.
So I tried another one which resolved the Player 1 Start problem and only that.
So I just swapped the switches for up and down and everything works fine now.
99.8%. For 100% you have to insert backup batteries and put it on free play.
The monitor has very weak red, had to crank it all the way up and the
convergence was way off, but I managed to fix both. But the CRT will need to
be rejuvenated sooner or later.
And just a few months before one of the Stargate RAMs died, this machine
stopped working even though nobody uses it. There are a few static pixels in
the rug pattern, indication of bad RAM. But all it ever says is RAM ERROR.
So to know which one is bad, I need to go to the service menu. Which is hard
enough so you need a manual.  And self test says 36, which is bank 3 chip 6.
Looks like counting starts at 1 here. But left to right, top to bottom?
Which way round am I supposed to hold the board? Let's try 1,1=top left with
the board inside the cabinet. And... that was correct.
At least it's in the museum now, but still no one plays it because of Corona.
But when we reopened, after two hours I get an endless loop with rug pattern,
flickery pixels and RAM ERROR and enough dead or flickery pixels to suspect
more than one bad RAM chip. Self test says 31, changing it didn't change
anything. Weird. Then I noticed two red LEDs lighting on the power supply.
Should be three. Weird that the RAM works this well missing a power rail.
Well that was quick, just the 7905 regulator and I had lots of spares so that
machine had a downtime of like half an hour.

Defender (from a customer of a customer)
Says "RAM FAILURE" and has missing pixel columns (black).
So I go to the test menu to see which RAM is bad. 14. Swapped it, no
difference, now it says 11. Replaced, no difference, now 18. Replaced, no
difference, now 14 again. Someone has installed turned pin sockets, but these
are the exact opposite of the ones I remember - the chips just float out, so
probably a bad contact. Anyway, I tested all the ICs in Qix, all OK, beeped
them through and it appears all contacts are OK, so that's not the issue.
So let's now triage the RAM glue logic chips with the chip tester. Stupidly,
they all pass. The upper of the three 374s doesn't have data, that belongs to
the upper bank of RAM, which isn't outputting data, but it's not getting fed
any. So I took another look in the schematics again in RAM/Video section.
In between all the tested chips there is a bit of glue logic, especially two
7432 sitting in a very important spot. Checked them and one had an output
stuck low (easily visible because High or High mustn't be Low).
And that was it then and the machine's back up and running.

Super Shot:
(a rarity from Model Racing, monochrome shooting range).
Everything disconnected, "has to be rebuilt", the ROMs are working in MAME...
...what? It works all right! As long as you leave the sound board disconnected.
But then you can't coin up since the coin circuit is on the sound board.
It does count credits if you put a coin in, but never registers. It says
1 credit, but you can't start a game. The circuit to pass the coin signal to
the mainboard is broken. Did they put in an extra part just so it can break?
Put in a credit button that is directly wired to the mainboard.
Sadly, the machine doesn't have any sound even when operating with the sound
module. There are 2 ROMs on the sound board (which even undumpedwiki knows
nothing of) which get brutally hot, maybe it's them.
But before that I must fix the voltage regulator, that only gives out enough
power for the mainboard itself. When you connect the sound board, voltage
drops to 4V and the game crashes. The gun (using potentiometers, not a light
gun) needs to be calibrated). Put the coin board on the bench and adjusted
some resistors, so that the coin signal gets through. But as soon as you put
it back into the machine, it stops working again. (though very quiet noises
in the background suggest that at least some oscillators are running)
Swapped the 7805 can style voltage regulator. My spare sadly had the same
exact problem (5V with no load, drops as soon as you hook up anything to it),
so I just used a TO220 style which works fine.
Trying to map the inputs to the soundboard, I killed the Blink Bulb "you hit
something" effect. But I suspect the Output ICs to be defective. LS374, almost
identical to Q*Bert. Clear is tied to ground, makes it easier to debug.
Swapped the LS374 (4 in total), it did... almost nothing. ONE sound effect now
works, but it rarely ever gets played.
So there are 2 LS374 to drive the two output LS374 which have output clear on
high almost all the time. If I short that to ground, I get an assault of
sound (polyphonic), even some melody. So the sound module should be OK for the
most part of it. Gotta check the interface on logic analyzer.
Tried to dump the 2 soundboard ROMs (74188, no LS, certain no Low Power here).
One doesn't have DB7 set but I suppose it's the one from the socket that
doesn't even have DB7 connected (supposedly the melody ROM), the other one is
super shot. Dead as a doorknob. There are some faint reactions from the
outputs, some millivolts but, crap. The last of its kind. But it's wired into
a resistor ladder DAC, so it probably just contains some kind of custom
waveform.
After killing the machine by putting the edge connector on the wrong way and
fixing it again, I found through ringing that one of the driver LS374 had one
contact unconnected. Put some solder in and see, most soundeffects now work!
The gun bang sound works, beer bottles go bing, light bulbs go bub, balloons
also make sounds. The extended game sound is missing, there's a sound effect
stuck on, but... not much missing.
Well I use a resistor (56Ohms) to drive one of the two sound enable lines
which the LS374 for some reason can't drive (the difference between attract
mode and game is that one sound enable line is high - otherwise it makes the
same sounds in attract mode as it would in the game).
That 56 Ohm resistor even gets pretty warm. The testofone says input is near-
short. The enable line goes to 4 chips: The live rom, one 7474 which had an
input pin internally stuck to ground, swapping barely changed anything), a
7404 and a 7410. Checked the 7410 and the near-short went away and the LS374
can now drive the sound module, the stuck sound is also gone.
The melody is still missing. If I put some 40 hours more into the machine, I
might get it going again, but I'm too lazy and don't get paid for fixing the
machines. It really should be paid.
After someone retreiving the schematics and me looking up I found that the
two ROM chips are completely different types, they're almost mirror images.
One has 5 address pins and 8 outputs, the other has 8 address pins and 4
outputs. And they are in the place of the other chip's inputs and vice versa,
so I measured the Inputs when I tried to read this ROM. Of course, I swapped
the two ROM chips. And also one of the LS374 still had a bad output.
Now the glasses and bottles make a synth glass breaking sound instead of bing,
there is a weird Game Over tune and the Extended Play tune is the cuckoo dance
from Stan&Ollie.
Gun shot sound still works intermittently but I found this was due to a bad
potentiometer.
For some weird reason, this game is really often played. More than 800 games
in 6 months.

Video Pinball
After noticing that the monitor just wasn't put together correctly we fetched
the machine from the external storage and I started trying to fix it.
Status: No picture, reset low, disable the watchdog, the CPU runs but the BUS
looks ugly, lots of pins stuck low (and other chips trying to fight them high).
The stuck pins go into an LS157 that gets awfully hot. Swapped it, BUS looks
better now, but still not running. So we go to my favourite spot, the address
generator. It has LS161, just like the Galaxian. The second one doesn't have
Ripple carry out or does it? These aren't from TI. The third does nothing.
That second one's ripple carry out has... a pulse so short and rare that it
only just barely shows up on the scope when you switch the timebase to near
stand-still, so the third one must be broken. Swapped it (socketed) and the
machine is running again. Sort of.
You can coin up, you can start a game, it beeps, but you can't engage the ball.
The second plunger optocoupler only drops to .9V, that's insufficient.
When the machine is warm, it starts to work and you can play, but no sound.
Some LEDs don't work, was a defective LS174 in the LED control circuit.
Couldn't fix the plunger by swapping chips or transistors, so I changed the
resistors. The pullup was less than 1k, swapping it with a bigger value fixed
the plunger. The plunger opto assembly is identical with the steering from
Night Driver (it's even curved).
Don't know why the sound doesn't work and according to Ari who stayed a day
longer, the picture has crapped up again. I hope it's not the 74H04 which I
suspected for nothing and swapped with a 74LS04 which got me an unsynched
mess for a picture. The problems, 74H04s don't exist. Never have. WTF?
If the chip dies, only MAME will help. I wonder if a HC or an F will do.
The chip is responsible for the clock signal
So.. it wasn't the H04, it was... another LS161. Easy to find since you could
burn your finger on it. Always these LS161, I thought only Texas Instrument's
LS161 suck, these here are all from Signetics.
Well, continue with the sound. After hours on checking the latches I thought,
let's see if Bong works. It does, but it only plays when you coin or in the
service menu. Bong my ass. It's the same sound that PONG makes when you score
a goal! So Bell. No, Bell doesn't play. It's an LS93, which gets clock but
doens't count, even if you pull the inputs low (or was it high?).
Swapped it, Bell works. Still no sound ingame. Turns out Bell only plays when
starting a game (and on booting, just like the old IBM PC). So we're back at
the beginning.
Next morning I found Ari doing what I usually do, with the same amount of
success. The latches don't get clock, where does that come from? On I/O write,
which happens rarely enough. With two probes I could see both be LOW at the
same time. So I swapped the OR gate combining the Write signal with the I/O
signal. Didn't change anything, probably unneccessary, but at least you can
now see I/O write on the scope. Latches are latching. After 3 more hours on
Ari's Logic analyzer I found it should make sound. The sound generator
consists of 2 LS161 (yeah great...). One works, the other... does nothing.
That I could have guessed earlier. So I swapped it and - SOUND! But the sounds
don't sound like those from MAME (if Ari hadn't had MAME running I would have
guessed the sounds are correct the way they are), they sound more like a real
pinball machine. (Note: What's coming from the speaker sounds as much like a
real pinball machine like a piano sounds like a whistle - the wrong sounds
sounded more like a piano compaired to a distorted electric guitar).
So I checked the 3 volume bits, they are always all high or all low. Checked
with the logic analyzer - they get written that way. They put in a feature
and don't use it!
The sounds sound normally, only way too high-pitched. Since I had to swap all
the LS161 except for 2 in the video circuit which appear to work, I swapped
the second sound LS161 and hear, the sounds are back to normal. After more
than 8 hours of work.
After 2 months a small, easily noticeable and easily correctable error shows
up: Every second column is doubled. Looks like H8.
After 3 hours on the scope: It should work! No Idea what the problem is.
Does the RAM get read wrong, does the RAM get written wrong? The machine has
DMA, so the video portion shares the BUS with the CPU and finding out which
signals come from where is hard. The machine SHOULD work, the game is playable,
meaning the lower 512 bytes must be controlled correctly.
The 3 LS161 left aren't the problem.
After despair, I ordered all the chips that could cause the error and firing
up the soldering gun.
Symptom: The H8 line only has the H8 signal which means the CPU never drives
this line. So I put a piggyback on the LS244 that connects the CPU to the
address bus and its output shows that the CPU would drive that line.
Changing the chip did nothing, so I put the old one back and change the chip
that connects the video circuit to the RAM and the machine now works fine.
So H8 permanently drove the address pin. How the CPU could run normally with
that error is a wonder (not a miracle).
Next problem: The screen doesn't feel like working anymore.
Removed it, hooked it up to another machine. Still doesn't work.
Swapped the monitor mainboard with the Space Invaders' Monitor mainboard,
works. Put board back, Space Invaders is dead, the voltage regulator's board's
diodes blew. Put in new (which didn't fix the wobbling, see above), monitor
works. Did I mention I did nothing to the monitor and now it works again?
Back in Video Pinball there was no vertical deflection for a short while, 
looks like a scratchy pot.
Nah, none of the pots is scratchy. You can make the monitor jiggle when it
only shows a horizontal line by jiggling the wires. Anyway. The line stays
longer and longer every time you turn it on.
Looking at the signals with my scope and half-knowledge I found that the
sync signal just slightly dips when there's no vertical deflection and the
signal almost goes to ground when there is. Removed some small cap and
replaced it with the one from the other monitor and it works.
Since I couldn't get it to work with any of my spare capacitors, I jump 
started the monitor with that other monitor's cap, put it back into the other
monitor and put that one back into the machine 5 minutes before the arcade
opened.
Turns out all I did was wrong. There are two signals on that line: The sync
signal which just barely dips and the vertical retrace signal that shows up
as some kind of vertical line on the scope. The PUT (see below) fires when the
two match. The PUT is an MPS 131.
Another problem the monitor has is sometimes no horizontal deflection (and no
HV also). H-Oscillator doesn't oscillate. If you remove the board and put it
back a couple of times it starts working. And it has the same problem as the
Space Invaders monitor - when all the other machines are turned on, the 
picture wobbles horribly.
Now the machine hasn't malfunctionned for quite a while.
Just february 2017 after the open saturday it stopped working. Back to the old
state: No deflection whatsoever. Downloaded the service manual so that I
don't have to poke around an unknown circuit blindly.
The sync separator output doesn't have horizontal pulses, only the vertical.
Swapped it (for a BC548) and what did it? Nothing (well the sync separator now
has H- and V-pulses on the output) but the monitor still doesn't start, even
though this model monitor should start with no signal.
So I take a measurement of the H-oscillator and the monitor starts as soon as
my probe touches the H-oscillator transistor's pins).
As usual, we have no vertical deflection... unless you crank the V-Hold pot to
one or the other extreme. It doesn't sync then, but at leat the picture is
more than one pixel high.
The game's LED drivers are working worse and worse.
After turning all the pots I noted that the picture started wobbling when I
turned up the monitor's +B voltage. So I slightly reduced it and the picture
was stable the next open saturday.
Put the vertical sync transistor MPU 131 Programmable Unijunction Transistor
(some kind of thyristor which was the next best thing in the mid-70's until
they found that no one really had a use for these and by the mid-80's no one
made these anymore) in the Space Invaders' monitor for fun and lo and behold,
no vertical deflection, so this thyristor's bad.
Ordered some off of eBay (for 5 euros a piece! Rare old shit).
Put it in and - what else, no H-oscillation. According to alltransistors the
BC547 fits as a replacement for the MPSA20, but not the BC548, even though the
BC548 is supposed to be backwards compatible. So I didn't care and put in a
BC548 (didn't have a BC547). And see there, it works great.
After a few months, the picture wobbles again a little, but independantly of
the voltage regulator. As long as it doesn't get worse, I don't care.
I also put in a BC558 for the one column of broken LEDs. It didn't work, the
LS42 driving the columns was bad, too. I first thought the 558 was a bad
choice since the LEDs were pretty dim. I tried different resistors, but only
managed to kill the transistor. Then I noticed I put it in the wrong way D'OH!
And after a few hours the machine went haywire. First it started playing with
itself (randomly firing inputs), then it kept rebooting (showing garbage on
the screen, flashing (memory test) and repeating). That stops when you turn on
the test switch and then everything appears to be alright. The chip reading
the inputs I had already socketed, but swapping didn't fix it. And very weird,
when I measured one of its inputs, the problem went away for as long as I
measured the input. It wasn't connected to anything and according to the
schematics it is used so they could use different kinds of buffer chips
(LS240, LS241, LS245). I bent the legs out and the problem went away, but an
input becoming sensitive could mean that this buffer chip is likely to die
soon. And it did, but that wasn't the reason why the machine kept crashing.
Jiggling around with the socketed chip brought it back, but the screen doesn't
work. AGAIN. Same H-oscillator problem. Since it starts when probing, I just
put a smallish picofarad capacitor in there since I have no real knowledge of
the exact workings of an oscillator let alone such a complicated one. And the
monitor works at the moment, so bite me.
Next week, the machine was dead again with staying in test mode and doing
weird things so I thought it's time to swap the chip connected to that line
I removed from that LS14. That didn't do anything except that I had one chip
less to swap since I put this chip next to the CPU to test it and the machine
was completely dead thereafter confirming this LS244 was indeed bad.
But now it's alternating between garbage screen and test mode. These three
LS244 in the input get IN0 and IN1 signals that don't look too healthy and
also the game starts working if I just MEASURE these signals. Where do they
come from? According to the schematics, there is an LS42 next to the CPU, its
label completely unreadable. Let's see what goes in. Looks OK. What comes out?
Dead, garbage, invalid signal, dead, invalid signal, garbage, garbage.
Er... this chip's toast. Swapped it out, machine now runs rock solid.
Also I once swapped the monitor's mainboards again and that jittering picture
didn't change so the fault is with the chassis (power transistors, capacitors).
But the wobble changes when wiggling the big capacitor. So I changed it and
the result is the picture is now wobbling solidly as soon as all the other
machines are turned on and I can't do anything about it, even though when
swapping the caps, I encountered a wire that was broken at a point and has
been like this since the monitor left the factory.
Maybe there's another bad wire in there somewhere.
And now, after a long phase of working, the plunger quit. Only one of the two
signals arrives at the board. I jiggle everything around, nothing.
I disassemble the plunger unit so I can see the signals at the source and
check what's broken and - it works. Never didn't. It wants to fool me.
Whoa! 12/31/2018: Scored more than 680000 points!
I then started swapping monitor chassis semiconductors (there's only three
which can be bad) with Space Invaders, starting with the big power transistor.
I noticed that the transistor's socket holds on pretty weakly and hoped this
could be the problem already. And the machine did the light switch thing
again (plunger). As soon as you disassemble it, it starts working.
OK one month later. The fault did not wander with the power transistor.
The picture just keeps on wobbling, gotta keep on swapping.
Next in line is a diode. That 1950's style free-to-air soldering kept me from
desoldering it, had to cut it out (literally). Testofon beeps different
frequencies both directions. Suspicious but not too out of the ordinary.
Multimeter shows GOOD both directions which is really weird. Could be some
exotic jobbie though. Removed Space Invader's diode (had to cut one side only)
and immediately put it back in after Testofoning it - beeps only one direction.
According to the service manual, it's a small signal general purpose diode
that can do 100V at 0.2A. Hmm the 1N4148 stops at 0.15A. And the 1N4004 is
overkill and maybe too slow who knows. Let's see what Reiner's father's boxes
have in stock. Lots of Zeners and early diodes that look like they could do
600V at 20A but too often only something like 60V at 200mA.
Found a BAY72 which according to its data sheet fits the job description
perfectly. Put it in, it works. I'll need to wait another month for
confirmation though. There was an event on sunday after I swapped the diode so
I went there, the picture was stable, but since the pinball machines in the
basement were turned off, that means next to nothing.
The next open saturday though the picture remained rock solid. Fixed!
It also had *cough* *whispering behind my hand* no problems starting up for
quite a while.
But the plunger problems keep randomly reappearing.
Now the machine was really unstable. You could see that it tried to boot, but
kept crashing and drawing play field elements in random changing positions.
Looked interesting. Jiggled the edge connector, problem disappeared.
But I got an eye on a few bus transceivers...
Nah, the problem's gotta be power supply. I only measure 4.27V even though the
regulator gets 15V. If I jiggle everything around, I get 4.9V, so probably
it's like Galaxian where it needs both voltage rails to regulate the 5V.
Ah, no it wasn't the BUS transceiver. Finally, the light switch problem
resurfaced. This time I thought, let's put in two LEDs like Night Driver has.
That didn't help because according to the LEDs, it works, but with my bodging
going on, it doesn't work at all.
Having Tobi on one side and the scope on the other, I traversed the path and,
unbelievably, the LS14 I replaced years ago is not reacting to its input!
Put a new one in and it works now.
And yet - after a week, the machine still doesn't plunge. The input to the
LS14 doesn't reach a clean low and is noisy around 1V. And finally the light
switch board measures the same and that means it's time for a new transistor.
My Testofon agrees. Let's see if it works. It does, but will it stay like that?
Now we gotta wait.
Nope, still not working reliably. But it got a little more reliable though.
Knowing the machine has an oscilloscope detection circuit (as soon as I drag
in a scope it starts working again) I put two pair of LEDs in front of and
behind the LS14. And one of the behind-LEDs doesn't toggle and one of the
front-LEDs is pretty dim. So first I tried adapting the pull-ups/-downs after
the transistor (ah btw I replaced the transistor) which both didn't do
anything. Instead of LOW I got 1.6V (the other pin has 0.3), so the fault must
be on the mainboard but there's only the LS14 and that's the second
replacement already! I hate faults like that!
So I need to find a way to get the transistors to switch stronger, but how?
With a 620k resistor on the base (I think it was a pull-up, not down).
With 200k I don't get a clean High anymore when the light switch is blocked.
At least it's enough to get the second LED to toggle.
Let's see how long it lasts.
And I swear, if it doesn't last, I'll use this transistor as a driver
transistor for something like a 2N3055 and then we'll see who can drive LOW.
At least a different problem manifested before that could happen, after half
an hour the screen starts tearing and then loses sync into a flurry of pixels.
That looks like the video address generator, afaik there's still an original
74161 in there. During a break I removed the board. No, there were even two
original 161s in there (actually 3 but the third not in the address generator).
It works on the bench though, at least what the scope tells me.
So I put it back and after a few minutes the jittering/tearing starts and a
few seconds later the synchronisation flies away. I pushed on both 161s and
jiggled the edge connector and the picture intermittently reappeared.
After some time I noticed that touching one of the 161s made the picture
return, touching the other did nothing (both were moderately hot), so the bad
one just has to be cooled by a few degrees to start working again!
So I swapped it, put the board back in while all the people were still running
around and it worked for the rest of the day.
And I haven't mentioned what was the cause for the intermittent plunger - it
was the edge connector!

GDR PolyPlay
We have 2 machines. Both the living room TVs (RFT Colormat 4506A) had corroded
fuse holders. The one doesn't degauss, but both have a very sharp and
geometrically correct picture. Astonishing. With one of these machines, the
middle board (with a single 2K ROM) (the graphics card) doesn't make a picture.
The other makes green stripes (the Char RAM board).
Now comes the great Fun, matching up GDR and Russian ICs with their West
counterparts and then checking where the problem is. At least one playable
PolyPlay - they're RARE (a little more than 2000 were made and maybe 40 still
exist - they were made from '85 to '90, using technology that would have been
competitive in 1979).
We have two of the last ones, with a CPU board that has 64k of RAM and a
serial interface to load programs off tape and store them in that huge RAM,
but it never happened and now we have 64k of RAM, only 2 of which are ever
used. Ari should write a driver to use the serial port to link 2 PolyPlays
together, so you can play the car race against the other machine instead of
the CPU - Daytona GDR!)
I gotta fix these TVs. With the first one, the picture gets darker after 30
minutes and it starts smelling, the other made "Bzjt!" every couple seconds,
but it got rarer and rarer and after maybe 2 hours it stopped completely and
now this monitor is quite reliable (don't tell it or it will instantly fail).
In the other TV, the LOPT gets quite hot, not alarming, but the other TV's
LOPT barely gets warm. The horizontal adjustment coil gets very warm in both
TVs.
So here's the curious (translation: boring) story of how I fixed the circuit
boards.
After 3 hours of googling all the chips and get the pinouts of those that 
didn't have one. That wasn't easy at all. There's one chip I didn't find
anything for except "general purpose 4 bit shift register". Sounds like LS94,
but for the LS94 you find even less information. The LS95 looks very similar,
but 2 control signals are different. After Ari brought a list of GDR 
comparative types, I could deduce the LS type (russky->GDR-DS->LS).
95 in the fact. Well whatever.
Another chip I couldn't quite make sure it is what it is, it looks like an
8 bit bus driver like the GDR DS8212 (Intel 8212) I have a bag of. So I put on
one of the DS8212 on that chip (the russian IC should have had one different
symbol if it were a DS8212). If it works, it is one, if not, it isn't.
Putting the chip on made no difference, so it's identical to the DS8212. Good.
So what does the bus do? Nothing. Well after power on it does something for a
short time, then nothing. Enable signals turn the bus drivers offline, the
chips producing these are probebly ok, so I have to measure the rest of the
board (no schematics).
My first attempt to find the bad guy was: Foreach type of logic chip: Check if
the outputs match the inputs. After 30 minutes and 4 chips I realized, this is
a stupid way. My next strategy was: Foreach chip: foreach pin: If it toggles,
is it an input or an output? If no further pins toggle, are these inputs?
Another 30 minutes later and having checked all but 9 chips I found a russian
LS193 whose step-up signal looks stuck high, but looking closer, it became low
very regularly for a very short time. Put a cap LS193 on that russian chip and
I could see the new chip fighting to get the outputs of the old chips to
toggle, so I replaced the chip and the picture came back.
It's a glitchfest, but better than a black screen by far. Some games are even
playable! Now at least, both PolyPlays can be turned on at the same time.
But... you can't put the one on free play and the joystick is busted.
I somewhat fixed the joystick, the shaft was bent and the circuit board had
lots of dry joints.
The credit board is below the power supply and can only be accessed if the
power supply is removed. That helps a lot, since the board must be measured in
operation. So I just stuck a piece of junk in the opto coupler as a makeshift
free play mode (the other machine's free play mode works fine).
To get rid of the glitchfest, I measured the signals, the video SRAMs get/put
out. Looks good. Damn. So I checked all the support chips, nothing. Counters
count, gates gate, flipflops - I put a cap on, no difference. After having
checked all the chips, I put caps on the SRAMs. First two made no difference,
the third brought back the lower half of the picture by some means, it looked
like the upper half of the picture - a glitchfest, but text strings were
readable (before they read "00000???//0000"), so it's gotta be the SRAM.
Swapped the chip (and socketed it) and... the lower half of the screen is
PERFECT! No glitches! The upper one still has lots of glitches. So put a cap
on the last SRAM and weirdly the glitches went away (I'd have bet that the
broken SRAM jams the BUS).
What's left now? Every 8 pixels the red is missing on some characters.
The character RAM consists of 48 of 4k RAM chips (what with the 8 64kBit chips
on the CPU board?), I located the bad guy, which appears to go to tri-state
sometimes even though it never should. Put in a resistor to get the logic
levels right, but now there are red stripes where they shouldn't be.
Ordered spares on eBay (I put in the western chip type and guess what I got?
The exact same kyrrillic chips that were in the machine - western is 1012,
russian is 595KP5 or something like that). 8 chips for 5 Euros incl. shipping.
The stripes are gone now, the machine is working perfectly now.
On to the TVs.
On the TV, where the LOPT and the H-Width coil are getting so awfully hot,
I measured a 5k resistor inside the HOT between collector and emmitter the
other HOT doesn't have and both shouldn't. I guess that makes some amount of
DC flow through the coil, turning it hotter.
Interestingly, the HOT doesn't get hot (ha ha - it doesn't even get warm).
So I salvaged a BC208 and it didn't make a difference. (I swapped a couple of
parts between the 2 monitors and couldn't get a clue as to why the picture
goes darker after half an hour). So I thought - leave it running and see where
the smell comes from. I was just gone 5 minutes, the entire arcade is smelly
with burning plastic, the LOPT drips brown liquid. OK the LOPT is shot.
(ALL the GDR TVs were notorious for their LOPT failing).
So I found a spare part on HR Diemen (a replacement) which was 100 Euros, and
they didn't have it in stock anymore! There's no one on eBay selling anything
that fits. And the other PolyPlay only worked for 3 1/2 hours on open saturday
before showing picture errors that looked like TV and bad capacitor and went
away after cooling off. Someone else had the same symptom and it was a bad cap,
so I replaced one that got something akin to warm and that was that, the TV
works since then, even in the summer heat, for 6 hours.
Finally I got a replacement for the replacement for the replacement for the
replacement of the LOPT. (Original: UHA 106, Replacement: HR5667. Replacement
for HR5667: HR5666. Bought it and got a replacement instead).
It works. The picture's width is still too low even with the width all turned
up it's missing 2cm on either side, but at least it doesn't smell up the joint
and it runs for hours and hours.
Ari burned new ROMs so one machine has different games than the other and now
have only 4 games in common, so that's 14 games, some of which are rare.
OK in the heat of the summer, one machine developed a vertical problem:
The picture slowly retreats into the bottom of the screen (decreasing in size
with the bottom staying where it is). The problem goes away when the machine
is cold. I swapped the vertical deflection circuit boards and the error
wandered, so at least I know where to find it. Good if you have 2 of a kind,
especially when they're rare (also you know where to get money if an emergency
arises).
Swapping the capacitors didn't help so I turned to the TDA2030. Swapping it
also made the fault wander so this chip is bad. Turns out I only have TDA2040
and not even really to spare but no current use for the amplifier they're in
and Google said that the 2040 is just a 2030 with moar powah so put that one
in and the monitor worked great and the heatsink doesn't get as warm anymore.
Hm now the picture has horizontal problems looking like bad capacitor.
Changed'em all and no difference. Swapping the sync boards between both
machines made clear that the problem's on the sync board itself.
After all that troubleshooting it turned out that the H-Hold potentiometer
was set to an extreme position where it would just barely sync. Turning it a
little fixed the problem.
One of the two machines now intermittently shows weird white garbage on
screen which looks a lot like the most significant bit from the character RAM
(which switches between monochrome text from ROM and addressable graphics from
CHR RAM) gets missing sometimes.
And now the machine in Offenbach Digital Retro Park died. First everything
looks normal. Then the picture starts breaking up as if it's losing sync, but
inserting short periods of black randomly mid scanline all over the screen.
These periods of black get bigger and bigger and the picture just falls apart
while the machine still seems to run (reacts to input).
I'm pretty certain the horizontal LS193 is bad. Or the sync circuit. Or the
LS193 gets reset randomly.
Whoa! Ari found the ROMs for the prototype machine that loads the games from
tape.
Well, I was totally wrong about the LS193. Right board, wrong place. Turns out
there is a quartz on the video board, but it camouflages as a ceramic 
capacitor. The GDR 74H04 equivalent gets a little warm. Like those high-speed
DRAMs in Galaxian (that don't need but should have heatsinks).
A hex inverter shouldn't get hot at all! That's suspicious.
But... with other chips it works worse or not at all.
I found out relatively quickly that the cap in the oscillator circuit is
borked, but swapping it didn't help much. It oscillates, but irregularly.
I tried different combinations of capacitors and resistors between cap and
ground, nothing runs stable, unless I touch the circuit traces with my finger.
So I went on the internet and looked at quartz oscillator circuits with LS04s
one of which fit the bill, but the GDR circuit uses asymetric resistors (1.2k
and 420 Ohms) and the capacitor doesn't match (measured 80pF, but is certainly
broken) and the optional cap between resistor and ground is missing.
Eventually I gave up and made it match the schematics from the Web with .1uF
and 2*330 Ohms. The signal looked like crap. Installed the 1pF cap between
resistor and ground and the signal looked great. It even has full TTL 
amplitude. In the machine, the picture was a tad bit shaky, like the wobble
some ancient VCRs sometimes exhibit. Interestingly the picture is much wider
than with the other graphics card, I always thought the monitor was the
reason.
I could probably now swap the card back in. But first let's measure the
oscillator with my cool 10 digit frequency counter. 10.36MHz. Should be 10.0.
Why did I only put in 2 digits after the point? Because the digits after that
were jumping all over the place. Woooooo!
Trying different 74xx04, got different results. With the original D204 I can
see that the frequency is constantly decreasing after switching on, after a
few minutes I thought this quartz can kiss the bottom of my garbage bin.
But I have another, waaaay cheaper frequency counter that just uses a 10 MHz
quartz. Which can be trimmed to 10 MHz with 6 stable digits behind the point.
So four orders of magnitude better! Let's put it in the PolyPlay!
But there I only get 9.996something MHz, but at least 7 digits behind the
point are stable. But how do I now power my garbage frequency counter?
Ah wait, I have a bunch of ancient (MFM) hard drive controller boards whose
mechanisms have long died. One happened to have a 10 MHz quartz. So put that
in the PolyPlay and give the frequency counter its quartz back.
And still 7 digit accuracy. And still 9.99something. I'll probably have to
touch the H-Hold control on the monitor. If I'm out of luck the picture will
flicker like hell (these machines, being made in east Germany, output 50Hz),
but geometrically it will be rock solid.
Well jep, but only in the FOA museum. At the DRP, I still got a black screen.
The CPU runs, though, I get the blip sounds if I move the joystick.
Removed the video board, the quartz wasn't soldered on correctly. Soldered it
on, checked everything with the Testofon, still dead. Plugged it in at home
(just a power supply and a scope), no output, but the chips do something.
Measured the other output. And D'OH, I was measuring (and plugged into) the
wrong connector. Which is weird since I remember that if you mix up the two
identical plugs right next to each other, the monitor gives off a loud 50Hz
buzz.
Well whatever it works now. I did have to readjust H-Hold because the new
quartz was definitely slower.
And now I get word that the white sparkles I sometimes got on one machine have
become permanent - the entire character layer is gone. Everything that doesn't
come from the character RAM is completely white. If you would remove the
Character ROM from the video board, it'd likely do the same.
(only thing that baffles me is that the car race has "shadows" on the road
which are shifted copies of the road border. Might be control signs for the
computer opponent)
And when you put the boards into the other machine, no problems, can't
reproduce. Must be the backplane (all wire wrap).
And I didn't mention that either and just now remember it while I'm synching
these translations with the German original, that I went and traced the Chip
Select line from the character ROM to see where it all went. And one resistor
along the way had its leg bent right up so another component made of metal.
And since I bent it back, there were no more reports of white flakies.
But that was because they haven't told me that the machine still does it! Grr!
Our machine had the marquee lighting fail in the meantime. Only 2 bulbs are
working even though only 3 bulbs were burnt out. Beeping through the control
board found me a burnt trace and a shorted Triac. We don't have triacs, only
thyristors. Gotta order some. Interestingly, now 5 of the bulbs are working
(all broken ones were swapped, so should have been 6 working bulbs).


Qix
So. We have 2 original Taito boardsets, 3 bootlegs that have video boards that
look identical (except for the Taito logo for obvious reasons) and 2 bootlegs
whose existence the internet confirms, but that's about all the information
you can get on these.
The 5 identical videoboards suffered from leaked batteries, especially the
Taitos. The bootlegs had different makes of batteries which were the crystal
kind of leaky and didn't damage the board as much.
Almost all of them only light up the LEDs in the 10100 pattern and all of them
give an impression of death (no picture anywhere whatsoever).
Since none of the boardsets were getting a reaction, I concentrate on the two
original Taitos (especially on the Backup board from our Qix machine where the
battery acid didn't drop onto the ROM board and also f*cked up the connectors).
One had a tantalum cap shorted. Scope says there's heavy fighting going on at
the data bus. Something's pulling the signals low (and chips try to put data
onto that). Pushing and pulling chips around does nothing, ringing through the
SCSI ribbon cables showed one defective, replacing it didn't do nothing.
One of the video boards, the 6845 CRTC doesn't show any signs of life (since
it's not getting any control signals or barely any). The board where the CRTC
runs (or appears to run) you only see some switching artifacts (ghosting).
The monitor itself is OK. There is also no sync signal.
In Karlsruhe, there is another arcade museum which has a working Qix (and not
a single MAME machine). Sadly, they're too far away for just visiting and
rarely have time for fixing stuff (tuesday evening so I gotta take a day off).
But I gotta go there so at least I can swap single boards and see if any are
working or workable in less time.
Oh I forgot to mention: The manual as well as the entire internet says that
the machine won't do anything without the battery. I haven't tried 
substituting the 3.6Vbatt yet.
OK. Preparing a visit to Karlsruhe (had a doctor appointment so I had to take
a day off and it happened to be a tuesday), I stole both Qix boardsets and
took'em home. At home I rang through all ribbon cables (had to clean some 
pins), rang through almost all sockets, tested all the 2114 and 2114 
compatible SRAMs (the machine has 2 ultra fast SRAMs for color RAM as well as
2 ultra low power SRAMs for battery backup, all of them are directly 
compatible to the 2114). I used the Commodore 8050 double drive floppy disk to
test, which even detects the Galaxian RAMs as defective even though they work
fine in the Arcadia 2001. The good boardset's 4116 video DRAM I tested in my
Apple ][. Also, 2 resistors and one transistor in the battery charging circuit
just fell apart when I touched them. The transistor is a 2N3095. Decided that
a BC558 would do the job. So. 2 of the standard 2114 were defective (good
because the weirdo versions of the chips I didn't have, just one set of 
spares) and one of the 32 4116s was broken, too. I just checked the 
replacement which I took from the other boardset, since unplugging, replugging,
unplugging again and replugging of 32 RAM chips is... not much fun. And they
bite if they get single pins stuck in their sockets. I ran around with 3 .1in
holes in my index finger for a few days...
I checked the traces that look as though they had acid damage, but all of them
were fine. The acid did make the solder look ugly (as well as kill the
charging circuit and some ICs in that area), but it didn't eat through the
board's protective layer (the one that make the traces look green).
After all those bad chips (so the work ram of both the main and the video CPU
had a bad RAM) and soldering on a battery, I thought, I'm gonna go to
Karlsruhe, put the board in there and it's gonna work right away.
Well, nope! No and no. In Karlsruhe, they had that 2 board bootleg where the
internet says it exists but no further data (*cough* pinouts *cough*).
So I at least could shoot some fotos of the edge connectors. You can't see
what goes in or out (except for power and video), but you can see which pins
have anything on them (less than half on the upper board connector and the
lower board connector is really just for power).
So back in Seligenstadt, still in high spirits, I put the boardset back in
and... just as dead as before. Still fighting on the BUS, the address bus
shows the CPU is executing 2 instructions in an endless loop. 2 of the BUS
transceivers (LS245) were already socketed and run very hot. Swapped them, no
difference but on one of them, the fighting stopped and the new ones run cold.
On the upper left of the video board, there's two more of them and that's all
that's connected to the main CPU's data bus. If I swap these LS245 out, at
least the main CPU data bus should come back.
Erm...so half-half, not really, maybe as a consequence to something different.
Forget it. Master reset is stuck low. With the other CPU board it starts up
and either crashes after a second or jumps to a random address and crashes,
but at least the LEDs blink (exactly once). And sometimes the machine reacts
to the test button and the LEDs change again. No, I didn't really find the
cause (btw. both CPU boards now have reset working) and all four 6809 CPUs
are working (tested both boardset's CPUs). But at least the following chips
are definitely defective: A 7408 next to the leaky battery (U100 I think),
one of the gates had no output despite input at different frequencies and an
LS299 which looks like an EPROM without a window. WTF!? I wonder what happens
if I replace them... Then we have 2 LS138 (one in the lower left on the video
board, one on the CPU board) which I think don't work as they should. Or not
at all. At least he sometimes bloops when you turn it on (doesn't stop, but at
least you can see from the scope, it's a software defined wave form so at
least the sound CPU appears to be running) and once it even produced a HSync
signal but no VSync and no picture.
Since I don't have LS299 or LS138, I continued socketing the upper bridge head
of bus transceivers (2 LS245 and 3 LS157, all of which were OK and still are),
but the BUS fights existed, because the subsystems were running but both CPUs
were stuck on RESET. So that didn't help.
Would be nice if I at least got the machine to boot into the self test so that
the LED blink codes make sense. What the LEDs currently display is completely
random and changes every time I turn it on. If it reacts to pressing the test
switch is completely random too and if it does, the LEDs change to something
different every time it lets me press the test switch.
Got the 2 chips (the one LS138 on the video board) and the LS299. Put them in
and the machine runs. Almost. To be precise, all the real fun just starts now.
With the new chips I get an unsynched blue screen. The CRTC does output H- and
Vsync. I knew it! The CSync chip is dead (LS86). Changed it, what's on the
screen? A bluescreen. MEMORY ERROR. OPEN COIN DOOR TO TEST. MAY REQUIRE 
SERVICE. It's alive (kinda)!
So I push the test button and the LEDs count up to binary 4 and stick there.
According to the manual that's the color RAM or its circuitry. I could even
get the machine to go into the game. It was running, but the colors didn't
match the slightest. The Qix itself was invisible (except when it left the
normal play area). Even sound worked. And was 100% identical to a youtube
video as well as MAME. SOMETHING works A HUNDERED PERCENT! Even if it's only
sound.
OK so the color RAM. We know it's good from testing it, so it must be one of
the surrounding chips that drives it. There's no A8 signal on the color RAM.
But A7 and A9 are pulsing. This leads to U57 (or was it 55?), an LS157.
Swapped it. Sadly, now there's more broken than before. Gotta check if I
killed or bridged any traces at that LS157 (note: I didn't).
It now crashes in attract mode, every time at the same spot. The characters
are missing half of them and the other half looks broken.
Trying to fix that lead to lots of skyscrapers from a birdseye perspective
appearing on screen (looked kinda like a higher resolution version of TARG).
And CGA snow. Lots and lots of it, especially when he draws something.
After swapping lots of chips back and forth I found, that an LS244 next to the
video CPU has a pin stuck low on one side that pulses on the other. Putting a
cap on yielded fighting on that pin. Swapped it and the picture looks normal
again, except for the CGA snow and now he won't draw any Stix (the lines that
you draw) or fillings. At least the Qix is visible now. To be exact, nothing
that should be white is visible, but white CGA snow. But Game Over and Qix
Kickers appears but are almost unreadable with white lines all over.
(NOTE the machine maps 2 different colors both to the same white - 256 colors 
out of 1024 and the machine uses maybe 12...)
At least the colors that are present look normal.
And the next problem, it doesn't reset anymore.
The main CPU does but the video board's doesn't.
Invalid signal on the reset pin. Swapped the LS244 which is connected to it
(an Input!?) as well as the only other chip connected to it (74C00) and it now
works again. Weird. (well that LS244 had a leg broken which the last owner
tried to fix).
But still no lines, no fillings, weird white artifacts on Qix itself, white
text with white lines on it, CGA-snow... maybe it's the VRAM read-out circuit,
since the self-test now passes. Although when testing the VRAM, you only see
32 different colors instead of 256.
At least you can play if you memorize where you've been.
There's no CGA snow on self test... oh wait - the color RAM maps it to dark
red so you can really just barely see it. This hardware has so many features
so little of which are used in the game... I really feel sad for these chips.
After a relatively short search I found an LS157 at U104 that didn't work
right. Swapped it and all the colors were back and looked right. Despite CGA
snow and weird artifacts around the Qix lines he played normally. But try to
make out the Sparx in that snowstorm. The end of the song was that I spent the
rest of the day putting caps on every chip on the video board with NO change.
I could just guess the ribbon cables have faults. If I can't find the problem,
I'll gotta build myself a SCART cable, feed it CSync as well as one color into
the maybe red channel and hook up the blue and green (shorted together) to a
makeshift video probe and poke it around the board until the snow signals
match.
At least it's playable and I'm gonna turn it on the next open saturday even if
I can't fix it (NOTE: I could fix it before the next open saturday).
It can only be the video data bus (VDB). But which bit?
New strategy: The problem started when I swapped the U55 LS157. What happens
if I remove it? The colors are crap and the snow is gone. So I bend up the
outputs one after the other. At pin 12 I found it. The snow disappears along
with all the colors that were faulty (blue Qix lines, white text, not the
playfield border or Qix logo, that's the different white). So bent up the
inputs (pin 13 and 14) one after the other. The bad one was output by U81, pin
7. This chip's inputs come from U110, but look clean.
(Look out, wrong track). Which bit does U110 (another LS299) serialize?
Bit 1 (starting at 0), which means U11,27,42,60. Look at inputs, it's VD6.
After measuring REALLY everything hooked up to VD6 (a clean signal looks
differently, but all VD signals look just as ugly) and couldn't reach a
conclusion (U34 is the only nonsocketed chip that has VD6 as an output), I
went to bed. In the morning I had the Idea to bend up the video RAM's inputs
and tie it to ground, let's see if the CGA snow persists. And YES! So the
outputs? Don't look too clean. The outputs go into that one LS299 (U110) and
4 LS373 and 4 LS374 each. Improbable that all four of these kind of chips have
the same gate broken (one test was if the CGA snow was only on one of the 
VRAM's outputs). So let's socket U34 and U110, these are the most likely
candidates. Especially since I measured garbage on the inputs of U110 even
when they should have been floating or tied to ground.
So... U110 was to blame. Swapped it, colors normal, video snow and artifacts
gone. Who would have thought that a bad chip leaks signals from one input to
another (or an output)?
So, proudly put the machine back together and it doesn't reset anymore.
Searched a long time. The Master Reset comes through the post connector on
EVERY board. The only board that states that is the Data/sound board.
The video board had a corroded pin for the reset signal.
So. I fixed a Qix with not less than 11 broken ICs (probably 13 since the
LS244 from the data board might have just been bad and ran real hot).
I'm a hero!
OK this machine is EXTRAORDINARILY TROUBLESOME. Basically every other week it
kills another chip.
The first thing that happened is on the first open saturday some yellow pixels
appeared and in other places, yellow pixels are missing. The ones that are
there but shouldn't disappear for a short time after wiping the screen so it
sure is a bad video DRAM. So after the day was over and the people were gone
I turned the machine back on (after it was off for 10 minutes max), the fault
was gone so I waited half an hour until they reappeared and pressed the self
test button. It said "U43" so I swapped it with U42 to be sure and of course
the fault disappeared again and the arcade now really closed so I had to wait
a week. Thank God the fault wandered (and sure enough the self test noted U42)
and I replaced the DRAM from the spare board set. But the bad pixels just got
worse! Tried the next DRAM and half the picture is gone. Machine crashes when
painting the start picture, half the picture is a black/white stripe pattern
and never gets written. The self test says everything's alright (see boardset
2 for explanation in case you wonder). The signal for which half of the RAM to
access comes from an LS139 together with some others (U87). Putting a cap on
fixed the problem so I swapped the LS139 and the machine was fixed again for
a short time. Then I went on vacation. When I came back, the machine was still
working, but reported to have crashed once. Then there was the Circuit Board
meeting and the machine was in a zombie state. You could see that only about
half of the drawing commands from the main CPU reach the video board. It just
continues drawing where it would normally have been if it didn't stop drawing
in the middle. So it looked a lot like an intermittent connection which became
even more clear when jiggling the connectors, because the machine worked for
2 hours before crashing again but sadly it erased the high scores.
Well the scores were not that much to look at... TL TL BSE TL TL TL BSE BSE...
just Tobi (DaddlerTL) and me. I tried to troubleshoot the connection, but the
fault got worse and worse until it became a permanent one. At best, the
machine brought up "MEMORY ERROR" but you couldn't enter the service menu OR
the self test. So I started to investigate the signals at the chips near the
battery, since they were eaten pretty badly and were suspect they could fail
any time. A 7400 had some outputs that didn't really correspond to the inputs.
Tried the 7400 from the Road Champions Coin input, but that didn't work.
Finally found the tube with the new 74LS00s, putting one on brought the game
back to life, so I swapped the chip and for good measure the one next to it
too (U102 and 103). Sadly I destroyed the good U102 when desoldering it.
(but for how long that chip would be good...) at least the machine ran stable
for many hours.
Until the next turn on, then some colors had colorful stripes through them.
Looked flashy, I almost wanted to leave it that way. The stripes are visible
in self test but no errors, so it must be another LS299. They all have normal
outputs, so put caps on. It was the one next to the wannabe-EPROM.
Since this machine keeps killing chips I want to know the reason. I brought a
storage oscilloscope to take a look at the +5V on turn on. There is a -2V/+2V
Spike when you turn the machine on, then the 5V slowly start to rise.
I put a fat diode accross the 5V and the spikes got less.
Sadly, the machine still keeps killing chips. Two weeks later I had white
stripes through the picture, since attract mode runs normal, it's gotta be
another LS299. Self test shows no errors, so definitely an LS299. Good that I
ordered more of them. Measuring the outputs again yielded no useful results,
putting caps on yielded no results at all, so I went to the color bar screen
in the service menu and I could see nice and clean rectangle waves of the
outputs of all the LS299 but one which was overlayn by some fuzzy stuff.
Swapped the chip and the game died. Again. Not again! It doesn't react to self
test button, well sometimes, but it turns the LEDs on and off again and just
crashes. After scoping for half an hour I noticed that one of the battery SRAM
chips gets pretty hot (I first thought I had touched the Color SRAMs).
The other one is cool. Replaced it with the replacement board (and replaced
the replacement board one with a Philips 5114 from 1992 which seems to work OK
but no clue about power dissipation) and the machine works again.
Sadly, my 65thousandsomething high score is gone.
No clue how that could happen, only a surge could kill these SRAM chips since
they're powered continuously by the battery.
So, a week later, hooray it still works. Couple hours later, eeeuuugghh it's
full of dead white pixels the kind you already know and love, another DRAM bit
the dust. Selftest says U27, fast and painless. I'll have to study the
schematics to find out how to replace the chips with 4164's that are much more
common, without wasting too many. It would be ideal if the game would work
with eight of them. At least I know how to get the game working with 16 of
them, wasting only half their capacity.
So the open saturday I wasn't there the machine went offline again, now
showing 4 vertical copies of the screen (so the H-address generator is to
blame). Would be easy if that address generator wasn't so goddamn complicated,
but the self test shows no errors, so it can only be the readout, so it's
either the CRTC or some of the LS86 next to it.
Hm weird the problem hasn't resurfaced and the machine has run for two
consecutive open saturdays on the same exact chipset. A miracle!
Whoa - the machine's working fine for almost a year now. Don't tell it, it'll
break down immediately.
I'm not totally getting how RAS/CAS works (on hardware level) but at least I
could replace two 4116 with one 4164 wasting only half of the RAM's capacity.
Well the quadrupled picture problem came back but went away after hitting the
machine. And it's the only problem the machine has had for a year.
It once again erased scores though - but Q*Bert does it more often.
Someone told the machine it's been running for too long without anything
breakting so after a year and a half (total, not from the last entry) it died.
Black screen. Self test either does nothing or stops consistently at 4.
So... black screen, if I remove the color RAMs it turns white and self test
says color RAM bad. Testing the 2148s in a tester reveals they're fine
(thought so), so there's like 4 or 5 ICs that can be the problem.
Testing with the other video board stops at 5 even though the video RAM test
completes. Which means that it can't communicate with the CPU board.
So I tested every single IC on the board and found nothing wrong. Put the
board back in and the self test half the time won't execute, other half stops
at 4. If I swap the cable between video board and ROM board, it is a little
more reliable.
The color RAM has a bad address bit (high with garbage), The LS157 driving it
is ok though. But now video board 2 is working again with both CPU boards.
Ooookay after wasting another day on trying lotsa stuff and not getting
anywhere I hooked up the second ROM board on the bench and the FI fuse blew
and I sat in the dark. TF? When I put the fuse back in I noticed the board
drawing way more current than before and the CPU getting really hot. Dammit!
Swapped chips with the other board to test. CPU dead, CRTC dead. Testing every
single chip AGAIN showed the two LS244 next to the fried CPU went crispy as
well. Replaced all four defective ICs and reworked the ribbon cable connector
and for whatever reason boardset 1 came back from the dead.
The ROM board from Qix 2 didn't take any damage in the incident.
I bought two 6845 CRTCs from soviet military surplus for rock bottom prices
as well as two 6850s (they didn't have single ones in stock), now I can
restore both QIX boardsets to 100% at the same time.
The 6845s work fine, no idea if the 6850s work, but the board requires them
to boot and it boots, so they work well enough for me.
And after God knows how many years and the pandemic where we were closed for
months, then half-closed (with Qix not running) and then me turning on ALL the
machines to see if any have failed amd Qix didn't (fail), now we're finally
ready to fully open (we did twice before the next lockdown light) and we turn
on all the machines and... Qix doesn't want to work. Weird.
Even weirder... CRTC gets correctly initialized, black screen, all LEDs are
blinking at the same time (at a frequency that's normal blinking for humans,
like near 1Hz).
If you push on the board in a certain way, it starts the self test, gets to
half the video RAM and freezes and/or starts flashing the LEDs again.
That was as far as I got the day before, so I had to put the spare videoboard
in, but I checked and the line to select which half of the video RAM to use is
coming from the last original LS244 next to the video CPU. The chip tester
says CONFLICT on two of its pins (including the one I'm talking about).
Since these aren't hooked up any different, that chip goes bye bye. Out of
circuit the chip tester doesn't say conflict any more, instead we get a clean
LOAD-0V short circuit FAIL.
I'm excited to see if it works...
I didn't find any broken traces (flexing the board).
Nope, not working even though the self test almost does something similar that
resembles passing instead of crashing mid VRAM test.
If I Ohm out the BUS, I don't get any kind of significant values (megOhms)
even though there's a 3k3 resistor network in there, should beep several
octaves higher. So I broke out the network (it was a ceramic wafer,
desoldering was impossible so I yanked it), put in a new one (I only had 10k)
and now it's up and running again.

