APW12 Red LED Fault and Diagnostic Codes

Hard AC-cycle for 60 seconds at the breaker, not just the AC switch on the PSU. APW12 protection latches in non-volatile state on some firmware revisions; only a full AC discharge clears the latch. Wait the full 60 seconds — premature reapplication can re-arm the same trip if the bulk cap has not fully discharged. If green returns and stays for 10 minutes under load, the trip was transient.

Pull the C19 cord and inspect the contact pins for blackening or oxidation; reinsert firmly until you feel the click. Do the same for every 8-pin PCIe output bullet — APW12 PCIe contacts wear after dozens of hot-swaps. Loose contacts arc, the arc damages contacts further, and the PSU eventually trips OCP on the now-resistive joint.

Verify ambient at the PSU intake. Put your hand at the intake — should feel cool. If the room is over 35 °C, or the PSU is intaking warm exhaust from another miner, or a piece of cardboard is blocking the grille, fix airflow before doing anything else. OTP trips on the APW12 are dominantly environmental.

Confirm 240V at the outlet under load. APW12 is 240V-only; on a 110V circuit the LED sits solid red because PFC never reaches its operating bus voltage. Multimeter under no load should read 235-245 V on a healthy 240V split-phase outlet, 215-225 V under heavy mining load.

Try the suspect PSU on a known-good miner, and a known-good PSU on the suspect miner. This single swap collapses 80% of `is it the PSU or the miner` ambiguity. Solid green on a known-good miner = problem is the miner. Blink-red on the suspect miner with a known-good PSU = miner is pulling out-of-spec current.

Multimeter on DC, probe red on a PCIe bullet centre pin and black on case ground. With no miner connected, expect 14.5-14.9 V steady. Under full hash, expect 13.8-14.2 V sustained. Brief 14.5 V then collapse to 0 V as the LED drops to red = unit is OVP-tripping on the secondary, route to Tier 3+.

Inspect every PCIe output cable for damage at the strain relief — APW12 pigtails split there if a miner has been moved frequently. A nicked positive conductor that bridges to negative under vibration triggers SCP latch on next boot. Replace any cable with visible damage. Never repair-tape a PCIe pigtail running 3 kW.

Confirm PSU casing fan operation. With LED red, listen for a fan tick on AC apply. With LED green, verify the fan ramps with PSU temperature within 30 seconds of load. Dead PSU fan = OTP latch is inevitable within 10-15 minutes. Replace the fan (60mm, 12V, 2-pin APW12 service part) before re-deploying.

Clean the PSU intake and exhaust. Compressed air at the intake, vacuum from the exhaust side, never run a brush *into* the PSU. Dust-clogged grilles raise internal temp by 8-15 °C; an APW12 running at 85 °C internal is one ambient spike from an OTP latch. Highest-ROI maintenance move on any APW12.

Test the C19 cable and outlet under heavy load. Plug a high-load resistive heater into the same outlet through the same cable. If voltage sags below 215 V at the outlet, your circuit is undersized — the PSU is OCP-tripping because supply voltage drops drag current up to compensate. Move to a dedicated 240V/30A circuit minimum.

Tier 3 — primary-side inspection: unplug AC, wait 5 minutes, discharge the bulk capacitor with a 10kΩ 5W resistor between its terminals. Open the casing. Inspect AC entry fuse, NTC inrush thermistor, bridge rectifier, PFC MOSFET, PFC boost diode, and bulk capacitor for bulging or leaking electrolyte. Visible damage = component-level repair candidate.

Auxiliary flyback diagnosis. With AC re-applied (cover off, line voltage exposed — only if competent), measure across the housekeeping flyback's secondary output filter cap. Expect ~12 V steady. Absent or oscillating = aux flyback failure. Common root cause: dried primary-side bulk cap on the aux. Replacement is ~$2 in parts and brings the unit back.

Visually inspect the opto-coupler near the secondary feedback resistor divider. Cracked package, burnt board, or a resistor drifted >5% will produce OVP latches. Reflow or replace the opto, and replace feedback resistors in matched pairs (1% precision is the spec).

ESR meter on the secondary 12V output capacitors — the bulk caps directly between LLC secondary rectifiers and PCIe bullets. Bulging tops, leaked electrolyte at the base, or ESR > 50 mΩ on any cap = replace the entire bank. Mismatched caps in a parallel bank cause uneven sharing, localized heating, and accelerated re-failure.

LLC half-bridge MOSFET test. With unit unplugged and bulk cap discharged, diode-test the LLC primary FETs. Drain-source short = catastrophic primary failure. Highest-value-at-risk repair on the APW12; get it wrong and you cascade-damage the PFC stage. Strongly consider Tier 4 if uncertain.

Communication-latch firmware ruleout. Some APW12 firmware revisions are documented to latch red on transient comms-line glitches that newer firmware tolerates. Bitmain's PSU update procedure is poorly documented — check support.bitmain.com for your specific build, and only re-flash with explicit guidance. Wrong firmware on the wrong hardware revision can permanently brick the protection MCU.

Tier 4 — when to stop DIY: bridge rectifier shorted, PFC FET shorted, aux flyback dead, LLC primary FETs damaged, or visibly bulged secondary cap bank. Any of these justify shipping rather than DIY-ing on a live 380V bulk rail. Component cost is low but line-voltage risk and primary trace-rework difficulty earn a Tier 4 ticket. Book a D-Central repair slot.

What D-Central does at the bench: full primary-side teardown, fuse + NTC + bridge replacement on a soldering station with proper line isolation, aux flyback rebuild with matched-spec capacitors, secondary cap bank refresh, opto and feedback-resistor pair replacement, full bench load-test from idle through 3.4 kW peak with thermal imaging on every stage, 24-hour soak under full S19/S21 load before return.

Ship safely. APW12 is heavy (~5 kg). Original Bitmain box if available, otherwise double-cardboard with at least 5 cm of foam on every side. Pull all PCIe cables, coil and zip-tie them separately. Include a note with observed LED behaviour, AC environment, and any diagnostic readings — every minute saved at the bench is a dollar saved on the invoice.