[solved] Powercom KIN-1000AP UPS repair, help needed

[shapirus]

So today my good old beloved Powercom KIN-1000AP decided that enough is enough and failed after 20 years of honest work.

What happened: after some ~12 hours of it being powered off (but connected to the wall outlet), I turned it on and waited, as usual, for the power-on battery test to start and finish. This time it took less time than normal. Then I turned on the load (a desktop PC). Everything seemed normal until suddenly it cut off the load and lit the "battery fail" LED. I then tried to turn it off and on again and it started to make unusual sounds and finally started to let the magic smoke out, at which point I turned it off and disconnected from the mains.

On further investigation I found out that when I tried to cold-start it (battery only, no mains power), it would begin its battery test, the inverter started operating, but current drawn from the battery (a 8s3p 32650 lifepo4, nominal 24V) was huge, above 100 A, which is basically a short circuit current.

Then I found out that the Q7 mosfet (IRF540N), which can be found in the lower right corner of the attached schematic diagram in the charging circuit, was shorted between every pair of pins. I don't have this same transistor at hand, so I picked a close enough replacement (lower Rds(on), same Ciss, lower, but sufficient, Vds(br)), but, alas, the new transistor was blown the same way, shorted on power on (again followed by the excessive current draw from the battery during the battery test).

Finally, after some more attempts, the fuse F1 failed open. At this point the UPS does not attempt to start the inverter, just clicks some relays, powers on the VCC (which is +5V) and +12V rails and powers off again. I have not tried to replace the fuse yet.

I also checked:

- output of U2 (L7812) is correct: 12V
- same for U1 (L7805): 5V
- U7 (ULN2004A) appears to be working properly, at least as far as the behavior of pins 7 (input) vs 10 (output and gate drive of Q7) goes; at some point gate voltage goes high for a brief period before the UPS powers off
- inverter output mosfets are fine, or at least not shorted
- nothing suspicious was found on visual inspection: no bulging caps or charred parts etc.

My next idea would be to remove Q7 altogether and try to power the UPS on without it, but I need to replace the F1 fuse first. My assumption is that this transistor gets turned on at an inappropriate time (e.g., when the charger secondary winding of the transformer is energized), conducts a very large current and gets fried. Why, I can't understand (other than some failure of the MCU).

The thing I don't understand is the purpose of this transistor.

Why in the world would anyone want to short the output of the D2/D3/D8/D9 diode bridge effectively to ground, save for the small 0.82Ω current sense resistor R51? CN5 is where one of the secondary windings of the transformer is connected, the wires are coming directly from it. As far as I can tell, there is always ~32V AC RMS across it, when mains power is present. I am not sure, though, that it's also the case when the inverter is working, but I'd assume it should be, in which case turning Q7 on should never be required or safe.

The PC2 output of the MCU (which is used, in the end, to turn Q7 on or off) is designated as "S3" in the table in the top right corner of the schematic, but I have no idea what that could possibly mean.

Another question would be regarding F1: what type of fuse can it be? It really looks like an inductor (photo attached). It's hard to tell the colors of the bands, but they look to me like brown, black, silver, gold, black.

Would it be appropriate to replace this fuse with, say, a polyswitch fuse rated for, I don't know, say 5 A? I guess that would be a sane choice considering that the charging linear regulator's (Q4, LM317T) output is limited to about 0.6 A.

What else should I check? Maybe someone has experience and/or more understanding how this UPS (or ones similar to this one) operates? Any input will be much appreciated.

Sorry for the schematic like this (no values), but it's all I could find. Right click->open image in new window (or tab) to view it full-size.

[shapirus]

All right, another finding: there is a short across D10, or across the LM317's input and output. I'm not sure if I overlooked it yesterday in the beginning (I think I did check the LM317 for shorts, but not 100% sure) or it happened after that.

Now I'm gonna order some polyfuses and identical replacement transistors for Q7 and wait a day or two for them to arrive, then continue working on it.

p.s. any idea whether replacing the inverter output MOSFETs (Q14, Q9, Q28, Q20, in the upper part of the schematic, part # RFP50N06) with some modern alternatives having a lower drain-source resistance will make sense in order to improve efficiency? Can there be any bad side effects, if I manage to find some with the same input capacitance so as not to affect the original switching times?

up: I've got a photo of the same fuse (which might actually be a fusible resistor?) from a different board. Black, gold, silver, black, brown.

[shapirus]

...so it was a blown LM317 in the charging circuit, input and output shorted. After replacing it (and Q7) the UPS is working like new again, currently with a glass tube 1.25A fuse in place of F1.

I'm not sure what was the cause and what was the consequence (Q7, LM317, something else probably), as it's not clear to me how exactly the shorted LM317 could have caused this problem, because D7 and D6 are operating correctly.

Anyway, hope this helps someone with a similar issue in the future.

Next I'm going to use a polyswitch fuse for F1 and -- cannot resist the itch -- change the inverter output stage mosfets from the original RFP50N06 to STP100N6F7. The latter has a somewhat less forgiving SOA, but I expect that it should be fine in this case. It also has almost the same Ciss and a much lower total gate charge and of course Rds(on), which I believe should be good for this application.