Qix Boardset 2
Wouldn't it be a great idea to fix the other one since I can now look at each
board individually and know where the fault is NOT?
So I start with the ROM/IO Board. It's missing the QQ3 ROM so I burnt a copy
from the other Qix board. It still doesn't work. But because the video board
works, I can enter self test. It shows QQ9. Removing the ROM left me in a
cloud of green dust. Well that is the board set which was in the machine when
it was stored and the battery leaked exactly onto QQ9. Swapped the socket, the
ROM itself looks stable enough to try, cleaned it, put it back in, works,
hooray.
And here's some weird trivia for you: The spare board set does a self.address
test when testing the video DRAM. It tests the left half of the screen
(the upper one in landscape oriented cabinets), then clears and tests the
right half. The board that was in the cabinet and what I am now fixing fills
the entire screen with some random pattern and it does that FAST, so there's
at least 2 different versions of the self test. Since there is no video of the
self test on Youtube (except for the one that Tobi recorded on this machine),
this might be helpful.
OK the CPU board is next. But I can't get it to work. First I suspected the
LS138 since there are signals at the outputs that just look bad. Swapping it
didn't help and testing the chip in the video board showed it works.
I desparately socketed ALL the chips on the CPU bus, nothing, they checked out
OK. According to the schematic, it could only be the 6821 so I socketed that
one too, but to no avail. Interestingly, without that 6821, the machine draws
the start screen (in cocktail mode) and crashes.
I then tested all interface chips to the sound CPU/PIA, nothing.
Well... it was the 6850 serial interface controller that is just there for
debugging. The machine won't work without it (even though I was sure it did)
and this one is bad. I can't replace it, but I put a sticker on it, so if the
CPU board fails, we have a spare that can be brought to life with just one
chip swap.
Now the video board is sure gonna be a bigger construction site.
Well... no. After the mandatory resistor/transistor swap in the battery
charging circuit, without swapping a single chip I got the same bluescreen as
the other board, but with sync. Note: Qix DOES run with no battery, but you're
pretty much limited to the self test. Getting the actual GAME to run with no
battery is a hard trickery of learning the right timings to turn the machine
off and on again, since it is only possible to enter the game from the self
test when the self test passes.
So I got the game and the flood fill algorithm just totally fucks up since
there are some video DRAMs missing. So I fetched a bootleg Qix with a
compatible video board to salvage its DRAMs, put them in and... the same
symptom I had when the other video board showed the first signs of life:
Random LED patterns followed by SOMETIMES other random patterns when you push
the test button. That's gotta be U100 next to the battery and also matches
other people's repair logs (that IC really doesn't like battery acid).
Swapped it (LS08), still lines in the picture, but an improvement.
Self test quits in the VRAM test which means the Color SRAM is OK, so the
weird colors must stem from U103 (LS157). Isn't it great when there are two
boards with the same exact error?
Since I need the Service Menu Color Bar diagram to test the LS299s, I must go
through the hassle of finding every bad VRAM chip. That was a wild goose chase,
I got one address for a bad VRAM chip, swapped it, next address and so on and
a total of FOUR chips were bad. Finally I see the color bars, but only 32 out
of 256 colors. So scoped the LS299s, one had no output, put a cap on and bent
the output up, it pulses, swapped the chip and the game works flawless.
Now I'm a super hero, we got 2 working Qix boardsets. Original TAITO!
When I had to use this set's video board as a replacement for a month (see
above) I noticed all the colors are garbage, almost everything is black, self
test passes so let's check the LS299. It was the one next to the battery.
One little bit of information: We have three Centuri bootlegs that have video
boards identical to Qix in every way and yes they're compatible. Mostly.
For some reason they use a slightly different horizontal frequency so I had to
readjust the monitor.
One board was already salvaged for video RAM (and apparently for CPU and CRTC)
another one is missing the CRTC, with it it throws a video RAM error (but
starts painting the test pattern) and another one is complete and also
completely working. Great!

Qix Boardset A1
Here we have one of those Qix boards the internet doesn't know the pinout for.
And here's why: EVERY BOARDSET HAS A DIFFERENT PINOUT (including numbers of
contacts). The Insanity!
The boards look absolutely identical. Except for the audio amplifier and the
edge connector. Like WTF?
This board set is dead. So I use the other one to fix it. And make lots of
mistakes doing so. For example I swapped the CPU board even though both CPU
boards work, making me mark one 6809 (which I swapped) as defective even
though they both work fine.
Well whatever. Both CPU boards and interconnect cables work.
So. Strangely it initializes the CRTC. If the board was completely dead, I
wouldn't even get anything out of the video port. But I do get sync.
Correct sync, to be exact. But then nothing happens and the screen stays black.
Test point 9 (the schematics only has 8, so no clue) had floating while the
working board set had Low. Took me a while to trace that. It comes from a PIA
from the CPU board. Great. Back to the video board. After a while I notice /WE
from the battery RAM is High with periodic garbage on 3V, so invalid signal.
After even more time searching I noticed the supply voltage rail on that RAM
(and the HC00) looks so too. Both have an identical circuit. Well if you short
the battery + terminal (the battery is missing), the board starts, passes the
self test and runs with no further errors.

Qix Boardset A2
This is fully functional [cue Twilight Zone music].
After noticing everything works, I want to check which pins have the video
signal. Turns out none of them, not even sync. But the sync is being generated,
the palette RAM outputs signals, the LS273 latching these has output, but the
LS17s after that don't. They're open collector. Hooking up a pullup resistor I
see output, so it's not the buffers. All the transistors after the buffers are
Low. Checked the schematics, there's ONE transistor Q1 supplying all the
others. And I don't measure jack crack. They had the wrong sort of transistors
on hand and had to cross the legs. And they soldered them in short, so the
transistor just lost his legs staring at it. Replaced it with a BC548. That
one you just have to insert the wrong way round, no crossy leggy, and it's all
back up and running.

Slither (uses Qix video Hardware)
We got three of them. And one's already salvaged to fix the first two Taito-
Qixes. And... none of them had any noteworthy problems.
Getting them to run outside of their original machine (which was wired WAY
weird) is a hassle though and gave me a few hours of headscratching.
- Reset is generated externally (like Qix). So you have to hook up an
alligator clip to the Reset line, then after each power cycle you hook it to
+5 and then to GND just to get the machine going.
- Full self test only works when you bridge the cash box switch (like Qix).
The cash box switch is a single two pole connector below the monitor connector
on the video board.
- The self test button becomes unresponsive once the machine booted to Audit
mode.
- The game does a "small" selftest on powerup and then either goes to Audits
or first draws the game's background graphics and then goes to Audits.
In short: You have a very small window in which the self test button actually
does something.
And the following came out of that:
All three boards are fully functional.
Board 1 had a bad ROM. Really bad. It had 18 occurrences of bits flipping from
1 to 0. So you can't burn it over. So I had to burn another one. But my GALEP
had blown a transistor and can only handle 5V...
Board 2 had a bad Video-RAM. The error codes are different from Qix so I had
to download the manual.
Board 2 also had two bad ROMs. One had the same problem as before, so I had to
burn yet another (and I'm running out of erasable and writable 27C16s), the
second luckily had ONE bit flipped to 1, so I could just burn it over.
The third board as I said was salvaged and missing the 6809 and 9 (nine) video
RAMs. And if you replace'em it just works.
So... what am I going to do with three Slithers? And what kinda game is it?
Well... it's Centipede. Not just like Centipede. It's Sue Me close to
Centipede. It has some nice background graphics though and does demonstrate
Qix' graphics capabilities a lot better than Qix. But... it's Centipede.
The spiders exist too and you control it with a trackball.
Problem: We're still under lockdown, it's Christmas (second day of holidays)
and from the box of broken game boards (one of which was actually working)
there's only ONE board left to fix... and we're running out of 4116 RAMs.
So what to do?
I would love to replace the 32 4116 in Qix HW with just 8 4164s.
To do that I have to generate a correct M7 signal.
Until now my opinion is to MUX ODD/EVEN and 0/1. Only... M0 to M6 are also
controlled by the CRTC - the CRTC can't know which 16k page is currently
active. And even weirder - all 32 RASes are shorted together and the CAS per
column (4 ICs). In short - I have to keep studying.
OK I did some studying and... the CRTC can only see 16k. The data output are
all individual, so the CRTC sees 16k address range at 32 bits and the CPU sees
64k address range at 8 bits. So the read circuitry differs drastically from
the write circuitry and I have to make sure that
- the CPU can write all 64k (I'm pretty sure I got that down)
- the CPU can read all 64k (--ditto--)
- the video circuitry can read all 64k (it can't - every chip is responsible
for one pixel in a 4 pixel wide area, shifted by the chip number)

Gorf/Space Zap (Astrocade)
First of all: Space Zap runs great since a long time.
But Gorf doesn't and it has the same hardware, so we can swap lots of boards.
So I put one identical board after the other in Space Zap and see if it's
still running and if it isn't, put the old board back and swap in chips with
the Board Under Test until it fails.
We have 2 Gorf PCB cages. Sadly, both have the same custom IC failing: a C2860
custom. It's hard to Google and probably even harder to eBay (although it's
easy to find a complete working GORF PCB on eBay, but the prices are 
competitive, meaning skyhigh). But I found a page that had a link to an
astrocade hardware on FPGA as a VHDL file (sadly 404, but will probably show
up if I look harder). And from the VHDL file, one should be able to extract
just the custom chip and put that one on an FPGA.
OK what else isn't working (properly): Out of 4 RAM modules, one works.
One gives graphic glitches (some lines and dead pixels on the screen) and two
are dead. They died of shorted capacitors. Replaced them and one looks just
like that slightly defective one and the other one looks exactly the opposite,
that is where the other module has dead pixels, this one has working pixels
and dead pixels everywhere else, meaning there's at least one chip that forgot
how to do tri-state (much more probable an address multiplexer like LS157 or
LS138 or 139).
So even if I take all the hardware from Space Zap and drop it into GORF, I
just get static pixel snow (a static version of static - every pixel is random
but few ever change), but at least the pallette data is correct and you can
see squares of snow moving in a way where the "sprites" would be if the game
was working and the hardware had sprites. It has a blitter, maybe that one's
cracked. You can even start a game, but it's completely unplayable.
GORF drives Astrocade to the limit - there are multiple pallettes that even
fade in between scenes and even the screen border has some kind of animation
(think C64 before they realized how to open the border for graphics).
I should check how many custom chips on the I/O board are compatible.
Sadly, GORF poses high stress on emulators, you need a pretty powerful machine
to emulate it without dropping frames. The Igel Thin Clients we use start
lagging horribly as soon as you enter the Galaxian level.
Well it makes no difference what you run the game on, a Pentium 3 with 450MHz
will run just as horribly as a 2.8 GHz Pentium 4.
So I gotta wait a few years until someone brings out the Data Custom as an
FPGA.
I guess the Blitter can write the video RAM but the CPU can't.
With my fancy new old In Circuit Tester I tested all chips having to do with
that signal, they're OK. Also the RAM board that is just completely shot does
not have any broken control logic chips, so there must really be several bad
RAM ICs which all screw up the image.
Btw. MAME thinks it's running fine even when it jerks up completely!
OK so I now have completely socketed two bad RAM modules hoping to be able to
make one working board. And it appears so many ICs are bad that I can't root
out which ones are good! So I had to socket some chips on a known good board.
This is annoying. If a chip dies completely, the entire board will be dead.
The other chips won't die though, but the machine stops booting. At least I
got ONE working VRAM board from three broken ones. One had EVERY chip bad,
another one is missing two to be fully functional.
At least enough for a complete machine.
So let's test the Pattern Boards in Space Zap. They both work. And Space Zap
actually uses them. For everything that could count as a sprite. Without you
just get SPACE ZAP and the starfield on the screen.
Checking the Power Supply shows one cap shorted. Which I found later wasn't.
But the one next to it looked like it was already using Air(tm) as the
electrolyte. Swapped it and powered up the machine. Voltages are fine, time to
hook up the boards. No picture, not even the garbage it showed last time.
If you turn it off and on again, you see a quickly fading green screen.
There's only 0.5V on the 12V rail. Hm. These black plastic drum style 
capacitors are notorious. Removed the RAM modules. Still shorted. Removed all
the other modules and the I/O/Blitter module was bad. Tried the other one and
it's also shorted. Both had the 12V cap die. And with a replacement capacitor
the game just WORKS! Awesome! GORF is up and running!
Only... no sound. And you can't fire. And sometimes the ship moves to the
right by itself. The joystick uses light switches and an LM339 which is also
known to fail very reliably.
I found fire quickly with the Testofon, broken trace. The 339, even though not
really bad, I socketed and the error went away.
Concerning sound... the amplifier module was shorted. Another one of these
black drum caps. They piss me off! But even with a "new" one it doesn't work
(I like if you take a capacitor from an assortment of ancient NOS capacitors
and it just works and I take two more and they just work).
Checked the power supply and the short is gone, but still no power.
According to the schematics the sound amp has its own winding on the power
transformer. And there's nothing coming out. Oh wait, there's a fuse.
And it's blown and I can't read its value due to corrosion, could be 1/2A,
could be T10 /2A and what's left of the wire is pretty thick.
And neither the schematics nor the internet show what type of fuse goes in
there. But my digital copy of the schematics says 2A.
And the sound came back.
The reason why the Space General light bulb didn't work was because the lamp
socket was bad.
Now we have a fully functional GORF. Time for a real game.
The German speech synthesizer strings are awesome, even more fun than the
original english one (literally translating some of them that differ from the
original "Your last hour has chimed" (German saying if you're about to kill
someone in a game), "You're a galactic failure", "Next time you'll bite your
teeth out" - I'm afraid it could be right.)
Well I made it to Space Colonel as the screen started crapping out. And here
starts a long story of chicken-egg-problem since it sounds like it comes from
the neck, but could also be the yoke or the neck board and sometimes it runs
for minutes without arcing. And I don't see anything arcing.
At least it's not the LOPT. After hours of fiddling it got easier to just pull
the chassis and test it in our only other compatible monitor, Astro Invaders.
Sucks that the CRT ground strap is soldered in both sides. At least I got the
tube out and that was pretty service-friendly  -  ground strap and degauss
coil are part of the monitor chassis so the tube will just plop out if you
remove the neck board, disconnect yoke and EHT (HV) anode cup.
And since the monitor didn't arc, sparkle or hiss in the other machine, I
swapped out the tube. And they liked each other, I even had to turn down all
the pots. Picture was too bright and too large.
Kinda sad, the old tube had a somewhat dim but still very usable and sharp
picture, the new one's kinda fuzzy.
And that entire thing took six hours. I even found time to wreck the YPbPr->
RGB board and then bridge the severed turns of the delay line.
There's still a small problem with the sound: The right channel starts 
crackling, sometimes horribly, sometimes it even starts playing sound effects
even though the machine didn't command it to. If you swap the I/O board, the
problem goes away, but the other board has really bad problems drawing the
starfield. In the border area, it completely craps out!
After a week, the right channel sound has gone completely. Swapped one of the
sound ICs (the one that only does sound, the other does I/O as well) and that
fixed the sound.
And after 8 hours of running, dead yellow pixels appear in the second to top
row of GALAXIANS. They're harmless though. Self test just flickers in an
endless loop telling me RAM is bad. Great, 32 chips to test. And none of them
was bad! So check the blitter board. Nothing. Swapped the ROM board and the
error went away. And instead of "SPIEL BEDNDET" it now reads correctly
"SPIEL BEENDET". And the selftest doesn't show the small "x" under the 
switches. And Galaxians is back to normal.
And even worse, I found the picture to be too oversaturated and washed out so
I turned down the screen and it started crapping out again so it wasn't the
tube! At least it ran for a day with no problems after turning the pot up
again. Next week I retested. It shortly did it after turn on but then no more.
And the replacement I/O board had a bad LS161 responsible for the starfield.
It was completely dead.
And reading the ROMs from the bad ROM board revealed no problems besides the 
defective X11 German localization IC and that just held the language strings
for the screen as well as the synthesizer.
And swapping just a single IC from this board makes the other stop working.
And because both ROM boards have that Galaxian glitches sometimes and I
noticed no connection on one pin of a ROM and the sockets being Augat I
replaced all the sockets on the ROM board that had chips in them.

Regular Donkey Kong
Ignored the machine for years, in all the noise from the arcade I never
noticed that Mario's walking sound isn't working, it just goes crack crack.
Very low priority, hard to notice when all the machines are running.
Well I suspected the 556 timer 8N for run and jump sounds and one side had
signals going in and out and the other side had just one signal going in,
putting a cap on made another signal appear, swapping the chip entirely
brought the walk sound back.

Another regular Donkey Kong.
Symptom: When there's a certain number of sprites on the screen at certain
positions, the picture or part of it gets shifted half a pixel or even an
entire pixel. And it's not the monitor. Found out that one of our DKs is
rock solid. Tried all the video address lines, but they look OK.
There's three LS163 which get hotter than most PALs and about as hot as
74188. Swapped them with LS161, but now the sprites glitch like mad.
Per single swapping it turned out the middle LS161 HAS TO BE an LS163.
Hmmm now I have to mutilate a working DK. If both glitch, I'll have to look
on, if not, the glitch will help me find the fault. Well, the other DK works
fine with an LS161, but... they have De-glitching-Caps populated and if I
populate them on this DK, I still have a sickly colored Mario on the title
screen where there should be no Mario, but at least the ingame sprites stop
glitching. Sadly, the half-pixel screen movement is still there, even if you
try playing with deglitching caps some more.
I give up. If I charge this guy my regular prices, I could buy myself a
gaming notebook with internal PCI-E GPU.

Another regular Donkey Kong (Hi-Score Hanover)
Weird fault - the level looks strange and you can play it like that. Instead
of ladders, there are more weird girders and you can jump up them, so the code
reacts to the different playfield. Also some text characters are messed up.
Looks like bad character RAM. We were at it with a phone conference and they
swapped the character RAMs to no avail. Should I mention I confused the line
RAM for the character RAM the first time, leading them to replace the fast
2148 with too slow 2114?
Well there ain't much in that circuit, just a BUS transceiver (LS245), the CHR
RAM and the CHR ROM. There's two invalid signals on the scope plus one that
looks weird. The invalid signals are OK, there isn't supposed to be data on
that on the first screen (there's normal looking data right on powerup).
Chiptester says BUS conflict on two other lines, even without the RAM. Weird.
I remove the ROM and the conflict goes away. Turns out, 3N has two low
impedance address lines! Who's gonna think of a fault like that!?
Now we have another problem - those people don't have a regular Donkey Kong -
they installed a High-Score-Save-And-Upload-It-To-The-Internet kit that also
includes a bunch of ROM hacks. Thus, all the ROMs contain slightly different
data, which I don't have. It's not on the internet either (or if, I won't be
able to find it). The defective ROM still reacts to the address line though,
but requires over 100mA to drive it. So I read it twice - once stuck high,
once shorted to ground. Then I stitched the two images together to get the
correct one, burned that to an EPROM and installed it and the machine is back
up and running.

Crazy Kong
We got tons of them. In German I say 5 but it's more like at least 15.
Most work, some are flaky and one was dead.
I could revive the dead one by swapping ROMs. Turns out, one ROM was bad, but
fixing it (it just lost a few bits to stray UV rays so I burnt it over) didn't
fix the game, but I notice different errors with different ROMs and pressing
on all of the sockets made the board work with whatever ROMs I put in.
The flaky boards also show no really reproducable faults and seem to be
generally working so I say - bad IC sockets.
So I bought 100 pieces of 24 pin dip sockets.
HOW -HIGH- BORED CAN YOU GET?

Vanguard
Intermittent speech. Searched a long time to find two tantalum caps which had
the rare failure mode of being random value resistors. Measuring them makes a
great random number generator, hooking them up to the Testofon made them play
an eerie tune.

Scramble Scramble Amidar
I bought a Scramble cocktail table and an Amidar boardset from some guy who
stores them in his garage which is exposed to the elements, it doesn't rain in
but the humidity matches the outdoors one exactly. There was fungus growing on
almost all the ROMs of all the game boards I saw. It grew on the paper 
stickers on the ROMs and their legs and some of the capacitors.
At the Arcade Museum I thought, let's steal another Scramble board so I have
something to check.
All three boardsets are (near) identical bootlegs of Scramble hardware.
So I put the "stolen" Scramble in and... Garbage Screen. Great. Both main and
sound board CPUs are extremely corroded. Turned out one of them was bad.
Swapped it, hooray, game runs. But the graphics columns are double, exactly
like Galaxian but 2 instead of 4. Almost measured the entire complex address
generator until I found that the fault only shows up in the machine, not on
the bench. It probably was a shorted trace. It now runs fine.
I had to cut the -5V trace even though I don't own the board, since the 12V
were shorted with the -5V on that board intentionally for some weird reason.
And many years later I needed the board as a placeholder in a Frogger cocktail
table that came without a board, but with a harness that was incompatible to
anything I've ever seen. Playing it, I noticed engine noise is missing
(background woobwoobwoob). It's missing no matter if you swap the AYs.
The 4066 are socketed, I remove the first one and the sound comes back, albeit
unfiltered. Swapped it, done.

Now that I have a working board, time to swap ROM chips. I already found the
roms from my Scramble boardset to get different results every time I read them
out. If I dump them four times, I get three different results...
So I put my ROMs into the "stolen" Scramble. It works fine. Hm.
Swapping the CPUs doesn't do nothing... oh wait, my sound CPU is as dead as a
nail. Put in a new one. The AY sound chips just chirp some garbage that has
nothing to do with the game sounds. If you swap them, the game works fine and
has definitely more than three sounds going at the same time, so one of the
AYs is knackered, but not completely dead.
OK so I know that all swappable chips are working (or at least not interfering
with the game), time to actually FIX my board set. It's just showing a black
screen with some colorful garbage travelling accross it and one of the 8255
PIOs is making weird noises. You can put in any other, no matter what brand,
they all make the same noises. I should write a program to play a tune on an
8255 PIO chip.
Measuring the CPU I found NMI to be floating, invalid signal. It's high on the
working board, so not connected is out of the question. Schematics say NMI
comes from an LS74 at 9L. Checked it, all four outputs are floating. Dead.
Swapped it, game runs! Oh I can't see my shots (shell), but they are hitting
their mark. Also the sound board is not working.
10J in the shell circuit was bad, no Ripple Carry Out. Putting a cap on also
gave no ripple carry out, which was weird, but only three signals went into
that chip and none of which resetted it so it SHOULD have given out RCO.
Swapping the chip brought all four of my shells back.
Now only sound is missing. So far three Z80s were defective (checking those of
the Amidar board too).
Trying the known working board again, there was a loud humming in the speaker
and no working sound and the audio amplifier just released its magic smoke.
Damn. Why? No clue. Even weirder, when taking the boards out again and
operating them outside of the machine the sound was working again. OK who knew
that LM377 amplifier chips work without their magic smoke?
The sound board also works when I put my sound board's ROMs in.
So... putting my broken sound board back in and just scoping around...
the entire BUS looks ugly. Barely anything goes down to logic LOW, neither
inputs nor outputs. They used a bag full of LS08s, tied one input each to HIGH
to just use these chips as a BUS driver. Sadly, I have only one LS08 spare
left. I'll swap the one with the ugliest signals and see, with my replacement,
the signals look fine, inputs like outputs. But the others still look bad.
The 82T245 (schematics say Ls245) has no output on one pin. Swapping it fixed
that but I still have like two further LS08 to deal with plus two that look
fine, but are of the same make as the failing ones.
So the easy fix for missing LS08 is to use LS32 and wire the input to logic
LOW instead of High. That oughta do it. Did that, still no sound.
Continued probing, found an LS28 with a gate that has no output but should.
What's that? An LS02 with driving capable outputs. Put in an LS02 instead.
Checked the "stolen" Scramble, it even has an LS02 in that location.
Still no sound and address line 15 has a characteristic pattern on the working
sound board, this one doesn't. After lots of measuring around I decided to
swap the two working LS08s and that's where I found out that using a 3k3
pulldown resistor to get the inputs to logic LOW didn't suffice so I just
shorted all the LS32 inputs to ground and... the pattern on A15 shows up.
But... still no sound! How does the sound board know it should play a sound?
Both 8255 are controlled by the mainboard. One of its output ports is read by
one of the AYs and if there's something on, an interrupt is thrown.
But no interrupt. It's generated by an LS74. Data and Clock goes in, nothing
comes out, swapped the chip and voila, it works. PHEW!
So we're almost complete. When the machine is cold, there are two malfunctions:
The ship's position lights are very dim and some sounds are odd.
When it warms up, both problems go away independantly.
I could confirm that the slave AY 8910 is definitely not OK and is also the
source of the sound problems, but it works well enough so this is not really
a problem, also after a few minutes it starts working great, all three
channels.
The position light were caused by the LS273 right next to the color PROM.
Scoping one output brought the position lights to full brightness, but putting
in resistors didn't change anything so I changed the entire chip, problem is
now gone. I know, that was unnecessary, but LS273 are still manufactured,
AY8910 are not.
After two weeks, scrolling stopped working. Well it did scroll, but there was
no new terrain data entered on the right of the screen, which then led rockets
to be fired all of a sudden out of the blue. Since there were lots of 
defective LS08 of one brand and lots of defective LS74 of another, it only
took half an hour to find the guilty chip, it was right next to the defective
9L (NMI). Putting a cap on made one output toggle and another fight, so
swapping the chip and the game works fine.

Trying the Amidar ROMs in either Scramble boardset doesn't work, just garbage
that just slightly changes.

So I hacked my Scramble board to Multirom and found at least three more games
that work on it and aren't minimal hacks of Scramble (Dude, there's like five
games that shout Lawsuit). Interestingly, the Konami Version of Amidar runs
fine, but the Konami Version of Scramble doesn't work at all.
Also, with Frogger and The End, I couldn't get the sound to work.
According to the hacker's manual, there is a LOT that has to be done to the
sound board to get Frogger's sound to run (and even then, it's a bootleg that
doesn't even have music!)

OK now we're ready to fix the Amidar board.
Boy, it's hacked. First I panicked when I saw a metal cap chip was missing.
But the other board has it and it's... a 741 opamp. Phew.
I put that in and also swapped all electrolytic capacitors.
The board just shows the classic Garbage Screen. Also, another Z80 quit 
working , now that's four out of six. I had to steal some from the museum.
(note when I say steal, I mean I take them, I don't pay for them, but I make
sure that everyone knows that I took them. Most people think it's OK if I do
that, since for one, some of the chips I take end up in machines which end up
in the museum, so that's OK, and also with all the repairs I'm giving and 
never charging anything for them - well I did charge for the spare Flyback
of that PolyPlay machine and I did charge for a wooden back plate for some
machine whose backside was open, but I don't usually charge for all the
replacement chips I install, so I guess it's fair)
OK. So the CPU is resetting sporadically. The watchdog depends on VBlank and
some signal from an LS138 which measures OK. The CPU just does what it likes
and the BUS is just Garbage. The CPU drivers have floating outputs most the 
time, Output Enable is high most the time. Comes from CPU !RD, so I'm stuck in
the same endless loop as with the Galaxian.
Not knowing where to start I just hunt the entire board for invalid signals.
The variety of chips is pretty low, mostly 00, 02, 08, 32, 74, 161, 36x.
Quickly found two LS161 in the shell circuit (unused in Amidar, but I want a
multirom, so gotta fix it), at least one LS74 for certain, an LS02 is suspect,
an LS367 dead, as is an LS283. Now I have two defective chips I don't have
spares for, so gotta order. If the 8216 are bad, I can replace two with an
LS245.
I could try and burn an Amidar Bootleg, I couldn't find one that was 1:1
compatible - hey I forgot to mention I couldn't find a compatible Scramble
image in MAME even though Romident identified it as a certain Scramble bootleg,
but the Scramble I "stole" was identical!
Replacing all these bad chips did nothing, although the LS283 full adder used
in the video output address generator (what!? why!?) made the garbage screen
less repetitive. On the bench I have a choice between a few select endless
loops, it even looks like it tries to show a title, but no sign of clean
booting and in the machine just garbage.
So I need to get the board into the same state as my Scramble - make scans
of all the hacks and then undo them.
So I undid the hacks, installed the Scramble ROMs and... it looks horrible,
but it's clearly running Scramble now.
So I installed the Scramble Character ROMs to asset the state of the graphics
subsystem and it still looks broken. First I had vertical lines that went away
when pressing on the ROM, then I put in new sockets and it's completely dead
now, just very colorful vertical lines with traces of the game running in
background. After a LONG time looking in several wrong places I found that the
color RAM doesn't get R/W, it's stuck high. Replacing the Clock chip which
also generates that signal didn't fix it. Testofone says it's shorted to 5V.
So I took a wire and shorted it to ground, shorting out whatever was causing
the short. Long story short I didn't want to remove all five color RAMs to
find the bad one, turns out that they all work fine. No clue where that short
came from. So the picture looks normal now except every character is twice its
left half. Which points to the address generator. The second LS283 full adder
had no output on one gate, so swapped that (I ordered two) and finally the
board runs Scramble like a champ. Including all the things that weren't used
in Amidar. But it still has the color PROM from Amidar. Whatever. You barely
notice it anyway.
So the next step is a Big ROM Mod since I want to play as many games on it as
possible. Or at least Scramble and Amidar for a start.
The Big ROM Mod requires two gates of an LS32 and a single inverter. Luckily
the board has lots of chips with unused gates lying around and the Big ROM Mod
was successful. All eight ROMs replaced with one single one that even has two
games on it. The Amidar Bootleg from MAME runs just fine.
Since I haven't installed multiroms for graphics yet I play Scramble with
Amidar graphics. This looks funny. I'm sure it must have been the inspiration
for Parodius. Your rocket ship is two dancing skeletons, instead of bombs you
drop color rollers, Rockets are teddy bears (or pigs), they even show a death
animation if you shoot them. Parodius. Amidarodius. Scramidar.
Scramidarodius...
Sadly, the Scramble Sounds don't fit Amidar. Well I still have to fix the
Amidar Soundboard. Gotta check which other games run on that platform.
Actually lots. But you have to do lots of modifications to get each one
working, so with just a ROM swap I get only Amidar, Scramble, some Frogger
bootleg and The End to run and the latter two need Sound Board modifications.
Big ROM HACK:
FYI: 4E is the chip select split chip which gets A11, A12 and A13 which will
be connected directly to the big ROM. Chip Select is then 4E Pin 4 OR 4E Pin 5
OR NOT 4E Pin 6. 5F is an LS32 with two gates to spare and 7C an LS04 where
only one gate is ever used. Boy that one will be happy.
Stupid thing the game likes to lock up when you switch between Amidar and
Scramble, so I'll put in a switching and reset circuit. The sound board has
an LS74 with one gate to spare. Set/Clear are tied to +5V, Data to !Q and
Clock to the push button which again is tied into the Reset circuit
(5C or 6C, there's an LS00 with Pin 4 and 5 shorted together. Cut the 
connection and solder the reset button to pin 4)
OK fixing the Amidar Sound Board.
First I had to undo the hacks again. I'll have to redo the address space
expansion hack since Scramble has 6k sound ROM space and Amidar has 12k, using
8k of which.
Hooray after undoing the hacks, the soundboard quit crashing the mainboard.
You can even control the ship, so the 8255 work. Just no sound.
Standard procedure, power it up and look for bad ICs.
Just one LS08 bad this time and 3 of 4 gates worked, the other one had 1V for
LOW. The one LS90 next to the sound chip had two signals missing, putting a
cap on brought these back, so swapped it. An LS28 and LS04 had unholy signals
but they're interwoven and I know that LS28 is pretty certainly bad so I
swapped the LS28 for an 02 and the LS04 came back too. The LS245 looked bad,
too, but putting a cap on did nothing, still the pattern on A15 isn't there.
Oh I overlooked an LS32 at the edge in the A column. LOW OR Pulse is LOW?
Put a cap on and A15 now shows the familiar pattern. Swapped all the dead
chips, put the old LS245 in which worked fine.
Put the sound board in and was able to play Scramble with sound just fine.
Now I had to redo the sound ROM address space hacks.
Burnt multiroms with Scramble adapted to the new address space configuration
and Amidar and it worked right away. Scramble has sound, Amidar has music
(and sound effects). And the music IS wicked. Can't get the tunes out of my
head. The two 4k ROMs from Amidar Sound I can now reuse as dual boot Character
ROMs for Amidar and Scramble. Did that, erasing and burning them worked just
fine - almost fine, the second ROM required two passes until all the bits
settled. It works and now I have Amidar and Scramble switchable but... again
no sound. The LS04 which had bad outputs now has bad outputs again. Swapped
it and it came back.
Well no sound on open saturday, hard to notice due to the noise from all these
machines, also doesn't help the speaker is on the bottom rear of the machine.
One of the remaining three LS08 had a bad output. Since I was at it, I
socketed the remaining two LS08 from that batch since maybe their days aren't
counted yet but their months sure are.
And would you believe it, one of them even survived the next open saturday.
And you STILL can't hear it, even though I turned it up as much as possible
without leaving the comfort zone when it's the only machine turned on.
For some weird reason, the LS08 survived the next chip on the machine.
But first we have the problem of no +5V. I've changed all the silicon in the
voltage regulator, no difference. Resistor values are OK, where's the problem?
Measuring a small transistor (BC256) makes the 5V come back for an hour or two.
I "fixed" the power supply by bending up the LM723's Current Sense pin.
But the machine didn't work with 5V there. CPU's running, what's the Video
Address Generator doing? Nothing. the first LS161 (5M) was broken.
And then it worked for a short time, then back to Garbage Screen.
CPU is watchdogging. BUS looks OK, but no Read/Write. The CPU outputs these,
though. The LS367 that puts these on the BUS happens to be the one I salvaged
from a mystery pinball CPU board and had green legs back when I salvaged it.
The chip was dead. No output. But now I have replacements so I can swap the
chip and the game works again.
Really weird is that on Scramble, Player 2 can't move the ship left.
On Amidar, moving left is possible for Player 2, so it has to be a software
bug. Weird. It worked on the cocktail table.
And sound dies if the machine is running hot. I can be rather sure that it's
one of the AY sound chips.
And just a little time later it just produced an unsynced garbage screen.
It appears as if the machine tries to draw the grid pattern, but crashes.
Since one of the horizontal address generator LS161 died, I checked the
second one first and sure enough it only counts to 7 and the other three
outputs have 1V for Low. Swapped it, runs.
The sound now lasts until the end, so it must have been the AY.
Hm - after three years, there's still one LS08 left from the bad batch.

Amidar (bootleg, Super Cobra hardware)
I got a stack of Scramble/Super Cobra hardware hoping that one of these might
have Frogger on it or will help me get the Frogger sound to work, because I'm
trying to convert a Scramble to Frogger for over a month with no success.
I got it to make dissonant beeps when you move, the beeping stops only when
you die and that's all she wrote.
Well let's see how far I can get here. I got 3 Super Cobra boards which are
all partially parted out (only the socketed chips), all have the CHR ROMs
missing, most are missing most the RAMs, some Z80s are missing and cross-
testing found a bad Z80 or two.
The cool thing about these boards is, that there's four 2732 instead of eight
2716, but six sockets are populated. And since the machine has 2k work RAM, I
only need to bend up WE and CS and can populate one of the empty sockets with
a 6116 and can save myself four 2114s. WE gets connected 1:1, and I'll have to
AND the chip selects, there's a 74LS08 on the left edge of the board (if the
CPU is top right), pretty much exactly in the middle, pins 1-6 are unused.
I used those on Frogger to get the water "effect" working, but here we only
have Amidar. A bit evil is that the LS32, which I need for the big ROM hack,
is unpopulated.
But before getting there, I have to get it to run first.
I got 5114 RAM from Jupitec's stash, that's too slow for sprite/character RAM,
but fine for CPU RAM.
And... if the game even ever boots, it immediately crashes.
Noticeably every 4th character repeats vertically (if the screen were 4:3),
just like the Galaxian long ago. And that took me a while to find...
Starting with the video address generator, nope. Well maybe the bank of LS86
for the cocktail table mode. Nope. So where does it go next? Well there's a
bunch of other chips involved, I checked them, but in the end I found an LS273
just below the CHR ROM that had an invalid signal on one output, High is only
1V. That goes to a bank of LS157 and if I short the signal to +5V, the LS157
gets hot, not the 273. So I snipped the leg off the 157 and the signal came
back. Swapped the 157 and the picture is normal again, but the machine crashes
when drawing the title screen, as soon as it wants to draw sprites.
Without the sound board, it goes into gameplay immediately, just to crash when
trying to draw sprites. And that make me take forever to find the fault...
until I swapped in ROMs from another board - I dumped the ROMs and checked
them, found a few faulty ones, but I didn't remember which ones were.
Well - SEVERAL of the ROMs of this board were bad and it works just fine with
those from another machine.

Revolution X
Well that was a quick one, even though it's from 1994. First I swapped all the
SMD electrolytics, which have all leaked. That didn't bring it back, it still
shows the diagnostics screen saying two SRAMs at 67 and 77 are bad.
So I measured them and there was no D0 signal, it was stuck high, which means
the RAM is very unlikely the culprit. Let's check the ALS245 transceiver.
It receives only high for D0. Pulling it low with a resistor makes me see
signals, but nothing changes. The SRAM side of the 245 doesn't reflect the
changes, just an ugly mess between 4.5 and 5V, so I changed the 245, betting
this doesn't work and of course it didn't. Also, some traces didn't survive,
the board is of poor quality. Probably an early multilayer.
So let's study schematics and continue with the BUS. The SRAM's outer bus has
lots of chips attached to it (including the gun recoil coils), maybe clipping
the D0 input pins, hmm, no wait, let's check where the data bus signals
originate. ANOTHER LS245, this time an SMD variant. Gnnnhhh! It also doesn't
get D0. But this time it originates from that huge Texas Instruments
TMS whatever which is probably a geometry processor. That means it should
originate, but I don't measure any connection. With my Testofone I can measure
a 100 legged SMD chip in a few seconds, no kidding, no connection.
D1 however has a connection, so does D2.
So let's triple check where D0 is supposed to go and lay a bypass.
Turned it on and it worked just fine. Except sound. After a few hours I asked
myself if I maybe forgot to connect the ribbon cable to the sound board.
I did. D'oh!
One recurring problem with the machine is the pots from the guns. The gears
are broken and are slipping. A job for a 3D printer. The replacements fit with
some drilling. Thanks to Daniel. The reason for the recoil not working on
player 1 was a blown fuse. Rated at 800mA, it blows after three shots.
Replaced it with a 1.2 and that seems to work.
Uhh and today (11/21) it shows weird graphical garbage, one of the RAMs
doesn't pass the test, it still goes into the game, it runs stable but the
graphics are snowy/distorted. And something smells like leaky electrolytic
capacitor. Unfortunately all the machines are running so I can't be sure what
it was and the inside of the machine smells less. The faults get less and less
until they stop if you leave the machine on long enough and a reboot passes.
Gotta check the power supply.
Nope, at 4,94V on the chip furthest from the edge connector the PSU is good.
So I gotta check again - since the colors in self test go on an intense LSD
trip and in gameplay sprites are smearing colors, I take a look at the palette
RAMs which were the victim last time already.
And their address lines have some illegal stuff in between the highs and lows,
looks kinda like a video signal - and check this out, the manual says they are
SUPPOSED to go floating from time to time. So I take a look at the /OE signal
on the ACT374 address latches and it's a floaty low with a chance of signal.
The schematics say that comes from that big CPLD... no it doesn't, because the
trace has been corroded off. Put another bodge wire in and the machine is back
up and running.

Mortal Kombat
Goes here because it has a similar board to Revolution X.
The self test flags ALL Graphics ROMs as bad.
The game plays OK though and if I wasn't told that the board has graphics
issues, I wouldn't have even noticed. Everything looks OK, except 3D 
backgrounds (the floor on some stages) and the status screen which is missing
lots of text and probably pictures.
The way the ROM test fails makes it look a lot like BUS drivers. The GPU gets
access, the CPU doesn't. I mean it flags the ROMs red after about a second per
ROM, as opposed to Revolution X where it flashes colors for tens of seconds
per ROM.
So, BUS driver, bad PAL, timing problem, I don't feel like doing this.
It's still more fun than my current job.
And the owner came and took the board away again.

Tutankham Bootleg
Another one that isn't ours. I fix it as a job. Maybe.
This is going to be a bigger job.
The guy had the edge connector connected backwards and from then on he only
got colorful garbage screen. No clue if the ROMs are OK. The RAMs are suspect
since they're 4116, also they need +5V, -5V and +12V, maybe ground is missing,
but I've measured two of them and they have a signal on their output, so maybe
they're not completely dead.
The bootleg has a couple Konami Custom chips rebuilt as 7400 logic boards.
Two LS161 attract my attention because hot. The outputs are bullshit.
The In Circuit Tester/Debugger flags one as OK, the other one I remove and
test it out of circuit. It's OK. Put in a new one and the signals look
identically bad. Invalid signal that never reaches 0V for Low. Weird.
Too bad there's no plan (and no one else on the planet has ever seen this
version of the board).
Well, let's put it into our Bubble Bobble: Colorful snowy and extraordinarily
unstable picture. One of the VRAMs is socketed, let's test it in our Qix.
Err... OK!?
Hmm. maybe Tetris likes this board better. Jup. And through all that static
I can make out PRESS START and CREDITS 2 and a bit more. If you press start,
the game runs, even with sound. But I can run through walls and it looks like
more than 3/4 of the video memory are bad.
So I desolder the first bank of VRAM and test it in Qix. They're ALL OK! WAT?
Maybe the read-out circuitry is bad? Measuring voltages... for 5V I get 4.6V.
Uhh. 12V are there. Where's the -5V? Missing. But I can ring it back to the
edge connector. Duuuuh Tetris' JAMMA adapter doesn't have the -5V equipped for
some reason (maybe lazyness). So put in a wire for the -5V and now I get
colorful stripes that seem to be on an LSD trip (color LUT switches constantly)
At least you can now make out more graphics.
I move the board and the stripes flicker and disappear. Then the machine
crashed. I jiggle everything (the daughterboard for the custom chips) and
still the picture flickers wildly. I restart the machine and the picture looks
way better. I push upon the sockets and tada - the picture's perfect.
For some weird reason I can jiggle as I want and barely anything happens.
The VRAM seems completely OK.
Hooked it up one week later, no picture. Pushed around the sockets again, it
works fine. Sound has no more distortions now whyever.
Well if it's turned off for a while the graphics garbage problem returns.
I changed the sockets for the daughterboard to something that is made for the
thicker pins from the interconnect. The sockets were single wipes and got bent
by these thick pins, no wonder it didn't make good contact.

Pub Time Dart machine (ca. 1987)
The machine has an 8 inch amber monitor. And a TMS 9129 video controller with
insane 64k of video RAM. What's burnt in on screen looks like 1k of VRAM would
have done the job just as well. Like my Loewe BTX. Originally, only the
monitor was broken. 16 kHz whine is there but no picture, no neck glow.
Trial&error with another tube shows they need about 12V to glow. This one gets
seven. Main voltage regulator should output 10.8V, but outputs also just 7V.
And that tiny monitor consumes 1 Ampere at 7V. What's going on? I don't
expect more than 500mA at 7V. Scoped the LOPT and I don't like what I see.
Bad LOPT. Let's see if the LOPT from another BTX phone fits. Physically it
fits, but electrically nope. Not doing anything now. No spare parts, I give up.
Now I have a few black&white TVs too many, among them a dirt cheap Grundig
from 1981 whose Tuner bit the dust a long long time ago, so let's salvage it.
Carefully raising the voltage... at 9V we're already at 1A... hooked the TV's
original yoke back up and the power consumption normalizes, but still no
image. Shouldn't I see a bright vertical line? I do see it when I turn off the
TV, but it should be a lot brighter. So I swapped the yoke for good measure,
this should be our main problem. Still no picture but the flash upon powerdown
looks more promising. And we also have neck glow. So I turn all the knobs that
have to do with brightness up to 11 and I see a super dark picture.
The tube is gone! Let's see if the original monitor board works now.
Yes and no. It now consumes a lot less power than before, even less than my
Grundig TV, but still no picture, so back to Grundig.
About time to reconnect the machine's main board instead of my test pattern
generator (a retro computer).
Hooked it up and no nothing. Not even sync. Scoped it and there is no more
video signal. Now I need the VDP's pinout. It does strobe the VRAM like a nice
guy, but looking at the control signals it looks like the video CPU just 
crashes. It locks up in an endless loop.
Suspecting the one PAL, a senile ROM and the battery SRAM as well as the VDP.
Dumped the ROMs, they look OK, but I can't compare them since I can't find'em
online. Gotta check undumpedwiki. But there's ASCII text strings in there.
Built a 6116-adapter for the battery SRAM so I can test it in my In-Circuit-
tester.
Then I removed my Colecovision's VDP (see below) and put in a socket, so I can
test the Dart machine's VDP.
The both chips aren't entirely compatible, all I get is colorful lines.
But they are COLORful. Not monochrome. And the chip DOES output a video signal.
And Colecovision's VDP also only outputs +5V for a video signal. So I take a
closer look and bend up the video output pin and - Output!
The transistor shorts the video signal to 5V. Swapped it, works now.
The 8k battery SRAM tested bad. Scope says most of the bits still seem to be
there and work OK, but not all of them, so the SRAM is senile.
I put in a generic 8K SRAM and every few seconds the picture changes to
something different. There seem to be three different screens in attract mode.
Jup, looks good on my 1701. The Grundig TV gets a decent and somewhat bright
picture with the original board, I can at least turn down the +B to normal
levels now. But boy this CRT is done for.
Let's see if the original monitor works... no, Fabi sold the machine. With my
Grundig in it. All I got left of it is the tube. I could put it in my 8032SK.
Or that Philips green monitor.

Asteroids 1:
Got this as an assignment (yay make some money of what I love doing).
What's the state? Reset is stuck low. Without even downloading the manual, I
look where it's coming from. An LS74 directly above it. It doesn't get input.
That comes from above, an LS393 which doesn't output on that pin. But gets
input and even outputs half the frequency on the first pin. Piggyback a working
chip, jup, this one's dead. Now it's watchdogging.
Let's see what it does in self-test mode. Nothing. Still watchdogging.
Not even self test is working.
Could be I've swapped the boards in the meantime and forgot to swap my brain.
The board now works in the machine... somewhat. It plays half blind.
Only full brightness objects show up. The player's ship glitches in certain
angles, but moving objects (shots, UFOs, player's ship) move linearly across
the screen. The remaining lives display is slanted upwards.
Self test shows the diagonal lines just perfectly, but nothing else.
I noticed that someone replaced one of the LS273s in the middle of the board
and did a pretty bad job. Well I've seen worse and read about way worse, but
this chip controls brightness and parts of the Y axis, so this is suspect.
Removed it, cleaned up, put in a socket, no difference. All the outputs seem
to work though.
Went through a standard search for bad simpleton gates (AND, OR, XOR, NAND,
etc) and found a bad LS32 pretty far up that had an output stuck high and
guess what - it controls the one LS273 I've just talked about.
Now the Z-axis looks like it would show a gradient (on my bench I have no way
to get X and Y axis to work so I could play it on an oscilloscope).
I wonder if the glitchy UFO problems are gone now too. Well, yes.
Plays 100% OK. And I even managed to set a new personal high score: ~45000.

