An AvalonMiner that won’t start — dead LEDs, no fan spin, or hashboards that never come up — is more often a failed power supply than a failed miner. The Avalon PSU is a two-stage switching supply: mains AC feeds an EMI/inrush front end, a Power Factor Correction (PFC) stage builds a high-voltage DC bus of roughly 390V, and a main LLC transformer (TR1) steps that down to the 12V rail that feeds the hashboards, plus a small auxiliary 12V rail for the fan and control logic. Diagnosis means walking that chain from AC in to 12V out, finding the first stage with no correct voltage, and repairing the component that broke there. This is component-level bench work that assumes you can solder and read a schematic; if you can’t, treat the PSU as a replaceable brick.
Safety first: only the primary side can kill you
There is exactly one lethal section in this supply: the PFC high-voltage DC bus and its bulk capacitors (~390V), plus everything on the mains/primary side of transformer TR1. Those caps hold a lethal charge for minutes after you unplug. The 12V output rail and low-voltage secondary side are not a shock hazard — probe them freely.
- Unplug the AC cord and wait several minutes before opening the case.
- With the case open, put your DMM on DC volts and measure across the two large PFC bulk capacitors. Do not touch anything until they read under ~30V.
- If a cap still holds charge, bleed it with a resistor (a few kΩ, 5W+) across the terminals — never a bare screwdriver, which pits the plates and can spit molten metal. Keep the unit unplugged for every measurement that isn’t a live power-on test.
Tools and materials
- True-RMS multimeter (continuity, DC volts, capacitance, diode test); an oscilloscope for ripple.
- Temperature-controlled iron and hot-air rework station for SMD and the big electrolytics.
- ESD wrist strap and mat; an isolation transformer or current-limited variac for safe primary-side live testing.
- Exact-spec replacement parts: capacitors of matching µF, voltage and low-ESR rating; MOSFETs, diodes and the NTC inrush thermistor by original part number.
Step 1 — Confirm it’s the PSU, not the miner
Before you open anything, prove the fault is in the supply. Swap in a known-good PSU of the same model, or move the suspect PSU to a known-good miner: dead on the bench is a supply fault; working elsewhere points back at the control board or hashboards. Then, unplugged, measure resistance across the AC input and the 12V output. A near-zero reading on either is a hard short — find it before you apply power.
Step 2 — Work the failure modes in order
No ~390V on the PFC bus (completely dead supply)
Cause: the front end or PFC stage isn’t building its DC bus, so nothing downstream can run. Diagnose: with the caps confirmed discharged, ohm out the input path — the F2 fuse, the RB1/RB2 bridge rectifier and the RT1 inrush thermistor. A blown F2 with a shorted bridge or PFC switch is the classic “plugged in, nothing happens” pattern. Fix: check the PFC switching MOSFETs and their drive network (Q32, Q33, Q34 with resistors R311, R309, R314, R329, R330) and the PFC diodes (D5, D9, D13); replace any shorted or open part like-for-like. Never replace only the fuse — a blown F2 is a symptom; fix the short that killed it first, or you’ll pop the new fuse instantly.
PFC bus present but no 12V output
Cause: the main or auxiliary DC/DC conversion after TR1 has failed while the PFC front end is healthy. Diagnose: confirm the ~390V bus is good, then check the main switching MOSFETs, their sense resistors, and the PWM/controller IC; separately verify the small auxiliary 12V rail — if the aux rail is dead, the main stage never gets its enable/PWM and stays silent. Fix: replace the failed MOSFET/driver/IC and confirm the enable and feedback signals are present. A supply that builds 390V but makes zero 12V is almost always a secondary-side switch or controller fault, not a PFC problem.
High ripple on the 12V rail (miner unstable, random hashboard drops)
Cause: aged or failed output filter capacitors let switching noise through, so the 12V averages fine on a DMM but swings badly under load — the classic “boots then throws intermittent hashboard/power errors” fault. Diagnose: scope the 12V rail under load; healthy ripple is tens of millivolts, not hundreds. Inspect the output electrolytics for bulging tops or crusty electrolyte. Fix: replace the output filter bank (in these units the C31/C32, C40/C41, C46–C49, C51/C52, C55–C57, C59/C60, C62–C66, C70–C73 group) with identical low-ESR parts. Match ESR and ripple-current rating, not just capacitance — right µF but wrong ESR will fail again fast.
Fan failure
Cause: a seized, blocked or wire-corroded fan lets the supply overheat and shut down. Diagnose: spin the fan by hand, inspect the leads for corrosion or breaks, and confirm ~12V reaches the fan header. Fix: clear blockages, repair corroded leads, or replace with the same voltage, size and airflow rating — a lower-CFM fan defeats the supply’s thermal design.
Input / on-off switch and cable faults
Cause: a failed input switch, damaged AC inlet, or a chafed/disconnected 12V output cable. Diagnose: ohm the switch through both positions, wiggle-test the AC inlet, and inspect the full 12V harness for crushed insulation, cracked lugs or a broken conductor at the strain relief. Fix: replace with the correct current-rated part. A partly-broken output conductor raises resistance, drops voltage under load and cooks the connector — check for browned or melted housings.
Repair discipline
- Stay on the ESD strap for the whole job, and replace parts with the identical model and spec — no unauthorized substitutions on a safety-critical mains supply.
- Bag and label every part you pull, and log the PSU serial, symptom, root-cause component and fix — that history is how you spot a batch-level failure across a fleet.
Verify before it goes back in a miner
- Re-check for shorts across the AC input and 12V output, unplugged.
- Bring the primary up through an isolation transformer or variac; confirm the PFC bus reaches ~390V without excessive current draw.
- Confirm the 12V rail is on-spec and, under load, that ripple is back to tens of millivolts on the scope.
- Load-test at the miner’s real draw to thermal steady state — many PSU faults only show up warm and loaded, not on a cold no-load bench.
When to stop and escalate
If the PFC caps won’t bleed down, shorts keep returning after replacing the obvious parts, or the board shows heat damage across multiple stages, retire the supply rather than chase a moving fault on a live 390V bus. A replacement PSU is cheaper than a burned hashcenter.
Related: Diagnose which unit actually failed with the ASIC fault finder, source the exact fuses, MOSFETs and capacitors from ASIC repair parts, and if you’d rather have it fixed on a bench, start a repair.
