• Please review our updated Terms and Rules here

Uncle Sherman doesn't like the cold

Scruit

Experienced Member
Joined
Apr 9, 2022
Messages
105
Working on "Uncle Sherman", my Mitsubishi MP-3200 386. Trying to run down some memory errors.

1024K of soldered 'base' memory on a daughterboard, and 1024mb upper memory that is socketed / populated. Total of 2Mb.

I have been trying to use CheckIT 3 to figure out why I often get a message "Parity errors detected, run a full scan" when I install the upper memory range, even when I am only testing base memory. I have been keeping a written log of every test, (range being tested, what chips were installed, where, what error message I got, and if I get the "parity error detected" message) Removing combinations of chips will create specific hard faults at particular memory addresses, but if all sockets are populated I get no hard faults.

After several hours of testing, broken into smaller sessions of 15-20 minutes, and being unable to find a pattern for the parity errors, I realized that I sometimes get the "Parity errors detected" when testing video memory. Sometimes not. Then I realized that if I go into the memory test routine and exit back to the menu I will often get a the "parity errors detected" message even if I did not run a memory test. IF I get the parity error message, it's always while going back to the main menu, not during a test.

And then it hit me. The thing I was not logging in my testing was uptime of the system.

The pattern is... If the PC reaches room temperature, then CheckIT3 will always report "parity errors detected" as it goes back to the main checkit3 menu from either "memory test" or "video ram test", regardless of what test I perform or don't perform. After the PC has been running for 10 minutes the parity error message goes away.


So I have two options now...

- It's a simple enough board that I was able to decode it and create a rough schematic. I understand which pins are the address lines for which combinations of chips, how it uses the supporting 74 logic chips to handles /RAS, /CAS, /Write, Data00-15 and how the 74F280's send parity feedback to the motherboard. I can build a test harness using an Arduino Mega to simulate the motherboard and write/read the memory chips and monitor the parity response. I figure this will take a few days to build and likely lead me to at least which column has the parity error. I'd hopefully be able to use heat/cold to make the parity error occur or not. I'd learn a lot about building a complex tester in arduino world.

- I have a desoldering gun, and a T866II+. Every logic chip on the board (11 of them) can be tested using logic check mode on the T866. I could probably do that in an hour. It'd be good experience in desoldering/soldering.

The little adhd kid in me is telling me to build the tester because I'd learn a ton about how these memory circuits work. The old guy in me who is seeing other projects pile up is telling me to take the quicker route and just desolder & test the ICs, then socket them back onto the board.

Thoughts?
 
I would run Memtest x86 next. Checkit can be really strange with memory errors, and report nonsense. Also I would take out each of the socketed chips, clean them off, and reinsert them. Me, I use 0000 steel wool to clean up chips, then hit them good with the air compressor, just to make sure none the steel wool strands short out anything.

As for the steel wool, i tear off a tiny piece and roll it into a ball, then flatten it. Just take your time and go down the legs towards the tip, so you don't catch the legs and bend them. Always works a treat for me. Make sure to wear a ground strap. That is where I would start.

As for the soldered on ram, what manufacture is it? Some brands are known for going bad, aka micron.

Link to older memtest to create a bootable floppy for ya
 
Last edited:
I'll try that once the new power supply is hooked up. I already recapped it last month, and a couple days ago it quit. 😠
 
I tried desoldering one of the parity chips (the ones that calculate parity, not the ones that store it) and I found it tough going with the thickness/layers of the board. As I was removing the chip I realized I was pulling a trace with it. Even going slowly I lifted about 3mm of the trace by the time I realized it, and I stopped to figure out what to do.

The trace was still connected to the leg, so I reversed course and put the chip back down and soldered it back down into place. I verified all the connections are as before and the PC behaves in exactly the same way as before.

So, I can tell that I need to improve my skills first. I have decided to set aside that part of the project until I can practice on hardware that is not quite as irreplaceable. :) I'll get back to it at some point.

The rest of the project is proceeding and I am now able to boot from hard drive and use the PC as normal - except I can't use any upper memory yet. That can wait, I have other projects to work on.

Also I am proceeding with building the memory board tester...
 
Back
Top