Whatsminer M50S – Power Supply Failure

Kill the PDU breaker first — never hot-disconnect a live M50S PSU. The integrated `P222` is not hot-swappable; interrupting AC while DC is loaded kills the PSU controller and can brick the control board. Wait 10 minutes before opening the chassis so the PFC bulk caps drain to a safe level. This is a Tier 1 safety step that applies before every subsequent action.

Cold-boot the PSU. With AC off for the full 10 minutes, re-power the PDU and watch the first 90 seconds of boot. Code `267` (watchdog protection) and hex `0x2000` (firmware-lock) clear on a cold boot roughly 60% of the time in the D-Central repair queue — faster than MicroBT's `contact a technician` path. If the same code reappears within 5 minutes, proceed to hex code capture.

Pull the PSU hex code with WhatsminerTool V9.0.1. Select the miner, `Remote Ctrl` → `ExportLog`, open `power.log` in a text editor, and search for lines matching `psu_status: 0x`. Record every hex value you see across the last hour. The decimal error shown in the UI is downstream of this hex — MicroBT's public docs map the hex bits poorly, so the hex itself is your real diagnostic handle.

Measure AC at the miner inlet under full load. Target `≥205 V` for North American 220 V single-phase, `≥235 V` for 240 V split-phase, `≥202 V` for 208 V commercial. A logging multimeter left on the inlet for 24 hours will catch the evening-sag pattern that drives a permanent `0x0001` or decimal `206`/`250`/`251`. If AC is sagging, the PSU is innocent — fix your grid or PDU before anything else.

Verify your branch circuit. M50S pulls `~14-15 A` at stock on 240 V. `NEMA 6-20` (20 A) is the absolute minimum; 6-30 is safer once you add any derate or share the circuit. A 15 A circuit will sag at every transient load and throw input-side hex bits regardless of PSU health. Check breaker rating, wire gauge, PDU rating, and outlet type before any hardware swap.

Check your firmware version. Open WhatsminerTool → Device Info and record the firmware string. Builds `20240509` and `20240612` are known to throw spurious code `272`/`273` (over-power output warning) on healthy miners — MicroBT published no release notes. If you're on one of those, upgrade to `20240815` or later. If you're on a newer build and still alarming, firmware is not the cause.

Re-seat the I2C communication cable between the PSU and control board. Pin oxidation on this 4-pin connector is the single most common cause of decimal `263`/`264` and intermittent hex `0x0400`. Kill the PDU, unplug the cable, drop one drop of DeoxIT D5 on the pins of both ends, mate fully until you hear the latch click. Re-power and watch if the comm warning clears on the next boot.

Torque the copper busbar screws to `2.0-2.5 Nm` (`18-22 in-lb`). MicroBT publishes no torque spec. Loose busbars arc under load, throw decimal `233-235` (power output over-temp), and cascade to `236-238` / `268` as the PSU compensates. A `$30` torque screwdriver pays for itself the first time it prevents an unnecessary hashboard recap. Do this with the PDU off and the miner fully cold.

Swap the PSU internal fan before swapping the PSU. Hex `0x0100` and decimals `253`/`254`/`274` are fan-related. The stock 120 mm sleeve-bearing fan is a `$15` generic 12 V 4-wire part; replacing it takes 20 minutes and saves a `$300-$550` PSU swap. If the PSU runs cool post-swap and the hex clears, you're done. Match the original fan's connector polarity and RPM range (3000-5000).

Clean the PSU intake and the chassis intake. Dust raises internal PSU temp, triggers hex `0x0080` / decimal `275` (over-temp warning), and accelerates PFC cap aging. Blow out with oil-free compressed air at breaker-off state — never with the PSU live. An ESD-safe shop vac works for heavier dust. Follow up with isopropyl alcohol on the front grille to remove oily residue if present.

Hashboard isolation (only if hex `0x0040` / decimal `236`-`238`/`268` is set). Using WhatsminerTool `Remote Ctrl`, disable hashboards one at a time — disable `SM0` and re-test, then try `SM1`, then `SM2`. If the fault clears when one specific board is removed, that board has a chip-level short and the PSU is innocent. Community field data across multiple sources shows this is the 95% case for these codes. Route to the hashboard-not-detected repair guide.