Asteroids 2:
Same box as Asteroids 1. I should have gotten 3 boards, but he sent just 2.
This one watchdogs without me needing to swap any ICs. 4 out of 8 of the CPU
BUS data bits look unholy. Erm... 2114? But which one of them? It could be
that this is normal for a BUS that goes to floating in between and the other
one lost his Tri-statedness. But if you crank up the scope (timebase), you
see this isn't floating, it's just sick!
Hmm in Self test mode it stops watchdogging. But I have no audio, because I'll
probably need the 12V or something. Let's scope the address bus instead and
look for patterns. Jep, after 2 seconds it changes. Which means one of the two
CPU RAM chips are bad. Just what the scope says. Removed both, both bad.
One had bit errors on 2 out of 4 lines, the other one had a christmas tree of
warning displays come on. What a coincidence since it's X-Mas right now!
This is the 8-Bit-X-mas decoration. Had to take a screenshot, but forgot my
camera. I'll keep the chip to take a photo next week.
Socketed and put in new RAM, put it in the machine, baaap baaap boooop.
Vector-RAM. Well, the RAMs are all from the same batch, could be they're all
dead, swapped them for sockets, put in new RAM, no difference (even though 2
of the vector RAMs were bad as well), so the CPU has trouble accessing the
vector RAM. Scoped the BUS: Ugly looks different. Better for once. This here
are chopped up zombie body parts still trying to eat one another. Where's that
come from? A bank of LS157 next to the CPU. The Chip Tester flags two of them
as bad. Swapped them and checked the other 11, since they're also from the
same batch, but these are OK. What's it doing now? Watchdogging still, even in
test mode. But it goes Bip! at every Reset. And the entire screen lights up
from stray electrons from a vector deflected way off the phosphor screen.
Turning the size all the way down I see one, two lonely lines.
What could disturb the BUS? Let's take a look at the vector ROM. I hold the IC
in my hand, but not all of the pins. Rusted straight off, nothing you can
solder back on. The other board already has all-EPROMs, so let's clone it and
see how far I can get.
The 4016s are notorious for going bad. Hm the other board has them original
and works fine. Ours has them original and works fine. But whatever, let's
swap'em and now there's more on screen. Yuck. At least it tries to draw ONE
diagonal line. And crashes. The picture appears to be six times the size of
the screen with everything turned down to minimum. At least there's the
gradient in the middle of the screen.
Only... on the test bench, it's completely dead now. No Z-axis and the most
part of the board is just dead.
Weeell... doing the standard search for bad components I came across an LS86,
one of the gates had the output stuck high, piggybacking it fixed the signal
so I changed the IC. I could have kept the original since only 1 gate was bad
and only 2 gates were used at all. Any difference? Nope.
Let's see what the timers are doing. Nothing. Oh God, they come from an LS273!
Which is getting latched (or isn't. None of the latch signals for the four
LS273s is coming) from an LS42 only two of its inputs do anything. These come
from an LS174 which gets its inputs from one of these weird LS288 mini-ROMs
whose address inputs come from the same LS174 - GAAAAH ENDLESS LOOP!
The LS42 is marginal, if I touch the outputs I get some hum across them.
The chip tester passed it with no problems at all.
I flag it "M" anyways ("M"arignal, "M"ight fail soon).
Another thing marginal is the 7404 at the Quartz. Its output amplitude starts
off weak when you turn it on and then gradually gets bigger and bigger until
it starts complying with 7400 logic levels and the rest of the board starts
reacting to it, which can take up to a minute. Removed it, socketed it and now
it starts right away, but it's still hot from desoldering. Let's cool it down
and yes, it's just like before. In the bin with it.
As said before, the timers are completely missing. They originate from the
LS273 which is controlled by the vector state machine. What's it doing?
Nothing. Stuck in a loop, maybe 2-3 instructions.
After socketing the three main ICs of the state machine without getting
anywhere, I piggybacked the LS74 at A7 (or thereabouts) out of sheer 
desparation and it fights! Which is weird since all the outputs of the chip
pulse! Swapped it, now we"re back to square one - a gradient in test mode and
broken lines and watchdogging, but only once per 1« seconds instead of every
half. Also it draws the gradient outside of test mode, but there it doesn't
watchdog.
OK so now I socketed L6 so I could bend up pin 1. And guess what - it
watchdogs in test mode even without the vector CPU attached, so the 6502 part
must have an error. But I couldn't find anything rotten (except for the IC
pins of the ROMs), not the ROMs, not the RAMs, not the 6502 itself, not the
remaining LS157. (or have they failed in the meantime?)
After not finding anything and getting distracted by the fact that one of the
bits on the data bus looks different than the others I swapped the ROMs.
Great, now it's only watchdogging with the vector CPU attached and the data
bus signals look a lot like attract mode (you can even see something blinking,
probably INSERT COIN or PRESS START). And if I try putting the old ROMs back
in their pins remain in my hand. Maybe I should try soldering them back on
since the ROMs don't appear to be completely bad.
Let's continue troubleshooting the vector CPU. The two LS367 of which the
chip tester already flagged one as bad and which has weird signals on the
scope (low, near-low, high, near-high as well as floating), behave funny.
These signals are coupling to the 4 bit register file. One of the chips has
been swapped by the previous owner. Swapped with an even older one!
Well, the signals don't all reach LOW completely, no matter the make.
Makes the LS367 more suspicious. If I piggyback them, nothing really changes.
Except for the Z-axis signal. Could be something's indeed wrong with one of
the LS367s. One of the LS42s that went past my radar until now I can see an
invalid signal on one of the inputs. Coming from an LS83 full adder. The
Jupitec Old Parts box I bought had a few of these in so I piggybacked in.
Now this signal is valid but others aren't, since the two chips are fighting
each other. Gotta desolder it tomorrow, getting late, damn job.
So what's the difference after swapping the '83? Nothing different on the only
measurable thing, the Z-axis.
In the meantime I noticed that even though I swapped the marginal LS04 the
board stays in RESET for quite a while after power up. The LS393 at A3 (or
someplace near) has signals on all pins except pin 9.
Well let's swap the suspicious LS367. Now the Z-axis looks even worse. Crap.
Also interesting: Both LS367 are the same make (Motorola), have the same date
code, but one has a shiny finish, the other doesn't. In self-test mode I get
two deep and two very deep notes (P11 pin 9 has audio with only the 5V
attached), so the CPU can't reach the vector RAM again! The Z-axis died
completely now and there's not even DMAGO anymore! Traced the RAM control 
lines and one of the remaining LS157 fights its cap, so in the trash it goes
and now - the Z Axis looks exactly like the other machine's Z-axis. Runs!
Too bad it's likely the remaining 8 LS157s fail soon. I only have one in
reserve (plus an unknown amount in the Jupitec box).
And the picture? One of the asteroids moves down at blazing speed, everything
else is stuck in whatever vertical position it initialized with. Hyperspace
works in both dimensions. Selftest flagged the newly burned ROMs as bad.
They are used old parts, so maybe they've lost a few bits. Using the other
machine's ROMs it works fine though, except for two sound issues: The saucer
sound is always the high-pitched one of the smaller saucers and the saucer
shots are seen but not heard. And the ship shot sound frequency decreases a
little too fast and stays at one frequency at the end of the sound and if I
hadn't mentioned that I bet you would never have noticed.
Let's check the schematics for where the signals come from. An LS259.
Chip tester says Yuck, so swap it. Huh? I ran out? Hm the box from Reiner's
father could have some. Nope. Literally hundreds of 7400s series, but no
259. But some you've never heard of and are probably made out of solid
unobtainium as well as exotic chips many of which you can't get anywhere
either. Couple hours later I find the last LS259 in the wrong drawer.
Tested it, the machine's back 100%.
Concerning the ROMs I burned - they're OK, they're just too slow for the 1.5
MHz clock, so just burned a few more EPROMs and these work fine.
Also I replaced the four 2114 of the vector RAM with a single 6116.
Diode-AND is hard on the limit though (0.6V is LOW) but it works and the board
doesn't have a single LS08 with any unused gates. Also this way my adapter can
be completly removed and swapped for 2114s.

Asteroids 3
Finally the guy I got the other boards from sent me this one over. He claims
the #1 board works, the #2 board works but on each start, the ship turns 180ø
as if someone hits the turn key for the exact amount of time it takes to turn.
But it is perfectly playable though.
So... first things first, the ground pin burnt off long ago and someone
professionally soldered a thick cable to ground. A cut trace next to a DAC has
been patched as well. Someone must've worked on this board before.
(after MANY days I noticed that some ICs must have been swapped, but it has
been done by someone at least as skilled as me)
The 4016s are by Toshiba, gives me hope, but not as much after Hang On...
So what does the board do? Watchdogging. R/W line pulses, at least it tries to
boot. Vector CPU is idle, Z axis is dead.
Switching to self test, I get one short note. But it stops watchdogging.
Could it be the RAM? Before heating up the desoldering station, let's check if
the addressing signals are OK.
Hm. There's a couple signals missing, eg. OUT from the LS42 at L6. The LS174
is on the far side of the board (really!), getting no clock which depends on
OUT, if I piggyback the glue logic, nothing, but also nothing different if I
piggyback the RAM which baffles me. And it stays on one beep.
Let's check CPU and ROM. The pins are very corroded, but all the chips pass.
The ROMs don't match MAME, but are identical to the ROMs from Asteroids #1
(and #2 after I had to clone them). BUS transceiver look good too.
Now I'm 100% sure it's the RAM. And was right, I could've saved me some time
if I just believed the self test (but whenever I do, it's never the RAM and
whenever I don't, it's the RAM).
So now... no watchdogging, no sound and... looks like a game to me! Test mode
shows the gradient, must be fixed then!
No. Dream on. One of the DACs is dead and no one got spares.
We just got a few 8 bit DACs there, maybe I can make a kludge.
At least I get to keep the board, so we can use it as a spare (gimme a DAC and
I'll work).
I bought two AD7355 (or so) which are 10 bit DACs and look a lot like the ones
from Asteroids and went to design an adapter.
Which didn't work. I get an analog signal out of it though and it doesn't look
half bad but the machine draws garbage. Also the signal doesn't go negative
and has ripple. Self test passes, but the game is unplayable and there are
sound problems.
The DAC makes a lovely signal, but it doesn't look like anything the chips
that come after it like. Maybe I need another op-amp to get it there.
Problem is, the signal has blips that go to zero all the time.
While I'm testing, the signal just collapses and comes back, but for ever
shorter times, I notice it watchdogging.
The reset signal isn't ok at all, running around at 1.5V. Since the sound
section is pretty knackered, I start looking for chips that use RESET there.
And I find an LS174 that has reset on pin 1 and shorts it to pin 2.
Desoldering this pin says the short is inside the IC, so remove it completely
and replace it, the machine is back to working.
So the new DAC inverts the picture and does some kind of stair effect that
makes the picture look like it got Jailbars. And still only half the
deflection. At least it's playable.
The sounds are all there except thrust, but they crackle madly, every time an
explosion happens, all the sounds (except thrust) go back to normal for a
short time. And the Saucer shots as well as the ship shots keep their
frequency. The final preamp is an LM324 at P11 and it also does thrust, so I
remove it and put a new one in. The sounds are now perfect except that the
shots still both keep their frequencies and sound like ding ding ding instead
of pew pew pew.
Concerning the jailbars... they're on both axis and don't change if I swap the
DACs around. Found out that if I short DB5 to ground using a 4.7uF capacitor,
one axis loses the stair effect. I checked all the ICs in the chain from
vector ROM to LS374, but all of them are OK. Except one of the LS374 is
missing its least significant bit and that didn't do a difference.
It's suspicious that the graybar only shows 6 lines instead of 12(I think).
It draws ones over the others, I checked with the scope. Looks like a missing
bit, but I couldn't find it after hours on the scope. At least I swapped the
second LM324 which brought all the sounds back to normal.
And now there's only four lines in the center of the self test.
And now the output of the remaining original DAC shows a staircase signal.
And it wanders with the chip, so it turns out both original DACs are bad.
The rest of the operation is just hampered by my lack of experience with
analog circuitry. So I go the way of least resistance (but maybe part overkill)
and use a TL084 (I only need two op-amps, but had no TL082 on hand).
I set up one axis with another op-amp as an impedance converter, but it turned
out that channel looked horrible whereas the channel I fed the signal
directly in looked OKish, so I removed that circuitry, back to one op-amp per
channel and the picture is definitely recognizable. The leftmost line is
jittering and not touching the rest of the diamond pattern. Diagonal lines
have lots of jiggling in them, but it's a lot more playable and can be
adjusted to fill the screen. Let's see what I can do. Forget, that is.
I forgot to supply the +15V to one of the DACs! How could that even work!
Now the geometry is perfect, but the jiggles are still there.
I put a .1æF capacitor between GND and Vref and the jiggles are almost gone.
If you haven't seen a perfectly working Asteroids, you won't notice it,
especially in game.
So if you want a perfect picture, there's some possibilities:
1) experiment with different capacitors, note that high capacities have
problems compensating /really/ fast current changes
2) if (1) fails, put a 78L05 directly next to Vref and supply it 15V.
3) instead of (2) you can always try a Zener.
One of these should give a perfect or at least so next to near perfect picture
that you won't be ashamed of displaying your left-for-dead Asteroids machine.

OK here's the circuitry description for the adapter AD561->AD7355.
These two DACs have a few differences.
-They are still being made as of 2019.
-They are pricey, but compared to AD561, that, if they ever even show up on
eBay, have insane prices.
-They are not compatible.
To tackle the latter, here's the wiring diagram to make an adapter.
The easiest thing to start with, are the data bits. They're just mirrored.
So if pin 1 is on the upper left, the lower left five pins go to the lower
right five pins and vice versa.
561     7355
4       13
13      4
5       12
12      5
6       11
11      6
7       10
10      7
8       9
9       8
1       2+3
2       nc
3       nc (will be connected to the TL082 and the offset adjustment pots)
14      15
15      1
16      16
Last thing we need is the +15V which we can get from the 7815 to pin 14 of
the 7355 and not to forget the cap between pins 2+3 and 15.
The offset adjustment pots go between +15V and -15V (but I think 0V and -15V
does it just as well). The TL082/TL084 needs +15V and -15V as well.
The output from the offset adjustment pots goes to the positive input of the
op-amp.
OK. Next thing is a little more difficult: Next to the axis amplitude pots
on the Asteroids board are test clips for each axis. These are actually
jumper wires and need to be removed. Careful, these are hard to get out
without removing the vias as well. The pin next to the board edge is the
output to the monitor and the pin below that has the board's op-amp output.
This lower pin goes, with a 1k resistor in between, to the negative input of
the op-amp and fed back from the output with a 3.3k resistor (2k-3.9k) for the
X-axis and a 1.5k resistor (1.5k-2k).
Op-Amp 1's output is pin 1, positive input is pin 3 and negative input is
pin 2. Google the pinout, if you're reading this, you're probably smart
enough to figure out the other channel yourself.
Just one important thing though: Center the offset adjustment pots before
turning the machine on, you don't want to fry the monitor.

Asteroids 4
Another assignment, this time by another guy. He bought the machine and says
there's only a dot halfway between center and top left.
What do I receive? Chips with black legs that fall off if you look at them
weird. Although this one looks almost passable. And it has decent sockets.
And it's watchdogging. Even in test.
Let's check our signals... it writes a lot of data, but it never selects the
RAM... the ROMs are often selected, so it tries to write to ROM!?
Why's the RAM never selected? That comes from an LS32 which has an input
hanging at 2V which comes from an LS139 (above the CPU, not that one next to
the vector ROM) (another noteworthy thing is if you remove the CPU you can
test almost everything without signals interfering with the tester and the
entire B-bank just passes).
The LS139's other outputs are between stuck and floating, a piggyback fights
but has no chance against the brokenness.
Where's my desoldering pump? Daaamn now I have to wait until someone is at the
museum again, they threw me out at 11:30pm on a Friday! (usually they never
close before 2am) GRR!
So next day, swapped the chip, no difference on the surface even though the
LS139s outputs are pulsing nicely. It does stop watchdogging if you go to test
mode. Audio unfortunately hangs on 3V and does nothing, so let's go to 8N
(I think) and on pin 3 I see a low frequency followed by a lower frequency.
When I piggyback the RAM that the manual states corresponds to the second beep
being low, the machine only beeps once and that's low. If I piggyback the
other one, it goes beep boop again. Wat lol? Whatever I remove both, besides
they look like a carpenter has installed them with 2mm diameter solder and
managed to break off three pins. One broke at the bend which makes it
irrepairable (unless you're skilled with a tiny Dremel or something).
Chiptester says "no way" anyway so it's goneski. The other one passes most the
time, but sometimes one bit gets nagged at.
With new work RAM installed I get 4 low and 1 super low beep. If I piggyback
two of the vector RAMs it even starts watchdogging again in test, but more
slowly. And no more thumps on 8N.
Testing in circuit says one RAM is OK, one has some bit errors (that one also
gets very hot) and two have a BUS conflict and are untestable. If I look at
the BUS, there's only two times four bits with nothing different and the scope
says one nybble is pure garbage, one of the chips kills the BUS.
So with only one known good RAM I have to remove at least three, and we're
running out of 2114 so I do the Big RAM hack again as on the second Asteroids.
This time I use a 2018 which is a 6116 in a slim case which we got a few more
of. Also instead of a diode logic gate I use the two unused gates from the
LS86 above the vector RAM.
Now the game appears to run, but it plays blind and deaf. At least from
the data signals on the BUS you can see the attract mode is running and you
can distinguish between game play attract and high score screen.
But there's no Z-axis and the X-axis only gets data from time to time.
Y-axis gets data all the time. Weird.
One's missing, K7 pin 9 (LS273) only has half height output and after getting
the signal into the good domain by piggybacking, now it's stuck low.
But that's not the root cause, so I mark the chip and search on.
The other signal of interest is BVLD, that's stuck low. XVLD is stuck low,
YVLD is stuck high. They come from different LS157 which are driven from
different sources. XVLD comes from an LS04 which has its input floating which
comes from n LS191 whose other outputs are OK. The bank of LS191 has a clock
signal that also looks iffy with a high of 2V so either one of the LS191 is
pulling it down or the LS04 is knackered (a piggyback gets the signal to full
strength). Another LS191 has QC and QD stuck low but a piggyback agrees when
it makes good contact which it rarely does as the pins are corroded.
So that's three chips with partial defects already. So I just swapped the 191
and the 273 and that appears to have done the job. Clock is at 3V now, still
bad, but according to X- Y- and Z-axis, the game runs.
And what does the machine say? There's a picture, but it's squiggly like my
incomplete DAC kludge from Asteroids 3.
So I start piggybacking everything and one of the LS191 two left of the bad
one (which was the rightmost one) makes the squigglies less pronounced.
But there's a new fault - the top of the screen keeps shaking, it jumps by 2cm
at a time, but without steps in between.
At least you can play, sound also works, but no thrust or explosion, so bad
noise generator.
Which looks pretty classic but does nothing. One of the gates of the LS86 is
bad! Even shorting the inputs to low or high or different doesn't make the
output toggle.
Stupid, now I have to remove my bodge wires for the BIG RAM chip select,
remove the LS86, put in a socket, replace the bodges and insert a chip.
Now the noise generator works.
The squiggly LS191 passes in the chip tester (but the squigglies are gone with
the replacement). But the 157 left to the 174 muxing the high X/Y signals
doesn't pass. Scope says OK though. But the spare Asteroids board passes this
chip even in circuit. So while we're at it, also remove the weakish LS04.
And the picture is perfect. Apart from the shakies. I gotta scope that in the
machine which isn't possible on an open Saturday.
OK now the board is home with the owner and he says it's dead. First the
player start LEDs lit up for a short time (which shouldn't be except in test)
then they turned off and it started buzzing and now that's all it does.
He suspects the big blue capacitor. I suspect the rectifier and then the big
blue.
Well it was the cap after all. Problem - his power supply and regulator/audio
board both work fine.
And the board's completely dead, it won't even watchdog!
The 393 has output, but the 7474 below it doesn't. A piggyback is fighting but
not toggling but the signal is fed back. Swapped it and the watchdog is up and
working again.
In Test mode it's also watchdogging slower, but no bip, no boop, no nothing.
(and no Z-axis)
So I swap ROMs and CPU with the working machine. The vector ROM crashes the
working machine, swapping it back doesn't make the broken one work.
Since all the chips in the CPU area are still passing, the fault must lay in
the vector CPU area and since that's still something I don't fully understand,
I just check every IC with the tester. Since every chip can make it fail
completely.
So... from bottom to top.
First up is the first 7402 which has a stuck output. Forcing the inputs to
toggle doesn't make the output toggle so that one's bad. Doesn't change a bit
apart from the output now having short blips every now and again.
So test on. There's a 74175 in the same row as the 273s whose outputs are
stuck. But it's getting inputs, but no clock. That comes from the rightmost
273 which has all its outputs stuck low. But since the chip tester doesn't say
it can't toggle the clock to the 175, I pull both. Turned out the 175 was OK
but the 273 was completely dead.
Phew. And now? Still watchdogging, but the activity on the Z-axis shows it at
least tries to render the playfield area *sigh* so test on...
And then I get to the 7474 at what was it A8? that gave me trouble on that
other machine by not working right. On this one, it's easier, it doesn't do
anything at all. Swapped it and the game is running again.
Although instead "ASTEROIDS BY ATARI" I get greeted with some weird language,
it's not French, it's not Spanish, it's not German and it's not even Klingon
(although it's close). In the self test you can see it starts counting at 1
now, so the dip switches get shown as 11122122 instead of 00011011.
But the diamond pattern in test mode is solid now, no shakies.
The game runs, but the scores are weird and the remaining lives are at an
angle. And half the asteroids are invisible although you can see them break
apart if hit and the fragments are visible most the time. The UFO doesn't get
rendered but its shots do. Somewhere there's a bit going lost.
Weird... within a week more chips died on that machine than in the last 20
years...
At least there wasn't much breaking after that... almost.
After putting the original vector ROM back in, the Y axis is amplified too
much, I had to turn down the size pot to almost zero. Everything that moves
vertically makes weird jumps every 3cm or so. If that was a raster game I'd
say it's video address line 32V. But since the machine counts to 1024, that
makes that 128V which is YA7 which I can find on C10 (I think - 74374) Pin 4
and there's nothing going in. It comes from the 157 to its left and if I
piggyback this, the signal comes back and stays when I remove the piggyback
but only until I turn it off.
So I replaced that chip and there were no more errors. Well almost.
I thought I had some trouble with the sound but that turned out to be a touchy
volume pot that I hit when flipping the test switch.
So the game works magnificiently! Only... I can't set it to Free Play.
Test menu shows DIP switch 7 always being 1. Free Play has to be 00 (manual
says 11, but it's inverted). But that can only be the 74253 next to the DIP
switch bank. Swapping that is kind of a waste, but better before it breaks
more and sets the machine to some weird foreign language...
And a few months later the customer brought it back saying it looked messy and
showed me a photo of the test screen. I forgot what chip it was, I think it
was the LS157 in the row below the analog part (with the DACs), I had already
swapped the other LS157 in that row, so I looked closely at the remaining and
one of the signals was crap, so I replaced that and that solved it.
There are still one or two of the originals remaining.
I think I was talking about the 74174, but the guy said the machine was
running for about half an hour since the last fix when it broke down again.
And again there's one bit to the X-DAC missing(or stuck high most of the time)
It comes... from a 74157 next to the last one I changed. One gate isn't
connected, it fails and so does the one above that has the missing bit.
A piggyback brings back the signal, so that IC was swapped.
And the board came back in, this time with squigglies on the X-axis, looks
more like analog circuitry than bad DAC, so I piggybacked the OP-Amps and the
4016, but that wasn't it, so back to the DAC signals, two were missing, traced
them to yet another 74157, one output was stuck high, the other had Low «V
lower than High, running the chip tester over it passed it and now the signals
are back and the squigglies are gone. Whatever, Imma toss it anyway.

Tempest
It's running. But it looks like Barf. You could think it's a bad DAC.
The DAC08 for the scaling effects is also not working.
I did doctor around for quite a while and the signals all look good, so I'm
pretty sure the DACs are to blame. My contraption from Asteroids 3 isn't
working here because the signal has to pass through so much more electronics
like a pincushion correction circuit, an integrator and last but not least the
DAC08 for scaling effects - which is only used by the title screen, all the
other scalings are done by the mathbox.
As a little aside, I had it happen that swapping two of the bit slicers, the
picture did look different, but before the DAC starts working, I can't say
anything. Even better, I don't have the machine - just my oscilloscope.
I did eventually desolder and socket the DACs and when you swap them, the
fault changes axis, confirming a bad DAC.
I managed to score some on eBay on second attempt. The Y-DAC was bad, the X-
DAC seems to be fine because I only swapped the clearly faulty one and the
image now looks like it should. The only error letter in test menu is E,
that's the EAROM on the mathbox board, unsurprising, as the mathbox has its
own 5V, +20V and -20V and I would need two more power supplies on my totally
overloaded bench, so I only had the 5V hooked up. C'mon now people, the EAROM
needs -20V to work, if it doesn't get it, it can't work, thus the test fails.
Back to the DAC08. I buy some replacements and don't see they're SMD.
So I soldered an adapter. Power supply goes into protect (I set it to 0.2A)
and the DAC gets hot. Checked the pinout... the SMD version has a completely
different pinout than the DIL version... WTF Dude! So... made another.
Looks crappy and works just as well as the old one or no one at all.
Too bad the test menu didn't have a test for the DAC08 as some later Atari
games do, so I have to wait for the title screen over and over again.
After a few dozen times of chip inout inout, it suddenly started working, but
when jiggling the chip, it stops. Turns out one pin was bent out of the socket
and doesn't make good contact. If you bend it away, it works, also with my
SMD adapter. There's nothing but a single glass 10nF cap hooked up to it.
My Testofon beeps a melody (with discrete notes, no portamento, but just 3 or
4 different notes). So I swapped the cap and now it works. I did some samples
of other like caps, but they were all ok.

Hang On
Another assignment and before I got Asteroids, nothing really has failed, so
I got something akin to time.
Guy says sprite errors. Sprites are duplicated and have jailbars.
Just before he sends me the board, he says there's only garbage screen now.
And that's how I got the board. Garbage screen.
Self test works fine though. It just flags every ROM and RAM of CPU 2 as bad,
so I suspect a communication problem. They're connected via a bank of LS157s
in the middle of the board. One of which was clearly defective.
Swapped it, still garbage screen. But the self test now only flags one single
RAM on CPU2 as bad. Knowing this is a 16 bit system, having one 8 bit RAM bad
in a self test means it really can be only the RAM itself, but it can't be bad,
it's from Toshiba, these just never ever go bad! Or do they? Removed it, put a
new one in, I lose my faith in Toshiba. What God allows something like this to
happen!?!?! (I know, some people don't find this funny, sorry to those. Get a
job. Hey by the way: Dracula says to a Hippie: "Get a job". Hippie says to
Dracula: "Get a life!")
OK we got a new problem. The machine boots now, shows the title screen and
plays the music, but the road is moved left half a screen and the machine
crashes before even starting to draw the overlay layer or any sprites.
I swapchecked all custom chips, they're all fine. I checked all the chips with
the chiptester, those that failed I scoped, they're fine.
The interconnector cables are OK as well. The videoboard is not the problem
either.

Speaking of video board.
I could reproduce the original error with our Hang On's CPU board.
Sprites are glitchy (no jailbars, just garbled pixels) that repeat horizontally
to the right, so as soon as a sprite starts drawing, it gets stuck in a loop.
Self test shows no errors. The only RAMs that don't go checked are 2148s which
by poking around turn out to be the color RAM.
The chiptester fails almost all the ICs with Mid Level. That board is so
power hungry that most chips have ground floating around at half a Volt.
Scope says in most cases "Okayish", but the sheer amount of chips that get
flagged is too much. Do I really have to change them all?

Going back to the CPU board, thinking I can easily solve the sprite problems
with a video probe. The second CPU has a few data bits stuck low. Not all the
time, but after the machine booted, they remain silent. There is no device on
the bus that wants to send a 1. The working CPU board has all the bits pulsing
vividly. The ROMs are OK though.
There are a few places if I measure them, the bits go high for a fraction of
a second. If I put a cap on a certain BUS transceiver, the game consistently
starts scrolling the road in an erratic way and renders some very broken
sprites, but swapping it didn't do a thing. Timing problems maybe?
Then I checked what's on the stuck data bits. Three BUS transceiver, one of
which is almost permanently active, the other one has the same signal inverted
and the third one sometimes interfering?
If I lift any of these transceiver's pin, I get data pulsing. Are there two
ICs fighting each other or what? But since all ICs on the BUS are working and
only the one RAM that passed the self test is left, I swap that RAM as well.
Let's swap the High and Low RAM. No difference. Wait, the sky's orange now!?
If I swap them back, it's blue. Back, orange. Hm that RAM can't be that OK
then. Installed a new RAM and the machine just ran away with the road and the
bikers. Hooray. F%#k the self test. So the CPU board is fixed now.
Back to the video board. Even with the video probe I couldn't really find
anything suspicious. I can't use the chip tester and the scope says all fine.
So let's start piggybacking. One LS194 near the Sprite RAM restored the
picture when putting a piggyback on, so I swapped it and the machine is now
100% fixed.

Super Hang On
Same assignment as Hang On. We got the non-suicide variant here which
explains why there are so many bootlegs.
What's the problem? We got a character layer missing and the sprites only
consist of random lines. Self test passes 100%.
The CRT test screen stays blank however, the gradients are just gone!
Swapping boards determines the CPU board works perfectly fine.
Hooray for multiboards where you can swap boards to localize the fault.
Boo for multiboards that disallow any kind of measurement when they're put
together.
I checked all of the 7400s on the video board. Those that fail pass as soon as
you short out the quartz. Swapping ROMs or PALs doesn't make a difference
either. Only the Sprite RAMs next to the huge custom ICs are left as are said
customs. Swapped them out and the random stripes change a little.
Testing the sprite RAM chips in the tester shows one is bad.
I've checked all traces for continuity (since it's impossible to remove a chip
on this four layer board without damaging a trace or two) and found nothing
wrong. I conclude that it can only be a bad custom chip or a hidden bad trace.

Super Hang On Bootleg
Only shows colorful stripes and/or garbage, but sometimes INSERT COIN flashes
on top of that. And sometimes you hear broken music. One of the CPUs doesn't
run, sometimes both. They try for a split second to execute code and then
HALT.
On the CPU next to the interconnect I stumbled upon an LS138 that has the
enable pin floating. It doesn't seem to be connected to anything. Bootlegs
often have that, but *shudder* doing so on a MUX is a really bad idea.
And yes, comparing to another identical IC they literally forgot to connect
the enable signal.
So first I dumped the ROMs. One of which reads mostly zeroes with sprinkles of
ones here and there which change position and address every time you read it.
Now for the fun of finding out which bootleg this is so I can get a 
replacement. And then I need a PC power supply to get it running.
And it doesn't make a difference. But there's a huge amount of 74HC356s on the
video board that are soldered in as 18 pin ICs, but they have 20 pins, so the
manufacturer cut off the clock and ground pin making these like 9 or so ICs
useless. Enable is floating, clock is grounded. I dare say that this can't
have worked ever. Which doesn't baffle me completely after fixing a Zippy Race
bootleg where one ROM IC was missing and you could see that it was never
populated which might have been OK if the bootleg was running a different
version of the code or had bigger ROMs...
According to Ari, the machine was actually working once, but didn't draw the
road. Now it doesn't boot most the time and if it does, all I get is colored
lines with INSERT COINS flashing on top of it until it crashes.
So. Out of desparation and boredom I removed all the RAMs and the one LS138
they forgot to hook up the enable pin to +5V. Testing the RAM chips revealed
one that was totally shot. Sadly I ran out of cheap 28 pin sockets
(I won't put the expensive machined sockets in a machine that's worth less
than the 50 cents for the socket...)
And yay instead of colored garbage lines, we get something that resembles a
normal garbage screen, but with lots of flickering garbage on it, some of
which resembling movements that reminds me of something that could well be
from the game. And it repeats when you power cycle the machine. Sometimes.
Since Ari said this machine worked once, I might undo my 74HC356 hacks.
And retry burning the ROMs I found have a few bytes with different contents.
(no pattern though except for one address where both match, so maybe this 16
bit word is correct, because there are differences with that bootleg and it's
not completely in MAME).

DECO Cassette System
Don't get me started on the tape drive. Afaik, there is currently only one
working machine on this planet and it's in Japan (surprise).
The screen has stripes every 8 pixels. They come after a few seconds.
I suspect video RAM or addressing.
It's not the board with the monitor connector, it just does color.
Desoldered 3 RAM chips just to swap them and the picture didn't change.
There are no schematics, since we've got a Model 3 and the only schematics
on the web are for Model 6 and they're pretty different.
The video circuitry is on the lower circuit board. The RAMs are socketed
and the problem never wanders with any of them. All the 7400s pass in the
chip tester.
I should try cold spray, but it looks like a cracked trace.
Another great thing is that someone made a multicart for DECO cassette system
and made 5 prototypes. And that was a year ago and I now kinda wait for the
second run. You basically swap it for the dongle and toss the tape drive out
of the window.
Well the machine's power supply is definitely knackered. Everything's wobbling
and I have lots of 50Hz ripple on the output, which on a switch mode supply
means the primary filter cap or regulation is bad.
So I swapped the big filter cap which didn't do much.
Next there is a battery (I think phalanx is a better word here) of beefy
resistors and the board is certainly well done here and all the solder had to
be reworked. There's three smaller caps in there which look toasty.
Nichicon. 125 degrees. No typo. As in a hundred and twenty five degrees
Celsius. 4.7 uF 50V and as big as a recent 470uF 25V.
I measure the first one and get a choir of cats out of my Testofon.
Out of circuit however it tests just fine. I measure the other two, choir of
cats. But also out of circuit. These two are beyond bad. Swapped them and...
not really a difference. The supply works great for about 8 seconds then the
voltage drops to 4.7V (well before fixing it was 4.2V) but still it has wobble
on it. No wobble on the 12V and -5V rails however (before fixing, the 12V also
had wobble, but not as bad).
So this machine needs a new power supply, because that was all the caps in the
primary section. You might think there's an overload situation, but the entire
cage just pulls 3.5A at 5V (albeit without the -5 and +12 attached)
Hm. The third of those wonky 125 degree caps wasn't 4.7uF, it was 47uF.
So I replaced that one as well and now it takes ten seconds before it starts
wobbling and over a minute before it becomes too noticeable.
Too bad the monitor is making a noise on powerup that sounds like arcing for
a second. Really angry hissing.
But the last time I turned it on, I left it on for a minute (sadly the hissing
is intermittent, not just once on startup) and it didn't wobble.
Kinda sad... all the hissy picture tubes produce the deepest colors...
(OK that's just bullshit because with GORF it wasn't the tube - although the
monitor doesn't hiss on the settings needed for the replacement tube... but
the colors are really washed out).
Hm well it does seem to only do that on startup, because I just had it running
for several hours on a Super Nintendo. (also B+ was way too high)
Well I finally took to it to repair this machine after a Youtuber (Jacklick's
Arcade and Tech Repair Journey or something like that) not just fixed his DECO
but also its tape drive so that there are now TWO working DECOs in the world.
My plan however is installing the Darksoft Multikit.
But before we can do that, the boardset has to work.
So I try to get the cage out (easier said than done) and look if at least
nothing new went bad - but the focus did. I had to turn it all the way up
(actually - down. I measured.) to get it tolerable, it kinda pops in and out
of focus, most the time it's a bit fuzzy, but sometimes it's super sharp.
No clue why, doesn't always happen, couldn't fix it, so *shrug*.
Back to the board. I noted you can swap the boards around as you please so you
can access most of whatever is on top. So let's put the BIO board up because
that's what caused the stripes in the video on Jacklick's machine.
But I couldn't find anything wrong. Then I came across a board with 48k of RAM
in 4116s. So I put that one on top and scoped them.
All the inputs look more or less the same. The outputs have a video signal
except one which is floating. That was easy. Piggyback makes it work, case
closed, chip swapped.
Next hurdle is gonna be the Multikit. And then we somehow gotta put The Tower
on there - because we have a copy or maybe it's the only one in existence
because it isn't dumped and thus it's not on the Multikit.
I do have to thank Darksoft for just sending us one for free! Cool stuff!
Well yeah - but the Test Tape isn't on the Multikit either even though
Jacklick had that on his (and ran it a few times)
Even worse, the EPROM on the Dongle isn't a 27C080 as advertised but a 27C0160
which is a 16 bit chip and also has 42 pins. I can't burn that!
Anyway - the Multikit I got came with a 4k BIOS EPROM which is OK if you have
a Version-8, but our ancient Version-3 has the BIOS on two 2K EPROMs so I had
the choice on splitting the ROM onto two 27C16 or modifying the board to take
a 27C32 instead (you can guess what I did)
Making a Diode-AND (see Asteroids 2) didn't work. What I didn't know is the
machine still had another hidden problem with a combination of a crappily
repaired power supply (that was me...) and an FPGA that wasn't soldered well
(well actually it turned out the 47pF cap that's required for the Dongle to
work with Version-3 boards only read 42pF and needed to be 56pF).
So the next "morning" I replaced the contraption with a broken 7408 where 3
gates were still working (I only need one) and now had a machine that shows
changing garbage instead of static garbage.
So I started fingering the Dongle and got the menu to show up.
So I loaded up Burglartime, err... Lock'n'Chase (which was installed on the
machine last) - and it works. I leave the machine running for a while and
notice the picture gets fuzzier and fuzzier and the focus slipped where it was
and I turned the focus pot somewhere to the middle and now it's perfect!
But when I wanted to play the game, the machine glitched out, lost HSync and
crashes. So load it again, more glitches and HSync got lost completely.
After a while with the Scope I see the 5V slip away. The power supply I was
having a hard time with doesn't feel like working anymore. Seems to be thermic
because it runs fine when cold. Up to 10 minutes.
And I couldn't find anything really wrong with it, so I replaced it with
another PSU I fixed and never got working right. Looks like I suck at fixing
PSUs. This one doesn't turn on unless you turn it off and on a few times.
And also it starts doing similar weird things after an hour or so - the 5V
slip. All in all the Multikit is a fine thing. I can play Burgertime
(Lock'n'Chase is dull - almost every game on the system is dull), but it does
have a few decent games on, so there's never total boredom. Still a rather
weak system (underutilized fits more - maybe I should code a demo for it.
Hey I do speak 6502 ASM so why not?).
And now the monitor spazzes out like the one from GORF, only much -less- more
dammit unless I put the original CRT back in GORF.
(NOTE when it spazzes, the FOCUS spark gap lights up, so here the CRT is shot)
(however it does work when warm, so unless it becomes permanent, the CRT will
stay in)
And a week later it doesn't wanna boot anymore. Now I got yellow lines across
the screen which looks like video RAM and it doesn't boot.
But now it's in the museum and I have a multimeter here and the 5V measure
4.66V. So turned it up and... it runs. For now.
So I loaded up Burgertime, played it and left it running for a while.
I come back and there's just garbage on the screen (the POST before the menu,
just in broken) and when I finger the dongle it boots, albeit with yellow
lines again. They're sometimes there, sometimes not. When they're there, you
won't be able to load a game, when they're gone, everything's fine and dandy.
And it still won't boot reliably, sometimes you have to finger the dongle.
(I already spoiled it here...) So with finer finger control I fingered every
cap on the board and found one that makes it always boot when fingered and
almost always crash when not fingered. Desoldered it and it read 42pF,
replaced it with a 56pF one and now it boots reliably without needing to
insert the magic finger. But Burgertime crashes again after a few minutes and
these f#*%ing yellow lines are back.
So I dusted off the Scope and measured, and the last chip in the middle row
had only High coming out, weird thing. Put a piggyback on, it fights, so put a
socket in and new RAM in and the yellow lines are gone for good. Burgertime
and Super Astrofighter run reliably, it boots reliably, only this stupid
monitor hisses and arcs and then it gets blurry and even crashes the machine.
The focus pot can't decide on a setting. After four hours of runtime, the pot
is turned all the way up (this time up for real, not down).
And after 5 hours it just froze. No glitches, no artifacts, but a reboot got
the frozen static screen like I had when I had the BIOS ROM select logic the
wrong way, so I first suspected the defective 7408, but its output looks
cleaner than its inputs and it does make sense.
Maybe another video RAM?
Nah... that took me some time (well not that long), the 6502 is stuck in RESET
(no clue what the other one does, it's stuck in an epoxy block). That I don't
have schematics for my -3 boards doesn't help because nothing is where you
expect it - in this case, it's not even on the same board - the 555 Reset
generator sits on the DSP-3 board with the epoxy block and not on the SND-3
board (that for some reason also has the video RAM).
Anyway, that 555 isn't doing what it should even though its cap is slowly
charging up on powerup. A piggyback gives certainty - the 555 is shot.
Swapped it and the machine's back up and running.
Now all we need is a temperature controlled servo on the focus potentiometer.
Dude I just played Manhattan for a prolonged period of time - felt a bit like
the AVGN Polybius episode... first it's kinda meh, even rather shitty
(and yes, it's easy getting 0 points on the first few credits...) but if you
play it for a while until the control feels kinda natural (and you found out
what's even the point of this game), it gets really addicting!
So the system has at least four good games
(Burger Time, Boulder Dash, Manhattan, Super Astrofighter)

Another DECO Cassette System
Some french arcade collector sends me his DECO boardset to get it fixed.
And it's the same exact model (with the same exact bodges and also black DIP
switches) so I can just swap boards and isolate errors.
So let's start with the bottom board (hehe the order doesn't matter), it's the
one with the four socketed RAMs and nothing but glue logic on it.
The machine boots, but as soon as you load a game, it glitches out, you get
weird things happen, funny colors. The socketed RAMs have really corroded legs
and the sockets are single wipes, I bet that's all.
The CPU board (I think it's BIO, the one with the epoxy block) is completely
dead, I get the same pattern as CPU stuck in Reset (even though it isn't).
The sound/VRAM board does a little more - you can see the "dead" pattern for a
fraction of a second, then it renders the first test screen (or whatever it is
the Multikit does) and crashes.
Unfortunately, Asteroids 4 had me busy (with the owner being in the same room)
and then everyone went home and no one came the next day.
And the weekend after I also had only one hour of time, but at least I managed
to find out that the IO board (which I call Glue board) had two of the four
4118s gone bad.
Studying the pinout, you can just replace them with 6116s.
Unfortunately they are completely independent so you can't replace two of them
with a single 6116 (4118 have 1k). Only the soundboard allows that.
So at least I took the CPU board with me because it's got a Quartz so I can
check the signals. And I scoped every single chip on that board and found
nothing wrong, except the 74LS10 in the bottom left corner, outputs are dead
(floaty with tiny amounts of garbage on them) and the legs are really rusty.
Yes, that was a Fujitsu, but it got more moisture than what'd kill a Hitachi.
I don't see any difference on the CPU Bus with the chip replaced, but the rest
of the boardset is missing. And the outputs of the LS10 are valid now.
Next time I checked the two 4118 on the CPU (DSP) board and they were OK, I
replaced the faulty 4118s on the BIO board with 6116 and that worked great!
Next I put the CPU (DSP) board in the machine together with the sound/VRAM
board and hooray - Bluescreen with Cassette Error 59 and that DECO stuff and
the error is flashing and no stripes and no dead pixels, but no sound.
The sound CPU doesn't get any clock whatsoever. That comes from an LS74
(the manual says LS92). The levels are garbage, the chip is dead. Put in a new
one and now the bleeepdidibleeep of the error message can be heard. Yay!
Maybe this board now works with the multikit? (cue Twilight Zone music)
Nope, it doesn't. So the plan was to check every single IC (except the RAM)
with the chip tester. There was an LS112 which fails the same way in both
boards. Carry on. There's an LS367 next to the dongle port which fails
completely, but passes on the good board. Changed it, it's trash. Found no
further defects, all chips pass (except that LS112).
Cross-swapped the socketed ICs and they all work fine except for the 80C41
which has some bad contacts but works once inserted right.
Maybe it was the dongle? Nope, still doesn't work. And zero progress.
With this sound board, it renders one screen of garbage and freezes.
The garbage always looks a bit different but always very similar. It does load
the palettes before crashing.
And here is where the odyssey against the brick wall started. I dumped like 16
or more hours with absolutely no result whatsoever.
I socketed the 4118 RAMs on the sound board. Because they passed the test.
I noticed a 220æF capacitor in parallel with the 47æ cap. You can guess how
much of a difference that made. I put piggybacks on the 4116. After 9 chips
tested and 2 replacements killed I gave up on that. Also all the outputs look
to be perfect.
After some time I started to remove the 56pF cap on the dongle, which lead to
the most interesting results - with 47pF or no cap at all it does the same,
but with the finger inserted I get funky colorful glitches, it does pretty
much the same the other machine did with the wrong/defective cap plus my
finger in it - it even looks like it's attempting to boot.
And this is where I'm still standing. That was my new year's weekend - I had
planned, like always, to play some SNES on some arcade machines, watch a movie
on a B/W arcade machine... nope.
To get out of the frustration, I modded the CPU (DSP) board to take the 27C32
BIOS ROM for the Multikit. Really big mistake. On the surface, it appears to
run fine. If you load Angler Dangler, the game loads, shows an intro, shows
the title screen, starts a demo and crashes back into the menu. Burger Time
and Bump'n'Jump run perfectly fine... at half the speed they should!
Flash Boy on the other hand runs too fast and Super Astro Fighters doesn't
show your own shots on the screen.
So I removed the quartz from the board and tested every single IC I could.
In the meantime I noticed my defective LS86 which I used as /CS-generator
(defective LS08s were out) had died some more. So I had to waste an entire
brand new LS08 for what could be done with 2 diodes (but can't since that
didn't work).
And what did I find? Nothing really bad. Chips that fail in the same way on
the working board, some should be scrutinized, but nothing solid except for
three LS367 which show BUS conflicts on the bad board but pass on the working
one.  And there's an LS74 in the same row with the quartz (or the rightmost
LS74 when the CPU is on the top left) which the chip tester complains about as
really bad and the scope agrees so I removed the three 367s and the 74 and the
367 were fine, the 74 flip flop was indeed bad. And I thought, bet that only
brings the shot back in Super Astro Fighter? Which is exactly what happened.
Ah - I just remembered, the working DSP board has a bodge LS00 next to the
quartz which on the faulty board only has an empty space, was never populated.
Let's concentrate on the RMS board because the DSP board is at least somewhat
kinda working and the RMS board is completely dead even though all ICs work.
Installing the clear all pixels mod (shorting the input to the RAMs to VCC or
GND and pushing a Write button), it appears that the machine can't write to
all the RAM (at least the RAM is good though, there wasn't a single low blip
to be seen when I wrote high and there wasn't a single high blip to be seen
when I wrote low). So I went and checked the inputs to the VRAM address muxes
(the four 153s). And there were a few signals with very poor logic low levels
(six to be exact...). I traced them to one single 74367 which passed the test.
I desoldered it anyway and lo and behold the ground pin was bent from the
factory and never made its way into the hole in the board! It had enough
solder on it to make contact and work for a decade, but now the logic levels
are really bad (like 1V for low). So I replaced the chip and if you expected
it to work, you can expect to have a pie shoved in your face. I put the
original chip back and it now has clean logic low levels (<0.2V) so it really
was the connection, not the chip.
Out of desparation I swap the two 374s interfacing to the dongle, sometimes
they get flagged by the chip tester and also High is only 3V on these.
I also piggybacked all the other chips interfacing the dongle and that didn't
do diddly.
The only thing that's left is desoldering all the video RAM chips for real.
Maybe I should pull their outputs low and see which one causes these short
sporadic black stripes in the lines.
And if removing all video RAMs fails, there aren't too many more ICs on the
board so I'd replace all of them and if that fails, I make a duplicate of the
working board. Only takes like forever.
Or maybe I check the ribbon connector, who knows if a pin has been broken.
That wasn't the case so the attack plan says to yank all the RAMs. One of them
behaved strange in the Qix self test. When the first pixel is set, a line of
8 pixel width appears. This gets filled with the real data, then when the next
pixel column&7==0 appears, another 8 pixels appear. You can't catch this
failure mode with "zero the entire chip" and "one the entire chip". Sneaky!
But how would such a subtle fault keep the machine from booting the menu,
especially after we've seen the menu getting loaded even with one completely
dead RAM chip (see yellow lines above). And even though I didn't believe in
the board working, it actually freaking did. After only (almost) 2 months of
work (like about 45 hours)...
Sound also works. So what's left is the Slow Motion malfunction of the DSP
board.
My first idea was to use a Dremel to get the chip legs free. Bad idea, because
through the dust I couldn't see how many traces I scythed through. And after
2« ICs I already spent one entire brush. Immediately thereafter someone gave
me a hot air rework station, what a coincidence! Unforutnately that didn't
work either, the only reason something happened at all was because the paint
expanded at a different rate as the circuit board when heated. This way pieces
of the paint near where I aimed the heat gun for a while started chipping.
So I just got enough paint off to fix the traces I annihilated. But the module
was still dead. If I test the board alone, I can see two signals missing.
So I took it home to get the CPU module out (the desoldering station at the
museum sucks - it doesn't transfer heat) and immerse it in Acetone.
My own desoldering station made short work of the CPU module...
The Acetone did next to nothing, but at least I could get off the paint on the
solder side now, which allowed me to desolder the IC where I killed the traces
which made another ghost of a trace appear really close to the IC.
I'm anxious to see if it works, the missing signals have reappeared though.
And I can now solder wires to every IC on the board.
And it works just as before, at least with its original ROM. So let's put in
that bodge IC that's missing and wire it up. Now it freezes at LOADING...
OK there's a tiny difference far away from the bodge-IC, does that have
anything to do with the bodge? Beeped it through and it's a direct short to
the bodge output. Sure, when 2 logic ICs have their outputs shorted together,
that can't work. So I fixed that and now it looks a lot more promising, but
there's an erratic pixel above the A in DECO CASSETTE SYSTEM and it freezes
there. Power cycle the machine and now it loads some game instantly. Power
cycle again, now it only says LOADING... and it flickers if you push the start
button. Power cycle again and from now on it doesn't react to the start button
any more at all. Everything else works though. Huh stupid. Removed the hack
and that didn't change it, so it's a new fault independant of the old one.
Too bad, a few seconds more and we would've known if the bodge IC fixed that
other fault.
Well OK back to the chip tester. On a 7474 I see a LOAD - 5V on pin 2 and I
don't remember that being there before, the tester can't drive this line.
Unfortunately I don't have the time to investigate that. And there's an LS367
that has 4 suspicious looking outputs, tester says bad, scope says eww and the
chip isn't marked - I marked suspicious ICs when I tested that thing a few
weeks ago, so that behaviour is new. Let's see what I can find out next week.
I found out I was wrong - the working board shows the same exact test patterns
on these ICs. *sigh* keep looking... Almost all the signals look identical on
both boards. On an LS193, there's a difference, some signals are missing on
the bad board, but that makes no sense if the IC was bad. So I went and fetch
the original BIOS ROMs so both boards executee the same software and now the
signals are identical again. But on the LS367s next to those two 4118 RAMs are
a couple of signals - 8 to be exact, floating which show some clear patterns
on the working machine. After some time of tracing traces and signals, I
notice one of the 4118s has only 1V on its VCC pin... duuuh bad socket!
So I swapped it and... still broken, but at least some games will load again
now - Burgertime shows CGA-snow when the screen changes and in the same scan
lines as there are sprites and the game crashes a few seconds into the attract
mode or the game if you started before that. But in those few seconds at least
I could see that the speed is correct now, so that Bodge Hack at least solved
that problem. OK so I swap the other sockets too, with no success.
The behaviour is kinda random, sometimes it won't react to the start button in
the menu, sometimes it crashes while loading, sometimes it crashes while
decompressing (the games on the dongle are compressed), sometimes it crashes
during gameplay. Some games soft-lock up where sprites stop moving in spaces
where you can't hit them. Interestingly, Super Astrofighter runs in cocktail
mode with the screen flipped and Zeroize is on Free Play giving me FF (=255)
lives. The fact that Zeroize works at all and was playable for like a minute,
shows that main RAM access works without problems (Zeroize is compressed
independantly from the compression in the dongle - so it's double compressed).
Annoying is that the signals on all ICs look identical on the bench.
The problem transpires when the machine is running software, accessing some
memory banks elsewhere. Sometimes even the menu gets corrupted.
I suppose one of the enable signals for one or some of the LS367 is bad.
Weird. The enables all look OK, there's a few that are stuck high, but if you
look closely, they do go low on bootup or game load. All suspicious signals
look exactly the same on the working board. At least - if you measure two of
the enables, the machine crashes more often and if you touch the wires while
measuring, the machine goes all wonky, even corrupts the menu...
Now I gotta investigate these two signals... they come from the LS10 in the
corner that I replaced first... grrr! Replacing it again makes it work even
worse and the inputs look healthy. So I removed the socket just to check if I
made bridges or scratched a trace and while I'm at it, socket two of the 367s
that were enabled by the originally bad LS10 as well as the three line RAMs I
can't test in circuit. Of course, no difference. But once I got the machine to
crash where I could just move my hand over the clock circuit to completely
distort the image, so there's a trace turning into an antenna or what?
And now the complete insanity:  Just for fun I put my customer's RMS board in
instead of ours and now it works. There's still CGA snow in some games like
Burger Time and Angler Dangler (!) and some others, but it doesn't crash
anymore and the games run at normal speed.
So on an occasion I will remove the bodge IC and all wires, because I believe
I made a solder bridge somewhere.
Well... no I didn't. And the crashes with our RMS board persist, really bad
with crashes and corruption in the menu already. Barely any game loads let
alone runs, everything sucks. Well okay then, let's swap the RMS and DSP board
so the DSP is on top and accessible to my scope probe.
I turn it on, menu is ok... I load Burger Time and it runs. In slow motion,
but it runs. No crashes, no CGA snow, it just works. With our RMS board. WTF?
I load Manhattan (which crashed wonderfully reliable before the title screen)
and it runs normal. What's going on? Board positions and order shouldn't make
a difference! So I put the bodge back in and notice I must've miswired two
connections. Did I do this at 2AM at night or really early in the morning, say
10AM? A German saying to express disbelief is "Somebody cook me a stork" -
because it's working. Burger Time is running perfectly. Correct speed, no CGA
snow, no crashes. And... Angler Dangler also runs. But I can make out a lot of
errors in the background graphics, there are lots of duplicate tiles in places
where there shouldn't be any. Unfortunately, Angler Dangler reads the tiles
back after generation to generate the mini map and run the game collision
logic. That impedes playability. Mission-X also has tile errors here and there
(I call them glitchies), for example there's a piece of runway in the middle
of an ocean and 8 blocks up (=left) there's a piece of ocean in the middle of
a runway. The only other game that showed such behaviour (and I tried all) was
Burnin'Rubber, where it didn't seem to affect the gameplay.
And after scoping for an entire day with no tangible results, I noticed that
once the machine is running for like half an hour, the errors start to vanish.
Even Angler Dangler runs fine now.
So I give up and say it works. Why don't I party? Because it doesn't feel like
a victory, it feels like defeat. There is still a fault hiding in the board
which I can't find and it has barely any effect on operations.
And in the middle of my non-existing celebration there appear electronic
interference noises. I first thought it was a bad ground contact, but the
soundboard generates these. After some searching I found a 555 timer
generating the noises which can be shut up by shorting the trigger input.
I swapped it out and instead of modem noises I now get a beep.
How is that wired up? The AY-3-8910 sound chip pulls the 555's Reset line to
ground, stopping it. But it's high. The other AY also has a 555 and it's
being held in reset. So I swapped the AYs around. Too bad, that was it.
With a new one, the problem's fixed. I gotta check if the broken one only has
that pin bad and still produce sound, because then I can still use it
somewhere where that pin (PA0...) isn't used.

