APC SUA2200RMI2U overload shutdown with minimal loads

[bombsquad]

This was going to be a request for help, but while continuing to experiment and collect information I managed to fix it. For reference, then - a strange failure mode of an APC SUA series rack mount UPS (2010 manufacture, 640-0799H board)

Something similar had been described elsewhere on the forum (see here) but not with any concrete resolution there.

Initial symptoms:
- functions normally in all respects with minimal loads of any type (resistive, inductive, PFC PSUs)
- intermittent high frequency noise/harmonics heard from transformers as load increases near 300W
- very noisy, LED indication of transient overcurrent conditions, oscillating between normal and maximal/overload current readings on the bar graph LEDs, eventually fails safe with the “overload” LED lit.

During initial tests:
- no dodgy-looking electrolytic capacitors or otherwise failed/discoloured components
- MOSFET banks desoldered and tested individually "out of circuit"; all operate normally, resistances/voltage drops fall within datasheet tolerances
- relay coil resistances and operation verified*
- tried two different sets of batteries for the 48V pack, no change
- battery cabling and contacts examined, 8 and 10 AWG throughout, minimal resistance (though I don’t have access to a true low ohms meter these days)
- battery and cables substituted for an SMPS; unable to test at troublesome wattages due to limitations of this PSU but confirmed the increasingly noisy transformers as before
- big transformers x2, coils intact, low winding resistance
- sensing / current transformers T1-3, CT1-2 appear intact cf datasheets, in terms of winding resistance
On the basis of this I cleared the battery pack and cabling. The issue must be the inverter or the onward/output mains AC stages.

Oscilloscopy findings:
- nice PWM inverter output verified at low wattage
- nice clean sine wave output  verified at low wattage
- overlaid high frequency noise seen during fault conditions, both in AC output (high frequency sine on the 50Hz one) and the inverter (really messy)

The images show the extremes of operation. Note the white maths trace on the scope (the CH1-CH2 difference) is the one to watch, it's not quite as nice and sharp as a real differential probe but it illustrates the inverter's ± 48V PWM output normal vs. crazy.

* I think I killed the RY3 (final AC output switching, after boost/trim + inverter/step-up stages) switching transistors when I was testing that relay's operation, so I had to replace both SMD 2N7000s with whatever I had - a couple of BC548Bs don’t look pretty but they do the job

At this point I was a bit stuck - this wasn’t anything simple, unless some passive components were dried out or crappy in a non-obvious way.

The fault condition seemed to be easiest to produce using a server PSU i.e. the intended type of load. I built a ~500W resistance box with some power resistors to test a mainly resistive load (although, as they are wire wound, I guess they're inductors too - who knew?) -- this seemed to get to the highest wattages before becoming unstable.

I tried placing my mains isolating transformer on the OUTPUT of the UPS to test a purely inductive load - and it was able to power the server PSUs on the secondary of this with no problem. This clue got me thinking - given the propensity of such a transformer to remove HF noise, was backchat from the SMPSes actually making its way back in to the UPS, into the inverter, and destabilising it?

I swapped out all the X/Y series AC filtering capacitors and the single electrolytic on the 240V side of the board - and it was fixed. I swapped out all the electrolytics on the board as well, while I was in there. I still have the bag of old capacitors to see exactly which one went bad, but I don't have a capacitance-capable multimeter as yet.

This post is probably describing a similar issue - I only found it while looking up bits for my own post..

A test with the other half's hair dryer, that can do variable fan and heating up to 1600W (thanks Dyson) suggests that the UPS can now happily power almost its maximum wattage without complaint - and no harmonics/noise. Certainly it has no issues with servers, and has run through several discharge tests and calibrations without issue.

Things I learned:
- exhaustive searching of forums can yield clues to one's woes
- test components 'out of circuit' wherever possible
- single ended vs differential measurement and the joys of oscilloscope ground being protective earth: think about what is referenced to what, and whether a current might flow!
- my isolating transformer's output socket earth pin is connected through to the normal protective earth

Casualties of this endeavour: one ground lead for the oscilloscope probe (the inverter had an unscheduled high current output test ) and one screw that I could find the home for when reassembling from the parts tray in reverse order. Given nothing is loose, I'll take this as a win overall.

Hopefully the above is of some utility to someone else in future.

[TheMG]

If I remember correctly from the last time I messed around with one of these APC Smart-UPS, there are a few film capacitors in there that are of particular importance. They are wired in parallel with the output of the step-up transformer, and have the important function of filtering out the PWM switching frequency in order to have only a nice 50/60Hz sine wave at the output.

As they are film capacitors, they lose a little bit of capacitance every time they are subjected to a transient which causes "self healing". Eventually, the capacitance gets so low they can no longer provide adequate filtering especially at higher loads, and having the PWM at the output of the UPS causes all sort of problems for the feedback and other measurement circuits which are normally only supposed to see 50/60Hz.

That being said, where do those transients come from that degraded those capacitors so much? Maybe the MOVs at the input of the UPS are worn and no longer providing protection? Maybe they came from the load side (unlikely if all this was powering was servers and such)? Or maybe some design flaw in the UPS where switching between mains operation and battery operation, or between taps on the transformer, is creating brief transients? Dodgy quality of the film capacitors?

Who knows, but it seems weird to have such a failure as I've seen many various models of Smart-UPS, which basically have very similar circuitry, that are well over 20 years old and still functioning perfectly.

[bombsquad]

Certainly those film caps you mentioned were part of the batch of replacements. I will replace the MOVs in a second round of maintenance.

I noted that the output was always a nice sine wave, but during the fault manifesting itself there was a high frequency sine noise superimposed upon the 50Hz sine.

The unit was a second hand purchase, initially sold as "tested/working" although I think this was probably just a power-on test without a load. I can't speak for the punishment it endured before I got it, though I think it sat in somebody's house for some time - initial cleaning and dusting inside revealed light residue with a certain fragrance reminiscent of deoderants or similar!

[ozorfis]

Awesome guys! My SUA2200RMI2U went into overload just by connecting a switched off (!) ATX power supply and sending it to battery. Resistive loads on the other hand were fine. So I read your posts and pulled the 2 4u7 film caps, which measured 0u3 and 0u8. I replaced them with provisionary caps and it seems to work. Still have to test with a proper load though.....   

Edit: Just tested it under full load and it works!

Thank you