The little class-D amplifier board that drives the speakers in my workshop died sometime before Christmas. No drama, no smoke, no smell. One day it was filling the garage with Radio 6, the next there was a faint thump on power-up and then nothing. I had a fortnight off and a soldering iron, so instead of binning a board that cost me about eight quid, I decided to find out what killed it. The repair was beside the point. I wanted the diagnosis.
First, look at it
Before any meter comes out, you look. Properly, under a lamp, with a loupe if you have one. Dead electronics very often tell you what happened if you give them the chance.
This one did. One of the electrolytic capacitors near the input had the faintest dome to its top, where it should be flat. Not a dramatic bulge, not the split-and-weep you see in photos of failed motherboards, just a tired little curve. A bulging electrolytic is a cap that has been cooked, slowly, by ripple current or heat or simple old age, and the electrolyte inside has started to gas off. When the top domes, the capacitance has usually collapsed and the equivalent series resistance has climbed through the roof.
That is a strong suspect but not a verdict. A bulging cap can be a symptom rather than the cause. So I kept looking, and found nothing else obviously wrong: no scorched resistors, no lifted traces, no cold joints gone grey and cracked. Just the one sad capacitor.
Then, measure it
With the board unplugged and given a few minutes to discharge through its own bleeder, I went round the obvious points with the multimeter.
- Continuity from the barrel jack through to the main filter cap: good.
- The main 5V regulator output: dead. Zero.
- Diode-check across the rectifier diodes: all sensible, no shorts.
A dead regulator output with a healthy input is interesting. Either the regulator itself has failed, or something downstream is pulling it to the floor, or its supporting capacitors have gone and it is oscillating itself into shutdown. Given the bulging cap sat right on that rail, the picture started to make sense. A filter cap that has lost most of its capacitance can let a switching regulator misbehave badly enough to trip its own thermal or over-current protection.
I desoldered the suspect cap. This is the part people dread and it really needn't be. Add fresh solder to the old joints first, even though that sounds backwards. The new solder melts the old, mixes its flux in, and the whole joint flows at a lower temperature than the tired oxidised stuff will on its own. Heat one leg, lift gently, heat the other, lift, and it walks out. Braid the pads clean afterwards.
Out of circuit, the meter told the rest of the story. The cap measured nothing useful on the capacitance range, and on a quick ESR check it was effectively a resistor. Dead as a brick. Its replacement, a 470µF 25V from the parts drawer, measured exactly as it should.
Solder it back, properly
Putting a through-hole part back is the easy bit, and the bit most likely to be done carelessly. A good joint is shiny, slightly concave, and wets right up the lead and across the pad like a tiny volcano. A bad joint is dull, blobby, or balled up because the pad never got hot enough. The classic mistake is heating the solder instead of the joint. You heat the pad and the lead together with the iron, then feed the solder into that, not onto the tip. The joint melts the solder, the iron melts the joint.
good: smooth, concave, shiny, lead clearly wetted
cold: dull, grainy, blobby, lead poking out of a ball
Get the iron in and out fast. A second or two on a through-hole pad is plenty. Lingering cooks the board and lifts traces, which turns a five-minute fix into a far worse afternoon. I trimmed the leg, cleaned the flux residue off with a bit of isopropyl and a cotton bud, and inspected the joint under the loupe. Shiny, concave, wetted. Good.
Power up, gingerly
I do not just plug a repaired board straight into the wall. If something is still shorted you want to find out quietly. I brought it up on a bench supply with the current limit set low, watching the meter. It drew a sane idle current, the 5V rail came up and stayed up, and nothing got warm that shouldn't. Then, and only then, the speakers, and the input.
Radio 6 again, in the garage, on an eight-pound board I'd written off a week earlier. The total cost was one capacitor I already owned and a couple of evenings I'd happily have spent worse. The lesson I keep relearning is that most "dead" consumer electronics are not dead, they have one tired component, and the time it takes to find it is mostly the time it takes to slow down and actually look. The iron is the trivial part. The looking is the craft.