Cook Race (Burger Time bootleg)
Hey cool we got a Burgertime (so it would be excusable if I'd hack that into
the DECO cassette system). And on the bench I can see one of the spacers is
longer than the others.
Turned it on, got some bad graphics glitching going on. I was hunting that for
a while, found nothing, but if you bend the board like it would have been bent
if it was mounted in the machine with the uneven spacers, the glitches cleared
and the game worked fine. (that's a hairline crack for sure)

Bomb Jack
This one has perfect graphics, but very poor sound.
Some voices are normal, some are very quiet and others are completely
missing. The machine has 3 (three) AY-3-8910 with a total of nine voices.
Scope shows there's a lot more going on than what arrives at the speaker,
but now you try getting four sound effects playing at the same time while
also holding the scope probe AND looking at the scope screen without dying
in game.
So I tried using the Testofon as a signal injector. And from nine channels,
only four produced a sound in the speaker. Going downstream I got all nine
working after the capacitors. So there are nine capacitors, each output
beeps, but not every input beeps. So I swapped the capacitors and measured
them for good measure. Five were electrically open and the other four weren't
too great either. There are nine more identical caps, all of these were
functional, but some were out of spec. And to top it off, I used 40 year old
NOS tantalum caps. Go crucify me. But to be honest, if they go bad, they
won't cause much damage. Also I measured them before putting them in.
Also I noticed the fingerboard of our cocktail table Bomb Jack getting
really hot. I cleaned the contacts and the graphics glitches went away.

Another Bomb Jack
This one looks waterlogged (good that means the electrolytics aren't dried
out :-) )
The rusty parts of the screws on disassembly shows the boardset must have
gotten wet more than once.
But it runs, it just looks like crap. Backgrounds and sprites are OK but
the foreground is just garbage.
You can just barely make out the self test showing everything is OK except the
last item on self test. But it's illegible. So I hook up the other Bomb Jack
so I know which is the last item. RAM4. OK which one is RAM 4? No idea.
Tried piggybacking RAM chips on the bottom board, nothing. But one of the top
RAMs if I piggyback it, there is some change. If I short the lines, the 
garbage changes and nothing else changes. And I see fights on the data bus.
So I desoldered it. Or did I? After desoldering I noticed I desoldered the
working Sound RAM instead of the broken graphics RAM. D'oh!
The socket I put in was already a little knackered so if in 40 years the
foreground graphics start glitching, that's the socket.
At least the machine is now fully working except for sound which is untested.
The amplifier and surrounding capacitors look ugly though.

And two more Bomb Jacks. One just loops the self test over and over (and
passes each time with no problems) and the other one is just dead.
Interchanging CPU-board and Video-board, the dead one worked again.
Putting everything back together, both boards now worked fine - the self-test
looper I think I put the dip switches like they were on the initially dead
board and it looped the test once and then jumped into the attract mode with
no more problems. So the interconnects must've had bad contacts.

Phoenix.
This one's mine (literally - I got it for free, it just didn't have a
monitor. Or game board. Just a power supply and a CRT that turned out to
be incompatible with nearly every chassis - and no isolation transformer,
so I MUST use a TV).
Reiner kindly gave me a Phoenix board I found in his crates of boards he
bought bulk without knowing which games were in there. I checked the
photos and there are at least three Phoenix boards and the one I picked
looked like it's been thrown in the mud and then driven over by a tank.
But it's working and looks and sounds ok and is playable.
But the colors are off. First, there's only eight and they're primaries.
So there's two palette ROMs, one makes primaries, the other pastels.
Maybe one is dead. Nope, they all have outputs. Maybe the resistors are bad?
Nope. Trying different resistors did nothing. So I removed one of the color
ROMs and it made no difference. So I put the other one back in and removed
the other one and now the picture is darker, but the colors are different.
Putting both back in - only KBGCRMYW. Then just for fun I swapped both
ROMs and all the colors matched the emulation...
And now I have a new (old) TV, because with the one I used, the CRT was a tad
too big and didn't fit. I found one on the curb, made in 1982, it's a
Blaupunkt (labeled Universum, which is a no-name brand). But the CRT fits like
a glove, only I put it in upside down and the TV doesn't like the 60Hz video
sync frequency. And what's even better, this TV doesn't have a V-Hold pot,
as opposed to ANY other TV until like 1985... (spoiler: Turns out this TV is
supposed to autodetect 50/60Hz, but doesn't appear to work on my exact
machine). Well there is a connector that is unpopulated and says V-SYNE [sic]
let's see if I can hook up a potentiometer and control V-Sync.
Well sortof. I have some influence on the sync rate (whoop whoop sink rate),
but can't get it to lock. For some reason, putting a 39k resistor between 12V
and V-sync input (pot input), it SOMETIMES locks on and stays on even if you
remove the resistor.
(if you scope it, there's just the V-sync pulse on there)
I could never get it to reproduce. It only SOMETIMES locks on. And suddenly,
the blue was gone. No more blue coming out of the TDA3300, but still going in.
Chip fried. Damn. Hm if I reroute the red output to the blue output, the TV
turns off in order to save itself. Weird. I found an electronically identical
TV on the curb just another two weeks ago, but this one is labeled Blaupunkt,
stereo, built 1981 and has a 26" 110 degree CRT, so probably some resistors
and power transistors are different. So let's test the chip in that one.
And no difference - both chips work fine.
And no clue why the first Testofon session I didn't find the short in the
color amplifier stage. It consists of a PNP/NPN pair, one has a short and the
other beeps completely differently than the one from a working stage.
Reiner's Father's box of spares just by a hair doesn't have these - closest
I could find is a pair that matches exactly, but only for 250V instead of 300V
and the TV runs on exactly 250V. And we're not in China, where zero tolerance
in components is the order of the day. I put them in anyway. And guess what,
they didn't blow, but didn't work neither.
OK out of desparation I swapped red and blue at the RGB preamp.
And the TV doesn't turn off and the screen goes full red with retraces.
Green looks normal, but still no blue.
Put it all back and no picture. Not even raster. But high voltage.
Noticed there's a wire snapped off the neck board. Removed all the wires, cut
them shorter and reattached them and the picture looks as before, no blue.
Hm. Oh wait the solder joints on the CRT socket look somewhat dry - and as I'm
doing it let's swap red and blue at the CRT. And for some reason, I've got all
three colors. White balance is way off, but it looks promising. Put the 
signals back and the picture is normal again.
Man, I bit my teeth out on a dry joint. But I don't know why the RGB output
stage let go off its magic smoke. Maybe because the capacitor leaked 
electrolyte next to the blue stage.
And whyever it syncs just fine to 60Hz now.
But only for a day... I've engineered a second RESET circuit so the machine
can turn on the TV, because the CPU Reset signal is too short it's gone before
the 12V are there. But the TV didn't sync before I put that in anyway.
Checking with my identical Blaupunkt chassis at home I can see it's synching
every one-and-a-half frame meaning that it's running at 40Hz.
If it syncs correctly or not is completely random.
The V_SYNC connection does nothing, because the H-stage produces the sync
pulse. But thanks to the schematics I have an idea on which capacitor to
change to change the base frequency. Idiodically the Blaupunkt died. But not
because of tinkering with it while it was running, but when putting it back
together after tinkering with it. It's completely dead now.
Not even standby. If I'm really outta luck, the remote control IC died.
And the TV died for no reason. My fix didn't work.
Next plan: Put no signal on and measure the free running frequency and then
determine how to increase it.
I found two 0.47uF caps and replaced them with 4*0.1u=40u, now it's free
running at 53Hz. Which will not sync with 50Hz anymore and I want the TV to
still work with both kinds of signals.
I rummaged through the assortments of unsorted capacitors to see what I can
get and I got 0.39 and 0.039 which gives me 0.44uF which the analog computer
in my brain tells me I should end up at 47Hz (free running with 0.47 is at
about 40Hz).
And it now syncs 60Hz perfectly fine every time I switch it on.
Now I need to check if there's somewhere a 74LS08 (EDIT: What I need is a 32)
with a spare gate so I can put in a credit button hack (press Player 1 and
Player 2 start to get a credit)
And then I have to get the front panel fixed in position. It's open and the
left lock is busted and the key's gone and the right lock is missing entirely.
And also the TV still gives a picture that's too bright, black is dark gray.
But at least it's Plug&Play.
And now for real. I put the hack in, so if you press player 1+2 start at the
same time, you will get a credit. I need an OR-gate and according to the
schematics, there are four LS32 but only 12 gates used.
In the machine I can only see three. But still one unused gate.
So now I can screw the door shut. The machine can now go on exhibitions.
The only thing missing is a black black. Otherwise I could have popped in an
LCD monitor like the other Phoenix.
After lots of Googling I know the Telefunken U334M IC is responsible for
volume, brightness and color (and contrast, but they didn't use that).
And after half an hour more Googling I got some incomplete pinout that at 
least tells me which four pins to look at.
And with a bit of scoping and fidgeting the controls I found that Pin 48 of
the main control module has the brightness information. It flows through R563
on the main board. So I put a 1.3k resistor to ground right after that 
resistor and now you can only see in total darkness if you know where you look
that there is still a beam tracing the raster. So I'll leave it at this
because if the marquee is lit, that's enough light to be brighter than this.
OK great, I'm done, but now the joystick keeps pulling right. Before it was
just mechanically stuck, now it generates the right-pulse when the stick is
centered. It's got a light switch joystick like GORF and after a couple of
measurements I see that the IR sender diode's voltage sometimes shoots up when
it doesn't work right and the solder joints don't look too great and my
Testofon also says it doesn't make good contact so I reworked the joints and
it's stable ever since.
And after a long long time someone told me that someone else plugged the
machine in and left it running for four days. Hmm. Maybe someone doesn't like
me. At least there's no screen burn yet and on the surface everything works,
but the shot sound and explosion sound are missing. Since I just had a Phoenix
on the bench for testing (fully functional, with one big color PROM instead of
two small ones) I know where to look (I thought). And yes, noise generator is
dead. But no clue why. There's a single transistor next to the 4006 and that
oscillates on the working but not on the defective board. And I've been looking
and scoping all the 555 timers and found nothing. After beeping through the
traces with the unreadable schematics I see I've overlooked a 555. And it's
dead! (according to the schematics the signal from that 555 go to an LS14 and
then to pin 7 from the 4006. But that's ground! Beeping it through, it's pin
3 - that doodle could have been a 7, maybe a 2, but never a 3).
Well there's a cap missing that's populated on the other board. But how should
that ever have worked? Well maybe I just stick a piggyback 555 on and see, it
fights. So I put a new 555 in and it's back in business.
And after like half a year somebody comes and tells me there's something wrong
with it, weird malfunctions. I see the 50Hz ripple typical of a bad capacitor,
bad 7805 or bad rectifier. When I piggyback the filter cap, the picture gets
better but still hums. Checked the cap, it's good. And it's a Sprague. And
it's overkill.
So I piggybacked the rectifier and that fixed it.
Yay finally I get rid of one of the 10 rectifiers my Boss gave me. He's got a
box with more than 30 of these.

Another Phoenix
The melody IC is missing on this one. So when you start a game, there's only
sound effects, no music. Also one of the color ROMs seems to be bad (like mine
originally). It's gotta be the least-significant-color IC since the primaries
are all there.
Also the foreground character layer has a problem with the 4H signal since the
characters have their (upper because 3:4) left half doubled.
And here's exactly the same as above - remove one of the color PROMs, no
difference, remove the other one, the levels go from 0V to 2V, if you plug in
a chip half way, the scope starts showing intermediate colors, but plug it in
all the way and you're left with 8 colors.
And again the same - just swap both ICs and there's your full 16 color palette.
(16 out of 64)
So and now I'm finally stocked on LS283 again (because the HC variant works in
Le Bagnard and I couldn't get the LS variant). So putting a piggyback on and
scoping it reveals nothing. Trying the others... if I piggyback the first
horizontal LS283 it glitches and then the picture turns normal. Even if you
remove the piggyback. And indefinitely, until you turn it off for a couple
minutes. Marked the chip, it's probably bad.
After getting my solder sucker back, I pulled the chip to no avail.
Turns out it's gotta be a hairline crack since bending the board makes the
fault come and go. I just couldn't find it though...

And another one
When playtesting, I found the Barrier button not working.
At the edge connector, the pin was shorted to ground. Traced it, it goes to a
resistor and a cap and... to ground. Checked against the other boards, it must
be an error with the bootleg. After cutting the trace and tracing the trace
back to IC12 pin 14 (where it should go), that lead to another pin on the edge
connector, so they made a whoopsie and "fixed" it by having a slightly
incompatible edge connector. I cut two traces and rerouted, now it works as it
should.

Rolling Thunder
After 5 years (when I joined FAO), the machine finally started making trouble,
because the monitor made bad contacts with the colors. I knew it's running
Windows XP with MAME since forever, but I never bothered. Now I had to open it
to get the picture fixed and lo and behold, the original mainboard was still
there! Let's see if the PSU is in spec, it wasn't disconnected the entire time
but running without load might have hurt it. Well, 5.6V, a little hot, but not
gonna fry anything instantly. So I hook up the board and 4.98V spot on!
Huh? What's that? I can hear the attract mode running. But the monitor makes
weird noises (spoiler alert: It always does, independant of the picture) and
there is absolutely nothing on the screen. Complete blackness.
It plays blind, so as complex as the board is, the problem must be somewhere
at the very end of the video chain, so there isn't much that can break
without the game completely crashing.
So I put the machine on the bench and started probing round with my video
probe (educated guessing). Weird, almost every signal has H_BUS&&15==0 on it.
But eventually I stumbled upon chips that have parts of the video signal with
no such glitches. What kind of chips are they? Google says they're exotic
PROMs. Since the outputs work just fine, they aren't the problem, but they
output unblanked signals, so if they don't go to black during retrace, the
picture will drown out. So there must be a chip for blanking. Checked the
schematics, after the PROMs comes an LS273 which gets CSYNC for enable and
Pixel Clock for Clock. These two signals are ok, the outputs however are stuck
low all the time. Piggybacking brought back part of the picture, so I replaced
the chip and that was it, the game works fine again. And even though you have
a huge life bar, 2 hits will sometimes kill you. Most the time you're one
hitted. Even on easy. Why do I even get a life bar with 8 segments that's
fluently animated and even changes colors?
Somehow I managed to beat level 3 on the second attempt!
And after quite a while, early 2023 it was out of order - switching it on gave
a Garbage Screen. The 5V measured at 3.8V and the power supply made noises and
some caps were puffy. I removed all the capacitors and put only one back, they
were either puffy or had leaked or measured bad. One cap and the mains cap
were OK, I didn't bother with the mains cap.

Gauntlet
The monitor had a small and wobbly picture since like forever. Well I saw a
photo from before I joined FAO where the picture was OK, but since it didn't
get any worse and/or threaten to blow up, I didn't care.
But now after 5 years it did blow up, so I have to act.
Hm no component looks damaged, but a completely dry joint on the H-linearity
coil which looked like it was responsible for the smell.
So reworked the joints and works but keeps wobbling.
It's a Hantarex 900E where I already have a little experience with the voltage
regulator (resistors like to go open). The +B pot does nothing.
Checking the resistors, all OK. Diodes, all OK. Big transistor... doesn't what
it should, but not entirely broken. Replaced and... got a feeling it wobbles
less, but the +B pot is still dead. At home I compared the transistor to a new
one and could see the curve tracer shows sharp edges on the new one and very
slurry curves on the one from the machine. So all that's left is that small
transistor. Schematics call for a BC237, populated was a BC547 which should do
the job, testofoned it and it's fried. Leakage current: Massive, Hfe:1.
So I put a new one in and the picture was waaay too big. Turned down +B, had
to turn it all the way down to minimum to get the picture to match the size.
It still wobbles a tiny tiny bit, but barely noticeable.
Testing the machine turned out that it has speech capability but it's disabled.
I enabled it and got greeted with Sxg'l%#ch!'ng k..t!
It's not the speech chip itself or the socket, and there's no schematics.
OK now there are schematics, out of nowhere.
There's only an LS273 before the speech chip which now is the main suspect
after checking that the -5V is present.
Unfortunately the LS273 seems to work. The other control signals seem to be OK
though. Piggybacking the LS273 doesn't work but when I start scoping the
LS273's clock signal, I sometimes get intelligible word fragments.
WhIch draws the attention to the bodge capacitor on the chip generating that
clock signal. It says 220, 500V. No unit for capacitance. So I measured it and
got 107nF. So I put a 220nF cap in and speech fell silent, the clock doesn't
go below 4V, so I measured the bodge cap from another Gauntlet board where
speech works and it has very close to 220pF. We don't have caps this small,
but I tried a 270pF that looked like it was salvaged from a tube TV (not
speaking of the CRT) and speech came back, but it says "Speech chip tedt"
instead of "Speech chip test", but in game, you don't notice.
But after a year it failed again and it looks like I can't solder since one
leg of the bodge cap was dangling freely.

Gauntlet 2
Tried this board to fix the other board's speech problem.
This machine has a tiny problem in the character layer, every 8 pixels there
are stripes in some of the colors.
Schematics or not, the video probe should take half an hour max to find it.
(btw. these stripes scroll with the playfield and are limited to it, and I
already checked continuity on the sockets).
The oscilloscope quickly found out that one of the bits of the graphics ROMs
doesn't go all the way to 0. It wasn't the ROMs though. Unfortunately, the
problem was in one of the Atari custom SLAG chips. If you swap the two of
these, the problem wanders. The defective SLAG also gets warm.
However, if you pull the input to ground using a good old short circuit, the
chip reacts accordingly. So I could hook that bit up to a BC547 (or two since
one inverts) to drive the signal. And if it gets worse, I take a 2N3055 :-)

Moon Patrol
It is decided: The Zaccaria machine with the Bomb Jack and Official Super
Bobble Bobble should become a Moon Patrol.
We have a box with 5 (five) Moon Patrols and a Zippy Race (same HW), let's see
what runs.
So since we don't have a harness, let's power them all at a bench and probe
the video outputs.
The boardset draws 4 amps. The Zaccaria Power supply should manage this.
But the bottommost board only has 4.1V left...
But still all 5 Moon Patrols produce a picture that looks an awful lot like
the attract mode of... Moon Patrol. This means they all have working CPU, ROM,
RAM and at least some of the video hardware.
But: On one I've seen the buggy disintegrate just after scrolling started,
one has the picture flipped and one has a shaky picture (but it might be the
scope since I don't have it synched), but with a digital oscilloscope I
wouldn't have even seen that there is a video signal.
So now I made a harness and tested them in the machine. Four of them work.
One has the picture upside down but not for player two, so flip the first two
dip switches and that fixes that.
The fifth has no sprites. Not just after power up, just never.
So whatever died while I scoped it, really died.
I'd fix it but we have three more fully functional Moon Patrols (except one
has some small explosions inverted and show up as pink squares with a ball of
transparency in the middle, maybe a bad GFX ROM) so I put a sticker on it
and called it a day.
But I couldn't leave it alone. Of course I started on the sprite board and 
went from one dead end to another. If you short out some chips (like J1 which
is a 74LS157 I believe), the sprites show up, so the sprites are there, they
just don't get rendered. After a long time of searching and lots of signals
that have 2V as their HIGH but a frequency so high it might be out of the
oscilloscope's reach I got a signal that looked like a very noisy LOW and just
by measuring it it made the left screen border appear.
I hunted the signal down and found it on the CPU board, there's an LS10 in the
timing chain that generates a high frequency signal and it goes into an LS04
that inverts it. The output from the LS10 looks good, the output from the LS04
is no good. If you piggyback it, the sprites return.
I swapped the chip. The chip tester says it's ok. Yeah but not at high
frequencies, not anymore.
The boardset's got another problem, I hear no sound. The amplifier works 
though, if I finger it, I get a nice hum. And by the way that's the closest I
can get to fingering anything... All I could find was that the 384kHz quartz
for the explosion DAC was hanging on one leg. I reattached it and the bit
patterns look just like a working machine and even though it's a separate
system, the sound works now. For whatever reason they used a 400kHz crystal.
The explosions sound way different, but I'll leave it like that because I
think it sounds great! The machine's shaking every time you hit a rock.
So let's test the entire board set. Looks great, the sprites too and it's even
a Moon Patrol, not a Moon Ranger (although the board set doesn't even try to
hide that it's a bootleg) so let's start a game.
Hm. I can't shoot. Everything else works though.
Measured input levels. Got invalid logic levels on both LS244s.
Swapped them out, machine works fine, measured it, looks like the machine got
zapped like 30 years ago. It doesn't look like that defect came with age,
more like electrostatic discharge or a taste of 12V.
And another boardset. And this one is also working with a little catch, one of
the colors is inverted. And since it doesn't get blanked as well, the monitor
doesn't display it.
After unhooking the outputs of each color PROM on every board and still 
getting a signal on the color channel and with another boardset could limit
the fault to the D board, I started looking for shorts and found one.
Since I will have to remove the interconnect and there's no one in the museum,
I will have to take the board home.
There's bodge wires under the interconnect, but the short was right next to it,
they cut the traces they bodged and the way they cut it, there's the trace
touching another.
And I dug up yet another one which has one of the colors bad, the signal looks
like it has to go through a really bad capacitor, super weak signal.
It actually was a bad interconnect that has a pin broken right between the
bottom of the connector and the board (so it's unfixable, need to replace the
header). Well I just clipped the pin and put a solid wire in place.
Looks ugly, but it works.

Traverse USA / Zippy race
Museum was closed and I was bored sitting in the storage building/second lab
with all these boxes of games. And the box of five Moon Patrols had a Zippy
Race in there which is mostly the same hardware.
The game was missing its CPU right away and I already replaced it, but it
doesn't run. I just get a garbage screen that randomly changes every second.
First thing I noticed is that the IRQ is stuck low. Awesome. That took me a
while to figure out. Namely: There are two interrupt sources and they trigger
an LS74 flipflop each which then creates an IRQ that stays until the CPU
services it. And since the CPU doesn't run, it never gets serviced and stays
stuck low.
Second thing the data BUS looks like a battlefield and the machine crashes if
I measure some of the data bits. The BUS is floating most of the time and if
not, you can see lots of half-assed transitions that go from low to floating
back to low, then it goes high, from high to floating and back to high and
there are several voltage levels for floating. Either some chips are fighting
each other or one or several chips are bad and can't drive the BUS anymore.
The Chip Selects look OK. Eventually I got to a BUS transceiver at M5 (LS245)
that has direction switched at high frequency 50% duty cycle square wave.
So it should drive the BUS both directions. If I put a piggyback on, they
fight. So removed the chip and put the new one in, which didn't change an
awful lot, it looks a bit different now and the garbage changes a bit 
differently, the BUS looks way better now, it just looks broken and not beyond
broken but not healthy and it still keeps crashing if I probe the data pins,
even at X10 probe setting.
And even though the CPU board is identical to Moon Patrol's CPU board down to
the chip types and positions, the ribbon connector is different and doesn't
fit.
I just had an idea: The machine isn't watchdogging. Maybe a VSync signal is
modified somewhere and triggers something and while VSync is there, the 
modified version of it isn't.
The LS245 passes in the chip tester. As do all the other ICs, so it can only
be the ROMs. And looking for more, I found two more Moon Patrols, so we have
7 now (edit: it's 8 now). Let's build one line with only Moon Patrols!
Oh wait, our lines are 10 machines wide.
Well I found two more Zippy Races that work both, but they aren't compatible
either! And they have 4 ROMs each instead of 3. If I put the ROMs from the
defective machine into a working one, that one doesn't run either.
After running out of ideas, I set the 32/64 wire jumper like the other board
has it and populated the fourth socket and put the known good ROMs in and
guess what? Of course. It just worked. So I gotta make a copy of that missing
ROM. Kinda weird that Traverse USA / Zippy race (btw. you just flip a dip
switch to change the title) even has bootlegs, the game looks like crap and
the ratings are meh.
Looks like Monaco GP on Valium with real obstacles on the road. More like the
SuperPONG car race, but with graphics.
I wonder if there's a hack for Moon Patrol to run on this hardware. I mean
there's just the parallax graphics layers missing.
And what I didn't mention: The three ROMs were identical to three of the ROMs
from the working machine. In the same place. It looks like they forgot to put
in the fourth one and never even checked. It can't ever have worked. Never.
Too bad my only 27C64 has a broken data line 4.
And now the "donor" boardset stopped working. Jiggling it turned out the
interconnects are a little flaky.

Kung Fu Master (almost the same hardware as Zippy Race)
Interesting fault... vertical garbage lines in the picture (the first pixel is
solid beginning somewhere around 1/4 of the screen), then the next 1-2 pixels
are jagged and sparkly. The stripes are standing still even when the
background scrolls and the sprites are behind them as well.
The stripes can be manipulated and even removed by pressing on the single SMD
custom (yes). After fixing that, I noticed stripes in the sprites.
That I have searched for a while because I didn't noticed that one of two
2018 (or something - slim versions of the 6116) has only a few wires remaining
in the circuit board - most of the chip is missing.
So - reworking the SMD and replacing the RAM fixed the machine, I tested it
and it works fine, even sound, even samples.

Turbo Outrun
It's running for years now without sound. Since Turbo Outrun is notorious for
its Sound Suicide SeePeeU whose battery dies long before the 68000s suicide
battery dies, I left the machine alone and was looking unsuccessfully for
decrypted ROMs on the internet every now and then, but besides finding hints
to their existence, I didn't find the ROMs themselves. Until I emailed some
other arcade repair guy and he sent me a link.
Now I'm standing behind the machine and think there's something wrong with the
world. This machine has a 68k suicide CPU and the sound CPU is a vanilla Z80.
Which also works just fine.
I check data and address BUS and see a pattern. If I go to service menu and
select a sound effect, the pattern changes and reverts after a few seconds.
Welp, that works. So next I checked the outputs of the YM and the DACs and
they have sound coming out of them. Which even goes to the final LM324 on the
board after which it goes into the amplifier.
But nothing comes out of the LM324. If I short circuit the inputs, I can hear
a crackle in the speakers. So swapped the chip, no difference.
Then I noticed that without the LM324 I have sound. A little quiet, but loud
enough for the arcade and the machine's puny speakers anyway.
So I gave up and left the chip out. The silk screen says clearly LM324 so my
successor will know what chip to put in there. I guess the inverted input has
a muting signal applied to it but I didn't check where it came from.
It makes a lot of racket in attract mode. But Hang On does that too.
Originally I planned on desuiciding the 68k, but measuring the battery voltage
got me 3.1V so plenty of time left. And every other Turbo Outrun should have
a normal 68k CPU in this place so even if I screw up, I can always put in the
decryted ROMs.

Blood Bros
That one used to be in the machine that now houses Bubble Bobble. Or was it
the Bloxeed machine?
Anyway the board had a habit of black screen every now and again and was even
left running for a week because nobody noticed it was turned on.
Back then (3-4 years ago) I noticed that the video DAC must be bad because
digital video signals go in and nothing goes out.
And isn't that cool, just a few months ago someone reverse-engineered the
original DAC and now sells replacements. And he didn't publish schematics and
the photos are just barely sharp enough so you can't make out everything.
Through combining the photos and the facts given in the text description and
huge amounts of guesswork, I could get most of the circuit reproducable.
So I sat down and made a replacement.
So, it is a five bit per color DAC. And the input pins are in order, so pin
6 = MSB . . . LSB | MSB . . . LSB | MSB . . . LSB = Pin 20
I forgot which is which, but I think it was RBG or GBR. Green is not in the
middle. Anyway you'll figure it out.
The resistors used are 220 Ohms for the MSB, 470R, 1k, 2k, 4k7 for LSB, I used
2k2 because I don't have 2k resistors and I think it makes more sense and you 
won't notice.
There's two LS374s whose inputs are connected to the color bits and each
output has the corresponding resistor on it. Each resistor of a color has one
end connected to the output of the LS374 and the other end shorted to the four
other resistors of that color. That output then probably goes to the HC368,
but I can't make out how from the poor photos. So I left it out.
So I tested it and got nothing. No signal. But blanking is stuck high.
Pulling it down gave me a signal on the scope that looks just like the game,
isn't even inverted and you can see the specific logo fades from color to
monochrome and then to I think white, but it makes perfectly sense on that
scope screen. I wonder what a monitor makes of this signal.
(possibly it will need a catch up amplifier which is likely what the HC368 is
for)
But now I must find out why blanking is stuck high. Turns out there is a PAL
generating it and just this output is dead. Unfortunately, no one has a JEDEC
so I'll probably have to reverse engineer that (just coming from the Space
Train Mac). But sadly, none of the inputs look even remotely like a sync or
blanking input, so I gotta need a logic analyzer for that.
Or check if Raiden has such a PAL that does the same job, as Raiden has the
same video DAC.
I'm kinda mad at myself for not looking which pins are used at all, because I
know at least one input is tied to ground and several outputs are unused.
After all, I only care for the one.
Well. I can read the GAL. And if I copy it, nothing changes and it seems to be
identical to the one off the web. So I check which of the input pins have a
solid signal that never toggles and see what happens if you toggle it with an
alligator clip. And pulling pin 6 high enables the screen - with a 16 pixel
high area of garbage below the playfield, but I could live with that.
So where does this signal originate from? A custom IC. And I beeped everything
to make sure it can't come from anything that's feasible to replace.
Awesome. Bad custom IC. So this is how it's gonna stay - perfect picture
(even the colors), just a Tad (pun intended) (splat) (who's throwing tomatoes
at me?) flickering garbage at the bottom.
Uhh and the colors aren't perfect and some palette animations are working and
others are missing. The river level has a non-moving white river, but the
background waterfall animation works normally. The video signals (btw. only
4 bit per color, not 5 like the DAC) are generated by the defective custom, so
this chip must have more wrong with it... well it is playable...
Well I found a repair log from someone where the backgrounds completely failed
which had the same exact custom IC bad. He -salvaged- stole one off of a
Raiden II (which is a crime and should be punishable - Blood Bros is a
thousand times better than Cabal, but Cabal is so incredibly crappy that this
makes Blood Bros being meh and Raiden is an awesome game!)

Green Beret Bootleg (1985 Konami)
Well I was looking for a needle in a haystack that didn't have a needle in it.
CPU isn't running. HALT stuck low. Where does that come from? From two 
opposing ends of the board. Both forks like mad and then after a few gates
come flipflops and latches and address decoders and I'm out.
Wait a minute, if you start the machine, it pulses for a few seconds and HALTs
then. And what's on the video output looks to orderly for garbage screen.
More like a raster. So I hooked up a monitor. It is a raster. And it says
RAM OK ROM 3 BAD on top of it. Duh. The piece of paper on the board only said
"Green Berret [sic] \n defect [sic]". If whoever wrote that copied that little
error message from the screen, I could have saved myself an hour of searching.
So I dumped the ROMs to match the bootleg and the only ROM that was 32k had 3
errors total, two of which went from 0 to 1. Burnt a new ROM and the game runs.
Sound also runs.

Jailbreak bootleg (Konami 1986 similar HW as Green Beret)
Pretty dead, classic garbage screen. Dumped the ROMs, hooray, one doesn't
match with MAME. Wait a minute, it's 8k? Oops I read it as a 32k. Since it
reads FF for the lower 24k, it can't match. OK so the ROMs do match.
So let's do a visual inspection of the solder side (I wanted to say backside
but I already fulfilled today's quota of terrible jokes), and just like the
other boards that were just thrown together in a cardboard box, they have pins
bent so they short to traces as well as cut traces.
This one doesn't have cut traces, but lots of bent pins.
And now the machine is running!
On the monitor however I see that the sprites have jail bars (how fitting!).
Shorting data lines I could locate the sprite RAMs. There's four next to each
other and it's a wild mix of Hitachi and NEC. And one data pin has signals not
going all the way down to ground. Testofon says short to +5V, multimeter says
12 Ohms. With a milliohmmeter I located the chip with the short. It's a
Hitachi! Hard to believe. I've seen it on Youtube, but couldn't quite believe
it. And now I experience it first hand - Hitachi chips CAN fail! So what's
left to trust in on this planet if not Hitachi ICs? Next I'm gonna wear a tin
foil hat! (there's no quota for jokes that are just bad).
But what's that? There's another bit that has problems reaching Low.
It's within limits, but I measure 136 Ohms while the other data pins don't
register (on the 200 Ohm scale . 5« digit display). So the fuse is lit on this
IC, meaning it has to go. I hope it's the same IC as the other broken bit.
But sadly it isn't. And I can't pinpoint it with my milliohmmeter (I get
oscillations). So all that's left is remove all RAMs and put sockets in.
And the evil doer was the second Hitachi. Since I don't have 8k RAM here I
thought what's to lose if I try to short the bad pin to ground?
(well... the other 7 bits)
I tried that. It sunk 500mA and didn't go away. So let's crank up the voltage
until it burns through. It worked (at ca. 8V).
Sadly, I stopped the process early on the chip that was almost OK with the
result of the bit being completely dead now. The other chip now goes to .8V
which is a lot better than before (4V for Low).
Wait a minute - I got 8k RAM. At the last Classic Computing, I got a bag of
Serial Interfaces for Starwriter needle printers from around 1986 for free!
And every one of them has an 8k EPROM (hooray!) and an 8k RAM (HOORAY!) as
well as a spider chip CPU, probably a microcontroller that's compatible to
something known, and a tiny bit of 7400 logic and a few level shifters.
So I can salvage these. And it worked!
And actually I found the datasheet for this spider CPU. And it has all the
commands including their opcodes and what they do. It seems to be the
predecessor of the ATMega. It has 8 ADCs (and can sample them at very very
close to 22kHz), a bunch of GPIOs, a UART, timers and what have you.

Dribbling (Model Racing/Olympia)
Just juice it up! No sync. But a picture. And yes indeed, the last chip before
the connector is just a buffer, signal goes in, nothing comes out.
Weird. I was positive I still had 74LS06 left. Well an LS05 does in a pinch
and I have tons of these.

Espial
Same problem. Picture (sort of) but no sync. This time it was a 74LS08 that
had one gate bad. Now I can choose between two faults: Constant reboots with
RAM CHECK OK on screen and a garbage screen with a vertical resolution of ca.
100 scan lines instead of the usual 240 and the TV won't sync it.
The underside of the board shows lots of bent pins but it doesn't get better
from straightening them.
At first I didn't even know what game this is. No labels on the board.
ROMident shows most the ROMs to be Espial, except 4H, so I got the MAME ROM
and compared 4H and it's identical, except some bytes have bits flipped from 0
to 1. So the board must've been stored in a sunny place.
(also interesting, the game is made by Orca, the board very obviously isn't,
the ROMs were detected as original though)
(we have another board that WAS made by Orca. It has the logo and everything
and the make quality is fairly unique)
(ah and the machine should show
RAM CHECK OK
ROM CHECK OK
when working)

COP 01 (Nichibutsu)
Runs, but has stripes in some sprites and parts of the title screen.
Whenever that happens, I short out data bits for the sprite ROMs and see what
happens. The upper bank has a data line where nothing happens if I pull it low.
Where's that connected? To an LS377. And two of its outputs are dead even
though all inputs look the same. Piggybacked it, much less stripes in the
sprites. The sprites now have weird stripes. Hmm. According to the ROM labels,
the leftmost ROM from upper and lower bank must've been swapped.
Putting them back restored the picture.

Popeye (Bootleg)
Major graphics glitches. Lots of striping. Shorting the data lines showed one
where shorting it made flowers appear over the entire screen - shorted to
address line! The solder side of the board showed many bent pins one of which
was responsible for that short. Now I got much better graphics, but still
stripes, but they aren't solid! Shorting all the databits made black lines
appear in the position of each data bit, so it can only be the ROMs.
If I dump the ROMs and look at their bitmaps I see the exact same stripe
pattern. And looking at the ROM itself shows there is a scratch in the sticker
on the UV window that has the same exact shape as the stripes in the data...
...I won't even have to burn a new ROM, just burn the image over this one.
And I did burn the image over the original ICs, didn't erase them, didn't use
new ones. The one that really got too much sun took several attempts (my
GALEP was a bit shy - Update: It blew a transistor) but eventually it worked
and there's only a few lines in a few sprites and then the color around
Popeye's eyes doesn't match his face but it looks like it's been sloppily
programmed - after all it's a bootleg.
Hehe - I found another Popeye bootleg in a different box and it has exactly
the same color mismatch!

Puck Man (bootleg?)
The Namco customs (only 2?) are broken out onto two nice little daughterboards
one of which covers the CPU...
And I get Garbage Screen. The CPU attempts to run and doesn't make a
difference if I swap it. After short testing I see RESET goes high immediately
after powerup. If you reset it manually, you get a black screen with 2 Pac Men
looking like part of a cut scene from Ms. Pac Man.
The Pac men have sprites through their graphics and the machine immediately
crashes. That behaviour is independant of the CPU used (unless you put in a
defective one ;-) )
And a week later, RESET just works and it goes to the "cut scene" immediately.
I dumped the ROMs, they all match. The data bus looks crappy. The back side of
the board has some bad scratches, one of which did cut a trace.
Patched it and that's it, no more problems.
Now all that's missing is a Ms Pac Man Gender changer. I'll do that on X-Mas.
Unfortunately I couldn't find any half-unused ICs, so I had to salvage an LS74.
But it works. For resetting the machine, 2 diodes did the job.

Nibbler
I'm stupid. I swapped +5V and GND and wondered at crummy logic levels, then I
noticed the 5V is only 4V, then I noticed I got 4V on ground, then I noticed
everything got damn hot and THEN I noticed I had the polarity reversed.
The pumping plant on whose hose I was standing has blown up by now, has been
rebuilt and blown up again in the time I took to notice (to stand on the hose
is a German saying for failing to understand, usually used when the mind is
elsewhere).
Really surprising though, the board still worked. After a few weeks though, it
was dead. No sync, no video, nothing. CPU is halfway running, video board is
quarter way running. Pretty soon I saw bad logic signals on two transceivers
next to the 6845 CRTC. Removed the CPU board, chiptester says one is OK, the
other one is bah! Swapped it, video board runs three quarters, still dead.
Well. There's 74LS283s hanging on the transceivers, maybe they got some.
Only one passes, the others fail partially or completely. With patterns that
look like "works bruh". At a certain spot I notice that the game is running
and produces a video signal that looks like attract. Turned it off, turned it
on, works.
Back in the machine in storage (I made an adapter and put the board in a
vacant wall mount arcade machine) it was dead again. CPU bus has 2 bad bits.
One rom is stuck low on chip select. Removed it, BUS dead. And then I see it
uses the same sockets as the early Nintendo games with the same problems.
Put the ROM back in and the game runs.
The game was flaky and did fail self test with one bad ROM. I jiggled it and
the fault cleared. These are deox-me sockets.

Elevator Action
This has me tense. I didn't get that one to do anything. And the manual says
it requires external reset, 5V, 12V, -5V and -12V. Great.
Plucked the boardset apart (you can lay them out to become one long string of
boards and still work). There's absolutely nothing apart from the amplifier
that requires anything but 5V.
And it doesn't require external reset since it has a watchdog (spoiler alert:
not all top boards (sound) are created equal and some might not have the WD).
The watchdog takes like 3 seconds for initial reset though.
And this is the behaviour: Blank white screen (maybe off-white). With a black
border around it, so blanking works. After Reset, the screen is erased and
never changed.
This thing is as complicated as can be (typical Taito, you gotta be a genius
on the brink of insanity to understand these machines, maybe they push you
over, there's a video sub-BUS whose data come from another BUS whose data come
from the first BUS only and there's no way of getting any data into there,
you can only sort the static that's there on powerup. Or it operates on magic.
Or the schematics are incomplete (they often are)).
And the game is so simple (graphics-wise), I could probably implement it on a
Galaxian if I spoke Z80 ASM.
And once upon probing I saw everything has changed! The video signal looked
like an attract mode and stuff moves and the screen scrolls and when I tried
to hook up the TV the magic has gone and never came back.
Now for the big fun of checking continuity everywhere... I could do that on a
Streets of Rage bootleg, but that at least runs with lots of graphical
corruption and the quality of the board makes it pretty clear I don't have to
look for defective ICs here (edit: I fixed the game for the most part by
bending pins straight and fixing a few bad traces. I don't know if the game
will get an entry here as it's really just shorts'n'opens).
Btw. we have 2 (Edit: at least 3) original Elevator Actions and a Bootleg...
guess what works... and also needs 5V only...
OK correction, Elevator Action only requires 5V to run.
I found another specimen with the crypto CPU (or something), the bottom board
is completely different (5 boards btw) and it works perfectly fine.
So I can check the other two boards in the non-working stack and they work.
Neat, the bottom board has hacks to the brim with almost finger thick cable
bundles (there's not a bodge wire. There are busses of bodge wires!).
And a funny lackquer coated daughterboard branded "Video Tuning" (seriously)
with a Z80 (Noooooooh that's not a Z80, believe us!) and a few interface ICs
(likely 245/244).
The daughterboard's outer 4 pins are soldered and since the board sticks out
of the stack and was handled rough, it broke off. So I bypassed them with
wires and now the entire boardset is working, giving us 2 (original) Elevator
Actions.
Finally it's time to install that boardset in a machine.
But it doesn't wanna run reliably. The CPU socket is wrecked though. With a
new one, it only runs if you hold the crypto-CPU board at a certain angle.
GRRR! So I soldered the crypto CPU to the board through a ribbon cable.
The game runs and is playable, but the blue is whacked, especially right of
the screen there's like 6 characters copied from elsewhere on the screen.
And the blue signal does look off on an oscilloscope.
Anyway you barely notice it. What you notice though is that the 5V are only
chilling at around 4.6V even with the PSU cranked all the way up. So I
eliminated one of the in-between-connects (the machine ran Kangaroo and
there's a Kangaroo-to-JAMMA adapter in there someone made) and now the machine
is native JAMMA and we're at 4,96V. Great.
After 20 minutes the game crashed though. If you turn it off and on, you can
choose between garbage and unicolor screens. After 10 minutes I turneed it on
and it worked, so that sounds thermic. Let's see which of the NSMs has a fan
and doesn't need it.
Amidar has one and it's been cut out, so I fixed it and built it in, it also
needed some oil.
Only... the machine doesn't even run cold anymore. Just garbage or unicolor
screens. So I got a spare from the storage (see below), soldered the wiring
harness around and put it in the machine. It works with small problems, see
below. It's running on the bench with variable framerates between 2fps and
30fps. It does work if you jiggle the interconnects.
But now that I have it at the bench I can care about the blue.
Nice, the outputs go from the 9 bit palette RAM straight to a resistor ladder
DAC... according to the schematics... here there's two 7417 buffers in between.
And one of the pins of one of the two ICs doesn't drive. Swapped it and blue
is now normal.

The third Elevator Action's gonna get hairy.
Got the same symptom as the first, but a different cause.
Swapping boards concludes that the sound board kills the entire thing and
there's not really anything on it the machine needs to boot, just the entire
sound CPU with 4 (four) AY chips (and it doesn't even qualify as elevator
music!) and a bunch of 7400 logic. The chip tester doesn't really fail
anything, there's just a PAL looking suspiciously silent...
Nope, had to look closer. There are needle pulses...
...I found a fourth Elevator Action that also works and uses the same PAL, and
it really is the same, you can swap them, so they're both good.
Now my GALEP stupidly enough can't read PALs, so I have to read it as an EPROM
andreverse engineer it. Why should I do that when both are working? Because
they aren't dumped! Well if it's nothing more... been there done that.
And here's the real deal: The game actually runs. You just can't see anything
as it's hidden behind a flurry of CPU data that somehow got on the video BUS
(it's not the sound CPU's data).
Ah btw. the chip tester did fail 3 chips in a way sufficient to mark them with
an X. Now I have two boards, let's test the other. The marked 283 fails in
exactly the same way (so false positive), the two others pass in circuit, so
let's get them out. Both pass out of circuit, but replacing the LS374 next to
the PAL made it pass in circuit too (which it didn't before) so I'm sure I
must've fixed something. And now the 367 below it also passes (with the
original one). Sadly it didn't make any difference.
And that's weird... the boards aren't compatible it seems. The "defective" one
has a 6MHz quartz along with the driving circuitry, which is completely
missing on the working board!
And then I found a Tarzan that has absolutely identical hardware (BL though),
including that quartz on the sound board. Maybe that board would work on this
hardware. That Tarzan seems to run though (sync is off, my TV won't sync).
Oh kay... wait for it... the video RAM (9 bit by 64) wasn't in properly, no
Ground. And now it's back to completely non-working. But it had... you could...
see the game running behind all that static flickering garbage.
So I removed the RAM, turned it on, plugged the RAM in while running and the
game runs. Picture is flipped, has a background layer address error which
sucks for the Score-layer since the entire screen is now filled with semi
transparent squares.
But... I could only reproduce that twice, now it's back to dead.
And now watch closely, you're never gonna believe what was the problem...
That one ROM-board sitting on the sound board was put into another machine for
testing and worked. Which is weird because the ROM board was the problem.
And no clue where the video garbage came from because either it works or not.
So after an hour of signature test without results (I had another boardset
where the interconnects were missing, the soundboard and both daughterboards
were working - instead of a PAL there was another daughterboard) I just for
fun put the ROM daughterboard of the other one on it, because it had a
different shape which allowed me to check more ICs on the main board and it
worked! So I swapped the ROMs, that wasn't the problem. The problem was that
the four LS193 are almost never used - just for super short periods of time,
and one of them was bad - they're all early OKIs.
And by now I made new interconnects for that last board set and it works.
After this board had to stand in for the board from the previous log,
everything looked OK, only... credit sound and and shot sound are missing.
Maybe this version doesn't have a credit sound since I swapped all the AY
chips and it didn't come back, I could only fix the shot sound, looks like the
middle one from the three next to the Z80.
And after I finally made it through level 1, level 2 had corrupted graphics.
And now everything had corrupted graphics. So I swapped the ROM board with the
machine above and everything works wonderfully again so I guess one of the
161s is knocking on heaven's door, but the chiptester says they're all OK.
And now the machine's power supply could only handle the 5 board stack for 5
hours before it died. Even with the cooling installed. The 5V are missing.
Uhh weird I coulda swore I wrote that down... there was a short circuit in the
board stack. I took the board stack apart and... no short. Put it back
together and it's running since then.

Jungle King (aka Jungle Hunt)
I'm fixing this one for Retrogames Karlsruhe. They originally wanted to give
me a stack of broken boards when I visit them next time, but that turned to
"Can't find it, can't find it, not allowed to give you this one".
"Hey how about that one?" - "Hey yeah he can fix that one".
He plugs it in and you can clearly see that video ground is missing and the
game is playing behind all the hum bars. On closer investigation, the video
ground pin on the Jamma adapter isn't connected to anything on this bootleg.
For short: I only got 2 boards to fix: Jungle King and Street Fight.
And Jungle King works (and Street Fight is unfixable due to bad crypto CPU).
Just the graphics are wrong. And the text is completely white (every pixel
set).
On the ROM daughterboard on the sound board (next to a discrete reproduction
of a PAL) there are a few High-levels looking ugly (you only need 6 years of
experience to see that...). The chip does occasionally get input pulses
(but... only when the screen totally changes, so only once every couple dozen
seconds), but all outputs are High-ish. On other chips, some of the outputs
toggle after inputs like these (LS193). Piggybacked and after the next screen
change the background and sprite graphics were fine but text was still all
white. Not even the title screen changed that.
Does that machine only ever load the font into RAM once at boot time?
Reboot... and it's working.

Street Fight
I just talked about it. I got two of them here, one from Wolfsoft, the other
from Retrogames Karlsruhe.
The Wolfsoft one has a ripped off... Transistor? Google says it's a Reset
Generator. Good, but the pins are ripped off at the case and the device is
unobtainium. Do we have spare gates somewhere? Yes, massively. So I tucked a
contraption with a cap in there and... nothing. I get the choice between
garbage screen and black screen.
When injecting signals I could get it to boot, only to freeze with a Coin
Circuit Error.
Later I find out that I could voltage glitch it to get it to boot, Attract
mode shows no faults, but no coin up and always crashes with the coin circuit
error.
The Karlsruhe one boots on its own at least, but doesn't coin up. Well it's
got the same crypto CPU as Bubble Bobble bootlegs that loves to break and
there too it won't coin up.
Swapping the crypto CPUs made the behaviour change boards too, so now
Wolfsofts is booting reliably and Karlsruhe's isn't.
So it's pretty clear that both crypto CPUs are shot. And since Street Fight
is about as popular as the paint on the inside of a Bubble Bobble cabinet,
there's no way someone's gonna reverse-engineer them and come up with a
replacement.
Unless... the Crypto CPU from a Bubble Bobble bootleg just works.
Aaaand... it doesn't.

