PSU Problems and Solutions: Tips for APW9/APW12 Power Supplies

When an Antminer goes dark, the power supply is guilty far more often than the hash boards. This guide walks the four failure signatures used to triage an APW9, APW9+, or APW12 — no fan, fan-but-no-output, cyclic stop/start, and output-but-no-fan — and tells you what to check, in what order, before you decide between a bench repair and a replacement. Work through them top to bottom; each rules out a whole class of cause.

Know which PSU you have first

The APW9 and APW12 families live in different worlds, and the same symptom can mean opposite things depending on which is on your bench. Confirm the model on the label first:

• APW9 — 14.5–21V DC output, 170A, ~3600W at 220V. Pairs with the S17, S17 Pro, T17.
• APW9+ — same 14.5–21V window but 200A, for the higher-draw S17+, S17e, T17+, T17e.
• APW12 (1215) — 12–15V DC, 233A, ~3600W. Pairs with the S19, S19 Pro, S19j, S19j Pro, T19.
• APW12 (1417) — 14–17V DC variant for the L7, K7, KA3 and similar.

Input voltage is the critical fork. The APW9/APW9+ are 200–240V units that will still start and run on 110/120V at sharply reduced output — low output there is expected, not a fault. The APW12 is 200–240V only: its undervoltage protection (UVP) trips when the AC input falls below roughly 180V, so it simply refuses to start on a 120V circuit. An S19 that won’t power up on 120V is a wiring decision, not a broken PSU.

Safety before you open anything

The 12–21V DC output rails and their copper bus bars are not a shock hazard — probe them live all you like. The danger is the AC side: the PFC bulk capacitors hold roughly 410–420V and stay charged for minutes after you pull the cord. If you crack the case, discharge them through a proper bleeder resistor and confirm 0V before touching the primary side. Never hot-plug the 4-pin I2C/signal connector; its 3.3V logic is easily killed.

The four PSU fault signatures

1. Fan does not spin, no output

This is almost always an AC-side problem — the PSU is not being powered at all.

1. Confirm the IEC cord is fully seated at both ends and rated for the load. A 3600W unit needs a heavy-gauge C19/C20 cord on a dedicated circuit, not a thin kettle lead.
2. Meter the wall outlet under load. Confirm you actually have grid voltage and that a breaker or PDU port has not tripped.
3. Swap in a known-good cord and outlet. Many “dead” PSUs are dead cords.
4. If AC is confirmed good at the inlet and the unit is still lifeless, the fault is internal (blown input fuse, failed PFC stage) — that is a bench repair, not a field fix.

2. Fan spins normally, but no DC output

The primary side is alive but the PSU is either starved for input or sitting in protection mode.

1. On an APW12, meter the AC input and confirm it is a solid 200V+ (ideally 205V or more) under load. A sagging 208V circuit shared with other miners can dip below the start threshold. On an APW9 running 120V, remember reduced output is normal.
2. Check for a short or overload on the output. A shorted hash board or a pinched bus-bar lug will drop the PSU straight into short-circuit (SCP) or overpower (OPP) protection.
3. If it tripped, power off for a full 10 minutes so the protection latch and thermal mass fully reset, then re-apply power with the load disconnected. If it now outputs cleanly with no load, the fault is downstream in the miner, not the PSU.

3. PSU stops for a few seconds, then resumes — cyclically

This hiccup pattern is the classic over-temperature (OTP) signature: the unit shuts the main output, cools slightly, auto-recovers, heats up, and trips again.

1. Watch the internal fan. If it is not spinning while the unit is trying to run, the PSU is cooking itself — stop and send it for repair.
2. Inspect the cooling air duct for blockage — cable spaghetti, a wall pushed against the intake, or a filthy filter will do it. Clear it.
3. Blow out accumulated dust. A dust-packed heatsink raises internal temps at any load.
4. Check that output load and ambient temperature stay within the unit’s power-limit curve. Running a 3600W PSU at the ragged edge in a hot, unventilated room invites OTP. Improve airflow or reduce load.

One caveat: if the fan is healthy and airflow is clear, a cyclic output dropout can also be a firmware watchdog timeout — a crashing control board that stops sending its heartbeat makes the PSU deliberately drop the hash-board rail (~60s timeout) as a safety feature. That points you at the control board, not the PSU.

4. Output is normal, but the fan does not spin

Here the power stage is fine; the cooling fan itself has failed. Left alone this becomes fault #3.

1. Check whether the fan is physically jammed by debris and clear it.
2. If the blades are free but the fan still will not turn, the fan or its internal 12V fan rail has failed. Do not keep mining — run it and it will overheat and trip OTP or die. Send it for fan replacement.

Protection mode: what trips and how it clears

The APW9/APW12 hardware protections are independent of firmware and cannot be disabled. Knowing the recovery path saves you a needless repair shipment:

• OVP (overvoltage) — immediate shutdown; requires a full power cycle to clear.
• UVP (undervoltage) — AC input below ~180V; restore proper voltage (the APW12-on-120V trap).
• OCP (overcurrent) — output above ~250A; current-limits or shuts down, usually auto-recovers or clears on a cycle.
• OTP (over-temperature) — shuts down, auto-recovers once cool (fault #3).
• SCP (short circuit) — immediate shutdown; remove the short, then power cycle.
• OPP (overpower) — total load above ~3600W; current-limits until you reduce load.

Protection exists to prevent fire and hardware damage when load current goes abnormal. A PSU that keeps tripping under a normal load is telling you the fault is real — do not defeat it, find it.

Repair or replace?

No output with confirmed-good AC, a dead internal fan, or a unit that trips protection under a known-good load all point to an internal fault — a blown fuse, a failed PFC/LLC stage, or a bad fan. Those are bench-level repairs with real parts, not field swaps. Before spending money either way, isolate the PSU from the miner and bench-test it against a known-good load: it confirms whether the fault is the PSU or the machine it was feeding.

Related: pin down the failure with the ASIC fault finder, source a bench-tested replacement or spares from ASIC repair parts, or ship the unit in through start a repair. New to the terminology? The PSU glossary entry covers the fundamentals.