NerdQAxe++ – PSU Error Guru Meditation #0000015

Verify the PSU spec on the sticker. It must read 12 V DC at 10 A minimum, 120 W minimum. An 8 A PSU — including the one some early NerdQAxe++ units shipped with — will silently trigger #0000015 under full load. A USB-PD trigger cable negotiating to 12 V on a 65 W laptop charger is not enough. A generic lab supply without a clear current spec is not enough. If your sticker reads below 10 A, you have found your cause; order a correct PSU before touching anything else on the miner.

Cold power-cycle at the XT30 connector for a full 30 seconds. Not 5. Not 10. Count out thirty. Soft reboots via the AxeOS restart endpoint do not reset the onboard DC-DC regulator state after a latched #0000015 fault — only full power removal does. Reconnect and observe the display for 3 minutes. On a NerdQAxe affected by a transient VREG fault rather than a hard PSU or thermal issue, a clean cold boot sometimes resolves the fault for the current session.

Move the unit to a flat, open surface with at least 10 cm of clearance on all sides. Shelf-mounted inside a closed cabinet or a mining rack with no side clearance traps heat around the onboard VREG and biases the temperature sensor into fault territory even before the ASICs get warm. This is a free diagnostic: a NerdQAxe that clears #0000015 when moved to an open desk has a cooling-headroom problem, not a PSU problem.

Power-cycle the WiFi router and retest. If the ESP32-S3 is hung in its networking stack during boot, the chip-init sequence for the four BM1370 ASICs drifts and the firmware's VREG-health check can fire prematurely. Not a common cause on its own, but a two-minute diagnostic that costs nothing and rules out an otherwise noisy variable before you start measuring rails.

Check the firmware build. Open AxeOS / NerdQAxe dashboard, note the build string from System Info. If it is the most recent shufps/ESP-Miner-NerdQAxePlus release and the problem started at that upgrade, use the shufps/nerdqaxe-web-flasher in Chrome or Edge to flash the prior stable release. Archive the prior .bin locally going forward — Nerd firmware history is a toolkit, not baggage, and some community releases have shipped with over-tight VREG trip thresholds.

Measure the 12 V rail at the XT30 under load. Multimeter on DC, positive probe on the XT30 + pin, negative on the XT30 GND pin. Power the unit and watch the meter for 30 seconds. Expect `11.8 V – 12.2 V` sustained. Anything below `11.4 V` at steady state = PSU cannot deliver. Swap to a verified 12 V / 10 A brick (the D-Central NerdQAxe PSU is spec'd tight for exactly this use case). This single measurement is the fastest way to confirm or rule out the PSU as the cause.

Inspect the XT30 connector and the pigtail solder joints under magnification. XT30 is rated to 30 A continuous but the pigtails are often soldered by hand at the factory. Look for cold solder joints, un-heatshrunk strands that can short pin-to-pin, or a loose crimp inside the shell. A cable with intermittent contact produces identical symptoms to a tired PSU but rides the current at the joint and will eventually overheat. Replace a suspect pigtail rather than try to rehabilitate it.

Strip the heatsink and inspect the thermal interface. Power off, disconnect the XT30, unscrew the four heatsink corner screws with a Phillips #1. Lift the heatsink straight up. You are looking for: (a) paste squeezed unevenly across the four BM1370 dies, (b) paste that is dry, cracked, or grainy rather than homogenous, or (c) one or two dies with no paste contact at all. Any of the three is your answer. A healthy install looks like a flat silver film on every die and a matching pattern on the heatsink face.

Clean both surfaces with isopropyl alcohol 99% on lint-free wipes — not paper towel. Remove every trace of old paste from the four dies and the matching heatsink pads. If the paste is dry enough that IPA is not cutting it, let a damp wipe sit on the surface for 30 seconds before wiping; do not scrape with metal tools. Do NOT use acetone — it attacks the PCB solder mask near the dies and can lift silkscreen or damage nearby components.

