Whatsminer M50+ – Hashboard Short Circuit Detected

AC power-cycle 60 seconds at the breaker, NOT the chassis switch. The MicroBT P21 PSU needs primary-side bulk caps to discharge fully for the OCP latch to clear cleanly. A short chassis-button reboot leaves the P21's MCU fault latch set on some firmware revisions and the error re-appears instantly. Wait the full minute. Power up. If the fault re-latches within 5 seconds of boot you have a real downstream short and need to escalate. If the miner stays up and hashes for hours you had a transient (brownout, surge, marginal contact) - log the event and monitor. Do not normalize repeated OCP retries.

Stop retrying after three failed boots. Each repeated OCP trip under a real short condition stresses the P21's output MOSFETs (the protection circuitry is not infinite-cycle-rated) and pumps another current spike into the failed component on the bad hashboard. Cap failures cascade - the dead cap was carrying neighbours that are now over-stressed. Power down, leave it down, and escalate to Tier 2 isolation work with the chassis powered off. Iterating by destroying gear is not the Mining Hacker workflow.

Read the BTMiner API from a laptop on the same subnet, not the Web UI. curl http://<miner-ip>:4028 -d '{"cmd":"status"}' and parse the error_codes array. 233 is the umbrella code; 234 indicates sustained output overcurrent (most aligned with hashboard short); 236 indicates a fast OCP trip (also short-aligned); 237 indicates slow OCP. The Web UI collapses these into 'Code: 233'; the API exposes the sub-code which steers diagnosis. Note any per-board status lines that survived to the API. If the API itself is unreachable (control board down with the rest of the miner), proceed assuming 233 umbrella.

Inventory recent events. Tune push in the last week? PSU swap? Lid-off service? Thunderstorm? Brownout? Miner relocation? Note each one - they steer the diagnosis. A short that began the day after a tune push from 3.4 kW to 3.7 kW on an M50+ that lived happily for two years is almost certainly a marginal output capacitor that finally gave up under the higher rail current. A short the morning after a thunderstorm points at surge damage to multiple boards simultaneously. A short after a PSU swap points at harness mis-seat or wrong-model PSU. Service log shortens diagnosis.

Kill AC at the breaker, wait 60 seconds, open the control-box cover. ESD strap on. Take photographs of the cabling layout BEFORE disconnecting anything - restoration photos save 20 minutes per board on reassembly. Inspect the 12V copper busbar for blackening, blueing, or melted ring terminals. Any visible damage on the busbar itself = the short already cooked something on the path; tag for Tier 4 inspection regardless of where else you find a fault. Smell test for burnt-electronics odor near each circuit area to localize the damage zone before any further disassembly.

Disconnect all three hashboards from the 12V busbar. Both ring terminals (M6 socket, MicroBT spec is 2.5-3.5 N-m per the self-service guide - note positions before loosening) and the data ribbons. Re-apply AC. Healthy PSU boots clean with zero hashboard load: fans spin, no OCP latch, no fault LEDs. If the PSU still latches OCP with all hashboards disconnected, the short is PSU-side or harness-side - skip ahead to Step 16 (Tier 4). If the PSU comes up clean with no load, the PSU is innocent and you've narrowed the suspect set to three boards. Proceed to Step 7.

Reconnect ONE hashboard at a time. AC off between each connection. Power up. The board that triggers OCP on connection is the bad one. Mark it SM0 / SM1 / SM2 corresponding to its slot. All three boards individually clean = the chassis short was elsewhere (mounting-screw bridge, debris on busbar, harness pinch); reinstall everything carefully and test. Two or three boards triggering individually = ship the chassis to D-Central; cascading multi-board failure is bench territory and almost always indicates a control-board issue or a thunderstorm-event cascade requiring full-chain inspection.

Pull the bad board, set on an ESD mat under bright light. Visual inspection at arm's length first: scorched components, swollen caps, discoloured PCB. Smell test localizes the damage zone. Then 10x magnification: swollen capacitor tops, hairline cracks in MLCCs, lifted pads on PMIC packages, debris bridging traces, blackened solder joints. Photograph any damage before further work - bench documentation pays back later if you escalate to D-Central. Look at the buck-converter output regions specifically - that's where caps cluster and where the highest failure incidence sits.

Multimeter check across 12V to ground at the board's input on the bench. Diode mode or low-ohm. Healthy board reads tens of ohms initially, climbing as the input caps charge. Shorted board reads <5 ohms stable - confirmed electrical short. If the bench reading is healthy but the board triggers OCP in-chassis, suspect a chassis-side short: a mounting screw bridging traces under the board, a thermal pad with conductive contaminant, a piece of conductive debris on the busbar. Reinstall clean (alcohol wipe, fresh thermal interface, verify no debris) and re-test. Surprising number of 'shorted boards' are actually chassis-side faults the bench doesn't reproduce.