Halley's Comet
Erm, ok, doing this again? Right after the last there was Another Halley's
Comet, so the article got deleted. From my hard drive. Which is in a 486.
Which is offline. The keyboard I have hooked up sometimes glitches though.
(what works often is that if you let go of Shift in a certain way while typing
a letter, the next letter will be capital too even though you're not pressing
Shift. And sometimes keys bounce and you get two. This is completely
independant of the keys. And sometimes after a space it inserts a copy of the
last non-space character you typed)
OK same problem as the one Elevator Action up there - the picture consists of
CPU data snow where you could catch glimpses of the game running. It's not
quite Elevator Action, this one uses 32 4116s.
I shoulda looked at the bottom of the board... looks like it went through a
meteoroid storm. For some strange reason, only two traces have been cut.
But these made all the difference, it works now.
And it really shows what Qix hardware on steroids can do. It looks like the
palette is 4096 colors because they do show off on the title screen.
And this game is a hidden jewel. A bit like Scramble Formation (also Taito)
but much more fun.
Sound doesn't work yet, but the sound board also has some scratches.
One chip was so badly scratched you couldn't make out what it was, I had to do
trial&error to find out it was an LS74 (and then discovered it was
silkscreened under the chip).
But none of the chips fails, so... there are just two volume pots to replace
and let's see... the data BUS on the oscilloscope looks like the typical
endless loop of a sound CPU having nothing to do but juggle empty channels.
So into a machine and... runs. And sound works, doesn't look totally broken,
no clue how it should sound, but it does sound cool, there's background music
for each level and appealing sound effects - everything done with AY-3-8910)
And in an arcade machine with a decent joystick and sound and a bit of
practice that game is a lot of fun.
This game is a small masterpiece!
What's also awesome, it's not too hard. When you die, you do level all the way
down, but firstly the enemies arend OP AF and secondly after dying it starts
raining asteroids containing powerups.
The end bosses are a bit tricky without adequate powerups.
Play some Whack-A-Mole when the moles are shooting at you AND the normal level
enemies are still coming like they haven't got the memo that you're in a boss
fight. But since the level of difficulty isn't *that* evil, this game is a lot
of fun!

Another Halley's Comet
It's dead. And two of the 32 video RAMs have no or bad output.
CPU is stuck in reset. And dude that massively overengineered reset circuit
had me puzzled. There's like 7 or 8 transistors, a cap (which was pretty shot
but that's not even the problem), tons of resistors and then I spotted the 2
Zener diodes. 5V and 12V. And the 12V from the video board don't go to the
CPU board. With another alligator clip bridging the 12V, RESET is released and
the game runs.
I wasn't sure if it's Halley's Comet because the TV didn't sync it - turns out
the TV's composite input half-died.

Space Cruiser (I think) (Original Taito)
This game has no title screen but a really long attract mode. It also has
quite similar hardware to Halley's Comet, but this one has speech.
There are stripes and other garbage through the picture. They change and
disappear if I jiggle the ribbon interconnects. Seems to be the cable itself.

Jackson (Zaxxon BL)
Doesn't clock. The clock generator circuitry is a WTF with a handwound toroid
transformer with ~15 turns primary and 2 (two) secondary. And that goes to an
exotic (ECL) IC and from there to the system.
It doesn't do anything though. The quartz might run, tested it with my flying
74LS04 "rig". Frequency is garbage, but I can't measure it with my 20MHz scope
and I suspect the slow LS04 clocks it at its base frequency instead of the 3rd
harmonics (or something). The thing has 48.66MHz. And I idiot wreck it by
breaking a leg off. And I can only buy 48 MHz quartzes. And it still doesn't
work (checking the transistors out of circuit, they look OK).
Cool, in the Arcade Workshop, there's a board labeled Future Spy that looks
suspiciously identical to this one - including the clock generator.
Now the Workshop doesn't have a PC PSU with alligator clips, so I check the
schematics if the clock circuitry needs anything besides 12V. And nope, 12V is
all it needs. So juice it and everything is oscillating. The secondary with
the home knitted ferrite ring is a little weak, but at two windings that had
to be expected.
So let's compare it to Jackson. And it clocks! And looks just the same!
(it's now running a wee bit slow though).
And still nothing out of the 10125 on the secondary. I wanted to order them,
but Reichelt didn't have these. So eBay. Hey wait, don't we have a box full of
ICs? And by chance there were dozens of them. So I piggybacked one.
No difference. Bent up the output pin, works. So removed the chip, put
the new one in (as an exception without a socket because timing critical),
runs. I never had that happening. The defective IC is a ceramic IC, which
rarely fail. And it was made by Hitachi, whose chips just never fail!
Well ok it clocks, let's see what the rest of the machine is doing.
It runs! Has a few minor glitches in the background, but there might just be
an address line missing, maybe a scratch between two ROMs or a bad socket.
Weird... I make a JAMMA adapter and test everything, it runs with no problems
at all, but the picture is upside down. Re-plugged the monitor (it had two
sockets for the yoke, one for 180ø rotation), now everything is awesome, but
some sprites look like on the NES when there's a lot of stuff going on with
sprites flickering or becoming invisible. That's really fair when you die
because something invisible hit you.
First I thought it was the sprite RAM and cross-swapped them to no avail.
So I measured the 5V rail and it came back as 4.25V. Strange enough that it
runs at all. Turned it up and now the picture is flipped again and has weird
glitches that look like someone tries rotating the graphics by 90ø which have
been pre-rendered to give a ~35ø isometric pseudo-3D look.
Plugged the monitor over, still bad. Jiggled everything, the picture flipped
again and the graphics are correct now and it's solid now.
After ~20 minutes of playing, the Shot sound drastically shortened.
(still sounds like a laser shot though). The internet says the caps on the
sound board suck. After swapping all the 2.2æF and took samples of the other
caps (which tested fine, the 2.2 ones were marginal) I checked the schematics
on the shot circuit and blindly replaced all the caps in there even though
they measured fine. Might have worked - it still sounded fine after 5 minutes
and lately the shot sound sounds normal for like 30 seconds after powerup.

Future Spy
1985? I gotta check that. It's exactly the same hardware as Zaxxon which came
out in 1982 (or 83?). (I checked. Zaxxon came out in 83 and Future Spy in 84,
but this board which is almost certainly a bootleg, says 85)
No one seems to know Future Spy, but what Zaxxon was to Scramble, Future Spy
is to Xevious. So yes, Future Spy is Xevious with isometric 3D graphics.
Has a bomb button too, so it plays like a mix of both. It would be a great
game if up/down wasn't reversed and the joystick was installed at a 45ø angle
like Q*Bert. So what's this boardset's problem? Garbage screen.
Sometimes parts of it change.
If you swap the video board with the newly fixed Jackson, it's clear the CPU
board is the problem.
And the CPU sits in a socket that sits in a socket. Instead of sitting in a
socket on a daughterboard that sits in the CPU socket. Removing the socket
shows that at least one of the pins is bent. And if you install the CPU
directly in that socket, it just falls out again if you flip the board.
So I changed the socket and the game just runs. Sounds broken but does sound,
so bite me.
(the game would be fun if you switch up/down and flip the joystick 45ø)

Time Pilot (BL)
It did take me some time to find that. The first boardset had some graphics
problems, especially the pixels were doubled, so the game ran at 128*120
resolution. Looks like the first pixelation step on the SNES.
Turned out one of the daughterboards replacing the Konami customs wasn't
plugged in right. After that, everything was OK.
Now I found another identical boardset (identical up to the chip level -
every IC in the same position on the other board was from the same batch from
the same manufacturer - I didn't check them all, but I did check like 10 of
them)
This one has bad graphics errors. Fragments of sprites move erratically all
over the screen. Of course I swapped the three daughterboards with no change.
I looked at all the signals from and to the daughterboards, Scope and video
probe. Cross swapped socketed ICs. Went over all latches and transceivers.
There are four 2114 RAMs in the middle of the board, not socketed. And every
one of them has its own 4 bit data bus. And three look the same and the fourth
looks different, less patterns, smaller "pixels" (on the video probe).
Checked, there's only this one RAM on the BUS and everything before and after
is 8 bit. Could only be the ram because the chance of a transceiver having
exactly those four pins go bad is rather unlikely. So the RAM went out and the
chip tester said PASS. Errrr... whaaa? So put the RAM back in (its new socket)
and no change. Put a new RAM in and of course it works just fine now.

1942 (Capcom) (probably BL)
Interesting... it draws the title screen with some garbage and crashes.
If you turn down the voltage it gets clear the garbage is made up of sprites.
Measuring Sprite RAM /WE, it's just stuck high. And have I been chasing my
tail chasing the signal's origin. In between I stumbled upon a russian LS08
with an output stuck low even though both inputs toggle on mostly high.
Since I'm slowly running dry on LS08s (yes that happens. I have ALS08 which I
could use, but they're already running low too, I have a bunch of HC08 and S08
and you can't just use these everywhere), so there's an unused gate in this IC
that still works so I just clipped the output of the bad gate and put three
jumper wires over that IC... of course that didn't change anything except that
chip's output is now toggling again. Someday someone's gonna find that and
think WTF?
So back to Sprite RAM /WE... I chose err chased that through dozens of gates
and muxes and flipflops until I got to an LS138 next to the CPU whose pin
controlling this circuit was stuck high with garbage on it. The piggyback
strongly agrees. And the output does go low when building the title screen.
Now I'm fed up and download the schematics. And after thinking for a long time
I noticed my handdrawn Z80 pinout in my handy book of pinouts has IRQ on pin
20 even though pin 20 is IORQ and IRQ is pin 16. That baffled me, since pin 20
is never low and if you pull it low you get garbage snowing over the screen.
But pin 16 is never low as well. Chased that through 3 flipflops and it comes
from an LS367 on the edge of the board next to an interconnect.
Its input is invalid, never fully goes low. It comes from the one and only
LS112 in the middle of the board. Interestingly, /Q is stuck low and Q is
toggling, albeit at invalid levels. With a piggyback, /Q is still stuck low,
but Q is OK and the game appears to run now. So I swapped it out and the game
is fixed with no further visible issues.

Jackrabbit (Zaccaria, original)
Interesting thing. Black screen. On powerup, you briefly see the garbage
screen which is quickly cleared, then nothing happens and after a couple of
seconds you can see on the BUS that the CPU crashed. And the BUS looks bad
with several bits toggling only between Low and Floating.
And boy was this a red herring that stole much of my time. For example, there
is a PAL on the ROM board that seems to have only 2 outputs and all 16 inputs
used (or something). And all inputs are invalid signals from the video board.
There are LS245, these aren't triggered or enabled. After some time I checked
IRQ/NMI/IORQ on the Z80 and saw nothing there. If I manually trigger an NMI,
I can see a different pattern on the data bus. And I can reproduce it.
And it looks the same every time. That looks sane! Huh, now there's something
on the video output? Plug in a monitor... that's not possible. It shows Player
1 and 2 score with that special Zaccaria font and if I trigger another NMI, it
displays some kind of test "menu" (it just shows the positions of the DIP
switches).
OK so it should get NMIs, probably once per frame, but it's just constantly
high. It comes from an LS74 on the CPU board whose inputs are all extremely
dead (and most wired to +5 or GND anyway). I follow one to an LS14 next to it
which also has nothing going in. Now that I follow... everywhere. It goes to
3 or 4 chips, all inputs, so I keep following, it comes from the video board.
And was hard to follow since it kept changing between parts side and solder
side. And there's no schematics online. And I arrive at an LS74 (hooray an
output!) whose inputs aren't dead but its outputs are. If I piggyback it, I
can see them fighting, so I swapped the chip and the game came back to life.
(the IC was located at the board edge opposing the interconnects, right at the
gap in between the sound and CPU boards)
Then it wasn't easy leaving the test mode, but I managed it by flipping the
DIP switches so their dirty sides are out (nice old trick)
I put the boardset into a machine. Everything appears to work except for the
controls. Speech isn't 100% lipsync, but damn close, so bite me again.
Also the speech quality is enormous, sounds like the entire thing has been
sampled, not just phonemes. And at a relatively high bitrate.
The joystick goes to the 8255 just like the DIP switches which I all tested in
that test menu. So no clue why the controls don't work. Coin and start work
though...
I bet it's set to cocktail and if I start a two player game, the controls of
player 2 are gonna work fine. Betcha!
Nope - simply put, to move, you have to hold the joystick in a diagonal
direction AND press fire to jump. The test machine had a 4-way joystick
installed.

Le Bagnard (aka Bagman)
We got a few of these, even a bootleg, most of which work passably.
Now here I have one with a really messed up picture (game runs though).
Each char repeats twice (8V), 32V also seems bad and while I measure,
characters doubled vertically (4V). So I checked the address generator which
appears to be OK (and none of the piggybacks fight) even though one of the
LS161s has lost most of its chip markings.
The LS86 look OK, at least they all have toggling outputs. But there's two
LS283 which I already know like to cause issues like that. So I piggyback the
first and the picture changes, I get illegal logic levels, Where's my solder
sucker? Uhh I left it at the museum and I don't have a key. So... next time.
Btw. I haven't mentioned in the German version that BOTH LS283 must be bad.
Yeah because by now I fixed it.

Another Le Bagnard, correction Super Bagman
I don't remember if that's the one running on Crazy Kong hardware. Don't think
so because that one was working fine I guess.
Well whatetver - here we have a Le Bagnard with massive graphics errors.
Looks like random video noise (not CPU data). The sprites are OK though.
Sadly, there's rarely more than two sprites on screen.
OK this IS the version running on Crazy Kong hardware. With lots of hacks and
four unused ROM sockets and no idea if the two 2101s are doing anything.
After looking around for a bit I found one data bit on the background RAM
looking very different than the others. And if you pull it high, the screen
goes from CPU noisy image to just broken, but statically broken.
If you piggyback the RAM, the scope trace shows changes.
If you piggyback the BUS transceiver, nothing happens.
Since I temporarily ran out of DIL20 sockets, I remove the RAM first.
And rip out 10 traces. Uhh. And the RAM wasn't even the problem.
Without the RAM, the one signal still goes ape-shit (and there's nothing on
that BUS except the transceiver and the RAM).
So with a lot more effort (and solder) I removed the transceiver and only one
trace. By now, the people at the museum have arrived and so I can sneak in and
get another DIL20 socket (and test the chips. The RAM passes, the LS245 fails
with only the one data bit).
And with a new BUS Transceiver, the game runs fine (I even played it, it's
pretty hard)

Super Bagman
(placeholder)
This one seems to run. On closer look however you notice a subtle sprite
error: When a sprite goes behind a "background" object, it just disappears -
partially or completely.
That will be fun. For another day. And as I said - it's subtle and you might
not even notice on a casual game.
Oh and this is a converted regular original Le Bagnard. And EVERY IC is
socketed.

Golden Axe
Graphics errors. Sprites are OK, title screen is 25% OK, 25% errors (most
tiles are wrong or slightly wrong), 25% OK and 25% utter garbage (a single
tile repeating for the rest of the screen). Some levels are flawless, others
have slight glitches in the background (parallax layers), some have the floor
comprised of one tile repeated endlessly.
Sounds like a RAM problem. Let's find out which ROM is holding the video
information that doesn't get drawn properly. I do that by shorting address
lines. Then I check if the ROMs are being driven correctly. Aha. All the
address lines are toggling, all the data lines look similar. All the ROM
Chip Select lines are toggling too. That almost rules out the ROMs.
Let's find the correct RAM by shorting their pins. There's two 8k SRAMs and
two 2k SRAMs (narrow version of the 6116). Those I suspected first.
I could even put a piggyback on and have it improve the screen, but not by
much and not reproducable. Then I put a piggyback on the left 8k chip which
removed the errors in some levels. Then I managed to reproduce the behavior
and even the title screen was OK. So I changed that 8k SRAM and that fixed the
game.

Tiger Heli
This looks massively better than the NES version, even with the massive sprite
errors this board has. Insane, there's two LS669s next to the RAMs, one of
which is hanging onto the board with only two pins. Looks like something heavy
must've hit it. So I removed the remains, put a socket in, no difference, just
fragments of sprites flickering erratically across the screen.
If you remove the IC, the sprites disappear completely.
Let's see... the first RAM next to the smashed 669 is socketed, so swap it.
No real difference. One of the address lines has a crappy looking signal on it.
How does the board look from below? Similar damage - something must've hurt it
big time. Interestingly only two traces were severed. Patched them, game runs.
Even with the original LS669 which now is an SMD version of itself.

Two more Tiger Helis
One of which works. The other has some slight sprite glitches (graphics, not
position). Boy did that thing take a while. By swapping boards with the
working one it could be established that the fault lays in the board with the
edge connector. Swapping ROMs didn't get me anywhere. Scoping showed that one
of the Turbo Variant 2114 has a data bit that next to never goes high and that
was a dead end I spent like 2 hours on...
Piggybacking did nothing. Scoping around some more showed an invalid signal on
an LS273 near the color DAC. The pin has connection to one of the two 373s
next to it but not the other. Looks like a Via has split from its trace.
(I think I have to mention the board has some liquid damage and it doesn't
look like water - at least not clean water - other boards are covered in what
looks like bird droppings). And what did fixing that do? Nothing. I suppose
the colors are more correct, no idea, I have only two of the RGB pins
connected, and those go to a NOS Black&White monitor built in 1987.
And now the other Tiger Heli died. It crashes, appears to watchdog (because it
does it in a loop, but Reset is never seen low). CSYNC is garbage. The 74S04
next to the quartz has rather weak signals but that's expected at these
frequencies. There's a 74S04 nearby whose outputs also look as bad, but on the
working board it looks the same. But then there's an LS163 which looks really
bad, two of its output pins should be high all the time,  but these are the
only ones which have a valid signal on them. The frequencies of the outputs
appear the same, so this chip doesn't appear to count. So I replaced it and
that made no difference at all, not even to the invalidness of the signals.
OK put it back and put a piggyback on the S04. I only have 7404s, HC04s and
LS04s. I try the HC04 first. And the machine works. If I remove the chip, it
stays working, until you switch it off for more than 30 seconds.
I can reproduce the fault / fix by piggybacking the S04, so that's bad, out
with it, in with the HC04. Wow these are nice strong signals... some are so
strong they distort... The game now doesn't crash anymore, but Sync is still
way off (it looks like 30fps instead of the ~120 it does with the bad IC).
With the 04 and LS04 it looks similar, it appears to have to be an S04
(or deglitching cap). Since I don't have one, I just put the bad S04 back in
and solder the HC04 on top since that seems to work fine...
So, back to the board with the (almost) subtle glitches...
An advantage of having a monochrome monitor with separate Sync input is that
you can use it as a video probe too.
Spoiler: What looks like bad RAM turned out to be bad RAM.
With the video probe I see garbage flickering on top of the enable signal of
the two LS244@P7 and P8. The working board has a clean signal there.
It comes from LS32@P1-LS08@T1-_>LS10@R1, only one of whose inputs has garbage,
that comes from LS00@M8/9->LS02@7, two outputs have garbage, all 4 inputs have
garbage, these come from an LS08@L7 that has garbage on all 4 inputs too
(one input of all 4 gates is common to a single signal that's clean), the 4
garbage inputs is fed in by all of the four LS166 strewn around the ROMs.
So it comes from the ROMs but these are good. Looking for garbage I find that
pin 4 (A6) has the garbage signal on it. That's connected to Pin 12 (DB 1)of
the 8K SRAM on J2/3/4. Some of the signals going in this RAM also have the
garbage signature on them, but they look almost the same on the working board!
Theres an LS374@J1 involved but that could be ruled out. So I removed the RAM
and - phew - that fixed the problem. Only took like 6 hours.

Typhoon (Konami)
Woo a modern board with scaler/rotator (part of SNES Mode 7).
Sadly it has lines through some of the sprites or character layers. I guess
they used character layers since that's easier to handle than ginormous
sprites.
There's one small part of the board that has one big scratch where traces used
to be. The rest of the board looks pristine.
I'm not writing about all the boards I encountered that had shorted pins or
cut traces...
Only... fixing the traces didn't quite fix the lines. Although bending the
board gets rid of them. Couldn't find it yet though. These traces are tiny.

Hyper Olympic (this is more about bad traces/shorted pins)
This is a recurring problem with all Hyper Olympics and/or Sports.
You get a screen full of white letters on startup? (often Z)
Remove the battery. These games have a battery, if it's deep discharged, it
will keep the battery RAM from getting powered. Without that RAM working, the
machine just crashes.
But even without the battery, it stays at that screen. So I take a look at the
bottom of the board and see lots of bent pins touching themselves or shorting
into traces and after shortening my fingernails by 2mm, the board ran fine.
Then there were 2 more boards where I had to bend pins back and they still
wouldn't run. One had a scratched trace to the Program ROMs (wandering garbage
screen), the other had a bad 2114 RAM, but instead of outputting an error
message (which would have been easy to implement and still decypherable), it
stayed on the screen full of white Ns.
Both have slightly glitchy sprites. If you flex the board of the one with the
bad trace, the glitches can be increased, decreased or removed fully.
Amd then there was a Wonder Boy I marked "Sprite errors, prob. video RAM" -
IIRC most sprites were invisible. I just wanted to check what kind of video
RAM was in there and beep if the bodge made contact, but I didn't get this far
because I saw two of the EPROMs had half their legs next to the socket instead
of in. Now there's no more sprite errors.

Hyper Olympic
OK here's two whose repair wasn't so easy.
Number 1 reports a white screen where there are some weak switching artifacts
that tell you the game is running.
And it's an Original Konami!
And what do I get? Only every second scan line is white, the others are OK
apart from missing sprites.
That was a major PITA, I had to look everywhere. Piggybacked ICs to no avail,
started trying stuff blindly, no idea.
I found two bad LS163s on A17 and A18, piggybacking them changed nothing even
though they now have output.
After a while I noted an LS139 in the middle of the board, if you short
circuit pins 2 and 3, the picture returns (without sprites).
I traced that back to an LS74, if you ground its data input the picture
appears completely (xept sprites), the data comes from the LS74 next to it and
it's looped back and if you short anything there not the entire picture comes
back so no clue.
After an even much longer time I found an LS21@F19 where if you short pin 12
to 5V, not only does the picture return, but also fragments of sprites, even
in the right places!
That comes from an LS375 which gets input to all 4 gates, the first one has no
output though. I don't have these so I have to order them.
Even more striking is this chip's made by Hitachi, they usually never go bad.
Seems to be identical to the LS75, just different pinout, maybe I hack
something.
Nah I wait until the spares arrive. And now the picture is perfect, only the
sprites are missing every second scanline. At least it's playable.
And then I found the two LS163 marked as bad, they're not doing anything even
though the inputs toggle happily. Piggybacks fight, swapped, sprites return to
normal.

Hyper Olympic
Number 2 was labeled "Video address generator faulty".
What I get from the video address generator however, is dead silence.
At least the quartz is oscillating and the LS04 next to it tries to make a
nice signal out of that. It goes into an LS161, wait, it's just a 74161 not
even LS. But Hitachi. But there's nothing coming out even though the chip is
wired in the cheapest possible way (just count, no load, no reset).
There's a deglitching cap preventing a piggyback from sticking, so I have to
pull that IC. And indeed, it's bad. This is the second broken Hitachi in one
day.
And... still dead. The CPU gets a clock now, but not on the pin it should.
So... there's another 74161 next to it and it's just as dead as the first one.
Piggyback fights, swapped it, now I get a screen full of white Ns and nothing
else happenes.
Remembering the pre-previous Hyper Olympic which was an identical bootleg,
where it took me quite a while to found the cause was a bad RAM.
So I take a look with the scope at the two RAMs in exactly that position.
One has no outputs. The other has outputs with a few signals on them and
floating in between. Since they're shorted to be always enabled, they can't
be floating. A piggyback gets the machine to boot, but crash immediately.
Swapped the RAM and it runs. Well the sprites look like thunderstorm, but
apart from that the machine is running. What happens if I put the deglitching
cap back? Oh hey the sprites look great now, so we're done.

Another Hyper Olympics belonging to Wolfsoft
Wow this board is filthy! It looks like it's been stored in a storage hangar
that's also used to store birds if you know what I mean.
For some reason some places are more corroded than others and Spoiler Alert
some totally corroded chips work better than some almost clean ones.
Let's start first by replacing the single wipe sockets which like corrosive
environments even less than their dual wipe counterparts. I let some non
corroded sockets in, they measured fine, also I ran out of 28 pin sockets.
The chips I pulled out of the sockets however, couldn't decide whether their
pins should be brown, green black or non existent. And some rotted away right
at the plastic (or ceramic).
So what have we got? D5 program ROM has a few rotten pins, some fell of, but
luckily just the pointy end. Soldered new pins to the chip, GALEP says the ROM
is empty. Beeped every pin, the Testofon says there's semiconductors at the
other end of each pin except pin 26 which on a 2764 is at 5V anyways.
Chip reads blank nevertheless. Just for shits&giggles I hit Program but of
course that didn't work, so I burned a new one.
Funny that was the first chip I tested and also the only bad one.
D3 has a total of 20 "bad" bytes, these are in a coherent spot and make sense
(not just a bit flipped or something) and the sticker on the ROM says J3
instead of D3, so that makes it clear, this is just a different revision of
the game, so I let that pass.
One of the graphics ROMs has the +5V pin broken off right at the ceramic edge,
no way to fix that, but if you hold a piece of wire to it and read it out -
Surprise, it still works!
The 6116 RAM between D5 (sticker, not board position, that would be A3)
and the battery does look like it was salvaged from the Titanic and had a pin
broke off right on the plastic. Konami for some reason scratched the label off
the chip so no one could get the idea that a mystery 24 pin chip that's hooked
1:1 to the address- and data BUS and gets its power from a battery in a game
that's known to save high scores, could be a RAM. Ooohh nooooo....
Long story short: I don't know if it still works and I don't wanna know.
Now for the Bad News Everyone:
From the four customs (two more are in the form of daughterboards even though
this board is NOT a bootleg!), only one and the CPU work (well OK maybe the
one in the corner, either row A or column 1, still works, I haven't fixed it
yet). One has a pin broke off right at the plastic which will require
dremeling it open so I have something to solder to. The other custom (address
generator) is in much better shape, only 3 pins broke off and the stubs are
still there.
Unfortunately that one's dead as a doorknob. It's the address generator custom
which gets a clock input and all the other pins seem to be outputs. And they
all float. Checked with the pointy tip of the scope probe that ground, 5V and
clock make good contact. That chip's gotta be dead.
Too bad.
OK nothing we can do about it, put the working one in there and continue
testing the CPU portion. It's watchdogging and the screen doesn't look much
better, gray background with some garbage sprites, nothing much happening.
Some of the address signals look pretty bad, one looks normal except it has 1V
as low. Checked which chip it comes from, that's an LS244 (Fujitsu), it's
always enabled but doesn't pass much to the other side. Piggybacked it,
signals look perfect now, but except there's a lot more flickering going on
with the garbage sprites, that didn't do much.
Checked what else looks broken. There's another 244 which gets CSYNC and
outputs CSYNC, but that's the only clean signal. It also outputs what looks
like VSYNC but only goes "low" to 4V, but that output has no corresponding
input. If I piggyback it, CSYNC becomes the only signal on this chip. Well.
Then we have an LS04 next to the battery, it has a lovely input on pin 3, pin
4 is floating. And pins 13 and 12 look ugly. And on the other side of the
board there are three LS157 which are all completely dead, all of them Fujitsu.
(although there are still plenty of perfectly fine working Fujitsus on the
board).
And now? Weeell it produces a yellow (I think) screen with garbage sprites,
but now you can recognize these as from the game (not just flickering pixels
of garbage anymore). And the RAM BUS for the backgrounds is completely
floating.
I let it be, because two bad customs and a third custom was needed in our
Gyruss bootleg.

Silk Worm(? Tecmo, couldn't quite make out the title screen, it was only there
for a split second and I read SILKWOOD)
Completely dead. Just from looking at it, half a dozen traces are cut.
Patched it, no clue if it's worth it, it's a bootleg comprised of almost
exclusively GoldStar ICs (these are terrible and have a habit of only working
correctly at *exactly* the right voltage, none of that 4.75-5.25 crap, more
like 5.13V precisely)
Juiced it, nothing. Doesn't even interfere with the radio. No signals, all the
BUSes are dead. Quartz is running but doesn't go far. There's a HCT04 and
another HCT04 after that and this one appears dead, no outputs.
Piggybacked it and the radio died. Now the board looks a lot more alive, we
got CSYNC but no picture.
So I removed the bad HCT04 (hooray thought of my solder sucker) and put a new
one in (yes I do stock HCT04s). So no picture and the CPU runs loops.
Let's do the finger test. And there's a Toshiba RAM? Mask ROM? which gets
hotter than it should get. None of the others even gets warm. When I remove it,
the machine jump starts. And as I hinted, the title screen graphics are
missing and in game there's only a few single pixels as sprites so I suspect
the sprite ROM is what I'm holding here in my hand.
And yes, I can't even read it out. All I get is zeroes.
I burned a new one and compared the other ROMs. The ROM next to the bad ROM
gives me a faulty byte every 256 locations compared to the original and on top
of the address range there's a bit more amiss, but that might be the bootleg.
Then there's a few 32k EPROMs where 64k should be in there. They do have the
right data, just only the first 32k...
And with the newly burnt EPROM... it still doesn't run.
You see lots of garbage, sometimes a really broken screen test.
If I flex the boards, it improves. After removing and putting back the
interconnects, the game runs fine except for a few stray pixels in the
background. Pretty subtle. You can leave it at that. Now you can also make out
the title screen to be Silk Worm.

Do! Run Run
Am I even going to put that down? Was too trivial.
We got two (edit: three) of those and one is running, the other only has
trace amounts of sprites. Dat thing has 128k of video RAM (at a time when 4k
was deemed sufficient by most other games).
There's a custom DIL 40 jobbie in the middle of the board. One of the pins has
an invalid signal. The other board has a nice TTL signal on there that's
mostly high with occasional lows.
Beeped it with the Testofon and it goes into IORQ on one of the Z80s.
Or more it gets lost just before it because the trace has a scratch through.
Patched it, now all the sprites are there, but they have jailbars.
The title screen looks even worse. Just for fun I put the video probe on the
output of each of the 4164 video RAMs and all their outputs have a part of the
video signal (a color layer). Except one which only has random garbage.
The input looks the same as the others though and the address signals are
identical, so that has to be the bad guy.
And it was. After swapping it, the output resembled the game video.

Pento (Pengo bootleg)
Now it gets interesting... there's an LS258 in the middle of the board that
actually caught fire - at least from the looks of the chip itself and the
scorch marks on the board. Several traces became incandescent before burning
through. Patched them, replaced the charred chip and tried to power it up.
My PSU limits to 2V. An LS367 next to the CPU gets very hot very fast.
And I forgot my solder sucker again!
Took the board to the museum because desoldering station.
Boy that thing is dead. Well at least the Quartz is running and so is the
horizontal counter after removing the hot LS367 and the LS367 next to that on
accident. I didn't want to remove that one, but it turned out to be bad too
anyway (just not AS bad).
And it's still doing nothing. All I can see on the BUS are stuck bits and
illegal logic levels and only rarely something toggling.
Btw. I tested all the RAM and ROM on the board, even the LS 289 and they're
all OK. And the ROMs match MAME except part of the graphics ROM.
Three of the traces I patched show a short to ground. So let's short them to
5V and it's a huge short circuit. Only the LS257 below the burnt up LS258
gets a little warm. So I check out the chips next to the blown up one with the
chip tester and mark all which are clearly bad. Shortly before that I stumbled
upon a defective LS74 by accident, and when I piggybacked the first vertical
counter, it didn't do anything, but when I bent up Q1, it counted.
So after all I removed the LS74, 161, the other 258 and two 257s and now the
board does a whole lot more. Lights are on, but still nobody home.
There's three signals on the edge connector that almost kinda sorta look like
video, but no sync and no blanking.
One of the patched traces (my guess is that everything that was connected to
these must've gotten a hell of a current, so it's gotta be fried) is still
stuck low, and with 5V applied to it, an LS157 in the same row two columns
further than the exploded one gets really hot. And now things look better.
Still nothing on the second vertical counter, but Count Enable is stuck low
and the LS74 I replaced also has one output stuck low and it toggles if you
bend it out. Better than the Galaga bootleg where it seems to be normal that
the inverted output is NOT equal to (NOT normal output).
So I put 5V on that pin and whatever it was has burned through, but at least
it got warm enough to identify it as an LS174 next to the graphics ROM.
Swapped it and now we have composite sync and a clean video signal with
blanking and finally the CPU watchdogs.
Interestingly, if you touch A10, there's a huge hum on it (really floating).
That doesn't really change with a different CPU. And with yet another CPU it
does something, even stopped watchdogging for a while, it even clears the
garbage screen and shows something, but it's not the game or an error message.
Putting the original CPU into another machine reveals it's toasted.
Put a new CPU in and went hunting for illegal signals and broken ICs.
And there's an LS377 with a high level of 2V on some inputs and nothing on
some outputs. Piggybacking that not only brings back the outputs, they even
stay working after I removed the piggyback. But only until the past next
restart (or until it is switched off for more than two seconds).
Definitely bad. Then there's bad signals on the LS289s (or whatever), these
originate from the last original LS367 and if I pull that signal to 5V, it
gets hot again...
So swapped these two chips (367 and 377) and still broken. At least the
signals look healthy now, so the hunt goes on a step at a time.
(I'm taking life 8 bits at a time)
And finally I find an LS138 with some very suspicious looking outputs.
They look less suspicious with a piggyback on, but still don't toggle
(dirty high). But one of the inputs has an illegal level. But instead of
pulling it to 5V (high was at «V), I chased it across the board to an LS04 and
pulled the signal to 5V THERE and guess what got hot? The LS138.
Swapped it and... unbelievable... it erases the garbage sreen.
Power on Reset doesn't work, but after ¬ second the watchdog kicks it and it
clears the screen! And VSync comes as NMI! And it stopped watchdogging.
But it doesn't do anything more.
So more illegal signal searching... and there's a bank of four LS253 which mux
the inputs (joysticks, DIP switches...) and the enable signal looks floaty.
If I pull it to 5V, two of these get warm (remember it's an INPUT), so they
all must go. Unfortunately I ran out of LS253 but I have two 153s which are
compatible and two of the original 253s seem OK.
And with these replaced, the machine runs...
The sprites are glitchy and sometimes hang at the left (=bottom) border of the
screen and no matter if they get rendered at the correct position or not, they
have sparklies (looks a little like the TV static noise circuit that GORF and
Wizard of Wor use, just unintentional).
And... according to the title screen the game's name is PENTO, but when the
bees start painting the big logo, it still says PENGO. And I thought I was
lazy.
If anything then PINTO. If you touch an enemy, you explode instead of getting
f#*ked (sorry look at the animation the snow bees make when your Pengo is
lying on its back and tell me what else it should look like).
OK when testing the machine a week after that, the sprite position problems
got a lot worse. The legs on the LS289 RAMs are horribly corroded. In, out,
all the sprite position problems are gone. Even on the title screen.
Only the sparklies are still there, but rather subtle. Certainly playable.

Tapper
I didn't know it was defective. The picture looked like a worn out CRT.
White things looked yellowy and black is dark blue and you couldn't really
control the blue gun.
Turned out, the 9 bit * 64 SRAM directly before the video output had a bit
stuck low (which gets inverted by a transistor logic). And curiously we even
have this type IC in stock, so I could just replace it.
I put the chip back into the spares (with the bad output bent up), because a
chip where one bit doesn't work is a lot better than a completely dead one
(and they are hard to get).
Now only the monitor has to run reliably - I changed so many caps, once I even
had to change the HOT since it somehow got fried, but now it doesn't work
again even though it did last week. It does that from time to time.
So the HOT is OK. But I notice that B+ is almost all the way up.
Turning it down didn't fix it on open saturday though.
So now I replaced a really disgusting looking ceramic cap (dripping... wax!?)
(0.047æ200V) and for whatever reason it works now.

Pac Land
I mistook it for a Rolling Thunder because of the style of PCB.
The PCBs do look similar, that stops if you put them next to each other, but
in a sea of bootlegs, an original Namco post Pac Man does stand out.
This one gives graphical garbage that only syncs up horizontally.
It looks like the screen in Yar's Revenge on Atari 2600 when you destroy that
face thingy. Looks very broken.
After finding a working Pac Land, I thought let's swap chips.
After swapping two customs to no avail I thought of something better and took
a look at the underside of the board. A few bent pins. That didn't fix it.
But there's a trace scythed through. And now the video output looks like
graphics - it starts the self test and then shows 2 0 0. Whereas the working
machine shows 0 0 0. Gotta be an error code.
And it means bad RAM (9N?). According to the Net there's another chip that can
be made reponsible, simple 7400 (5B?), I checked it and it was good. So I
replaced the RAM. No differences. The RAM passes in the chiptester which came
to no surprise as that thing passes a lot of RAM that has subtle problems...
So let's take a look at the signals going to that RAM. One of the /OE and two
of the address lines look bad. Low with low frequency noise (looks like
surface noise from a record).
Went on a rather short safari with my Testofon and found an LS32 at 10L which
had this garbage on all four of its outputs even though the inputs are happily
toggling. This is an OR, so whenever an input is high, the output must be high
too. Replaced the chip and the game runs.

Skylancer
I originally passed that one by because it came in as a hardware donation
that's supposed to be working, but the last test was 2010, 11 years ago so I
thought why not.
I should have ignored it. No I shouldn't, because it took me just half an hour
for the entire fix including chip replacement.
So the sprites are in the wrong place. On closer look they avoid four equally
wide stripes, reminds me on Galaxians when one of the video address signals
was dead, only here it's only on sprites.
So first I shorted everything to find out where the sprites are.
Then I got distracted from a couple LS283 which I know like to fail.
One of them had a high signal with a bit of noise on top as an input.
When I grounded it, the sprites snapped into the formerly unused positions and
proceeded to avoid the stripes they were usually in. Smoking gun!
I followed the signal to a RAM. It's on a BUS and it's not alone.
And that comes from an LS245. The inputs all look the same, but one output is
dead. I put a piggyback on and it looks a lot better now, not quite as
expected, but the sprites now use the entire screen.
Replaced the transceiver and after a reboot, everything looked allright.

Piranha
Black screeen. If you look closely you can see that there's a garbage screen
preceding the black screen. So at least it executes some code.
Trying to fix it a few months later, the initial test gives you a sort of grid.
Well it's full of zeroes and they're fading in and out, looks like a bad trace
or joint. A few solder joints were indeed bad and the game has a daughterboard
with those trustworthy looking (because totally corroded) wire-wrap sockets.
Couldn't find a fault there, though.
After dumping the ROMs and found nothing wrong I started looking at signals
and landed on the 6 2114s where 3 looked fine and 3 looked weird.
After making sure there's nothing else on the data BUS but the RAM and the
transceiver (which is OK) I piggybacked the first 2114. And lo and behold, the
game boots. Looks crappy with many glitchy characters, but hey it runs.
If you piggyback the second RAM, nothing happens. If you piggyback the third,
it gets crappy again but if you reset the game, everything is awesome.
So let's pull these two RAMs. One passes, the other one doesn't.
All 6 were made by Hitachi, but the one that was definitely bad was from a
different batch than the other 5.
The one 2114 which was passing the test but fighting its piggyback works just
fine and now sits in a Tazmi King Derby cocktail table whose boards have been
washed (otherwise it would've cost a fortune in cigarettes) and the original
chip turned out to be not quite waterproof.

Toki (identifies as Juju Denetsu) (still a bootleg, what else?)
Very broken if it does anything at all. The backgrounds work partially.
Later I looked on Youtube and see two of the background layers work (I think
there were four). The sprites consist of pure triple-distilled garbage.
But since the ROMs all do match and it looks like RAM, I gonna swap the RAM
first. But which one? By shorting lines I could make out the RAM in the middle
of the bottom board to be responsible for sprites (ey dat ding got 2 video
boards one of which also happens to contain the CPU... and FIVE quasi
independent pixel pusher pipelines with four LS194 each)
It took me a while to even make out the name of the game - first I misdumped
the ROMs as 27C080s even though they were 010s, so now at least I know what
game it is - the title screen is on one of the non-working layers...
Back to the sprites - pulled the RAM, ripped two traces, fixed those and it
got me... not much really. You can at least make out what the sprites are
supposed to show now. But the sprites are repeating blocks (every block
appears 16 pixels further down again if the sprite does have a block there)
and every second scan line is missing.
After a long search during which I managed to melt two traces (PC power
supply - you want 20 Amps?) and got two RAMs to self-desoldering temperatures
(one of which works fine, the other is a little glitchy), I noticed the RAMs
next to it were the real bad guys. So that restored every second scan line of
sprites to perfect working order. Every other line is missing.
And that was hard as hell. The crazy thing was I beeped everything and after
soldering in the socket I beeped everything again. But now that a chip is in
the socket, the fragile rest of the trace cleared and made no contact to the
rest of the chips in the chain or the signal source. After that I had my
broken barely recognizable sprites back.
Temporarily I checked all the Not-PALs, Not-RAMs and Not-ROMs on the board
with the chip tester which flagged almost all the LS174 as totally bad as well
as the LS175 and the two 173. I couldn't find anything wrong with them during
gameplay, the scope says all's well and piggybacks make no difference.
And after some time I flexed the board and see the game has crashed.
But the backgrounds were back and the sprites were complete.
After a while of jiggling everything it appears to be a combination of
hairline cracks, poor interconnects and chip sockets.
Finally I got the game working perfectly with all the backgrounds and sprites.
Damn is it hard - and it's really unfair. You gotta memorize every single jump
because: You have to jump but there's an enemy coming you can only see when
you can't avoid him anymore. So next time you're there, you have to shoot in
exactly the right direction (multiple times!) to kill the enemy.
Also I didn't have sound, but I noticed if I measure the amp in a certain way,
sound comes back. Then I noticed I was shorting two pins.
Soldered in a permanent short circuit and sound works now.

Euro League (the label says "Viesel Soccer")
After a newly arrived unknown board with no label turned out to be a working
duplicate of this broken soccer board, I finally dared trying to fix this.
Which I did without really refering back to the working one.
What was striking however is that all the Toshiba 2K SRAMs get cooking hot.
And the board produces a garbage screen, but only on one color. So first I
took apart the "DAC"s (4 resistors per channel) (I managed to put them back
together). But there's nothing going in (there's traces running in between the
legs where the solder mask is missing and I was measuring those).
There's nothing going into the latch chip. Which comes from an LS245. Which is
never enabled and looks the other way (direction pin).
Taking a second look, it also goes to one of these Toshiba SRAMs. The one
channel that's working goes to what looks like a 2114 clone. Which makes it
clear. If you measure the board cold, you can see the two other color channels
briefly have outputs which go to nothing after a few seconds.
So I pulled all four of the Toshiba SRAMs and one was OK. The others are so
bad they even short address lines and get really hot even in the tester.
And with new chips I still get a garbage screen (with a broken Techno sound
loop), but at least it's in color now. If I turn it off and on again, I get a
different garbage screen after a second. And if I turn it off and on again,
the machine boots. And I got vertical jailbars. If you bend the board in the
area of the graphics ROMs, the jailbars disappear and the game runs fine.
Sound also works even though the machine only plays 2 different samples in
attract. But also music.
So if I turn it off and on again I'm back to garbage. If you apply Reset
manually, it works. So the power on reset circuit must be bad.
After looking for it I see a missing capacitor in the middle of the board.
Now I do need the other board for comparison. OK with a value of 10æF and a
known polarity I can put a new one in and now it works fine.

Street Fighter 2 (Bootleg? You bet!)
Weird graphics errors, as if a counter wasn't working right.
As soon as the background scrolls past a certain position, the textures change
into the holodeck grid. Sprites glitch, the title glitches and after fixing it
I see the copyright Wall Of Text you get when turning the machine on is
missing, instead it stays on a black screen for half a minute.
The self test passes...
The fat custom IC is marked with a big X and I didn't do it!
Anyway I think I'll try my luck.
Beeping traces->Nope. Bent pins->Nope. Piggybacking RAM->Nope.
What about ROMs? One is missing the sticker (and after noticing lots of boards
with no stickers on the ROM having been semi-erased by the sun I know they're
suspicious) amd some have scratched stickers, so let's dump'em.
The graphics ROMs match. Of the 3 CPU-ROMs only one matches... wait a minute,
this is a 16 bit CPU, shouldn't that have an even number of ROMs?
OK the two that don't match have "errors" in exactly the same addresses and
these can't be explained with erased bits. And most of the code is identical.
So it's an undumped bootleg (probably with a slightly different copyright
message or something). Sincee the ROM accompanying the missing ROM matched, I
burn the missing ROM->Nope. The "You lose" message is going on my nerves!
OK so the faults look like counters. Piggybacking the LS163->Nope, not even
fighting. What's the next chip in line? An LS157? The outputs look clean
though. And suddenly the picture goes SNAP! and everything looks perfect now.
And it's running ever since. A sound amp gets really hot, there's a capacitor
missing (electrically, the other leg is still soldered, but it leaked).
With a new cap the sound now works a lot better and the amp merely gets very
warm. And even after turning it off for a few hours it still works.
No fault found, nothing fixed (just burned a missing ROM and it worked before
without it), it just Christined itself.
And yes - I flexed that board all the way and nothing happens.
And even after a few weeks of it being back in storage, it still works.
Just - playing with a Competition Pro isn't much fun even though the stick was
modded for 2 buttons, but that's still 4 too few.

Aero Fighters
There's a label dangling from it "bought as defective". And it's completely
dead.
On the solder side a few pins are touching but that can't be the reason it's
so incredibly dead. An Eyeball Inspection shows that in a corner of the board
there's a diode that's split in half and a capacitor is missing. A missing cap
usually doesn't do much, neither does a broken diode (unless you want to play)
but beeping shows they're connected. And they look like the Power On Reset
circuit. Since there was only 2mm of the capacitor's leg left I checked
another board for a typical value of cap and decided on 10æF. I used a bog
standard 1N4148 for a diode and when I turned it on, I could see the video
signal changing. So I hooked up the TV and it doesn't sync. But it's JAMMA and
there's a JAMMA machine next to it and... it works perfectly fine.
What a missing RESET signal can do...

Goindol
It's got a lovely picture. Just the attract screen drags on for an eternity
without much happening and the game looks like Arkanoid with no ball.
Now that board has a big epoxy block so I thought it's gotta be a suicide CPU
killing the sprites.
So I thought, let's use the video probe to see if I can get it at least
playable.
And I did! There's a 6116 that has all the sprite data in a usable form.
Where's that going to? A 244 and a 374 the latter of which only outputs high.
Clock is there and Enable is shorted to ground.
Piggyback does nothing, bending the pins out shows signal so that chip has to
go. And now I only get a gray screen.
After a while I see the short between Enable and ground is gone so I restored
it and now everything is back. Including bonus items and a weird caveman
running around on top of the bricks as well as the title screen.

Future Flash
This one says it's got an evil short circuit from -5V to +5V but only when
it's running. But I couldn't reproduce it, all the voltages are there.
Only there's no video signal, not even sync, even though the machine has sync
littered throughout the board (I was measuring a weak sync pulse on some of
the output transistors). And turning it off and on a few times I could make
out three pins with a video signal (frozen) but none with sync.
The board only has sync in inverted form (but there are chips with unused
gates).
To fix that, I need schematics.
So let's check first why it doesn't run.
The CPU is stuck in IRQ - IRQ is stuck low.
That comes from an LS08 from two places, one comes from the video board and
there it's from an LS74 whose inputs look legal but don't make it toggle
because a signal appears to be missing (and then it's just a blip).
That calls for a complete test-over with the chip tester because I do want to
make a video of that thing, so why now with this board?
Weeell... checked the schematics and now I'm kinda sure it's gotta be a ROM
issue (since the Z80 has to acknowledge the IRQ for it to go back high),
unfortunately my GALEP Laptop (which I also use for VC4000 development) has
decided to die today.
So I dumped them a week later. What's noticeable is, they're 4k ROMs, but A11
is shorted to 5V (cue Twilight Zone music). And the ROMs are... flaky.
Most read their lower 2k reliably but the upper 2k which is the only thing the
machine can see is flaky, one has it the other way round - the upper 2k are
solid and the bottom 2k flaky. And then there's a ROM that doesn't match the
others (first of all it's a mask ROM, the others are EPROMs, then it only
contains graphics data and doesn't match anything).
So - there's a switch in the schematics that allows to connect A11 to GND, VCC
or A11 - but -5V is also running along there...
And the ROMs all read as if they're a bit deaf on A11 - or ignore it
completely. So my guess is that's what the -5V went to.
Err, no. In Texas, things are handled a little differently. These are TMS2532
which have A11 where others have /CS. These only have a single /CS (/OE) and
pin 21 is VPP which should be +5V when operating normally, so everything is
wired up correctly. And when I tell my GALEP that these are TMS2532, I can
even read them back error free. Which brings us back to square one, the
machine doesn't run and no clue why.
At least it initializes the CRTC and clears the screen, but the endless loops
it lands in aren't the same every time, sometimes it even acknowledges
interrupts.
Sometimes I even get an image. Sometimes I even see patterns getting
overwritten with other patterns. But I can see reliably it's only ever writing
to half of its video RAM. But that in an endless loop. Maybe one of the RAMs
has a similarly subtle fault as the DECO.
Changing sockets and freeze spray didn't do nuffn... at least after swapping
the CPU sockets I got CREDIT 01 and PRESS 1 PLAYER START BUTTON and the
typical high score stuff as well as a little display of colorful aliens under
the start button message that looked like a score advance table, but that was
most the intelligence I got out of it. But in short: Yes it can access all its
video RAM. Also interesting, when there's a picture, there's always 8 pixels
blue, 8 pixels black or video data (like text).
After shoving the board ahead of me for a while (strategic procrastination),
on one open Saturday I removed ALL the video RAM and replaced them with known
good which made absolutely no difference whatsoever (except that I can now
test them in Qix - which btw turned out to be broken again, I'm talking about
the spare Qix boardset not the one in the machine which still works).
So I removed the RAM from the CPU board (which passed in circuit without the
quartz) as well as the 8253s. These passed, but one of the RAMs now fails.
And with fresh RAM installed, the machine (almost) always boots to a solid
khaki screen and just sits there doing nothing. None of the switches
(including TEST) do anything. Shorting the palette RAM's inputs show static
garbage. What's cool is that the machine always boots to the solid khaki
screen. If you crash it by shorting stuff (like the pallette RAM's inputs) and
hit RESET, the machine always goes to the khaki screen after about half a
second, so it does other stuff before.
So the only untested chips are a few analog ones in the sound section.
All the chips are tested, all the RAM and ROM are working and have known good
data, and the machine still isn't working, or at least not booting. I at least
expect self test to work!

Bosconian
We have two - one that works and one that has missing interconnects and
doesn't even with ones. So I checked if Reichelt had stuff to make new ones.
And yes they did, so I made one for this machine to see if I can get it
working.
Conveniently I can swap top and bottom board with the known working set, so
let's start with the known working CPU board.
So - it's running but sync looks like shit and the video signal looks like
there's an entire layer stuck to opaque, but the game appears to run.
And sync? There's bodge wires around so the video signals can go to the edge
connector and if you remove sync, it looks fine. And if you put it back, it
still looks fine. ???
Concerning the video signals, it's just a missing input. That comes from a
custom BOS1-8 or something, it's a ceramic 20 pin DIP next to a pallette ROM
and if you swap it out with the other one the picture is normal.
Input pin 1 is low impedance. With an 18 Ohms resistor I can almost pull the
signal somewhere usable. The chip has an internal short. I tried burning the
short out, but that only made it worse, so that's a death sentence. Too bad.
It almost works. Well I can try duplicating the working one...
And the CPU with a known working video board? Throws an error message, but my
TV doesn't wanna sync it and I can't read it.
Stupid question - why didn't I use the Konami adapter from the other board?
And with that I see monochrome flickering garbage instead of yellow/red
flickering garbage, followed by the message RAM 1L flashing for a quarter
second. There is only one RAM on this board... swapped it, runs.

