Volcminer D1 High Reject Rate and Stale Shares

Power-cycle the miner for a full 30 seconds at the cord — Web UI soft reboot does not flush stuck stratum state on the D1 control board. After power-up, observe reject for fifteen minutes. If reject collapses, you cleared a transient firmware state. Common after firmware updates, partial config saves, and brownout-induced reboots that survived as inconsistent NVRAM.

Save your current OC/UV settings to a notes file, then revert to the factory stock profile in Web UI > Configuration > Miner (no autotune, default frequency, default `VCORE`). Save, power-cycle, observe reject for fifteen minutes. If reject collapses below 1%, your tuning was the culprit — rebuild the curve in Tier 2 with smaller increments. If reject stays elevated at stock, you have a non-tuning cause and the rest of the diagnosis tree applies.

Switch to the regional pool endpoint closest to your ISP's geography. Litecoinpool: `us.litecoinpool.org:3333` (US East), `us2.litecoinpool.org:3333` (US West), `eu.litecoinpool.org:3333` (EU). F2Pool: `ltc.f2pool.com:8888`. ViaBTC: `ltc.viabtc.com:3333` or `:25`/`:443`. Kryptex: `ltc.kryptex.network:7016` (US) / `ltc-eu.kryptex.network:7016` (EU). Save, power-cycle, observe stale percentage for fifteen minutes. Stale is a direct function of round-trip time — closer pool, lower stale.

Override the router's DNS to `1.1.1.1` (Cloudflare) and `8.8.8.8` (Google). Save the router config, reboot the router, then power-cycle the miner. This is the single highest-yield fix on Canadian residential setups for stale-share elevation: Bell, Rogers, Vidéotron, Cogeco, and Shaw all serve mining-pool DNS slowly enough at peak hours to push the D1 to a sub-optimal regional cluster. Set at the router level so every device on the LAN benefits.

Clean intake filters and verify nothing is restricting airflow. Shop-vac the front grille, wipe down the intake, verify there is at least 15 cm of clear space in front of the miner. Hot intake = hotter chip junction = elevated reject rate. Low-effort, high-yield, and a good baseline before any deeper work.

Multimeter on DC, probe at the PSU-to-board connector while the miner is fully hashing (not at idle — sag manifests under load). Expect `12.0 V ± 0.2 V` sustained on the `12 V` rail. Anything below `11.7 V` under load = PSU sagging. Swap PSU with a known-good unit (server-grade `1500 W` in parallel, or a quality bundled VolcMiner PSU). The cheap `3,900 W` units that ship with grey-market D1s are the most common offender D-Central sees on this exact ticket.

Rebuild OC/UV slowly from stock. Start at factory frequency, factory `VCORE`. Add `+50 MHz` (or `-0.01 V` if you are hunting efficiency). Watch reject stabilize for ten minutes. Add another increment. Stop at the step before reject crosses 2%. That is this miner's silicon-lottery ceiling — and it differs per unit. Two D1s out of the same shipment can hold different stable frequencies.

Undervolt for stability, not efficiency. A small `+0.02 V` to `+0.04 V` bump on `VCORE` at stock frequency is the cleanest fix when chip drift is suspected. Yes, it costs you a few watts. Yes, it pulls the chip back from the edge of its stable curve. The trade is worth it: cooler chips, lower reject, longer hashboard life. Mining Hacker default: undervolt-AND-underclock for longevity, not efficiency.

Replace thermal paste on each board. Power off, disconnect, remove the heatsink from one hashboard. Clean old paste with `99 %` IPA and lint-free wipes. Apply a uniform thin layer of Arctic MX-6 or Thermal Grizzly Kryonaut. Reassemble. Observe chip temperatures and reject rate for thirty minutes. If chip temps drop `5-10 °C`, you found a real cause and reject will follow down. Refresh paste every 12-18 months on a continuously-running D1.

Check line voltage at the panel under load with a multimeter or voltage logger on the same circuit as the miner. Targets: `235-245 V` on `240 V` split-phase, `202-212 V` on `208 V` commercial, `115-125 V` on `120 V`. Below those bands at peak hours = under-sized circuit driving PSU sag driving reject. Real fix: dedicated `240 V` circuit (preferred for any D1-class miner) or move the miner to a stronger circuit.

Scope the `12 V` rail under load for ripple. AC-coupled, full bandwidth, probe ground close to the PSU connector. Healthy ripple `<100 mV pk-pk`. Above `150 mV pk-pk` and you have a PSU on its way out. Replace PSU. While the rig is open, inspect the bulk capacitors near the PSU input on each hashboard for bulging or discoloration — bulged caps drift first, bring rail noise into the digital domain, and surface as elevated reject months before they fail outright.

Inspect and replace bulk caps if visibly drifted. Bulging electrolytics or cracked MLCCs on the input rails of the hashboard are the second most common board-level cause of reject creep on a year-plus D1. This is a soldering-iron + hot-air job, not a reflow job. Match cap value, voltage rating, and ESR class. Test the board on a known-good PSU after replacement before committing it back to the rig.

Per-board isolation: swap suspect hashboard between slots. Power off. Label slots with tape (board 0 / 1 / 2 on multi-board D1s). Move the suspect-bad board to a known-good slot. Power up, observe reject for thirty minutes. If elevated reject follows the board to the new slot = the board is the problem. If reject stays in the slot = the control-board / cable path is the problem. Either way, search space is cut in half.

Reflow the worst chip on a board with localized chip drift. If per-board isolation has identified one chip on one board with higher temperature or asymmetric behavior, remove the heatsink, flux the chip, preheat the bottom side to `~150 °C`, and hot-air the top side at `310-330 °C` for `~30 s`. Let it cool naturally, refresh paste, reassemble, burn in for an hour. BGA-package Scrypt ASICs tolerate a single reflow cycle well; multiple cycles on the same chip stress the package. If reflow does not bring reject inline within an hour, that chip is a chip-replacement candidate, not a reflow candidate.

Roll firmware to a known-good version. Save every config field to a notes file first — VolcMiner's firmware update flow has a documented behavior where un-checking the `keep configuration` checkbox wipes pool, password, and network settings. Cross-check the firmware bundle hash against the manufacturer's published hash before writing. Flash over Ethernet only, with a UPS on the circuit. Roll back one version if you suspect a regression on the latest build, or roll forward if you are more than two builds behind.

Stop DIY when chip-level reflow has been attempted on the worst chip and reject returned within 30 days, or when scope-level rail noise traces to a board-component fault deeper than a bulk-cap swap can address (PMIC failure, voltage-domain IC drift, trace damage), or when two different hashboards in the same rig show identical reject signatures (rules out individual chip silicon failure, points at controller / cable / PSU). You are now in test-fixture territory — the next move is the bench.

D-Central bench process for D1 reject diagnosis: programmable load on the PSU rail to isolate sag from ripple, board-level chip-by-chip behavior characterization with a Scrypt-job feeder, replacement of failing chips with sourced parts (Scrypt ASICs are a more limited secondary market than SHA-256 silicon — supplier turnaround can be longer), full reflow + reseal + 24-hour burn-in at nameplate frequency, and a post-repair reject-rate report so you know what you are getting back.

Ship safely. Anti-static bags around hashboards. Double-box with at least 5 cm of foam every side. Include a written note: observed reject pattern (sustained vs intermittent, time-of-day pattern if any), firmware version, OC/UV profile if any, PSU model, what you have already tried. Saves D-Central's bench 2-3x triage time, saves you billing time. Book at https://d-central.tech/services/asic-repair/.