Repaste with Arctic MX-6 or Thermal Grizzly Kryonaut Extreme. Pea-sized dot on each of the four BM1370 dies. Arctic MX-4 works too but MX-6 has materially better longevity at the 45-60 °C steady-state these chips run. Do NOT glop paste on — excess squeezes out and contaminates pads. Reseat the heatsink square to the board and tighten the four corner screws in a diagonal cross pattern (top-left, bottom-right, top-right, bottom-left), snug but not over-torqued. Power up and retest for a full 10 minutes before declaring victory.

Capture the ESP32-S3 serial log on boot. Connect USB-C from the NerdQAxe to a laptop. Open a serial terminal at 115200 baud — on Linux/Mac: `screen /dev/ttyACM0 115200` (adjust device path); on Windows: PuTTY, Serial, COMx, 115200 baud. Power the unit via XT30 and capture the first 60 seconds. Save the log to disk — it is the single most useful artifact if you end up shipping the board for repair, and it disambiguates firmware panic from hardware fault in about 15 seconds of reading.

Flash the current stable release via the shufps web flasher. Open Chrome or Edge (WebSerial is not supported in Firefox or Safari). Navigate to the shufps/nerdqaxe-web-flasher page. Connect USB-C, hold BOOT on the ESP32-S3, tap RST, release BOOT to enter download mode. Select the correct .bin for your exact board revision (rev 4, rev 6, rev 6.1, or Hydro — confirm against the silkscreen near the XT30 input), flash, reboot. Reconfigure WiFi and pool from the AxeOS AP after first boot.

Inspect the VREG step-down regulator and associated passives under a 10x loupe or USB microscope. Check the onboard DC-DC buck near the XT30 input for cracked solder flanks, a lifted inductor, discolored caps near the VREG feedback pin, or visible heat damage. Shipping impact can fracture the feedback divider resistors — small 0402 parts are particularly vulnerable — and a fractured feedback resistor produces exactly the voltage-readback-fault pattern that trips #0000015 regardless of what the input rail looks like.

Measure individual BM1370 VCORE if you can reach a test point. Boards with exposed VCORE test points let you probe core voltage at each of the four chips under load. Healthy VCORE sits at `1.10 V – 1.20 V` per chip at stock 600 MHz. A chip reading `0 V` or `>1.3 V` is faulted or the shared VREG is misregulating for that domain. This is a diagnostic check, not a repair — use it to inform whether you ship the board to the bench or keep chasing firmware.

Erase NVS and factory reflash if stuck state persists across clean web-flasher flashes. With esptool.py and USB-C: `esptool.py --chip esp32s3 erase_region 0x9000 0x6000`. Then flash the factory .bin for your board revision. NVS corruption after an interrupted OTA is a documented Nerd-family gotcha and erase_region recovery is the canonical fix — see the shufps/ESP-Miner-NerdQAxePlus repo README. This is the last-resort firmware maneuver before ruling that the fault is hardware.

Stop DIY and ship to D-Central when: (a) you have swapped to a known-good 12 V / 10 A PSU and cleanly repasted the four dies and #0000015 still appears; (b) the VREG package or surrounding passives show visible damage under 20x; (c) any BM1370 die surface is chipped, cracked, or heat-discolored; (d) you have flashed current stable firmware via the web flasher and erased NVS with no change; or (e) you have burned more than 2 hours of diagnostic time. Past that line, our bench is materially cheaper than another evening of yours.

D-Central Nerd-family bench process: USB-C serial diagnostic first (5 minutes, saves you repair cost); current-draw profile on bench supply at 12.0 V, current-limited to 12 A; VREG inspection under 20x with thermal-imaging cross-check under load; known-good BM1370 swap from graded inventory if a chip failed its init sequence; thermal interface material re-seated with Kryonaut Extreme; 24-hour burn-in at nameplate hashrate (4.8 TH/s on rev 6 or 6.0 TH/s on rev 6.1) on the D-Central bench stratum before return-shipping. We track board revision, PSU variant, and firmware build per repair.

Ship safely. Anti-static bag, double-boxed with at least 3 cm of foam on every side — especially over the heatsink, which is exactly what caused your fault in the first place. Include a note: symptoms observed, PSU brand and rating, firmware build last running, and which tiers you have already tried. Optionally include the PSU you are using so we can test the full power path end-to-end. NerdQAxe units ship Canada-wide for roughly CAD $15-25; US and international welcomed.