Inspect the 12V output harness end-to-end. With the bad board pulled, walk the harness from the PSU to the busbar to the now-empty board slot under bright light. Look for jacket discolouration, heat marks, kinks, pinch points, exposed copper, greenish corrosion at the ring terminals. Squeeze progressively along the length - soft spots indicate internal conductor damage. A genuine MicroBT replacement harness is CAD $30-60; cheaper than the time chasing a phantom board fault that's actually a damaged cable. Reseat the harness with proper torque (2.5-3.5 N-m on M6 lugs) and replace it if any damage is visible.

Bench-isolate the failed component on the bad board with a current-limited supply. Bench DC supply set to 12V with current limit 2 A (NOT full board current - you're locating the short, not destroying it further). Apply to the board's 12V input on the bench. The shorted component sinks current and heats fastest. Use a thermal camera (FLIR ONE or Seek) or a careful tactile within 30 seconds to locate the hot-spot. Most common: a single output capacitor across a buck converter output. Second most common: a single PMIC package. Third: a single chip's power input. The current limit is critical - without it you destroy more components in your search than you find.

Capacitor replacement (the ~60% root-cause case). If the hot-spot is a single output capacitor, hot-air rework station with bottom-side preheat. Preheat the board to 120-150C from below (avoids thermal shock to the rest of the assembly). Hot-air at 350C on the cap, lift gently with tweezers. Clean the pads with flux + braid. Solder a same-package, same-value, same-voltage replacement (verify against silkscreen and the MicroBT self-service guide; common values are 100uF / 16V ceramic and 47uF / 16V polymer for buck-converter outputs but VERIFY for your specific board revision - MicroBT shifted parts across the M50 family). Re-check resistance across the rail before re-installing - a clean swap restores tens of ohms.

Capacitor chain replacement (when one cap failed, neighbours are suspect). Output capacitor failures usually mean the rest of the parallel chain has been running over-stressed compensating for the dying member. Replacing only the visibly-failed cap leaves the next one to fail in 30-90 days. If you're already in there with hot-air rework set up, replace the full output cap chain on that buck converter. Adds 15 minutes and CAD $5-10 in parts; saves a return trip and prevents the cascade-failure pattern that takes out the buck converter PMIC after the second cap failure.

PMIC replacement (if the hot-spot is the buck-converter PMIC, not a cap). This is where Tier 3 starts to push beyond home-bench. PMIC packages are typically QFN or LGA and require preheat-from-below (150C), top-side hot air (310-330C for 30 seconds), and an exact part-number match from the silkscreen. Reflow the new PMIC, wait for natural cool-down (forced cooling cracks the package), re-test resistance across the rail. If you do not have a calibrated preheater, stop here and escalate to Tier 4 - iron-only PMIC rework lifts pads and converts a $40 fix into a $300+ board scrap. Mining Hacker rule: don't reach beyond your tools.

Reassemble and bench-load test BEFORE returning the board to the chassis. Once the short is cleared, the board reads healthy resistance on the bench. Apply bench 12V at full current limit (still bench, not in chassis), watch for 30 seconds, no thermal hot-spots, no resistance collapse. Only then return the board to the chassis, reconnect (torque to 2.5-3.5 N-m on the M6 lugs), power up, and watch the BTMiner API for 15 minutes - clean boot, all chips enumerated, hashing at expected rate confirms the repair held. Going straight from 'short cleared' to 'in-chassis power-on' skips the verification step that catches incomplete repairs.

Stop DIY and ship to D-Central when ANY of: PMIC short identified but you don't have preheat + hot-air rework capability; multiple components on the same board are damaged (cascade); two or three boards failing the same way (not coincidence); chip-level failure isolated to a specific position; visible burn marks on the PSU board; OCP latches with all hashboards disconnected (PSU board fault); swollen caps / blued copper / cracked packages anywhere - all signal cascade damage where bench-level full-chain inspection is the only reliable repair path. [Book a D-Central ASIC Repair slot.](https://d-central.tech/services/asic-repair/) The Tier 4 boundary is well-defined; cross it without ego.

What D-Central does at the bench: programmable DC load tests the PSU isolated against a synthetic duty cycle (catches marginal output MOSFETs that pass static tests but fail under real load profiles). A hashboard test fixture applies 12V at full current and instruments per-domain current draw, isolating shorts to the failed converter or chip in one pass instead of trial-and-error. SMD rework station with bottom-side preheat for PMIC and chip-level work. Salvaged-grade or new-old-stock SHA3-class chips for chip-level replacement. Full burn-in at nameplate for 24 hours post-repair before ship-back, so you receive a board that hashes in chassis the same day you reinstall it.

Ship safely with full context. Anti-static bag for each hashboard, double-box with >=5 cm foam on every side. Include: chassis serial, PSU serial, your last BTMiner API status JSON, any WhatsMinerTool PSU telemetry log captured, tune profile in use at time of failure, recent service history (PSU swaps, tune changes, relocations), intake-ambient log if available. Match chassis serial to PSU serial in your ship note - both are on metal plates, do not guess. Ship the chassis AND the PSU together so we can test your exact stack; cross-stack substitution misses marginal interactions and adds a return trip. Canada-wide shipping; US / international welcomed. Turnaround 5-10 business days for an isolated single-board short.