Bitmain APW PSU Bulging Electrolytic Capacitor Replacement
Bulged or vented electrolytic capacitors inside Bitmain APW-family PSUs (APW3+, APW3++, APW7, APW9, APW9+, APW12). Aluminum electrolytics dry out predictably at the 3-5 year mark of continuous-duty mining, gain ESR, lose capacitance, and either visibly bulge/vent or silently drift. Output rail ripple climbs above the hashboard PMIC's tolerance window, the rail sags 0.5-1.5 V at load, the miner trips PMIC undervoltage / volt-low / volt-range, and the operator chases a phantom hashboard fault until someone opens the PSU.
Warning — Should be addressed soon
Symptoms
- Visible bulging, doming, or convex tops on aluminum electrolytic cans inside the PSU chassis
- Brown crystalline residue, dried electrolyte, or chemical smell around the base of any electrolytic
- Vented rubber bung at the top of an electrolytic — Y or K score line has split open
- PSU output rail measures 0.5-1.5 V below spec at the load connector under full mining load
- PSU rail ripple exceeds 200 mV peak-to-peak on a scope (clean rebuilt PSU should sit under 100 mV)
- Miner UI throws hashboard PMIC undervoltage / `volt low` / `volt range` faults despite a healthy hashboard swap
- PSU is 3+ years old and has been deployed in continuous-duty mining at warm ambient
- Audible 60 Hz / 120 Hz buzz from the PSU under load — classic capacitor-aging signature
- PSU fan is at full RPM and chassis runs hot to touch even at moderate load
- Rail recovers to spec when miner is undervolted -10% or more
- ESR meter reads above 0.5 Ohm on bulk primary caps that should be well under 0.2 Ohm
- ESR meter reads above 0.3 Ohm on secondary low-ESR filter caps that should be below 0.1 Ohm
- Issue follows the PSU when swapped between miner chassis (rules out hashboard or control board)
- Repeated PMIC undervoltage trips followed by clean operation when PSU has cooled — cap dielectric reforms briefly when cool
Step-by-Step Fix
Confirm the symptom with a DMM at the PSU output under full mining load. Compare against spec rail for your specific APW model: `12.0-12.5 V` on `APW3++` / `APW7` / `APW9`, `14.5-15.0 V` on `APW9+`, `12.5-13.5 V` (or programmed value) on `APW12`. A `0.5-1.5 V` shortfall at load is the gross signature of cap aging. If the rail is in spec, this is not the page you need.
Measure AC line voltage at the panel under full load. Expect `235-245 V` on 240 V split-phase, `202-212 V` on 208 V commercial, `115-125 V` on 120 V residential. Loaded readings 10-15 V below nominal mean the circuit is the problem. Move the rig to a dedicated `240 V` circuit before spending money on PSU parts — AC undervoltage presents identically to cap aging.
Drop the miner profile by 10-15% and re-test the rail. If the rail recovers to spec at the lower load, caps are marginal but not yet dead — undervolting buys you `6-18 months` while you plan a re-cap. If the rail still sags at lower load, caps are likely dead. Either way, document the result before continuing.
Vacuum the PSU intake and exhaust grilles. `3-5 years` of continuous duty puts a measurable dust film on the heatsinks; clearing it drops internal temperature `5-10 °C` and slows remaining cap aging. Free intervention, mandatory before opening the chassis.
Add an external `120 mm` cooling fan blowing across the PSU heatsink as a Tier-1 holdover. Drops internal temperature `5-15 °C` and frequently restores rail voltage on a marginal APW for `6-12 months` while you order parts. `$15-$30 CAD`.
Capture a thermal image of the PSU under load. `FLIR ONE Pro` clipped to a phone or any IR thermometer. Bulk caps glowing above `75 °C` while heatsinks and FETs sit at `45-55 °C` is the visual fingerprint of cap-driven failure. Uniformly hot = airflow or load problem; cap hotspot = re-cap.
Tighten and clean the PSU output load connector. Power off, IPA on a lint-free wipe, firm reseat. Loose high-current connectors drop voltage at the contact and present identically to a tired PSU.
Inspect and replace the PSU fan if bearing-worn. `60 mm 12 V` on smaller APW units, `80 mm` on `APW12`. `$8-$15 CAD`. Hot PSU = fast cap aging — a tired fan is often the upstream root cause that pushed caps over the edge.
Listen for buzz under load. `60 Hz` or `120 Hz` audible hum from the PSU at full load is the classic cap-aging signature — secondary filter caps with elevated ESR can't suppress switching ripple, and the inductors and transformer windings sing the unsuppressed frequency. Buzz = re-cap.
