Amiga 4000 - Corrupted graphics after use of broken CPU-card

[Cav]

Hi,

I have an odd issue on one of my Amiga 4000.
The issue arise after trying a faulty CPU-card in my machine..
(issue is verified to be caused by the card, because same issues happend, but not persistent (phew), in another A4000).

The machine boots OK to workbench (but no picture).
It finishes all memory tests in, CIA, IRQ, disk etc in DiagROM without errors

Raster-test displays ok but without the text.

Audio test with module plays very very very VERY bad with 030 card.
If I use my 040 card it's just quiet.. test-tones is possible to play but seem to be with distortion.

None of the graphics tests works..

The graphics is HEAVILY corrupted... see attached picture which should display the kickstart picture...

So what signal, chip, component could cause this....
Or is it more likely to be several issues?

CPU should be OK .. ?
IDE works OK
Memory is tested OK
Serial port works OK
Keyboard, mouse, joysticks works OK

Graphics and sound seem to be the only things f***d up..

Have no issues with replacing the chips on the motherboard, but would appreciate to be pointed in a plausible direction...

[grelbfarlk]

Sounds like you have multiple issues going on.
More history would be helpful, has the motherboard been recapped?
What CPU card was it, and what condition?

If you've had an A3640 rotting away with leaking caps for some time and tried to power the system up, it's reasonable to expect the rotting caps on the audio amp area are probably damaged as well. Or if it's not an A3640 what was it?

I have an A4000 motherboard which had much of the same problems, I fixed the audio first with Castellen's guide. Then recapped the rest. This still left the video problems which I cleaned thoroughly the 2MB SIMM socket and replaced the 2MB SIMM. Then after loading the system up the video issue started to creep back in and I changed out the power supply with a new one. This was in an A4000 Desktop mounted in an Elbox tower.

[Cav]

Hi,

Thanks for the reply.

The 'funny thing' is that I've managed to reproduce the issue on another board, by using the same accelerator... (A3660)
But it wasn't permanent on that one..

The board is recapped and has been working flawless until use of the broken Cpu card.
But I agree that sound and graphic issue might be two separate issues though.

As the memory test success it shouldn't be the chipram... Or am I wrong? For a while it actually reported 1 mb chip... I then tried another simm (unsure of the size, but it reported 2 MB but a lot of errors, changed back to the original and got 2 MB and successful memory test I diagrom).

I've tried to change the psu, no change.

I've found a spare Lisa for sale that I'm thinking of buying... But I'm unsure of if the issue is related to Lisa, Alice, chipram or something else... + Paula...!?
Kinda lost here..

[grelbfarlk]

Originally Posted by Cav

Hi,

Thanks for the reply.

The 'funny thing' is that I've managed to reproduce the issue on another board, by using the same accelerator... (A3660)
But it wasn't permanent on that one..

The board is recapped and has been working flawless until use of the broken Cpu card.
But I agree that sound and graphic issue might be two separate issues though.

As the memory test success it shouldn't be the chipram... Or am I wrong? For a while it actually reported 1 mb chip... I then tried another simm (unsure of the size, but it reported 2 MB but a lot of errors, changed back to the original and got 2 MB and successful memory test I diagrom).

I've tried to change the psu, no change.

I've found a spare Lisa for sale that I'm thinking of buying... But I'm unsure of if the issue is related to Lisa, Alice, chipram or something else... + Paula...!?
Kinda lost here..

I'm not sure what memory test on chipram you're using specifically but I used MBRTest-2 on my Chip RAM when I'm pretty sure it had some problem and it still passed.



If you're sure all this is correct and the Accelerator nuked a motherboard, I would suspect an issue along the data or address bus. First easiest thing to check is the row of 2.7k pullup resistors next to the CPU socket.



After that the check the Address and Data bus pins from the accelerator socket to some point where they terminate on the other side of the board, after you check continuity, check resistance to 5V and Ground. Maybe you'll find a short somewhere. Though this is really time consuming, that's 192 measurements, more if you check multiple points.
It might be even more helpful to do this with the Accelerator installed, you might find exactly what shorted that way, while the damage is not easily detectable without the Accelerator installed. Of course you're doing all of this with the machine powered off, so further damage isn't likely.



Break down the issues into separate parts and start with say, fixing audio. Audio issues are generally pretty easy, this is the most bestest guide by Anthony Hoffman/Castellen:
http://amiga.serveftp.net/audio_repair.html

[hooverphonique]

Note that even if a piece of software successfully tests chipram (using the cpu), there could still be a problem with custom chip chipram access.

[Cav]

Originally Posted by grelbfarlk

I'm not sure what memory test on chipram you're using specifically but I used MBRTest-2 on my Chip RAM when I'm pretty sure it had some problem and it still passed.

I've been using DiagROM. As i don't get any video-output that's my only option..

Originally Posted by grelbfarlk

If you're sure all this is correct and the Accelerator nuked a motherboard, I would suspect an issue along the data or address bus. First easiest thing to check is the row of 2.7k pullup resistors next to the CPU socket.

I've measured all the 2.7k pullup's and they are ok.

