Testing an ASIC power supply with a multimeter answers one question fast: is the PSU dead, or is the fault downstream on a hashboard or control board? The measurement is simple — set the meter to DC volts, probe the main output between a positive lug and ground, and read the number. The catch is that a mining PSU is not a computer PSU: it does not run on 12V ATX wires, and on most modern Antminer supplies the main output stays at 0V until you tell it to turn on. Read the whole procedure before condemning a PSU on one reading — both a flat 0V and a surprisingly high voltage are often normal on the bench.
Know what voltage the PSU is supposed to output
Bitmain APW-family supplies output very different rails depending on which miner they feed. Probe with the wrong expectation and a healthy PSU looks broken. Match your unit first:
- APW3 / APW3++ / APW7 (S9, T9, L3+, D3): roughly 11.6–13.0V DC. These are always-on — apply AC and the output is live immediately, with no enable pin to assert.
- APW9 / APW9+ (S17, S17 Pro/+, T17 family): adjustable 14.5–21V DC. With no control board it comes up at its default maximum (~21V); firmware lowers it to the hashboard target at boot.
- APW12 (S19, S19 Pro, S19j, S19j Pro, T19): adjustable 12–15V DC; default max ~15.2V with no I2C.
- APW17 (S21, S21 Pro, S21 XP): adjustable 12–15V DC.
On the adjustable supplies (APW9 and newer), the running voltage is commanded by the control board over I2C while hashing. Off the miner, the PSU falls back to its default output at the TOP of its range (~15.2V on an APW12, ~21.3V on an APW9); firmware then lowers that rail to the hashboard target at boot. It does not climb up under load, so a high open-circuit reading is expected, not an over-volt fault.
The 0V trap: enable-gated outputs
The APW9, APW12 and APW17 generation gate their main output behind an enable signal. The 4-pin communication connector carries I2C (SDA/SCL), ground, and an EN pin that is active-low — the miner pulls it to ground to switch the main rail on. With nothing plugged in, the big output lugs sit at 0V by design, so a 0V reading on a bench-isolated APW12 tells you nothing until you assert enable.
To wake the output on the bench, jumper the EN pin to GND at the signal connector (a short wire between the two correct pins). It snaps up to its default — the top of the range, because no I2C master is present to command anything lower. A ~15V reading on an APW12 or ~21V on an APW9 is the maximum-safe default, exactly what firmware would step down — not a fault. If you lack a verified pinout for your exact revision, do not guess: a mis-identified jumper can short the enable logic. The always-on APW3/APW7 have no gate and read voltage the instant AC is applied.
Tools you need
- A digital multimeter rated for at least 600V CAT III.
- The correct AC feed for the PSU (see the 120V note below).
- A short jumper wire for the enable pin on gated PSUs.
- The pinout reference for your specific PSU model and revision.
Safety: where the danger actually is
Never open the PSU shell. Inside the sealed case the mains input feeds PFC bulk capacitors that hold roughly 410–420V DC and stay charged after AC is removed — seconds to a minute with a healthy bleeder resistor, longer if it has failed. That is the shock hazard, and it lives behind the case; nothing a multimeter test requires you to open.
The low-voltage DC output is not a shock hazard. A 12–21V rail will not hurt you through the skin. What it will do is deliver enormous current — an APW12 pushes up to 233A. A dropped tool or a metal watchband across the output lugs will arc, weld, and vaporize metal instantly. Remove jewelry, keep one hand off the chassis, and never let a probe slip between the + and – terminals.
Step-by-step measurement
- Isolate the PSU. Disconnect it from the miner so you test the supply alone, not a hashboard fault feeding back into your reading.
- Set the meter to DC volts (V∡). Auto-ranging is fine; on a manual meter use the 200V scale — the no-I2C default is the top of the range (~21V on an APW9), so give yourself headroom.
- Connect AC power to the PSU input. On dual-input supplies (APW9/12/17 have two independent AC inlets), feed both inlets from the same source — running one leg only leaves the supply under-powered and unable to reach full output.
- Assert enable if the PSU is gated: jumper EN to GND at the signal connector. Skip this for always-on APW3/APW7.
- Probe the main output. Black probe on a ground lug, red on a positive lug. Bitmain marks polarity — bolted copper bus-bars on S17/S19/S21-generation supplies, or per-board connectors on an S9’s APW3/APW7. Read the voltage.
- Check the auxiliary rail if present. APW12/APW17 provide a separate fixed 12V / 15A output for the control board and fans. A dead aux rail with a healthy main rail (or vice-versa) narrows the fault to one section.
Reading the result
- Voltage at the top of the range (~15V on an APW12, ~21V on an APW9, EN grounded, no I2C): this is the default maximum, and it is NORMAL — not an over-volt fault. It proves the supply comes up and holds its rail open-circuit; firmware would step it down to the hashboard target on a real boot.
- 0V after enabling: now the reading is meaningful. A gated PSU still showing 0V with EN grounded and both AC legs live has an internal fault — failed PFC, blown output stage, or a tripped protection latch.
- Voltage present but sags hard under load: the failure a bench test hides. Reconnect the miner (or a load) and watch the rail. A PSU that reads clean open but collapses the moment a hashboard draws current is failing under load — a classic cause of a miner that boots, hashes for seconds, then drops boards or reboots.
A voltage test is a screen, not a verdict. The cleanest confirmation is a swap test: fit a known-good, correctly-rated PSU to the suspect miner. If it runs, the original supply was the fault; if it still fails, the problem was never the power supply — look downstream.
120V vs 240V — a common false failure
Input rating varies across the family, and it matters. The APW3/APW3++ and APW7 are wide-input supplies (100–264V AC) — they natively run on 110/120V, just at reduced output (an APW7 makes ~1800W at 220V but only ~1000W at 110V). The APW8 and APW9 are rated 200–240V yet still run on 110/120V with derated output. The APW12 and the S21-generation APW17 require 200–240V. Either way, a supply on a 120V bench can read correct voltage open-circuit yet fail to sustain a big miner: a single 15A/120V circuit cannot supply a 3000W+ ASIC’s full draw. If your test looks fine but the miner starves under load, suspect the supply circuit, not the PSU, before replacing anything.
Related
Walk the whole symptom tree with the ASIC fault finder, source a replacement supply or output connector from ASIC repair parts, and if the unit needs bench time beyond a multimeter, start a repair and we will diagnose it properly.
