APW12 Fan Stalled But PSU Still Outputting

Cut AC at the breaker immediately on confirmed fan stall. Not the AC switch on the PSU casing — the panel breaker. Walk away for 30 minutes. Every minute of continued operation accelerates capacitor degradation. The single highest-impact action on this entire page is the one you take in the first 60 seconds. Tag the cord 'do not energise' and let the chassis cool fully before any handling.

Visually inspect the fan grille for foreign objects with AC off and chassis cool. Shine a flashlight through both grilles. Check for cardboard sucked against the intake, cable ties caught in the rotor, dust mats collapsed onto the blades, or insects. If you can remove a blockage from outside the chassis, do it, then bench-test before re-deploying. Most 'dead fan' reports trace to mechanical obstruction, not motor failure.

Confirm input AMP-draw trend on the dashboard. Pull the last 7 days of input current from your smart plug, PDU, Home Assistant, Grafana, or whatever monitoring you run. A slow upward drift of 1-3% on a constant-load miner = the fan has been stalled or marginal for hours-to-days; treat the unit as suspect even with green LED, replace caps proactively. No trend = the stall is acute and component damage is minimal.

Smell the casing carefully with AC off, chassis cool but not cold. Sniff intake and exhaust. Burnt-plastic, sweet-acrid (electrolyte out-gassing), or scorched-PCB odours all confirm internal damage from the overheat. Any of these → unit ships to bench, do not re-energise. Clean neutral smell = damage may be limited and bench triage is worth doing before declaring the unit dead.

Document and tag the unit before pulling from the rack. Photograph the rack position, dashboard input-AMP-draw history, LED state at fault, casing temperature with an IR thermometer, and any audible cues. This evidence pack saves bench-diagnostic time at D-Central or your own bench by ~50%, which directly saves repair-cost dollars. Tag the unit physically with masking tape and a Sharpie note.

Open the casing with AC unplugged and bulk cap discharged. Discharge the bulk cap with a 10kΩ 5W resistor across its terminals; wait 5 minutes minimum. Cover off. Look for: bulging or leaking secondary capacitors, discoloured PCB around the LLC FETs or bulk cap, melted or scorched plastic on the fan or its connector, dust caked on the heatsinks. Photograph everything. Any visible damage → Tier 3 regardless of fan diagnosis.

Bench-test the fan on a known 12V supply. Disconnect the fan's 2-pin connector from the PSU PCB. Connect to a benchtop 12V supply or a 12V wall-wart pigtail. Fan runs cleanly = motor is fine, fault is in the PSU's fan rail (Tier 3 territory). Fan does not run = motor dead, replace it (Step 8). Fan runs erratically = motor degraded, also replace.

Replace the chassis fan: 60mm × 25mm thickness, 12V, 2-pin, ball-bearing preferred over sleeve-bearing for mining-duty longevity. Match connector polarity carefully — APW12 fan connectors are not universally keyed across hardware revisions. Re-route the cable through the same factory path; do not pinch under the cover. AC-apply briefly (no load, 30 seconds) and confirm the fan starts within 5 seconds and runs steady.

Clean the intake and exhaust grilles thoroughly. Compressed air at the intake direction (so dust exits the way it came in), vacuum from the exhaust side, never run a brush into the chassis. Wipe the grille mesh with a dry lint-free wipe. Clogged grilles raise internal temperature by 8-15 degrees C even with a healthy fan and are a leading cause of secondary fan stalls. Highest-ROI maintenance step on any APW-family PSU.

Bench-test under representative load before re-deploying. Casing closed, programmable DC electronic load or known-good miner connected, run 60 minutes from idle through 3.0 kW. IR thermometer or thermal camera every 10 minutes on: LLC heatsink, bulk capacitor surface, secondary cap bank surface, output bullet bases. All surfaces should stabilise below 75 degrees C in 22-25 degrees C ambient. Anything above = component damage, route to Tier 3.

Replace the secondary capacitor bank if any cap shows distress. ESR meter on every secondary cap with the unit unplugged and bulk cap discharged. Replace the entire bank in matched pairs — same brand, same series, same voltage rating, 105 degrees C low-ESR series strongly preferred. Mismatched caps in a parallel bank cause uneven sharing and re-failure within months. Reflow with proper through-hole technique; flux-clean with isopropyl 99% afterward.

Inspect and replace the bulk capacitor if cooked. The primary-side electrolytic between bridge rectifier and PFC is the single most heat-sensitive part on the APW12. Bulging top, leaked electrolyte at the base, or ESR drift above spec = replacement. Match voltage rating (typically 450V), capacitance (typically 470µF or 680µF on APW12 hardware revisions), and 105 degrees C low-ESR series. Wrong cap here causes PFC instability and walks the unit toward the next failure.

PWM control transistor or fan-rail regulator repair. If the bench check showed 0V at the fan connector with the LED green, the fault is on the fan-rail control circuit. Identify the fan PWM transistor or 12V regulator IC on the control PCB, verify with multimeter and oscilloscope (PWM signal at the gate during normal operation), replace the suspect part. Straightforward for a competent technician — brings the fan rail back without a full PSU replacement.

Refresh thermal pads on the secondary heatsink. The pads between the secondary FETs / synchronous rectifiers and the heatsink dry out and lose conductivity after 3-5 years in mining duty. A unit that just survived a fan-stall episode is a candidate for pad refresh even if no component is visibly damaged — the heat exposure dries marginal pads. Use 2 W/m·K thermal pads minimum, cut to size, replace in matched pairs across the heatsink.

Reflow any solder joint that shows hairline cracks under 10× magnification. Years of thermal cycling crack joints around heavy-current paths — secondary rectifier pins, fan connector pins, output bullet bases. A loupe inspection after a thermal episode is mandatory; reflow any joint showing a ring crack. Use leaded solder for repair of leaded joints (don't mix tin-lead and lead-free unnecessarily). Flux-clean and inspect post-reflow.

Document component failures and retain the failed parts. Bag and label every replaced component with date, observed failure mode, and unit serial. This builds a private failure-population dataset that informs future maintenance scheduling — D-Central's own repair-bench data on APW12 capacitor lifetimes comes from exactly this discipline applied across hundreds of repairs over the years.

Stop DIY when: bulging primary bulk cap, multiple secondary caps showing distress, visible PCB scorch in the LLC area, audible 60 Hz hum that doesn't quiet after cap refresh, or any uncertainty about line-voltage safety procedure. The APW12 primary side runs on a 380V bulk rail; this is not the place to learn line-voltage safety. Book a D-Central ASIC repair slot at d-central.tech/services/asic-repair/.

What D-Central does at the bench: full chassis teardown on a soldering station with proper line isolation, fan replacement with a known-good ball-bearing OEM-spec part, full secondary cap bank refresh in matched pairs, bulk cap replacement if compromised, thermal-pad refresh on every heatsink interface, PWM-rail repair if applicable, full bench load-test from idle through 3.4 kW with thermal imaging on every component surface, 24-hour soak under simulated S19/S21 load before return.

Ship safely. APW12 is heavy (~5 kg) and the secondary cap bank is fragile. Original Bitmain box if you have it; otherwise double-cardboard with at least 5 cm of foam on every side. Pull all PCIe cables, coil them, zip-tie them separately in their own bag inside the carton. Include a printed note with: observed LED state at fault, dashboard input-AMP-draw history (screenshot), how long the unit ran with stalled fan before shutdown, ambient conditions, your serial number and contact info. Every minute saved at the bench is a dollar saved on the invoice.