UC3843 IC keeps shorting FET

[Vovk_Z]

Resistor burned, FET was fine

Mosfet may seems fine but it is not. Typically PWM controller don't have ability to destroy gate resistor (if it's own power supply works ok). Typically, high voltage from MOSFET drain can do it. But an opposite can happen too.

[Dannyx]

Well I'll be replacing the FET with a lower capacitance one anyway. Already found one but can't remember what it was right now. I'll also try a 20ohm resistor as its gate resistor.

[Dannyx]

Ok, the FET I chose is a 2SK2645 which features a capacitance of 800pF (ideally). I used a 20ohm resistor for R5 and also replaced C2. Still not working and we're sort-of back where we started: the filament glows briefly as the main cap charges up, it stays off for like 0.5 then the IC tries to start up. It pulses the transistor apparently, because the filament pulses in cycles of around 0.5s, so current IS flowing through the transistor, but something causes the IC to loop.

Starting to think that the IC may be bad, based on the idea that when the FET failed (several times so far), a large current flowed through the CS resistor, possibly causing a voltage drop in excess of 1v, which is what the CS pin is rated at, if I'm not mistaken. That pin doesn't appear to be clamped in any way, so I don't know how well it took that shock. Also that gate resistor increasing in value last time (this 20ohm one survived).

Could the high-ish gate threshold of 3.5v be too high for the IC to supply ? Doubt it, because the datasheet gives  Von HIGH on page 8 as 13.5v which should be plenty, since it's coming from the VCC pin, through the internal transistor and out the OUT pin

[Vovk_Z]

If primary circuit tries to start and can't do it you may look at secondary circuit. There may be something wrong at the feedback circuir or some opamp (if there is one) or etc.

[magic]

Repetitive restarts could be a failure of power supply to the PWM controller.
You could carefully connect a DMM to it, set it to volts and try to turn on the PSU again.

edit
But this still doesn't explain how it blew a few FETs already...

[Dannyx]

If primary circuit tries to start and can't do it you may look at secondary circuit. There may be something wrong at the feedback circuir or some opamp (if there is one) or etc.

There is this LM393 and TL431 in the corner here. Already replaced LM393 once because the outputs were stuck at 0v, regardless of the inputs (tested it off the board). Not knowing precisely what it does in this circuit, I can only assume it's functional now based on the readings I take with my DMM. It gets 12v at its VCC pin and one of its outputs is also 12v, which is normal because the inverting input corresponding to it is low, while the N/I input is "high". I'll do some more digging in this area, as I think it's time to start drawing up a schematic of this thing, since it's getting fairly complicated...

Repetitive restarts could be a failure of power supply to the PWM controller.
You could carefully connect a DMM to it, set it to volts and try to turn on the PSU again.

edit
But this still doesn't explain how it blew a few FETs already...

What I did was lift one leg of D13. This separated the relays from the 12v output so I can run it completely unloaded. Now, with no load at all, the bulb comes on fairly bright and steady for half a second, then goes off and I get an output, along with a faint high frequency switching whine. The problem is that the output is only 9.7v. Still, the components survived this test. So the problems we're faced with right now is the low output voltage and the failure to start under load. The first could be a bad feedback loop...or caused by the current limit of the bulb, though I remember checking multiple adapters with a 25w bulb in series like this and they all gave me the correct output (when unloaded of course), even one based on the UC3842 which I recently patched up...

[Dannyx]

Ok guys, I made some progress: I took a day and I drafted up the schematic of this board, at least partially. I excluded the high voltage side. Couldn't work out precisely what Q2 and Q3 are. I'm not even sure they're NPN to begin with - they're marked as SS1 or perhaps SSI and I've drawn a bit of a blank on trying to figure out what they are, so I did the next best thing and just speculated based on how they're connected. Aside from the values of the caps which I didn't bother to remove to check, everything else should be correctly marked and measured.

Also, after a bit of digging around my colleague's desk, I managed to find the original current sense resistor. Big surprise: he must've got the color code a mixed up when he got these big ceramic ones, because the original is a 1.5ohm resistor, whereas the one I found on the board is a 0.15ohm one. He probably messed up at the last ring. Anyway, the original was indeed open, so I just replaced it and sure enough: now the board comes on as expected, with just a single flash of the bulb. I even took it one step further and plugged it in directly and it's stable....well sort-of: it doesn't regulate properly. The output voltage is too low, at an unsatisfactory 9.7v. I imagine it would be even worse with all 3 relays trying to pull in at once - might not even start at all then. I replaced UC3843 again and the same thing happens.

I took a look at the board with a thermal camera and the IC does get a bit warm at around 37c, at least that's what the camera tells me, both on mains and powered directly with 12v. R10 and R1 get hot too (when on mains), which makes sense, because by my calculations, they're dissipating around 1w of power due to the internal zener of UC3843.

