Here's a scene every field tech knows too well. Machine goes dark. You pull the power supply, slap in a spare, machine boots up. Dead unit goes onto a shelf. Nobody asks what actually failed.
And honestly? Most of the time, you could have fixed it. For under fifteen bucks.
Our CTO has been cracking open ATM power supplies for 23 years — across NCR, Diebold, and Wincor platforms. After a while, you stop seeing mystery faults. You see the same five things failing over and over again. They account for roughly 80% of what lands on our bench. Here they are, in no particular order of pain.
---
- Switching MOSFET Shorts — Fuse Blows With It
The MOSFET on the primary side switches at high frequency, every second, for years. Eventually the junction breaks down short-circuit. When it does, the current spike takes the main fuse instantly. Machine goes completely dark — no LED, no fan, nothing.
What to check: Fuse first. If it's open, pull the MOSFET and test drain-to-source in diode mode. Near 0Ω confirms the short.
What to replace: MOSFET and fuse. Together. Never just the fuse. I've seen techs do this — they swap the fuse, power up, and it pops before they can even look at the meter. The shorted MOSFET is still sitting there. Also check the gate resistor and current-sense resistor while you're in there. A blown MOSFET tends to take its neighbors with it. If the gate resistor reads open, your new MOSFET won't switch and you'll be back on the bench before the week is out.
One thing that trips people up: fuse is blown, but MOSFET tests fine. If that happens, the problem moved downstream. Check the Schottky rectifier on the secondary side before you waste time staring at a MOSFET that isn't the culprit.
---
- Startup Resistor Goes Open — UC3842 Never Wakes Up
The UC3842 PWM controller is the brain of the supply. It needs a kick of voltage on pin 7 to start switching. That voltage comes through a high-value resistor that feeds rectified DC. After years of running 24/7, this resistor drifts open. The controller never gets its wake-up call. Dead silent. No switching, no output, no nothing.
What to check: Measure pin 7 to ground with power applied. If it reads 0V, the startup resistor is open. Normal startup voltage is 12–16V. Once the auxiliary winding takes over in run mode, it settles around 10–14V.
What to replace: The startup resistor. And then — this step gets skipped a lot — trace the VCC auxiliary winding path. If the controller starts switching but shuts down after half a second, the auxiliary supply is failing. Check the rectifier diode and filter cap between the aux winding and pin 7. A dried-out filter cap on that rail causes exactly this: one chirp, then silence. Don't confuse this with a shorted output pulling the controller into hiccup mode. Check all output rails for shorts first, then go after the aux supply.
These resistors cost literal pennies. They kill more power supplies than any $50 component.
---
- Aged Bus Capacitor — The Ghost in the Machine
This is the one that made me want to quit field service once.
The large electrolytic on the primary side smooths rectified mains to roughly 300V DC (on 220–240V mains). As it ages — and they all age — ESR creeps up, capacitance drops, ripple increases. The machine starts rebooting at random. Monday morning: fine. Tuesday afternoon: three reboots. Wednesday: runs all day without a hiccup.
No error code. No log entry that makes sense. Just intermittent chaos that correlates loosely with temperature — cold mornings after a hot day, or peak afternoon heat inside the enclosure. You show up with tools and the machine is running perfectly. You leave. Thirty minutes later, it reboots again.
What to check: Visual first. Bulging top, domed vent, electrolyte crust — those are dead giveaways and you don't need any meter. If the cap looks clean, measure ESR. For a 100–220µF 400V capacitor, anything above 2–3Ω is suspect. Compare to the datasheet if you have one — some low-ESR series should be well under 1Ω when new.
What to replace: Same capacitance, same voltage rating, and absolutely use 105°C rated parts. 85°C capacitors will cook inside an ATM enclosure. I've measured 70°C inside cabinets in desert climates. An 85°C cap running at 70°C has a life measured in months, not years. While you're swapping the main bus cap, look at the smaller electrolytics on the primary side too. The startup cap on the UC3842 VCC rail often ages at the same rate as the bus cap — they've been baking in the same oven.
---
- TL431 Feedback Drift — Output Goes Rogue
The TL431 programmable reference works with the optocoupler to tell the PWM controller what the output voltage is doing. The reference pin should sit at exactly 2.5V. When the divider resistors around it drift with age, that reference point walks. Output regulation follows it right off a cliff.
Here's what makes this one nasty: under light load — machine idle — output voltages look perfect. When the dispenser motor fires up and current draw jumps, voltage sags or overshoots. Card reader errors. Dispenser timeouts. System resets that only happen during peak transaction hours. And of course, the bench test at idle shows nothing wrong.
What to check: Measure TL431 R pin to output ground under light load first. 2.5V ±0.1V is normal. If it's off, the divider has drifted or the TL431 itself is going bad. Then scope the output under a step load if you have an oscilloscope. Ringing or slow recovery means the compensation network capacitors are probably aged too.
