Lou - N2MIY
Veteran Member
The Chrislin CI-1173 is a very nice 2MW QBUS memory. Here is a good overall picture: http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44991 . All the talk a few weeks ago about qbus memory made me want to make an inventory of my loose memory boards. I came across this one that Will gave me a few years ago. It was broken as long as I've had it. But wow, it is fully loaded with 2MW. I decided I should try to fix it.
There is almost nothing on the internet about this memory. There is a bit of text from an announcement in an old computer magazine, and some of the resellers have this board, but there is nothing else. No manuals, no prints, no nothing. However, it looks (and is) very simple. There are no gate arrays, it's only bus transceivers, DRAM drivers, and 7400 series TTL. I figured I was up for the challenge.
The behavior seen reading and depositing in ODT made me encouraged that I could fix it. It behaved as if the lower nine address bits were stuck. The lowest bit is not so special since the memory is 16 bits wide and words are handled as words, so it struck me as signficant that the eight (useful) low order bits were stuck. When I would read or write, one entry appeared to affect 1000(o) words.
So I decided to get into the job, starting with a thorough visual inspection with the magnifying glass. There was nothing bad, so I decided to get the lay of the land by looking at the organization of the devices on the board. The memory covers 3/4 of the board, and all of the logic is by the A fingers. See here: http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44994 . The memory drivers are the three AM2966s that drive the DRAM RAS and CAS lines. The bus transceivers are the AM2908s by the B fingers. There are some other 8640/8641/8837 bus drivers/receivers around. Of course, 373, 244 latches, 682 comparators by jumpers (must be for starting address selection.) So I put the card in a card extender, in a machine, and started scoping with the scope, first looking at the AM2966 outputs to see what looked stuck. I quickly zeroed in on the 2966 closest to the fingers as being awfully quiet. To cut to the chase, I scoped upstream, moved to the logic analyzer, pulled the card and reverse engineered my way back to the BDALnn bus lines (of course focusing on BDAL1 through 8 ). Once received by the 2908s, four each go through two 74LS197s being used as latches (I still don't quite know why.) The data is then latched into a 373 (the one in the picture with "C" on it in pencil. I labeled the ICs for my sketches as I reverse-engineered). Then from that 373 to the 2966. So with the logic analyzer it was clear that everything came to a screeching halt at the LS197s. (http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44993).
Watching the load and clear on the 197s (which are tied together between both of them) the problem became evident on the logic analyzer http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44988 . Sure enough, load and clear were shorted together somewhere, as confirmed with the continuity checker on the VOM. A few nights later I had chased the traces everywhere that the load and clear went on the whole board.
Wouldn't you know that they both arrived at two adjacent pins (2 and 3) on a 74LS123. Yup, the pins were shorted inside the IC (not supposed to be that way!). I wish I could have it x-rayed, maybe a wire bond failed at the die and sprang across? So, we replace that IC and bingo, it's all good. http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44992 .
It took only a few minutes to read this, but two weeks worth of evenings to actually finish everything. I did go down some other ratholes, but learned things along the way so it was not all for naught.
I thought I would share the experience since I am a fixer and not a board swapper. And, this is a hobby so it's OK if I blow two weeks of evenings.
Lou
There is almost nothing on the internet about this memory. There is a bit of text from an announcement in an old computer magazine, and some of the resellers have this board, but there is nothing else. No manuals, no prints, no nothing. However, it looks (and is) very simple. There are no gate arrays, it's only bus transceivers, DRAM drivers, and 7400 series TTL. I figured I was up for the challenge.
The behavior seen reading and depositing in ODT made me encouraged that I could fix it. It behaved as if the lower nine address bits were stuck. The lowest bit is not so special since the memory is 16 bits wide and words are handled as words, so it struck me as signficant that the eight (useful) low order bits were stuck. When I would read or write, one entry appeared to affect 1000(o) words.
So I decided to get into the job, starting with a thorough visual inspection with the magnifying glass. There was nothing bad, so I decided to get the lay of the land by looking at the organization of the devices on the board. The memory covers 3/4 of the board, and all of the logic is by the A fingers. See here: http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44994 . The memory drivers are the three AM2966s that drive the DRAM RAS and CAS lines. The bus transceivers are the AM2908s by the B fingers. There are some other 8640/8641/8837 bus drivers/receivers around. Of course, 373, 244 latches, 682 comparators by jumpers (must be for starting address selection.) So I put the card in a card extender, in a machine, and started scoping with the scope, first looking at the AM2966 outputs to see what looked stuck. I quickly zeroed in on the 2966 closest to the fingers as being awfully quiet. To cut to the chase, I scoped upstream, moved to the logic analyzer, pulled the card and reverse engineered my way back to the BDALnn bus lines (of course focusing on BDAL1 through 8 ). Once received by the 2908s, four each go through two 74LS197s being used as latches (I still don't quite know why.) The data is then latched into a 373 (the one in the picture with "C" on it in pencil. I labeled the ICs for my sketches as I reverse-engineered). Then from that 373 to the 2966. So with the logic analyzer it was clear that everything came to a screeching halt at the LS197s. (http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44993).
Watching the load and clear on the 197s (which are tied together between both of them) the problem became evident on the logic analyzer http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44988 . Sure enough, load and clear were shorted together somewhere, as confirmed with the continuity checker on the VOM. A few nights later I had chased the traces everywhere that the load and clear went on the whole board.
Wouldn't you know that they both arrived at two adjacent pins (2 and 3) on a 74LS123. Yup, the pins were shorted inside the IC (not supposed to be that way!). I wish I could have it x-rayed, maybe a wire bond failed at the die and sprang across? So, we replace that IC and bingo, it's all good. http://www.vcfed.org/forum/album.php?albumid=406&attachmentid=44992 .
It took only a few minutes to read this, but two weeks worth of evenings to actually finish everything. I did go down some other ratholes, but learned things along the way so it was not all for naught.
I thought I would share the experience since I am a fixer and not a board swapper. And, this is a hobby so it's OK if I blow two weeks of evenings.
Lou
