Bitaxe – Fan Not Spinning / PWM Stuck at Zero

Manual fan override + hard power-cycle. AxeOS > Settings > disable `Auto Fan Control` > `Fan Speed: 100%` > save. Power-cycle at the wall outlet, not the AxeOS soft-reboot button (soft reboot keeps peripheral state wedged). Watch the fan for 60 seconds post-boot. If it spins, the auto-fan curve is the entire problem - skip to Step 6 for a firmware fix, or leave on manual 100% as a stopgap. Manual 100% is louder but keeps the ASIC alive - on an unheated Bitaxe that is the right trade-off every time.

Reseat the fan connector. Power off at the wall. Unplug the fan from its header, inspect pins for bent contacts, flux residue, or a crushed shell. On GT/Hex it is a 4-pin PWM header and the connector only seats one way - verify polarity against silkscreen (V+, GND, TACH, PWM). On Supra/Ultra/Gamma it is a 2-pin; if someone reversed polarity during a mod, the fan simply will not spin. Re-click firmly until flush, power on, test. This alone fixes about 20% of `dead fan` tickets D-Central sees on shipped-in boards.

Blow out the fan. Canned air, held upright, short bursts. Hold the blade stationary with a plastic probe while blasting - overspinning the motor through the blade induces back-EMF that stresses the bearing. Dust buildup between the blade hub and the motor stalls a tired fan. Pair this with a visual check for hair, lint, or the classic Canadian-basement failure, cat fur.

Power-cycle after warming the room. Bitaxe Gamma 601/602 have documented I2C race conditions that worsen below 15 C ambient. If your shop is cold, move the miner to a warmer room for 2 minutes, then power-cycle at the wall. If the fan now spins, you have a cold-boot timing issue on the EMC2101 I2C bus - read the I2C handle-not-initialized playbook next for a permanent firmware fix.

Check AxeOS firmware version against known-bad releases. AxeOS > System > Firmware Version. Cross-reference against ESP-Miner #66 (v2.0.4 fan at 0 RPM), #599 (auto-fan reset loop on v2.4.2), and any v2.8.0+ release before a fan-curve fix lands. If you are on a flagged version, jump to Step 6 for a firmware flash.

Flash one release back or forward, erase NVS, reconfigure. Download the previous stable AxeOS release from the ESP-Miner releases page. Flash via USB using the Bitaxe Tool or `esptool.py write_flash`. Before rebooting, run `esptool.py erase_flash` to wipe NVS - stale fan-curve settings can survive a firmware downgrade if NVS is not cleared. Reconnect to AxeOS, reconfigure Wi-Fi and pool credentials, verify fan behaviour for 15 minutes. Pin this release in your notes; do not auto-update without checking the changelog for fan-curve fixes.

Replace the fan with a known-good unit. Stock Bitaxe fan is a 40 x 40 x 10 mm 5 V axial on Supra/Ultra/Gamma (2-pin JST); GT and Hex run a 40 x 40 x 20 mm PWM 12 V axial (4-pin). D-Central stocks both. Before soldering or crimping a new connector, pin-check polarity: V+ / GND on 2-pin; GND / V+ / TACH / PWM on the 4-pin. Power on, set manual 100%, verify spin and tach reading. If the new fan spins, the old one was dead. If the new fan also will not spin, the fault is upstream - continue to Step 8.

Measure the fan header under command. Power on. Fan unplugged. Manual 100% in AxeOS. Multimeter on DC, probes at the header pins. Expect 4.8 V (2-pin) or 12 V (4-pin) on the V+ rail, steady. On 4-pin also probe PWM - expect 5 V steady-high or a 25 kHz square wave at 100% duty. 0 V on V+ means supply-rail fault or dead MOSFET (2-pin). 0 V on PWM means controller is not driving. Record which pin is dead and carry that into Tier 3.

MOSFET / gate-resistor rework on 2-pin models. Advanced users whose multimeter shows V_GS = 0 V under command can replace the gate resistor or the N-channel MOSFET itself. This is a SOT-23 or SOT-23-3 part - 5-10 min rework with hot air and flux. Verify the designator against the Bitaxe open-hardware schematic on the bitaxeorg GitHub before desoldering - the part varies between Supra and Gamma revisions.

Reflow the EMC2101. If scope showed no PWM activity and serial log showed `EMC2101 write failed` or handle-not-initialized, the chip or its solder joints are suspect. Flux the SOT-23-10 package, hot air at 290-310 C for 15-20 seconds, let cool naturally. Do not dwell - the EMC2101 tolerates reflow but is small enough to desolder unintentionally. Practice on a Bitaxe Hex first - same EMC2101 package, same I2C topology, much lower blast radius if you slip. D-Central pioneered Bitaxe Hex heatsinks specifically because the Hex is the ideal teaching bench for this family of rework.

Replace the EMC2101 chip. Part is Microchip EMC2101-ACZL-TR, SOT-23-10, about 3 CAD. Lift the old chip with hot air, clean pads with solder wick and flux, place the new chip matching pin 1 to the board silkscreen marker, reflow at 290 C. Verify orientation before applying heat. Re-power, re-flash AxeOS if NVS was corrupted during the dead-controller period, and run the verification loop in Diagnostic Step 7.

Build custom AxeOS firmware with an explicit EMC2101 init + duty floor. Fork ESP-Miner. Patch the fan-init path to: (a) retry the EMC2101 probe three times with 50 ms delays if the first fails, (b) write a minimum 10% duty to register 0x4C at init regardless of auto-curve output, and (c) log actual register values read back over serial. Build with `idf.py build`, flash with `idf.py flash`. This is the Mining Hacker workaround for the 0% auto-curve bug - D-Central ships this patch on every repaired Bitaxe.

Add an external pull-down on the PWM line. Rare but seen: if the fan PWM pin is floating because of a damaged trace or unpopulated component, tack-solder a 10 kOhm 0603 from PWM to GND on the bottom of the board to force a defined off state on the MOSFET gate. Does not fix a 0% auto-curve bug but does fix flaky `sometimes the fan runs, sometimes it does not` behaviour on units with cold solder on the PWM trace.

Full fan-subsystem rework: EMC2101 + MOSFET + fan together. If you have chased the problem through the controller, the gate, and the fan itself on the same board, stop burning single parts - swap the whole subsystem in one pass. Replace EMC2101, MOSFET (on 2-pin models), and the fan. Verify with scope and multimeter at each step. This is the repair D-Central ships when a Bitaxe arrives with stacked failures - cheaper than chasing one fault at a time and proves the fix on the bench.

Stop DIY and ship to D-Central. Triggers: EMC2101 replacement did not restore PWM, scope shows ESP32-S3 is not clocking SDA/SCL at all, you lifted a pad during rework, you see burnt-component discoloration around the fan subsystem, or multiple Bitaxe units in your rack show the fault simultaneously. That is SoC-level or batch-QC debug territory. Turnaround 5-10 business days, Canada-wide, US/international welcomed. D-Central has been in the Bitaxe ecosystem since the first production PCBs - we built the Mesh Stand, pioneered the Hex heatsinks, and publish open-source fixes back.

Confirm the fix. After any tier applied, run the miner at full hash for 15 minutes. Verify Fan RPM > 3000 (2-pin Supra/Ultra/Gamma) or > 5000 (GT/Hex 4-pin) and ASIC junction temp settles below 60 C. Capture serial log to confirm zero `EMC2101` errors across the run. Anything else - intermittent spin, tach reading zero while fan is visibly moving, temp above 70 C - means the fault is still present. Escalate to the next tier.