Whatsminer M50S – Fan Speed Error

Hard-power-cycle the miner at the PSU switch or breaker for 30 seconds, not a soft reboot. Clears wedged state in the PWM driver and occasionally re-negotiates a marginal fan header. If FAN_ERR clears on restart and doesn't return within an hour, continue monitoring — but log the event; recurring FAN_ERR without mechanical cause almost always escalates within weeks.

Visual check through the intake grille while the miner is running — do not remove the cover yet. Fan not spinning? Spinning much slower than its sibling? Debris visible? Mark the failed fan position (FAN1 front-left, FAN2 front-right, FAN3 rear-left, FAN4 rear-right) and plan a powered-off inspection. One second with a flashlight often identifies the failed unit.

Shop-vac the intake filter and grille with a soft-brush attachment. Dust cake restricts airflow enough to pull tach readings out of spec under the same PWM command — you can trigger 120/121 deviation codes on a mechanically healthy fan if the grille is 60% occluded. Vacuum first, then re-seat the miner and monitor for fault recurrence over 15 minutes of steady-state hashing.

Confirm ambient intake temperature with an IR thermometer at the front grille — not room-middle. Target below 30 degrees C for stock 26 W/T profile. Above 35 degrees C drives fans to sustained maximum duty, collapses bearing life, and can present 140 (fan speed too high) as a pseudo-fault when the miner is asking the fan to exceed rated curve.

Check WhatsminerTool firmware version against current stable release. Buggy firmware builds have miscounted tach pulses on M50S units in past release cycles. Rolling one version back or forward through official authorized builds is a legitimate Tier-1 diagnostic before ordering hardware. Never cross-flash non-MicroBT firmware — integrity-check errors 100001/100002/100003 can hard-brick.

Power off at the PSU. Remove the top cover (four M3 x 6mm Phillips #2 screws). Re-seat all four fan connectors on the control board — labeled J1-J4 or FAN0-FAN3 depending on revision. Unplug each 4-pin header, visually inspect pins for blackening or oxidation, reconnect firmly until you hear the click. Vibration loosens these headers over thousands of hours — re-seating alone clears roughly 25-35% of incoming M50S FAN_ERR tickets in D-Central's repair queue.

With cover off and power still off, spin each of the four fans by hand using a plastic stick. Healthy M50S 12038 = 5-9 second coast, silent, no hub wobble. Failing bearing = gritty, 1-3 second coast, perceptible wobble, or won't rotate. Check all four — bearing failures come in waves because the fans ran identical hours. Also inspect for chipped blades, embedded debris, dust cake, or oil-staining on the hub (grease migration confirmed).

Replace the failed fan with a matched 12038 unit — Yoeda YD12038HS12, Protechnic MGT12038XB-W25, Delta AFB1212SHE, or D-Central-supplied M50S replacement. Do not substitute quieter 120x25mm fans — static pressure will be insufficient for the push-pull topology and you'll trigger 120/121 under the same PWM command. Mounting: M4 x 30mm machine screws. Watch the airflow-direction arrow — intake fans blow INTO the chassis, exhaust blow OUT.

Consider replacing all four fans if two or more show more than 18,000 hours of operation. They shipped on the same date, ran the same duty cycle, and live in the same thermal environment — bearing failure comes in waves. A full-set swap at CAD $70-140 is cheaper than a second service call in four months, and it resets the airflow envelope back to factory spec.

Multimeter on DC, probe 12V and GND at the fan header on the control board while the miner is hashing at stock profile. Expect 11.8-12.4 V sustained. Below 11 V under load = PSU sag feeding the control board (route to Whatsminer M50S Power Supply Failure). Voltage in spec but fan still spins slow = fan itself is bad, swap it.

Advanced bench only: verify PWM signal integrity with an oscilloscope. PWM should be a clean 25 kHz square wave, 0-5 V amplitude, duty cycle scaling with thermal demand. Ringing, collapsed amplitude, or distorted waveform indicates a failing PWM driver IC on the control board — route to Tier 4 for bench-level rework.

Firmware reflash via WhatsminerTool. Back up miner config, reflash the current authorized stable release, watch for fault recurrence over 24 hours. M50S firmware signature verification can hard-brick on downgrade beyond a certain version window — stick to authorized Whatsminer builds only. Do NOT cross-flash non-MicroBT firmware on production M50S; integrity check errors 100001/100002/100003 can require manufacturer JTAG recovery.

Cable pigtail repair if chafed or damaged. 4-pin housing is standard 2.54mm JST-XH compatible. Cut the damaged section, crimp new contacts, reassemble. In practice, replacing the fan is usually faster and cheaper than splicing pigtails — labor premium on a clean splice outweighs the CAD $20 of a new fan. Listed because the option exists, not because we recommend it as first-line.

Clean fan header pins on the control board. Oxidation raises contact resistance enough to drop 12V rail voltage to the fan. Power off, unseat connector, scrub pins with 99% isopropyl on a cotton swab, dry fully, re-seat. Last resort before considering control board replacement. If a pin is visibly bent or burnt rather than just oxidized, route to Tier 4.

Control-board PWM driver reflow or replacement — bench-work only. Visible heat damage around the PWM controller (discoloration, lifted solder mask, charred IC body) means the driver itself is cooked. Reflow-station territory, not soldering iron. If not already comfortable with BGA rework on production boards, stop and ship the board to D-Central.

Stop DIY when: (a) replacement fan also faults at the same position within 24 hours confirming a bad control board tach input or 12V rail; (b) multiple fan headers show heat damage or burnt insulation; or (c) FAN_ERR co-occurs with 264 power comm error, 267 critical power fault, or 710 control-board unexpected reboot. Those combinations point to escalating control-board failure — replace the board before adjacent components take damage.

D-Central bench process on M50S control boards: test fixture with programmable PWM generator, each fan channel driven independently with tachometer response logged against target. Isolate bad header vs bad 12V rail vs bad PWM driver IC. Rework or replace board from parts inventory. Full 24-hour burn-in at nameplate before return. Board-level rework typically CAD $120-260 including labor; full board replacement CAD $140-320.

Ship safely to D-Central. Full miner: original MicroBT box with all foam inserts, or equivalent; double-box preferred for international freight. Control board only: anti-static bag, rigid mailer, at least 5 cm foam every side. Include a note listing observed error codes, affected fan position, firmware version, and your contact info. Saves diagnostic time and keeps your invoice lower.