ASIC Miner – PDU Overload

Power off the rack at the upstream panel breaker before touching anything downstream — not just the PDU, the panel feed. Let the miners' APW PSUs fully discharge for at least 60 seconds before unplugging anything. A PDU under live load arcs when you pull a `C19` cord, and the arc damages both the PSU-side and PDU-side contacts, which often shows up as a 'dead PSU' ticket weeks later.

Run the math on paper. List each miner, its nameplate wattage, its measured wall draw, sum the wattage, and divide by line voltage (`240`, `208`, or `120`) to get total amperage. Compare to the PDU's continuous rating: breaker rating × `0.8`. A 30 A PDU = `24 A` continuous. If your total exceeds 80% of the branch rating you have a design-level overload and no amount of troubleshooting will fix it — remove miners or upgrade the PDU.

Redistribute miners across the PDU's outlets to balance branches evenly. On split-phase PDUs put equal load on each branch; on 3-phase units rotate L1-L2-L3-L1-L2-L3 as you plug in. Most racks we see in the wild have all the heavy miners on one branch and the other empty — that guarantees a trip even when total rack load is fine. Target within 20% balance across branches.

Touch every cord plug after 20 minutes of full load. Warm is OK; anything above mildly warm is suspect. Use an IR thermometer if you have one (a `$35 CAD` Canadian Tire unit is plenty). Note which plugs read hot — those are your Tier 2 candidates. Readings above `55-60 °C` when ambient is `25 °C` indicate high contact resistance that will escalate to a scorched outlet.

Visually inspect every outlet for scorching. Unplug one cord at a time with power off and look inside each outlet with a flashlight. Blackened brass, melted plastic at the contact, or any discoloration means the outlet is compromised — do not re-use it. If more than one outlet shows damage, the PDU is end-of-life and must be replaced rather than patched.

Measure line voltage at the PDU output under full load with a true-RMS multimeter on AC. Probe L-to-N at an unused outlet or the PDU terminal block while every miner is hashing at nameplate. Target within `3 %` of nominal: `232-248 V` on 240 V, `202-214 V` on 208 V, `116-124 V` on 120 V. Sag greater than `4 %` means either the upstream cord or the PDU itself is the bottleneck — turn into heat at every connection, then into damage.

Replace the upstream `L6-30P` / `L14-30P` cord if it is more than 10 ft long, not 10 AWG, or visibly scorched at either plug. A quality 10 AWG 6-8 ft cord with molded `L6-30P` ends is roughly `$45-90 CAD` and solves a surprising number of mystery voltage-sag tickets. If the receptacle at the panel is also hot, stop — that is building wiring and requires a licensed electrician.

Swap any hot `C13` / `C19` cord for a new rated one. `C13/C14` is 10 A continuous, `C19/C20` is 16 A continuous, and the newer S21 `P13` is 20 A and requires a 20 A-rated `C19` outlet. Scorched lugs are not repairable — the brass has annealed. Replace with rated cords from a reputable vendor; cheap no-name `C19` cords are a fire hazard we see on the repair bench monthly.

Load-balance branches by relocating miners across outlets while watching the metered PDU display (if equipped). Target all branches within `20 %` of each other at steady-state hashrate. This alone clears many 'random breaker trip' tickets where aggregate rack load was fine but one leg was at `30 A` and another at `12 A`. Unbalanced phases are the #1 reason a single branch trips while the rack total looks OK.

Verify breaker rating against real continuous load. If the rack genuinely needs `30 A` continuous, you need a `40 A` breaker + PDU (`40 × 0.8 = 32 A` continuous). This isn't negotiable; it's NEC 210.20(A). Upgrading the breaker requires a licensed electrician at the panel — measurement itself is DIY, the physical upgrade is not.

Open the PDU chassis with power off and locked out. Inspect the internal bus bar, breaker terminals, and cord-entry gland. Discoloration or corrosion on the bus bar = retire the PDU. Loose terminal screws are common after years of thermal cycling — re-torque to the manufacturer spec (typically `~25 in-lb` for 10 AWG terminals, verify on your PDU's label). Never guess: loose is arcing, too tight is stripped threads and arcing.

Flash DCENT_OS (D-Central's open-source Antminer firmware — Mining Hackers' option with full per-chip diagnostics, tuning, autotuning, and real-time input voltage monitoring) on every Antminer fed by the PDU. See https://d-central.tech/dcent-os/ or GitHub source. Alternatives: Braiins OS+, LuxOS, Vnish. All expose input voltage to the dashboard so you can watch sag happen across the rack and correlate with PDU branch load. Stock Bitmain firmware hides the number; custom firmware makes PDU overload visible in five minutes.

Underclock every miner 3-5% as permanent mitigation on any rack running above 70% of continuous branch rating. Small hashrate loss, large thermal and electrical margin gain. D-Central's repair queue shows PDU-caused PSU failures drop by roughly half on rigs that sit at `60-70 %` of continuous branch rating vs rigs at `80-90 %`. Apply via autotune on DCENT_OS / Braiins OS+ / LuxOS / Vnish, or by profile on stock Bitmain firmware.

Rework the rack's power layout so each branch carries the same miner profile. Don't mix an S9 (~`1400 W`) with an S21 (~`3500 W`) on the same branch — the S21 dominates and masks its own overload. Put matched miners on matched branches: 2× S19 XP per 30 A branch at 240 V is roughly `75 %` continuous, a reasonable sustained configuration. Plan the layout on paper before plugging anything in.

Install per-outlet current monitoring on any PDU that doesn't have it. Aftermarket clamp-on current sensors with an SNMP or MQTT gateway cost `$200-400 CAD` and turn your rack into a monitored asset. The alternative — finding out about overload via a scorched plug — costs more every time. The first prevented failure pays for the upgrade, and you catch creeping load increases before they trip.

Stop DIY if you see any of: melted outlet plastic with ongoing visible scorching, evidence of fire inside the PDU (char marks, soot, smell), a burnt-smelling receptacle at the building panel, or any conductor showing heat damage beyond a single outlet. You are in electrician territory — continuing DIY risks fire, arc-flash injury, or insurance-invalidating damage. Book D-Central for the downstream miner repair and a certified electrician for the PDU / panel / wiring side.

D-Central bench workflow for post-PDU-event miners: full PSU test on a programmable load (catches weak caps that pass idle but sag under load), control-board SD recovery, per-chip HW% isolation via DCENT_OS on the bench, `PIC` reflash where corrupted, and a 24-hour burn-in at nameplate before return. Turnaround 5-10 business days, Canada / US / international. This is not DIY-recoverable without a test fixture.

Ship safely after a PDU event. Anti-static bag the hashboards, double-box with ≥ 5 cm foam on every side. Include a written note with observed PDU current, line-voltage sag measurement, firmware version, and suspected overload duration — that saves hours of bench diagnostic time and therefore your repair cost. Don't ship the PDU back unless it's the subject of the repair; ship the miners.