Originally Posted by grelbfarlk

After that the check the Address and Data bus pins from the accelerator socket to some point where they terminate on the other side of the board, after you check continuity, check resistance to 5V and Ground. Maybe you'll find a short somewhere. Though this is really time consuming, that's 192 measurements, more if you check multiple points.
It might be even more helpful to do this with the Accelerator installed, you might find exactly what shorted that way, while the damage is not easily detectable without the Accelerator installed. Of course you're doing all of this with the machine powered off, so further damage isn't likely.



Break down the issues into separate parts and start with say, fixing audio. Audio issues are generally pretty easy, this is the most bestest guide by Anthony Hoffman/Castellen:
http://amiga.serveftp.net/audio_repair.html

Thanks for link and ideas!
I will try to check all traces on data and adress bus..
But one odd thing that bothers me is that the same issue occurred (temporary on my other board) which leads me to believe it was an IC that has been damaged and maybe just overheated or something on the other board, as it wasnt permanent there? Just a thought...
That's why I'm thinking of replacing Lisa / Alice..

But now I will give the audio-part a try!

Again, big thanks!

[Cav]

Originally Posted by hooverphonique

Note that even if a piece of software successfully tests chipram (using the cpu), there could still be a problem with custom chip chipram access.

That sounds like a very interesting idea as audio and graphics are corrupt...
But how would I intercept this idea... ? :/

[grelbfarlk]

Originally Posted by Cav

That sounds like a very interesting idea as audio and graphics are corrupt...
But how would I intercept this idea... ? :/

Well sure, what he means maybe is that the Analog output could be broken or corrupted in some what that the CPU accessing the ChipRAM would not be aware of a problem, the ChipRAM contents are still going to Lisa and Agnus so the memory test might not detect an error from the CPU's perspective. Though if it is also affecting Paula, those are separate Digital to Analog paths on the A4000 motherboard.
For instance an error on U455, U456 and U457 can corrupt the output but the CPU will still think the ChipRAM access is ok.
Maybe look into the DRD(31:0) signals those go from Bridgette to Lisa and Paula (and Alice).

But without more info on what kind of output you're getting it's hard to say.



Is it possible you didn't change the jumpers on the motherboard for INT/EXT clock when you mounted the A3660? Was it running at 66MHz or 50MHz?



How about a bunch of high resolution pictures of the motherboard, with some close ups of the whole Accelerator side and the Custom chip side.

[Cav]

Originally Posted by grelbfarlk

Well sure, what he means maybe is that the Analog output could be broken or corrupted in some what that the CPU accessing the ChipRAM would not be aware of a problem, the ChipRAM contents are still going to Lisa and Agnus so the memory test might not detect an error from the CPU's perspective. Though if it is also affecting Paula, those are separate Digital to Analog paths on the A4000 motherboard.
For instance an error on U455, U456 and U457 can corrupt the output but the CPU will still think the ChipRAM access is ok.
Maybe look into the DRD(31:0) signals those go from Bridgette to Lisa and Paula (and Alice).

But without more info on what kind of output you're getting it's hard to say.


Is it possible you didn't change the jumpers on the motherboard for INT/EXT clock when you mounted the A3660? Was it running at 66MHz or 50MHz?


How about a bunch of high resolution pictures of the motherboard, with some close ups of the whole Accelerator side and the Custom chip side.

Hi,

Unsure if I temporarily forgot the jumpers. Might have done.. would'nt have been the first time while switching between 030/040. Are there cases when this have caused problems except failure to boot in wrong "mode" ?

I've only run it in 50 MHz.

I've shared some pictures below.

Motherboard pictures: (a little jumpwire on the DiagROM because one leg is broken)

https://photos.app.goo.gl/y1NoWZU9xHmTkoht7

Video of DiagROM: (as you can see the only thing displaying correct is the raster test...?)

https://photos.app.goo.gl/bUuCMr9ieC3kEFpi8

Video of chipram test:
https://photos.app.goo.gl/Pkab7gs8eVS4oagZ6 (start)
https://photos.app.goo.gl/Y9iRHBAB6atH7vcC9 (finalize)

Video of IRQ and CIA test:
https://photos.app.goo.gl/Wvs9joLdtuuQS1Eu6

Video of kickstart animation:
https://photos.app.goo.gl/8bST5Gnsxr3RGiiy6

Video of booting to Workbench:
https://photos.app.goo.gl/HT52yUrAiTAemfz96

[Cav]

@grelbfarlk and others!

I REALLY appreciate your effort to help!

[grelbfarlk]

Originally Posted by Cav

Hi,

Unsure if I temporarily forgot the jumpers. Might have done.. would'nt have been the first time while switching between 030/040. Are there cases when this have caused problems except failure to boot in wrong "mode" ?

No, not really not with an A3640 at least.
An A3640 with a 66MHz clock will run the motherboard logic quicker, which could -but-is-really-unlikely-to- cause lasting damage, but if you only ran at 50MHz then it's probably ok.

I notice some odd marks on Bridgette is that chip getting hot by chance?

The trace to the via that is right next to R176 and C198 appears to be gone, is it?