Double Dragon
This is part of a joblot of three crates from Wolfsoft (the Vectrex guy).
After setting the boards aside that lay on top of it (all dead bootlegs with
way too many PALs, soldered in ROMs and daughterboards as well as a Konami
Hyper Olympics that spent the last two decades in a puddle of acid), many of
the boards appeared to work - there was a Viesel Soccer where I could see on
the scope that there are only sprites being rendered, no backgrounds. There
are two graphics ROMs missing, that could explain it.
Err... back to Double Dragon. This one doesn't work. No video signal and no
V-Sync component on the C-Sync. Looked at the LS161s on the video board and
one got high with some bit of droop (.2V less than high) on it.
Traced it to an LS00 (Pin 3). It gets floaty noise on pin 1 and outputs 2,3
and 4 are illegal. Swapped the chip out and now C-Sync is back (I wanted to
write "looks clean", but there's some crosstalk on the high level, so it
doesn't look clean at all, but well it's a bootleg, what can you do) and it
has a nice video signal with lots of moving objects.

Pitfall 2
Here a PAL is missing. The internet doesn't have it. But that joblot of boards
had a Wonder Boy which looked identical to our Pitfall 2. So I nicked the PAL
and tested it and... dead. (at least CSYNC now exists and looks good)
And at least I could test the daughterboard which was almost as big as the
mainboard, it works fine. So maybe they're incompatible, let's take the PAL
from a completely different looking Choplifter and test its PAL. It has
graphics glitches, but it runs - compared to Dead. OK we don't have video and
the data BUS looks bad, really noisy and no real Low anywhere.
Swapping the CPU (Crypto with PAL) and the ROMs didn't bring it back to life,
but that wasn't unexpected as I had dumped the ROMs previously and they all
match. So why do I get a black screen and not even garbage? There's two 175
Flipflops which create the video signals. Clear is stuck, that's Blanking.
That comes from an LS109 JK Flipflop whose Clear is also stuck. That comes
from a 7425 which has two inputs floating. These (actually one input doubled
because 4 inputs, only 3 needed) come from the Z80 PIO, so it's CPU controlled.
The PIO never gets selected, /CS is stuck high. The MUX generating that signal
has three of its four inputs floating. These come from an LS244 right next to
the CPU, these mustn't be disabled but they are! On the back of the board I
see the trace for the enable coming from a ground plane and the trace looks
suspicious. I scratch at it with the scope probe and it disintegrates.
Put a wire bridge in and the game now works wine. Now all I have to do is
reverse-engineer the PAL.
Oh cool, no I don't. When I look online for SEGA System 1 PALs, I don't find
it. If I google Pitfall 2 PALs, I don't find it. But if I google Wonder Boy
PALs, bam, there it is! Someone even took the time to re-code it to work with
GAL16V8s, so I can burn that!
Unfortunately, while it generates the signals and the scope says the game is
running, the sync signals are irregular, the screen is shaking like mad!
So they aren't fully compatible (bootleg...), so I guess I have to reverse-
engineer it anyway...
Nah, his Wonder Boy runs great with the burned PAL, so I put his PAL in our
Pitfall 2 - win-win!

Tetris (Original)
Phew, that was five years ago... I had 2 boards, one of which worked, the
other one was from Winni who abused it as a test pattern generator until it
stopped working. That board had crappy IC sockets and started working once I
replaced those.
Incidentally the other board stopped working after that, only when it was cold
it would work. Turned out one of the POKEYs was bad.
Unfortunately, the SLAPSTIC also died, so that's a write-off.

Tetris (BL)
Draws a black screen. There's some garbage on powerup which gets cleared and
then it crashes and watchdogs.
The EEPROM is missing, but I tried it with a working board, it should work
without it. And if you swap the EEPROM from the working board in, it doesn't
make it work either.
So let's look for invalid/abnormal signals.
And there's an LS374 next to the edge connector. 7 of its inputs look exactly
the same, floating with single bits every now and then, they're either 1 or 0.
The 8th input only changes between low and a bit above floating. What else is
hooked up to that?
Two LS240s right next to it. One is already socketed. If I remove it, the
signal doesn't change. If I plop it on top of the other, the machine appears
to boot! So out with that, new socket and new IC in and now it's working!

Cosmic Invader (Universal)
Here we have not one, but two fifth-class Galaxian wannabes.
One has the game running with garbage running through the image and the
sprites are missing, the other one is just dead, garbage screen, never boots.
Already on the screenshots I could see that the garbage through the image on
the first one ARE the sprites, so first I check the sprite RAM on the first
board. One chip has a leg bent and not making it in the socket, but that can't
be it, as there are two rows of RAMs (4x 7489 and 4x 74201).
The '89s have their Write Enable stuck high, so the machine can't write to
this bank of sprite RAM. The 201's Write Enable is... floating. Testofon says
that's coming from a 7402 on P4 and three of its outputs are floating (and the
'02 is not a tri stated device...) and the third is low with both inputs being
low, but that's a NOR, it should be high then. Whatever that chip's dead, new
one in and my monitor isn't synching. So I jiggle the CSYNC alligator clip and
BOOM the 4116 RAM blows up. F#*k. Removed the RAM, tested it, phew, five of
them still work, still bad. At least two of the three marked bad RAMs from the
QIX I used for piggybacking in the DECO still worked. Well anyway, the game
now works, no idea if sound works, but the sprites are there.
So to the next one - Z80 goes to HALT pretty much immediately after reset.
The Z80 doesn't get warm either. Hm. But it is working in the other board.
I see funny stripes through the picture, every 16 pixels are filled, 16 pixels
are black (except for starfield). It does that without the CPU too.
One of the RAMs (C8) has the output stuck high. The input looks like crap
though. It's DB5 and looks crappy everywhere except on the CPU. But DB4 on the
CPU is stuck low while it's happily toggling everywhere else. That makes me
suspect the BUS transceiver (2x 74244@J7/J8). Pretty much all the pins on J8
look crappy, even pin 1 which is the enable input, but that looks good enough
for the chip to work. With a piggyback, the signals look way better and the
CPU now at least tries to do stuff, the screen changes. Btw. the 244 gets very
hot while the one above it "only" gets very warm. Anyway, out with them.
The chip tester says J7 works OK, but the current draw seems to be at 200mA,
that's excessive, and J8 has the entire christmas tree of error lights on, so
since they're from the same batch, I toss J8 and keep J7 for me as I'm running
low on 244s at the moment, but who knows when that one's gonna fail.
Back on the test bench it's still dead. So I remove the CPU and... Pin bent.
Straightened it, put it back in, game runs.

Sprint2 (Atari/Kee)
Hm cool that thing has 2k ROMs already... waidaminute all datecodes are from
1978? I thought that game was from '76! Did they produce it for that long!?
So what's it doing? Dead with Flaky. It clears the first 256 bytes of RAM and
starts drawing the road (at least the part that fits in these 256 bytes).
It crashes and watchdogs (and the track has flaky pixels too).
Selftest doesn't work either.
I removed the board and took it to the museum to test the RAM and other chips.
The RAM test passes, only bit 4 took two retries. The standard glue logic also
passes.
Uuhh what now? Desolder the ROMs, the schematic says the pinout is unlike any
other ROM I've seen so far, so I gotta make an adapter to read it as a 2716.
The ROMs match except for the boot ROM, so I uploaded it to Romident and it
comes back as Sprint 2 Revision 2, so it's OK (most Sprint downloads don't
include any of the BPROMs and/or are only for the original 4 Bit ROMs).
You don't really need the sprite ROM and the character ROM to get the system
working, but they used a 256x4 bit PROM for timing generation and a 32x8 bit
PROM as an address decoder (like a PAL). So I desoldered these and tested the
address decoder (it matched). I couldn't test the 4 bit PROM at home, so I
leave that for later. I noticed VRESET never comes on.
So with the ROMs, RAMs and glue logic all working, time to test the CPU in a
working machine (I tested it at the museum in an Atari Tetris with a bad
SLAPSTIC and it behaved the same as its CPU), so I put it in my CBM 8032SK and
played a round of Kalawaum and Canabalt (in Kalawaum, there was a 132HP+-10
enemy behind EVERY SINGLE DOOR ON EVERY LEVEL, rendering the game unbeatable,
but that's just one unlucky Random Number Generator seed, so everything's all
right), so CPU is OK.
On my workbench, the board is even more broken, it doesn't even render the
beginning of the road. And sometimes it watchdogs, other times it doesn't.
Sometimes the picture changes, sometimes it doesn't.
I even made a NOP generator to test if all address decoders have output and
they do.
With a video probe I see the RAMs bit 1 and 7 have pretty disgusting outputs
that flicker like mad even if I hold down the Reset button.
I even clipped bit 7's output leg off to see if anything's interfering with
the BUS... although that's not a BUS, the output goes to an inverter which
goes back to the CPU and to the display latch.
The only reliable thing that happens is that the CPU clears the first 256 byte
of video RAM (which happens to be the main RAM).
After throwing everything at the board for days, I googled a repair log and it
says that you have to connect both coin switches or the game won't boot.
Thanks. Now it does the same thing it did in the cabinet - clears ¬ of the
screen (256 bytes), draws part of the road with flickering glitchies in it,
crashes and starts over.
*SOMETIMES* it starts the self test when I short the corresponding pin, but
only sometimes. Holding and resetting doesn't work, power cycling works
sometimes. Wow that's flaky.
Sometimes the self test has flickering garbage across the screen, sometimes
a few characters flicker and it shows
01234567
  BB  BB
And a few more B's all over the screen, no idea if any of those mean anything.
Sometimes (actually long before I shorted the coin switches) it tried to draw
PUSH START BUTTON, which flashed and turned into PUQH QTAPT  UTTMN (bit 1 is
faulty).
I noticed that the 4 bit DIP switch that says "DO NOT ADJUST", switch 4
"kills" the CPU, so you can troubleshoot the video circuit and part of the RAM
more easily. The RAM is chip selected all the time (stuck low) in that mode
and the inputs are all zero, so let's do a write and check for ones.
And there aren't any!
When cycling 4-block DIP switch 4, I could get the RAM to fill with all kinds
of random patterns (same random byte over the entire screen), I got it to show
all "0"s once. And sometimes you could see some "2"s showing up sporadically,
hinting at bit 1 again. Before I couldn't quite make out that there was always
the same bit flaky when stuff flickered (which it did with or without the CPU).
So I remove bit 1 and bit 7 (where I cut the pin) RAMs, put sockets in, give
bit 7 a new leg and then... I swapped them just to see what happens.
And would you believe that? The game boots. And runs (with minor issues).
And it reliably reacts to self test, which now shows no errors.
The drivers are totally drunk though. The cars go all over the screen with no
rhyme or reason or (barely any) correlation to the track currently on display
(of which it only shows like 4 for some reason), jitter in place and/or
teleport randomly across the screen.
What's also striking is that instead of flickering 2s (or any other flickering
letters), I now get flickering "colors" (background is gray, foreground
objects/text can be black or white), stuff that's supposed to be black is
white and vice versa or flickering, characters change color mid-word...
Guess what bit 7 is for... it's the 1-bit-"color"-RAM. And since the RAM is
used to position the sprites, maybe that'll also solve our sprite position
issue.
Well... I don't have 2102 RAMs, but I got a bunch of 2125 RAM for cheap.
These are 100% identical but with a different pinout. So I make an adapter and
use a 2125. And that's it. The game runs perfectly fine (and goes through all
tracks in attract).
(Dude. Two different self tests have tested this RAM as good unrelated to each
other)
(I suspect the RAM has an address fault, if a certain address bit is set,
it'll read the wrong value every now and then, i suspect one addressing gate
on the die is faulty)
And independantly from that, the monitor has a problem... two problems...
for once, it does sync horizontally, but where, that's up to chance, but never
even near where it should. The second fault goes away after a few minutes...
there's a spot in the bottom left corner where the cathode ray doesn't reach,
it looks like a very strong magnetic field is bending the beam around that
corner. After a minute you can see it getting less and less noticeable until
it goes away completely and the picture is normal (except horizontal phase is
whereever it wants to be)
Studying the schematics showed it would be most sensible to check the power
supply of the sync comparator... which was crap. As I thought, one of the 4
sections of that big multi bucket capacitor was completely flat. Which of
course means that now the horizontal frequency was twacked because of that and
since I screwdrivered it somewhat back, it's gonna be screwed again now.
And yes, it was. And even worse - the ferrite core from this stupid coil broke
(because why use a potentiometer like everyone else?) and you can't change the
frequency anymore. So I desperately threw different resistors at the circuit
until I found a spot where a resistor would significantly impact the frequency
and soldered in a potentiometer and now it locks the picture horizontally
where it should be.
And I don't know what was more fun - playing Super Mario Kart or Sprint 2.
One more problem with the board turned out that player 2's sound channel was
very quiet even when you turn it way up and the amplifier chip gets scorching
hot. So I gotta replace that, the caps measure OK.
So on my next visit, not only didn't the customer buy the replacement chip,
the machine stopped working entirely. Black screen and motor on in idle.
Since I didn't have my scope with me, we unboxed his (unfortunately a non-
noteworthy modern thing with a color TFT like everyone uses nowadays).
And it tells me the data bus is dead, CPU is dead. There is a something on the
CPU clock input, but that's not TTL. There's spikes looking like something
trying to drive into a short circuit. Without the CPU, the signal looks fine.
So I shorted it to +5V which seems to work, funny, the clock signal outputs
seem to toggle when I short it. Unfortunately, the short stays. Where does it
come from? A 7432 way on up. Piggybacked it, now high is at 0.5V. Another
piggyback on top of that (with 3 chips driving the signal together) and now
the signal looks normal and the game runs fine.
That's an evil failure mode - the 6502 is working just fine, but the clock
input is low impedance (or high? anyway near shorted)!
And now I bought a replacement TDA1004 amp and so did my customer.
Mine had a leg snapped off, so we used his and it only gets as warm as the
other channel, so that fixed that, but not the low volume... and you can only
turn up the volume around 20%, after that only the tone changes.
After some time I wanted to go back into test mode and saw the two volume pots
under the test switch... the right one was turned down almost all the way...
facepalm well deserved...

DragonNinja (Bad Dudes BL) DECO
The picture got stripes through it, every second scan line is just gray.
At least on my monochrome monitor.
There's a ton of narrow 2K SRAMs on the board, two of which are in a different
orientation and from a different manufacturer and one of them gets a lot
hotter than the other. But the signals look OK and piggybacking didn't help.
So for months I'm searching myself crazy off and on every couple weeks a
couple of hours until I arrive at a flipflop where when I short the input, the
screen either goes all gray or all perfect except for missing layers.
Using a video probe I can only conclude that the signal is for prioritizing
graphics layers ("this pixel is a sprite, this pixel is background") which
leads me back to the hot RAM. So I socket them both, hooray I can get stripes
if I plug the hot one in either position and I don't get stripes if I plug the
working one in either position. I don't have 2018 (narrow 6116) so I adapt an
8K Cache SRAM from some long gone 386 mainboard and now the stripes are gone
and all the layers are there. Or something.
Unfortunately the machine developed a new problem, the letters and the
graphics of some layers are swapped by pairs (instead of ABCD it's BADC), so
instead of BEST PLAYERS I get EBTSP ALEYSR.
Looks like a cocktail table issue, maybe one of the LS86 took a hit, but it's
only on some layers, so to find the layer, I short out address pins of every
SRAM I see (except the CPU area). And it's on the bottom board. And I found
the correct pin. I traced it to a MUX that gets a square wave (probably 8H)
and some CPU data because the CPU has to access the RAM too. So I follow the
square to an LS86 (hooray) and its other input is floating (hooray but also
WTF?). If I pull that low, the letters sort themselves out.
So where does that trace go to? An unpopulated IC which has never been
populated. I still know this has worked before. The other inputs to the other
gates of that flipflop are floating too and also go there. But if I pull these
low, the graphics get garbled, so these must be floating (read high by the IC).
On the unpopulated footprint, two vias have flux on them. One is ground, the
other goes to the one gate that works when low (8H). That means there must've
been a piece of wire between the two fluxed pins that must've gone missing.
Whatever, new short in and now it works.
The machine has another huge problem I haven't talked about... it doesn't go
into play. Title screen works with all animations, looking good except the
heroes' trousers which have a few cells with the wrong palette which I even
can see on the monochrome sceren, then it shows the highscore table and then
either a sandstorm of glitches (with a few frames of level graphics) or it
goes directly to some guy saying something in Japanese or Chinese.
I don't remember if the machine always did that or only most the time, it's
hard to tell with such a bland beat-em-up I have a whole other crate full of
them.
Well it doesn't go into game play.
And it uses a lot of juice, I had to parallel two bench power supplies and had
to turn them to 6V just so I could get around 5.1V at the machine.
(Amperage at approximately 6 to 7). After a while, the machine suddenly
started working perfectly. I measured 5.5V. Turned it down to 5.2V and it
kept running! Turned it off and on again - still working. Waited a little
longer, still running. The voltage sagged down to 5.0V and it was still
running still. I flex the boards, nothing. It just runs now.
And at the museum it reliably runs   not.
Drat.
Crap, I completely forgot about the wall.
After the title screen, there's a high score screen with black background.
But there should be a wall there. When there's a wall, the game runs. When the
wall's missing, the game crashes. I even knew how to make the wall appear.
I think. But I didn't make a note.
Well anyway. I tried... lots of things and found lots of funny things.
Like an LS148 priority encoder which has an illegal signal on pin 10 (input 0)
Shining a light through the board shows, that's simply not connected.
Elsewhere, there's an LS00 that has something connected to pin 1, but nothing
to pin 2 and 3. (where the bootleggers drunk when they designed this?)
And after lots of searching I found an LS08 whose output on pin 3 wasn't low
enough for my taste. Piggyback - the screen changes! Hooray!
Clipped the pin, it wasn't the chip, there's a signal on the other end!
That mustn't be (08s aren't tristate). So I beeped through everything and
there was a via parked a little too close to another trace.
Scratched it through and - still no wall behind the scores and still some
bad color squares in the pants of the heroes on the title screen, but now the
game runs!
I got a bit panicked when I tried to start a game and saw that japanese or
chinese speaking guy again who usually only shows up when the game had crashed,
but it did go into play after that and I had a player character that reacted
to my inputs.

Carnival sound CPU
The game is working, all the sounds are working (I think), but no music.
Not that anyone would miss it, the music is so annoying that the programmers
put a way into the game to turn it off.
After getting access to a Carnival cocktail table (with Carnival inlays,
pretty substantial Carnival screen burn but some space shooter from '86 and no
Carnival CPU board, but both sound boards were still there), I borrowed its
Sound CPU. Thereby I notice - this thing uses an AY-3-8912, same as in the
Vectrex. I was sure this thing abused a Timer IC for PC speaker style beeps.
And sure enough, ours had an 8253 timer IC.
Well it does play music with the other sound CPU board. And some kind of
jingle to count the bullets after the end of the round.
I put that 8253 in the chip tester which failed it badly, which was surprising
because I would have bet the machine doesn't have that IC in the database and/
or can't test it anyway.
Anyway, let's see what's in our IC box.
There's an 82C53 I scavenged off a mystery Siemens board and it was one of two
ICs that didn't have Siemens numbers instead of regular markings (like
Williams and HP), rendering them ungoogleable (I did manage to guess at what
they are and managed to figure out all but two). So let's put it in.
Still no music. But we also got an 8035 CPU in the box. Let's swap that in.
Sounds exactly like the AY, at least ingame. The plings at the end of the
round are missing the decay. Let's put the original 8253 back, it only plays
one voice which is part of the bass accompaniment. Boop boop... boop boop...
boop boop... boop boop...
There's three ICs on that board and two were bad... ok 3 more 74xx but those
don't count.

Street Smart
Well this uninspired could be named "Pretty damn stupid".
Curiously, every fourth pixel in a 2D pattern is set
..
.X
in some backgrounds as well as some sprites. First I thought of a bad ROM,
there's a daughterboard covering 2/3 of the main board and it's littered with
ROMs. But checking that got me nowhere. So our next suspect is the RAM,
unfortunately most of the RAM is under the daughterboard.
Piggybacked one, now I get an error on boot so the machine can test the RAM
under the daughterboard and the piggyback isn't sitting on right.
So I started shorting RAMs to see which ones do not give an error message.
The smartest thing about the game is that the error actually tells me the
position of the RAM on the screen (like it says RAM ERROR R15) and the RAM I
shorted was at position R15. There's a bank of eight 2018 (narrow 6116) I can
short without the machine giving an error. Video probe shows nothing
suspicious. Scope says there's a weird mains hum interference on all data
lines (but just those) and all the other signals are OK so it isn't these, so
it can't be the ground connection. What other tools have we at our disposal?
Good old finger test. One of the RAMs is getting noticeably hotter than the
others. Piggyback it, graphics errors disappear.

Vendetta
An original Konami board!
Uhh where do I even start?
The machine is dead. It doesn't even generate CSYNC while it's in reset.
Not even CPU clock... wat? Err that seems to be intentional.
One output pin of both PALs has a floating signal. The other outputs do work.
Below it is another PAL, all outputs are stuck Low, three of the inputs are
low too, maybe that's supposed to be like that.
So what else is the board doing?
It's suspicious, there are two SRAMs and their BUS transceivers getting very
hot. They form a 16 Bit BUS (palette RAM or something, pretty far up the video
chain) and 8 of these look ugly. A piggyback makes them look nice. So I swap
those. And that didn't do less than nothing, I even scratched a few traces and
had to bridge these.
So... the CPU's RAM doesn't get warm. Suspicious. The only RAM to not get warm.
And the BUS transceivers from the RAM I swapped are too hot to touch.
So I swap these too, only killed one trace this time and that near its pin, so
easy to fix.
So. Now I'm at my wits end. Piggybacking the Z80 tells me both CPUs agree to
100% on what to do next.
So only now I get the schematics.
And visited system16.com which told me the Z80 isn't the CPU, but an unnocuous
(is that the right word?) custom IC on the other side of the board!
8 bit data but 24 bit address... whoa! So armed with the great new circuit
diagram, I checked where the broken signal from the upper PAL went... nowhere.
There's a footprint for some SMD component which isn't populated and that
trace goes nowhere else. Maybe it's configured as input.
And the other PAL? Err... that controls all the chip selects in the CPU area
like the PLA of the Commodore 64. So never must there be more than one output
Low at a time. If you overvolt the board to 6V, some of the outputs come back,
but only up to floating. That PAL's toast. And as luck would have it, someone
bruteforced this PAL. Now all I need is a GAL and off we go.
And with the new GAL burned and installed it works a treat - except the colors
are weird and flickery. Maybe the other color RAM is bad as well. Swapped it,
now it works fine.
Now I got the board sent back because "sounds are distorted".
It's not the amp, because music sounds good and the distortions sound digital.
Some of the sample ROM databits look yucky. I remove the sound ROM and put it
in my adapter. It appears to not be reacting to some address bits, the
contents is half identical, half wrong, but weird. Ordered one from Reichelt
and when I wanted to burn it, I noticed I ordered the wrong one. Oops.
Well, next time.
Before burning the replacement, I did dare to push the original ROM deeper in
the adapter socket and now it comes back 100% perfect. Hmm... you know what,
I burn the ROM anyways! Dammit, ripped another trace desoldering the original
one, gotta fix that too.
And what's it doing with the absolutely identical ROM? Er... the sound effects
sound perfect now... but what's that buzz droning out everything?
It changes frequency and seems to correspond with the motorbike sounds.
With the original ROM back in (everything's distorted, so that's affirmative
the original ROM has a problem) I can correlate the buzz to the motorcycle
sound, it's the same sample, but instead of actually playing a sample, it
plays a square wave (likely it's in an area of the ROM filled with 0xFF),
but why? All the other sounds are fine. Maybe I have mixed up the higher banks
or the pinout isn't the same.
Ah - must've burnt it wrong. But if I solder a bridge from pin 1 to pin 42,
ALL the sounds are correct. So finally I got this board off my bench.

Monster Bash (Sega)
Unfortunately not the pinball machine. Pins are below.
Tested, found no problems (joblot of untested boards), so I put it in a
machine for a playtest.
Looks great. Self test passes, all sounds work (and that evil laughter!).
Second attempt, all the wave sounds are missing. I gotta investigate that at
home.
Then I played it until I could get to level 3.
At home on the bench I wanted to investigate the missing wave sounds, but the
backgrounds are garbled. They're not moving or flickering, but look different
on each screen.
Man have I ripped my hair out over this.
Schematics only exist for the G80 card cage version, this is a two board
version (at least the sound board looks like G80).
And the schematics are illegible. You can make out just enough to see my two
board version there doesn't have much left to do with the schematics.
There's no background RAM like the G80s have. Instead there's Name Tables in
ROM whose data output drives the address inputs to the character ROMs.
In the schematics, these get 1H..128H and 1V..128V from the video address
generator. The two board version has its own address generator just for the
backgrounds. That goes through a bank of LS86 for Cocktail table mode and then
a bank of LS283s so it can scroll the playfield if it should ever decide to do
so (hint: it won't. ever.). WTF?
And then there's two LS161 being fed and feeding back into some kind of sync
PROM (8 bit x 256). These seem to play a major part in graphics generation.
Since they get Load and Clear, you can piggyback'em. And they agree!
Then I went to the museum to test every single IC, I even went and wrote a
program to dump these weird sync PROMs I talked about (guess what - they are
100% identical to MAME), I found nothing tangible, but the two LS161s and the
sync-PROM were among the "closer look" candidates. One barely toggles, but the
piggyback doesn't fight... unless I put ANOTHER piggyback on top. Now I can
see some fighting. Nothing going on on the screen. OK that's my only straw
besides an LS125 which showed up as completely borked, but in the machine it's
almost permanently disabled and thus can't be the problem).
So I socketed the suspect LS161 and... it's back up and running.
Back to the wave sounds... finally I can test this.
What enormous fun... First you have to start the self test. Which takes some.
Then you have to wait for the RGB test. That takes some more. Then you gotta
short one of the start buttons to ground to go to the sound test.
Then you have to endure all these (how do you call these wind up things that
play melodies?) really shitty annoying music jingles until finally the wave
sounds are being played or not. How can you see if the sound CPU works?
When the wave sounds are working. It sets itself to sleep when there's nothing
going on, it's a microcontroller, not a CPU, so you don't see a difference
between idle and dead.
Close inspection of the sound board yields two ripped caps and one that leaked.
Swapping them made no difference. Then I measured a small electrolytic near
the sound CPU and see the level jump and hear the speaker click.
Looks like it's the power on reset cap.
Swapped it, now wave sounds work reliably.

Frisky Tom
Wow - it's not a bootleg! I procrastinated this one together with a Crazy
Climber (also an original) forever, since they had all the ROMs stolen.
Crazy Climber still had the graphics ROMs in, which allowed me to ROMIDENT the
game, Frisky Tom had an original Nichibutsu license sticker that said Frisky
Tom on it, making things slightly easier for me. Frisky Tom still had the
color PROMs installed where as Crazy Climber didn't. Which is weird since you
can't do anything with these PROMs except install them in a Crazy Climber
(OK technically you could install them in a Galaxian to get weird different
colors)! Humans do weird things.
So OK let's try to restore this guy, ideally with a Big ROM hack. The four GFX
ROMs aren't all hooked together, so it's more sensible to just install four
2732s.
Beeping the 40 pin DIP sockets allowed me to find out where the Z80 goes and
where the AY-3-8910 goes.
And here's where the fun begins - there are 9 ROM sockets next to each other!
Ah nay, number 9 is for a 6116 2k SRAM. Phew.
And here's where the fun begins - that RAM shares the upper 4k with the last
of the ROMs of the 32k Program ROM bank which is a 2k chip while the other 7
are 4k! Argh! And that makes the chip select logic just a teeny tiny tad more
complicated...
I think the formula was A15 | (A14 & A13 & A12 & A11). I implemented the AND
with 4 diodes and for the OR I had to scratch a few traces on a CMOS chip
because the engineers followed the rules and pulled unused inputs to ground.
And does it work? Well yes but actually no. The screen says RAM TEST...
RAM TEST OK. ROM TEST... ROM TEST OK. And then it watchdogs once and the fun
starts over. In some rare cases the machine goes into game, the title screen
does look cool, it gets broken by enemies and the player tries to fix it
Fix It Felix style, then an enemy drops a bomb, the player kicks it off screen
where it explodes. Then the attract mode starts, looks like Super Pipeline on
the Commodore 64, but on steroids. If the player dies here, the game freezes
(so far so normal), but then nothing else happens except the timer continues
to count down. When it reaches zero, the sreen redraws and attract continues.
Then a baddie throws a bomb which explodes, the screen goes black apart from
the dead player sprite and then it sits there forever.
Did I mention that resetting doesn't work, it just goes back into the infinite
self test?
Well ok let's see what the DIP switches do. They do nothing, apparently.
Then there's two weird inputs on the edge connector: TEST and CONECTION [sic]
ERROR. If you pull the latter to +5V (instead of GND), the screen shows
exactly this in red on the screen, even containing the same spelling error.
If you pull TEST to GND, it displays (after the test IIRC) a crosshatch
pattern together with input states (joystick test) and you can test music and
sound effects which even works!
But - as I said - game doesn't work most the time and even then only rarely
and it freezes in attract reliably (and I just remember that - yes it gets
IRQs). If you remove the custom IC, it never goes to game.
With my wits at its end, I went to the Crazy Climber. That one was easier, I
didn't have to burn GFX ROMs. This one has 5 2732 PRG ROMs and then two more
in the sound section that form a 4 bit DAC. These are guaranteed to be for the
samples for the speech output ("GO FOR IT!"), so we don't need those for
testing, just as the color PROMs because I can hook the video probe to their
inputs to see if the game runs.
OK so 5 PRG ROMs, not 4. But the 4 are mapped dead simple, I don't even need
extra logic. Oh wait. I gotta use A15 | A14 as chip select or the ROM wraps
over to where ROM #5 sits. Well whatever, two diodes and a resistor because
all the logic ICs are completely used up. For #5 I use a 2732 as intended, the
other 4 I cram into a 27128 (which needed 21V to burn).
Well whatever. I hooked the video probe to the color PROMs inputs and the game
was clearly running perfectly, sprites and all. And rebooting shows that the
game does boot reliably.
Then I can burn the other ROMs too. The sample ROMs are super simple, one chip
select is grounded, the other is mutually NOTted, guaranteeing that one chip
is always selected, so I use that as the upper address line for a 2764 and
ground its second chip select too.
Concerning the color PROMs... I did build my own burner a while ago after
studying the datasheet, since I had a Galaxian color PROM to burn and 6 to
spare... ok so after this I only have two more to spare.
And I managed to burn one bit on one address on one chip wrong. Oops. Can't
undo that, now something lavender or violet is now orange. My bad.
I don't think that's the end of the world, since the other 95 values are
correct and if you put this boardset into a machine, it will work.
So back to Frisky Tom... by now I got a few inspirations. Maybe it's my wonky
chip select logic. So I replaced the diodes and the CMOS OR with some proper
74LS logic and... nothing. So the next thing I might wanna do... we have so
many spare EPROMs at the museum, let's steal a bunch and give the machine the
chips it wants. And that... also did nothing! Oops there was a scratched chip
select line on the one 2k ROM. Fixing that also didn't do anything.
OK next.
How about replacing the #4 socket with a 28 pin socket that gets another 28
pin socket on top that has the extra lines so I can just swap it with the
EPROM simulator? And try different games that run on the same hardware?
I mean there's like 3 different sets of Frisky Tom in the first place!
Let's try them!
After some starting difficulties I managed to get to the endless RAM/ROM test
screen with the simulator, so I assembled friskyta and friskytb and they do
differ quite a lot on a binary level, but they behave so much the same that I
doubted that the file transfer worked! OK what does Radical Radial do?
Black screen, wait what was that, it's gone, now there again? Depending on how
long I hold the Z80's reset pin low, I can either get a black screen or the
game boots - halfways. But more than Press Start Button and the score info
stuff on the right of the screen it doesn't do. You can start a game, which
flashes garbage in the center, then there's garbage on the bottom and it
freezes there. Hmm... don't we have a Radical Radial somewhere? Maybe I can
swap the custom? Or it's a bootleg? I gotta check that.
Anyways, onwards.
Seicross runs! It doesn't boot reliably, you have to reset it once or twice,
but it runs! And if you didn't know how the hardware does the parallax
scrolling, you'd be impressed! But Scramble (and probably even Galaxian) can
do it too, by scrolling every text column pixel precise individually.
At least the attract mode runs and you can coin up and (with alligator clips
only somewhat) control it and you can die and go Game Over and the machine
goes back to attract, so that's a clear win! So what happens if I remove the
custom? Nothing, it keeps running! But if I reset it now, it doesn't boot even
after 20 tries. If I put the custom back in, it works almost first try.
The machine apparently only needs the custom to boot, which implies the custom
is damaged or at least flaky. Now I read somewhere this is just a relabeled
6802, gotta check if there's something to that.
Worst case the machine will run Seicross, that looks awesome!
Best case, the bootleg has a daughterboard I can copy and get Frisky Tom to
run as well.
Even better - Radical Radial is nearly identical to Frisky Tom and has the
same custom IC. I swap'em - no difference!
I set the DIP switches on the mini-board  the same - nothing.
I swap the mini-board - nope!
I take the Frisky Tom video board and hook it up to Radical Radial - the
graphics are garbled (because of course it's running on Frisky Tom GFX ROMs),
but it works just fine.
I re-check the 7400 logic... the one that fails fails the same on the working
board. Last straw I swap the AYs. Nope again. I put the Frisky Tom ROMs in
Radical Radial - finally I can see the entire game run even though with
garbled graphics because wrong ROMs.
Weeeelll what now? There's nothing left it can be. The sockets are crap but
beep out OK. Frisky Tom has an LS193 that's missing in Radical Radial which
has a jumper shorting two pins of an LS08 together.
The only thing I can think of is (if I can't get Seicross to work anymore),
would be to "just" make a copy of Radical Radial's CPU board and populate it
with the parts from Frisky Tom, because slowly I think of a cracked trace!
Hmm - I found something interesting on the bench - there's a 555 timer that's
fed an AC signal (which I don't have) and probably makes an interrupt source.
That goes in an LS193 and turns off its output. Really interesting is, that
the custom IC then really lives up - and on Radical Radial that thing's
running permanently.
And pulling the input to the 555 to +5V, I can get Radical Radial running on
Frisky Tom - somewhat. It keeps crashing or doing weird nonsensical things
drawing the road. It can run somewhat stable and do what it should - until you
look at the board weird. Then it either crashes or does weird things.
It is pressure sensitive, so I'm pretty certain now it's a cracked trace. 
Ah - and my funky ROM socket construction is similarly flimsy. Have to do it
over.
So I beeped through all the traces and found no problems...
And with the new ROM socket construction and the shorted 555 it runs.
It won't run without it and Radical Radial still performs Erratical, but
Frisky Tom runs! And it doesn't freeze! So bite me, I'll leave it like that.

Silent Scope
No I'm not gonna write about the Hantarex Polo 2 that smoked off and now kills
every HOT I install., even after replacing that one capacitor which only had
half of its capacity left. And not the replacement monitor that only fills
half the screen horizontally.
OK so we got 2 machines, 1 working boardset and 3 defective ones.
The working one only works somewhat as the RTC chip has a flat battery which
causes the machine to always boot to service menu. Swapping the CPU board into
other machines I could make out at least one fully working video boardset
stack, so if I can fix another CPU board, we got two fully working machines.
One seems to have different ROMs fitted, so the CPU flags those video boards
bad with ROM checksum errors (they're Silent Scope 2s).
OK that leaves us with two CPU boards that crash in different parts of the
boot process (that thing has a POST code display), but none even reaches the
part where the video gets initialized, so both sit on a black screen.
One CPU board watchdogs. Some butcher desoldered the RTC chip, drilled it up,
removed the battery and installed a battery holder. The latter was done really
well and looks good, but the former not so much so, looking really bad with
several cut and bent traces and one that's "fixed". But why would anyone
solder a data/address line to Ground?
So what does the Testofon say to the socket? A2 is shorted to ground and D4 is
missing. Turns out, A2 is the one that was "fixed", but the line should have
gone somewhere else instead, so I fixed the fix. Then there's still one or two
broken traces. Let's see if it works now.
I did also desolder two of the other RTCs (without breaking anything) so I can
install external batteries on those too.
So the good news everyone is that the fixed board now boots through! The bad
news is that drilling open RTC chips and hooking up external batteries doesn't
work, these chips self destruct even if you don't dig into the silicon.
You always get Hardware Error P11 or something. But now we have two complete
cages where we only need new RTC modules for! If we order these, that makes 2
completely working Silent Scopes!
--------------------------------Off Topic--------------------------------
Schneider EuroPC (8088, 512k RAM)
RTC battery leaked and has eaten away lots of traces (unlike Qix where it just
has eaten away the leads of some components and left the traces alone).
When you turn it on, it just beeps. For 10 seconds or more.
Scoped different chips. Graphics subsystem does nothing (it has that popular
Paradise CGA/Hercules chip). RAM signals look almost OK (2 are weird) and an
LS245 near the battery has 2 floating lines, one of which is Direction.
Downloaded the schematics. Now it's just a matter of ringing through the
traces and replacing the ones that don't buzz. The two pins should go to the
CPU controller IC. But they don't. Scratched up the traces near the pins of
the LS245, they have contact to the controller but none to the LS245.
Socketed that IC (the pins were green with acid remains and one pin was bent
when the chip was installed in the board) and bent up the wires that made no
contact and connected them to the scratched up trace.
Now the computer beeps once long (1 second) and once short on power on.
Graphics system is still dead. Graphics enable signal is on inactive.
Where does that come from? A JIM 5010 system controller (looks like Toshiba
made it). If I probe it, it does a lot more than back before I fixed the
LS245 which is a bus transceiver that connects the CPU to the rest of the
machine. There are two dead pins in the middle of actively toggling pins.
According to the schematics, these are DB2 and 3 and those next to them are
DB0,1,4-7, so they must pulse. Aaaha! Scratched open the traces right before
the chip and again, they made no contact. Bridged them, the machine now beeps
a concert on power up. But the graphics system is working now. Completely!
I go fetch the monitor (MDA version). It makes a disgusting whining sound and
shows an out-of-focus blob on screen. Looks like it really doesn't like CGA
frequencies. So I turn the machine off, hope for the best, turn it back on
again, hoping it will read out the monitor's MONO switch and yes. With every
beep I got a readable error message, most of them concerning the dead RTC.
But... It's alive! Just, it doesn't memorize BIOS data, not even when it stays
on. So I really have to fix the RTC too. As luck would have it, most the RTC
control traces go into a resistor network on the bottom of the board where the
acid didn't reach and then on the top of the board from the resistors to the
chip. I just had to connect one single longer wire. Now the cursor does weird
things in BIOS and it still won't store BIOS data. Well so I have to fix the
Quartz circuit as well. But it doesn't oscillate. Fixed the entire battery
circuit, now it oscillates. And it now stores BIOS data (unless you turn it
off). And boots reliably and way faster than in 4.77MHz mode. So I put in a
battery, but it won't accept a date after 1999 - MILLENIUM BUG!
The guy who was right next to me at Classic Computing 2011 (kpanic with his
10 MHz 65816 in a CBM 8032) wrote a Y2K bugfix for Schneider EuroPC.
Burned it, but it isn't compatible with my model of EuroPC (even though it
said). Only detects 64k of RAM and then crashes. Put the old bios back.
I didn't test serial and parallel port. I was happy with Works and who would
hook up periphery to such an old machine? Well, Ari of course and she
complained, one of the chips doesn't work right and the other doesn't even get
detected.

BiFi Box
This is a computer on which the biggest part of the German version of this
file was typed on (this version is typed on a Compaq SLT/286).
It's a super mini 40 MHz 386 with Opti 82C495 chipset just like the board I
replaced it with, but this one has a socketed 386 which is even populated with
a 486. And runs twice as fast at the same clock speed, so it has double as
much time to get bored between two keystrokes (just think about it, if I type
4 chars per second, there's 10 million cycles in between!).
Problem is the machine died mid-typing (probably of boredom). Rebooted, BIOS
memory test crashes at 12 MB (16 installed) and then says GATE A20.
This has to do with the keyboard controller and usually is a sure sign the RTC
battery has gnawed through the traces. The acid damage doesn't look too bad on
this board and the only bad looking trace measures OK, so...
The cache's chip sockets are rusty&mouldy but not as bad as those on the
board I use as a replacement, so this can't be it and it doesn't run without
the cache either.
What I found though was that the 14,3C39 BUS quartz is bad. If I swap it with
the replacement board, both boards are equally dead.
Measuring the BIOS ROM's Chip Select line looks identically on both boards to
the point where it initializes hardware, then it lands in an endless loop in
the ROM BIOS.
Hmm if I drag my finger nail across the pins on the corroded chipset IC, some
pins move along. So it needs some rework done.
And that didn't help a lot. At least it beeps thrice when you remove the RAM,
which it didn't do before. Maybe it does need the 12V, -5V and the RTC quartz
(which I stole for I forgot what).
Damn it, while trying to resolder more of the IC (I now got a flux pen), I got
some pads to come off and now I can only trash the board. Sad.

Nordmende Film scanner 451A CCS (1975)
A very interesting and massively overengineered piece of technology.
It takes a Super 8 film and creates a PAL video signal out of that, watchable
on the tiny TV that was once the pride of your living room or recordable on a
VCR (they did exist in '75. And they *were* called VCR. Then came Betamax,
then came VHS. VCR then evolved into Video 2000 but the race was already run).
(ah yeah I forgot U-Matic was already out).
And it turns out, the machine (if working properly) is capable of delivering
significantly higher image quality than one of those new-fangled full-HD film
scanners you can buy new now.
And how does it work? Nope - no bright light and a camera.
It basically works like Motorolas EVR - in that there's a monochrome 5 inch
rear projection CRT underclocked at 16kV (but if it was a TV, it would've got
8kV max) drawing a blank raster and three flying spot scanners and some
dichroic mirrors. The flying spot scanners were made by Fernseh GmbH in
Darmstadt (they folded around the 80s and I live in a suburb of this city)
So the machine has the following problems: The picture is jittering like mad
(every second frame is offset by like 100 scanlines, the anti-jitter
potentiometer only makes it worse) and sometimes the film runs way too fast
and stops after a few seconds.
Looking at the raster on the screen it is evident there's a bright white line
on the bottom of the screen which already has burned in a little and the black
bar that appears in the center of the screen when playing is offset to the
side with the white line (when playing, instead of still doing 50 scans per
second, it does 25 scans per second, but not in the same spot, instead on the
top and bottom half of the CRT so every film frame gets scanned twice and the
cathode ray runs quasi synchronous to the film).
From the looks of it, this is almost guaranteed to be bad capacitors
(cue twilight zone music)
Inside the machine I can see the usual suspects, but they're OK (most of them).
Some orange tantalum barrels have started leaking (electrically because solid
electrolyte) but nothing which could be the cause.
In the deflection circuit you can see the signal sit on a voltage for a while
before starting to deflect.
Turning potentiometers didn't help.
I calibrated the waveform from the deflection board  to specification, looks
great. Then I notice one of the main deflection caps on the main board isn't
what's called for in the schematics - there should be a BD135.
(I kinda remember the complementing BD136 was inserted the wrong way round,
but it survived) (the transistor that replaced the BD135 should have actually
worked according to alltransistors.com) (I even scoped the signal with the
deflection transistors out of circuit and then it looked fine).
Well, nevertheless, the white line is gone and the black bar is now centered
and the picture is much more stable when it syncs up.
Because that's the second problem - sometimes it doesn't sync up and stops.
Reason being the light switch for the perforation which not only tells the
machine that the film is running, but also to synchronize the capstan speed to
the CRT.
And here's the problem: The signals don't make it all the way through the
circuit to the takeover capacitor (there's two complex circuits, one runs on
5V, the other on 24V, in between there's a coupling cap).
Through trial&error I found out at which voltage the signal needs to be for
the circuit to detect it and then I used a 1megOhm resistor to 5V to pull it
up into the comfort range of the first transistor.
After ten minutes of tweakign I got the machine calibrated so that you can get
a stable image with the flicker potentiometer centered. Well, stable in the
center, but not on top&bottom. I just wanted to use the manual to calibrate it
(there's like 9 potentiometers for that), when the perforation detection light
switch circuit failed again. This time the tape kept running, but way too fast.
Took me forever to find that.
I took the machine to the arcade museum because there's lots of 7400 logic in
there, basically half a PONG machine including the video address generator,
which I can test with the chip tester. These ICs are from 1974.
And while I'm at it, I swapped a few numb orange tantalums in the mechanism
control circuit, it didn't help but it couldn't hurt.
After taking everything out of circuit that could interfere with the capstan
speed control I could state that the æA709 op-amp is still doing something,
but not the way it's supposed to. For example none of the potentiometers on
the negative input had any influence and the signal on the positive input was
matching the manual 100%. When fidgeting with the output signal, you could go
to fast forward.
And here's the cool thing - in that that radio/TV repair shop surplus was
exactly one such op-amp. And after replacing the one in the machine with this
one, the machine went back to working perfectly. I couldn't get the picture to
stop shaking unless you move the flicker fixer pot to the maximum position,
but then it's solid top to bottom (only horizontally it's jittering a little,
sign of a crappy camera - spoiler alert: No, it isn't).
Another interesting thing... I got the machine to produce a really awesome
picture, but as soon as you put the bottom metal sheet back on, the picture
goes back to grayish and dull.
So... operate it without that metal cover.

ANOTHER Nordmende Film scanner, same model, later revision
Now we have LEDs instead of light bulbs (heck even a green one, bleeding edge
in '75) and the little circuit board for shaping the perforation light switch
pulses which I desparately looked for on the other machine is here on this
machine but not on the other.
Symptom: I think he said a color is missing. Not sure. By now the picture gets
darker and darker and by now it's inverted and so weak my monitor has trouble
syncing to it.
So let's see if any capacitors are getting warm...
And there's one on the LOPT board in the corner that gets way warmer than it
should. A new cap gets even warmer. It's part of the horizontal deflection
circuit, so not part of our problem. It needs a tone frequency cap (ultra low
ESR) so it won't get hot.
One of the power supply caps is a little lukewarm (with nothing nearby heating
it up). Doesn't help, the power rails are OK.
Swapping the deflection control boards with the known good ones from the other
machine didn't do anything either.
And what's coming from the photo multipliers (flying spot scanners) looks the
same on both machines. Only one of the RGB boards doesn't have an output.
Flatline! If you swap the channels, the fault wanders, so the board must be
bad. The caps have the same signature on the Testofon, so I gotta do an in
depth search for the fault.
My instinct guided me to the only tantalum that isn't an orange barrel
(even though these have horrible standings) and it's good on the surface, but
has some tiny (but measureable) leakage current, so we check the voltages on
the transistors against the working board and against the schematics.
And the signal on the transistor that's connected to the suspect cap is
floating around 1V, whereas on a working board it's at 100mV.
So I replaced the cap and now it looks the same and when I put the module back
in the machine, the signal looks the same as the other two, so it is fixed.
Unfortunately the picture doesn't change at all.
After a long and unsuccessfull poking around inside the machine (manual says
there should be 10V here. I measure 5. I'm not gonna get any higher than 5V
when twisting the pot. And the working machine only has 4V there...) on the
video board I see a drastic difference between two identical video amplifiers
which are next to each other. (especially the defective machine has next to
no video signal going into the UHF modulator). One of the two amps has a
signal while the other only outputs like 24V or thereabouts.
There's two transistors in series. The input to the first one looks the same
on both amps. But the output from the first is at a high level with no signal.
Removed the transistor, Testofon says dead, it's a BC213. I only have 212s.
According to alltransistors the 213 has like 2pF less parasitic capacitance,
but apart from that the 212 has identical values, even a higher voltage
rating, so in with it!
And I'm greeted with a picture that looks fantastic. No sarcasm.
All the colors are there and they're vibrant like in the commercials.
Only there are two faint 50Hz hum bars, one bluish, one reddish.
It took me an hour to find out that the optics for the spot scanners is aimed
right at my ceiling flourescent fixture... Turned off the lights, the hum bars
are gone.
I press play, it runs, it syncs and it barely flickers. But it stops after a
few seconds for no reason. Measurements show that the impulses from the LDR
from the take up reel light switch are too weak to trigger the IC reliably.
If the reel turns too slow, there's no impulses. And the pulses look dirty
compared to the working machine.
So I went on a recapping frenzy (lots of bad orange barrels - you can reform
them if you got the patience and if you know the machine gets lots of use).
And the problem was the LDR itself. Checked all the components in the circuit,
even swapped boards, no difference. The LDR is cast in epoxy which has
discolored from the light bulb's heat and the huge amount of use the machine
has got (the CRT is well burnt in). So the only remedy would be a brighter or
more directed light source. So I put in a white LED which I bought 20 years
ago and overvolted no-idea-how-often so the yellow phosphor is a little burnt
out and the light color is more aquamarine. It's not a lot better, maybe
almost twice the amplitude, but enough so it works reliably even at 16fps.
Ah - the horizontal jitter is untolerable. 1cm on a 42cm telly.
After trying EVERYTHING in the horizontal deflection circuit (which isn't
much) I turned the CRT yoke and that fixed it.
(wow the image really is awesome.)

