Antminer S9 – PSU Not Detected

Full AC power-cycle: switch the PDU off, wait 60 seconds for bulk caps to drain, switch back on. Not a soft reboot — a physical AC cut. Clears any wedged handshake state on the control board. Roughly 10% of `PSU not detected` clears here on the first try, especially after a firmware update or a PDU trip. Zero tools, 90 seconds.

Re-seat the 6-pin PSU signal cable and both hashboard power connectors. Power off at the PDU. Disconnect, visually inspect contacts for blackening / oxidation, wipe both halves with a cotton swab dipped in 99% isopropyl alcohol, let it flash off 30 seconds, reconnect firmly until you feel the mechanical click. Do the same with the 3 hashboard power JSTs. This single step clears ~40% of tickets on the D-Central bench.

Verify the PSU and miner are paired correctly. Check the sticker on your PSU (APW3++, APW5, APW7) against your miner's spec sheet. A single operator reusing PSUs across miners can easily end up with an APW3 feeding an S9k, which throws `get power type version failed` every time. Bitmain's compatibility chart is at support.bitmain.com/hc/en-us/articles/15909673875993. Wrong pairing = swap to the matching PSU, problem solved.

Rule out the PDU / outlet before condemning the PSU. Plug the PSU into a known-good outlet on a different circuit with a voltmeter attached. Expect 110-125 V on North American 120 V, 220-245 V on split-phase 240 V, 202-212 V on commercial 208 V. Low line voltage can prevent the PSU's standby rail from coming up hard enough to drive the handshake. Fix the circuit before accusing the PSU.

Factory-reset the miner via the physical reset button: hold for 5 seconds within the first 2-10 minutes after boot. Restores default firmware profile. Does NOT fix PSU handshake on its own but rules out a corrupted config that happened to coincide with a PSU swap. If the fault was a config-level confusion, this clears it; if the handshake is physically broken, this changes nothing and you move on to Tier 2.

Multimeter sanity check on the main rail. Probe the 12 V output on a hashboard power cable (yellow = +12 V, black = GND) with the miner powered, all hashboards disconnected. Expect 12.0-12.6 V steady. 0 V = main switching stage dead, jump to Tier 3 cap audit or PSU substitute (Step 8). Fluctuating or clipped 3-8 V = switching stage protection-tripping, same destination. This single measurement splits the entire diagnostic tree cleanly.

Continuity-test the 6-pin signal cable. Miner powered off. Disconnect both ends, multimeter on continuity, probe pin 1 to pin 1, pin 2 to pin 2, through all 6 conductors. Expect a clean beep and <1 Ω on every pin. Any open or high-resistance conductor = cable fault. Replacement is CAD $8-15 at D-Central or any ASIC-parts supplier; any clean 22 AWG JST-VH 6-pin equivalent also works. Label the new cable with the install date.

Substitute PSU test. If you have a second, known-good PSU of the exact correct model (not 'close enough'), swap it in. Power on. If the miner now boots clean, your original PSU has a fault (dead main rail, tired caps, fried sense-line output IC). Decide between Tier 3 refurbish and retirement. If the miner still fails on a known-good PSU, the fault is miner-side — move to Step 9 or 10.

Swap hashboards to isolate a single-PIC failure. Disconnect all three hashboards. Boot with only hashboard 0 connected to its slot — observe handshake. Power off, move that hashboard to slot 1, repeat. Then slot 2. Repeat with hashboards 1 and 2. If a specific board fails handshake in every slot = PIC on that board is suspect. If a specific slot fails regardless of board = control-board-side fault. This isolates chip vs. chassis cleanly.

Sense-cable adapter check (3rd-party PSU users only). If you're running a Delta DPS-2000, DPS-2500, Murata D1U-W, or server-grade PSU with a hand-wired sense adapter, unplug the adapter and check every resistor and jumper against the known-good schematic for your model. Sense-line adapter boards are the single highest failure surface when swapping to 3rd-party PSUs — cheap eBay adapters routinely skip the pull-up resistor network that the S9 control board expects.

