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

a tenner of capacitors versus a new power supply

A desktop PSU that would not start reliably turned out to have bulging output capacitors, and a careful repair with a soldering iron and a discharge resistor brought it back rather than the bin.

A circuit board with a soldering iron and replacement capacitors

The symptom was a desktop that needed two or three presses of the power button before it would actually come up, and once it was running it was fine for days. Intermittent power problems are miserable to diagnose because the machine spends most of its time working, which makes you doubt yourself. But "won't start cleanly, fine once running" pointed at the power supply before anything else, and a PSU that is on its way out is a fire risk you do not want to ignore.

I swapped in a known-good spare and the fault vanished, confirming the suspect. The old me would have recycled the bad one and moved on. But these supplies are not magic, and the failure mode is almost always the same: cheap electrolytic capacitors on the low-voltage output rails that dry out, bulge, and lose capacitance over years of heat. So before binning it I opened it up to look.

Safety first, because this one bites

A power supply is not an Arduino. The big primary-side capacitors can hold a lethal charge long after the thing is unplugged, and they do not care how confident you are. Before touching anything I unplugged it, left it overnight, and then discharged the primary caps through a resistor with insulated leads while keeping one hand in my pocket. If that sentence does not already make sense to you, this is a repair to read about rather than attempt.

Once it was safe to handle, the culprits were obvious. Three of the output-side electrolytics had domed tops where they should have been flat, the classic bulge of a capacitor that has given up. A couple had the faint crust of dried electrolyte around the vent.

A close-up of a circuit board showing bulging electrolytic capacitors

The actual repair

The repair itself is unremarkable once the diagnosis is right. Note the polarity and value of each bad cap, desolder it, and fit a replacement of the same capacitance and voltage rating, ideally a low-ESR part from a brand I trust rather than whatever was cheapest the day this supply was built. The numbers matter:

  • Match the capacitance, in microfarads.
  • Match or exceed the voltage rating, never go lower.
  • Respect polarity. Electrolytics fitted backwards fail dramatically and loudly.

A 105°C-rated low-ESR capacitor in a spot that runs hot will outlast the bargain part that died, which is the small upgrade you get almost for free when you do this. I replaced the three obvious ones and, while I had the iron hot, a fourth that measured low on the meter even though it looked fine.

Buttoned back up, the supply now starts first time every time and has run steady since. The replacement caps cost about a tenner. A new supply of the same quality would have been the better part of sixty, plus the guilt of a working-bar-four-capacitors lump of copper and steel going in the bin.

I am not romantic about repair for its own sake. Sometimes the right answer genuinely is a new part, and a no-name PSU on a machine that matters is not where I would economise. But a known-good design with a known, common, cheap failure mode is exactly the case where an hour with a soldering iron beats landfill. This was one of those. The iron has more than paid for itself in supplies like this one.