Waveform not centered on -∞dB line

I’m seeing an issue on this file where the waveform isn’t centered along the middle dB line: the bottom peaks are much higher than the top ones, but the RMS seems to be higher on the top side. I normalized the DC offset, but it didn’t make a difference.

I don’t hear obvious issues with it, but maybe I’m missing something. When dealing with other projects, I’ve noticed that when it doesn’t visually look right (even if it sounds OK), eventually I do find something that I need to dig deeper and fix, but in this case I don’t know what to look for.

This is from a CD, but I also see the same thing on a digitized cassette tape version of the same exact recording of the (audio book) content.

I have a few questions about it:

  1. How can a sound wave have higher peaks on one side or the other? Shouldn’t they be symmetrical across the center axis?
  2. What would cause this? It is related to the original recording environment, equipment, etc.?
  3. Is there any way to fix it? Does it need to be fixed?

Cannot hear asymmetry … https://www.soundonsound.com/sound-advice/q-why-do-waveforms-sometimes-look-lop-sided

Great information, thank you. This explains the issue, since it is a voice recording, the positive air pressure bias makes total sense.

Seeing how out of balance this one is (the very high peak around -1dB on the bottom), do you think it’s worth pursuing a phase rotator to correct it? Or if anything, maybe some limiting or compression?

Audacity’s native high pass-filter applies phase rotation, (a side effect).

gif rot

Trial and error is required : try 20Hz, 30Hz, 40Hz … 80Hz @ 48dB/octave.

May still need to apply a (soft) limiter, (without make-up gain).

Usually it’s not worth it, particularly if you intend to mix with other sounds or apply further processing, or export in any format other than PCM, because any of these processes can reintroduce asymmetry, and may push the peaks into clipping.

Where it may be worth it, is if this is the final mix and you want to “master” it as loud as possible (which is not generally recommended these days). Making the waveform more even can allow you to increase the overall level by a couple of dB while avoiding digital clipping.

The RMS looks pretty even (on average) to me, which is expected when DC offset is absent.

A tell-tail sign of DC offset is when the RMS level is noticeably offset in the same direction as the peaks.

This topic was automatically closed after 30 days. New replies are no longer allowed.