Gorf/WOW/Space Zap/Roto sound amp tantalums!

Here are my observations and thoughts.... YMMV

I have repaired over 200 of these board sets (Gorf, WoW, Robby, Space Zap, and Extra Bases). The tantalums fail often at this point, but it is almost always the caps on the 12v rail that fail. This happens with both the original PS as well as switchers. On my bench supply the voltage is set precisely at 12.00v, and it still happens.

It's gotten to the point where I will not even test a board set anymore on my bench without first replacing the caps on the 12v rail. It's just not worth the headache for if the rupture of the cap is pointed down towards the board it can actually do some significant damage to the board and/or traces. I'm too old for fireworks on my bench (it reminds me of a "flower" firework).

I've replaced all the tants (16 on Gorf, for example, not including the one on the Amp board), on every set I've done. I replace them with electrolytics and have never had an issue nor complaint. I'm sure we could argue the original intent of why the engineers chose the tants for this application, and many other Bally Midway applications, and there probably was a valid reason, but that would be out of my knowledge area or interest quite frankly. But again, the electrolytics work fine.

As for the customs, I'm not sure I am a big believer of the mildly "over-voltage" theory. I'm sure if the 12v is cranked way up over 12.00v then it can cook them, but it's my experience that switchers do not typically run super high on the 12v as long as the 5v is dialed in. The on-board regulation circuit is very crude, but it usually holds fine. Even when the customs die the voltage is still at the prescribed setting if the 12v input is accurate.

Moreover, the single I/O applications (Space Zap, Extra Bases) run that I/O custom at 10v whereas the dual I/O' applications (Gorf, WoW, Robby) pull that down to 8.5v. As such, I suspect the VGG voltage on these customs have an acceptable range and are a bit more lenient then we give them credit for.

That all said, I am a firm believer in heat killing these customs (and, yes, I do understand that heat can be related to voltage), and especially a killer for the Data custom. It's odd that the earliest production CPU board version ("C") and every later release has the through-holes. Could that have been for a heat sink to be zip-tied to the Data custom? or maybe that was for a daughter board instead of a custom?? but either way Bally did not install a heat sink on the Data custom until the later production run of the later games (Gorf). Either they got lazy or cheap but it eventually caught up to them. Of all the dead Data customs I have seen (probably at least 80-100 of that variety), I think only two had the factory heat sink installed. So, that heat sink really did/does make a difference.

Since there are almost as many failures on the other two chip varieties (Address and I/O), I also have been wondering lately if the fact that AMI was the partner here (and not another vendor) also contributes to the failure rate. This is pure speculation but if I use the high failure rate of AMI's 6821 PIA chips as compared to other vendors, perhaps they just didn't have the best quality in their 40-pin DIP manufacturing process?? Again, pure speculation here but it does make you ponder. Maybe if another partner was chosen by Bally we would have less failures..... maybe....
Great write up, Kevin! I've been testing some of my extra boards as well, but I don't install them until the tantalums are replaced.

On the topic of heatsinks, do you suggest all 4 customs should have heatsinks or just the data custom on the CPU board?

Thanks,
Jason
 
Great write up, Kevin! I've been testing some of my extra boards as well, but I don't install them until the tantalums are replaced.

On the topic of heatsinks, do you suggest all 4 customs should have heatsinks or just the data custom on the CPU board?

Thanks,
Jason
I guess it theoretically wouldn't hurt, but I don't believe those other two varieties get as hot as the Data. I used to install heat sinks on both the Data and Address when I went through a set but stopped doing it on the Address as I felt it wasn't likely necessary. The Address custom actually fails the least out of the three.

The other issues you run into if you go heat sink happy is:
  • being able to probe the legs for troubleshooting if you use a design like the original heat sink
  • fitting all 6 boards in the cage if you use the more modern raised fin variety heat sink
  • being able to rotate boards to front position (for probing) if you use certain heat sink varieties
  • installing Brian's FPGA Data custom if you use certain heat sinks on the Address chip (or run a HSS kit on the Z80)
I just mostly focus on the Data chip for heat sinks these days. If one comes in that does not have a heat sink I install one identical to the original design as that allows for the CPU board to be moved within the cage. I still advise people to run that board in the front position to maximize cooling.
 