Measure hashboard rail voltage at the copper busbar under full load. Healthy M50S reads `14.8-15.4 V` DC sustained. Below `14.5 V` means PSU output caps are tired or the feedback loop is drifting — swap PSU or ship to bench. Above `15.6 V` means OVP is about to trip on a hashboard and you risk chip damage if you keep running. Pause mining the moment you see a high reading.

Substitute a known-good `P222` PSU. Mechanical fit is not electrical fit — `P222A`, `P222B`, and `P222C` differ by batch and SKU. Verify the part number sticker on the chassis matches the replacement. If the same hex code appears on a confirmed-good substitute, the fault is upstream — control board, I2C cable, adapter plate, or harness — and you return to steps 7 and 14. Do not keep throwing PSUs at the problem.

Check for code `8700` (PSU/miner model mismatch). A `P222A` in an `M50S++` config throws this; so does a salvaged PSU from a different production batch. The control board reads PSU model ID over I2C at boot and enforces a hard-stop mismatch even with a physically identical connector. Match the chassis sticker exactly when sourcing replacement PSUs.

Update firmware only over a known-good wired network. Decimal codes `800`/`801`/`802` (checksum errors) and hex `0x2000` can be triggered by interrupted firmware flashes over flaky WiFi or a bad switch port. Use WhatsminerTool's firmware upgrade on a hardwired connection to the miner — never over WiFi — and power-cycle via the PDU after the flash completes. Confirm the new version in Device Info.

Bench-level PFC capacitor recap (Tier 3). If hex `0x0200` persists after fan and I2C fixes, the primary-side PFC capacitors are drying out — typically `Nichicon PW` or `Rubycon ZLH` `470 µF 450 V` electrolytics. Replace with `105°C`-rated equivalents, dress the leads cleanly, recap both bulk caps in matched pairs. This is a hot-air + iron job with mains-level voltage risk — short the bulk caps to ground through a bleed resistor before touching anything.

Bench-level output-stage diagnosis (Tier 3/4). Output under-voltage (`0x0008`) or over-voltage (`0x0010`) that survives a fan swap and cap recap points at the feedback optocoupler or the DC-DC controller IC. This is not a garage-tier job — the `P222` still stores hundreds of volts even after AC is removed and the bleed resistor is applied. Disconnect, label, box, and ship to D-Central.

Rule out the control board as a PSU masquerade. Decimal code `329` (control-board temp sensor comm error) and persistent hex `0x0400` often look like PSU faults but are actually the control board. Swap in a known-good control board if you have one; if the fault disappears, the PSU was innocent. MicroBT's published fix for code `329` is `replace PSU` — community data across Zeus Mining, BiXBiT, and BitcoinTalk is unanimous that this is wrong.

Inspect the adapter plate. The small PCB between the PSU and hashboards (`button board` / `adapter plate`) carries both power and sensor signals. Cracks, solder fractures, or bent pins here masquerade as PSU output faults and reproduce across PSU swaps. Visually inspect under magnification, reflow any suspect joints, and re-seat every connector. Zeus Mining and BiXBiT both flag this as an under-diagnosed failure mode.

Cold-climate sanity check (Canadian-specific). If your M50S intake is below `0°C` — a garage in a Quebec January — warm the miner to room temperature for 30 minutes before energizing. Electrolytic caps in the `P222` lose ESR margin at cold start and can trigger spurious hex `0x0080` / `0x0200`. MicroBT documents nothing below `0°C`. D-Central does — because we mine through Canadian winters and the caps hate cold inrush.

Stop and ship to D-Central when: (a) hex `0x0008`, `0x0010`, `0x0800`, `0x1000` or `0x2000` survives a cold boot; (b) you see bulged caps, discolouration, or smell burnt electrolyte; (c) busbar output stays below `14.5 V` under full load with a known-good control board; (d) a known-good P222 reproduces the same hex code (upstream fault); or (e) any mains-side repair is outside your comfort zone. Pack the miner with the PSU bolted in, include the hex codes, decimal codes, inlet voltage, and breaker size on the packing slip, and book a slot at d-central.tech/services/asic-repair/.