[magic]

Also, after a bit of digging around my colleague's desk, I managed to find the original current sense resistor. Big surprise: he must've got the color code a mixed up when he got these big ceramic ones, because the original is a 1.5ohm resistor, whereas the one I found on the board is a 0.15ohm one. He probably messed up at the last ring. Anyway, the original was indeed open, so I just replaced it and sure enough: now the board comes on as expected, with just a single flash of the bulb.

I should have thought about it; peak current limiter triggers at 1V, corresponding to 6A through 0.15Ω which is a lot for such a small PSU.

I took a quick glance at your schematic and it looks a little weird, I'm not sure if it's correct. You also omitted the rectifiers and where +12V is coming from. HV pullup is one thing, but usually it doesn't have enough power for continuous operation and the UC3843 is also supplied from the transformer.

edit
Are these HV resistors really 33k and not 330k?
Were they replaced and what was the value of the originals?

[Dannyx]

Yes it looks weird because I didn't bother to lay it out nicely. Ok, so it needs some more work, like I said - it's a work in progress. I'll draw the 12v bus too since it's fairly important indeed, so now it looks like THIS.

Those resistors are 33k (3 orange rings) and only R1 was replaced, so I doubt it was a different value in that forest of 33k

Ironically, VREF pin IS stable at 2.5v even when the output is 9.7v, so the IC thinks it's doing a bang up job and it's satisfied. I COULD start modifying resistor values in an attempt to "trim" it higher, but I shouldn't have to do that considering the values are fine and they're factory.

[magic]

Yes it looks weird because I didn't bother to lay it out nicely.

Yes, but it's more weird than that.

You missed R24, which pulls up Q3 collector. It wouldn't make sense without it.

R15,R14 look like they could suffice as a voltage divider to regulate the rail you call "+12V" to about 9.4V.
D12,Q3,Q2 look like thy could suffice to regulate output voltage at D12 zener voltage + 0.7V, except that there should be a resistor from D12 anode to ground and R25+R26 are counterproductive and may make this circuit prone to triggering too early
Why are both of the above circuits included 

U2B appears to be doing nothing interesting, it goes down when U2A goes down.
U2A appears to simply go down when HT voltage exceeds about 425V.
Why is this circuit so complicated

[Dannyx]

Thanks for that. Well, aside from any mistakes I made in drawing the thing (which I will only be able to correct on Monday, since it's on my PC at work) the board is what it is - you have the photos right there as proof - not making this up myself  Yeah, OVERLY complex for a stupid task IMO. Ironically, the 9.4v you described is in line with my readings, though it doesn't change the fact that the relays ARE rated for 12v. They MIGHT pull in at 9.4v indeed, who knows.....There's a possibility the transformer may have been damaged and is no longer behaving properly. My only hope is that I can alter the value of some components slightly to bump up the voltage enough to get it going, WHILE still preventing the FET from dying. The thermal camera shows everything as stone cold so far, with no load of course, except the IC itself and those resistors which get slightly hot, at least through the eye of the camera, as I can't feel it with my finger (don't worry - I know the board is not isolated from mains )

It's mostly for trying to work out how it works and improve my skills. I haven't analyzed your....analysis yet, but that's the kind of stuff I'm interested about: getting a simplified explanation in layman terms (changing X causes Y), as opposed to sifting through datasheets....which I CAN follow to a degree, but once you start throwing all kinds of formulae and math at me, I'm starting to lose track

[magic]

Actually, the effect of the transistor circuit is that at low voltages Q2 is on and adds R23 in parallel with R15. So the rail should sit at ~14V. (There is also R11 pulling FB down, but it's negligible unless it's damaged and less than 30kΩ).

So if you get 9.4V it's possible that something is wrong with transistors or the zener or the idiotic resistors in parallel with it (why??? ). I completely don't see the point of this circuit, unless it's connected to something else that you missed. Perhaps to the second rail? I think there are two outputs from the transformer, or at least two SMD diodes and two electrolytics near it...

[Dannyx]

Yeah, like I said, there may still be mistakes in the schematic - already found one, which I'll be sure to check and update as necessary. Still, I'm pretty certain that's all there is to it: no other output off that transformer and no other diodes - I checked. I see some traces going UNDER the transformer indeed, but they're some GND traces for the relays. I tried buzzing out the other pins of the transformer and they're not connected. There are some photos of the whole board amongst the first posts, so you can check those out too, just in case someone over the internet is able to spot something I missed with my own eyes and the board in my hand....and the multimeter......and the magnifying glass, who knows, might happen

It could be I misrepresented something, like a diode being flipped around, as you said...

