SuperMac Spectrum/24 Series III Display Artifact Issue

[Bolle]

But how would a read damage a chip?

You have to apply programming voltage to one pin of the chip to put it into prgramming mode - however it’s totally not standardized which pin that’s going to be across different GALs/PALs or PALCEs.
GALs have a signature that tells the programmer what voltage and timings to use but not all (especially cheap ones - looking at you 886 and derivates) do not obey to that.
For plain old PALs it’s even worse, even between different revisions of chips from the same manufacturer the programming algorithms and pinouts are different enough that I killed chips before by selecting the wrong type.
So there’s some care to be taken when poking unknown sanded chips.
With the chips that we know to be Lattice 22V10s I‘d see it to be extremely unlikely that one gets hurt.

I do have a Spectrum/24 III without the addon board so I‘d actually only need the accelerator boards.

 

[jmacz]

For plain old PALs it’s even worse, even between different revisions of chips from the same manufacturer the programming algorithms and pinouts are different enough that I killed chips before by selecting the wrong type.

Ok, that's enough for me to not touch these chips anymore myself.

Will work with you via PMs on getting these to you for reverse engineering.

 

[jmacz]

I couldn't resist ... I pulled the working chip from the working accelerator / daughter board and tested it in original accelerator board, and it works. It works without issue. I put it through some tests and I don't see the artifacts. So it does look like this particular chip (the one that failed, the one that worked when heated previously) seems to have been the culprit.

But I'm not going to attempt another read.

Both accelerator boards will be leaving for @Bolle next week. Hopefully he can reverse engineer the GALs so that I get the original accelerator working.

 

[MacOSMonkey]

It will be exciting to see what happens this week on...Bolle-ing for Diodes!

Regarding my earlier recollection about NS PALs vs. Lattice GALs, the fact that the parts are sanded does not necessarily mean that they were the NS PALs, but it's possible.

I wouldn't beat yourself up too much about the bad part. Based on all your other testing, it seems like it was marginal and whatever happened probably just kicked it into an unmarked silicon grave.

One advantage to reading the GALs would be that there were later device update(s). .For example, I think there was an 8●24GC compatibility issue -- maybe bus master- or interrupt-related. Anyway, there's no guarantee that your original board has any updates, but it's more likely that the GAL board will. I think there was also a respin to eliminate a rework jumper. The labeled GAL board is 2 rev steps (Rev G) away from your original board (Rev E) and at least one of those steps could be programming-related. If it doesn't have a little jumper trace on it (on the connector maybe? - don't have a board handy) and your original board does, then that is the other step.

 

[jmacz]

If it doesn't have a little jumper trace on it (on the connector maybe? - don't have a board handy) and your original board does, then that is the other step.

Both accelerator boards look identical except the sanded markings and the serial/revision indicators. They both have jumpers on the delay chip (kappa st08cb500) and no where else on the board. They removed two of the delay taps on that chip and then used jumper traces to bridge the signals to the other taps.

 

[MacOSMonkey]

I thought there was a respin at some point, but maybe not. Maybe just the 8*24 fix and PAL->GAL cost reduction (that would have also required different/recompiled device files). I will find my boards and check them for the rework.

 

[Bolle]

It's getting ugly...


I could confirm that all chips from the problematic board are indeed Lattice 22V10 as well, so the dead chip wasn't killed by a wrong programming algorithm at least but must have decided to go on its own.

Looking at the second board that still has markings on the chips there's 12 GAL22V10 in total, four of them are 15ns parts, 8 of them 25ns parts.
I was able to glitch around the security fuse on all the 15ns parts - luckily the bad chip was one of those...

Will keep going poking at the rest to see if I can get the 25ns chips to talk to me as well.

 

[jmacz]

Nice!!! Sounds like at minimum the dead chip can be replaced now. Would be great to get the 25ns chips too allowing clone boards

 

[MacOSMonkey]

Great job, Bolle!

 

[jmacz]

Circling back on this thread. Bolle was able to read the four 15ns parts but unfortunately he has not had luck on the eight slower 25ns parts. He's going to leave it for now but try in the future as he uncovers new techniques to read them. Both daughter cards are alive but hopefully some point in the future he will be able to clone them. We'll see.

 

[Bolle]

One of the accelerator boards will stay with me in trade for some other goodies, so if at some point someone figures out another way to read those slower 25ns chips we can continue right where we left off now.

Just noticed I haven't posted the files publicly yet for anyone who might be interested - let's fix that: