DC Offset Checker

How the DC Offset Checker works

DC offset checker: how it works

A healthy audio waveform swings equally above and below zero, and its average over any reasonable stretch is zero. When a recording chain leaks a little constant voltage into the signal — a cheap interface, a DC-coupled synthesiser output, a damaged capacitor, a software bug in a generator — the whole waveform shifts up or down. That shift is DC offset. You cannot hear it directly, because a constant is not a sound, but it causes a series of small problems that are worth catching early.

What it does to a file. The most direct cost is headroom: if the waveform is sitting 2 % above centre, the positive peaks reach the clipping ceiling 2 % sooner than the negative ones, and the file cannot be normalised as loud as it could otherwise be. The Headroom lost line shows this in decibels. The second cost is clicks: any edit, cut, loop point or fade that starts from a non-zero level produces a step, and a step is a click. A file with a large offset will click at every cut in an editor. The third is that processors which react to level — compressors, gates, saturation, some limiters — see the offset as signal and behave asymmetrically. Finally, a big offset pushes a loudspeaker cone permanently off-centre, which is harmless at small values and hard on the driver at large ones.

How it is measured. Every sample of each channel is summed and divided by the count, giving the mean. The mean as a fraction of full scale is the offset percentage; the same value in dBFS puts it on the scale audio engineers think in. The verdict is graded: under 0.01 % is negligible and normal for any real recording; up to 0.5 % is small and harmless for playback, though worth removing before editing; up to 3 % is noticeable and should be removed; above that the source hardware is probably faulty. Each channel is reported separately, since a bad cable or a faulty preamp channel typically affects only one side.

Subtract the mean is the exact fix: the measured offset is subtracted from every sample, leaving the audio otherwise bit-for-bit untouched. It is the right choice when the offset is constant through the file, which is the usual case.

High-pass filter is the choice when the offset drifts — a slowly wandering baseline from an unstable power supply, or a recording that was silent and centred for a minute and then shifted. A gentle high-pass at a few hertz removes the constant and the drift together while leaving everything above the cutoff alone; at 5 Hz the lowest audible bass, around 20 Hz, is not affected. Raise the cutoff to 20 or 30 Hz if the file also carries sub-sonic rumble from handling noise or wind.

No change exports the file as it is, which is handy when you only wanted the measurement but the page is set up to save the result.

The export is a WAV at the original sample rate and channel count. Run it through the checker again to confirm the offset now reads zero.

Related tools