This is not the post I had planned to write.
This is not the board I had planned to repair.
But here we are, let's get on with it.
This is a Commodore PET 3032, sometimes badged a 2001N-32.
The N is important, this is the "New technology" 2001, an almost complete redesign of the PET 2001.
(not to be confused with the 4032N where the N was "Normal" keyboard, and the 8032B was "Business" keyboard)
This has dynamic RAM, 16x 4116 DRAM chips giving 32K.
This has 7x normal 24 pin mask ROMs, 4 or 5 for the OS (BASIC 2 or 4 respectively) and 2 expansion sockets.
(both replacing the rather unfortunate 6540 ROMs and 6550 ROMs of the original 2001).
Or at least it would have those DRAMs and ROMs if they hadn't been removed, along with the 4x 40 pin 65xx chips.
I am afraid I am unsure of the origin of this board. It was in the "untested" pile (actually untested, rather than ebay "untested").
It was either an ebay purchase, a donation, or a part-exchange, but it was a long time ago, so the details elude me.
Oh dear, that's not good.
The electrolyte has leaked out of that capacitor. That doesn't normally happen. I rarely replace caps on boards of this vintage, as they are generally large reservoir capacitors on linear supplies. They only have to deal with 100Hz or 120Hz, rather than the more strenuous tens of KHz in switch modes.
I have not seen one just give up and decide to leak one day. I wonder if it was over voltage?
Musings on a Possible Cause?
I was sent some screenshots of an ebay listing once, a PET for sale in the UK with a US label on the back stating the voltage as 110V. That's fine if you use a step-down transformer, but in the listing in question, the 2 prong US mains plug was fitted with a travel adapter to plug direct into a UK 230V socket.
Nice.
Apparently it was working, although not sure for how long. The transformer would have been getting twice the input voltage, so would be giving twice the output voltage. The 9V that feeds the 7805s to generate 5V would now be more like 18V, and the 16V for the 12V rail more like 32V. The regulators would be cooking, and the capacitors would be over their rated voltage. This one was rated 16V and on the -9V rail, so if it got more like -18V and the regulator next to it was putting out a lot of heat, I can see why it might have given up the ghost and let out the magic juice.
Capacitor Replacement
Yes, that's not good.
It's not like leaky batteries, this generally cleans up fine.
There is a nasty puddle of it under the capacitor.
That cleaned up quite easily. I was trying not to flood the board with IPA as I didn't want to lose the hand-written serial number from 1979.
Speaking of which, 1978 and 1979 must have been tricky years for voltage regulators. That is a 7905, a -5V regulator. It also says 7810, but that is the date code, 10th week 1978, and not a +10V regulator. That same thing must have happened with 7400 series logic 5 years earlier.
The original was 470µF 16V. I have replaced it was 470µF 25V, a nice Nichicon. It never hurts to go for a slightly higher voltage rating. They tend to be physically larger and should last longer.
That all looks quite neat. Right, now why was I looking at this board?
Well, I was in the middle of another repair, and I wanted to check some of the signals on another board, and this one was the same revision.
The only thing was, it also had the same problem.
Well, to be precise, it had the same symptom, but not necessarily the same cause.
Both boards had no video output, no pixel data or vertical sync.
Back to the Other Board
Oh well, never mind, I put this board to one side and went back to the first one.
I eventually found the problem on the first board (more on that in it's own post to follow), but it had several other issues, every time I fixed one thing, I found something else broken.
I decided to leave that first board for a while and thought I may as well have a go at fixing this one.
Back to This Board Again
The video signals on this era of PET were generated from a lot of discrete logic chips, dividing down the 16MHz clock to generate the 15.625KHz horizontal sync and 60Hz vertical sync (all PETs of this era were 60Hz, even the UK ones).
Most of this half of the board is related to generating the video and DRAM refresh timing signals.
The horizontal sync output was correct, but the vertical sync was stuck high.
With something like this, it is best to start at the output and trace back until you find a chip with valid looking inputs, and invalid looking outputs. (or sometimes go the other way, trace the signal from the clock through all the dividers and combinational logic until the signal goes flat).
In this case, there wasn't far to go.
Vertical sync is generated from two halves of a 74LS107 JK flip-flop, via a 74LS00 gate.
I checked the 74LS00 and it's inputs were flat.
I checked the 74LS107 and it's inputs were also flat.
Onto the next page, the next signal, or in this case it's actually called NEXT.
This is generated from a 74S74 D-type flip-flop (shown as 7474 on the schematic).
The inputs to this were wiggling all over the place, but both outputs were flat.
Bingo.
Checking with the scope, it almost looked like the clock signals were always getting blocked by the clear, so I wired up the logic analyser to be sure.
It is not obvious at first glance, but the top chip, G9, is a 74LS74, but the bottom one, G8, the problem chip is actually a 74S74 (Schottky rather than Low-power Schottky).
There must have been a performance gain on the 74S chip that outweighed the extra cost over using another 74LS chip.
