Antminer S19 XP – PSU Overheating

Hard power-cycle at the breaker for 30 seconds (not a soft reboot). Clears latched OTP state, gives bulk caps a real discharge moment, resets firmware supervisor flags. Re-power and watch for 30 min — the real test is a 4-hour soak without `ERROR_POWER_LOST` recurring.

Verify intake ambient at or below `30 C` with an IR thermometer or USB thermo-hygrometer taped to the intake grille. Not room-middle. Not the hallway. Intake-specific. The S19 XP's thermal margin is tighter than the S19 Pro's — above `30 C` intake, the APW12 will trip OTP on any aggressive profile and many stock ones.

Vacuum the PSU intake grille with a shop-vac to pull off the dust mat, then compressed air (canned duster or 30 PSI shop compressor) blown *out* through the exhaust side to clear anything that moved into the fin stack. PSU must be off and unplugged. Do the same on the miner chassis intakes — they feed the PSU too.

Pull the miner 30 cm off walls on the intake side and 60 cm on the exhaust. XP exhaust recirculation — its own hot air finding the intake — is a silent ambient killer. Re-orient the chassis if airflow is obstructed; add a 20-inch box fan blowing across the intake if you can't fix the path.

Revert to stock Bitmain firmware + stock profile (no OC, no autotune, target `3010 W` at the wall). Watch for 4 hours. If `ERROR_POWER_LOST` stops, you were past the APW12's thermal design point on OC/autotune. Decision: live with stock, upgrade the cooling environment, or ship the PSU for a thermal refurbish before OC'ing again.

Measure the PSU output rail under full load. Multimeter on DC, probe at the PSU output lugs to the hashboards (not the control-board feed). Healthy APW12-`15` reads `14.2-14.8 V` under full load, `14.8-15.0 V` at idle. Below `13.6 V` sustained confirms the PSU is sagging. Record the number for the repair ticket.

Measure mains line voltage at the PSU input with a multimeter. 240 V split-phase: `235-245 V`. 208 V commercial: `202-212 V`. 120 V residential: `115-125 V`. Measure under load, during a bad event if you can catch one. Low line voltage is the most common *cause* of apparent PSU overheating — fix the circuit and you fix the PSU.

Reseat every power connector in the chain. Power off at the breaker, wait 60 s for bulk caps to discharge. Unseat and reseat the PSU input cable, both output lugs to the miner, the PSU data/control cable, and all hashboard 12-pin connectors. Inspect every contact for oxidation, black staining, or bent pins. Clean corroded contacts with a pencil eraser or DeoxIT D5, reconnect firmly.

Move the miner to a dedicated 240 V / 20 A circuit. If you're on shared 120 V / 15 A residential, the XP is already at 22 A duty-cycle-limited and line voltage sags whenever anything significant cycles in the building. A dedicated 240 V circuit halves the current for the same power, dramatically reduces PSU thermal stress, eliminates voltage-sag trips. Single biggest environmental upgrade for a home XP — expect $300-700 CAD for the electrician.

Substitute with a known-good PSU. Swap in a second APW12-`15` (or a compatibility-listed Delta DPS-3600 or Great Wall equivalent). Run for 4 hours under full load. If the substitute runs cool and stable, the original PSU is degraded — keep going to Tier 3 or replace. If the substitute also trips, the problem is environmental (revisit Steps 7 + 9) or the miner is drawing above spec.

Replace the APW12 internal fan. If Step 2 identified bearing whine or RPM that never ramps up, the fan is the fix. Remove 8 Torx T10 screws on the top shell, note the 2- or 3-pin fan connector (3rd pin is RPM tach — don't lose it), replace with a `12010` (120×120×10 mm) dual-ball-bearing `12 V` axial fan rated `1.5 A+` for airflow margin. Zip-tie the wiring clear of the fin stack, close the shell, torque fasteners evenly.

Re-paste the PSU primary-side heatsinks. DANGER: discharge bulk caps first — `400+ V DC` lives on those caps for 30+ s after unplug. Short each cap through a `1 kΩ / 5 W` resistor for 10 s, verify with DMM before touching. Disassemble PSU, clean old paste with 99% isopropyl, apply Arctic MX-6 or Kryonaut in a thin uniform layer on the primary MOSFETs and PFC diodes against the heatsink, reassemble with fasteners torqued evenly.

Inspect and replace aging bulk electrolytic capacitors. Visual: any cap with bulged top, vented rubber plug, or brown residue needs replacement. Full audit: ESR-meter every electrolytic, replace any >30% above stock ESR. Stock primary-side caps on the APW12 are typically `330 µF / 450 V` (verify against your PCB — photograph before ordering). Use `105 C`-rated replacements minimum, `135 C` if available. This is the single most impactful repair for a thermally-degraded PSU.

Flash DCENT_OS (D-Central's own open-source Antminer firmware — the Mining Hackers' option, preferred) for per-rail telemetry — PSU input voltage, rail voltage, per-hashboard current metrics that stock Bitmain firmware hides. Alternatives: Braiins OS+, LuxOS, or Vnish. Run for 48 hours, graph the rail voltage — the sag pattern proves whether the PSU is the culprit or whether the mains circuit upstream is.

Undervolt and underclock the miner as a thermal buffer. Using DCENT_OS (or Braiins OS+ / LuxOS / Vnish), dial the miner back to `2700-2850 W` wall draw while keeping most of the hashrate. The PSU runs noticeably cooler at `2800 W` than at `3300 W` — often enough margin to eliminate thermal trips without new hardware. Mining Hacker tax: 10% less revenue, 3-5x the PSU lifespan.

Stop DIY and book a D-Central repair slot when: (a) rail sags below `13.6 V` with a known-good substitute PSU also failing, (b) visible cap bulging / discoloration / vented rubber / burnt-component smell, or (c) line voltage is in spec and the environment is clean but the PSU still overheats. You're in test-fixture territory. Book: https://d-central.tech/asic-repair/

What D-Central does at the bench: APW12 test fixture with programmable electronic load up to `300 A` at `12-15 V`, thermal chamber for OTP trip verification at `25 / 40 / 55 C` ambient, full capacitor audit (ESR + capacitance + leakage), primary-side semiconductor check (MOSFETs, PFC diode, rectifier bridge), thermal re-paste, fan replacement, 24-hour full-load burn-in at nameplate before ship-back. Miner itself gets thermal paste refresh, fan audit, and firmware at customer preference.

Ship safely. Wrap the PSU separately from the miner. Anti-static bag, double-boxed, `5 cm+` of foam on every side. Include a note with: observed symptoms, firmware version, error codes from `kern.log`, and your measured rail/mains/ambient values from Tier 2. A five-minute note saves us 45 minutes of re-diagnostic, which saves you repair cost.