Open the PSU casing and discharge the bulk caps. Five minutes minimum after power-off and unplug. `10 kOhm 5 W` resistor across each bulk cap's terminals to bleed residual voltage. Verify zero volts with a DMM before reaching in. The primary rail at `~385 V DC` on a 240 V supply is lethal. Take it seriously.
ESR-test every electrolytic with a `Peak ESR70` or equivalent meter. PSU unpowered and discharged. Pull at least one lead per cap if the meter is not true in-circuit. Bulk primaries should read below `0.2 Ohm`; secondary low-ESR filter caps below `0.1 Ohm`. Replace anything above `0.5 Ohm` primary or `0.3 Ohm` secondary regardless of how it looks.
Source replacement primary bulk caps to original spec. Common APW configurations: `APW3++` / `APW7` / `APW9` use `2 x 470 uF 450 V 105 C` low-ESR; `APW9+` uses `2 x 470 uF` or `2 x 560 uF 450 V`; `APW12` uses `2 x 1000 uF` or `2 x 2200 uF 450 V` depending on rev. Use `5000-hour 105 C` parts only: `Nichicon LGW`, `Rubycon BXC`, `Panasonic EE`, or `United Chemi-Con KMW` series. Voltage must match or exceed; capacitance within +/-20%; do not downgrade hours or temperature.
Source replacement secondary `12 V` rail filter caps. Common values: `1000 uF`, `2200 uF`, or `3300 uF 25 V` low-ESR aluminum electrolytic, `105 C 5000-hour`. `Rubycon ZLH`, `Panasonic FR`, or `Nichicon HE` series. These caps are the most common proximate cause of rail droop, and replacing them at the same time as primaries is non-negotiable for a 3-year rebuild.
De-solder the old caps. Through-hole, solder mass is significant on primaries. Temperature-controlled iron at `380 C`, plus solder wick or desoldering pump. Mark polarity before pulling. Pull straight up to avoid lifting pads. A lifted pad on the primary rail makes the unit a Tier-4 ship-to-bench job — do not scab-jumper a `385 V DC` pad.
Solder fresh caps in and verify polarity. Polarity is absolute on the primary rail — a reversed electrolytic on `385 V DC` vents catastrophically the moment power is applied, with real injury risk from the explosive blowout. Triple-check before energizing. Trim leads. Inspect every joint under `10x` magnification for cold solder, bridging, or whisker.
Bring the rebuilt PSU up on a current-limited bench supply or a `100 W` incandescent bulb in series with the AC line for first power-up. If anything is wrong (reversed cap, solder bridge, lifted pad, dead FET), the bulb glows bright and the PSU shows zero output instead of venting a cap. After 5 minutes of clean idle, remove the limiter and verify rail voltage at no-load before reconnecting to the miner.
Verify the rebuilt rail under load. Reconnect the miner, mine at stock profile, probe the PSU output for `30 minutes` continuous full-power operation. Healthy rebuild: rated rail sustained, `<100 mV peak-to-peak` ripple on a scope. Anything above `200 mV` ripple means a secondary cap is still bad or the PWM feedback loop is unhappy — pull every secondary electrolytic and ESR-re-test.
Stop DIY and book a D-Central PSU bench when: re-cap did not restore the rail, switching FET / SR diode / feedback opto measure shorted or visibly scorched, scorch or burnt smell present, PWM controller IC reads abnormal, pads lifted during extraction, vented cap took adjacent components with it, or you don't have the bench safety setup to service `385 V DC` rails. You're now in test-fixture-and-schematic territory.
D-Central bench process: full PSU strip-down, ESR sweep on every electrolytic, FET / diode / opto verification, PWM controller IC sanity check, full cap replacement with `5000-hour 105 C` parts across primary and secondary, post-repair `4-hour` burn-in at full nameplate load on a programmable test fixture. We document the rail trace and ripple before and after — you get the data, not just a green light. Bench rebuild `$95-$185 CAD` plus parts; `5-10 business day` turnaround; Canada / US / international.
The retire-or-rebuild call. Re-cap is worth doing if (a) chassis, transformer, FETs, and PWM controller are all healthy, (b) you have a functional miner the PSU is paired with, and (c) parts cost stays under 30% of the going used-PSU price. A vented `APW3++` paired with a 2018-vintage `S9` frequently fails this math — at that point harvest the fan and chassis as spare parts and step up to a fresh `APW9+` or `APW12` for your next miner. Run the math before ordering parts.
When to Seek Professional Repair
If the steps above do not resolve the issue, or if you are not comfortable performing these repairs yourself, professional service is recommended. Attempting advanced repairs without proper equipment can cause further damage.
Related Error Codes
Still Having Issues?
Our team of Bitcoin Mining Hackers has been repairing ASIC miners since 2016. We have seen it all and fixed it all. Get a professional diagnosis.