Amiga 4000 With Lots of Little Problems

I’ve had a few people send me things in to repair lately. Amongst these was an Amiga 4000 motherboard, RAM and CPU card with lots of minor issues. Here is how I diagnosed and fixed them.

The motherboard

This is what I received. It was reported to have a black screen, the owner had partially recapped it.

Some oddities straight away with this one:

• The LEDs and wires are not stock. Not that it really matters for testing.
• The Fast RAM SIMM sockets were vertically orientated. I’m guessing these were replacements a long time ago.
• It is missing the bulk decoupling capacitor for the RAM area, likely due to the missing ground pad.
• One of the A3640 capacitors had been replaced with a physically bigger one, orientation corrected, but the other two incorrectly orientated capacitors had not been fixed.
• The ROM chips were swapped.
• It unfortunately had the odour of cigarettes and related grime on the board.
• One other problem, see if you can spot it in the motherboard photo before I give a spoiler below.

Diagnosis

I swapped out the ROMs for DiagROM and fixed the other immediate problem I spotted.

This jumper was designed to configure an architecture that was never actually released, an 8MB Chip RAM option is not possible, and selecting it will likely stop the machine booting. So, this was swapped back to the 2MB option.

I removed the customer ROMs and popped in DiagROM, that is when I noticed the next problem.

One of the ROM pins has been bent, likely not making contact with the socket. Not that it would matter too much because the ROMs were in the wrong sockets anyway.

I used Hafnium to test the SIMM sockets, and they all showed as functional. So, I popped the CPU card on, inserted the RAM, wired things up and powered it up… Black screen.

Time to swap things out for known-good parts. First the CPU card. When powering on this time, the boot-time RAM test failed. I therefore swapped the RAM with a known-good SIMM.

That is a great sign, it means that a vast majority of the motherboard is functional. I needed to take a look at the RAM and CPU card.

RAM

I inspected the RAM under a microscope, it looks like a wire brush has been taken to it to remove corrosion, I didn’t see any obvious breaks in traces, and then I spotted this.

Lovely corrosion on one of the RAM chips. My hopes were not high, but I used a fibre pen to clean off the corrosion and reflowed the solder joints. I also used some isopropyl alcohol to make sure the contacts were clean.

Surprisingly for me, this was enough. The machine booted properly like this.

CPU Card

The CPU card definitely needed a recap of the remaining capacitors, so I decided to replace all of them on the card with new ones. It is a good job I did because as soon as I applied heat to the capacitor that had been replaced, fluid boiled out of it.

Whilst I was at it, I replaced the missing capacitor on the motherboard RAM area, attaching the ground of the capacitor to the ground pin of U891 nearby.

Unfortunately, the CPU card still didn’t work. I figured the next step was to clean everything, the grime was definitely in the CPU card connector and it was likely causing issues. I had a collection of boards that needed to go in batches into the ultrasonic cleaner, so these were added.

I removed the CPU from the card so that it didn’t dissolve the thermal glue holding the heatsink on.

Once clean and dried, I connected everything up to test again.

It booted into DiagROM! Fantastic! Next up, let’s try the customer’s Kickstart ROMs.

Bonus!

More testing

From here I ran Amiga Test Kit. I tested the RAM, RTC and chipset, all good. I then tested the audio, the quality wasn’t brilliant, but more importantly, the low pass filter switch was not working.

The capacitors in the audio area had a high ESR, so I replaced these, I also replaced the transistor that handles the low-pass filter switch.

The low-pass filter still didn’t work. I then noticed that the power LED was not changing with the switch. The way it works is that a register is set on the CIA which turns the power LED dim/bright. This signal is picked up by the audio circuit to switch on/off the low-pass filter.

There are really only a couple of things that can cause this. One is a faulty CIA chip, but since the other parts of the CIA were working, I thought this unlikely. The next is a bad 7407 chip which is near the RAM sockets. This acts as a buffer for the LED signals to power LEDs themselves, but if it has failed, it could be back-feeding a signal.

I desoldered this chip to test things.

Well, that’s not good, I replaced this chip. But the power LED, and low pass switch still were not working. Which meant tracing things back a bit more.

The power LED signal also goes to a 1488 chip at U304. This is actually the chip which, amongst other things, takes the power LED signal and buffers it for the audio circuit switch. I wondered if this chip had gone bad and was pinning the LED signal.

You can see how the signal runs here in green. From the CIA to U304 and U311 (which I’ve already replaced). The red is the output of U311 to the resistor for the LED itself.

I removed that chip and powered on the board. I was definitely onto something because the telltale dim ⇾ bright sequence upon boot started to work. I soldered a replacement for U304, and we had fully working audio!

Another problem?!

The owner of the board did not supply any Fast RAM SIMMs, but I figured I should test the area just to make sure it was good. Hafnium showed that all traces were connected up, so I wasn’t expecting any issues. Unfortunately, I was wrong.

The machine booted up fine, and detected the RAM, but when I ran the RAM test in Amiga Test Kit, it failed.

This shows a total failure, which was very confusing. I figured I would test in DiagROM, hoping to narrow it down.

Address errors? But if address lines 16-18 were bad, the machine definitely wouldn’t boot, or it would at least show other signs of failure. I popped in a full 16MB of RAM and ran a different test, this helped give me a clue.

Just A19 bad, that is plausible, I guess. I stuck a logic probe on A18 and A19 of RAMSEY, which is the RAM controller for the Fast RAM area.

OK, something is definitely not right there. A18 looks normal, but A19 has also unmeasurable transitions. A quick test with a multimeter and I found that the A19 pin of RAMSEY was not connected to the rest of A19 on the board. Somewhere in the red trace below, from the ROM chips to RAMSEY, there was a break.

I patched this up underneath the board, and it worked!

With that, everything was working, and the board was ready to go back to its owner.