Avalon Series – A3203 ASIC Pin Short Diagnosis

Cut AC at the breaker for 60 seconds, then attempt one cold boot. If the PSU trips OCP again immediately, STOP. Do not keep cycling power. Each OCP trip pumps inrush energy into the fault path and can promote a single-chip verdict into a chip + power-sequence MOSFET + adjacent regulator verdict in three or four power cycles. Document the PSU LED behaviour, fan blip duration, and any audible click in your service log.

Pull cgminer estats output if the chassis briefly enumerates. From any laptop on the same LAN: `curl http://<ip>:4028 -d '{"command":"estats"}'`. Save PS[0..2], ECMM, ECHU[0..2], MWx arrays, and MM_STATUS if printed. PS = 64 (OC_out) or any OC_IOSx bit set tells you the PSU saw a hard short downstream of that channel. Screenshot the response before any further action.

Multimeter cold-resistance probe at each hashboard input. Power off, AC disconnected, all boards still installed. Probe each hashboard's main DC input pad to chassis ground on the lowest resistance range. Hold each probe for 10-15 seconds (bulk caps slowly charge the meter). Healthy = tens to hundreds of ohms. Shorted = single digits or sub-ohm. Record all three readings. This is the single highest-yield diagnostic on this fault.

Visual inspection under bright light. Look for discoloured silkscreen, scorched solder mask, bulged or burst capacitors, missing components, conductive contamination (white crusty residue, dark stains under chips), burnt-component smell. Photograph any visible damage. A burnt board's damage pattern often points directly at the failed chip or the failed domain regulator.

Verify PSU compatibility with the chassis generation. A1056 / A1066 / A1126 Pro-S / A1146 Pro / A1166 Pro / A1246 / A1266 use one Canaan PSU family per Canaan's official product page. A1326+ uses a different family with different rail voltage and connector pinout. Wrong PSU on chassis is a documented cause of pin shorts. Photograph the PSU label and confirm against Canaan's compatibility matrix before re-applying power.

Pull the suspect hashboard. Power off, AC disconnected. Remove side cover. Disconnect the AUC ribbon and main DC harness from the suspect board. Photograph the connector pinout BEFORE disconnection so you can put it back exactly the way it came out. Slide the board out and place on an anti-static work surface.

Bench-test on a current-limited supply. Lab supply set to the board's rated input voltage (12V on A11/A12/A13 generations, higher on A14/A15 — check the chassis label) with the current limit set to 2-3A. NOT the full chassis rating. Connect the board. If the supply folds back into current limit immediately = hard short. If the supply rails up cleanly = failure is intermittent or chassis-side; re-probe in the chassis.

Thermal-camera or IR-thermometer hot-spot localization. With the bench supply folding back to current limit, leave the board powered for 5-10 seconds while sweeping the thermal camera across the chip array. The shorted chip warms visibly while the rest of the board stays cold. Record the chip position (board reference designator if printed, or row-column if not).

Re-probe cold resistance directly at the suspect chip. Identify the chip's supply pins (corner BGA pads or the nearest decoupling capacitor pads on the package's perimeter). Probe pin-to-VSS. Sub-ohm directly at the chip = chip-side short. Tens of ohms at the chip but sub-ohm at the board input = short is elsewhere on the board, walk the board to the next hot spot or suspect a regulator IC.

Stop here and ship if you've confirmed a single chip-side short. A32xx package replacement is not a Tier-2 repair. Continue to Tier-3 only if you have hot-air rework, bottom-side preheat, microscope, and matched-bin replacement chips. Otherwise book a D-Central ASIC Repair slot — your work to this point will save us 60-90 minutes of bench bisection, which directly reduces your repair invoice.

Document the failed chip's exact position and orientation. Photograph from above with a ruler in frame. Mark pin-1 orientation (corner dot or asymmetric pad layout). Record the chip's marking — A3206 / A3210 / A3218 / A3228 / A3229 plus the date code and lot. Replacement chips MUST come from a matched bin per the Zeus Mining A11/A12 repair guide. BIN mismatch on Avalon hashboards causes the firmware to refuse to enumerate the chain — a chip from a different bin will be cosmetically identical but will not boot.

Preheat the board from below to ~150 °C. Use a controlled-temperature preheater. Hot-air-from-above without bottom-side preheat will warp the PCB on an Avalon hashboard and lift adjacent pads. Allow the board to soak at 150 °C for 3-5 minutes before any top-side heat. The Avalon hashboard PCB is multi-layer copper-heavy and acts as a giant heatsink — without preheat, your hot-air gun is fighting the entire board.

Remove the failed chip with hot air at 310-330 °C for ~30 seconds. Watch for the first sign of solder reflow (the chip will visibly settle or shift slightly under gentle vacuum-pen pressure). Lift the chip with the vacuum pen, do not pry. Inspect the freed pads under magnification. If any pads lifted with the chip, the board is Tier-4 only — stop and ship.

Clean the freed pads. Inspect for solder-bridges, conductive residue, lifted vias. Wick excess solder with copper braid + flux. IPA-clean. Re-flux. If the freed pad area looks healthy, place the matched-bin replacement chip with pin-1 alignment confirmed against your photograph. Re-flow at 310-330 °C from above with bottom-side preheat held at 150 °C. Cool naturally. Inspect under magnification — every pad should show a visible solder fillet.

Re-probe cold resistance before re-installing. Pin-to-VSS should now read in the expected tens-to-hundreds-of-ohms range. If still sub-ohm, you either installed the chip mis-oriented, or there's a PCB-side short underneath that the chip-replacement didn't fix. Diagnose before re-install. If healthy, re-install the board, bring up the chassis on a current-limited supply if available, and burn in 24 hours under nameplate load.

Stop DIY and book a D-Central ASIC Repair slot when: two or more chips on the same board are diagnosed as shorted, suspected solder-bridge or PCB-side short under a chip, visible surge damage on the AUC or input filter, lifted pads from a previous repair attempt, cracked PCB, or no matched-bin donor chip available. Multi-chip rework on A32xx silicon is bench-only — yield drops sharply once you're chasing more than one position.

D-Central bench process: programmable load + bench supply with controllable inrush, full PS / MM_STATUS capture, thermal camera with chip-level resolution, chip removal under controlled hot-air with bottom-side preheat, matched-bin chip placement from a graded donor inventory, trace repair (jumper-wire or copper foil) where pads have lifted, full chassis re-burn-in under nameplate load for 24 hours before release. On older generations where board repair is uneconomic, we quote salvaged-grade replacement boards.

Ship the chassis with the PSU and your service log. We test your exact stack — the PSU is a contributing diagnostic, not just a power source. Include screenshots of the failing PS[0..2] / ECMM / MM_STATUS, your cold-resistance readings, photos of any visible damage, full service history (when, by whom, what was swapped), and contact info. Pack each hashboard in an anti-static bag, double-box with ≥5 cm foam on every side. Saves diagnostic time and saves you money.