What to replace: TL431 and both divider resistors. All three, as a set, every time. If the TL431 died, it stressed the resistors. If the resistors drifted, they'll eventually kill the replacement TL431. Replacing one without the other buys you a callback. Also check the optocoupler CTR while you're in the feedback section — if the optocoupler is degraded, regulation will drift under temperature even with a brand new TL431 in place.
---
- Schottky Diode Degradation — Fine at Idle, Falls Apart Under Load
Schottky rectifiers on the secondary side handle high current at low voltage. They cycle hot during operation, cool down when idle, heat up again — every day, for years. That thermal cycling gradually increases forward voltage. At no load, output reads dead-on spec. Under load, voltage sags.
This causes some of the most frustrating field issues I've encountered. Card readers that work intermittently. Dispensers that timeout when transaction volume is high. System resets during peak hours only. Nobody looks at the power supply because it tested fine on the bench at idle.
What to check: Load the supply to its rated current. If output drops more than 5% from no-load, the Schottky is degraded. An in-circuit diode test might still read "OK" because Vf degradation only shows up under actual current. Thermal camera makes this one obvious — a degraded Schottky runs noticeably hotter than its neighbors on the same heatsink.
What to replace: The Schottky rectifier. If the original part was running hot (and discoloration on the PCB around it is a dead giveaway), step up to a higher current rating. Don't stop there — check the output filter capacitors downstream of the Schottky. A degraded Schottky stresses them with higher ripple current, and dried-out output caps make the voltage sag even worse. Replace both if either one looks questionable.
---
The Three-Step Bench Method That Saves Time
When a dead PSU hits our bench, we follow three steps. Nothing fancy. Just a sequence that catches most failures without wasting an hour on a $5 fix.
Step 1 — Open it and look. Most failures leave evidence. Bulging capacitors. Scorched PCB around the switching transistor. Cracked solder joints on transformer pins or connectors (thermal expansion loosens them over time). Discolored board around anything that runs hot. Burnt or cracked resistors. Five minutes with good light beats thirty minutes with a multimeter. Take photos before touching anything — they help when you need a sanity check or a second opinion from someone who isn't standing in front of the board.
Step 2 — Measure, in order. Power off: check fuse, MOSFET drain-to-source, startup resistor. Power on (through an isolation transformer and current limiter — do not skip this): check UC3842 pin 7 voltage, 300V DC bus, and TL431 R pin. These six measurements will catch all five failure modes above. If you've got an oscilloscope, scope the MOSFET drain waveform. A distorted waveform points to snubber problems or transformer issues that a multimeter alone won't reveal.
Step 3 — Fix more than the obvious. Replaced the MOSFET? Check the gate resistor and current-sense resistor. Replaced the fuse? Make sure the Schottky on the secondary didn't fail first. Swapped the bus cap? The startup cap is probably just as old. A failure doesn't happen in isolation — it's almost always a chain. Find the chain. On first power-up after repair, run it through the current limiter. If the bulb glows bright and stays bright, you've still got a short. Kill power and go back to Step 2.
---
Repair or Replace? A Practical Call
Not every power supply deserves a second life.
Repair when it's a single, clear failure. Blown fuse + shorted MOSFET. Open startup resistor. Aged bus capacitor with no other damage. TL431 feedback drift. These are $5 to $15 in parts and less than an hour on the bench. Even the TL431 repair — parts cost is under two dollars. The economics are obvious.
Replace when the damage is widespread. Multiple failures across primary and secondary. Burnt PCB traces where the FR4 itself might be compromised. Shorted transformer windings (you won't find a replacement for custom-wound ATM transformers). When the bench time exceeds the cost of a replacement unit, stop. Your time has value beyond the parts you're saving.
And the no-exceptions rule: if the PCB is charred black under the MOSFET, don't even try. Carbonized FR4 is conductive. You'll chase intermittent shorts until you lose your mind.
---
Preventive Habits That Keep PSUs Alive
Most ATM power supply failures aren't sudden. They cook slowly — capacitors drying out, resistors drifting, solder joints cracking from thousands of thermal cycles.
- Every PM visit, open the PSU bay and look inside. Bulging caps announce themselves months before they fail.
- Dust is a power supply's slow death. It insulates, traps heat, cooks capacitors. Blow out the PSU compartment every service visit. Takes two minutes.
- Machines in hot locations — outdoor ATMs, desert climates, direct sun — expect shorter PSU life. Heat is the number one accelerator. If one supply failed, another will. Consider ventilation or shade.
- Keep one known-good spare per machine model. When a unit acts suspicious, swap it. Diagnose the pulled unit on your bench, with proper tools, not at the ATM at 2 AM with a phone flashlight.
---
We stock tested power supplies and repair components for NCR, Diebold Nixdorf, Wincor, and Hyosung machines. If you're managing an ATM fleet and want to spend less on replacement PSUs, send us your part numbers. We'll tell you what's in stock and what's repairable.
📧 linda.zhou@microlinkpro.com | 📱 WhatsApp: +852-93572135
*CTO, MicroLinkpro — 23 years on the bench, across every major ATM brand.*
© 2026 Guangzhou MicroLinkpro Technology Company. All rights reserved.
