Bitaxe – Thermal Shutdown at 75C / Self-Protect Mode

Power-cycle for 60 seconds. Unplug main PSU and USB-C, wait a full 60 seconds, reconnect. This drains input capacitors and forces a clean boot - some firmware versions do not reliably auto-clear the self-protect latch without a full discharge. Check AxeOS: is the Device Overheat banner gone? If yes, the chip has cooled and the latch cleared. Now figure out why it tripped before it trips again.

Relocate the miner to a cooler, better-ventilated spot. A Bitaxe tripping self-protect in July in a south-facing office is telling you something about the room, not the miner. Target ambient at the fan intake is 25 C or less, with at least 15 cm of clear space above the heatsink and nothing blocking the exhaust side. A Canadian basement in summer, a home-office floor (air pools cooler there), or a server closet with a return-air vent all work. Stacking on top of another miner, tucked behind a monitor, or sitting on carpet are all classic anti-patterns.

Set fan to Manual 100% in AxeOS. Settings -> Fan Control -> Manual -> 100%. Auto-fan is fine in winter; in summer at aggressive OC it lags the temperature rise and lets the ASIC reach trip threshold before the fan reaches max. Manual 100% is around 3 dB louder but gives you constant maximum heat rejection - the single easiest Tier-1 fix if you are not bothered by fan noise.

Drop to stock frequency and stock Vcore. Settings -> ASIC: Ultra 485 MHz / 1.25 V, Supra 525 MHz / 1.25 V, Gamma 575 MHz / 1.15 V, GT 575 MHz / 1.15 V, Hex 485 MHz / 1.25 V. Apply, reboot. If self-protect stops tripping at stock, your OC profile was too aggressive for your current thermal environment. You can tune back up later (Tier 2 step 8), but verify stock stability first.

Update AxeOS to the latest release. Several self-protect-related bugs have been fixed post-v2.9.1: the stuck-overheat-message latch (#181, #342), the auto-fan spurious-trigger (#599), and the I2C/EMC2101 race condition (#1291). Check the ESP-Miner releases page, flash the latest stable via Bitaxe Web Flasher, power-cycle, re-check. Full factory-reset NVS between flashes if issues persist.

Clean the fan and intake path. Unplug, remove the fan (usually 4 Phillips #1 screws), blow compressed air through the heatsink fins from the exhaust side (pushes dust OUT, not deeper in), wipe the fan blades with IPA on a lint-free wipe. Reassemble. A restricted airflow path can drop effective thermal rejection by 30-50% - enough to trigger self-protect on a miner that was fine 18 months ago.

Re-paste the thermal interface. Power off, remove heatsink, clean die and heatsink mating surfaces with IPA 99% + lint-free wipes until completely clean. Apply a thin uniform layer of Arctic MX-6 or Thermal Grizzly Kryonaut - rice-grain quantity spread to full die coverage, not glopped on. Reseat heatsink with even pressure, tighten screws in a star pattern to finger-tight-plus-quarter-turn. Do NOT over-torque - bending the PCB cracks BGA joints. Cold-boot and observe steady-state. Expected improvement: 5-15 C drop on a miner where the TIM had dried.

Rebuild OC back up slowly, in small steps. If stock is stable at Step 4: add +10 MHz at a time, observe 10 minutes of stability per step, stop at the step BEFORE AxeOS reports a temperature above 70 C at steady state. That is your chip's silicon-lottery + thermal-environment ceiling. Tune Vcore independently: add +0.01 V per step, same stability window. Log your final profile - you will want it if firmware updates reset NVS.

Replace the fan with a known-good unit. If RPM is below nameplate or if the fan is older than 2 years of continuous operation, replace it preemptively. Noctua NF-A4x20 PWM (Ultra/Supra/Gamma) or NF-A12x25 PWM (Hex/GT) are quieter + higher-quality than most stock fans, 5-7 years MTBF, same 4-pin PWM connector. D-Central stocks Bitaxe-compatible fans. Plug, boot, verify RPM reads in AxeOS.

Upgrade the heatsink to a D-Central Bitaxe heatsink. Stock Bitaxe heatsinks are adequate for stock thermal loads in moderate ambient. For aggressive OC or warm ambients, D-Central's purpose-designed Bitaxe and Bitaxe Hex heatsinks move measurably more heat - we pioneered the Bitaxe heatsink category. Upgrade a stock heatsink, re-paste, cold-boot, recheck. Typical improvement: 5-10 C at the same OC profile.

