In the last week, an Amiga 1200 was dropped off at my workshop to be repaired. That isn’t unusual, but the faults were unusual enough that I thought I should blog about them.

The Problem

Just the motherboard was brought to me, and I was told that it would work for 15 minutes and then crash to a black screen.

Sure enough, when I tested it, I could only run it for about 15-20 minutes at most. Then weird crashes would happen and it would sometimes generate random noise on the audio. Running Amiga Test Kit showed pretty much everything was fine (until it crashes), but the left audio was very quiet. In addition, the IDE port was not functioning correctly.

Inspection

I was told that this board had been recapped. But, it wasn’t the best work, and some of the capacitors were either originals, or bad. Including the left audio.

I assumed at the time that the gunk was flux. There were a couple of pads missing that had some metal soldered on to work around the problem. The IDE port was also a replacement one, I’m guessing the original was eaten.

After replacing the capacitors, the audio worked fine. I also found a tiny break in a trace just before a plated hole for the IDE, after a quick patch, that worked too. But the crashing was still happening.

Diagnosis

At this point the problem was getting worse, I could maybe run for a minute or two when it was cold, once the machine was warm, it wouldn’t boot at all. Thermal imaging showed everything at normal temperatures though, on both sides of the board.

I tried using DiagROM, but things would crash before I would get useful information. I then decided to use PiStorm with a combination of Buptest and DiagROM.

The advantage of using the PiStorm for DiagROM is that I can completely bypass the Chip RAM, and everything connected to it. Which means I can test that a little deeper.

Buptest is a tool that can be run during the PiStorm’s boot. It basically runs a Chip RAM test from the PiStorm, intended to test that the PiStorm is soldered correctly, but it can diagnose Chip RAM issues as well.

The PiStorm would run fine when the machine was cold, but issues started appearing again when warm, Buptest was a great early indication.

These three all indicated the same thing, a difference of 0x80. Which is an indication of a data line which is stuck low. Due to byte order, in this case it is D31.

This was confirmed when I disabled Buptest and ran DiagROM from the PiStorm.

Sure enough, when running a resistance test between D31 and GND, it was showing as 100ohm and slowly increasing as it cooled. It should be in the mega ohm range. In fact D30 was also very low, just not enough to trigger a problem. This would also explain the audio issues, data shorting to ground would basically be shifting the audio levels around as data was sent. Causing the noise I was hearing.

This led me to think of something: What if the new IDE connector was added but the capacitor leakage was not cleaned first?

I very carefully desoldered it, when it was replaced the first time, the plated holes took damage. Sure enough the sizzle sound and smell when desoldering it indicated there was corrosive material still under the connector. As soon as it was removed the resistance returned to the ranges I would expect.

I cleaned everything up and put a new connector on. From then everything worked perfectly!

I think it looks a bit nicer now too.

The moral of the story: clean up the corrosive stuff the capacitors spew, it can become conductive, particularly when warmed up.