ASIC Miner – Underpowered PSU Symptoms

Verify PSU continuous wattage exceeds miner wall draw by at least 10 percent. Look up the miner's nameplate wall-draw number in the manufacturer's spec sheet and the PSU's continuous rating (not peak). A 3250 W S19 needs a PSU rated at or above 3575 W continuous at your line voltage. Anything under that headroom runs underpowered from day one, regardless of any other fix. If the numbers do not clear, stop diagnosing and order the correct PSU.

Confirm wall voltage and circuit capacity on paper. Residential 120 V / 15 A: NEC continuous budget 1440 W. 240 V / 20 A: 3840 W. 240 V / 30 A: 5760 W. An S19-class miner needs a dedicated 240 V circuit, period. If it is on 120 V, its PSU is derated 30-40 percent regardless of the sticker and you will never hit nameplate hashrate. Plan the circuit upgrade before anything else.

Check for other loads on the same branch circuit. Space heaters, dryers, kettles, EV chargers, second miners all steal headroom and cause sag under simultaneous load. Unplug everything else, retest hashrate for 30 minutes with the miner alone on the circuit, and compare. If hashrate recovers, the circuit was being shared and a dedicated branch is the fix.

Set the miner back to stock profile. Remove any overclock, undervolt, or autotuning preset and reboot. If hashrate recovers to nameplate, your OC was exceeding PSU headroom; rebuild the OC slower, with a voltmeter on the PSU output rail, stopping at the last step before rail voltage sags below the healthy threshold in Step 6. A tired PSU has less tuning headroom than the factory sticker implies.

Visual and audible inspection. With the miner hashing, carefully place a hand on the PSU case (warm, not hot-to-injury). Listen for buzz, hum, or coil-whine that tracks with load. Look for scorching around AC inlets and DC outlets, discolouration on the C13 connector, swollen case vents. Any of these flag a PSU running past its comfort zone and tell you where to measure next.

Measure PSU DC output under full load with a multimeter on DC, probes at the PSU to hashboard connector while the miner is hashing at 100 percent. Targets: S9 above 12.8 V; S19 / S19 Pro / S19j Pro above 13.8 V; S19 XP / S21 above 14.5 V; Whatsminer M30S/M50S above 12.5 V; Bitaxe Hex 11.5-12.6 V; Bitaxe Gamma 4.8-5.3 V. Sag below threshold = PSU tired or undersized. Record the number.

Measure wall AC under load. Multimeter on AC, probes in the outlet while the miner is fully hashing. Expect 235-245 V on 240 V residential, 202-212 V on 208 V commercial, 115-125 V on 120 V. Sag more than 5 V below nominal under load = branch circuit or house service is the bottleneck, not the PSU. Fix upstream with an electrician before swapping PSUs.

Swap to a known-good, correctly-rated PSU pulled from a confirmed-healthy miner of the same class. Reconnect data and power cables listening for the click. Power on, let stabilise 15 minutes, re-measure voltage and hashrate. If everything clears, the original PSU is the fault. Retire it, rebuild it (Tier 3), or order a replacement; never leave a marginal PSU in service.

Move the miner to a dedicated 240 V / 20-30 A circuit run by a licensed electrician. 240 V doubles capacity, halves current for the same wattage, cuts wiring loss, and runs the PSU in its designed efficiency band. Expect CAD $250-$600 per circuit; pays for itself in recovered hashrate and in not burning down the garage. For any miner >= 1.5 kW wall draw on 120 V, this is the single highest-leverage upgrade on the whole rig.

Re-terminate or replace the kettle / C13 lead. Standard IEC C13/C14 cables are rated 10 A at 250 V; an S19 on 240 V at full draw pulls around 14 A. You need C19/C20 or a purpose-built 15 A+ lead. Check the printed rating, replace if under-rated, warm in use, or visibly discoloured at the connector. Cheap, fast, and has prevented more than one garage fire on the repair queue.

Open the PSU and inspect electrolytic capacitors. Power off, unplug AC, wait at least 5 minutes for bleed-off, then open. Inspect caps on the primary side for bulging tops, leaking electrolyte, or brown discolouration on the PCB beneath. Any of these = failed cap; replace with the same capacitance and voltage (or next step up in voltage) from a reputable supplier (Nichicon / Rubycon / UCC). This restores APW3++ or APW12 ripple to factory spec in most cases. Safety: primary caps hold mains voltage for minutes, discharge before touching.

Reflow solder joints on the PSU output stage and switcher. Three years of thermal cycling cracks joints on high-current paths. Reflow the main MOSFETs, output inductor pins, and bulk output caps with a hot-air station at 300-320 C for 20-30 seconds. Let cool naturally, re-seat. Often recovers PSUs that were fluctuating under load from micro-cracked joints invisible under a magnifier.

Flash DCENT_OS on Antminer (D-Central's own open-source firmware with per-chain voltage, per-chip HW%, tuning, autotuning, stratum v2). Alternatives: Braiins OS+, LuxOS, Vnish. Observe per-chip HW% and per-chain voltage. If sag isolates to one voltage domain, you have a hashboard-side issue masquerading as a PSU issue and should route to antminer-s19-hashboard-voltage-error. This is the single most valuable diagnostic on an Antminer complaint.

Log ripple and transient response with a USB scope (Hantek 6022BE or similar) on the PSU output rail during a hash burst. Healthy APW12: under 200 mV peak-to-peak ripple at full load, transient recovery under 5 ms. Anything worse is a PSU you cannot trust long-term even if the DC average looks fine on a multimeter. Scope data is the evidence a repair bench needs.

Recalibrate or replace the PSU voltage-sense wire on Antminer APW family. The small 3-pin sense lead from PSU to control board carries voltage programming and feedback. Damaged or oxidised sense leads make the PSU deliver the wrong rail voltage (usually low) under load. Replace with a fresh OEM lead; on APW9/APW12 the harness is under CAD $15 and fixes more 'mystery sag' reports than anyone admits.

Stop DIY and ship to D-Central when: the PSU shows visible heat damage (discoloured PCB, burnt varnish, cracked thermal pad); voltage sag persists after cap replacement and reflow; multiple PSUs behave the same way in the same rig (upstream wiring or control-board damage); the miner is tripping breakers or GFCIs under load. These are past the safe boundary of a home bench.

D-Central bench process: programmable DC load for real continuous-load testing (not the fan-spin proxy Bitmain's guide implies), full ESR sweep on every electrolytic, scope-based ripple and transient response characterisation, MOSFET and rectifier diode validation, full cap / magnetics / thermal-pad refresh on APW3++ / APW9 / APW12 / APW17 and Whatsminer / Avalon / Innosilicon integrated PSUs, and 24-hour post-repair burn-in at full load.

Ship the PSU safely. Anti-static bag, double-boxed with 5 cm+ foam on every side. Drain any residual mains voltage (wait 10+ minutes after unplug) before boxing. Include a note with the symptoms observed, the miner model it was paired with, firmware version if relevant, and the wall voltage / circuit rating you ran it on. Context saves diagnostic time, which saves you money.