Crest Factor

Crest factor is the ratio of the peak value of a waveform to its RMS (root-mean-square) value. For a perfect sine wave the crest factor is about 1.414 — the square root of two. The metric matters because it captures how peaky a current draw is, which a simple RMS reading completely hides: two loads can pull identical RMS amps while one draws smoothly and the other draws in violent spikes. Every piece of equipment upstream of the load has to survive the peaks, not the average, and crest factor is the number that tells you how far apart those two figures are.

Why nonlinear loads have high crest factor

Switch-mode power supplies — the kind inside every modern ASIC miner — do not draw current smoothly across the AC cycle. A simple rectifier front-end charges its input capacitors only during the brief window near each voltage peak, so the current waveform becomes a train of tall, narrow pulses riding on nothing. The resulting crest factor can reach 2 to 3 or more, meaning instantaneous peak current runs two to three times the RMS value. The saving grace in mining is that server-grade supplies like the APW12 include active power-factor correction, which forces the input current to track the voltage waveform and pulls crest factor back toward the sine-wave ideal. Cheap supplies without PFC — bargain-bin ATX units, some consumer electronics sharing your circuits — remain the peaky offenders.

Implications for sizing equipment

High crest factor forces upstream equipment to handle peak currents far above what the RMS load suggests. Generators, UPS units, and transformers all publish crest-factor ratings precisely because a load can be within their RMS capacity yet still clip, distort, or trip on its peaks. Ignoring crest factor when sizing a backup generator for a mining site is a classic way to buy a machine that carries the load on paper and stumbles in practice — nuisance trips, flat-topped voltage waveforms, and overheating despite an apparently comfortable margin. The peaks also stress components directly: conductors, breakers, and capacitors heat according to the square of instantaneous current, so pulsed draw ages them faster than the same RMS delivered smoothly. When in doubt, size for the peaks or choose loads with PFC — with modern mining PSUs on a clean three-phase or 240V split-phase service, the miner is usually the best-behaved load in the building.

Crest factor on the repair bench

The concept earns its keep at small scale too. A bench power supply driving a hashboard under test sees pulsed demand as chips clock up; a bench unit sized only to the average current can droop on the peaks and produce test results that look like board faults. Inexpensive clamp meters report RMS (and cheap non-true-RMS meters mis-report even that on distorted waveforms), so a board that "only draws 8 amps" may be spiking well past the supply's limit. If a test setup behaves erratically at high frequency but cleanly at low, measure the peaks before condemning the board.

Crest factor is closely tied to harmonics — the same pulsed current that raises crest factor injects harmonic distortion into the supply — and both belong to the broader study of power quality. Related sag behavior under load steps is covered in voltage sag.

Crest factor applies to voltage as well as current: a heavily loaded network's flat-topped voltage waveform has a reduced crest factor, and rectifier front-ends charge to that lower peak, eroding the very holdup margin that rides through sags. Measuring both sides with a true-RMS meter that also reports peaks is the fastest way to characterize an unfamiliar site before hardware ships to it. For a home installation the checklist is short: prefer PFC-equipped supplies, avoid sharing miner circuits with peaky consumer loads, and size any generator with its published crest-factor rating in hand.