As for the customs, I'm not sure I am a big believer of the mildly "over-voltage" theory. I'm sure if the 12v is cranked way up over 12.00v then it can cook them, but it's my experience that switchers do not typically run super high on the 12v as long as the 5v is dialed in. The on-board regulation circuit is very crude, but it usually holds fine. Even when the customs die the voltage is still at the prescribed setting if the 12v input is accurate.
There "on-board regulator" doesn't regulate, and that's the issue.
It's just a resistor divider going into a darlington, so you get Vin * DivRatio - 2 Vth, and it tracks up when the +12V increases.

Putting in a 7808 with a diode in the ground leg gives you far better ~8.5V regulation

Since there are almost as many failures on the other two chip varieties (Address and I/O), I also have been wondering lately if the fact that AMI was the partner here (and not another vendor) also contributes to the failure rate. This is pure speculation but if I use the high failure rate of AMI's 6821 PIA chips as compared to other vendors, perhaps they just didn't have the best quality in their 40-pin DIP manufacturing process?? Again, pure speculation here but it does make you ponder. Maybe if another partner was chosen by Bally we would have less failures..... maybe....

40-pin DIP isn't a "manufacturing process" -- it's just packaging encapsulation.

I don't know what process node was used at AMI for the 6821 or Pokey (likely both NMOS only processes at the time) but they're different from the process used for astrocade which required VGG for the enhancement loads.
 
40-pin DIP isn't a "manufacturing process" -- it's just packaging encapsulation.

I don't know what process node was used at AMI for the 6821 or Pokey (likely both NMOS only processes at the time) but they're different from the process used for astrocade which required VGG for the enhancement loads.
I was using "manufacturing process" as a generic catch all assuming that 40pin DIP packages from the same vendor may use the same machines, material, clean rooms, plants, employees, etc., 🤷‍♂️ 🤷‍♂️ and any or all of that may have be problematic for AMI back in that time. My entire theory is all supposition so I cannot defend nor explain any of the details, nor do I have any strong understanding on how these chips were made to improve my chosen language. Maybe I'll just leave it as, "AMI appears to suck".
 
I was using "manufacturing process" as a generic catch all assuming that 40pin DIP packages from the same vendor may use the same machines, material, clean rooms, plants, employees, etc., 🤷‍♂️ 🤷‍♂️ and any or all of that may have be problematic for AMI back in that time. My entire theory is all supposition so I cannot defend nor explain any of the details, nor do I have any strong understanding on how these chips were made to improve my chosen language. Maybe I'll just leave it as, "AMI appears to suck".

Packaging is generally completely separate from wafer fab.
Once the dice are passivated you no long need a clean room -- it's just wire bonding into the lead frame at that point.

In any case, we're long, long past the designed MTBF for any of these devices.
 
Packaging is generally completely separate from wafer fab.
Once the dice are passivated you no long need a clean room -- it's just wire bonding into the lead frame at that point.

In any case, we're long, long past the designed MTBF for any of these devices.
Yeah, Mark, that may be a better term to express my sentiment. I would suggest that AMI's MTBF is lower than others for the same spec'd part.

In the case of 6821's we actually have other vendors to compare AMI to, and AMI seems to have a higher failure rate (lower MTBF).

For these Bally customs AMI was the sole vendor, so I can only guess another vendor *may* have been better. The fact that Bally installed heat sinks on the later production Data customs *might* also suggest that the actual MTBF in the field fell short of the planned MTBF.

Again, all complete guesses and suppositions here but either way these customs are all slowly failing. Thanks to Brian for creating the FPGA chip so the original hardware can still be used.

Note, I did have a brief exchange with Jeff back in 2022 to see if he may be able to help in custom reproductions. He initially said he'd try and help but then went dark. I was hoping he'd have his old engineering docs in a file drawer somewhere, but alas Brian was able to succeed on his own.
 
Note, I did have a brief exchange with Jeff back in 2022 to see if he may be able to help in custom reproductions. He initially said he'd try and help but then went dark. I was hoping he'd have his old engineering docs in a file drawer somewhere, but alas Brian was able to succeed on his own.

The schematics for the customs are in the Bally patents...
I've written verilog models and made rudimentary test benches for all of them but never put in the work to make the designs physical.

There's really not all that much to it, but some of the structures they used don't naturally map to modern fpga logic.
 
Back
Top Bottom