I post this here, not as a question but more of a note for "future me" when I encounter this one again and have long forgotten the why's and wherefores.
My trusty SU2200 shut itself off a week or so ago with the classic "Relay Weld" fault. It's 19 years old, so I decided to treat it to a new set of relays all round. While I was in there I replaced the remainder of the electros I'd not done a couple of years ago when I did the movs.
Imagine my surprise when firing it up and it happened again immediately. Brain-dead reset and it was all working. That lasted an hour and it happened again.
I then had a look at the last years power graphs and noticed in late April this UPS started reporting voltages 10-20V lower than the actual input. I bodged up a script to compare the UPS reported line voltage against the local power meter every minute and noticed that it would wander around some +/- 20V over a 24hr period and it was getting slowly worse by the day.
So the UPS measures the input voltage using T1 and the output voltage using T2, compares them and if they meet a certain criteria it assumes one of the AVR relays has welded and shuts it down. Nice. It doesn't actually have to be a welded relay to trigger the fault, it just has to look like it.
Putting a scope on the output of T1 and comparing it against an AC plugpack showed an odd clipping/sagging on the trailing edge of the top of the sine wave and there was no rhyme or reason to it. Skip forward past hours of pulling parts off the board to isolate the transformer to just having 240V in and a CRO on the output and the problem remained.
So I pulled the transformer. Primary open circuit. Not really open, just more than my handheld meter could measure (>20M). If you put 240V across it it conducted enough to put a pretty impressive sine wave on the output, with the odd result of the trailing edge sagging badly. This was throwing off the line voltage measuring circuit causing the UPS to believe the incoming mains was anywhere +/- 40V from where it really was, and comparing that against the correct output voltage. Thus triggering the relay-weld.
Lucky I had an old SU700 of the same vintage handy which has the same pair of transformers. Swapped T1 and we're back in business.
Correlating the fault against the environment would indicate that it was responding to huge local swings in humidity (the UPS is outside) and for whatever reason a break in the primary was conducting enough to be mostly useful. It has been operating in this state for some 4 months now until it finally gave out.
A measurement on a good bench meter shows the primary at ~147M Ohms to a couple of volts DC, but put 240V on it and it mostly works. A most frustrating and insidious fault, and one I'm ashamed to admit took me a couple of sessions to nail.
These are custom parts for APC by LEI in Taiwan, but it looks very much like mostly-equivalents are available at Digikey, so I might try one to see if I can resurrect the SU700 when I next need to put a parts order in.
Choose a location