APW7 Low Line Voltage Standby Cycling Under 205V

Unplug the APW7 from the wall and leave it disconnected for a full 10 minutes. This is the Bitmain-documented procedure for clearing latched protection mode after an OCP, OVP, or undervoltage trip. A quick 30-second cycle is NOT enough — the bulk capacitors and the latching protection IC need time to fully discharge. Don't power-cycle at the breaker only; pull the C19 cord from the PSU itself.

While the PSU is resting, plug a Kill-a-Watt or any clamp meter into the same outlet under load. If you don't own a meter, plug in a 1500W space heater on max and watch for outlet voltage on a multimeter. Read voltage with the load active. US 110V/120V outlets routinely sag to 105-115V under household load — that's nowhere near the APW7's 205V minimum. 240V split-phase outlets should hold 235-245V; if yours sits below 220V under load, something else on the circuit is competing.

Inspect the C19/C20 PSU power cord for melted pins, brown discoloration, or arcing damage on the receptacle. APW7s pull 8-10A at 220V and 16-18A at 110V; cheap or undersized power cords introduce voltage drop right at the inlet. Swap to a 14 AWG (or thicker) C19 cord rated for the full sustained current. This is the single cheapest and most-overlooked fix.

If you're on a 110V/120V circuit and you see cycling, the answer is not to fix the PSU — it's to feed the PSU a 240V leg. North American homes have 240V available at the panel; an electrician can install a NEMA 6-15, 6-20, or L6-30 receptacle on a dedicated 20A or 30A double-pole breaker for $200-400 CAD. APW7 was designed to deliver full 1800W only on the 220-240V leg. On 110V it's electrically capped near 1000-1200W and electrically much closer to its undervoltage threshold.

Move the APW7 onto a confirmed 240V dedicated circuit. Re-test under full miner load for 30 minutes. If cycling stops cleanly: the PSU and the miner are healthy — your original outlet was the problem. Re-test the original outlet with a voltage logger over 24 hours before condemning it; sometimes only evening peak hours sag.

If cycling persists on a confirmed 240V circuit reading 235-245V under load, you've isolated the fault to the PSU itself. Power off, unplug the miner-side DC connectors from the PSU (all 10 of them on an APW7), and re-test with the PSU running into open circuit. Some APW7 protection circuits latch on a high-side load fault — a single shorted miner-side cable is enough to keep the PSU in standby. Inspect each PCIe 6-pin for burnt pins or a stuck retention clip.

With the PSU at no-load on a 240V circuit, multimeter the DC output at the connector. Healthy APW7: 12.10-12.50V steady. If it reads 0V or pulses (jumps to 12V then drops to ~4V repeatedly), the auxiliary 12V housekeeping supply is failing — the fan can run from primary-side aux but the main rail can't sustain. This is bench-repair territory; see Tier 3.

Re-attach miner cables one hashboard at a time. Start with one hashboard and the control board only. If the PSU holds 12V under that partial load but cycles when the second/third hashboard is added, you don't have a PSU brownout — you have an OCP trip from a hashboard pulling more than its share. Investigate that specific hashboard for shorted MOSFETs or a bad PMIC before blaming the PSU.

Use a voltage datalogger (Kill-a-Watt with logging, Hioki PW3360, or even a $40 Aliexpress data-logging meter) on the input outlet for a full 24-48 hour cycle. Look for the dip pattern: 4-9 PM neighbourhood peak load, mornings during ice-storm grid stress, late-night sag during commercial off-peak transitions. If the log shows any minute below 205V, you have your answer — the APW7 is doing its job and refusing to deliver 12V into a sagging grid. Move it to a stiffer circuit.

Open the APW7 case (FOUR Phillips screws on the chassis, then the top sheet metal lifts off). HIGH VOLTAGE WARNING: bulk capacitors hold 375-385V DC for several minutes after unplug. Discharge with a 5-10kΩ resistor across the cap terminals BEFORE touching anything. Visually inspect: any bulging electrolytic caps near the PFC stage, any scorch marks around resistor R33, any blackened windings on the small auxiliary transformer T1. Photo-document everything before further work.

