Antminer – APW PSU Not Powering On
Critical — Immediate action required
Symptoms
- PSU does not turn on when AC is applied — no LED, no fan, no click
- Green PWR / OK LED on the APW case does not illuminate
- PSU fan never spins — not even the startup sweep
- Miner completely dead downstream — control-board LED dark, no beep, no hashboard init
- Multimeter on DC at PSU 12 V output screws reads 0.00 V on both main and standby rails
- kern.log from a previous boot shows ERROR_POWER_LOST, get power type version failed, or V:0 as last entries
- Single thunk or faint click when mains is applied, then silence — inrush NTC or soft-start relay signature
- Input fuse (T6.3AL250V on APW3 family, T10AL250V on APW9 family) reads open on continuity test
- Known-good PSU powers the same miner — fault isolated to this APW
- Known-good miner will not power on with this APW — fault follows the PSU
- PSU was running normally then cut out mid-hash and never returned
- Bulk electrolytics visibly bulged, crusted, or leaking brown residue
- Burnt smell, browned PCB, or black splatter pattern near PFC MOSFET or flyback transformer
- APW9/APW9+/APW12 fed 120 V (single-range, will refuse) or dual-range APW brown-out on 120 V
Step-by-Step Fix
Confirm the breaker isn't tripped and the outlet is live. Kill-A-Watt or multimeter at the outlet: expect 235-245 V AC on 240 V split-phase or 118-124 V AC on 120 V household circuits. If 0 V, reset the breaker; if it immediately re-trips, the fault may be downstream of the PSU — check your miner's PDU and circuit-breaker-tripping troubleshooting before blaming the APW. Three to five minutes of work eliminates 15% of 'dead PSU' tickets. Do not open the PSU before confirming mains are good.
Swap the C19 power cord for a known-good 16 AWG / 20 A cable rated for the PSU's full load. Generic 14 AWG computer-style C19 cords are not rated for a fully loaded APW9 / APW12 — they overheat, char at the inlet, and present as 'PSU won't come on' when the pins have melted. A real 20 A C19 is CAD $15-30. Verify continuity live-to-live, neutral-to-neutral, ground-to-ground with the cable disconnected before re-plugging.
Try the PSU on a second known-good outlet on a different circuit. Eliminates circuit-specific brownouts, tripped GFCIs upstream, or a bad receptacle. If the PSU powers up on the second outlet, the fault is in your electrical, not the APW. Dedicated 240 V / 20 A circuits on NEMA 6-20 or L6-20 are the correct home-mining setup for anything past an S9 — not shared 120 V / 15 A kitchen branches. Ten minutes including retest.
Verify input-voltage selector (APW3 / APW3++ / APW7 only) and the 10 A / 250 V signal-path jumper (APW9 / APW9+ / APW12 only). Older APWs with a physical voltage switch must match the mains. APW9 / APW12 need the sense-cord jumper between the control board and the PSU's signal header seated — missing jumper equals no enable equals silent PSU. If the jumper is missing or damaged, replace with a Bitmain-compatible sense cord (CAD $20-45) before digging deeper.
Test the PSU on a known-good miner, and the original miner on a known-good PSU. This single cross-swap isolates 'PSU fault' versus 'miner fault' faster than any other test. If the suspect APW powers a known-good miner — you don't have a dead PSU, you have a dead miner (often a shorted hashboard pulling the PSU into protection). If the original miner refuses to boot with a known-good APW, you have both problems. Fifteen to twenty minutes including re-seating cables.
Discharge the bulk caps before opening the PSU case. Wait 10 minutes after unplugging AC. Open the case. Find the two largest cylindrical caps (labelled 450 V or 400 V, 390-680 µF). Meter DC across each; if anything above 50 V, clip a 10 kΩ / 10 W ceramic-body resistor across the terminals for 60 seconds. Re-meter. Repeat until below 5 V. Do NOT short caps with a screwdriver — that welds metal onto the PCB and destroys adjacent parts. This step is non-optional. 390 V DC on a 470 µF cap is 47 J, enough to stop a heart.
Continuity-test the input fuse. With AC off and caps discharged, locate the glass-tube fuse near the IEC inlet (T6.3AL250V on APW3 family, T10AL250V on APW9 family). Multimeter in continuity mode, one probe each end. Beep equals good, silent equals blown. If blown, DO NOT simply replace and retry — a blown fuse is almost always a symptom of a shorted MOSFET, bridge rectifier, or bulk cap. Re-fusing into a downstream short blows the new fuse instantly and can start a fire. Proceed to step 8 before replacement.
Diode-check the PFC MOSFET and the bridge rectifier. Multimeter in diode-check mode, probe each MOSFET pin pair in turn then each bridge-rectifier pin pair. Expected: healthy diode drop one direction (~0.5 V), open (OL) the other. Shorted both directions equals dead part. Visual tells: a matte burn stripe or crater on the MOSFET top; cracking or black corners on the bridge rectifier. Replacement MOSFET (IRFP460-class): CAD $3-8. Replacement bridge (GBU806-class): CAD $2-5. Fifteen minutes per test.