[xavier60]

When the PSU first starts, Q2 would be on causing regulation at ~14V for some short time until the capacitors at the Base of  Q3  charge enough to turn Q3 on and turning off Q2 causing regulation to go down to ~9V.
Relays are usually happy to hold at much lower than pull-in voltage. Saves power and less heat.
 

[magic]

Interesting theory, but still weird.

Q3 collector load is 10kΩ to the supply rail, which means ~880μA in "hold mode" at 9.4V.
R25+R26=1.6MΩ and they provide 5.5μA base current under the same conditions.
The transistor needs to have β>160 for this to function. Maybe it does, but it seems marginal
And what's the deal with D12, then?

What's the capacitance of C21+C22? (I see that C23 is absent).

[xavier60]

Q3 could be a MOSFET,
D12 would be there to quickly discharge the capacitors when the PSU stops.

[magic]

Hmm, valid point.

I guess OP needs to check these transistors with DMM diode mode to find out what they are.
And check if they aren't shorted or something.

Test D12 too.

[Dannyx]

Thanks for digging into this thing. Starts to make a bit more sense now. I tested just about every component in-circuit and found no immediate defects, although I WILL pull one of the caps out to measure it and work out the value of the last one in that trio. Seems like there could be a soft-start type of deal there. I shall report back somewhere around Tuesday.

The theory that it might switch to a lower voltage on purpose sounds clever indeed.....excessive if you ask me, but clever. Wish I had a scope to confirm. I guess I can try the relays nonetheless and see what happens. If it suddenly becomes unstable, then there's still issues with it....

[magic]

This board's "ground" appears to be mains negative, i.e. it's live and not at earth potential at all.
It wouldn't be easy to scope and it's dangerous to touch when plugged into power.

With enough capacitance it could be slowed down to the point that even a DVM would see the initial 14V peak. If it works at all.

[Dannyx]

An isolation transformer or differential probe would take care of that, wouldn't it ? (provided the scope is not battery powered, OFC).

[Dannyx]

Ok guys, I updated the schematic to include the missing R24 and also tried my best to confirm everything else is drawn correctly, which it is. I also did what I should've probably done a long time ago and actually reinstall the diode that sits between the other diode and the relays and guess what: it works. Filament doesn't even glow. The 2 relays pull in with a good "thunk" and the voltage remains a stable(*) 9.7v (the relays only see around 8.8v after the 2 diodes though).

*One thing I don't quite like is that the least significant digits of my DMM's voltage range seem to be increasing very slowly the more the board runs. I only left it on (with the bulb) for like 30-40s max and the output had gone from like 9.710v to 9.721v. Don't know if this is because something is heating up, yet it's acceptable, or something is slowly breaking down and it's going to pop...I'll borrow the thermal camera from my colleague again and have a look. FET is stone cold. The last time I checked without the relays, the IC was getting "warm", or at least that's how the camera was showing it. It appeared to be burning hot on the screen, but that's because it referenced it to everything else around it which was cold, so in reality it was like 40c or thereabouts - definitely couldn't feel it with my finger.

It looks like it's solved (asterisk) for now, but I'll keep you posted in case something goes pop

[Dannyx]

I left the board plugged straight into mains for a solid minute and while it does work, the resistor R1 gets hotter than I'd like it to be - it even smells hot (don't have the thermal camera to check). So much so that after unplugging it, cap C10 to the right of R1 was hot to the touch, so it'll likely get destroyed if left like this....wonder if it was designed like this or if it can be improved.

Volage on main cap C13 is 317v and the whole contraption draws around 6.2w at the wall continuous with all 3 relays installed

[magic]

These resistors have 80kΩ total resistance so they draw ~4mA from the HT supply and dissipate over 1W in total.

It is possible that they cannot be increased by much because there is plenty of stuff on the board - LM393, TL431, Q3, UC3843 - which consumes power during startup. Before UC3843 reaches its turn-on threshold and begins to switch, this resistor must supply enough current to keep everything working and still charge 12V rail capacitors until UC3843 threshold is reached.

It was probably like that from the factory.

edit
If R1 is hotter than R10, check if they are still equal.

[Dannyx]

Yes, R1 is still within spec. I checked. I'm concerned about the other stuff around it becoming too hot, but it is what it is...the fans of the UPS might improve things once this thing is actually installed - haven't checked to see how they're positioned...

[Dannyx]

In the meantime, my colleague somehow managed to source another relay board. It's almost identical, with some subtle differences: the 4 33k resistors are now 2 5w 47k resistors and the two diodes in series with the relays are just wires now. We're obviously going to install this one, but this at least gave me the opportunity to compare my repairs to the factory-made one: I pretty much nailed all the values of the mystery resistors, except the gate one (R25 in my schematic) - they used a 100ohm one, whereas mine is 22ohm. I don't think this should cause any issues though.