I had similar experiences "back in the day" with PDP-11 based voice banking systems.? Thousands and thousands of TTL chips many with sockets then huge wire-wrapped backplanes.? A nightmare of millions of connections any one of which could turn intermittent at any time.? Countless frustrating hours with the old huge Gould logic analyzers.? The big improvements in reliability of more modern systems came about due to the elimination of so many connection points being replaced by much higher integration in chips.
Not sure if you saw my later post about how removing the fast processor board eliminates the unlock and CPU error messages. There's a lot of interaction and the fast processor has tentacles throughout the instrument.? As a goof I pulled the fast processor address latches (a couple of 74S374 chips) and tried running the thing, but it really didn't like that and hung up pretty early on so it doesn't tell me much.? I put a couple of machined pin sockets in those locations in case I might need to pull those chips again.? I'm not sure this troubleshooting route will be fruitful.
I don't have extenders so probing is very unwieldy.? I have a good setup of scopes and logic analyzers but such troubleshooting can be VERY time consuming.? I have a line on securing a replacement unit and will go that route.? After I do the experiments I needed this for in the first place, I will try board swapping to at least know which board it is, then maybe spot a replacement somewhere.
There is also the possibility things are going wrong because of the bad trace memory chips so I will see if anything is different once the replacements arrive and are installed.
Peter
toggle quoted message
Show quoted text
On 12/16/2024 8:57 AM, Dave_G0WBX via groups.io wrote:
Re:-
I wonder if I could pull them, read them out and reprogram onto new chips.? If one is on the edge that might fix it.? EPROMs do go bad...
Good, if you can be sure you read the EPROM as originally programmed, not while it is in it's faulty state.
As others have said, it could also be a bad bus transceiver/buffer, either not pulling it's output all the way up or down, or "leaking" significantly when it's output is supposed to be in it's high z state, or it's input thresholds have changed.
A memory address decoder problem can cause mayhem too, for example two chips intermittently being selected at the same time will cause trouble.? One of the more difficult faults to trace.
I'd get a decent 'scope, trigger it from a bus access qualifier related with instruction fetching, and look for bus contention, or "weak" signals.? (Analogue 'scopes are often better at that sort of work.? DSO's often lie like a cheap watch, unless you take "Greate Care" to ensure you have one with sufficient sampling rate, and trace memory to catch the fastest signal on the board in full fidelity, often the CPU clock, but not always.? Even then some miss glitches.
If the general logic is all on sockets, just "refreshing" the IC seating can often help, but doing that one chip at a time takes too long, so general the whole board is done and tried again.
At worst, the "fault" might then be permanent, and may be easier to track down.
I learnt that, when I used to fault find to component level on the old Data General Nova 3 and 4 series minicomputers "Back in the day", as well as the third party peripherals they used. Disc controllers and a semi autonomous Mass Spectrometer I/O board.? Often with some 200 TTL IC's on them!? (15 inch square multi-layer boards!)
One particularly troublesome mixed signal board I recall, over many days of odd bits of spare time, to prove a point, I removed each analogue and digital IC and it's socket in turn, and refitted that IC directly soldered to the board in the same position whence it had come from.
The exception was any IC that had a direct connection off-board.? They got expensive turned pin sockets. ? ALL the problems then vanished and never came back!? The result was so profound (the board in question had been officially scrapped as BER. ) The company started making those boards in that style, even using the same style turned pin sockets!? Product reliability with the end customers went up enormously, and our in house service workload reduced to sensible proportions.
Regards.
Dave 'KBV