Multimeter the PFC bulk cap with the PSU plugged in and running into open circuit. Expected: 375-385V DC across the cap. Below 350V = PFC stage is failing or input is sagging despite outlet reading nominal — re-check input under load. Above 400V = PFC overshoot, immediate shutdown. If the PFC bus is healthy but the 12V output stays low, the fault is downstream — auxiliary supply (U5/T1/Q5/D11), the main DC-DC converter, or the secondary-side rectification.

On bench, isolate the auxiliary 12V housekeeping supply: with PSU unplugged from wall and bulk caps discharged, in-circuit-test U5 (PWM controller IC), Q5 (aux MOSFET), D11 (aux rectifier diode), and T1 (aux transformer) for shorts and opens. A failed Q5 or D11 produces exactly the 'fan runs, no main 12V' symptom the BitcoinTalk community thread documents. Replacement parts are $5-15 CAD each from Mouser/Digikey for stock through-hole equivalents. This is 30-45 minutes of solder-station work.

Test under load: on a stable 240V bench feed, load the APW7 to ~80% of rated output (1440W) for 2 hours minimum. Use a DC electronic load if you have one, or two known-good Antminer S9 hashboards as a dummy load. Watch DC output stay within 12.10-12.50V band, no cycling, no thermal foldback. This is the standard post-repair burn-in procedure. If it passes, deploy. If it fails, you missed a component.

Decision point: replace vs repair. If the APW7 needed primary-side work (PFC IC, main switching MOSFETs, or magnetics), you're at $80-120 CAD in parts plus 2-3 hours bench time. A new APW9 or used APW12 is often the better economic call — APW9 is rated for higher-power miners anyway and has a slightly more forgiving low-voltage characteristic. Stop pouring labour into an APW7 if a single PSU replacement closes the ticket.

Replacement option A — stay on APW7: source a known-good APW7 from a trusted reseller (D-Central stocks tested PSUs). $120-200 CAD used, $250-320 CAD new-old-stock. Same C19 cord, same DC connector layout, drop-in for any S9/L3+ era miner. The 205V minimum still applies — if your installation is on 110V you'll be back here in 90 days.

Replacement option B — upgrade to APW9 or APW12: APW9 (3000W, 220V only) and APW12 (5500W, 220V only) have stiffer regulation under sag and are designed for the higher-pull S17/S19/S21-class miners. Adapter cables are needed for the older 6-pin ecosystem (S9/L3+). Cost: APW9 $250-400 CAD, APW12 $350-600 CAD depending on condition. Both are FORCE 240V — do not attempt 110V operation, they will cycle harder than the APW7.

Replacement option C — Mean Well RSP/SE/HSP series for home-mining flexibility: A Mean Well HSP-2400-12 or RSP-3000-12 on a 240V leg gives you industrial-grade regulation and a wider input voltage tolerance (180-264V on most series). Pair with breakout boards from D-Central or open-source designs. This is the Mining Hacker's pick for long-term home reliability — not Bitmain-branded, but engineered for continuous duty in less-than-perfect grid conditions.

Pro repair option: ship the APW7 to D-Central's bench. We test on programmable AC source, rebuild auxiliary supply chain (U5/T1/Q5/D11) with verified parts, replace PFC magnetics if needed, run 2-hour 80% burn-in, and return the unit fully tested. $85-180 CAD typical, 5-10 business day turnaround. Worth it for the rare APW7 (specific S9 vintage, custom configurations) where a drop-in replacement isn't available.

Pre-ship checklist: drain bulk caps with a bleeder resistor before packing, remove the C19 cord and all DC output cables, anti-static-bag the PSU, double-box with 5cm of foam minimum on every side, include a note describing the symptom (cycling vs dead vs reduced output), the input voltage you observed, and your contact info. Saves diagnostic time and your repair cost.