Goldshell – PSU Unexpected Shutdown / Brownout

Tier 1 — True cold boot at the wall. Pull AC at the wall outlet (not the IEC inlet on the miner — pulling at the inlet does not drain the bulk caps reliably). Wait a full `60 seconds` so the protection latch resets and the PSU output caps drain. Plug back in. Watch the boot sequence: fans should spin up, dashboard should reach the Miner page, hashing should begin within `90 seconds`. If the miner stays up for more than `30 minutes` under full load, the trip was a one-time event from upstream and you can monitor for recurrence. If it dies again, proceed to Step 2.

Tier 1 — Measure mains AC voltage at the outlet under load. With the miner running at full hashrate, put a multimeter on `AC volts` mode at the wall outlet feeding the miner. Expected: `115–125 V AC` in North America or `220–240 V AC` in the EU / commercial single-phase. If reading is below `110 V AC` (NA) or `215 V AC` (EU), you have a line-side problem — circuit shared with another appliance, undersized extension cord, oxidised wall outlet, or a marginal breaker. The fix is electrical, not the miner. Move the miner to a different breaker on a different circuit and re-test.

Tier 1 — Move the miner to a known-good dedicated circuit. A `KD5` / `KD6` / `LT5` needs a dedicated `15 A` (NA) or `10 A` (EU) breaker with no other appliances on the circuit. An `LT6` / `CK6` / `KD-MAX` needs a dedicated `20 A` (NA) or `16 A` (EU) circuit. If the miner runs stable on a dedicated circuit but trips on the original circuit, the original circuit is the fault — call an electrician, do not blame the miner. If the miner still trips on a clean circuit, the fault is internal to the miner and you proceed to Tier 2.

Tier 1 — Check ambient air temperature at the miner intake. With a thermometer at the intake side of the miner during full load, confirm ambient is below `30 °C`. Goldshells have no per-PSU thermal headroom and trip on `OTP` once the PSU heatsink crosses `~75–80 °C`, which corresponds to about `30–35 °C` ambient. If ambient is above `30 °C`, cool the room (open a window, add a fan, relocate the miner away from a south-facing wall) and return to Step 1. If ambient is in spec but the miner still trips, proceed to Tier 2.

Tier 2 — Open the case, inspect the PSU and hashboard bus. Power the miner off, unplug AC, wait `5 minutes` for caps to drain, then open the chassis. Inspect the PSU board for swollen / leaking / domed caps (instant fail — jump to Step 9 PSU swap), check the PSU output bus and the hashboard input bus for loose screws / oxidation / scorching. Retorque every bus screw with a Phillips #2 to manufacturer spec (snug, not gorilla-tight — about `1.0–1.5 N·m`). Clean any oxidised contact with `99 %` isopropyl on a lint-free wipe.

Tier 2 — Inspect and test the PSU fan. The PSU fan inside the Goldshell case is the most common preventive failure on `KD6` / `LT6` / `CK6`. Power the miner briefly with the case open and listen for bearing whine; observe whether the PSU fan spins at full RPM. A seized or noisy PSU fan is the leading cause of `OTP` trips on these models. If the fan is failing, replace it with an identical-spec replacement before doing anything else — a `$15` fan saves a `$300` PSU.

Tier 2 — Remove all surge protectors and PDUs from the AC path. Plug the miner directly into the wall outlet. Cheap surge protectors and consumer PDUs drop `5–10 V` under continuous high-current load and are a common contributor to `UVP` trips on Goldshell PSUs. If the miner runs stable when plugged directly into the wall, the fault was the strip / PDU, not the miner. Replace the surge protector with a properly-rated `15 A` or `20 A` industrial PDU, or just leave the miner on a direct wall connection.

Tier 3 — Cold resistance test on the `12 V` rail. Miner off, AC unplugged, bulk caps drained (wait `5 minutes`). Multimeter on the `200 Ω` resistance range, probes between the `12 V` bus and ground at the hashboard input. Healthy reading is above `~50 Ω` static (record your model's baseline before the failure if you can — values vary by hashboard generation). A reading below `5 Ω` indicates a downstream short on the hashboard. Do not power the miner on. Continued power-on attempts will turn a recoverable PSU trip into a dead PSU plus a damaged hashboard. Ship to D-Central.