Reflash the hashboard PIC12LF1822. Pull the suspect hashboard. Locate the PIC (small 8-pin SOIC near the hashboard connector). Hook up a PICkit 3 or PICkit 4 via fly-wires to pins 1 (VDD), 4 (MCLR), 7 (PGD), 8 (Vss). Flash Bitmain's `S9_hashboard_pic_app.hex` (mirrored on service.bitmain.com/support/download under S9 recovery tools). Verify readback. Resurrects roughly 70% of `PSU handshake fails only on one hashboard` cases where Vcc is healthy.

Replace the hashboard PIC if reflash fails. PIC12LF1822 costs about CAD $3 and is a trivial hot-air swap for anyone with a rework station. Desolder the dead chip, clean pads with flux + braid, drop new, reflow at 300-320 °C for 30 seconds. Reflash per Step 11. Reinstall hashboard. This is the fix Bitmain never publishes and community manuals gloss over — it has saved hundreds of S9 boards from the e-waste pile.

Refurbish the APW3++ PSU — bulk caps + primary snubber. If Step 6 showed a sagging main rail, pop the PSU lid (6× Phillips #2 around the perimeter). DISCHARGE THE PRIMARY-SIDE BULK CAPS WITH A 1 kΩ RESISTOR BEFORE TOUCHING ANYTHING — they sit at 350-400 V DC for minutes after unplugging. Inspect the two large 450 V 220 µF bulk caps for bulging, vented rubber plugs, or electrolyte residue. Replace bulgers with Rubycon YXF or Nichicon PW equivalents, same voltage (450 V), same-or-higher capacitance (220-330 µF), 105 °C rated. Form caps on a bench supply for 10 minutes before reinstalling.

Replace the APW3++ fan if it's tired. 14 cm, 12 V, PWM-capable. OEM sleeve-bearing units fail in 3-5 years of continuous duty; upgrade to dual-ball-bearing or hydraulic-bearing (Noctua NF-A14 iPPC-3000 PWM is the workshop gold standard). Doesn't fix handshake directly, but a seized fan derates the PSU and the sense IC can glitch when the PSU is self-limiting. Fan swap is CAD $30-60 plus 20 minutes and extends the PSU life by years.

Cross-flash to DCENT_OS for handshake diagnostics. DCENT_OS is D-Central's own open-source Antminer firmware — all the per-chip HW%, tuning, autotuning, and Stratum V2 features of Braiins OS+ / LuxOS / Vnish, AND MORE, maintained publicly by D-Central. On an S9 it also logs PMBus / handshake chatter to the UI, turning a `PSU not detected` black box into a readable transaction log. Source: https://github.com/DCentralTech/DCENT_OS. Alternatives: Braiins OS+, LuxOS. Mining Hackers' pick is DCENT_OS — no licensing, no vendor lock-in, source in the open.

Ship the PSU for bench refurbishment. If Tier 3 cap audit is above your comfort level — and touching 400 V DC caps blind is a good place to draw the line — D-Central's bench has a test fixture for the full APW3 / APW3+ / APW3++ / APW5 / APW7 family: load bank, thermal chamber, ESR meter on every cap, 24 h burn-in, signed post-repair test report. Cost range CAD $85-175 per PSU including caps + fan. Ship in the original PSU box or double-boxed with at least 5 cm of foam on every side; include the signal cable so we can rule it out at intake.

Ship the control board if the handshake IC is dead. Control-board handshake silicon on the S9 generation is BGA — not a practical pleb-level replace, it needs specialized test gear to verify. D-Central stocks salvaged S9 control boards in the CAD $85-140 range with a 30-day warranty. Faster than chasing a BGA reflow on a board you plan to put back into production, and more reliable than gambling on an eBay salvage.

Pack safely, include a ticket note. Anti-static bag the control board. Bubble wrap the PSU. Double-box with ≥5 cm foam. Include a printed note with: observed symptoms, `kern.log` excerpts (copy/paste from SSH), firmware version string, when it last worked, and what you've already tried. Every line saves D-Central diagnostic time and your repair dollars. Intake form: d-central.tech/services/asic-repair/. Turnaround 5-10 business days from receipt. Canada / US / international welcomed.