DC offset: how much is too much?

WL can show the DC offset of an audio file.
but how much DC offset is tolerable and where do i start considering to remove it?
what is the technical guideline?


From the WWW

DC offset is just offset on the y (amplitude) axis, but removing it doesn’t mean that the waveform will automatically become symmetrical (nor that it should be), it only means the wave will start and end at 0 on the y-axis.

If you remove the DC offset from is a simple waveform like a sine shifted by a certain offset, then the result is of course symmetrical… but a real-world complex waveform can be asymmetrical regardless of any dc offset (because their lack of symmetry is not caused by any dc offset, but from the sum of their partials). For example a male voice is likely to show some asymmetry.

Actually, if a waveform is asymmetrical, you could make the peaks more symmetrical around 0 by adding some DC offset (if you imagine a graph drawn on a piece of paper with a complex waveform peaking at +5 and -3 without any DC, then adding a -1 DC offset would shift the peaks to +4 and -4)… In my opinion it is a bad thing to do in practice, this is just a theoretic example to show DC offset and symmetry are not directly related.

Still, removing DC is a good practice (even if it means making the output waveform more asymmetrical!). A very small (and I mean really small) amount of DC is negligible in my opinion and I usually don’t care, as long as its contribution gets buried by the sounds and the noise floor. If larger, it should be taken care as soon as possible (by finding and, if possible, fixing the cause of the DC offset).

Dealing with asymmetrical waveforms is a different subject in my book. A way to deal with them (when a more symmetrical waveform is needed) is to use a phase rotator (a bunch of all-pass filters). As any other processing, it’s not a free lunch (you’re trading the symmetry of the waveform for something else…), but it’s still a tool that can be useful.

DC offset can be, of course, a cause to the asymmetry of a waveform, but in general it’s not the only (and main) contributor.

Just my 2 cents, of course!


thanks for the info.
i do understand the theory of it and have come across it on occasion. i was never able to “hear” and actual effect, maybe because the amounts were too small.
WL shows a % of DC which is probably a very sensible thing. so if i have 0.05% DC it is a very small amount that is probably negligible.
and i am attempting to get the “probably” out of that assessment. but it seems there is no clear amount that i could find on the www anywhere.
so my best guess is that as long as i stay below 1% it should be technically ok and inaudible.

IIRC, WL will not remove DC offset below a certain threshold, but I don’t know the value. I would think anything below 1% is no issue.