Tier 3 — PSU swap with an identical-model donor. If the cold resistance test passed, swap the PSU with a known-good donor of the **identical Goldshell PSU model** — cross-model swaps are not safe. `KD5`-series PSU does not interchange with `LT5`-series PSU. Power the miner up with the donor PSU and run for at least `30 minutes` under full load. If the miner runs clean with the donor, the original PSU is the fault — replace it. If the donor PSU also trips, the fault is downstream (hashboard load or control-board mis-driving the PSU) and the unit ships to D-Central.

Tier 3 — Thermal-camera scan of the PSU under load. With a thermal camera (or even a `$30` IR thermometer for spot checks), scan the PSU board after `15 minutes` of full-load hashing. Look for hot spots above `~85 °C` on individual components — a single bulging output cap, a single hot-running rectifier, a hot-running transformer winding. Hot spots indicate component-level failure on the PSU. With a hot-air station, the right `105 °C` low-ESR replacement caps, and bench experience, this is recoverable; without that toolset, ship the unit to D-Central for component-level rework.

Tier 3 — Oscilloscope ripple check on the `12 V` rail. With a scope on the `12 V` output of the PSU under full load, healthy ripple is below `~120 mV peak-to-peak`. Higher than that indicates failing output caps or a failing output stage on the PSU. This is a confirmation diagnostic, not a fix — a PSU with high `12 V` ripple is also a PSU that is closer to its trip threshold, and the corrective action is either component-level cap replacement (advanced) or full PSU swap (Tier 3 step 9 or Tier 4).

Tier 3 — Component-level cap replacement on the PSU board. For benchers with a hot-air station, an ESR meter, and the right replacement caps. Identify the failing caps (typically the `100 µF` / `220 µF` / `470 µF` low-ESR electrolytics on the `12 V` output stage) by visual inspection and ESR measurement. Desolder, replace with `105 °C` low-ESR equivalents of the same or higher voltage rating, reflow, retest under load. This is the cheapest path to a recovered Goldshell PSU but it requires bench discipline — the input side of the PSU runs at mains potential, and a probe slip is a hospital trip.

Tier 3 — Hashboard isolation diagnostic. With the donor PSU in place, disconnect the hashboard data cable (leave the power bus connected if you must, but ideally disconnect both) and power the miner up. The control board should boot without a hashboard, the PSU should stay on (no trip). If the PSU stays on with the hashboard isolated, the load-side fault is on the hashboard — ship for component-level repair. If the PSU still trips with the hashboard isolated, the fault is in the control board or the wiring harness — also a ship-to-bench item.

Tier 4 — Ship to D-Central ASIC Repair. Book the [D-Central ASIC Repair service](https://d-central.tech/services/asic-repair/), include in your repair note: the symptom timeline (when the trip started, how often it recurs, what triggers it), the line-voltage measurements you took, any thermal-camera images, and the model and PSU revision (the PSU has a sticker on the case with the model number and date code). D-Central bench workflow: intake photos, AC line-voltage check on the original circuit, cold resistance on every rail, donor PSU swap if available, full thermal scan under sustained load for `30 minutes`, and a quoted repair (or quoted replacement) with transparent pricing.

Tier 4 — Goldshell warranty check (in parallel). If your Goldshell is within Goldshell's warranty window (`12 months` from purchase, requires original invoice), open a ticket with [Goldshell support](https://goldshellhelp.zendesk.com) before shipping anywhere. Goldshell will sometimes RMA a tripping PSU under warranty, especially on younger units. Out-of-warranty units (most that arrive on the D-Central bench — these miners are typically `12+ months` old when the trip starts) are on you. The D-Central path is faster for out-of-warranty work; Goldshell RMA is slower but free if the unit qualifies.