IME 84
Continuing with really ancient, rare but huge boxes...
I don't even know if I legally own this thing, they're currently demolishing
these old army barracks and the demolishing company was unceremoniously
throwing anything that was still left in there on a big pile of junk.
And among that junk heap was about a dozen typewriters mostly from the 50s,
lots of vacuum cleaners from the 30s to 50s, sewing machines from all kinds of
eras but probably mostly 50s, electromechanical calculators (like the old ones
with a crank, but the crank was replaced with a motor, that's all), two
microfilm viewers, a couple of 1950s space heaters and... a table calculator
from the 60s. The IME 84. Google says this machine was among the first
transistorised table calculators ever.
And there's not too much known about the machine apart from the marketing from
the day (as they were understandably proud of it). No schematics and nothing
on Youtube. Looks like I gotta film this myself.
So before first powerup it's important to assess the power supply. For me that
means using the Testofon to check if any diodes or transistors are shorted and
how the capacitors test. Well almost. There's 30 Ohms on the main voltage rail
and some transistors measure near shorted, 8.2 Ohms, but there's a resistor in
parallel and... it has 8.2 Ohms. Printed on the case.
Next I lifted one leg of each cap and put them under the Testofon. Most of
them are... usable, leakage beyond measuring range and an ESR of... still
existant. Some were bad, one 500æ12V is extremely leaky, after some rounds
with the Testofon it measures almost OK but testing the ESR showed there isn't
any, but there's a 500æ6V next to it that still has some left, so that cap's
for the bin. Both the 100æ100V for the nixie power supply (which happens to be
a switch mode power supply!) measured poorly, so out with them.
So now's the problem: No schematics and no way of powering up the supply
without the rest of the machine without desoldering tons of wires.
Well. The mains power switch is a slide switch small enough for a mid 70s
pocket calculator, nothing you would expect in a power hungry beast with 16
Nixie tubes (sixteen!). And it switches the primary. So it can serve as a
current limiter since it's pretty corroded.
(OK it turned out to not be so bad)
So, said a prayer, held my breath and juiced it up with the new caps in.
And one Nixie tube lights up showing a 0. Voltage regulator measures 12V.
On the second pin as well. The third pin has 18V. Switched to AC. Got 0.04V on
the 12V pins so that's gotta regulate. By now, more Nixies have lit up and
show all zeroes. Looks promising. I type in numbers and nothing happens.
Since I did some research on the internet and know the four keys on the left
with the light bulbs are my registers. I switch them through and on the red
one I see my input. The 1 key works, the 5 and the 7 key work. And the 0 key
works. I shoulda tested the keys when I had the keyboard off to test the
voltage regulator.
Well at least all the digits are working and the numbers are rolling cleanly
to the left.
I get my camera, switch it on and all 7s turn into 5s. Crap. Well.
Since it all happened at the same time, the problem must be in the readout
circuit.
So under the keyboard we go again. The problem with the keys sticking I will
fix later (tried oiling them but that only makes them plop out when you
unstick them from below, they still stick). While I'm in here I check the
debounce caps, 5 of these were borked. Cleaned and tested the keys and just as
expected, the 2 bit was defectivee, so instead of
0123456789 I get
0101454589
Now the important question: Is the fault in the calculating section or in the
display section?
First I gotta find out how to operate this thing. After some while I got to
this method:
Select red register. Enter 4+4+, select green register, the result (8) shows
up on display. So it's adding correctly using numbers not using the defective
bit. Now for the interesting question: Will adding 1+3 equal 0 (because
1+1=2=0) or will it be 4?
The answer: 1+3 (which displays as 1+1) equals 4. Hooray, it's calculating
right, it just can't show it.
OK now the problem: I could now point with my finger on the bad transistor in
the schematics, but there aren't any and I don't even know on which of the 32
cards I have to look. I gotta follow the Nixie wires to circle the fault.
It can be assumed certain there's a diode based decoder matrix for binary to
Nixie somewhere and the Nixie drivers are the only transistors not using
2N1035 or whatever it was (apart from the PSU).
This gives me a wishbone.
So follow the wires without schematics. And it's the first card on the left.
Which has a line of 8 transistors and two rows of 5 transistors and a matrix
of diodes. Looks like a decoder. So what does the Testofon say?
These transistors have some pretty liberal factory tolerances, but one just
fails completely, there's a diode direction missing.
Even better, there's photos of a few circuit boards on the web and one shows a
different version of this exact card and someone annotated the edge connector
and indeed - binary in, 1-of-10 Nixie cathode driver out. And the 2^1 bit goes
straigt into the transistor my Testofon failed.
Let's see what I can get from the Arcade museum (they have bags of old
transistors, some even period correct).
There we have 2N3404 (Si), 2N2905, ACY17 and ACY39.
Let's see if a silicon transistor even works as a substitute... yes, perfectly.
Momentarily, all 0s and 8s lit up at the same time but that was unrelated and
went away by itself.
I still ended up putting an ACY17 in there, it's wimpy but we have a bunch of
them and its job is logic so it doesn't have to do hard work.
So let's test some functions... Addition seems to work, sometimes I get wrong
results but if I do the exact same calculation again I get it right.
Also of note is that the reset button sometimes doesn't reset everything, but
just parts. And if there's parts of temporary results hanging in a register,
the final result can't be right.
And if I subtract, leading zeroes will become leading 9s.
The rest of the result seems to be correct: 2000-663=9999999999991337
(at least I think it was, and other calculators from the early 70s behave the
same, you need an extra step to get the right result)
In the meantime, another bit broke and that makes
0123456789 to
0123012389
So now the 2^2 bit has taken a hit. Testofon says transistor is bad, so "new"
ACY17 in there and off we go. Only... sometimes it takes 5 minutes until all
the Nixies strike (sometimes only 1 minute) and in latest news the 5 digit
doesn't come up anymore at all. So probably one of the column drivers went
belly up.
Unfortunately, the next cards in the stack aren't the column drivers and I
can't really see where the wires go. After some time, I unscrew the other side
of the hinge of the card cages and THIS is the side that can flip out.
That makes life easier, now I can beep through every wire. They were so nice
to use standard resistor color codes for the wires for the columns and numbers.
Now I can just follow the yellow cable (5th digit, counting starts with
0=black). And then I wondered... there isn't a column driver card. There's
four of them! Each driving four columns. Well whatever at least I have found
it. Scope says the column select pulse has half the amplitude of the other
signals, that's not enough to strike the tube. Testofon also says this
transistor is a goner. And here's the next problem: The Net says there should
be T6055 and T6056 used. But mine uses T6057. And they're all ungoogleable!
No datasheet, no chance. But you can see where collector, base and emitter
are. And knowing they need to switch 100V, I check what I got there.
Unfortunately I didn't get (or look for) high voltage transistors.
So I take another look at the ACY39... it's rated for 110V... uuuuhhh that's
very near the red line, but I have no choice and... it works great and has the
same pinout.
The machine survived the Classic Computing without incident (all Nixies strike
and I can enter 1234567890, 99999+1=100000 hooray), but no one really helped
me fixing or explaining the leading 9s when subtracting or teaching me how to
multiply or divide. Maybe there's something wrong in the two's complement
circuit (really helpful if you don't have schematics), so my only choice is a
Testofon Safari testing every single transistor, every single diode.
The Testofon has proven inexpendable, without it I would have no chance
(because I absolutely don't wanna test hundreds of transistors with a
multimeter, where I need a minute instead of 3 seconds).
And guess what? I didn't find a single bad transistor and only one diode that
was kinda outta spec. And even better - it's not working anymore at all!
If you enter any other value but 0, it blanks it, so when typing, the display
goes dark from the right. Scoping the display data BUS you can see it keys in
hexadecimal F. If I add any two arbitrary single digit numbers, I get 18.
The explanation is simple, parts are expensive, keep the decoding as simple as
possible. If the 8-line is set, there are only two possibilities left: The 1-
line is set or not. So if the machine gets an F, it sees it as 8 + 1 = 9.
I check the diode matrix on the keyboard. The decimal to binary encoding is
completely passively done with diodes on a board under the keyboard.
These diodes are OK. I check where the bits go and remove the card again for
testing. Helpful that they're stuck in there really tight, every time I remove
them I break something off and had to fix some traces already. And again I
find nothing. The wires on the edge connectors are also tight and not broken.
So I put it in again. And now I can enter powers of two. So 0, 1, 2, 4 and 8.
3, 5, 6, 7 and 9 get me 0xF again. Oh man that's gonna be fun.
You know what? Let's fix the stuck keys first, the problem seems to be the
width, not the heigth (the brass parts are flat). If you file off something on
the left and right, they work again.
And now the big surprise... When I turned it on next time, only the 7 key
would give me 0xF. And just jiggling the cards made  the machine work right
again.
---------------------------------Vectrices------------------------------------
Quick and painless
1) Weird picture. No text, just lines. Weirdly shaped lines and bright spot
in center of picture.
No signal from pin 19 from CIA. Bad CIA.

2) Cracked case, picture tube destroyed.
Put in an alternative tube, no picture. The reason was the Quad Op-Amp that
also does the deflection and sits right next to the XYZ connector.
It now has a picture, but awfully dim. I put another 100k resistor in line
with the brightness pot, now it works fine.
Horrible wobble on the picture until unplayable. After venturing through the
entire vector generator, hooking several stages of the amplifier up to my
oscilloscope it turned out to be the transformer.
I rotated the transformer until the wobble stopped. This Vectrex, together
with a broken IBM 5151 MDA monitor, became the Wrectrex you can watch on
Youtube.

3) like 1), but readable text.
Brightness was set too high.

4) Doesn't turn off. Had to disassemble the power switch and give it a
lubrication

5) Doesn't turn on. Same problem as 4.

6) No sound, no vector chattering.
After long search and guesswork I found the CPU to be defective.
Now I have sound, but no picture. If you crank up the brightness you see a
weak line. Turned out that the XYZ-cable from the logic board to the power
board was bad. Now the image is there but it looks weird. Geometries are
warped. After swapping more ICs, the DAC was found to be guilty.

7) Picture consists of a diagonal line. Text gets rendered onto that line and
is legible for the most part. I'm suspecting the sample&hold circuitry or the
X/Y multiplexer or a short circuit. Hm it's using a 4052 for X/Y multiplexing.
Does it get switching signals? Yes. Swapping the CIA for fun did nothing.
I'm not good enough in reading schematics of complex analog circuitry to be
sure that that chip really multiplexes X/Y. There's only one analog signal
coming out of the 4052. Measuring a known good Vectrex showed two, so I
swapped the 4052 (with one from 1975) and the picture returned to normal.
Need to fix the volume control, but at least I won't need a buzzfix kit since
this is a later revision Vectrex that has the amp on the logic board and
barely buzzes at all.

8) Weak point halfway between center and lower left corner, completely dead.
No idea what that was. The scope says it's stuck in a very small loop 
somewhere in ROM space. The builtin ROM's Chip select is stuck low, but all
the others are stuck high. Checking the data bits I can see a suppressed bit
that is kept low. If I remove the CIA, it's gone. I put the CIA back in and
the machine just works perfectly as if nothing had happened.

9) Still in its original box, not much use, only one game (Pinball).
No picture, no buzzing but plays blind.
After some chip swap and measurement orgies I saw the 13V are missing.
The transformer is missing a rail, there's 10-0-10V AC, the left one was
missing. And suddenly it clicked, buzzed and the picture appeared.
Another one with a sticky power switch.

10) Dead. No picture, no sound (just volume pot scratching). Neck glow.
First I swapped the CIA and that got me vector chattering and sound, but no
picture, it plays blind. If you crank up the brightness, a dot appears halfway
between center and bottom left.
For some reason, instead of an LF423 there are two MC34001.
All ICs are severely corroded, so I remove them and put sockets in.
If you replace the MC34001, there are parts of the image appearing in the
bottom left corner and look really broken. If you replace the MC34004 (with a
TL084), it looks worse. If you replace the DAC, you can almost make out what
the picture is supposed to look like, but still garbage. If I put the old
MC34001 back, there is no difference with the new DAC. The scope says (for the
entire time) that the CD4052 is sucky. But I don't have that one in stock even
though I have CD4051 and 4053...

11) Plays blind, no picture. Same fault as 9)

12) Doesn't turn on. Same fault as 4)
The picture has foldover on the right edge and on first powerup it had severe
foldover on the top but here you could see the capacitors reforming which
tells me the problem already.

Colecovisions:
Here we have two french Colecovisions (the few machines that had SECAM used
SECAM generator boards almost the size of the mainboard so Coleco decided
screw that, let's just output RGB) and a regular PAL one.
1) Said to have no red - everything red is purple, yellow has weird problems,
the problems start at around 5 to 15 minutes.
Couldn't reproduce, works fine on my TVs.

2) Dead. Not outputting any video. Was found in a barn with a leaky roof, the
RF shield turned into a lump of rust. If you turn it off and on again, it
outputs a black screen instead of no video at all. It does clear the video RAM
and the glue logic appears to be fine (even though the chips look like swamp).
Work RAM Write Enable never stays high, it keeps pulsing low.
A healthy Coleco builds up the greetings screen and then enters an endless
NOP loop, so it stops all RAM writes.
(with the scope you could see it actually clear the video RAM)
So the desoldering/testing order will be Work RAM, CPU, logic.
Well I tested the work RAM in our Arcade Museum's PolyPlay and one of the two
2114s was completely dead. Swapped it, done. Works great, generates a picture
the first time turning it on every time, games work, play OK, sound works.

NOTE: The controller's diode swamp schematics you find on the internet has
every single diode in the wrong way, but only those belonging to the keyboard
matrix (you know the diodes that are drawn diagonally).

Interton Video 2000
(has nothing to do with that VCR system where you can turn the tapes over for
another four hours of recording and playback at any speed from -5x to +7x
including slo-mo both directions without any picture disturbances even though
these machines also have vast amounts of logic ICs)
It's a PONG console before the AY-3-8500 hit the scene.
So it should be in the logic graveyard section of this manual.
Enough talk. It gives out a picture, but it's got no balls. Or paddles.
Just the net and sometimes the score. Or are these infinite vertical lines the
paddles? What's the chip tester say? These are 4000s series logic, so the
chip tester has trouble with these but still passes most the ICs on the main
board. Those that fail look good on the scope except for two of the 4069s
where most outputs don't correlate with their inputs. And another one gets
so hot it smells like burning plastic. The outputs look good and the chip
passes though. Turning it off and on again left the chip cold and working.
So I swap the 4069s (simple NOT gates, had to buy them, I was sure I had them
stocked and even then Rainer's father's boxes MUST have had some of these
since they're really common, the CMOS version of the 74LS04!) and that's it,
the paddles and the ball are back, it plays, it counts up the score if you
miss.
Some weird facts about this console: It's ultra rare, even in Germany where it
was made. It uses CMOS chips and uses them partially analog, so it needs just
two 4069 chips for the balls and paddles. Even though the ball position is
purely analog, it has an address counter. It uses cartridges which also
contain lots of 4000s series chips, so it should be possible to create games
that have nothing to do with PONG but still use the internal circuitry even
though there's no CPU. You can quite possibly add one.
Also the previous owner wanted me to make the controllers interface with the
Interton VC 4000 so now I have no controllers for this console.
You can still play through skin contact though. Or bumbling pots.

Tech CreatiVision aka Dick Smith Wizzard
This machine is actually a Colecovision, just with a 6502 CPU instead of a Z80.
Same graphics, same sound, same RAM. Just a 6821 PIA for the inputs.
And apparently no BIOS ROM?
And I got one here that doesn't boot. The CPU appears dead in the water.
The signals going in look good, nothing keeps the CPU from running.
Piggybacking a 6502 from a 1541 floppy doesn't make it work either, but you
can actually get it to fight, so this CPU is mega sus.
And with a new 6502 from a 1541 spares board we're back in business.
Except the left joypad has a few keys unresponsive, eg. the T without which a
PRINT command won't be issueable. The joypads have more keys so you can have
a QWERTY keyboard for programming. Neat idea.
So I disassembled the joypad, found nothing, put it back together and it works.
Well the typical 10 PRINT"COCK FORGE SOURCERER "; 20 GOTO 10 works.
So let's load the demo from tape. Nice stuff with the sound from tape, stereo,
data left, voice right. And the second lesson asks if I wanna start over (Y/N).
And throws a SYNTAX ERROR on INPUT A$. WTF?
INPUT A works though. And string functions work fine too.
So I loaded the rest of the demo, everything works unless you get to an
INPUT x$.
I even programmed workarounds (no I'm not so stupid and erase the original!)
to get through the entire demo, I even put it on VHS, but the owner wants me
to put it on Youtube and I'd rather have a direct grab than a VHS copy.
And now not even INPUT A works. So I desoldered the three ROMs and compared
them to the Internet.
Uhh stupid, it's saved as a single 12k file. And there's four of them so I
have to split them all and find out which is which and what version I got.
Well. And so I found out I got BASIC82 Rev B (there's 82A and B and 83A and B).
And one of the ROMs has *ONE* bit flipped. From 0 to 1. And these aren't Mask
ROMs, they're OTP EPROMs (without the window).
So - I burned over it. And it worked! But putting the chips back in, the cart
didn't wanna work at all, completely dead even though the ROM now matches.
After some swearing I beeped the traces on the cart and it had a bunch of
hairline cracks. And now everything works.

Atari Jaguar
Symptom: Not reacting to controller input - but the machine works!
The joystick signals look OK. Also weird, it seems to use a scheme similar to
the NES, but with a matrix, even though the controller port has 15 pins, well,
whatever.
Against the light, one of the HC244s on the board seems to have a boil over
where the die is. But it doesn't get hot.
Two of the HC244s have to do with controllers. The enable signals are a little
weak, they only go down to 0.4V for Low which is enough but doesn't look clean.
But the one IC that has that faint boil has enable signals only going to 1V.
For TTL, that's deadly. Let's see what does the spec say for HC? 0.9V for 4.5V
VCC. That *might* just work at 5V, but if I take a 500 Ohm resistor and ground
the signal, it goes to... what 2.4V!? So I desolder both enable pins (damn
SMD!) and now the signals look a lot better, with a low of 0.2V and if I hold
a resistor to the desoldered legs of the chip to ground I can get the one to
only 2.4V, the other one sloooowly drops to 0 like I'm charging a big cap.

Another Jaguar
Oh my god it's a parts machine. OK the other one was also tampered with, but
it was alive at least.
Here two big RAMs are missing. And according to the note inside there was lots
of work done, lots of components replaced... by someone who's as good as me
when it comes to SMD soldering. Have I mentioned I suck terribly at SMD?
OK ok my Game Boy Asterix cartridge has a different opinion, but I don't allow
that.
OK in the bag with screws are two RAMs... that have a big X marked on them
with a sharpie and they seem to be the missing ones from the Jaguar.
They look suspiciously like RAMs from old PCI SVGA / early 3D accelerator
cards. Crazy that the Jaguar had this in the early 90s... That is if that's
the same chip. After tons of research I find out they have the same capacity
(wow) and very similar pinout, but there's always one pin on the Jaguar's
memory that the VGA cards don't have - it allows the RAM to be used as one 16
Bit RAM or two 8 Bit RAMs. Where you can read 8 bits while you write the other
8. And none of the more modern VGA card memory can do that. So the Jaguar is
actually MORE advanced than those late 90s video cards!
Unfortunately, the Jaguar actively uses this one extra pin, so replacing it
with video card RAM wouldn't work.

Fairchild Channel F
Sometimes the image looks weird and the right half of the screen gets repeated.
That leads to a stripe of garbage in the middle of the screen, because there's
pixels that never get displayed. Some of them are used to determine the
background color.
It took me a while to locate, even with shorting some outputs. Since the CPU
can't read the video RAM, there's nothing that can happen.
The 7493 work. The glue logic works. The 74153 seem to work too, but one of
them is responsible for 128H. I piggyback it and the problem goes away.
I remove the piggyback and the problem comes back after a few minutes.
One of the pins of that IC looks badly bent. I remove the chip and for one,
the pin was broken and after some lengthy staring at it, the pin fell out of
that IC! The die on the chip works, but the bonding wire makes no contact.
------------------------------Pinball machines--------------------------------
So it has come to this. Now I'm fixing pinball machines too, but from the
late 1970s on, they too have computer boards with logic chips in them.

Here's some interesting facts (to me interesting, you'll be probably bored to
death by them).
There's lots of different 8 bit CPUs, especially in the arcade, where you see
mostly the 6502 and the Z80, but also 680x, and the 8080, exotic ones are just
the Signetics 2650 which is used in many Zaccaria machines (as well as their
pinball machines) as well as some really rare other ones.
Well Model Racing's Super Shot uses the what was it called SC/RM CPU which has
like just a 16k address space and has a cycle per operation ratio like
something handcranked? And then there's Pole Position, the only machine ever
to use Z8002s.
But in pinball machines, you see only 680x and everything else is the
exception, but you basically DO see everything else. Z80s were used by spanish
companies as well as Game Plan. Zaccaria (at a time #3 pinball manufacturer)
uses the 2650 as said, but UFO-X from Playmatic uses two RCA 1802s (one for
sound and that's good, the composer must have been bored, this game has a
very rich sound atmosphere and great chiptune with two AY 8900s) and NSM's
Hot Fire Birds uses a TMS9995 (16 bit) which otherwise just got used on some
prototype home computers (and the Tomy Tutor).
Zaccaria once experimented with the SC/MP. There's a winter sport themed pin
that exists in three versions: mechanical, SC/MP and S2650.

Genesis
Symptom: Jet bumpers don't count. Reason: Invalid logic signal on switch
matrix. Probably a defective pull up resistor Signal gets better if you
deactivate some dip switches in the same row of the matrix. This doesn't let
you set the scoring entirely the way you want to, but at least the game works
again. So now the machine's out of order with switch matrix problems and the
driver board is missing. Whatever was edgy has now completely kicked the
bucket. So let's just measure for fun. After some poking around I measured a
big bunch of bad bits on an LS32, especially the outputs. Let's pick one.
It goes to the cathode of a diode and there is nothing else attached and the
diode checks OK, but the output checks shorted to GND. Putting a cap on made
the signals look a little less broken (still way off like 20% fixed).
Removing the chip made the short go away, also on the chip itself, maybe the
heat, but with the new LS32, all the signals look nice.
Can't really test it though without the driver board.

Robot Wars (another Gottlieb)
The machine was found in a barn (although one with an intact roof) and just
worked fine. Until the flasher next to the spinner lit up, which had a short
circuit into the spinner switch, sending dozens of Volts into the switch
matrix killing three switch matrix ICs (LS32 and two LS00) and even blew
through these, killing the 6532 RIOT as well.

Flash
Symptom: Could be used as a thermometer. Below 15 degrees it wouldn't turn on
at all. 15-16 it would reset permanently. 17 degrees it would reset as soon as
you start a game. 18-19 it would reset randomly through the game. 20 or more
it would work fine, but if it got too warm, would start resetting again.
Cause: I found a huge capacitor in the power supply to get a little warmer
than it should (there's 2 power diodes next to it heating it a little, but not
enough for it to get that warm that fast).
Replaced the cap and the machine is rock solid ever since.

Soccer Kings.
Symptom: Randomly deactivates the flipper paddles during a game and turns them
back on at random intervals. Pinball maintenance staff says some lights don't
flash the way they should. Chased the flipper relais signal through several
components until I found a bus multiplexer. The signals going in are clean,
the signals coming out are clean. The signals going in control 7 or 8 of these
bus-muxes and come from a 2101 (or 02? Same as Galaxian sprite RAM) dual port
RAM. Either that RAM has some senile bits or the bus mux is defective.
I don't have spares for that MUX, but I do have spares for that RAM - and the
self test runs without errors, but a dual port RAM... if the outputs are
connected to a different subsystem than the inputs, the CPU can't even check
this chip. Also, to check the bus muxes, I would have to desolder two and swap
them and see if the error wanders. And it wouldn't be the first time that an
almost 40 year old chip doesn't like the heat of desoldering.
So I try swapping the 2101 RAM first and see there, the problems went away.
I fixed a pinball machine the pinball maintenance experts are chewing their
teeth out on for years! The pinball people took the rest of the day to fix the
machine's other problems which were mechanical or electrical.
The next day the machine was fully functional.

Odin (Peyper, Spain, 1985)
Doesn't work, but after 15 minutes it catches itself and maybe it lets you
play or not.
The quartz works. The RAM that's supposed to be fed by the battery (which had
to be wiped up from the board) gets not enough power, power drops regularly.
A diode had no contact to the trace (everything's corroded), now the RAM gets
+5V from PSU. Let's test the socketed chips. The CPU does weird things in
Galaxian, I'd venture to say it is shot. Swapped it, now the machine starts
doing weird things right away instead of waiting 15 minutes.
So now the ROMs are suspect. The ROM-LS138 I've already scoped, when the CPU
was busy NOPping or doing something weird, all 8 outputs strobed.
The RAM is OK, tested it in Q*Bert, +5V is ok, too.
So I burned new ROMs from the Internet. These don't match the contents of our
ROMs at all. The first 3 bytes match, but that's it. If it were a stuck data
bit or a stuck address bit, I would have caught that, but no. The ROMs show
the same every time you read them out, so no random number generator here.
With the new ROMs, the machine does a little more. Coil test, but it doesn't
react to buttons and doesn't strobe displays.
I want my Galaxian CPU back and we have a crate of spare parts for arcade
machines with like 20 Z80 CPUs in there, even gold/purple ceramic ones, so I
chose upon a period correct (1985) one and the machine sprang to life.
Obvoiusly, the CPU gets clocket at 4MHz, which the 1979 Galaxian Z80 couldn't
perform stable at. Well the Lamps on the playfield now blink happy patterns.
The old ROMs still do nothing. It still doesn't control the LEDs, oh wait it
does, but very slowly, sporadically and randomly. It reacts to the start
and credit button and counts credits, so I checked the clock signal to the
data/sound board and the trace was corroded away, let's see if more traces are
bad, nothing found on the quick. So turned it on and hooray... not completely,
but further along the way. Some LEDs show numbers (credits display) and the
machine plays a melody on turn on, credit and start, but still gives no ball,
doesn't control coils, doesn't count points. After intense searching for bad
traces I've found a few and fixed them, but no difference. Maybe the fault is
on another board. The flipper relais doesn't engage.
According to the schematics, the coil driver controller IC is wired 
differently than the one on the board in front of me, ie. it sits at a
different memory segment. Since other people found more differences in the
schematics as opposed to the machine, the suspicion arised we have a prototype
machine. That would explain the different ROMs, maybe ours was compiled in
DEBUG mode.
Out of desparation I tried putting the original ROMs in, after reading them
out with a different eprommer yielded the same result as the first one and...
it does watchdog a few times after the coil test, but the machine plays a
slightly different tune (which in my opinion sounds better, but they both suck,
the composer should've got lessons from Playmatic's composer, same sound synth
hardware) and goes to attract mode. The lamps now blink in random patterns
instead in counting patterns. And he gives the ball when you start a game.
And he counts points (there were 2 or 3 shorts in the switch matrix), and even 
hits the coils (bumpers, slingshots, drop target reset). Two of the bumpers
had their wires swapped (WTF!?) and the right flipper finger stays up when you
press the button just once. When you try to move it down (both stick
mechanically), it buzzes and if you don't let go the fuse burns.
So... the coil driver controller sits at another address, the ROM is OK even
though it has a COMPLETELY different contents and the display board has a
trace that was scratched away since the drawing person overshot.
Now I'm not just fixing rare pinball machines, I'm fixing prototypes of rare
pinball machines. That sure looks good in my CV. Certainly better than
"Bricked the Voyager probes". (I have an idea for them, when the power fails
and only the computer can be powered, to prevent them from becoming totally
useless, let them mine bitcoin)
After 6 hours of investigating and drawing the entire flipper finger schematic
with nothing but my Testofone and 8 coil drivers for 10 coils, I solve:
The flipper fingers are only controlled by the coil relais and the buttons, no
software here. The contact finger of the right flipper button was too long and
when the playfield was lowered, it bent the contact so it wouldn't break after
a press. The machine is now fixed except for lots of mechanical problems and
the displays still not working.
The Extra Ball/Extra Game displays were just put in backwards, now they work
too, but the scores just read nothing. They flash for a split second when you
score some points, but most the time they are blank.
After 6 hours on the schematics I found nothing. The displays get controlled
by a display/keyboard controller chip. No ROM in there, but still not easy to
obtain. But the switches work, so what commands does this chip get?
The CPU writes the current score to the chip and forgets about it while the
chip cares about strobing the displays to display that score. There's even a
facility to read the score back so the CPU could save RAM if programmed
appropriately. So I studied that chip's datasheet and built a Latch-O-Matic to
show the commands to the chip, then record that with an old iconoscope camera
but I went back to the museum and first tried the dip switches that weren't
labelled and one of them wasn't properly set to off, even though it was
definitely not set to on. It clicked audibly when moving it all the way to off
so that's a clear sign and the displays now work fine.
And the pinball people took just 4 more hours to fix the mechanics.
One year later the pinball people decided to put the mechanics in a good-as-
new state. A week later, the machine died. Wouldn't boot at all.
All I get on the Scope is different endless loops on every power up.
It never accesses the RAM at all (not even reading), no IORQ, my fancy new
In Circuit Chip Tester/Debugger says there's something fishy with that LS161
(which is corroded from the battery acid). Testing it reveals it outputs the
same frequency on every pin. I removed it, it checked out OK. The schematics
show that the first output is tied to RESET so the outputs make sense.
The only other thing I found was that the right ROM gets extremely hot
(well you can't burn your finger but if you put it on your lip you'll burn it)
and the left ROM which is identical barely gets warm.
The ROM even gets hot when not in a circuit. Good thing I dumped it a year ago
so I'll just burn a copy and try this.
Verified the ROM against its older self and there's a few errors in there, two
data lines are stuck high and at some points there was different data than
what was stored, maybe bad address lines?).
Even with a new ROM it didn't work, but noted that the watchdog doesn't work
either, so jiggled around some plugs and it started watchdogging.
Turning the machine off and on again brought it back to life.

Foxy Lady (Game Plan, 1978).
Not one of the Big 3 (or 4) and it uses a CPU nonetheless. Not much difference
to Odin (both use a Z80, some ROMs and some RAMs). They both had the same
battery problem, but this board got chewed even worse.
The owner already removed all components near the battery and put sockets for
the glue logic in. I tested the components I could, the Z80 and most the glue
logic is OK. 
Reset circuit doesn't work. Swapped the transistors (for BC548/558), now
instead of fixed low, Reset is fixed high.
The ROMs are OK but hard to read since they're wired in parallel and have
internally masked Chip Selects.
We have a Pinball Lizard which has almost the same CPU. Here I could test the
rest of the chips (sadly killing Pinball Lizard's High scores), but they were
all OK.
So I read another repair plan. The game board needs 5V and 12V just so it can
blink only once on the bench. The reset circuit needs 12V but still doesn't
work on the bench. Wait a minute... did I replace NPN transistors with PNPs?
*facepalm* now Reset works. One problem less. But still the CPU doesn't work.
Also, lots of resistors are missing, the previous owner only preserved the
ICs. So the LED's resistor as well as a pullup were missing, but still nothing,
but the LED now lights up when you remove the 8255.
The 8255 doesn't get Chip Select. The Z80 must dispatch an IO Request for that.
So the ROM. Burnt a 2732 with both images on top in the correct order, nothing.
If I remove the CS pin from the 8255 (defaults to LOW), the LED sometimes
blinks once, but not reproducible.
So I despair and ring through ALL CPU-side traces with a LIGHT BULB.
And since I'm so desparate, I even fix 3 broken traces that went NOWHERE.
The only other thing I found was that there was a 100k Resistor in the parts
plan where the schematic diagram called for a 4.7k one so I swapped with the
4.7k one. Also I resoldered the diodes and dip switches.
Then I populated the CPU side components. The board is cleanly separated into
CPU-side and 8255 PIO side. And see there it blinks. And when I touch pin 3
with 12V, it blinks 5 more times. And it can be reproduced.
So I populate all the other parts and still it works!
The current owner of the machine has already fixed all the mechanics (except
for the roulette wheel which rarely gives out the advertised winnings) so when
I put the board back in, the machine was playable.

Fabis boss bought two more Game Plan MPU boards, excellent condition, no acid
damage... wow do they lie on eBay... or was that sarcasm. One board was still
wet with battery acid, the other one has already dried.
And the RAM chips just fall out of their sockets if you flip the board over.
And there's only a quarter of the pins remaining and these are green.
And the ROMs are missing. Excellent condition my ass!

Spac Train Mac (another spanish machine)
More like Space Train Wreck.
Doesn't do much - well attract mode works, blinks lights in patterns but does
nothing else. We have another CPU board which is completely dead.
Great thing both boards have ALL chips socketed.
So replacement board: Battery SRAM dead (and I don't mean the battery),
Keyboard/Display controller dead, LS04 clock generator dead, one of the two
ROMs is dead, too and the other guy who tried to fix the board changed a dead
7805 for a 3.3V one. After replacing all these, the spare board now behaves
the same as the original one.
Biggest problem so far: The ROMs don't match. Both boards have one ROM 
slightly different and the other one I can't compare since one of them is dead
and the Internet only has one 16k image whereas this machine sports two 32k
ROMs which are nearly completely used.
Two of the four display strobe lines from the display controller are garbage.
Removing the display board and the machine comes to life. If you try to
start a game a voice with a strong spanish accent says "No credits", you can
coin it up, it even starts a game but if you hit certain targets or close
certain switches, the game will reset. And the error is nowhere in the switch
matrix (and there are two independant switch matrix circuits all on the same
bus so the problem is elsewhere).
After lots of checking I found that the problem exists on the F-rail of the
switch matrix (strobes go from D-H, data returns go from 0 to 6).
You can even get credits by rapidly pushing a playfield switch in the F-rail,
except 0F and 1F, they don't reset the machine.
After some time with the oscilloscope it was clear that the LS42 which strobes
the ... strobes was bad. Replacing it had no effect though, but I could see
there's something fishy going on on another rail. There was a shorted switch,
but that didn't do anything either.
When studying the 8279's pinout I noticed that the data return lines are
inverted. Someone's put the plug on the wrong way.
Sadly, the machine now doesn't even get that far. In attract mode, it keeps
crashing after starting to light half the playfield's bulbs.
It appears to crash always at or near the same point.
With the other ROM, the light patterns are inverted, the lights start top-to-
bottom instead of bottom-to-top, but the machine still crashes when it reaches
the middle.
Swapping driver boards or main boards doesn't fix it, so there probably is a
short in the lamp matrix now.
Sadly, that appears to not be the case. It also crashes if you connect nothing
to the CPU board but the speakers. The Display/Keyboard controller also works
OK in Odin (which a few weeks later died... No I'm quite sure that's not the
reason since it first strobes the coils and then initializes the displays).
I give up.
After Odin's ROM going bad I tried reading Space Train's ROMs back. No errors.
Now it's Game Over. Only way to fix it is trying to recreate the program on an
Arduino or Raspberry Pi.
Hm not so fast, if you put the ball in one of the saucers, it beeps much
earlier. I bet the solenoid voltages are missing.
And yes, they are. The fuse blew. 5A for solenoids? I'd used 10.
Now the attract mode just keeps on running forever.
Let's start a game. It actually gives out the ball! And makes music.
But nothing happens, doesn't detect targets, pop bumpers don't pop (flipper
fingers work though) but after a while it plays the reset melody.
OK the playfield switches are falling apart. If I push the one at the end of
the start ramp manually, the tinny sparse background music unfolds to a
chiptune opera in stereo, but still, it tries to give you four more balls and
then plays the reset melody (which is almost certainly the Game Over melody).
Well we're back at the beginning, but it's a huge step ahead.
Just replace ALL the playfield switches and see what's happening.
One note: The reset melody plays from the right speaker. The error/game over
melody plays from the left, but besides that, they're identical.
I made a replica display board using the little reverse-engineering skills
I had. Turned out both custom display controller chips are fried (one is
cracked), so from the schematics it is quite clear that a 74LS48 would do the
job, but finding out when to blank the lines will be a little harder.
I found one line to be active most the time, so I hooked up the LS48's enable
bit and now you can barely see the "garbage" that goes to the lamps behind it,
which makes what the display should really show nice and easy to read.
But two out of the three enable/multiplex/I have no clue what they do but the 
custom chips need them-signals were bad (as in voltage). I found out that the
CD4514 on both display boards looked OK but was cooked anyway. So I guess that
whatever has melted the chips on the display board, worked itself through the
display multiplex IC and back into the AY-3-8910. The board is really tiny and
they used the AY-3-8910 100% with all their PIO lines.
So replaced the AY, noting that the other still produces sound and will be
just fine in a machine that doesn't need its PIOs.
Now I have to find out which combination of the three signals makes the lamps
(I used LEDs of course, so my replica board needs only 5V) show sensible data.
After I swapped the AY, the lines looked OK, but the attract mode freezes
again and it beeps the high-low-high-low tone. But now there's something on
the display. It's complete garbage, but from the higher-than-9 you can see the
characters in the spots can only really spell "Error2".
And the manual says, Error2 means no coil voltage. Again?
But the fuse is there, the 40V are there, and all components from the 40V test
circuit are OK. Yet I couldn't get it to test OK. After a couple of hours over
a couple of weeks I gave up and shorted the pin to ground. Now it thinks it
has coil voltage even if it doesn't. Let's see if the AY fixed the game.
Nope. Still slamtilting when you give a credit on the left coin chute.
Then I had an idea: Maybe the two working ROMs aren't compatible and maybe
I mixed them up. Swapped them and now the machine keeps saying "Try again",
even in attract mode, and at completely random intervals.
If I remove the little coin box board, it stops.
If I start a game and then quickly remove the coin box board, the machine
now plays. For real this time! It gives one ball and one ball only (unless
you lose it), it reacts to contacts, pop bumpers work, the two saucers work,
only the slingshots don't. I realised almost too late that the right slingshot
was stuck on, it already started smoking when I turned it off, but the coil
has survived. I guess that the switches in both slingshots are bad.
I hope that they are. Because if yes, the machine will be working fine after
the coin box board has been fixed and the playfield switches have been
reworked.
Testing the coin box board showed it's working flawlessly.
And then I noticed, TILT! The tilt pendulum was misadjusted and was leaning to
the contacts. So this thing gave sporadic resets. The ROM I had in first does
the same as slam tilt, going to Game Over as soon as you tilt, the other ROM
does the "Try again" thing.
The right slingshot's contacts were adjusted too tightly but that didn't fix
the locked-on fault and the left slingshot counts points and makes sounds.
Both slingshots get controlled by one AY-3-8910 sound chip. Swapping it made
all the solenoids (including slingshots) work fine and brought the machine to
99% working.
The only thing left to do was reverse engineer the lamp driver circuit on the
display board. And this is where I got stuck last time: ST0 goes to the RIGHT
half of the display board and not to the LEFT. That's why I got stable signals
that lit the wrong LEDs.
So this time (and knowing that using TWO LS154 is beyond overkill, especially
since they're becoming rare) I'm using an LS138 whose output (pin 10 in my
case, depending on which order you hook up the control signals in) ORed with
ST1 goes into the LEFT LS273 and the LS138s output ORed with ST0 goes into the
RIGHT LS273 and that's it! The lamps now make sense, you can see which player
is up and if he's rolled the counter over, so hooray!
Info if you're stuck in the same situation:
There's DB, EL and ED coming from the mainboard (and RESET in a four line
bundle) which are immediately next to the power supply lines on the display
board ribbon cable. These you hook up to the three inputs of an LS138 which is
wired to be always active. Then you take an LS32, hook two of its outputs to
each of the LS273s clock input and one of each of its inputs to ST0 and ST1
respectively (note again ST1 goes left and ST0 goes right which is opposite as
you'd expect). Now you hook a wire to both remaining inputs of the two gates
you decided on using. This wire you touch with each output of the LS138.
You will almost immediately see the lamps making sense when you got the right
one. (Note if you turn on the machine without starting a game, CREDITS,
GAME OVER, TILT (turns off as soon as you start a game) and MATCH light up
constantly and when showing high score, the x10 lamp on player 1-4 flashes)
And what's really cool is that now Space Train Mac is finally 100% working.
And what's really cool is that for fixing it, I get another Space Train Mac
(the owner has two - correction: the owner HAD two) for free! It's mine now!
Isn't that pure awesome? And by the way: Space Train Mac is a really fun
machine. Weird, but fun to play. Let's see if it's better than the Blackout
I bought. All I need to do to finally make this machine mine is fixing the
second CPU board. I already stole some hopefully 100% working AY chips by
swapping the defective AY chips with the slave AY chips from two Scramble
boards (they don't use a single line of I/O so no one will notice, and all 3
voices work fine), all I need now is cloning the entire ROM set and order a
DSKY (i8279). And that should be it!
GRR! It only lasted two weeks, at least long enough to completely pass the
test for the Pinball Competition finals with no errors (!)
Now it does nothing. The displays get strobed with all zeroes, but doesn't
react to anything and doesn't do the reset beep.
If you remove the 12V bridge short on the display board plug, it starts
running, however without 12V it can't engage the Game relais and can't amplify
sound. Weird. 
If you plug it back in, it just stops. And if you unplug it, it just continues 
where it stopped. More weird.
And if you let it run for a while, it stops and the amplifier
starts to buzz, which is weird since without +12V it shouldn't be able to do
anything. Measuring the 7812, I get 7V on the input! Even if I unplug the lamp
driver board (the only other 12V source). Even morer weirderer.
If I measure diodes, I get even weirder values. Sadly I forgot my Testofon at
work. But looking at the multimeter shows there is at least one diode leaking
in the wrong direction, making the 9VAC go places it shouldn't. And then 
there are these weird test circuits.
Trying the other CPU board, the 12V immediately die shortly after turning it
on. Regulator or capacitor.
So. The Testofone also beeps weirdly. One diode measures 0.8V if I try to send
power through it the wrong way. Should be 0. The other diode next to it
sometimes measures 7V, but only sometimes. I swapped all the diodes to no
avail, I was breeding on it for all saturday. The 50Hz detection circuit 
doesn't go low enough to trigger the ICs generating the timer interrupts.
After having ghosts in the machine as the last reason standing, I removed one
leg of each component on that rail, soldered them to a wire replacing that
rail and the machine works! So now it can only be leaked electrolyte from some
capacitor!
And concerning the other board... there's a Thyristor in the 12V circuit 
instead of a diode. And it turns off for whatever reason. So I replaced it
with a diode and... it works. Somewhat. Displays get strobed, lamps blink,
I can give credits, it beeps a little, but no Game relais. And the amplifier
distorts. At least there's no ERROR2. But I bet that's the electrolyte.
So back to board 1, I had that cap out just a few weeks ago and it was OK
(16V 470u). Now it has indeed leaked and yes, the trace directly under the cap
is the exact one I replaced with an air wire.
And not only does that rail work again, I don't need to fake the coil voltage
test to prevent getting the ERROR2 message!
And my CPU board now works fine after replacing that thyristor with a diode
and finding out that the AY chip controlling it was not 100% working and also
the transistor driving the Game relais was bad.
My machine still has no Speech, because one of the two AY chips has one or two
outputs bad and it happens to be the one triggering the speech hardware.
Now all I need to do is duplicate the ROMs, buy a DSKY and find a 100% working
AY 8910 and I got my own working Space train Mac. Oh wait I need to build
another display module.
Uhh the machine broke again. Oh wait, it's just a broken plastic on a target.
So I bought a DSKY from old soviet military surplus for cheap and found a real
DSKY in my attic a few days later. Duuh...
Well the military chip isn't quite compatible. Or broken.
It only strobes the digits that have recently changed. (it appears to control
the strobe outputs differently, since the digits appear on both sides even if
one doesn't change). So if you turn on the machine, only the credits display
shows anything, the rest are dark.
After you start to play it, one digit after the other starts showing up until
eventually all the digits are there. But you can't be sure and sometimes 
digits turn off again. I guess this chip doesn't like the "replace score with
lamp controls every now and then" command.
At least the attic find works perfectly fine. I should test if Odin works OK
with the soviet clone. Btw. keyboard matrix works just fine on that chip.
After a rather long time of working there was a short time where speech wasn't
working for a while. No idea why, but a few hours later it came back. And a
few weeks later, it was dead except for GI. A 10A fuse has blown. 
Someone told me, the machine did weird things with solenoids and crashed.
Indeed a few hours later it only showed garbage on the displays.
I prodded the CPU board and it came back to life.
And here's why I had so much trouble with the machine:
The beedoobeedoo beep is used for: Startup-Sound, no coil voltage, coin
rejected, "you need two coins to play", TILT (the other game ROM), Slam Tilt,
Game Over. Just like the "Generic car fault" light that Microsoft Car comes
with if Microsoft built cars. Or if you ask a woman what's wrong and she says
"nothing" and of course means "everything".
And as I said, the tilt pendulum making sporadic contacts didn't help.

Cool thing: The display board that didn't melt down, works. Well partially.
The PALs are broken. And the CD4514 is completely dead, not only that, it even
messes up its inputs.
For some weird reason you can only get CD4514 as SMD in my country, so
swallowed the toad and soldered it to some wires. You can't change it if it
fries itself, but for now it works. And I got a nice book from my part time
job when studying. GAL Programmierung by Franzis Verlag. So the goal of
getting a complete reproduction PAL is within reach!
The most important thing was getting the LED enable signals error free.
Which happens when you take pin 12 from the LS138 and invert it.
Now the LEDs are perfect, but dimmer still.
So I sat down and filled in the KV diagrams with the seven segment decoder
portion (E, r and o are conveniently mapped to 10, 11 and 12).
Then I combined the product terms derived from the KV diagrams with the enable
logic and added the lamp strobe product term.
Now comes the easy part of finding a program that generates a JEDEC file from
that. Or so I thought. They don't exist! Not even the program the book keeps
talking about. So apparently no one ever posted any of these programs online
(and the program from the book was for Atari ST, so they sure existed for
Amiga and PC and very likely MAC, too, likely even Apple 2)
So since no one ever uploaded any of these, they might never have existed.
Mr. Tryschuk (my former boss at that part time job) also didn't have any
program, just a black box of proprietary hardware that takes the product terms
and spits out a finished PAL already (and probably read protects it, too).
The book has the complete datasheet of the internals of PALs and GALs and also
shows the entire way from product terms to fuse matrix to JEDEC file even with
all these special modes of Operation I don't need like register outputs,
redefining outputs as inputs...
I even once tried using CD4511 for the display decoders, but couldn't get them
to work at all with any combination of enable signals.
So I went and scanned the fuse matrix schematic from the book and filled in
all the dots with MS Paint, made some errors doing it, then hand-compiled a
JEDEC file from that and made some errors doing it.
And it still didn't work. After everything I tried I couldn't get the displays
to show anything but 8s, I thought I could at least get the lamps to work.
My biggest mistake was that the LS154 is low active while the CD4514 is high
active. But inverting the output wasn't enough, I had to invert the input as
well. So the lamps work now. But no matter if I invert the lamp signals or not
I always get 8s. So remembering the UDN6118 weren't quite fresh anymore and
testing them I did have segments that stuck on I put new ones into the segment
control (for the price of like 20 Euros for 2 chips or the likes) and lo and
behold I see numbers that make sense. Sort of.
0 to 3 are OK. 4 gets rendered as H. 8 and 6 are both 6. 5 and 9 are swapped.
The "o" from Error is complete and utter garbage. The blank character has the
A segment set.
The latter was an error in my numbering of the KV diagram. Luckily I only had
to change one product term by adding !D. Close examination of the fuse matrix
and JEDEC file of the other defective segments show three errors (or slips of
the eye), and now I'm excited whether it works.
And it works right away. Even Error2.
It worked!
For completeness reasons I now have to reverse-engineer the orange chip select
PAL from the CPU board as well. So I made an adapter to be able to read it out
like a 2k EPROM. I also wrote a program in TestBasic for the In Circuit Chip
Tester that should give me the ranges where each output pin is Low.
Well I took some notes, but the ranges I got were... there's definitely bugs
in my code.
I dumped the PAL with the EPROM programmer and wrote a bitslicer program with
which I could determine the product terms really quick. Took like 2-3 hours,
not accounting for the time it took to fix the 386 with the Eprom programmer
card. (and writing the synthesis program that synthesized the EPROM dump from
the derived product terms as a sanity check. It did catch an error on the last
pin).
And the PAL I made worked just fine in the machine! So I put the old one back
in and the machine is dead. Jiggled it around, reseated it, no dice.
Put my reverse-engineered version back in and it works.
That chip has picked exactly the right time to die. Now the machine runs on
100% reverse-engineered PALs.

Last Lap (Playmatic)
Someone got the message that some Simon fixes rare exotic pinball machines and
brought us another Last Lap. We have one already and I can say this game is a
blowout. Boring as hell. Too bad it uses the entire RCA 1801 chipset (not just
the CPU) so it'll be one of the few pinball machines that will work after a
nuclear war (the RCA Studio II being the only video game console to also do
so, hooray for that, it's always among the top 3 worst video game consoles).
So OK what does it do? Well what doesn't it do? It doesn't boot.
One shorted capacitor was already swapped, no difference. After swapping the
chips with the working one (with no results) and half an hour of measuring
around, I found that one of the two RAM chips doesn't get a supply voltage.
According to the schematics, a big capacitor is charged through a diode, so
the power will stay on after you turn the machine off. Sorry to say, but with
all the battery backup solutions all the other pinball machine manufacturers
came up with, only Playmatic got it right. If you're lucky, it'll work
forever and even in the worst case of all the electrolytic isn't as 
destructive as that of a battery. We had the second to worst case where the
cap was just shorted. Replaced it and the CPU booted.
(the tradeoff of using a 2200 microfarad capacitor is that you'll have to turn
on the machine at least once a month - but seriously - back then they made
pinball machines to last a year or two, so a cap would be perfectly sensible)
OK it boots, but only one score display, the credits and match displays work
which is good because it uses these to display the boot progress.
Btw. gotta mention that Playmatic incorporated a super simple logic probe into
their early solid state pinballs. Not entirely useful, but you can spot really
simple problems like stuck switches or chips that have no output real quick.
And you don't even have to break out the oscilloscope.
OK we have three broken displays. Each had a different fault:
One had a cracked circuit board with broken traces, another has a faulty plug
and the last one had one or two of the 4511 bcd decoders bad.
Next problem: Sound has lots of hum, there are massive amounts of ripple on
the main filter capacitor, especially compared to the working one.
Since the mainboard isn't to easy to get to, I'll just recap the entier board,
no big deal. When measuring the finished product, I slip and short a fuse to
ground. It blew, but *somehow* also kills the 11V zener diode and adjacent
voltage regulator transistor. Since I didn't find the spot in the schematics,
I measured the other machine and guessed the diode to be 11V.
Much later, after having ordered a replacement, I found the spot in the
schematics and the diode was in fact 11V. Without it, I just put in a 7809
voltage regulator (since the big transistor driving the even bigger transistor
was dead as well) so the machine would work again and I could continue working
on it. The playfield had lots of misaligned and thus closed leaf switches,
which confused the CPU during game play.
I then stuck the replacement diode and transistor inside and it worked.

Big Town (Playmatic)
The CPU board is identical to Last Lap. Down to the last bit in the ROM.
The caps are OK. The machine boots, but self test hangs at 5, indicating a
communication problem with the periphery.
What an odyssey. Turns out the In Circuit Debugger is worthless for 4000s CMOS
chips as long as they're in circuit. Out of circuit works though.
Tried socketing the 4028 first, nope, nothing.
Noticed that if you press a button in the coin door, the GI gets darker.
Compared it to the wiring diagramm, no contact. Then I saw that a plug with
two contacts spans tightly across another big plug. Swapped the two big plugs,
re-did the measurement and sure enough it's OK now.
Now booting stops right after 6. The digits turn off, but so does the entire
CPU. It just stops.
The B Switch Matrix of Last Lap has a 50Hz wobble on it which appears to
generate the CPU's Timer Interrupt signal. The wobble on Big Town is so weak
that it doesn't trigger the transistor. And this dead-end cost me three hours
because I so really wanted to know where the wobble came from and it comes
from exactly nowhere. Probably induced from being too close to a transformer.
I was just going to cut open Last Lap's wires in stages to see which one adds
the 50Hz when Floh (another Pinball repair guy) showed me a fault search
diagram that had the 50Hz interrupt circuitry. And it showed clearly that it
takes each of the 50 Hz half waves and puts them in a flipflop made of two
gates of a 4001. Measured it, only one half wave there. Looking for the cause
I see that the power plug wasn't plugged in correctly. The machine now runs.
Couple months later it's been completely restored mechanically and playable.
Ignoring a few blown lamp drivers, it does strange things. Like giving a ball
and instantly going to Game Over or Ball 2 or 3, so per game you get between
0 and 1 balls. It doesn't do that when cold. Also it doesn't save high scores,
not even with a brand new cap, voltage drops from 10 to 2 Volts within a few
seconds. Also if you reset the highscore, it goes to 500000, but after a few
seconds it says 502000. This plus the quick sagging of the backup voltage
suggests that there's something wrong with the RAM. Sadly these chips were
made out of solid unobtainium (let's say they weren't popular back when they
were new, also they aren't compatible to anything - 10V CMOS levels)
NOTE TO SELF: Donkey Kong uses funky RAMs that require level shifters.
And there's a 4000 series IC that can do TTL<->CMOS (just not both at a time?)
These pins use CDP1824 which have 8 bits * 32, so not much in terms of memory.
But these chips also run at 5V so I can test them if I succeed to write a test
program for my chiptester.
(I tried writing one for the 6810, but it doesn't work because I can't get the
tester to stop outputting the data when it's supposed to read it back from the
IC)

Pinball Champ 82 (Zaccaria)
Too bad, this machine worked great all the time and I really love it, it's one
of the greatest machines I've played (I'd put it on my #3 spot after
Monster Bash (Williams 1996) and Triple Action (Williams, 1973)).
Now it doesn't drive the Game relay, leaving the playfield coils unenergized.
That's exactly the spot I was already measuring at the Soccer Kings machine.
But this machine works fine except for no power on the coils. Switches 
register, lamps light up, so no brain damage. The Game signal even goes to the
transistor that switches on the transistor. Only the transistor has nothing to
work with. I guess that there's plus 12V, then the coil relay, then this
transistor, then ground. When the transistor gets its signal, it turns into a
short circuit, pulling the relay to ground, energizing it. But there is no
+12V. So I look at the power supply board. Big big dry joint on one of the
main rectifiers. Anyone could have fixed that machine but since that Simon guy
fixes rare exotic pinball machines of course I had to do it.

