Testing diodes and transistors to predict failure ?

[MathWizard]

I was watching a video on PSU's with cheap or over burdened mosfets blowing up, and I'm just thinking about my own computer PSU. I will at some point take it apart and check all the electrolytic's for ESR/DF (it must be over 5yo, if not closer to 6 or 7yo).

But what about simple tests to see if fet's or diodes are near giving up? Can you just put them on a curve tracer or a like a source meter, or insulation tester for gates, and see the difference between a new part, that matches the datasheet, and some old part that is wearing out ? (not that I have those tester/meters or $thousands for them).

What about checking mains voltage x-caps ?

I guess when the day comes, I should go by their age, and replace them if I want to keep using it and feel safe.

[Vovk_Z]

If you see burned PCB around them then they worked hard and may be replaced. There is not any other specific parameter which will show a health of a semiconductor.
Usually electrolytes are the weakest part of ATX power supply, so if you replace em, you'll be fine. I suggest to use polymer ones (at the output) instead of usual low-esr type.
From my experience 5-6 years of work may be enough, so it is definitely a time to service it. If PS worked hard it may be fine to give it no longer then 3-4 year to service.

[David Hess]

Even a curve tracer will only occasionally reveal impending failure like poor hfe, low Vceo, high Vsat, or high leakage, and you would need to know what to look for by having good parts for comparison.

[Vovk_Z]

What about checking mains voltage x-caps ?

Though they may degrade gradually over time (because of transients) they usually live a long life and they are not of concern usually. I don't remember dead or very tired ones. The only typical thing about X2 caps - they may be only partially installed by the manufacturer of a cheap PS.

[Benta]

Though they may degrade gradually over time (because of transients) they usually live a long life and they are not of concern usually. I don't remember dead or very tired ones. The only typical thing about X2 caps - they may be only partially installed by the manufacturer of a cheap PS.

Exception: the notorious Rifa caps.

[Vovk_Z]

Exception: the notorious Rifa caps.

Yes, of cause. I just add they can be found in quite old equipment (like old analog Tectronix scope) but not in actual ATX PS.

[wraper]

You cannot really test if semiconductors will soon give up until they actually do so. At least without extensive parameter measurement out of circuit. I would look with thermal camera into those crappy PSUs to see if things are overheating.

[wraper]

Though they may degrade gradually over time (because of transients) they usually live a long life and they are not of concern usually. I don't remember dead or very tired ones. The only typical thing about X2 caps - they may be only partially installed by the manufacturer of a cheap PS.

Exception: the notorious Rifa caps.

They actually fail quite often. What normally happens is that metal coating evaporates when there are breaches in insulator (self healing) and capacitor loses capacitance over time as more insulation failures happen. You lose high frequency filtering but device still works. Rifa just explodes in magic smoke.

[Gyro]

Actually Rifa (self-healing metallized paper) capacitors are one of the easier ones to predict. Package cracking is easily visible on inspection, indicating moisture ingress and imminent failure.

Edit: Actually not that imminent with mild visible cracking - they can, and do, go on for years like that, but it's an indicator for immediate preventive replacement.

[MathWizard]

 A few years back I bought a few "broken" computer PSU's, and I mapped out them, in full, as best I could in LTspice. And I learned a pile of EE from that. I fixed a few, and still have few I need to teardown, and a few others I need to put back together.

Money-wise there's no point to sell them really, output re-capping, is the only way I'd trust any of them if I was to sell them (most read ok but they are 10yop caps by now for sure) , and I don't have a 1500W load tester, nor do I want to use my gaming PC as the test load.....so

But just watching the video of Gamer's Nexus blowing up more Gigabyte PSU's, I can make out the basic layout of the PCB from the other videos, and I wish they'd send me a fried unit to make a schematic's

It blows up just after 17min utube length
In my experience with the PFC/BOOST IC's I have seen, I have not seen a max power limit as such, but separate OV and OC limits. But maybe it's built into their op-amps as-well, IDK, yet. Maybe in some of the DC-DC converter chips, 12 down to 5V, some of them they internally calculate power, IDK. So again I'd love to teardown 1 of these.

[strawberry]

X2 capacitors dry out overtime(capacitance drop)
transistors parameters degrade at high temperatures(150C doesn't mean can run at full power)
when overloaded silicon can overheat(150C) due to junction to heatsink thermal resistance.

meanwhile in china, we got 100W psu circuit from IBM, lets make it run at 150W
after 20y, lets ad big fan and cool printed sleeves on capacitors and run it at 500W ..to be shore it will deliver 500W lets ad 10A fuse

[Gyro]

X2 capacitors dry out overtime(capacitance drop)

They don't dry out - they are dry (normally polypropylene film) in the first place. The mechanism for capacitance loss is self-healing, as indicated earlier in the thread.