PSU faults: capacitor vs MOSFET failure

Most consumer-electronics dead-PSU symptoms come down to a surprisingly small set of fault patterns. A TV that clicks but won't start, a monitor that's gone completely dark, an amplifier that hums but won't come alive — under the lid these almost always trace back to one of five things going wrong on the switch-mode board. The hard part isn't fixing them; it's working out which one you've got before you start swapping parts.

Knowing which fault pattern you're looking at is what decides whether the board is a £40 fix or a write-off. A bulged capacitor is a fifteen-minute job. A blown primary MOSFET that's taken the controller IC with it on a board where the IC is unobtainable is the end of the road. Below is what we actually look at on the bench, in the order we look at it, on a typical mains-input switch-mode supply.

This is descriptive, not a how-to-poke-mains guide. The bulk cap on the primary side of any mains PSU sits at 170V DC or more and will happily stop your heart for several minutes after the kettle lead's been pulled. If you're reading this to decide whether to attempt it yourself or send it in, that's the right instinct.

The five visible symptoms and what they typically mean

PSU clicks (relay or PWM controller restart). Rhythmic clicking — once a second, maybe twice — is almost always the controller trying to start, detecting something wrong, and shutting back down. The fault is usually an output-rail short pulling the standby line down, or a feedback-loop component (opto-isolator, TL431, or one of the feedback resistors) telling the controller a lie. Sometimes a downstream cap on a secondary rail has gone short and the PSU is correctly refusing to drive into it.

No power at all, fuse intact. Dead silent, no LEDs, no click, fuse measures continuous. This points at the input side: the bridge rectifier has gone open, the inrush thermistor (NTC) has died open, or the bulk cap has dried out so badly the PSU can't sustain its own start-up. A failed standby rail will also do this on modern TVs.

Fuse blown. A blown fuse means a hard short pulled enough current to clear it. That's a primary MOSFET shorted drain-to-source, a bridge rectifier shorted, or — less commonly — a secondary-side rectifier diode shorted with the short propagating back. Don't just replace the fuse. The fuse blew because something else failed; find it first.

Hums, output present but won't start up. You can hear the transformer working, you can measure some output, but the set won't come up. This is the textbook symptom of high-ESR filter caps on the output rails. The cap looks fine, often measures the right capacitance, but its equivalent series resistance has crept up to the point where it can't hold the rail steady under load.

Output dim/wrong voltage, then dies. Display comes on at half brightness, or the amp powers up but cuts out after a minute. Bad regulation on the feedback side: the opto-isolator's CTR has degraded, a feedback resistor has drifted, or the TL431 reference is off. The PSU is running, it's just not running at the right voltage, and the protection circuit eventually trips.

Quick visual triage

Before any meter comes out, look at the board with a torch. Cap failure announces itself: bulged tops on the can (the X-scored vent has lifted), brown crusty electrolyte leaked out of the base, or the cap leaning over because the leads have corroded. Check every electrolytic, not just the obvious ones — the worst offender is often a small 10µF next to the controller IC, not the big 1000µF on the output.

Scorched PCB tells you where the magic smoke came from. A blackened area around a TO-220 transistor means the MOSFET let go. A burned trace radiating out from a small SMD package means a controller or driver failed. If you see a crater where a component used to be, that's a MOSFET that exploded — usually because a cap upstream of it failed open and let the switching node ring up to destructive voltage.

Smoke pattern matters: vented caps leave a fishy-smelling brown residue; failed MOSFETs leave black carbon and sometimes a hole in the plastic body. They look nothing alike.

Before you touch anything: the input-side bulk cap is the big 400V-rated electrolytic, usually 100-400µF. It can sit at 320V+ for an hour after unplugging. Bleed it through a 10kΩ resistor across the terminals, then verify with a meter. This is non-negotiable.

Multimeter checks that disambiguate cap vs MOSFET

With the board out, bulk cap discharged, and the PSU unpowered:

MOSFET check. DMM in diode mode, black probe on source, red on drain. A healthy MOSFET reads the body diode in one direction (around 0.5-0.7V) and open the other way. A short in both directions — particularly near zero — means it's gone. Also check gate-to-source: should read open. A shorted gate means the driver took it out, or static did.

Cap ESR check. A dedicated ESR meter (the cheap ones are fine) tells you in seconds. For a 1000µF output cap, anything over about 0.1Ω is suspect; over 0.5Ω is dead. Capacitance meters on a generic DMM lie about caps — they'll read in spec while the cap is incapable of passing AC. ESR is the test that matters.

Bridge rectifier. Diode mode across each of the four diodes inside the bridge package. Anything shorted means the bridge is done.

Why cap failure is more common but MOSFET failure looks more dramatic

Caps fail slowly. Electrolyte dries out over years, ESR creeps up, and one day the PSU stops starting. Heat is the killer — caps near hot components, or in poorly ventilated TVs, die first. The 105°C-rated parts last roughly twice as long as 85°C ones, which is why those are what we fit on the bench.

MOSFETs fail suddenly and noisily, but they almost always fail because a cap upstream gave up first. A snubber cap dries out, the switching node rings high, the MOSFET sees voltage beyond its V_DS rating, and it shorts. So you'll often find both bad on the same board: the cap that started it, and the MOSFET that finished it. Replace only the MOSFET and you'll be back in three months.

When it's a controller IC

Third tier of fault. We rule out caps and MOSFETs first because they account for maybe 85% of dead PSUs that come in. The remaining 15% includes failed PWM controllers (UC384x family, NCP1xxx, common TL431 feedback references), opto-isolators with degraded CTR, and the occasional shorted secondary diode.

Controller-IC failure is genuinely rare unless the set's been hit by a surge — lightning, a dodgy supply transient, or someone plugging a 240V appliance into a 110V outlet. After a lightning event you'll often find the controller IC, the standby MOSFET, and the input MOV all dead together.

What we charge

TV and monitor PSU board repair runs £60-£150 depending on how many parts need replacing and how awkward the board is to get to. Vintage amplifier PSU work is £80-£200 — older boards take longer because the parts often aren't drop-in replacements and through-hole rework on 1970s phenolic board needs more care.

Minimum bench fee is £40. Unusual or industrial PSUs are £80/hour, quoted before we start.

If your set is dead and you'd rather not poke a 400V cap yourself, send it in: /soldering-repair.html.

See also

• Capacitor plague: spot a swollen or leaking cap
• When recapping won’t save your vintage gear