Whatsminer P21 PSU No Output — Loose Copper Bar Screws

Confirm AC at the wall. Multimeter on AC at the outlet - 115-125V for 120V nominal, 230-250V for 240V nominal. AC out of spec or absent means fix the breaker / outlet / wiring upstream of the PSU. The PSU's AC input LED should be solid green; if it is dark or flickering, the PSU is not seeing clean AC. Many 'dead PSU' reports are actually a tripped breaker or damaged cord - 60 seconds of measurement saves a long evening of mis-diagnosis.

Power-cycle at the AC mains, not the chassis switch. Kill AC at the PDU or wall breaker, wait a full 60 seconds. Bulk caps on the primary side and output caps on the secondary side need that long to discharge cleanly. Some PSU MCU revisions latch a no-output protection flag that only clears on a full AC cycle - a chassis-button reboot leaves the latch set and the rail stays dark. Re-apply AC after 60s and listen for the PSU fan ramp-up.

Inspect the chassis externally before opening. Walk around the miner. Any burning smell? Any visibly damaged cables, melted jackets, scorched paint near the PSU? Any rodent droppings near the intake or output? Any signs of impact from a recent move? If the unit looks externally healthy and there are no burn or smell indicators, you almost certainly have a contact / mechanical fault rather than a real silicon failure - the fault category this page covers.

Check service history. Has the miner been moved, transported, opened, or had its PSU replaced in the last 30-90 days? Any service event is a torque-reset opportunity. Even careful repair techs vary in how they tighten copper-bar bolts; the failure mode shows up weeks later as the joint settles. Note the date of last service in your maintenance log so the next torque-check is scheduled, not reactive.

Verify ambient and intake airflow are reasonable. Intake >35C does not directly cause P21_CONTACT (this is a mechanical fault, not a thermal one), but it accelerates the underlying thermal-cycle wear on the bolts. If the miner sits in a warm closet, fix the airflow before you re-torque - otherwise you will be back here in 6-9 months instead of 18-24. A Bitcoin Space Heater ducting setup turns the miner's heat into useful home heat instead of fighting it.

Probe the 12V output at the busbars with a DMM. Kill AC, set DMM to DC volts, clip the negative probe to the chassis ground, touch positive to each of the four 12V output bars in sequence. AC up. Healthy: all four read 12.0-12.3V continuous. Fault: one or more reads 0V OR oscillates 0V <-> 12V. The oscillation pattern usually means the protection IC is retrying because a downstream load looks like a near-short, OR a busbar contact is making and breaking under load. Either way, the conclusion is the same: continue to Step 7.

Re-torque the four hashboard-side M6 bolts to 3.0 N-m. Kill AC, wait 60s, remove the PSU-side cover (typically 4-6 Phillips screws). Use a calibrated torque screwdriver, NOT a hand-tightened Phillips - eyeball torque on a 12V / 95A joint is a fire hazard. Loosen each bolt 1/8 turn first (resets the metal contact pressure), then re-torque to 3.0 N-m. Spec range is 2.5-3.5 N-m; aim for the middle. Watch for cocked or cross-threaded hardware - replace any bolt that does not seat flat.

Inspect each bolt, washer, ring terminal, and contact surface during the re-torque. While the bolt is loose, look at the lock washer (split-ring should be present, not flattened beyond use), the ring terminal on the output harness (no cracks at the crimp, no greenish corrosion), the bolt threads (clean, not stripped or cross-threaded), and the contact surface on the busbar (no blackening or bluing - if there is, clean with brass brush + 99% IPA before re-torque). Replace any compromised hardware. A CAD $1 lock washer is cheap insurance against a CAD $300 PSU returning to the bench.

Re-assemble carefully, check for pinched cables, AC up under load. Replace the cover, route the output harness without pinching it against the chassis, snug the cover screws. Apply AC. The miner should boot through its normal sequence: PSU fan ramps, control-board LEDs come up, hashboards self-test, hash starts within 2-5 minutes. If the rail is clean and the miner reaches full hash, the fault is fixed. Run for at least 1 hour before declaring victory - some marginal contacts hold under static load and fail under thermal load.

Set a re-torque calendar reminder for 6-12 months out. This is the single highest-yield preventive maintenance action for any P21-class Whatsminer. 5 minutes of work twice a year prevents the next no-output ticket. D-Central's repair queue data shows miners on a scheduled torque-check cadence see ~10x lower P21_CONTACT rate than miners that never get re-torqued. The cadence applies equally to P221 / P222 units on M50 / M60 / M66 chassis - same root cause, same fix.

