Antminer T19 – Fan Speed Error

Kill power at the wall breaker (not the miner's soft button), wait 60 seconds, then power back up. This clears any latched-firmware fan fault caused by a brief tach glitch. If `ERR_FAN_SPEED` does not return within 10 minutes of hashing, monitor the miner and move on. This step resolves a meaningful share of tickets that are not hardware faults at all — just transient tach misreads after a reboot or a firmware update.

Record which fan number failed (`fan1`-`fan4`) from the dashboard Miner Status page. Photograph the status page, then open Miner Logs and capture the most recent `fan_speed_err` or `ERROR_FAN_LOST` line with its timestamp. This single data point determines whether you clean, swap, or open the control board later — do not skip it even if you plan to just replace the fan. Tape-label each fan (front-intake top/bottom, rear-exhaust top/bottom) before you touch anything.

Shop-vac the intake filter, each fan's blades and hub, the heatsink fins between the fans and hashboards, and the rear exhaust grille. A felt of dust on a 12038 bearing stalls it; lint wrapped around a blade hub unbalances the fan and accelerates bearing wear. Use a shop-vac only — no brushes (push dust deeper into the bearings), no compressed air indoors (redistributes dust through the miner). This single step resolves the largest bucket of T19 fan-speed tickets on D-Central's bench.

Verify intake ambient air temperature is at or below 35 °C using an IR thermometer or a cheap digital probe held at the front grille — not room-middle, not the hallway, at the intake. High ambient forces all four fans to 100% duty continuously, and the next bearing is weeks away from failure. Move the miner, open a door, run a circulation fan, or duct outside air in. If ambient is above 35 °C in the room, fix that before you replace any hardware.

Check firmware version at support.bitmain.com/downloads against your T19's hardware revision (visible on the control-board sticker). If you are on a build known for false fan faults on edge-case T19 units (several late-2022 builds are suspect), flash one version back or forward via the dashboard firmware page. Observe 30 minutes. This is a 15-minute diagnostic that rules firmware in or out cheaply — and sets you up to consider DCENT_OS for proper per-fan RPM history going forward.

Power off at the breaker, open the top chassis (Phillips #2 on the top four screws), and locate the failing fan's 3-pin JST connector at the control board. Unplug it, inspect both male and female pins for bent, blackened, or oxidized contacts, clean with IPA on a lint-free wipe if needed, and re-plug firmly until you feel the click. Re-route the cable so chassis vibration does not pull on the connector. A meaningful percentage of intermittent tach faults are bad JST contacts, not bad fans.

Cross-swap the failing fan's cable with a known-good fan's cable at the control board headers (e.g., if `fan1` errors, plug its cable into `fan3`'s header and vice versa). Power up, observe which fan number now errors on the dashboard. If the fault follows the fan itself, the fan is bad — proceed to Step 8. If the fault stays in the original header slot, the control board's fan-drive circuit is suspect — jump to Tier 3 (Step 13) or Tier 4 (Step 16). This is the single highest-value 5-minute diagnostic on this fault class.

Order a matching 12038 replacement fan: 120 × 120 × 38 mm, +12 V DC, 3-pin with tach, rated roughly 6,000 RPM at full duty, high static pressure. Direct replacements: Delta AFB1212SHE (6,000 RPM, high static pressure), Sunon PSD1212PMB1-A (6,800 RPM). Source from D-Central's accessory catalogue or a reputable electronics supplier — avoid no-brand Amazon 120 mm `PC case fans`, which are low static pressure and will fail the T19 tach threshold immediately.

Install the replacement fan. Four Phillips screws per fan, lift the old fan out, unplug the JST, plug the new fan in verifying pin orientation (red/yellow/blue typical — mismatch fails safely but will not spin). Seat the new fan with the airflow arrow pointing the same direction as the other three fans (front fans blow in, rear fans blow out). Reassemble the chassis, tighten evenly, verify no cables are pinched or touching blades.

Power up and burn in 60 minutes. Watch the dashboard for the new fan's RPM stabilizing in the 5,500-6,100 RPM range at 100% duty. Verify the error does not return. If the new fan's RPM is below 4,500 at full duty, you sourced the wrong fan — not enough static pressure to push air through the T19 heatsink stack. Swap for a proper high-static-pressure 12038 and re-burn-in. Log the install date in sharpie on the chassis for the next preventive cycle.

Measure the +12 V rail at the fan header under load. Back-probe the header with a multimeter while the miner is hashing; verify 11.8-12.2 V DC. Below 11.5 V sustained indicates either a control-board rail issue (Tier 3) or upstream APW12 sag (rule out the PSU first with a known-good unit). This step is cheap, fast, and rules out an entire class of `control-board-blamed` faults that are actually PSU-side.

Bench-test the suspected-bad fan at +12 V direct. Hook the fan to a bench DC supply, bypassing the control board. Measure current (0.5-0.7 A healthy) and RPM by tach wire or by timing the blade. If the fan spins healthy on the bench but fails in the miner, the fan itself is fine — the control-board fan-drive circuit for that specific header is faulty. Proceed to Step 13.

Isolate the fan-drive FET on the control board. Remove the control board from the chassis (disconnect hashboard ribbons first, then board-to-chassis screws). Each fan header has a small N-channel FET driving its +12 V line, typically SOT-23 or SOT-89. Probe with a scope or DMM: gate should swing with PWM duty, source at ground, drain carries +12 V under load. If the drain sits stuck low or the FET reads shorted on a diode test, it needs replacement.

Replace the failing fan-drive FET. Hot-air rework station at 310-330 °C on the SOT-23 pad, flux, remove the old FET, clean the pads, place the new FET, reflow. Verify gate-drive signal on the scope if you have one; otherwise reassemble and power up. Any generic +30 V, ~3 A-capable N-channel SOT-23 (e.g., AO3400A) works for bench repair — match the original part number where possible before substituting.

Roll firmware to the last-known-good version for your specific T19 hardware revision. Verify the hardware revision against the control-board silkscreen and sticker before flashing — wrong firmware for a late-rev board can brick the control board permanently. Use SD-card flash rather than the dashboard upgrade path, since SD is the recovery-capable route if the flash goes sideways.

Stop DIY. Book a D-Central repair when: multiple fan headers on the same control board are failing simultaneously; the +12 V at a fan header reads below 11.5 V under load with a fresh APW12; a replacement fan dies within 30 days; the control board shows visible burn marks, cracked MLCCs, or bulging caps near the fan-driver section. At that point you are in test-fixture territory, not bench-rework territory.

D-Central bench process: full control-board diagnostic on the test fixture with scope-trace of the fan-drive PWM and +12 V rails under simulated load, reflow or replace suspect FETs, cap-refresh the fan-driver filter network if indicated, and 24-hour burn-in on a loaded test rig before return. Typical bench turnaround is 5-10 business days for a T19 control-board repair. Salvaged-grade control-board replacement is an alternative if component-level repair is not economical.

Ship safely. Pack the miner (or the separated control board if you are confident) in anti-static bags, double-boxed with at least 5 cm of foam on every side. Include a note with observed symptoms, firmware version, which fan numbers errored, when the fault started, what you have already tried, and your contact. That note saves D-Central diagnostic time, which directly saves you money on the repair bill.