SMPS feedback taking too long to switch back

[nunojpg]

I'm going a bit crazy repairing a friends fridge board 

The SMPS in this schematic (from a fridge) outputs the right 12 and 16V and then goes off for 300ms and the voltages drop about 50%. The cycle repeats consistently.

I've replaced PC817, KA431 and STR-G6352 and measure all components as much as possible.

Once the SMPS started to work fine again, just to go bad after a reset.

This kind of 300ms cycle points to some RC constant, but all components appear to be fine.

I've also tried to re-melt every solder point and check continuity on the traces, and it appears fine.

I don't have a differential probe to test effectively the mains side.

Some idea?

[nunojpg]

After finding other posts recomending to check caps with lower capacity I replaced CE2 which was down from 22uF to 19uF and it apparently fixed the problem.

I wonder why a 15% drop can cause a failure...

[MathWizard]

It looks like that cap is for the aux. power rail from the lower winding, that goes to pin4 which should be Vcc of the chip. If you had an ESR meter, I bet something would read worse than it should be. And then maybe a lot of the delay is from the chip, sensing Vcc under-voltage, and timing out for a bit? Or maybe it's upsetting the feedback loop more than the Vcc, but something like that.

[T3sl4co1l]

The behavior sounds like the regulator isn't receiving enough supply voltage to keep running, so it conks out during startup.  Then it waits for R3 to charge it back up, it kicks on, and the cycle repeats.  (This is normal "hiccup" mode protection.)  Too much ESR means CE2 doesn't retain enough charge during startup, and it fails to stay on.

Why does it run eventually?  When components heat up, ESR falls, and CE2 operates normally again.  It should "reset" on a thermal time constant i.e. it takes some seconds at least, but probably minutes, to return to the failing condition.

Tim

[nunojpg]

Your answers make all the sense. The fridge stopped working after a power failure. The fault (high ESR) could exist for months or years, but only after the PCB cooled down it was unable to start again.