DEC H7822 power supply

[Peter Coghlan]

I'm having a lot of problems with DEC power supplies.  Newly failed ones
are getting added to the end of the queue quicker than I am managing to
pop fixed ones off the top of the queue :-(

The latest one to give trouble is a H7822 in a MicroVAX 3100.  This machine
apparantly never got deployed for the purpose it was purchased for so it
has had very little use and it is clean and pretty much dust free inside,
including in the power supply.  The two SCSI buses only worked intermittently
when I got the machine first many years ago.  Now, they seem to have failed
completely.  A single common fault seems to be responsible for both failures
but I haven't been able to find it so I use it by accessing disks across the
network instead which is a pity because it has lots of room inside for disks.

A couple of years ago, the two fans in the H7822 (which are the only fans
in the machine) stopped working leading to overheating and shutdown due
overtemperature if it was left on for too long.  This turned out to be due
to a faulty 12V zener diode in the fan control circuit.  I replaced the
zener diode and that seemed to fix this particular problem.

More recently, the machine was working away nicely when it suddenly stopped.
Powering it off and back on later didn't help.  Months later when I finally
got around to looking at it, it still failed to do anything noticable when
powered on.  I took out the power supply, opened it up, checked for blown
fuses or anything else obvious, found nothing, put it back together, applied
power (without a load) and the fans started spinning indicating it now seemed
to be working.  Thinking there might be a short on the system board, I put it
back into the MicroVAX, connected it up and powered it on.  The fans spun,
all the diagnostic LEDs lit and stayed on but the green LED on the power supply
did not light.  The machine appeared to be held in reset.  There's probably
a voltage missing from the power supply I thought.  While I was getting the
meter to check this theory, the green LED came on, the diagnostic LEDs began
counting down and the output of the POST appeared on the console terminal.

While I was making notes so that I didn't forget what happened, I noticed
the fans stopping making noise, the green LED and diagnostic LEDs had all
gone out and the console was no longer responsive.  It had failed again.
I put the meter on the front disk drive connector to check the voltages
from the power supply.  It showed +5V and +12V as if everything was normal.
I put it on the rear disk drive connector and both lines were varying around
one or two volts.  I plugged in an AUI tranceiver which happened to be to
hand and the LEDs on it varied with the meter readings.  I tried connecting a
test lamp to the +5V and +12V lines on the front disk drive connector to load
it a bit.  They held steady while the LEDs on the AUI tranceiver continued to
vary.  It seems there are at least two independent regulation systems in this
power supply.

I left things as they were and went to make more notes.  A few minutes
later, I heard the fans starting up again.  I looked at the machine and the
green LED on the power supply was on, the diagnostic LEDs were counting down,
output was starting to appear on the console and the LEDs on the AUI
tranceiver were lighting normally.

I've tried power cycling it several times since and it doesn't seem to want
to fail now.

Access to the H7822 is not great.  There are two boards (each with permanently
connected daughter boards), one mounted in the base and one mounted in the lid.
The boards are connected together at one end by about twenty flexible wires
and at the other end by two two core cables which can be unplugged to allow
the two boards to be laid out flat end to end.  I suppose I could make up
extensions to the two two core cables to enable better access for testing.
However, it doesn't look easy to devise a layout where everything can be
cooled by the fans for testing over an extended period of time.  At least
the H7822 does not seem to suffer from the leaky capacitor problem like the
H7821 does.

I have another MicroVAX 3100 which looks like it should have a H7822 in it
too but I'm not sure how this could be helpful given the symptoms.

Any suggestions on how to further diagnose this nightmare?

Regards,
Peter Coghlan.