Replace the 12V output harness if a ring terminal is damaged. If Step 8's inspection shows a cracked or corroded ring terminal, do not band-aid it - replace the harness as a unit. Genuine MicroBT replacement harnesses run CAD $30-60 and ship from WhatsMiner support directly or via D-Central's parts inventory. Compare the new harness gauge and ring terminal size before installation; a mismatched replacement creates exactly the same problem you're trying to fix. Re-torque to 3.0 N-m on installation.

Clean a heat-damaged contact surface properly. A blackened busbar surface from a previous overheated joint cannot just be re-torqued - the oxide layer keeps resistance high. Use a brass brush (not steel - steel scratches the copper) to polish the contact zone to bright copper. Wipe with 99% IPA to remove brass residue. Apply a thin film of no-ox conductive grease (Noalox or equivalent) - prevents re-oxidation under thermal cycling without affecting conductivity. Re-torque to 3.0 N-m. Skip the grease and the joint will black-out again within 12 months.

Log V/I telemetry via WhatsMinerTool once the unit is back online. Once the miner is hashing again, the WhatsMinerTool desktop app exposes live PSU telemetry (V_out, I_out, P_out, internal temperature) the Web UI hides. Log 30-60 minutes under full load. Healthy P21: V_out at 12.0-12.3V with <0.1V ripple, I_out stable within +/-3% of expected for your tune, no oscillation. If V_out is drooping >0.3V under load even after a clean re-torque, you have a deeper PSU issue (aged silicon, drifting sense resistor) and the fix bought you time, not a permanent solution.

Thermal-scan the busbars under load with an IR camera. A FLIR One or Seek Thermal pointed at the 12V output bars during full hash will show a uniform 35-50C rise above ambient on a healthy assembly, and a clear hot-spot at any marginal joint. This is the cleanest visual confirmation that your re-torque held - and that no other joint is loose. Sustained >80C at a single bolt = re-torque did not take, or a different bolt is the actual problem. Cycle back to Step 7 with that information.

Audit the control-box-side busbars at the same time. The output cable terminates at the control box on the other end, where another set of M6 bolts feeds the hashboards from the harness. The same fault mode applies - those bolts loosen with thermal cycling too, and a loose joint there presents electrically identically to a loose PSU-side joint. Most DIYers torque only the PSU side. Don't be a most DIYer. 5 minutes of additional work, identical procedure, both sides.

Stop DIY when the rail still won't come up after a clean re-torque + harness check + hashboard isolation. At that point you have a genuine PSU internal failure: output MOSFET shorted or open, sense resistor drifted high, protection IC mis-triggering, OVP reference component cracked. Component-level rework on the P21 PSU board needs hot-air, preheater, SMD replacement parts, and a programmable DC load to verify the fix. None of that is home-shop gear. Ship to D-Central - bench rework runs CAD $80-180 depending on what's failed.

Stop DIY when there is visible damage. Scorched output MOSFETs, blued copper at a busbar (joint hit >250C at some point), swollen capacitor tops on the PSU board, burn marks on silkscreen - any of these means cascade damage. Replacing only the visibly-broken part leaves adjacent damaged-but-not-yet-failed silicon vulnerable to the next event. D-Central's bench process inspects the full input-protection chain, replaces damaged MOSFETs and caps, and tests against a programmable DC load before discharging.

Ship with full context. Pack the PSU and ideally the chassis it pairs with - cross-stack substitution misses marginal interactions between specific PSU and control-board pairings. Include service history (recent moves, PSU swaps, last torque-check date), any DMM measurements you took, photos of the busbar joints before and after re-torque if captured. Match chassis serial to PSU serial on the ship note. Anti-static bag for the PSU. Double-box with >5 cm foam on every side. Canada-wide shipping standard, US / international welcomed.

Decide repair-vs-replace with the bench report in hand. A new MicroBT P21 runs CAD $220-320 (graded-salvage stock at ~60% of new); a bench rework runs CAD $80-180. If the unit is <3 years old and the failure is one component, repair almost always wins. If the unit is 5+ years old and the output silicon is beat up, replacement might be smarter - adjacent parts are on the same aging curve and the next failure will follow the first by months, not years. D-Central provides a photo-documented diagnostic report so you can choose with data, not a guess.