Amiga 4000 corrupted graphics and stability issue

[samaron]

EDIT:
Using DiagROM, I have found that the corrupted graphics occur only when movement happens. Static images are fine, scroll tests immediately throw problems and eventually freezes the system. RAM test in DiagROM show no errors.

Hi!

I put together an A4000TX board that a user under the alias "hese" has created. From my understanding, this is essentially a cost reduced Amiga 4000 in an E-ATX shape and some additional features.

Here's a short TL;DR of my issues:

• Stability issue like random crashes or strange behavior
• Corrupted graphics on screen
• RAM tests report errors on chip RAM randomly

I'll continue with a more thorough explanation now. Please bear with me.

This board has been soldered together by me, and the work has been double checked, and had a couple of other set of eyes inspect it. Everything appear to be in order with regards to the solder work performed.




When soldering together the board, I chose to install sockets for most chips. The chipset is a NOS set except the Super Buster and CIA chips.
Power comes from a new Corsair RM750 power supply, modified with a 15 ohm load on the 12 volt.
Inputs are done with a PS/2 keyboard and mouse.
Software is loaded with an Amiga disk drive.

The board is mostly complete. There are some parts missing, which are the video DAC and RTC chip.
Other optional components missing are the FPU, U180 (which is the logic for larger custom kickstart ROMs) and U893 (which is for fast RAM above 64MB).
Since the video DAC is missing, I'm using a ZZ9000 zorro card to output video.
8MB fast RAM is added to the motherboard slots.
Apparently modern power supplies no longer have -12 volt. This is currently missing as well.

On the first smoke test, the computer had a lot worse graphical issues. It would also crash if I pushed/tapped on the board. This was caused by oxidized pins on the chipset. Removed all chips, inspected and cleaned them under a microscope and reinstalled them.
On the second attempt it worked a lot better and no longer locks up if I push/tap on the board. However, some graphical issues remained. Almost looks like it appears as a pattern on the screen.



Booting a disk to run some test software, I find errors are reported on the chip RAM. Using AmigaTestKit, I find that the chip RAM range show errors randomly. During the RAM tests, graphical corruption appear on the screen, and it will show errors at random points. The rest can report a problem almost instantly, and rerunning the rest, it can do multiple full passes before it reports a problem again. Errors appear randomly on all ranges, both 16 and 32 bit, except D23-D16. The graphical corruption is usually cleared away when exiting the test.





Running the RAM test on the range for fast RAM, no errors are reported. However, the corrupted graphics gradually appears and sometimes AmigaTestKit crashes after a while. Either with a screen showing an exception error, or the background color covers the whole screen.

The boot disk I have is a custom disk where the tools are archived. Without fast RAM, it will report some files as corrupted and fail when extracting. Same files every time. Adding some fast RAM causes it to work every time. This should be able to extract all the contents on 2MB RAM.



I am a bit lost on where to begin the troubleshooting process.
Chip RAM certainly seem suspect, but seems a bit unlikely at least 3/4 of the chips are bad. These are NOS NEC UPD424800-70 chips. Maybe some compatibility issue I'm not aware of?
Perhaps any of the NOS chipset chips are faulty? Or some other chips related to the bus?
Hopefully someone with more experience than me can point me in the right direction. Any suggestions are very welcome and I'll do my best to answer any questions.

[samaron]

Got a pair of ROMs programmed with DiagROM.

I removed the fast RAM from the system and inserted the ROMs. Started the memory test and let it run for a good while. It detects the 2MB chip RAM, and after nearly 60 passes, it still hadn't reported any errors. Much longer than it ever has been able to run in AmigaTestKit.



Tinkering around in the DiagROM, I tried the various graphics tests. Running the scroll test, it immediately started to corrupt the graphics. It would gradually get worse before it eventually froze.



Since it had been running for at least an hour at this point, I did some temperature checks around the board as well. The hottest chips were the CPU and Alice, both measuring around 40-50 degrees. I would expect the CPU to get that warm, but is that normal for the Alice chip?

This reinforces my suspicion that there's nothing wrong with the RAM it self. Reading around, it appears Alice handles things like chip RAM and some graphics operations. Perhaps this chip is faulty in such a way that it moves things around in RAM to the wrong places?

[samaron]

Did some more testing today.

Measured all pins on the chip have contact with the socket. Used a microscope to be sure I were touching the actual chip pins and not the socket itself when testing for continuity.

Since it doesn't appear to be a contact issue, I swapped out the Alice chip with a different one that I were fortunate to be allowed to borrow for testing. This unfortunately did not change anything. The scroll and RAM test did the same as before.

Any suggestions on where to proceed from here? Since the behavior did not change, I am inclined to say Alice isn't the culprit either.

[hooverphonique]

My A4000 has a similar problem.. it's not unstable at all (with fast ram installed), but .e.g. when playing mods, it's like samples get criss-crossed, and there's also a bit of graphics corruption sometimes. I suspect an address bus issue somewhere between chip ram and the custom chips, probably due to a broken trace or via.

[samaron]

Thank you for your reply, hooverphonique.

My board is a brand new PCB that I purchased in February or so. Certainly can't rule out that it has a manufacturing defect with a trace or via. I have to start somewhere, so I suppose probing every address line between the Alice chip and chip RAM is where it happens.

Are there any forums better suited to continue this thread on? Response hasn't been great so far, and would like to get to the bottom of this issue.

Perhaps it would have been a much better option to have bought an original 4000 rather than building a clone with new and NOS parts.

[hooverphonique]

I've heard that the a1k forum is quite good for hw, but is mostly in german, and can be tricky to be allowed in?

My A4k is an original from the 90's, recapped, but has a bit of damage from caps and NiCd, so I'm not sure a new pcb is such a bad idea, actually..

Is your pcb a replica of the original? because then the amiga pcb explorer might be of help..

[samaron]

I have been told about A1K, but it doesn't appear to be welcoming to English speaking users. Could always brute force my way with translation software, but I still find it a bit hard to confirm I have read and understood the rules.

The PCB I put together is not a replica of the original design. It is done as an ATX compatible motherboard. I thought this was the most sensible way to go as I do not own any Amiga 4000 cases or other peripherals. I doubt PCB explorer will be of any help in my case. Although the schematics have the same designations as what is on the silkscreen. At least for the major/important components and jumpers.

[zipper]

Been years on a1k - both in German and then translated to English which works quite well on Chrome.