Middle Earth (Atari)
I bought a pinball machine. For 1000 Euros. Together with two rather rare
arcade machines and even got another cabaret arcade and a one-armed-bandit for
free. The cabaret had everything imaginable broken (see Scramble Formation)
but easy to repair.
It's in a beautiful condition, the playfield color is mostly intact, not bad
for >73000 games. The rubber belts were in bad shape though, some were
ripped and they all looked horrible.
OK that circuit board does NOTHING.
What CPU is this? An Atari custom. That can't be. It's gotta be a 6502 or a
6800, they never used anything else until the 1980s (the Quiz game doesn't
count). Checked the pinouts, gotta be a 6800. After long studies I could
exclude 680x with x being anything else but 0. And then the Internet told me
that Atari did use nothing but 6800 CPUs in their pinballs.
So there's no clock to the CPU. The clock circuit is a really funky thing,
they were really smoking some funky stuff back then:
Quartz->7400->7493->9602->7474->7414->CPU.
Why they used a '00 to drive the quartz, I don't know but that's not the first
time I've seen that configuration. The '93 divides the signal by four to get
1 MHz. The 9602 is a pulse shaper and unobtainium chip, you can't even get a
datasheet for it (there is one for 9600, 01, 02 but it doesn't even state that
the 02 has TWO monostable multivibrators, so the Atari Schematics is actually
the most complete documentation on that chip). OK dual monostable multivibrator
and looking at the pin description, the good ol' LS123 comes to mind, but the
pinout is different. Built an adapter (there's a difference: Input A needs to
go to LOW instead of HIGH) and it does output a signal. But it looks like a
regular square wave with a 50% duty cycle (from the circuit I was really
expecting something different), but at least it does something.
The 7474 outputs a clean signal on Q, !Q hovever is stuck high, a piggyback
fights, so I swapped the chip. It was working fine, though.
The '14 it drove was bad, had high impedance inputs.
Well I have this fancy new old In Circuit Chip Tester/Debugger here, so let's
test all known chips on the board. Some have false negatives, but in circuit
that has to be expected, the manual even has a section on how to spot clear
false negatives in circuit. It showed me that that '14 was beyond useless no
matter how often I test it (many false negatives sometimes pass if tested in
a loop). The chip tester only found another '14 with only one bad output gate
(sound enable) sparing me the hours it would have taken to find this (see
Video Pinball).
But even with a clock signal, the BUS is like died out. Reset is stuck low.
If I push Reset, it goes high for a little while, the BUS has signals whirring
around and then Reset goes Low and stays there forever.
Every time. WTF?
At least something's happening.
Wake-up reset is stuck high (part of the watchdog, part of the power fail
mechanic that was incorporated in later Middle Earth machines but not
obvious in the schematics). Can't get it low. It comes from a 9301 which is a
7442 with different pinout. I even tried an LS42 deadbug style but it showed
the same result.
OK screw it, I'm gonna tie Wake Up Reset to ground and now the machine is
running, strobing the displays and the switches. The lamp strobes look like
garbage though. The BUS shows an endless loop that is different on each power
up. Someone on the Internet complained about bad RAMs so I had to test them.
My test gear can't do 2111 RAMs, but it can do 2112 RAMs. The only difference
is that the 2111 has two extra chip selects, so I tie them up and hook them to
the regular chip select and put'em in my tester. It said the first chip was
completely dead. Of course, my adaptation method could be bad, but the second
RAM chip failed differently. Turned on the scope to see what I'm testing and
well, the first chip is just dead, on the second chip one of the four bits was
just as dead and the other three just sometimes pulsed part of whatever the
tester had written into them, so this chip is horribly senile.
Try getting 2111 RAMs... time to make another adapter.
But to what? 5101 would be the most fitting choice, one of the chip selects
has to be inverted. But 5101 are f*$#%ing rare as well. 2114 are obtainable
and 6116 are ubiquitous, having been manufactured until 2003 I've learned.
But anyway, 2114 we have more of and it's not as big a waste of RAM space, so
I settle on that. I just diode-ORred the two extra chip selects together and
omitted Output Enable (since on Atari, /OE is just //WE and the 2114 has this
exact behaviour built in).
With the new RAM in place, the BUS looks slightly different - now I see an
endless loop with some counting going on, but it still doesn't look good.
I dumped the ROMs and compared them to the Internet, but my revision is not to
be found on the Internet. There are a few similarities like you could
basically see service routine entries on the same addresses, but most of the
chip's contents differed. The other ROM however had two address lines stuck,
so I can't successfully dump it. This is also the reason why the machine
doesn't work. I didn't mention that if you leave it running for about a 
minute, it crashes. And it does that with great repeatability.
OK I can't get a copy of this version of the ROM set, so I burnt the set I
found. And now the lamp strobe signals look like the standard light show you
see on every solid state pinball machine. And it doesn't crash after a minute.
Let's see if it plays. Anticipating the weekend... I even stocked up on
74LS145 for the switches (which are just open collector LS42) just in case...
Putting it into the machines, but... all the switches worked (ALL of them),
all the coils worked (all eight of them) and only three of the lamps didn't
work and one of them came on after tapping on it so since the lamps are on a
matrix, I can be sure that there is not a single defective IC on this board
anymore. Yes, sound works. Sounds a lot better than Video Pinball.
The score display has bad bad contacts though. It won't be able to play a
tournament, since you can bet that segments will randomly come off and on,
even entire columns sometimes.
Well the machine shares the same problem as Video Pinball: On open saturdays
or when the hut is full it crashes. Or doesn't even boot.
The watchdog doesn't work either, gets a permanent reset. From the Power Fail
circuitry. So as long as there's a Power Fail, the watchdog doesn't work.
When the signal comes, it holds RESET low so that the watchdog can't work.
Arcade Meeting. Hut isn't full by far, all machines are running and my Middle
Earth crashed again. Finally I can, although only with the pocket scope,
measure the machine while it's running (or having crashed).
And what do I measure? Nothing. There's nothing oscillating. What's my
homebrew pulse shaper replacement doing? It's doing good. And the flipflop
below it? Jup, no output. But input. Swapped it, machine works now. Forever.
Half an hour later it's broken again, but hasn't crashed. It just doesn't
turn on any solenoids. Except the coin door.
Next week it worked again. Turned it on, went to Solenoid test and again only
the coin door worked but during the second loop three more coils turned on and
next loop they were all back. But one line of lamps died. Jiggling the plug
brought them back, but they turn off again as soon as anything moves.
Gotta clean the contacts. The machine works fine except for these lights, but
they're all GI, so I don't care enough. And after about half a year they came
on and stayed on so I consider that fixed.
But the bottom left flipper finger is sluggish. The upper left one is fine
though, so it must be the solenoid itself since they share their wiring and
power transistor.
Fabi fixed it - someone tried to oil it. The oil mixed with the old dust and
what's shaving off the coil's plastic insert and jammed it a little.
Uhh Floh made more than 130000 points. My best is something around 115k...
OK gotta correct that. I managed 130000 lately, but just barely.
And after many years the machine started giving out 5x the scores. Just when
they used the machine for a tournament. But no one touched the settings.
And self test doesn't work!
Self test is hooked up to switchbank 1 where scoring is set, too.
So the LS145 gotta be defective. And I can't find mine even though I ordered a
bunch of them. And the In Circuit Chiptester doesn't like them either.
Anyway I remove them and put in sockets. And both LS145 at the two dip 
switches are fine. But the machine runs on the bench even though it should be
held in reset by the aux-board communications transistor.
But at least I see one of the LS145's input strobe lines stuck high.
It's coming from a 9301. All the other lines are pulsing, all the inputs are
pulsing, just that single line is dead. I don't have a spare, but the 9301 is
just an LS42 with different pinout. So lacking a breadboard I put in a crazy
free-form-floating socket contraption, put in a 74LS42 and the machine works
fine again.
What a piece of shit! Last week it was working. If you turn it on now, a
bumper locks on, the displays show garbage for a fraction of a second and then
turn off, the lamps don't get strobed.
Since it was late Saturday Night and the last repairman with a key wanted to
go, all I could do was check the 5V, the fuses and the connectors but none of
this was bad. There's gotta be a fault early in the frequency generator chain.
It doesn't seem to be completely dead, when you push reset, the lamp strobe
goes to the next line or some random line.
Wow did I sink time in this. Reset is stuck low.
Was looking for that Power Fail transistor forever, but it's not in the
schematics (oh wait it is, but really well hidden) and if you force its output
high or low that makes no difference.
So I checked several ICs, all OK or within what's expected.
Grounding that one pin from F1 (4?) didn't do anything. I do get different
behaviour on resetting. It's not the RAMs, one of the 93whateveritwas also has
signal on all outputs, it's not the CPU, can only be the ROMs.
Stupidly my GALEP laptop died.
The ROMs are OK. But the machine still doesn't boot, I get an endless loop
most of the time and whatever I do, nothing makes it work.
So I have a choice of endless loops which do nothing or next to nothing.
So I even removed the quartz so the Chiptester can get better results, only 3
of the 7493 don't pass but I could get 2 of them to eventually pass.
I even removed the 93 next to the quartz because I could never get that one to
pass. I even replaced the second 9301 because I can't test it.
Nothing did anything.
One hint is A2 is stuck low. But it does toggle if it sits in a "I'm doing
something" endless loop.
But when I bend up A2 on the CPU, it always toggles. So I look where it goes.
There's a buffer and a 74367 of which there are two. Both pass in circuit(!),
but both have conflicts. Hmm. Since I sunk about 12 hours for just this last
fault already and am desparate I just yank them. And the Boss comes and says
he wants to close. But the one 367 was in fact broken and shows LOAD-0V out of
circuit. And now the machine is working again!
The credit/ball in play display is glitching a little, but I replaced the RAM
with 5114s because we're running low on 2114 and not every machine likes 5114.
Looks like this is also one of them.

Pinball Lizard Sound board
Awesome shit. Game Plan uses a Z80 including CTC in all their pins (see Foxy
Lady). The soundboard has a 680x though. Including 6821 PIA. Was it designed
by a different company?
The problem is, when you start a game, you hear thunder in an endless loop,
but no more sound effects. If you pull the data plug in a very certain way,
you can get it to play sound effects.
The error was in the Tongue Board which does a PWM fade effect for the entire
Backglass GI as well as four lamps behind the lizard's tongue, also generating
the hissing sound effect that goes with lighting those four lamps.
The tongue board also tells the Sound board to play the psiieoouuu-sound when
you shoot the ball. The signal doesn't come, neither does the hiss.
Through piggybacking and In-Circuit-testing I could find two bad chips.
After that, all sound effects worked, but only as long as the machine was cold.
After an hour, the tongue board swallows the shoot-signal and so the sound
board stays in the thunder-loop.
So U5 pin 1-3 and U10 pin 1 have the shooter signal. It's clear as day at U10
pin 1. U5 is completely stuck (but doesn't matter).
Rang through the wiring harness since I can't open the schematics with the
Acrobat 5 on my ancient machine I had there. It goes into an 6821 on the 
sound board. Swapped it, 1-2 hours later dead again. Tried swapping it yet
again, no dice. Pulled input low manually, Bzioouuuu!
So the pulse coming from U10 pin 1 is too short. Let's see if it's a bad
capacitor or one of the LS123 hanging around there.
Nope, the first thing hanging around is an LS121 and it's working fine.
But there's 2 tantalum caps and the 10 microfarad one responsible for the
timing of the shooter signal makes the same random beeps like the ones from
Vanguard. Just not as random or melodic as the Vangard ones.
Now the sound works the entire day.

Spiderman
Assignment from Fabi. Says it doesn't boot, displays just show garbage.
Funny too - he has these machines for months and now he tells me I got ONE
week to fix these two (the other one being Bally's Lost World, see below).
So it's gotta be a quickie. Just dumbly test all the 74LS with my chip tester.
All OK except one of the three LS48 for the displays which has only ONE output
working. This explains the broken displays. The others show pulses on all
segments except one  - and since the displays are strobed and should all show
just 000000 after booting that segment's gotta be the middle -.
The fact that the machine does this is telling me that the board's gotta work.
(especially since there's no hardware strobing the displays which means the
CPU must strobe them in software or per Interrupt from one of the RIOTs)
tells me this machine's gotta be alive.
Now I just gotta socket the dual port RAM. No idea if the RIOTs work and I'm
not an expert in System 80. Thank Gottlieb it's not System 1. These spider
chips are made out of solid unobtainium.
Generally both boards (the owner threw in another System 80 CPU board) appear
to be working. After noticing that the ROM sockets eat pins and having to
replace them both boards work now. Convenient.
The right bumper isn't engaged. The right drop target bank at one point 
stopped working. Gottlieb - they never made good contact.
The sound is wonky. Let's put it like this - this is a chimes board on Super
Steroids. Maybe Hyper steroids. You could say Asteroids. That deserves me a
hit in the face. But this joke hurts me as much as it hurt you.
There's a 6503 (the 6502's little brother) next to a 6530 RRIOT (which is a
RIOT with extra ROM). Next to that is a ROM version of a 2114. One of the
RRIOTs ports is hooked up to a DAC (same model as the Vectrex!) and out come
sinus waves with aliasing artifacts and software ADSR. Beauty!
The problem - it sounds weird and is hard to describe and even harder to
reproduce. Let's conclude the 6-8 hours of scoping and fidgeting and swearing
with: Contact problems. After lots of sandpaper, reworking solder joints and
flexing of many boards I got it to work half way stable and sounds almost like
one on Youtube. Btw. the CPU is clocked with a capacitor to an LS04, so if you
throw a potentiometer in the mix you can control the CPU speed arbitrarily and
since the pitch and speed of the sounds depends on the CPU pumping the values
into the RRIOT at the speed the program is running, you can bend the sound to
your will!

Lost World
As said, one week to fix both. Oh wait the deadline just got extended.
MPU not booting. We've got enough of these and I haven't fixed one yet, but I
can see that this board here has two 6800 CPUs. Weird. Oh, there should be a
6820 PIA in that slot, no wonder the LED doesn't work.
Hm I don't even get the LED to light up at all? Oh wait the driver transistor
falls apart.  The LED still doesn't light up on the bench.
I've already tested the chips in the tester. Let's put it in Harlem Globe
Trotters. Whoopsie, it
's booting now and blinks lamps. It just isn't compatible. If I hit the
saucer, it tries to re-set the drop targets.
Back in the Lost World I think - let's finally test all these 5101, but nope,
it stopped booting. Put the old RAM back, works. Put the DUT RAM in and...
crashes and doesn't even boot with the known working RAM. After an hour of
swearing I flexed a connector on the power supply and the board just came up.
Removed the PSU and sanded the pins, now it boots reliably. And all the coils
work, except the bumpers. Also some contact is stuck.
Fabi fixed the bumpers and I found the stuck contact after an hour of 
searching: Dragon's Den Target capacitor is defective even though it passes
the Testofon test, but without the cap I only see switch 22 closed every now
and then instead of permanently. Secondly, one of the switches in the chain
was wired incorrectly and had the diode across the inputs as opposed to that
weird configuration all the other buttons have. Removed the diode and now all
the signals look OK and the machine plays normally.
(except for the obligatory bad Bally lamps)

Haunted House (Gottlieb)
Another assignment. Symptom: If you press any button from attract mode, "24"
shows up in player 1's display and glitches around a little. It is said that
many buttons don't work. Probably a bad RIOT. Of course they're all soldered
in.
Bench test shows... nothing? What's RESET doing? Stuck high all the time.
Pulled RESET low, the CPU is now running. Displays show 0 (according to the
scope), BUS is alive, RESET circuit is bad. Bad capacitor.
Checked the 7400s logic, they all pass. So it's gotta be a bad RIOT. Or two.
Or all? Since I tossed the RIOTs around after desoldering, it's hard to say
which one was bad, but one was definitely bad, testing in Q'Bert's sound
board got me "Heeeeeeee..." or "Heeellloooooooooooo..." instead of "Hello,
I'm turned on". I can't test the board with a new RIOT though so I just send
it back. And the guy says except for the few odd lights and switches it works
perfectly fine.

Cobra (Nueva Bell Games - Spain)
Weird, when the machine was stored away it was in known working state.
Now it does weird things, counting up credits, crashing, getting reset by the
watchdog and strobing the displays waaay too slow.
The guy who checked it said he maybe made a short circuit.
Since the machine is Motorola 68xx based and the PIAs are socketed, I just
swap'em and the error looks way different now. So I got a new PIA and swapped
it in, no difference. Swapped the PIA that's now on my hand with the remaining
one and the machine came back almost completely.
Two lamps and one switch don't work.
Something corrosive dripped on one of the lamp transistors and severed its
connections. A second transistor is just defective (but the Testofone noted no
difference to a working one, usually you can find bad lamp drivers with it
real quick).

Unknown Nueva Bell Games board - only the CPU board
Someone told me Fabi wanted the board fixed. Nothing else is known.
So I put power to it and looked what happened. Nothing happened. The LED stays
on (note they copied the design from Bally). There was a relatively new 
looking battery on the board, maybe it's discharged and prevents the RAM from
getting power. No and yes. The battery was shot and thus the RAM got no power.
But still nothing. Reset is stuck low. After testing all the chips and even
swapping one, I found out that Reset must be generated externally.
So I pulled it up and the LED blinks 6 times. Fabi supplied the lamp driver
board as well so I tested its chips, the LS154 failed horribly, removed it,
works perfectly fine, shrug, whatever...

Lost in Space (SEGA)
Got an intermittent Reset problem for years, but now it resets all the time.
If you press on the circuit board, it stops resetting and runs.
It also resets on the bench even though the reset circuit is on another board
that's still in the machine!
The only chip that has Reset as an output is a BUS transceiver that has
nothing connected to the other side. There's a PAL that has outputs stuck and
looking like fighting and I can't even get them down if I short the inputs.
Now let's have a closer look at the part of the board that suffered battery
damage and was fixed years ago by someone else.
There's zero-ohm-resistors going accross naked traces some of which have been
patched with pieces of wire. So took the Testofone to the zero-ohm-resistor
and the traces and one rang that shouldn't. Lifted the zero-ohm-resistor and
the short went away. I could see a chip of paint missing from the bottom of
the "resistor" exactly where that wire patched trace was. These two shorted.
Fixated the zero-ohm-resistor in its lifted state so there will be no more
shorts and the machine runs fine now.
Lesson learned: Zero ohm resistors aren't just a wire with some ceramics
sprayed to the outside.

Blackout (Williams)
I bought another pinball machine for 500 euros, untested but with unwanted
extra pets, most of them eight-legged. The electronics look complete and no
battery damage though. It's got a blackout though, just looking at the charred
resistors, one power transistor was just glued (cold-soldered) to the board
and the solenoid fuse is blown, but at least many chips are socketed and
according to the date codes someone must have fixed it in 1997. And put it in
a barn in 1989.
The CPU itself is working. Turning on the machine with just the power board
shows the voltages are OK (including the +-100V) and all well within tolerance
and no magic smoke. Hooking up the CPU board shows one flash from the LEDs.
No displays (static orange glow on one digit per display) and no attract mode.
I slipped while scoping the data bus and that made one (!) controlled lamp
turn on and a few seconds later off again (left outlane special).
If you push diagnostic, the lower LED stays lit. According to the manual that
means that the 6810 RAMs are bad. For some reason I don't see CS on any of the
operating system ROMs, only the game specific ROM sometimes gets one.
Hm. The one socketed 6810 is working. And if I plug it into Firepower, it goes
into bookkeeping, but keeps complaining.
OK if I test the driver board, the machine keeps saying "FIREPOWER" and blows
the coil fuse. If I swap the middle and left PIA, it stops that, but now all
the lamps that were blinking are off. So I replaced the older of the two PIAs
(one was from 1986) and if I install the driverboard in Firepower and have a
play, I don't notice any difference to the Firepower driver board. Hooray!
Ow ow ow my brain just farted and I have to write that one down:
How do ghosts check if they fulfilled their haunt quota? - Spookkeeping. Ow!
Trying to test the 5101 SRAM in Galaxian fails due to the original not being
socketed, but if I piggyback it, the graphics go glitchy and if I piggyback
the other one I have laying around, they don't.
Better idea: The Defender machine in the storage has a socketed 5101 and a
test menu where I can check'em.
The 5101 from my pinball machine is so dead, I only see garbage in bookkeeping.
But the one I have laying around from Bally's Lost World works fine.
So let's put it in and see what happens. Well, nothing. Until I press the
diagnostics key. The LEDs flash twice. SUCCESS!
So I put batteries in and power it up, down and up again and - Attract Mode!
And... displays 0 and 1 were defective, the only displays showing anything in
bookkeeping mode.
But it doesn't react to credits and I can't play a game.
(I don't dare plug in the coil fuse yet)
A truckload of playfield lamps don't work, but the sockets are all horribly
corroded and I got all working except the 20k Bonus. Now I just take the sound
board home and fix that next.
Checked the caps, they're all still relatively usable, the big one is a bit
leaky, but that has to be expected (and is within limits due to the sheer size
of the cap). Still no sound.
Beeped the sockets, some make no contact on several pins. When desoldering,
the desoldering gun slurped in some of the pins through the circuit board.
They were THIS broken! Still no sound. Sound CPU seems to run though.
I dumped the sound OS ROM and got all $FFs. Dead as a nail. Hm you *could*
interpret the schematics this chip may need active high chip selects.
Tried that, $FF all the way. OK let's burn an EPROM, I got enough.
Put it in, hit test, and there I went. Even speech works. Flawless!
So, let's put the machine back together and see if the coils are ok and the
diodes only conduct one way.
So. The most switches work. Two controlled coils don't: Left slingshot, middle
bumper. In Firepower, they all worked (and Firepower has more solenoids).
A string of light bulbs (mostly backbox) doesn't work, the transistor and
resistor driving these stays cold. Next time: Scope driving signals.
Lamp matrix: Control signals go, the pre-driver does nothing. It can't, since
it has a broken leg. Soldered it together - blinkyblink.
The lowest green target lamp still doesn't work. The diode's OK according to
my Testofon, but bridging it brought the lamp back, so the diode's bad.
What are the solenoids doing? Have to bend over the pre-drivers to measure
and they just work like nothing happened. Hm why am I complaining?
So why don't the flippers work? Because I measure a resistance of 100k when I
press the flipper buttons. Sanded the contacts clean (I know this is the wrong
way to do it), now the flippers work.
Now all that's left are the two displays and somewhere there's a problem in
the switch matrix because I get 1010 points from just starting a game.
Of course not all the switches work because of corrosion, but wiggling them
brought them all back.
Still, getting the master display to work proved difficult. First I had to
separate the tube from the display and the 40 year old sticky tape didn't want
to separate and I even broke off a few glass shards. Temporarily soldered on
a known working tube using a bag of wires, doesn't work. I clip the salvaged
tube over the one other display with the broken tube and it flickers to life.
So I did the swapparoo and now I have 3 perfectly working displays and one
where the one's digit doesn't work (it always shows a zero anyway).
Masterdisplay is still dead. Schematics show there's a single chip just 
strobing that display and nothing else so that must be bad. It's a UDN6148
(or something like that) darlington array. And only 4 out of 8 used. Hooray!
So I just solder the unused ones in. But no luck, still dead (with the known
good tube). So I need to get a replacement tube AND chip (since I put the tube
into a formerly broken player display).
Or I take a look at the driving signals and build my own LED display (wouldn't
be the first time, mind you), especially after noticing these chips are
obscenely expensive! Like 20-30 bucks a pop. Or you buy them for 6 bucks a pop
from china and get a relabeled ULN2804 which does the same thing but goes
nowhere near 100V.
OK LED displays look relatively easy according to schematics, so I'm going for
LEDs (also I have all the parts I need in my junk bin!).
Sadly the solenoid fuse blew because the machine still likes to crash and pull
them on all at the same time (NOTE: one PIA kept popping out its socket).
Ouh man damn! I was building something nice even with drivers for the LEDs and
still they are so dim (SODIMM that's RAM) that I didn't even noticed they're
working at first.
Going from here, running the switch test. Weird stuff happens. Sometimes
(only sometimes) I get an entire string of switches. Scoped the matrix, one of
the lines is shorted to ground. One of the credit buttons was the cause, for
some reason it doesn't get strobed and is stuck low, so when it shorts, it
wreaks havoc on the switch matrix.
So, the next version of the LED board will have its very own latching decoder
(CD4511) for each digit, where the strobe latches the data and the digit stays
on all the time! Or I take a look at our Black Knight which has LEDs and they
are really bright. Why reinvent the wheel? Because it's fun!
OK this works better but is still too dim. But if I short out the resistor
arrays I put in, it becomes bright enough to read in well lit environments.
I can definitely leave it like that.
But still, if I start a four player game I get 24040 points each time I hit
the start button.
Oh wait in between testing the boards from Peter's Blackout (see below), the
displays weren't working for a while and I kept the machine on and it started
to smell like solder, so I looked at the machine and one string of lights was
extremely bright. Turned it off and three of the big lamp resistors have
desoldered themselves and were lying in the bottom of the machine.
The transistors survived though.
And Friday before Open Saturday it's out of order again. Turned it on, dead.
Put the other CPU board in, works, swapped chips, nothing happened.
Put it on the bench and plugged into the chip tester. It says the LS139
chip select generator has output 0 stuck high - so the very first ROM never
gets selected. Ergo the machine can't boot. Changed it, works.
Let's try to figure out the switch matrix. As stated, one line never strobes.
It's column 1. Checked the driver board: The 74LS06 has the input floating.
Beeped it, the socket doesn't make contact. Even better, the socket already
has a mark on that pin (and that pin only). Changed the socket and everything
strobes now. And if I hit the playfield, I only get two switches instead of
four. But it still sometimes jumps back to the second to last closed contact.
Let's see how it fares - during gameplay it looks like everything is fixed and
working fine now. It even survived the first open Saturday with no problems.
But now there's Corona virus, which means no more open Saturdays and no more
stress test. But if you read that, the spook is over and the world supply of
toilet paper has restored. I once made two orbits in a game (hard because the
left spinner doesn't turn easily and takes some speed off the ball).
I also got Blackout a couple times, so I beat the machine.

Blackout (Peter Popp's machine)
Now we got two Blackouts and one is working (somewhat) so we can swap parts.
Let's start with the master display. No problems here.
Continuing with the driverboard. In Circuit Chip Tester complains about a lot
of chips, so I checked the chips from mine in circuit and get the same results
except for IC6 (LS08) which has Mid Level on one output.
I just put the board into the machine and it works just fine. Except for
solenoid 19 which is stuck on. Checked the schematics and that goes to Q6
controlled by an LS02 which gets controlled by the very pin of the very LS08
that the chip tester complained about. Replaced it without testing because
if two independant sources come to the same result it's gotta be right.
(don't use that as an universal truth though)
Continuing with the CPU board. On the bench it just lights up. No reset.
Forced reset high. LEDs went off and turned on again. According to the 
internet that means 404 error code not found. So it probably needs the driver
board to work. So I plugged it into my machine. Nothing happens and the LEDs
now blink Bad 5101. Damn, they start to get expensive. Swapped with mine, test
passes. Reboot, it showed its version number for a second and brightly 
glitched away because of a short circuit with the board below.
So I put something in between and the displays are broken, every second digit
doesn't work. But the CPU board seems to work. Plugged my old one back in and
the problem persists. Damn it, this thing just blew my master display.
For some weird reason my CPU board passes with the other board's 5101.
Let's plug Peter's Master Display back in. Everything works fine. That's how
it should - my master display's blown. I plug it back in to diagnose it and -
it just works. WTF?
One string of lights isn't working. After not being able to find anything in
the schematics and on the board I jiggled the wire connectors and the string
came on and another one went off. Jiggled until all the lights worked.
Now we can actually put this machine back together and test it without help of
my machine. Only the power supply has to be tested before hooking it up to
anything and then off we go. Sound module hasn't been tested yet either.
Eventually I tested the sound module. It works perfectly fine besides having a
very high background noise level. The culprit was unsurprisingly the blocking
capacitor next to the 7805 voltage regulator. Replaced it and now the sound
module only makes sound when instructed.
Putting the machine together was fun... NOT!
The left nut for putting the left Head unit screw through was missing. And
even the longest screws we had were just barely long enough. And the previous
owner changed the wiring loom's contact plugs to something one-dimensional and
non-professional looking. And there's a wire going nowhere. Checking with the
other machine confirmed it literally goes nowhere.
And then there's a 24V transformer on top of it, that has something to do
with the GI and does God knows what. The GI also gets supplied 6.5V and is
relatively bright. At least the power module's output voltages are correct
except for the 28V for the solenoids which are at 38V, 10V higher than they
should (but I heard that's a common modification).
The machine is in much nicer condition than mine and works relatively decent.
Except: BOTH drop target reset solenoids have been "lubricated" and get stuck
now. Also someone completely changed the middle pop bumper's circuitry. 
The way it's wired up, the solenoid voltage goes through the contact switch.
That makes no sense unless... remember the bad IC in the driver board? It
controlled exactly this pop bumper and had it on all the time. However now
that the fault has been fixed, it doesn't work at all. So using the other two
pop bumpers as a template, I restored the middle bumper's circuitry and it
works fine now.
Kinda weird that all switches work. Some of the lamps had bad contacts, also
these homebrew contacts for the lamps have bad contacts, that's gonna be fun.
The pinball repair people have fixed the drop target solenoids and I can see
that my machine has a problem with the spinners. Here, you give them a short
stare and they turn for minutes. Mine barely spin at all!
And then the displays for player 3 and 4 and most of player 2 went dark.
And when trying to fix it, they randomly turn off and on and it appears to be
a mechanical problem. So I took the master control board out and the solder
joints looked pretty bad, so I reworked them and left the machine running for
hours without anything happening.
But it did come back a few weeks later...

Williams World Cup (1978)
Looking for System 3 schematics, found none. The CPU board does look different
than my Blackout which is system 6. And then I stumbled upon the fact that the
CPU and PIA have swapped positions. That takes something to get used to.
Helps especially if both chips only have cryptic Williams codes on them and
not 6800 or 6821...
Then I find a forum where someone complains about the bad quality of the scans
of the system 3 schematics. He was so pissed that he drew new schematics in
MS Paint, which are incomplete. I'd rather have the bad scans than no scans at
all, but at least I can make do with the MS Paint schematics.
Well, if you pull Reset high, the machine starts running. Complains about the
5101 being bad. Tested it, it's OK. One of the chip select signals is stuck
low, it comes from the reset circuitry, so it's a power on garbage supressor
memory protect circuit. I don't have 12V here. So I put it in my Blackout.
And the LEDs flash twice if I press Test, so yippie, all's well.
The displays  just show a weak garbagey flicker, though. After power cycling
I got 160000 displayed as the high score. Service menu works fine, too.
The driver board had a bad transistor which was swapped before.
So with the intermittent boot in my machine, I swapped out all the big chip's
sockets (apart from the ROM sockets which had been changed earlier)
and the machine should be rock solid now.
Hm according to Fabi it doesn't run at all.
He brought the board back like a year later. Yes I recognize my old work.
He says the CPU doesn't boot reliably. I take a look at the power supply board
and don't ask why. I asked why it did anything at all. It's full of dry joints!
The 5V fuse holder is corroded, completely rusted through, there's a fuse
holder that's populated but not connected to anything so I use it as a spare.
And then I sunk a ton of solder into the board. I put it in Flash for testing
and it works better than Flashs power board. Even with the original
electrolytics. The high voltage ones bit me even a minute after removing the
board, so the leakage current is negligible.
What's the driver board doing? I fixed and tested that one, so there can't be
much broken especially since the power supply board died of thirst.
And of course it works. But after I started coil test, the right slingshot
locked on after it got triggered. And now it stays on forever.
Driver transistor gets hot, so that's gotta be it. Testofon says he's
innocent. Pre-driver passes too. Must be a logic problem. Yes, the logic IC
sends an active signal. Why? Because an active signal goes in.
That comes from a 7408 and the chip tester says the input is shorted to ground.
So I short it to +5V and check if anything gets warm. The 7408 itself gets
warm. Replaced it, fixed it. Weird. Last week I had the same problem with
Gorgar, same symptom (right slingshot triggers, then locks on and driver gets
hot), but here it really was the driver.
OK on to the MPU board. That should still be working, not?
Nope. LEDs are stuck on, scope says endless loop and if I press the diagnostic
button, different endless loop. At least some life.
Then I saw someone has plugged the RAM chip in the wrong way round!
Couldn'tve been me, because when I handed in the board, it was working.
Curiously the RAM survived. I'm not too sure the MPU works now, but it
switches between 0 and 350000 points and blinks the high score lamp.
And I can go into audits and check stuff and the display, lamp and coil test
work (that's when the 7408 died), so it sounds working to me.

Space Race (Recel) (Post edit, I fixed this machine at the Gamescom 2018)
Space Race at some time quit counting down the Bonus. It gets stuck in a loop
and won't give a ball. That evening, we were standing around the machine, 5 or
6 people, all trying to fix it without any noticeable success.
Next day, everyone said screw it and went to lunch, so here I was, all alone,
everyone gone, hall closed to visitors, and never had fixed a mechanical pin
in my life.
I noticed the bonus relais engaging just very very briefly. If you hold it
closed, the bonus does count down and the ball comes out. So I went hunting
for the hold signal. I even went to the head unit, looking for the 100 points
signal which worked fine and didn't helped at all. After lots of thinking and
beeping wires with the Testofon it became clear that the bonus relay holds
itself until the bonus is at zero. So that self hold contact must be bad.
I removed the structure holding a bank of relays with the bonus relay being
one of them, I could see a contact tongue lying on the bottom of the machine
right away. Close examination of the relay showed that there was indeed a
contact tongue broken off. I soldered both ends together and even though that
was just a quick&dirty fix, the pinball machine not only worked fine after
that, my skimpy fix was never rectified and the machine runs with it even now.
(2 years later).
OK here comes a rant that has nothing to do with anything and probably doesn't
apply to your country, but for the sake of completeness, I translate it
anyway. Here goes.
As opposed to the skimpy fix on Space Race I was accused of doing a really bad
job on fixing the capacitor on Flash (see above). Because I have swapped out
a 12000 uF capacitor with a 10000 one and soldered a 2200 one on top of it.
Even though these solder joints are decent and the cap doesn't really flap
around much, for some individuals it is a red flag improvising the replacement
of hard to find or unobtainable spare parts. And even if I did get a 12000uF
cap (Mouser should have them), the replacement will be 40 years smaller in
size and won't be able to be tightened down with a cable tie like the original
was, so it will be mechanically less safe than my "bad work".
And while we're at it... the people making the laws often have no idea about
the subject, especially in electrical engineering. Example electric security.
Every metal surface of any machine MUST have at max a 0.3 Ohms connection to
ground. Great. Good news is that if you have a non-conductive surface with
metal screws sticking out, those need not be grounded. Also moving parts are
exempted as well.
Bad news:
- We have metals that have a high resistance finish on them.
- We have many machines that DO have a direct path to ground on all metal
surfaces, but due to age, the resistance went up to 1 Ohm and the machine has
to be completely redone and even that often doesn't fix it - think rust - I
had to route a wire around the front of the control panel of Galaxians.
The panel is painted, but since it's made of metal, has to be grounded anyway
and there was no way to do that on the back, since it's mounted on a wooden
panel. Now if you look closely you can see that cable on the front. It's ugly.
- Example Qix. We don't have keys for the coin door. It's completely painted
and the paint hasn't chipped so it doesn't need to be grounded. But the lock
is made of metal and it's exposed so it must be grounded come what will.
Now the coin door is already grounded via the hinge, but age and rust 
decreased resistance to 10 Ohms. So we would have to drill open the hole
(while we're at it, we could just remove it, negating the need to ground the
door), drill another hole in the coin door, file off the painting around it,
connect a ground wire to it and MAN will that look ugly. Can't I just stick a
piece of duct tape on the lock? It's way down there in the darkness and nobody
is gonna see or even lick at it.
The idea of grounding metal surfaces is that if something goes wrong with the
machine and it puts juice on said surfaces, it'll trip the FI fuse.
And for that to happen, only 10mA are necessary. Which happens just the same
if the resistance is 100 Ohms.
Hello? Have you ever changed channels on a 1970s TV with sensor touch buttons?
These are hot chassis TVs, so there is a direct path from your finger to the
socket! And it's desinged like that!
That makes as much sense as decreeing per law that all vehicles must adhere to
the latest exhaust pollution norms. ALL of them. No exceptions. You got a car
from the 1920s? You got two options: Either the exhaust must not contain more
than the maximum allowed amount of pollutants or you have to destroy the
vehicle instantly. And to keep it original, you have to use period correct
parts to achieve the exhaust cleanliness.
And if you have an electric car, you'll have to install a campfire or
whatever, because there's also a MINIMUM amount on the pollutants a car must
produce. Think it makes sense? This is how much sense the law makes to me now.
A small hint: Generations of people survived these illegally unsafe electical
appliances. And those that died often had nothing to blame but their own
stupidity. Quit interfering with Darwin, dammit! (OK I'm not really serious on
this statement).
If I made that law I would have stated that any mechanical surface must be 
able to allow at least 500mA to flow to the ground connector in the socket.
With the exception of metal parts that have no contact to the rest of the
machine like screws and rivets connecting wood or plastic parts. And of course
moving parts. I'd probably even go as far as stating that if there's no wire
with more than 48V within 30cm of the object, it doesn't need to be grounded.
This would get rid of having to ground dozens of coin doors.

Take Five (Allied Leisure, 1977)
Cool a pinball that has the same name as the band my father plays in!
It looks like any other Game Plan cocktail table, but the circuit boards are
completely different (EDIT: I just found out they've been designed by Chuck
Peddle just before he did the PET - that explains the insane amount of MOS
parts).
This one uses a MOS 6504 CPU (I think the MPS801 1-needle-printer uses this
one) and a bank of 6520 PIAs and a bank of 6530 RRIOTs (so no RAM or ROM,
instead an overabundance of I/Os in a chip that essentially makes a
microcontroller out of a 650x). Too bad no 6522. (but the 6520 are 
pincompatible to the 6820 PIA). And the RRIOTs (like what, 4 of them) are
unobtainium and even if you get one, how will you be able to program it!?
Well someone designed an adapter for 6532 which are hard to get but not as
impossible as a 6530 with the correct ROM.
Well, what are we up against here? The machine gives out no ball and doesn't
count points, but apart from that it appears to play just fine and react
normally. So if you clear the drop target bank, it resets them. After ringing
the bell 5 times. But the displays stay at 000000.
Since the machine appears to be otherwise working normally and the schematics
show the displays are connected serially (with CLK and DATA) there's either a
problem with the connection or the encoder or one of the three decoders per
display.
But... this is a 43 year old SS pinball machine that hasn't been turned on in
decades, so check the power supply health. Two big capacitors are getting hot.
One is hanging above a rectifier, but that one is lukewarm and the cap is hot.
Then there are two big bucket capacitors, one stays cold, the other one gets
warm. After changing the two suspect capacitors, the solenoids have more oomph
to them. While we're at it. The ball shooter coil is completely toasted. But a
new one doesn't remedy the situation. The driver transistor looks melty...
But my Testofon only says there's a difference in the pre-driver.
With a different pre-driver, it still doesn't work, but at least the control
signals up to the driver transistor are OK now. So the driver gets changed,
too (Reiner's father's spare box had three of those). And now when you start a
game, the ball appears in the shooter lane.
Concerning the displays, I see neither data nor clock on the scope.
The machine doesn't have a serial interface, so that gets solved with discrete
components and software, which is good because I know I don't have to change
any unobtainium parts.
Let's beep through the wires where the signals originate (mainboard pin 
numbers are illegible in the schematics) and there's 2 ICs, a 7404 and a 7442.
The 04 passes and the 42 fails completely. There's another 7442 next to it and
it passes no problemo. Removing the 42 was huge fun because of a ton of bodge
wires running past it on the solder side...
With the new LS42 the credit display and the player 1 score display works.
The player 2 display only has the tens digit working...
...the ultimative asshole machine for the tournament final. Player 1: 295830
points. Player 2: ... 40 points. With 5 balls. That's an accomplishment.
Swapped the displays and the fault moves with them.
The In Circuit Chiptester fails all chips in both boards.
But the first chip in line fails differently. Replaced that one and both
displays now work.
That means the only thing not working now is the knocker (too lazy and too
unimportant) and the switches are all in need of a cleaning.
Well. Matthias keeps talking about the knocker and I got nothing to do so I
gotta care about it. Ringing through where it goes. Nowhere. Unplugging the
plug. Ah, the green plug. Circuit board out and now I know which transistor it
is. But that one didn't beep. Hm. That plug is painted green, but the contacts
are green as well... so it didn't make good contact. Well it beeps now, but
still doesn't work. A closer look reveals a blown resistor. And the pre-driver
is shorted. So I put a new resistor in and a new pre-driver and luckily the
big driver transistor was still OK and now it works.
The displays have a few segments missing, but here the LEDs were bad and not
the LS47 driving them. A few LEDs actually died while testing them.

Cosmic Flash (NSM)
Dunno if I should write that down at all.
The owner says the flipper and slingshot transistors went up in smoke before
him and the 74LS06 next to the CPU got so hot that it desoldered itself.
Although when he removed it, he nicked all but two of the ringlets. Or vias.
Or how they're called. Yeah but they could've been delaminated from the heat
of the IC. Interestingly he said the machine was working for an hour after
swapping a new LS06 in there. So I removed the IC just to see where the traces
go. Took a photo, put a socket in and in 3-4 hours of work beeped EVERYTHING
for continuity, had to patch tons of traces with wires. The IC drives the
displays (data in/latch) and lots of important functions and I had to fix lots
of that on two fronts... well... no clue if it does. But if it does, I'm gonna
go for the Hot Fire Birds spare board first.
Well it seems it does because he wrote that his machine now works fine.
(Edit: The Hot Fire Birds spare has different ROMs)

Tri Zone (Williams)
One of the big trannies on the driverboard failed the Testofon test.
Is a lamp driver, so it has to handle 6V. I don't have TIP122, but I have
TIP120 which can "only" handle 60V so that'll do.
CPU board doesn't do much in the machine. The red LEDs turn on and off, but
besides that the machine is dead. On the bench it reacts to the Diagnosis
switch and then the lower LED stays on so one or both of the 6810 are bad.
Testing in Blackout didn't work as desired, so I wrote a test for the Chip
tester. Sadly after many hours it turns out it doesn't like bidirectional pins
and always reads back whatever was on the BUS the last time. Most tests pass
without the IC, just the self.address test fails. I wonder if I can get this
to work.
The IC sockets are SCANBE so get rid of them. Two 6889 next to the CPU are
said to become super hot, so they're suspect right away.
After the board still not working with new sockets and the RAMs are known good,
let's socket the 6889s. Blackout doesn't even have chips in there, just solder
jumpers. So instead of the ICs I put in some wire jumpers and it doesn't work.
So I make a Blackout out of it - put the Blackout ROMs in, works. Self test
passes, so ROMs back and it's dead. Through swapping of single ROMs I could
point it down to the GAME ROM and it seems to be the only bad IC left. That
ROM has bit 7 stuck high. For shits n giggles let's pretend it's an EPROM and
burn it. Now all I get is FF for everything. Meh whatever. Chip was bad to
begin with. So it's guaranteed a new ROM will get the machine working.
The 6889s still get hot and keep the machine from booting.
Speaking of FF... I wanted to burn a replacement ROM, but I couldn't burn the
first byte. But I could write the rest of the ROM fine. But the ROM has only
1k used I mirrored it and shorted A11 to +5V. Seems to work, although the
machine does fail self diagnosis now.
And just a year later I get it back, broke again.
Only goes to bookkeeping, the spare ROM doesn't work, but it doesn't work in a
different way so both appear to be bad. Maybe the tiny little bit of UV the
lamps put out erased it.
Also there are three transistors broken (physically) in the lamp driver board
and only two of them were to be found so let's see what we got in our spares
bin and decided upon a BD646 no idea if it's gonna work.
Also the customer says the soundboard starts humming like mad after half an
hour, so I checked the caps and the big 12000æF got chilly 20mA leakage at 15V
so that's gonna generate a little bit of heat which increases the leakage a
little which generates more heat and so on until the cap gets toasty.
Well yes no the EPROMs only lost one bit each at the start of the address
range, the other only on the mirroring.
The board seems to be completely dead. Pressing diagnostic switch turns on the
bottom LED, so I swapped the first 6810, nothing. Swapped the second one and
now it goes into attract. Put the original one back in, no difference.
It doesn't like that second one. If I test both RAMs in my machine, they both
work fine.
So I put my game Blackout ROM in and get two flashes. Weird.
And if I go to audits, I can't do anything. I can't even start a game, not
even with the Blackout ROM. And one string of lights is copulatingly bright.
Test shows something wrong with the 4049 chips in the keyboard matrix.
So I swapped them and now I can start a game and operate the audits menu.
But the lamps are still too bright. The owner put a 10 Ohm resistor on that
string instead of 27 Ohms. But that doesn't help. The transistor also sounds a
little different. Replaced it, nothing. The data is coming from a 7408,
swapped it, it didn't survive but it wasn't the fault, so the PIA's gotta be
bad. Shit. Of course I could try synthesizing the missing lamp strobe from the
others. Nah I didn't think of that anymore and put a new PIA in.
Problem - the CPU board doesn't wanna work for the life of it.
On the game ROM socket, the A7 line showed open. So I replaced the replacement
socket again. I can swap RAMs and ROMs... nothing works.
Finally even the CPU died, but only the reset pin was bent... stupid.
Finally I beep the ROMs and... forgot to solder the +5V... gotta be stupid.
Now it works again.
I'm just not 100% sure about the self-burned game ROM, one has the first byte
bad, the other one went missing and is nowhere to be found. I had put it next
to the soldering gun because a pin had broken off.
And now the middle yellow target tilts the machine but so does the tilt
pendulum. But the roll ball tilt doesn't. Well it appears the matrix
assignment is different because the owner says it's working fine.

Data East Star Wars
This machine isn't that old, why are the solder joints crap?
Unfortunately, the fault description is super wonky with the message going
something like "Uh I downloaded a hack ROM from the internet, now there's a
feature missing and then it started smoking and the lower GI failed" yeah
great starting point for looking for faults.
So I do what I always do. All the chips go to the chip tester, all trannies
go to the Testofon and one TIP42 in the lamp circuit is shorted. Must've got
zapped with solenoid power. So the owner must check for a short on the
playfield...
The pre-driver is still OK so there's a good chance the PIA is still OK
(there's like 6 PIAs in the machine).
I don't have TIP42, but tons of BD244 and since it goes in the lamp matrix,
55V maximum rated voltage isn't a problem there.
Fabi says I can test it in Jurassic Park. It runs, seemingly fine, some lamps
don't turn on, but you can start a game and all the coils are working - but
the flipper fingers don't! Driver transistor for the flipper relay was bad
(Testofon says Hfe of about 1).
With the original board in, I can see the missing lamps only turn on in game,
not in attract.
I still have to test the power board, but not in Jurassic Park, since I'm too
small, too fat and too clumsy. And I have to be on guard for every tiny thing
because even the tiniest scratch can cost a thousand Euros. The market is
insane. Btw. the original box for the Vectrex (with arousing game modules)
is probably worth more than the Vectrex itself which goes for 250.

Cybernaut
Weird problem and no solution.
It's dead - well not quite. The LED flickers on power up but never flashes
which is a ROM checksum error.
But if you leave it running for 10 minutes and reboot the machine, it just
works! The waveform on the 5V cap isn't to my liking so I checked the
rectifier but to no avail. Swapping the cap also made no difference.
So the operating system ROM is in a precision socket and the game ROM (if you
can call it that) is in an Augat socket.
So I replaced that socket and instant Bam - it works right away.
Unfortunately the next day it's back to its usual weirdness.
The CPU/PIA sockets are RS and make a similarly bad impression. I could get
the LED to blink once by jiggling, but only once.
And now the machine is gone so I can't continue.

Robot (Zaccaria)
That one worked perfectly. Until it suddenly became very broken.
The techie only noticed the knocker coil burned up and the player 1 display
had 4 digits missing. It appeared to have completely gone past him that the
machine was completely brain dead despite the clear lack of an attract mode.
I noticed that neither the sound nor main CPU appeared to boot.
If you pull the cable between both, the sound CPU comes up and can be tested.
After that it took me quite a while to realise that 2650 CPUs have their RESET
active high. Then I noticed that the original battery (which has been removed)
has eaten one of the RAM chips (which has been replaced a long time ago), left
the other RAM chip worse for wear and also drooled onto a 74LS00.
The latter had outputs on all gates but one, whose inputs looked like the
output should toggle. It's the RAM chip select. So I replaced that one (with
one from 1973...) and the CPU now seems to boot or at least behaves more
logical than before.
Back in the machine, it stops playing glitchy sounds all the time and it now
drives the displays properly and there's attract mode running, but the fuse
for the 39V blows as soon as you turn the machine on (the fuse was blown
before I started fixing the CPU board).
Unfortunately, I can't measure a short on the plug that causes the fuse to
blow. Only when I started measuring from Ground I found a short... it's the
knocker coil. If I remove it, the short is gone. The machine now plays.
Well, why's there a short? Because the transistor driving the knocker is dead
shorted. I put a replacement in and forgot to put the coil back in, but that's
unimportant. More important is I noticed some lights didn't work even though
their bulbs are good. After beeping through all the thyristors and not finding
anything that's way off, I followed the beeps from the bulbs to the thyristors
and noticed the control signals all come from the same DEMUX IC. Its chip
select is missing. I noticed it should come from the CPU board but doesn't
make it through the cable to the driver board. Turned out to be a bad contact
in the ribbon cable, so I made a new one and now it runs perfectly.
(as short as this story reads, it's 10 hours of work...)