• Please review our updated Terms and Rules here

First PET post of October. 4032 with dead RAM subsystem

Witchy

Witchy

Experienced Member
Joined
Oct 11, 2015
Messages
376
Location
Flatlands, UK
Hi folks,

Long time no post, life and non-PET projects got in the way. I'm slimming down my PET collection because nobody needs 12 and I'm moving house, and was just going to list them on e*ay as faulty/untested.

Then I realised my repairing conscience wouldn't let me do that* and have already fixed up an 8096 (dead Editor ROM) and discovered to my surprise a spare 3032 worked fine without me doing anything to it. The next one is a 4032 with 40 col 8032080 board, and after replacing the white CPU socket I fitted my Tynemouth Software ROM/RAM board and slowly eliminated faulty items until it's now just providing the lower 16K RAM to boot, so I have a RAM fault.

Half of the RAM chips are socketed already so I've tested those in my 4116 RAM tester and also cleaned up and refitted R10 and R11 which had been badly butchered by whoever worked on this last. Both CAS0 and CAS1 measure 27ohms back to those resistors from each chip so they're still good despite how they look. The power is OK at 12.05V, -5.06V and 4.88V and there's continuity along both banks though I've just realised I haven't tested back to and beyond UE8/9/10 for the address bus. There's a task for tomorrow.

I've watched CAS/RAS/WE/DI/DO with a logic analyser and they all seem stable with no dead spots.

I feel lazy asking for help at this point but I bow to the superior board experience of people here. I put DaveR's excellent PETTEST V4 onto a 2516 and it's showing 'b' for 99% of the memory locations but crucially not all of them. Just for the hells of it I enabled the lower 16K of the ROM/RAM board just to watch PETTEST complete :D

Any tips for where to look next before I start checking the LS244s for dead gates?

Cheers!

Witchy



*apart from an 8032 that's missing half a dozen chips and has a few lifted traces plus cracked CRT neck, hate to think what the previous owner did to THAT one
 
Thanks for the vote!

Any chance of posting a photograph/video of your G/B screen? With a knowledge of what is stored where and the results it may be possible to work out whether you have a data fault or an address fault and where they are. Since the first tests only check pages 0 and 1 all address faults can't be detected at this time though.

Is the G/B display static or do some of the results change as the test loops?

One possibility is the address multiplexer chips as well as the data buffers.

Another section to add to the documentation...

Dave
 
You could "cross" R10 and R11 (lift one end of both and swap their end's connections to the pads) so you could identify if you have a good 16K bank or both are faulty. If you have a good bank, then you can boot off of it and write a simple RAM test in basic to identify the bad bit(s).
I've done more or less this here:
https://www.youtube.com/watch?v=H5_mTETwWJk
 
If you look at how R10 and R11 were when I first looked at the board at the weekend (first pic) I'd guess someone already did that and couldn't be bothered to put things back together properly. This machine did work when I got it back in the early 00's but obviously that was a long time ago in terms of silicon not dying in the meanwhile.

Dave, yes the display changes while the test loops but only for areas that initially show as 'g' which as you can see isn't a whole bunch. Middle pic is the tidied up R10/11 and new Kernal.

IMG_4645.jpgIMG_4662.jpgIMG_4670.jpg
 
Another possibility you have is to run a diagnostic image in place of the kernel ROM if you can burn your own 2532.
Frank
 
I'm already running Dave's PETTEST V4, that's what's produced the third pic in my previous post showing bad RAM everywhere, pretty much :) I'm now going through the schematic to look at likely culprits for producing this fault while at the same time thinking that the first PET I fixed back in 2015 had this exact problem and it was down to a 74LS00 or similar. I also did a breakdown image of the universal board showing which chip did what and to whom, need to dig that out too.
 
Witchy said:
I'm already running Dave's PETTEST V4, that's what's produced the third pic in my previous post showing bad RAM everywhere, pretty much :) I'm now going through the schematic to look at likely culprits for producing this fault while at the same time thinking that the first PET I fixed back in 2015 had this exact problem and it was down to a 74LS00 or similar. I also did a breakdown image of the universal board showing which chip did what and to whom, need to dig that out too.
Click to expand...

on a "FAT40" universal board, I've been luckier with this diagnostic:
https://drive.google.com/file/d/1bPR4ZWENU6_TbtqdkT3_W0AuK3l5Ewf7/view?usp=sharing

It goes in the kernel position.
YMMV

Frank IZ8DWF
 
Witchy,

Thanks for uncovering a bug in V4! I will fix it at the weekend...

The '.' should change to a character that depends upon the value that was actually in the memory cell that was read if the value read incorrectly. It does actually do this... but then overwrites the value with the '.' in error :-(!

Dave
 
My apologies for that...

I have uploaded a fix to my google drive.

It runs under VICE - but I haven't had time to test the fix with a duff memory location though - I have to persuaded VICE to do that and I need to remember how to do it...

Dave
 
Well that was a good evening's work thanks to both you and Frank! PET is alive again and running without the ROM/RAM board for the first time :D

Only on 16K though, Frank, is that the official CBM diagnostic for a 40xx? It mentions 'removing the clip' so I assume it is.

Dave, I'll download the new V4 tomorrow and run it up, given that 4x 4116 had failed in bank 0 I'll be expecting a similar number in bank 1!

Night all.
 
Early start today because I'm at home doing an HPE course so during breaks I get to keep working on the 4032, excellent. First problem discovered in the cassette circuits. First off I hooked up my Pet MicroSD+ which takes 5V from the rear cassette port and that works as expected. There's a RAM tester on there which for some reason only goes up to 16387 and it fails at 16385/6/7 on bits 0 and 1 no matter what I do to those two locations on the board.

No matter, I have a RAM test on tape so I connected a known working C2N to the right hand cassette port and the motor starts but tape doesn't read. I know both tape and C2N are good because I was using them last weekend. Remove the MicroSD+ and put C2N on the rear cassette port and boy did something smoke! Powered off immediately and did a finger test on the two transistors as well as all the chips and nothing was hot to touch despite SOMETHING obviously pulling far too much current.

Disconnected the C2N and powered up again, all is fine. Voltage checks on the 3 IEEE chips etc show nothing is awry so I hook up the MicroSD+ again and that works too, so I suspect the smoky component is a resistor and I'll need to do more testing later.

Nice way to start the day, the house stinks now :shock:
 
OK, embarrassed face time :oops:

While searching for 'why did my cassette circuitry go up in smoke' I found a post from JonB back in 2016 which I was part of but never saw the last page. There he says he'd smoked a trace under the board by not having it secured while plugging cassette players into it and it had caused a short on the unregulated 9V rail.

Well.

My board is now screwed in while testing and the smoked trace is good. Ahem.

ANYWAY. The short RAM test which happily goes up to 32767 on an 8032 barfs at 16384 on this machine which implies that the upper bank is missing altogether. Time to do the R10/R11 swap and see what the diags give me.
 
Another 3-and-a-bit hours' work and it's up and running again :D

First pic is what happens when you accidentally ground the 9V line. Fortunately it cleaned up OK. 2nd pic is first power up with all its own chips again, 3rd is obviously FUN TIME. After I'd stripped and cleaned the keyboard.

Total replacement parts - 1 Kernal ROM, 1 CPU socket and 8x 4116s. There's only one original one left in there now and I'm pondering whether to replace that too.

Thanks for the help chaps. Sorry I didn't get to test your revised tester with this one Dave, but tomorrow I'll have another machine on the bench so I'll hopefully be able to use it then.

IMG_4695.jpgIMG_4702.jpgIMG_4714.jpg
 
You must log in or register to reply here.
Back
Top