The 74LS74 needs a minimum pulse length of 30ns, but the 74S74 fitted only needs 6ns
It did look like the rising edges of the clock were going to be immediately cleared by the falling edge of /CLR, but it was at least 100ns later, so it should definitely have changed the Q output before it was cleared again.
I removed the suspect IC and fitted a socket, since the board already had quite a lot of sockets. It seems to have had a lot of work before. Some a little dubious as there are some bodge wires presumably repairing damage caused with the 74LS153s were all replaced.
Testing the chip in the Mini Pro / T48, and it shows a fail. That's always good to see when you get proof you found the fault (or at least one of them).
I didn't have a 74S74, so I fitted a new 74LS74. It should be fast enough, but I could potentially use a 74HCT74 if it proves not?
I wired it back up to the logic analyser, and there was the Q output, changing when it should.
Yay, it works.
Time to Test The Video Out
I don't like using PET CRT monitors when there are video problems, I don't want bright lines or spots potentially damaging the CRT, so I am using a PET Video Out board to generate NTSC composite video to display on an LCD monitor.
I am expecting a screen of random characters.
Hmm, that's not right.
It is normally random characters with any of 256 values, but there is a distinct pattern there. No inverted characters, no numbers or graphics. Hmm, I think the RAM chip that handles the highest 4 bits is bad. It is a bit unusual, or at least unlikely, that it would randomly pick 1000 values that were all 0-15 from a range of 0-255 (1 in 16 ^ 1000 ?)
Looking at this, the white sockets are original (awful, but original), but I don't think either of the video RAM chips are. The soldered chips are mostly dated early 1979. The MOS 2114 is possibly original, early 1980, but the Fairchild part is from 1983. It is also F7, the one which handles bits 4-7, the one possibly at fault.
A simple test is to swap them around.
Ah, there you go, it is now the lower 3 bits that are stuck low.
If the problem had not changed with the chips swapped, then I would have been looking at the associated buffers and latches (or probably just blaming the white sockets).
It is difficult to find a matching set of 2114 chips that will work fast enough for video RAM, so instead I removed both chips and fitted a 2114 video RAM replacement.
That's better. That is what it is meant to look like.
Testing the Board
Since the ROM and RAM were absent, I fitted a PET ROM/RAM and enabled both.
I also fitted two new WDC W65C21N and a WDC W65C22N, and set it up for testing with a keyboard and datasette and the PET Video out.
Ready for instruction.
A quick test loading from tape.
All looks good.
It initially failed to load from SD2PET, that might be as simple as the edge connector needs cleaning or one of the MC3446 drivers may have failed.
I will address that and do some more testing on this one later, make sure everything is stable etc. But for now the video problem has been fixed and I can leave it running memtest for a bit.
Side Note: 9V PET
It is worth noting, that the power supply on these PETs is two AC supplies, one about 16V, one about 9V. These are rectified, smoothed and feed linear regulators that generate 5V, 12V and -5V. Only the DRAM needs 12V and -5V. The rest of the PET is 5V only.
(other than the datasette which is powered from 6V that is derives from the unregulated 9V rail).
This one has had the DRAM removed and socketed. I presume it didn't come from the factory with a mixture of 16 pin sockets and pairs of 8 pins when the 16 pins ran out.
When you have a PET like this which has the DRAM removed, and you are using a PET ROM/RAM and a PET Video Out, it opens an interesting possibility. You can run it from 9V DC.
I have a PET 2001 board I use for testing sometimes, and I run that from 9V DC, as it never had DRAM.
This board could be used the same, it only requires 0V and 9V on pins 3 and 2 of the power supply connector.
I use a cable I made with a DC jack socket, centre negative obviously.
The bottom 5 pins of the 9 pin connector are the same as the PET 2001. The other pins can be ignored here. (I did make a 9 way version of this cable, but I will not find it until shortly after I put this post live, when it will probably fall off a shelf and hit me)
A setup like this draws about 1.2A at 9V.
You just need to remember to not fit any DRAM chips, but then again, how often are you going to have sixteen working 4116 chips going spare and looking for a home?
Adverts
The PET Video Out, 2114 video RAM replacement and PET ROM/RAM are available form my Tindie Store.
- https://www.tindie.com/products/tynemouth/pet-romram-board/
- https://www.tindie.com/products/tynemouth/pet-composite-video-out/
- https://www.tindie.com/products/tynemouth/pet-2114-video-ram-replacement/
Or, if your PET board has more faults or missing parts than these two I have been looking at, you can always take the easy route and get a Mini PET 40/80 drop in replacement.
If you don't want to use Tindie, you can always contact me using the link at the top of the page and let me know what you want and which country you are in, and I will get you a price if you want to buy direct.
Patreon
If you enjoy posts like these, you can support me via Patreon, and get access to advance previews of blog posts (except when I am running behind, like today, sorry), and exclusive posts (like the extended update I posted earlier in the week).
You also get progress updates on new projects and other behind the scenes updates, as well as access to my Patreon only Discord server for even more regular updates, and to discuss your own projects.
