Sharp LC-46LE824E - switches itself off when warm - Delta DPS-141CP1A - defect optocoupler in power supply

A friend brought me his Sharp TV with the symptom that it switches itself off when it gets warm. After cooling down it would restart and the problem would repeat itself.

I suspected a cold solder joint, but it was much more detective work. Thermal defects are the hardest.

A look inside:

Nice job, Sharp. Double speakers for each channel and one sub-woofer. Everything easily accessible.

We started testing and after pointing a hot air gun at the output part of the power supply, the device switched off. It recovered after a while and a more precise hot air attack located the area to the down right as most sensitive.

I removed the PS and checked for cold solder joints. None. Everything in good condition. The area around the three semiconductors under the heat sink was very sensitive to heating. The diode in the middle seemed to be the culprit, at least that's what I thought. So I removed the three, unscrewed them from the heat sink, soldered them back in without heat sink and then tested each in isolation. I replaced the diode with no success. The whole area was sensitive, with a tendency to the top towards the optocouplers. I spent a lot of time heating and shooting icing spray without any luck finding the precise spot.

In this Forum article I read about problems with optocouplers, so I narrowed my tests down. And lo and behold, the bottom coupler's sender diode freaked out when heated up. Part number was PC7703. It showed 1.9V break-through voltage at room temperature and quickly lost it when heated up slightly. A cross check with its neighbors settled the case.

Even a slight hot air wind made it fail, that's why I was fooled all the time when I pointed the hot air at other parts in its vicinity. 

I had plenty of spares from scrap boards and soldered in a new one. TV working fine again!

I never had a failed optocoupler. This is very rare. So what was wrong here? Look closely. The coupler sits in the main heat stream of the three semiconductors. And those get very hot! It literally gets grilled. I think this is the reason for its failure.

I glued a little piece of plastic onto the board to guide the heat from below around the couplers. This will relax the temperature situation.

Hallo Frau Johansson!

What I have learned from this case: A part with a thermal defect will reveal it measurably. If a suspected part behaves flawlessly, leave it and move on, even though it is hard to drop a promising fault hypothesis. I measured a double diode isolated from the circuit (forward voltage, reverse voltage) and it did not show any defect. Still I replaced it "just to be sure". It was a waste of time. The optocoupler however immediately revealed its problem very clearly.