Measure the NTC thermistor. Desolder one leg (or measure in-circuit with caps fully discharged and PSU isolated). Expected cold resistance 3-11 Ω depending on part (5D-11, 10D-13, etc). Open equals failed; near-zero ohms equals absorbed an over-inrush event and is now a fire risk. Replacement NTCs are CAD $2-5. Do not substitute a lower rating — undersized NTC lets too much inrush through and kills the next thing downstream.
ESR-audit the primary-side bulk caps. Use an ESR meter (MESR-100, Peak ESR70, or equivalent). Healthy 450 V / 470 µF Nichicon reads under 0.3 Ω ESR. Anything above 1 Ω ESR equals replace. Always replace as a matched pair even if only one reads bad — the sibling is hours behind on the same aging curve. Match voltage rating, match or exceed capacitance, match or exceed temp rating (105 °C), strongly prefer Japanese brands (Nichicon, Rubycon, Panasonic) over no-name replacements. Pair cost CAD $15-40.
Replace blown input-side parts as a SET, never piecemeal. If fuse blew and MOSFET is shorted: replace fuse, MOSFET, bridge rectifier, AND NTC thermistor together. The NTC absorbed the inrush that killed the MOSFET, so it's compromised even if still in-spec. Total parts APW3-family CAD $15-25; APW9/APW12-family CAD $25-45. After replacement, apply AC through a current-limited bench supply or a variac — do not go straight to full mains on a freshly-repaired PSU without a slow ramp if you own the bench for one.
Recap the primary side. If ESR audit failed the bulk caps, pull both and install matched-pair replacements (Nichicon UCS series, Rubycon USG, Panasonic EEU-FC — voltage rating 450 V or higher, capacitance matched or +20%, 105 °C rated, radial lead spacing matching the PCB). Observe polarity — reversed electrolytics explode violently. Reflow cleanly; cap lead height must leave clearance for the heatsink or case cover. 60-120 minutes depending on desolder difficulty.
Replace the PIC18F2550 or auxiliary-rail regulator on APW9 / APW9+ / APW12 if standby rail is absent with fuse, MOSFET, and caps all good. A D-Central pre-programmed PIC replacement is CAD $35-65 including the labour to flash the correct firmware image for your APW hardware revision. Bitmain does not sell these parts publicly. This is a hot-air rework job on an SOIC or QFP package — not beginner territory. If you don't own a hot-air station, escalate to Tier 4 rather than cook the PCB.
Reflow or replace the secondary-side Schottky rectifiers. If standby rail and enable are good but the 12 V rail collapses under load, the secondary Schottky dual-diodes (MBR4045PT or equivalent 40-60 A) are a common culprit — especially after a downstream hashboard short event. Diode-check each pair; replace any that short. Heatsink coupling and thermal paste must be correct on reassembly — under-torqued Schottkys overheat and fail open within weeks. Parts CAD $8-20 per package.
Post-repair bench burn-in on a programmable load before reinstalling. Before putting the repaired PSU back on a CAD $2,500-8,000 miner, bench-test it on a 12 V / 300 W or greater programmable electronic load for 4-24 hours at nameplate current with thermal monitoring. A PSU that just had fuse, MOSFET, and caps replaced either flies through burn-in cleanly or reveals a second fault that would have killed the miner on first boot. Monitor with a thermal cam if available.
Stop DIY and ship to D-Central Repair when: (a) PFC MOSFET blew with visibly damaged PCB (burnt traces, lifted pads, cratering); (b) PIC18F2550 suspected and you lack a hot-air station plus programmer with APW9 firmware image; (c) input fuse has blown twice in a row after your first MOSFET replacement; (d) capacitor bulging plus burnt-electrolyte smell — damage may extend beyond the caps. Shipping the PSU alone (not the whole miner) is cheaper than turning a repairable APW into scrap.
D-Central bench process on an APW: programmable AC-source ramp from 0 V to mains (catches a short before it blows parts). Full diagnostic — fuse, NTC, bridge, PFC MOSFET, boost diode, PFC controller, bulk cap ESR, standby rail, PIC firmware, secondary rectifier, output rail. Parts swap (we stock salvaged-grade and new-old-stock parts for every APW variant back to APW3). 24-hour burn-in on programmable electronic load at nameplate current with thermal imaging. Return with a one-page report. Turnaround 5-10 business days.
Ship safely. Pack the PSU in a double-wall box with at least 5 cm of foam on all sides. Bulk caps must be discharged before packing — heat from shipping can hurt a stressed cap. Include a one-line note with observed symptoms, APW model plus serial, and your contact. Do NOT ship the hashboards or control board — send only the PSU. Include the exact C19 cord that failed, if relevant, so we can confirm the cord wasn't the actual fault. Anti-static bag optional for PSU, mandatory if you also ship a hashboard.
When to Seek Professional Repair
If the steps above do not resolve the issue, or if you are not comfortable performing these repairs yourself, professional service is recommended. Attempting advanced repairs without proper equipment can cause further damage.
Related Error Codes
Still Having Issues?
Our team of Bitcoin Mining Hackers has been repairing ASIC miners since 2016. We have seen it all and fixed it all. Get a professional diagnosis.
