"Peter Hucker" wrote in message news:opsbhnptolaiowgp@blue...

Ok, maybe watercooling the power supply was not a good idea.

It certainly kept the power transistors with the big heatsinks
cool, but what of the poor little diodes and a transformer,
which didn't get the airflow they expected?

Been working fine for a few weeks, then suddenly..... about
15 bangs, flashes, sparks etc flew out of it, as though I had
lit an entire box of fireworks under the desk. Strangely the
PC continued to run while this happened (for about 10 seconds,
at which point the PSU gave up and it went off. Fuses all
intact! Replaced the PSU, and the PC booted ok! Just one
drive of the mirror/stripe appeared to be blank/corrupted,
but it's autorebuilding it in the background.

http://80.229.155.158/temp/psufail

PSUFAIL1.JPG seems to show an LM339, a voltage comparator and not a
chip that normally handles high amounts of power. So I believe it
failed due to excessive voltage, not heat. That's not to say that the
high voltage wasn't caused by excessive heat somewhere else, and the
burned capacitor in PSUFAIL3.JPG could indicate that the main
transformeer got too hot and saturated, which can cause the current
through its coupling capacitor to increase greatly. I'm not sure what
PSUFAIL2.JPG is, but it looks like a transformer, and in PSUFAIL4.JPG,
it's possible that heavily-burned resistor R7 is either a load
resistor (some power supplies won't start without one, and excessive
voltage can burn it out) or part of a snubber (filter to eliminate
unwanted oscillations -- too much oscillation can burn it out).

In the process of testing the water cooling, did you put a temperature
probe on each of the power components? This can be risky because of
the high voltage, but there are probes with metal exposed only at the
tip, or for the more daring a dial thermometer (like a meat
thermometer) can be used if it's covered with a few layers of
heatshrink tubing. Transformer saturation is a big concern among
power supply designers, and heat makes them saturate at lower power
levels.