0

I was listening to a song and noticed the drums sounding very compressed/clipped, but when I actually took a look at the waveform, this is how it looks: waveform Now I'm wondering whether there is some kind of effect that causes this and also made me believe the sound is clipping, or whether I heard wrong and this is just how the drums are sounding. The red parts look flat, but they still have a lot of high frequencies in them, unlike clipped waveforms would.

Could this be caused by some very aggressive compression?

Improve this question
2
  • It is difficult to answer to your question without the actual sound file.
    – audionuma
    Commented Nov 19, 2016 at 18:06
  • Yes, in afterthought this was a really ambiguous question. :| I don't think it'd be legal to upload the sound file. I was listening to this song by Zack Hemsey and first thought it was clipped but was then puzzled by the waveform.
    – DeinFreund
    Commented Nov 23, 2016 at 13:56

4 Answers 4

Reset to default
1

A hihat will produce quite a bit of sound beyond 24kHz, the highest representable frequency when working with 48kHz sampling frequency. It's not really audible content but can lead to clipping in the analog circuitry nevertheless. Since the analog content (including the clipping artifacts) is still being low-pass filtered in both analog and digital domain and subsampled before being processed at the final sampling rate, most clipping in the analog domain will not lead to recognizably flat lines in the result, particularly for clipping involving signal components outside the hearable range.

Improve this answer
1
  • I'm not quite sure how this could happen, considering that in this case this was a sole drum without many pronounced higher order frequencies.
    – DeinFreund
    Commented Oct 15, 2017 at 12:22
-1

yes, this could be because it was compressed very aggressively, and then the input volume was decreased, lowering the volume, but not changing the compression curves.

don't quote me on that, this is how I think it works, but I'm pretty sure about my knowlege

Improve this answer
-1

A limiter can compress signal and keep the signal bellow an ansignable treshold:

enter image description here

Improve this answer
5
  • Wouldn't a limiter usually decrease the amplitude instead of "cutting off" peaks? The length of the sample I posted is about 30 ms, the limiter would need to have an attack of ~3ms, is this actually ever done?
    – DeinFreund
    Commented Nov 17, 2016 at 21:52
  • this isn't as much a limiter as a compressor
    – Grey
    Commented Nov 20, 2016 at 12:32
  • @Grey 1) This is a limiter (look at the name at the top) and not a compressor (sort of, read 3). 2) A limiter/compressor/gate/expander are all things that work on the same principle. They are called dynamic range processors, and they all work by having a threshold, and reduction of amplitude when the level passes that threshold. 3) The difference between a limiter and compressor is just he ratio of attenuation. Limiters are usually 20:1 and above, while compressors are generally around 2:1 to 8:1. You CAN make a limiter with a compressor. ovnilab.com/articles/limiter.shtml
    – user22688
    Commented Oct 14, 2017 at 17:17
  • 1
    @AytAyt woops, yes, my mistake, I guess I was super tired while writing that comment
    – Grey
    Commented Oct 15, 2017 at 18:09
  • 1
    Just wanted to add that a practical use of a limiter is by having one on master bus just to avoid any hard cliping (esspecialy when working in a digital workstation). You could use a limiter as a compressor but without the level of customization you would get when having one dedicated for that.
    – Dalv Olan
    Commented Oct 17, 2017 at 6:43
-1

Clipping leads to the strict form you describe only when it occurs during A/D conversion and no processing/mixing/equalizing whatsoever is done afterwards. Not even proper D/A conversion (which involves reconstruction filtering).

Looking at unprocessed time-domain waveforms for that reason is less reliable than actually listening since "listening" involves a lot of multiband filtering and analysis and pattern recognition that is a lot more suited for auditory recognition problems than a visual look at unprocessed graphs.

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.