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

the dead board that just had a tired capacitor

Bringing a dead network switch back to life by reading the symptoms, finding a bulging electrolytic capacitor, and replacing it, plus the unglamorous skills that make board-level repair possible at all.

A circuit board under a magnifier with a soldering iron alongside

A small managed switch in my rack died over the weekend. Not flickering, not flaky, just dark: power light off, no fans, nothing on the console port. The kind of dead where the polite assumption is "it's done, buy another one". And for a fiver more than the postage, I might have. But it had been getting unreliable for weeks before it went, dropping links and rebooting itself at random, and that pattern, working-but-degrading then suddenly-nothing, has a usual suspect. So before I binned it I took the lid off.

The usual suspect is the power supply, and specifically the electrolytic capacitors in it. They are the part of a board most likely to age out. They dry up, their capacitance drops, their internal resistance climbs, and a supply that used to be clean starts to sag and ripple. A board fed dirty power doesn't fail honestly; it gets flaky first, glitching under load when the supply can't hold up, before finally not starting at all. That description matched my switch exactly, so I knew where to look before I'd even found the magnifier.

reading the board

You can find a bulging capacitor with your eyes most of the time, and this was one of those times. Electrolytics are meant to have a flat top scored with a cross. When they fail they often dome upward, sometimes weeping a little crust of dried electrolyte. There it was: one 470µF can on the low-voltage rail, top rounded like a tiny pie, sitting right next to three identical flat ones. You don't need a meter to be suspicious of the one that's the wrong shape.

A close-up of a circuit board with one swollen capacitor among its neighbours

I confirmed it anyway, because confirming is cheap and being wrong is annoying. The board out of circuit, an ESR meter across the suspect cap read far higher than its flat siblings. Capacitance was down too. That's the signature of an electrolytic that's dried out: not a dead short, not an open, just tired, which is the most insidious kind of fault because the board half-works and sends you chasing software ghosts. I had genuinely spent an evening weeks earlier wondering if a firmware update had made the switch unstable. It had not. A 25p component had.

getting the old one off

This is the part nobody tells you about board-level repair: the soldering is the easy bit, getting the dead part off cleanly is the hard bit. Through-hole electrolytics on a multi-layer board are soldered into plated holes that wick heat away into internal ground planes, so a small iron just can't get the joint hot enough to flow. You stand there melting the surface of the solder while the core stays solid and the leg won't budge, and if you push, you lift a pad and turn a 25p repair into scrap.

The trick is heat, patience, and flux:

- crank the iron to ~370C for the leaded joints
- add fresh leaded solder to the old joint first
  (it carries heat in and gets things flowing)
- hit each leg with the iron and pull a little with pliers,
  alternating legs, never forcing a cold joint
- clear the holes with desoldering braid and flux,
  not by jamming the new leg through molten metal

That fresh-solder trick is counterintuitive, you add solder to remove a part, but it's the single thing that turns this from a pad-lifting nightmare into a thirty-second job. The old electrolytic in modern boards is often lead-free, which melts hotter and flows worse; reflowing it with leaded solder lowers the melting point of the whole joint and it lets go like it wants to.

A word on the tooling, because this is where people give up and shouldn't. You do not need a hot-air rework station or anything that costs more than the kit it'll fix. You need a temperature-controlled iron with a decent-sized chisel tip, because a fat tip moves heat better than a fine one and finesse here is the enemy. You need flux, a syringe or a pen of it, applied liberally and without guilt; flux is what lets solder flow where you want and refuse to bridge where you don't. And you need solder braid for clearing the holes once the part is out. That's the whole kit, maybe thirty quid all in, and it pays for itself on the first repair you'd otherwise have thrown away.

New cap in, same capacitance, same voltage rating, low-ESR type because that rail clearly cares, and crucially the right way round. Electrolytics are polarised and a backwards one will, given enough provocation, vent with a bang and an unforgettable smell. I checked the stripe twice. Powered it up on the bench behind nothing important, and the power light came on, the fans spun, and the console printed its boot banner like nothing had ever been wrong.

While I had the lid off and the iron hot, I did one more thing that costs nothing and saves a future repeat: I replaced the other three capacitors on the same rail. They tested fine, but they're the same age, the same make, and they've sat next to the same heat for the same years. A capacitor that's failed is just the first of a cohort to go, and chasing them one at a time means opening the case again in six months. Recapping the whole rail at once is the difference between a repair and a recurring chore. The parts were pennies and I was already there.

It's been running clean for three days now, including the random reboots that haven't happened. The honest economics are daft: my time was worth more than the switch. But that was never the point. The point is that "dead electronics" is very often "one tired component on an otherwise perfect board", and the difference between scrap and a working switch was a magnifier, a quid of capacitor, and the willingness to take the lid off before giving up. I find that genuinely cheering. Most things aren't as broken as they look.