PDA

View Full Version : Diagnosing A Bad Power Supply



tomasont
January 29th, 2012, 02:37 PM
A recent power outage seems to have damaged one of my power supplies. This was an ATX power supply that was powering an external SCSI array. I probably won't repair the power supply, but I'd like to understand what happened to it.

The power outage was a little unusual. We had two or three quick brownouts, then the power failed completely and was off for several hours. When I powered the array back up, I had one failed drive, but was able to recover the data from the other drives. (Gotta love RAID 5.) When I moved the array over to the workbench, I couldn't get the drives to spin up. (These are 5 1/4 inch full-height Seagate drives.) They would spin up part way, then spin back down. I have the jumpers set on these drives so they do a staggered spin-up, so there shouldn't have been an excessive load on the power supply. My multi-meter showed a proper 12V and 5V output. I replaced the power supply and the drives spun up correctly.

I hooked it up to my power supply tester (one of the simple ones with LED's) and everything looks good. In particular, the power good LED was lit up. I take that to mean that the power supply itself hasn't detected a problem.

Next I hooked up my oscilloscope. I'm a little rusty. I haven't used one in about 25 years. I just bought it a few months ago, so I was looking forward to using it. If I'm reading it correctly, I've got noise on the 12V output that's 264mv at 55.56 KHz. I think the spec for ATX is 50mv, so that's way out of spec.

So here's my question. Is this enough to explain the symptoms I was seeing?

Chuck(G)
January 29th, 2012, 05:42 PM
Well, assuming that you didn't blow a diode in the power supply, the first thing that I'd check are the electrolytic caps on the +12. That ripple points to at least one being bad. You probably were cooking them anyway and the outage was the last straw.

RickNel
February 2nd, 2012, 05:58 PM
Agree with Chuck. If the cap is failing, the PSU won't have enough reserve to smooth over transient high current demands that you are likely to be getting with those 5.25 drives spinning up - they can briefly draw a few amps. You might be seeing reasonably stable voltage on test, but the transients might be dropping just below tolerance for the raid controller or individual drive logic cards. Often a failing cap will show a "doming" or swelling of the top of the cylinder.
Rick