Clipped waveforms -> amplitude information ?

Sooooo, yesterday I’ve recorded a concert from Shobaleader One at the Nijmeegse Vierdaagse festival using my Zoom H2 recorder with a OKM soundman plugin mic. Unfortunately I couldn’t adjust the recording input level low enough to avoid clipping. Digital clipping sound which seems to be most prominent when the drummer strikes the snare drum.

When you look at the hour long recordings waveform in Renoises sample editor, you’ll notice the waveform doesn’t take full advantage of the complete amplitude range, rather the half and looking truncated.

However, when you process the waveform with something like waves x-hum, removing dc and low rumble frequencies (guess this can be done with any eq or filtering plugin), magically deemed lost amplitude peeks appear, having the waveform now occupy the complete amplitude range.

While the clipped snare drums are still sounding after processing, I wonder how the phenomena works in that the shape is retrieved, how it is stored inside the clipped recording?

Can anyone explain this in laymen terms? :slight_smile:

(oh, and recording can be found here; maybe some of you mastering wizards can process out the clipping?)

Adobe Audition has a specialised declipper algorithm to restore clipped away information. Seems to work pretty nice.

Check this vid:

Adobe Audition has a specialised declipper algorithm to restore clipped away information. Seems to work pretty nice.

Check this vid:

thanks, will investigate.

Look at a square wave. You know it can be viewed as being composed out of a series of sinus waves with different amplitude and starting point?

So you apply an effect which will modify the starting points (phases) and amplitudes of that sines. The square will not have that nice square shape any more. And as they were partially cancelling each other optimally to make up that shape, they might now even overshoot over the old straight line of the square.

This can also happen if you only change phases of frequencies. Like DC filter in renoise does this in a great way, apply it many times in series to a sample (sqare wave or hard clipped) with headroom and behold. Then things might also begin to overshoot over the old maximum amplitude - though they will to great extend still sound the same. So your snare is still fucked up though it looks different.

You can view the clipping distortion a bit like imposing certain characteristics of a square wave onto random audio. So same stuff will happen there with the overshoot. I normally, if I process samples, leave or add certain headroom, anticipating the effect I apply and also the possible overshoot.

IDK clipping removing filters probably do some magic to smoothen the discontinuities added by the clipping - they might not restore the original, but will make the sound a lot smoother to listen to, to the point where you won’t notice the grit any more.