Whatsminer PSU Firmware Upgrade Failed via Test Fixture

Hard power-cycle: AC off at the wall, wait 60 seconds for bulk capacitors to bleed, AC back on. The PSU MCU watchdog will roll back to the last-known-good image when a partial write is detected. This alone clears a surprising fraction of stuck-after-flash reports without further intervention.

Re-attempt the upgrade from BTMiner web UI: `Configuration → Miner Configuration → PSU Firmware → Upgrade PSU Firmware`. If the previous attempt was interrupted by AC glitch or comm hiccup, the second pass usually completes cleanly. Watch the version string read after completion to confirm commit.

Verify the firmware image matches your exact PSU model and cooling type. `P21`, `P21E`, and `P221B` are different families with different bootloaders. Air-cooled vs hydro vs immersion images are NOT interchangeable. The single most common cause of WM_PSU_FW is the wrong binary. Read the brick label twice.

Read the PSU version string after every flash attempt. BTMiner: `Miner Configuration → PSU Firmware`. Confirm the new version (e.g. `20220420` for the test-power image). If unchanged, the commit failed and you must escalate to Tier 2.

Reboot the miner once after a successful flash. Some PSU bootloaders re-read the version flag only on cold boot, not warm boot. A clean reboot clears `Power: N/A` style misreports and lets the host re-enumerate the PSU.

Re-tighten copper-bar screws between PSU and control board to the manufacturer torque spec from the P21 series manual. Loose copper bars are the single most-cited cause of every PSU communication error in the Whatsminer line. Hand-tight is not tight enough — use a torque driver. 95% of `Error 263` / `Error 264` clusters clear here alone.

Re-seat the data cable / ribbon between PSU and control board. Visually inspect for blackening, corrosion, or bent pins. A drop of contact cleaner on a lint-free wipe, then reconnect firmly. Listen for a positive click on JST-class connectors.

Multimeter the AC inlet under hashing load. Probe at the IEC inlet on the PSU while the miner is hashing at full power. Expect 200-240 V on a 220 V brick, 100-120 V on 110 V. Sag of more than 10% from nominal under load corrupts multi-megabyte flash writes — fix the circuit (dedicated, properly sized) before retrying.

Verify the host BTMiner firmware is current. A version mismatch between host firmware and the target PSU image is rare but documented — old host firmware may fail to handshake with newer PSU bootloader. Update host BTMiner first, then retry the PSU flash.

Verify line voltage at the panel under load: 235-245 V on 240 V split-phase North American, 202-212 V on 208 V commercial, 220-245 V on 230 V European. Brownouts during the flash window corrupt writes — your neighbour's A/C kicking on at the wrong moment is a documented cause.

Move the PSU to a workshop bench fixture: a known-good H3 / H6OS / H616 control board, network-attached to the PC running `WhatsPowerTool`, AC mains, AND a fan simulator (BiXBiT documents this explicitly — fan-stall protection trips during bench work without one, killing the flash mid-write). Build the simulator from a 12 V hobby fan plus a tach-spoofing PWM module, or buy a commercial unit.

Run `WhatsPowerTool` (or `whatsminer tool 8.1.30` for the test-power workflow) on a Windows workshop PC on the bench subnet. Scan, locate the brick, click `Upgrade`, select the firmware file, click `Start Upgrade`. Typical successful flash: 30-90 seconds. Watch the progress bar; abort and restart if it hangs past 3 minutes.

Verify post-flash with the version readout in `WhatsPowerTool`. Confirm the version matches the expected build (e.g. `20220420` for the test-power image, or whatever build your replacement firmware specifies). If the version reads correctly here but BTMiner won't surface it in-miner, the problem is host-side comm, not PSU.

If you intend to use the brick for hydro/immersion operation going forward, install a fan simulator permanently. BiXBiT documents that the immersion firmware DISABLES the fan-stall check and there is no documented rollback to air-cooled PSU firmware. Flashing immersion onto an air-cooled brick is one-way.

Re-install in miner, re-tighten copper bars to torque spec, re-seat data ribbon, re-energize, and observe under hashing load for at least 30 minutes of continuous operation. Verify there are no follow-on `Error 263` / `Error 264` reports and no hex `0x0040` / `0x0800` / `0x2000` status surfaces.

Stop DIY when: the brick won't enumerate in `WhatsPowerTool` on two different known-good control boards with two different `WhatsPowerTool` versions, or visible damage is present (bulged caps, scorched PCB, burnt smell). That's bootloader corruption, MCU local-supply failure, or dead silicon — beyond what flash retries can fix.

D-Central bench process: ICT/SWD probe on the PSU MCU to read bootloader state, MCU-level re-flash if the bootloader is alive, MCU replacement if silicon is dead, local-supply (small buck/LDO feeding MCU rail) inspection and repair, full burn-in test under programmable load, post-repair full firmware load and version verification.

Ship safely: PSU in anti-static foam, double-boxed with at least 5 cm of foam on every side. Include a note with PSU model, miner host model, attempted firmware version + filename, observed symptoms, and your contact info. Saves diagnostic time on our bench, saves you money on the invoice.