ASIC Miner – PSU Buzzing Noise

Power-cycle the miner at the wall breaker for 60 seconds. Not a soft reboot — the full kill. Let the PSU bulk caps bleed down, then power back up. Transient buzzes from brown-outs or momentary overcurrent often don't recur. If the buzz returns within 5 minutes of full-load mining, it's a real fault and you're into Tier 2+. If quiet, monitor for 24 hours and log the timestamp — this is how you catch trending issues before they become failures.

Shop-vac the PSU intake and exhaust grills. A dust-caked PSU runs `5-10 °C` hotter internally, which directly accelerates capacitor aging and sometimes changes the acoustic signature of the unit. Vacuum — do not blow compressed air, which pushes dust deeper into the chassis. Target the intake-side fan; inspect the exhaust for visible dust clumps. Canadian basement workshops collect a surprising amount of drywall dust and cat hair in a year.

Verify the PSU isn't sitting on a vibrating surface. A PSU on a bare metal shelf or glass table couples mechanical vibration from the fan and transformer into a larger resonator — the shelf itself becomes a speaker for the buzz. Move the PSU onto a rubber mat, foam pad, or felt feet. If the buzz drops noticeably, the PSU is electrically healthy and the mount was the speaker. Cheap trick, surprisingly common root cause.

Record a phone-mic audio sample and run it through a free spectrum analyzer (Spectroid on Android, SpectrumView on iOS). Note the dominant frequency. `~120 Hz` = input electrolytic caps failing. `15-25 kHz` = DC-DC stage mechanical. Irregular wide-band = MOSFET distress (Tier 4). This 2-minute test rules out half the diagnostic tree and gives you concrete data to include in a D-Central repair ticket.

Measure wall voltage under full miner load using a multimeter in AC mode at the outlet. Target within `3 %` of nominal: `232-248 V` on 240 V split-phase, `202-212 V` on 208 V commercial, `116-124 V` on 120 V. Anything outside = upstream problem; fix before blaming the PSU. A starved PSU always buzzes and a fresh one on the same bad circuit will buzz identically. Cross-reference the `ASIC Miner — PDU Overload` and `Antminer — PSU Voltage Out of Range` pages if wall voltage is the fault.

Measure PSU output voltage on DC at the PSU-to-miner terminal while the miner is hashing at full power. Probe at the bolt heads, never into the connector body. Expected: APW3++ = `12.0 V`; APW9 = `12-17 V` (typically ~`14.5 V` under S17-class load); APW12 = `13.8 V` sustained; APW17 = up to `15 V` on S21. Sag > `4 %` below expected indicates input-cap failure. Confirms Step 4's 120 Hz finding.

Swap the suspect PSU with a known-good identical unit. Fastest definitive test. If the buzz disappears with the swap, the original PSU is at fault — proceed to Tier 3 repair or Tier 4 ship. If the buzz transfers to the new unit, the miner is drawing an abnormal load profile that stresses any PSU; look downstream at hashboards for a shorted voltage domain. This step alone resolves 40% of ambiguous buzz tickets at the D-Central bench.

Reseat every power connector in the chain: PSU output, miner input, wall plug, any `C13`/`C19` junction. Buzzing can originate at a loose connector — contact resistance climbs, heat dissipates at the joint, and the PSU compensates by working harder. Power off, unplug, inspect for blackening or oxidation, reseat firmly, listen for the click. A melted or blackened plug is a full replacement, never a reseat.

IR-thermometer the PSU chassis under load. Target top cover `< 55 °C` at `25 °C` ambient. Hot top + buzz = internal electrical issue; hot bottom + buzz = output-side filter cap failure (bottom-heavy caps run the ripple). Photograph the hot-spots — useful evidence for a Tier 4 ship-to-bench decision. Thermal imaging reveals what visual inspection can miss.

