Fixing Amiga 4000

[kyberias]

I've been a proud owner of the original Amiga 4000 (040) for the past 15 years or so. Several years ago I stopped using it and stored in a basement. Soon after that I realized the risk of battery leakage and removed the leaking battery from the motherboard. Unfortunately, it had caused some corrosion to the motherboard near the battery location. Namely the shift-registers related to RTC, joystick and that mysterious jumper. I have since removed the shift-register chip related to RTC. Right after I removed the battery I checked that the device booted at least to kickstart. That was about six years ago.

A couple years ago I noticed that the Amiga no longer boots to the kickstart but rather ends up in yellow or green screen, no sure any more. This time I replaced the original PSU with ATX PSU and was able to boot to the kickstart a couple of times but mostly it ended up in yellow or green screen. I've also replaced some capacitors of the processor board. It didn't help.

Today the screen stays black and I would appreciate some advice what to try. I have some basic troubleshooting skills and have access to an oscilloscope etc.

When powering it on with only PSU, keyboard, CPU and memory connected, this is the behavior:

- Black screen with no visible colour changes (no white, gray or anything)
- Keyboard caps-lock flashes during startup and I can switch it on/off repeatedly but it freezes after 6-7. Some people have mentioned that this is a test for CPU. Somehow I believe CPU card is fine. What is the mechanism for this test?
- Lisa generates SCLK to read the shift-registers mentioned earlier (measured with a scope)
- All the other reset seems to be high but _IORST is not stable 5V or 0V but rather ~4V with visible (scope) ripple in the voltage. What could cause this? The signal from the reset circuitry is fine, the ripple is generated elsewhere. Dunno whether this is relevant.
- Resistance between GND and 5V is about 180 ohm without PSU connected.

I've also tried it with another processor card (rev 3.1) with same results. Also, whether I connect the riser card or not does not seem to have any effect.

I would appreciate any measurement-based (voltage, signals) troubleshooting tips. I've tried most mechanical ("push chips down") operations already. I have the schematics available as well. For example, how to test that the CIAs are working etc? Most chips are soldered down and I don't want to replaces them without good reasons.

Feel free to ask any details. I appreciate any help or advice.

[kyberias]

Some additional info: There seems to be some valid video sync signals coming out from pins 10-12 (~15kHz and 50Hz). This is a PAL machine.

[kyberias]

You need to check the area around the RTC quite thoroughly. If they get confused, they can affect the reset line quite easily!
What are the voltages on the board for +5V, +12V and -5V?

Voltages for +-12V and +-5V (5.04V) are good. I think I'll try changing the caps on the MB and test the CPU card on a known good A4k. Meanwhile, I'm still open for any tips or hints applicable to this situation.

Especially details about the caps-lock-test-procedure interest me.

[kyberias]

...Yours probably doesn't do this, as if the reset line is being held low it won't get this far...

I'm not sure what I said, but it seems the reset lines are HIGH not LOW, except _IORST which isn't exactly low either (~4V) but with visible jitter. I'm not convinced it could trigger a reset.

Have you checked all the traces around the SMT chips next to the battery? The tracks can look good but can actually have lost connection from the battery residue.

I should do that more thoroughly.

Does Ctrl+A+A work, by the way?

It does. It seems at least to reset the keyboard so that I can toggle caps-lock again (for 6-7 times).

[kyberias]

clean the contacts on the chip and fast ram sockets and reinstall or replace the ram with good known simms.

Oh, I'm testing it without any memory. I'm expecting to see a yellow or green screen without chip RAM. This is correct?

PS. Ofcourse I've tested it with many different memory configurations but I think it's good to minimize the number of components now.

[kyberias]

4V isn't a TTL level. I'm not sure that's what you want. Having said that I vaguely remember that mine may have been like that, but I'm not sure.

It doesn't always give a green screen with no Chip RAM module in. Test it with a known good Chip RAM SIMM.

Edit: It does sound like the problem I had with a battery damaged A4000. The mouse chips U975 and U976 had a faulty chip, and one of the tracks had dissolved from the battery. Changing the chip made the CLK line tick again, which allowed the system to start working. Take a very close look at the tracks around those two chips with a multimeter, checking for continuity with the other chips in the chipset which use the SCLK line.
What schematic are you using? http://www.amigawiki.org has fantastic schematics.

Thanks for the tips. The ~4V _IORST is certainly something I need to investigate. Are your _IORST lines stable in your Amigas?

SCLK and the returning MDAT seems to be fine. The signal is clear with the correct bits set (used oscilloscope to check that).

The schematics I'm using are pretty complete PDFs and match the mobo revision number, so I think I'm good.

[kyberias]

Certainly replace U975/U976 if there has been any visible corrosion around these shift registers.  I've seen faults where the computer fails to boot because the MDAT line from U975 to Lisa/U450 is floating which causes U450 not to generate certain clocks and the system gets stuck in this weird half-reset state at power on.

If the RTC U178 has been corroded it can cause hangups on the system data bus and prevent booting.  A fairly sure sign is if there's any heat at all detected with U178, which indicates it's defective.  If in doubt, remove both U177 and U178, the computer will boot fine without them.  U975 and U976 are required else the computer can fail to boot, or at best, the system software will fail to read the 'AGA capable' hardware flags as read from U976, so you'll only have ECS screenmodes available.

I've written various repair guides such as the ones here and here which might help further.  Details on the booting process and "caps lock test" here which is written for the A3000, but it's virtually identical in the A4000.

Failing that, I'll be resuming Amiga hardware repairs in November when I'm back in the real world.  Won't be a problem to have a look then if you get stuck.

Valuable tips, thanks! I'll definitely send this junk to you in November if I fail which will quite likely happen.

As mentioned above, SCLK and MDAT look fine and MDAT seems to contain meaningful data (two hard-coded bits and the bits from the undocumented header).

Checking the signals from RTC chip:
- _R/_W signals are not set at any point, so I guess kickstart never gets to read RTC.
- Address lines seem to be OK and there seems to be some activity, but:
- Data lines on RTC look weird with all kind of transient spikes. Doesn't look right to me.

Based on these observations I assume RTC chip might be somehow fried and I will try to remove it to see what happens. Perhaps CPU tries to boot from ROM but gets all kind of garbage from the data bus because RTC. Let's see what happens...