[Cav]

Originally Posted by grelbfarlk

No, not really not with an A3640 at least.
An A3640 with a 66MHz clock will run the motherboard logic quicker, which could -but-is-really-unlikely-to- cause lasting damage, but if you only ran at 50MHz then it's probably ok.

I notice some odd marks on Bridgette is that chip getting hot by chance?

The trace to the via that is right next to R176 and C198 appears to be gone, is it?

The marks on Bridgette was just glue left overs... removed it now.
And the trace is dark... but still ok.
Compared it to an old picture I hade on the motherboard from where i recapped and cleaned it and back then it was green ...
So I figure I missed some electrolyte when I did it..
Doing another clean now and will lift and check the traces to U402 + U403..

Sharp sighted man!

[Cav]

Cleaned up and checked traces to Q400 + Q440 + U402 + U403.
Without U402 & U403 the behaviour is same but lower volume. Same distortion.

All traces is ok.
Haven't had chance to change the IC's though.

[grelbfarlk]

Sorry I was wrong, I just don't see how this is visually possible in the little yellow arrows, this black line looks like the depression left by the the trace.
Measure the voltage on the red circle then on the blue circles. Check continuity between blue/red and blue/blue.

The larger orange arrow on the left points to another trace that looks not too good.

[Cav]

Originally Posted by grelbfarlk

Sorry I was wrong, I just don't see how this is visually possible in the little yellow arrows, this black line looks like the depression left by the the trace.
Measure the voltage on the red circle then on the blue circles. Check continuity between blue/red and blue/blue.

The larger orange arrow on the left points to another trace that looks not too good.

Red to blue is 0v.
Red/blue to GND is -12v (-11.86v)
Continuity between them is OK.

Can't explain the color of the trace but if you look at the pictures attached it is easier to see that it's not a depression, rather a misscolouring and something that looks like glue. (it doesn't seem to be flux because i cannot wash it away with isopropanol)

[Cav]

The signals DRD(31:0) is common between CHIPRAM, BRIDGETTE, LISA, ALICE & PAULA.
I'm about to go through continuity on all this signals.

Does it make sense or is it more likely to be a chip(s?) failure in regards to the characteristics of the issue... ?
Suspecting: U891-894, U212, U213 or U216.. Plausible?

Edit:

And that's done...

DRD(31:0) - (PAULA, ALICE, LISA, CHIPRAM & BRIDGETTE)

31 - OK
30 - OK
29 - OK
28 - OK
27 - OK
26 - OK
25 - OK
24 - OK
23 - OK
22 - OK
21 - OK
20 - OK
19 - OK
18 - OK
17 - OK
16 - OK


RGA(8:0) - (PAULA, ALICE & LISA)

8 - OK
7 - OK
6 - OK
5 - OK
4 - OK
3 - OK
2 - OK
1 - OK

[grelbfarlk]

Originally Posted by Cav

Red to blue is 0v.
Red/blue to GND is -12v (-11.86v)
Continuity between them is OK.

Can't explain the color of the trace but if you look at the pictures attached it is easier to see that it's not a depression, rather a misscolouring and something that looks like glue. (it doesn't seem to be flux because i cannot wash it away with isopropanol)

Yeah that is much clearer.



Ok how about the -5V output of the VRM, the pin just below the red circle?

[Cav]

Originally Posted by grelbfarlk

Ok how about the -5V output of the VRM, the pin just below the red circle?

Pin 1 is 0V, GND
Pin 2 is -11.85V
Pin 3 is -4.95V

In regards to the characteristics of the issue and how it arise (and temporary showed up on another board), would you point your suspicion in the direction of a chip or a trace...?

I am suspecting: U891-894, U212, U213 or U216.. what's your input on that?
Going to order those chips that are available as common parts and make an exchange..

[grelbfarlk]

Originally Posted by Cav

Pin 1 is 0V, GND
Pin 2 is -11.85V
Pin 3 is -4.95V

In regards to the characteristics of the issue and how it arise (and temporary showed up on another board), would you point your suspicion in the direction of a chip or a trace...?

I am suspecting: U891-894, U212, U213 or U216.. what's your input on that?
Going to order those chips that are available as common parts and make an exchange..

I would say: sure it's probably one of the 74F245s or one of the other common (non custom chips) but chances are just as good a resistor or a cap burnt out somewhere along the way.

That A3640 doesn't look like it's in particularly good shape, could it have leaked electrolyte in the accelerator socket? That's even possible and transferable between machines-in theory. But if it's on a new build of A3660 that you're talking about that caused the issue between the two machines, then you could at least rule that out.



Another possibility is one of the PAL/GAL chips on the custom chip side of the board, but those are common enough and the code is available to burn replacements.

[Cav]

Will start with the 'easy ones'..
The 3640 is from a third A4000 and is working flawless since I Recapped and repaired it. (broken trace).
The issue is because of the 3660. And as I said, the exact same behavior showed up in my other A4000 when I plugged it in... Luckily it wasn't permanent there... Maybe a temporary overheat...? :/

I'll start with 74F245 and try to see if I can source the gals...