Ramblings of an aging IT geek
← Ramblings of an aging IT geek
hardware

Fixing A PSU Instead Of Binning It

A dead ATX power supply turned out to be two bulged capacitors and a fifteen-minute repair, with the usual loud caveat about the lethal bits inside.

A soldering iron and electronics on a bench

A home server's power supply died last week. Not dramatically: no bang, no smell, just a machine that wouldn't power on and a faint, tired clicking from the PSU when I shorted the green wire to ground to test it. The clicking is the giveaway. It's the supply trying to start, failing a self-check, and retreating, over and over. Something inside couldn't hold a rail steady.

The lazy move is to bin it and buy a new one, and honestly that's the correct move for most people, for reasons I'll get to. But this was a decent 550W unit, out of warranty, and I had a spare evening and a soldering iron. So I opened it up.

the obligatory warning, which I mean

Before anything else: a power supply holds lethal charge in its primary-side capacitors long after it's unplugged. We're talking hundreds of volts sitting in a big can waiting for a careless finger. If you don't know how to discharge those safely, with a resistor across the terminals and a multimeter to confirm, do not open a PSU. This is the one piece of computer hardware that can genuinely kill you, and it is not being dramatic to say so. I unplugged it, left it overnight, and still checked every large capacitor read near zero volts before I touched a thing.

A close-up of a circuit board

what was actually wrong

With that out of the way, the fault was almost boringly common. Two electrolytic capacitors on the secondary side, the low-voltage output stage, had bulged tops. Capacitors are meant to be flat on top. When the top domes upward or the vent has split and crusted, the electrolyte has cooked off and the cap is failing. It's the single most common reason consumer power supplies die, a legacy of years of cheap capacitors that age badly under heat.

Both bad caps were 1000µF 16V parts. I had a couple of 105°C-rated replacements in a drawer, slightly higher temperature tolerance than the originals, which is the direction you want to err. The repair itself was the easy part: desolder the old ones, noting polarity because electrolytics very much care which way round they go, and solder the new ones in. Fifteen minutes, most of it spent fighting the wick to clear the old solder from the through-holes.

back from the dead, with a caveat

Reassembled, plugged into a mains lead I could yank in a hurry, green wire shorted to ground. The fan spun. All rails measured within tolerance on the multimeter. Back into the server, and it's been up for four days without complaint.

Was it worth it? For me, yes: an hour of my time, two capacitors that cost pennies, and a working PSU instead of landfill. But I'll be honest about the caveat. A PSU that's failed once is a PSU I trust slightly less, so this one runs a home box with good backups and nothing irreplaceable, not anything that matters. And the time-versus-cost maths only works because I enjoy the bench work and already own the kit.

If you don't, buy a new one. There's no shame in it, and the lethal capacitors mean the bar for "have a go" is genuinely higher here than for almost any other repair. But if you do enjoy it, a bulged capacitor is one of the most satisfying faults in all of electronics: visible, cheap, and fixable in the time it takes the kettle to boil twice.