Check line voltage and PSU under load. A tired PSU that sags under load can cause the regulator to ramp harder, which dissipates more heat at the VREG - and on Gamma/GT the VREG is monitored by the same EMC2101, so a sagging PSU can trip self-protect via VREG temperature. DMM at the plug under load: Ultra 4.8 V or better at 2-3 A, Hex 11.5 V or better at 4-5 A. If PSU sags, swap it.

Read the serial console during boot and the first 30 minutes of hashing. 115200 baud, 8N1 over USB-C. Look for i2c_master_transmit_receive errors (EMC2101 bus issue), OVERHEAT_TRIGGERED, raw temperature register readouts, and the TPS546_init / VCORE_init lines on Gamma/GT. Screenshot or paste. This is the engineering-grade diagnostic that narrows sensor failure vs actual thermal failure in minutes - and it is the first piece of information any D-Central bench triage or bitaxeorg Discord support will ask for.

Scope the EMC2101 I2C bus during a trip. With a cheap 2-channel scope on SDA and SCL, watch transactions during the trip event. Healthy: clean transactions at 100 kHz standard mode with no NACKs, consistent timing. Broken: stretched clocks, missed ACKs, noisy edges. If the I2C bus is damaged (damaged pull-up resistor, ESD-damaged EMC2101 input), the sensor reports garbage and AxeOS trips on phantom readings. Visual inspection of pull-up resistors and the EMC2101 IC for scorching, under magnification.

Bisect firmware to find the regression. If self-protect started tripping right after a firmware update on a miner that was previously stable, bisect: flash the previous release via Web Flasher, observe stability for a full day, flash the current release, compare. Document which specific version introduced the regression and file a GitHub issue on bitaxeorg/ESP-Miner. This is how the auto-fan spurious-trigger (#599) and the TPS546 ID mismatch (#1291) were both tracked down - user bisection.

EMC2101 replacement via hot-air rework. If Step 12 confirmed the EMC2101 is returning invalid readings AND the I2C bus checks out clean, the EMC2101 is damaged. Package is DFN-10 or QFN-10, 3x3 mm. Desolder with hot-air at around 300 C, clean pads with braid + flux, place new EMC2101 (from Digi-Key / Mouser - verify exact part number against your board's silkscreen), reflow down. Retest. Tier 3 if you have hot-air + microscope + steady hands; otherwise Tier 4.

Gamma/GT-specific: TPS546 VREG thermal check. On Gamma and GT, the TPS546D24x VREG is monitored by the EMC2101 via a thermistor on its thermal pad. A TPS546 with degraded thermal coupling to the board (cold solder on the exposed pad) can run hot enough to trip self-protect even when the ASIC is cool. Under magnification, inspect the TPS546's thermal pad for voids or lift. Reflow if damaged. See #1291 context.

Stop DIY when the EMC2101 or TPS546 need board-level rework and you do not have hot-air + preheater + microscope. These are DFN / QFN parts in tight pad spacing. Iron-only rework will lift pads and turn a CAD $45 repair into a CAD $180 board replacement. Ship the board to D-Central - we have the rework stack + the chip inventory for both EMC2101 and TPS546 variants.

Stop DIY when the diagnostic is ambiguous across multiple reboots. If self-protect trips at unpredictable temperatures (trips at 60 C once, 75 C another reboot), the sensor or I2C bus is intermittently damaged. Intermittent faults need scope-level diagnostics over time - not a Tier-3 DIY. Ship it with the serial log from multiple failed boots + your diagnostic notes.

Stop DIY on Hex multi-chip trip patterns. If one of the six BM1368s on a Hex runs 10+ C hotter than its neighbours and trips self-protect while the others are fine, the thermal coupling between that specific chip and the heatsink has failed - usually from a cracked solder joint on the thermal vias. Single-chip reflow in a 6-chip daisy chain with a shared heatsink is bench work. We pioneered the Bitaxe Hex heatsink and have rework history on dozens of them.

Ship to D-Central with the serial log, the AxeOS screenshot of the trip event, your PSU, and context. Pack the board in anti-static bag, include the PSU (we will test your exact stack), attach a note with: board revision (silkscreen), AxeOS version, the specific profile that was running when it tripped (frequency, voltage, fan mode), ambient at intake, serial console log from a failed boot. Canada-wide, US/international welcomed. Expected turnaround 5-10 business days.

Consider repair-vs-replacement economics. An EMC2101 replacement at D-Central runs CAD $45-$90 typical. A TPS546 replacement on Gamma/GT runs CAD $75-$140. A full Bitaxe Ultra replacement board is CAD $130-$180. If the diagnostic shows multi-component damage (EMC2101 + TPS546 + fan + dried TIM all needing attention), a board replacement may be the better economic path - we will quote honestly up front, no lock-in.