Replace input-side bulk electrolytic capacitors. On APW12 this is typically 2× `470 µF` / `400 V` or 2× `680 µF` / `400 V` Nichicon or Rubycon low-ESR `105 °C` minimum (prefer `125 °C`). Discharge the PSU for at least 5 minutes before soldering — 400 V across 680 µF is a nontrivial zap. De-solder with a hot iron + braid, install new caps with correct polarity, reflow cleanly. Test with the cover off before buttoning up. Zeus Mining's APW12 repair guide is the community reference; our bench follows a variant with more thorough post-repair burn-in.

Replace output-side electrolytics. Typically multiple parallel `3300 µF` / `25 V` or `6800 µF` / `25 V` low-ESR caps on the `12 V` / `13.8 V` rail. Same process: discharge, desolder, replace, reflow. Polarity matters — output caps see the worst ripple and will fail loudly (sometimes spectacularly) if installed backwards. Test output ripple on a scope after replacement; clean PSU is `< 100 mV` peak-to-peak at full load.

Re-pot loose ferrite cores with thermal-conductive epoxy (Arctic Alumina Epoxy or equivalent). Apply a small bead at each mount point on the output inductor and transformer cores. Cure per spec (12-24 h). Dampens mechanical vibration without affecting electrical performance. Cosmetic-only fix but it's the difference between a PSU you can sleep near and one you can't. Only do this if Diagnostic Step 5 confirmed mechanical buzz.

Flash DCENT_OS for per-rail voltage telemetry. DCENT_OS is D-Central's own open-source Antminer firmware (recommended on Antminer hardware) and exposes real-time input voltage logging and trend alerts. Alternatives: Braiins OS+, LuxOS, Vnish. Trending input rail voltage across days catches intermittent buzz events and correlates them with wall-side sag, ambient temperature, or neighborhood peak load times. Doesn't fix the PSU, but it's the best diagnostic layer for recurring ticket patterns. Stock Bitmain firmware hides this data.

Replace the PSU fan with a quality `12 V` / `120 mm` high-static-pressure unit (Delta, Sanyo Denki, NMB Mat). Match the connector and RPM rating (5000+ RPM at full load typical). Cheap fans shorten PSU life by running internals hotter, which accelerates cap aging. Budget `$25-40 CAD`. Only do this if Diagnostic Step 1 isolated fan-mechanical noise as the cause, not electrical.

Stop DIY immediately if you see visible scorching, carbon tracking between pins, a fractured MOSFET case, burnt-PCB smell, or a PSU that cycles on-off-on-off in under 5 seconds after power-up. You're in arc / short territory and restarting risks an adjacent-component failure chain. A bench test fixture is mandatory to isolate the fault without turning a $50 cap replacement into a $300 control-board follow-up. Book a D-Central ASIC Repair slot.

D-Central bench process on a shipped PSU: programmable AC input sweep (`90-264 V`), programmable DC load (up to `3500 W` sustained), oscilloscope capture of output ripple and transient response, full thermal imaging under load, in-circuit and lifted ESR measurement on every electrolytic, MOSFET gate-drive signal check, board-level cleaning, re-flow of any suspect joints, replacement of any cap with measured ESR above spec, 24-hour nameplate burn-in. Turnaround: 5-10 business days, Canada-wide, US and international welcomed.

When swap-not-repair is the right call. APW3++ units from pre-2018 are at or past the Arrhenius point for most of their electrolytics. A $50-90 cap rebuild buys another 2-3 years; a $200-300 new APW12 pairing is often better economics for an S17+ generation or newer miner. Our bench will call you with a recommendation before committing to repair — we don't rebuild units that are past the economic threshold. Accept the assessment; don't fight sunk-cost fallacy.

Ship safely to D-Central. Anti-static bag, double-boxed, at least 5 cm of foam on every side. PSUs don't urgently need ESD protection, but they absolutely need padding — the bulk caps and transformer are fragile under impact. Include a note inside the box with: PSU model, serial, observed symptoms, spectrum-analyzer audio capture if you took one, paired miner model, and a contact number. Saves diagnostic time, which saves you money on the invoice.