APC Smart UPS SU3000RM Overloaded when started from batteries

[Mikek400]

I have a  SU3000 that works from AC power and inverter will start if you remove power, but it wont start up from battery power.

Originally I didn't have batteries to test it so I used a couple of small power supplies in series to produce the 48V it requires to operate. They were only a 1 amp supply or so but it worked.

Now with nothing plugged in, it draws a lot more amperage on startup from battery and instantly goes to overload..
When using the power supplies I originally used to start it up, it draws too much power from them and they shut down, and it doesn't get a chance to do anything but beep once and the lights blink once.


So obviously the unit is using more power than before and not just a software issue.


I am thinking it may have a bad mosfet.  Cant remove them until I get my hot air station. Does this seem like a reasonable place to start troubleshooting? Maybe the mosfets were damaged from using the power supply to power the unit?


Schematic is attached below for reference

[BradC]

Oooo. My favourite series of UPS.

You want to be looking around the high side mosfet drivers to start with. R38 & R43 are known for cooking and if one pops it often takes out most of the high side bank and the transistors in the driver. Schematic calls for 1K, but I've seen as low as 470 Ohm out of the factory. I always replace them with 1K. This applies to *all* 48V SU series.

You can do the mosfets without hot air. I did the last lot with a spring loaded solder sucker and wick, but a desoldering gun makes it easy. They are held to the heatsinks with self tapping allen screws. So start with the first row and remove all the screws, then go through those holes to take out the second row and so on. That way you can pick and choose what to replace and you don't have to remove the heatsinks. It also means if you can identify them, you can just target which ones you need to remove and not pull all the screws out. You will need long allen keys though.

When my 2200 went it took out all the high side on one side and half of the other. No low side mosfets were injured (I still have a bag of them left over as I had initially bought enough to replace all of them). It did go bang/sizzle though so it was quite obvious something was amiss.

Do replace the aforementioned resistors with at least 1W rated units and stand them off the board. Also the bootstrap caps (C34 & C35) cop a hammering, replace them while you are in there. The schematic lists them as 16V units, but all but one board I've seen as 50V. I replace them with 50V. Check C48 in the AC current sense circuit. I've seen them cause calibration drift as they age.

I also do C111 & C112 (caps in the SMPS for the +24 & +12V rail). Stuff gets wonky as they age, although I've probably only found one that actually needed replacement in all reality. The symptoms are not fun to sort out.

I also replace all the MOVs if I'm going in that deep. These things are awesome at soaking up surges. Might as well tickle it back up to spec.

[Mikek400]

Thanks for the reply!

Unfortunately my Hakko 936 isn't quite up for the task of desoldering the FETs in the smart ups. The traces are  quite large and double sided.  Should be easy enough even with a cheap hot air station though. I had one before but waiting to get a new one now.

Will update once I have some info to share.

[Mikek400]

Looking around I realized I never updated this to say I managed to fix it.

There was a bad electrolytic capacitor hiding that I didn't find originally. Want to say it was a 100uF and was in the circuit for the current transformer. Seems like it measured good but changing it fixed the problem. Now as a matter of course I replace all the small electrolytics in these UPSes when I get them even in working condition.

And after re-reading the reply from BradC sounds like I should be replacing R38 & R43 while I'm at it! MOVs might be fun too just to have a more or less fully refurbished unit.

If you find this looking how to fix yours. Track down a schematic. Even a similar model wil do. After working on many of these they are almost all built the same. Even newer models.