Worn ATX PSU still works but with reduced power peak

[davidebaldini]

The model of this 500W ATX PSU is "Cooler Master rs-500-psap-j3".

It appears that after an estimated 30,000 hours of mostly uninterrupted usage, its output power peak dropped below the minimum required to keep my computer on.

Diagnosis: the PSU ran stable w/o issues for a few years permanently attached to the same PC hardware (PC #1). All hardware remained the same up to several months before the PSU failed. I found my PC abnormally powered off without any UPS event recorded; this occurred again after ~12 hours of uptime since the previous abnormal shutdown; this second time I inspected the PC and it wouldn't complete the boot due to sudden power off _after_ POST; a few attempts later the computer stopped POST-ing at all. It remained like that since. I replaced the PSU with a new one — a same power 500W PSU, branded — and PC runs fine. Testing again with the old PSU doesn't produce POST. Testing the same old PSU on a much less demanding PC (PC #2) works fine.
It appears that after sustained usage, my PSU suddenly and rapidly lost its power capacity.

Hardware description:

• PC #1: The PC on which this old PSU was mounted on is a dual Xeon workstation with presumably high power drain.
• PC #2: The smaller PC I used to test if the PSU worked at all is a very small Pentium 3 or 4 which works fine on its own with a 300W unbranded Chinese PSU.

PSU Inspection:
The PSU PCB appears perfect n the components side: no bulging or leaked caps, no discoloration due to heat, no cooked glue, all normal at sight. I measured the two large (1W ?) resistors on the lower right corned of pic #1: they are within ~7 percent tolerance, but they have a gold tolerance band (5%). I didn't unsolder them to take the measure, so the rest of the circuit might have interfered.

Then I flip the PCB: the tracks side of the PCB doesn't show any overheat discoloration, but there is an abundant solid deposit more or less around all capacitors' pins, so I suspect capacitors could be a culprit. The deposit is abundant also around the two primary 680uF caps and around the yellow parallelepiped on the bottom of pic #1 (a capacitor?). So I bought ~13 € of capacitors and replaced all caps circled in red in pic #1, excluding the yellow thing that I couldn't undoubtedly identify (pic #7).

There is also minor deposit on the tracks side of the vertical PCB mounted on the main PCB, too. I didn't work on that.

Recapping:
I replaced the caps circled in red on primary and secondary sides:



The non-circled cap in the middle of the picture appears to be connected to 5V or 3.3V rail; I couldn't cheaply buy a matched replacement for this one.

Results:
After recapping, the PSU continues to behave as described above under "PSU inspection": it still works fine under minor loads, but fails to boot the same PC that used to run for years.

Notes: all pictures taken after PSU was serviced.

[MHTSOS]

Re: Worn ATX PSU still works but with reduced power peak

This PSU looks like it has BJTs as primary switchers. Very close to the 2 BJTs you will find 2 caps. They must be 50V 1uF to 10uF. It depends on the design. If you have an ESR meter measure these 2 caps. If not, replace them and test it again.

[davidebaldini]

Re: Worn ATX PSU still works but with reduced power peak

I could find two — and only two — electrolytic caps close to the switches, they are: 2.2uF 50V and 22uF 50V.

I don't have an ESR meter so I'm placing an order on ebay for the caps. Which other caps are likely to need replacement beside these two? I already replaced most of the larger ones.

Also, why these two capacitors are the first ones to check?

[CapLeaker]

Re: Worn ATX PSU still works but with reduced power peak

with all that run time on the PSU, i'd be replacing all of these caps. Had a similar issue on a 12V PSU with a continuous run time of 7 years (that is over 60.000 hrs). Ended up replacing all caps in order for it to handle the current again as it did, when it was new, including the main filter capacitor.

[RukyCon]

Re: Worn ATX PSU still works but with reduced power peak

This appears that it might be a capacitor issue.
I'm guessing the secondary caps might be failing but it could also be the tiny caps failing.
And i recommend not using ebay for electrolytic capacitors.

[MHTSOS]

Re: Worn ATX PSU still works but with reduced power peak

These capacitors are in the circuit that drives the bases of the switchers. They dry from prolonged use because they are next to the hot heatsink and when the switchers have to be driven hard to provide more power then there is not enough current to drive the base. I have seen this a few times in some very old and used PSUs and they would power on fine with light load but the output voltage would collapse when you tried to pull more power. Replace them and check. It would be wise to replace all electrolytic capacitors since it had so many hours of use.

[momaka]

Re: Worn ATX PSU still works but with reduced power peak

These capacitors are in the circuit that drives the bases of the switchers. They dry from prolonged use because they are next to the hot heatsink and when the switchers have to be driven hard to provide more power then there is not enough current to drive the base. I have seen this a few times in some very old and used PSUs and they would power on fine with light load but the output voltage would collapse when you tried to pull more power. Replace them and check. It would be wise to replace all electrolytic capacitors since it had so many hours of use.

+1

And yes, replace ALL electrolytic caps, including the smallest ones. The only exception is the big 200V caps on the primary side, as those are not under a lot of stress and rarely fail (save for CapXon).

By the way, this PSU appears to be made by FSP, and an older design at that. Generally their PSUs are pretty good, especially the older ones. So definitely worth recapping.

As others suggested, you might want to avoid ebay or Ali Express for your cap purchace, as there are too many counterfeit / goofy no-name caps that could quickly fail again. Contact member behemot, as he is based in Czech Republic and can probably ship to you. Alternatively, see if Mouser, RS Components, or Farnell have the caps you need. TopCat (the owner of this site) also sells caps, but he is based in USA, so shipping may or may not be too expensive to you.

[davidebaldini]

Re: Worn ATX PSU still works but with reduced power peak

Thank you all, I repaired the PSU and it works fine even on high load now. Here's what I did.

I ordered Panasonic and Nichicon capacitors from this French seller on ebay; the caps appear legit at sight and the capacitances I measured match their nominal values. Before ebay I checked Mouser, Digikey and rs-online but their postage fee was high to Italy. The seller I found on ebay appears in good standing regarding caps as per his feedbacks.

As suggested, I desoldered and measured these two capacitors, which sit next to a heatsink cooling two D209L NPN transistors:



Their nominal values are 2.2uF and 22uF, but they measured 0.88uF and 20+uF (my multimeter cannot measure above 20uF). I replaced both, and in the process I also replaced about five other small capacitors for which I had a matching replacement; in my previous recap I neglected to replace any of the smaller ones.

Then I oiled the fan, reassembled the chassis and positively tested the PSU on the larger computer.

The major pain was to access and extract the smaller caps circled in picture, hidden beneath the primary heatsink and behind the two large caps. Another painful extraction was a capacitor next to the secondary heatsink, which turned out the be a cylindrical coil mounted vertically concealed with a black heat-shrink sleeve.

[momaka]

Re: Worn ATX PSU still works but with reduced power peak

Nice, good to hear the PSU is properly working again!

Their nominal values are 2.2uF and 22uF, but they measured 0.88uF and 20+uF (my multimeter cannot measure above 20uF).

Just beware that capacitance is not telling you the whole truth weather a capacitor is good or not. It's actually happen fairly often that when a capacitor fails, its capacitance shows normal value, but its ESR is way too high. When that happens, the capacitance becomes irrelevant, because the large ESR acts like a large series resistance and makes the capacitor ineffective. Thus, you need an ESR meter to check capacitors. Those cheap component/transistor testers on eBay for less than $10 are actually pretty good for that, as they have ESR measuring feature (and capacitance up to 20-40 mF, too! IIRC).