Hashboard Fault Localizer: Find the Broken Chip, Domain or Regulator

A hash board that reports fewer chips than it should is not guessing material — it is telling you, to the chip, where the fault sits. This tool encodes the D-Central bench procedure: pick your Antminer model, describe the symptom the miner log or test fixture reports, and it computes the break location, ranks the likely faults in the order a technician would probe them, and names the next test point with the expected values for your exact board.

Why chip count and domain voltage decide everything

Every Antminer hash board is a daisy chain: each ASIC forwards clock, data and reset to the next, and the chips are powered in series-stacked voltage domains — the rail is regulated per domain, never per chip. Those two facts produce the triage logic this tool runs. A dead chip stops forwarding its signals, so every chip downstream goes invisible: the chain always breaks at the chip immediately after the last one detected. A whole chain at zero with healthy domain rails means power is fine and nothing answers — a clock, level-shifter or supervisor problem, not thirty dead ASICs. No domain voltages at all points upstream, at the boost or buck stage. And an EEPROM or CRC complaint is a support-circuit fault; the compute silicon is usually innocent.

From differential to fix

The ranked list is a probing order, not a verdict — confirm before you desolder. The classic confirmation for a chain break is the binary-search (dichotomy) method: inject a response mid-chain, re-enumerate, and watch whether the count grows; three or four iterations corner the fault to a single chip. Component-level values live in the companion references: per-model chip counts, domains and rail voltages in the hashboard architecture reference, factory diode-mode pass/fail resistances for S17/S19 boards in the diode & voltage reference, failure modes per component class in repair parts & common failures, and bench equipment in the test-fixture guide. If the log itself is your starting point, run it through the kernel-log decoder first and come back with the chip count it reports.

Some diagnoses end at a soldering iron you may not own. Chain-break chip swaps and BGA reballs need hot-air rework and (for newer boards) a test fixture; boost-stage rebuilds need schematic-level tracing. That work is what the D-Central repair bench in Laval does daily, at flat-rate pricing with diagnosis included — start a repair and ship us the board instead of gambling on parts.