Capacitor

A capacitor stores electrical energy as charge on two conductive plates separated by an insulating dielectric. It opposes changes in voltage, making it the partner to the inductor in smoothing power and the first line of defense against electrical noise on every rail of a mining machine. Capacitance is measured in farads (F), though practical parts span microfarads (uF) for bulk storage down to nanofarads (nF) and picofarads (pF) for high-frequency filtering. A modern hashboard carries hundreds of them, and statistically they are among the most likely components to be the reason a board sits dead on your bench.

Decoupling, bulk filtering, and ESR

Small ceramic (MLCC) capacitors sit right next to the ASIC chips as decoupling caps, supplying instantaneous current during fast switching transients and shunting high-frequency noise to ground. An ASIC hashing at full tilt draws current in sharp gulps billions of times per second; without local decoupling on each voltage domain, the rail would sag on every transient and chips would compute garbage. Larger electrolytic or polymer caps provide bulk filtering, smoothing the ripple left over after the rectifier and switching stages inside the PSU. The key hidden spec is ESR (equivalent series resistance): high ESR means poor filtering and self-heating under ripple current, which is how a capacitor cooks itself in a warm mining environment.

How capacitors fail

Electrolytics dry out with heat and age, quietly losing capacitance until ripple gets through and downstream electronics start misbehaving — a classic slow failure in PSUs that have run hot for years. MLCCs fail differently: they crack from board flexing during rough handling or from repeated thermal cycling, and a cracked ceramic often fails short, dragging its entire rail to ground. A single shorted decoupling cap on a core rail will keep a whole hashboard from powering up, and because there may be dozens of identical caps in parallel on that rail, finding the guilty one is the real work.

Finding the short on the bench

The standard workflow: measure each rail's test point to ground with a multimeter in resistance mode. A healthy core rail reads low but not zero (the ASICs themselves are a load); a dead short reads near zero ohms and tells you a cap, a chip, or a converter has failed short. From there, technicians isolate the culprit by injecting a low, current-limited voltage from a bench power supply and feeling or imaging for the component that warms up — a thermal camera turns this from an hour of probing into a thirty-second find. Freeze spray works in reverse: the shorted part thaws first. Once located, the cap is removed and the rail re-measured; if the short clears, replace the cap and retest before celebrating, since a shorted cap sometimes died protecting you from a deeper fault.

Replacement discipline matters as much as diagnosis. Match or exceed the original's voltage rating and temperature grade — mining hardware lives hot, and a 105-degree-rated part swapped for an 85-degree bargain is a repair with an expiry date. Keep capacitance and dielectric class in the same family, and on dense ceramic banks resist the urge to simply delete a failed cap and move on: the rail may power up, but you have quietly raised its ripple and stressed every neighbour that now covers the missing share of transient current.

The craftsman's takeaway is that a shorted capacitor can mimic a far more serious failure — a board that looks like it has a dead ASIC or a blown power stage is often one ten-cent ceramic away from full hashrate. That is why capacitors are checked alongside the switching transistors and the domain they serve before any chip-level verdict is made. If you would rather not hunt shorts yourself, D-Central's repair bench does this daily, and boards written off as scrap routinely come back to life.