How is compression quality related to sample rate?

3

The higher the sample rate is, the better the compression quality will be. That seems to be the conclusion to draw from several dynamics plugins I've worked with that implement oversampling (in this answer it is mentioned too).

I can fully understand why the bitdepth is important when speaking compression, since it directly controls the number of posssible steps or levels within one sample.

But how is compression quality related to sample rate?

Is it because it allows faster attack times and more accurate release slopes, or what is going on?

EDIT: We normally say that 44.1 kHz is more than enough to cover the audible range of the average human being (~ 20 Hz .. 20 kHz). So in that regard, there is no insentive to go higher. I can understand why going in a little higher will reduce the accumulation of aliasing errors, something that would affect any effect or process of the signal, not just compression.

Michael Hansen Buur

Posted 2016-01-27T09:15:12.707

Reputation: 3 612

Answers

1

It has nothing to do with the Nyquist frequency, because you correctly state that there is no significant energy above 20 kHz, and we wouldn't be able to hear that anyway. If you over-sample a waveform, you can get a better representation of the waveform, so it indeed yields a better response to peaks in the compression algorithm. That's why something like the R128 loudness standard includes 4x oversampling: that way you get a "True Peak" measurement, which may be significantly higher (over 0 dBFS i.e. possibly distorting the DAC) than the peak measured at 1x sampling rate.

user2235980

Posted 2016-01-27T09:15:12.707

Reputation: 104

Informative. So we think of the samplerate more along the lines of image resolution: not only will details stand better out; general properties of larger features will appear smoother. I.e. here it is not about the ability to show minor details - we want a more precise temporal canvas to place events and a smoother representation of waveforms in general? – Michael Hansen Buur – 2016-02-01T10:58:16.010

If you can edit the answer a little to address the topics in question better, I'll mark it the answer for now. – Michael Hansen Buur – 2016-02-01T10:59:06.513

1As for aliasing, I don't think you will practically hear that. It's just that the 44.1 kHz signal contains "hidden peaks" that are only revealed with oversampling. I've done the experiment myself with lots of input files to compare the highest peaks at 1x and at 4x the sampling rate. There are lots and lots of occasions where the dBFS value at 1x is much lower than the value at 4x oversampling. It makes sense that a compressor can use this extra information to more accurately react to the signal. – user2235980 – 2016-02-01T13:30:25.647

Thats a good explanation - you should add that to the answer – Michael Hansen Buur – 2016-02-01T14:58:29.873

1

basically if the sound you have recorded is across the sound spectrum as a whole then that would translate to the sample rate.so fuller sound more possibilities of compression without disturbing the character of that sound.

Nithin Ganpati

Posted 2016-01-27T09:15:12.707

Reputation: 17

Thanks I can understand that. What if the recorded material has no significant energy above 20 kHz, and certainly not enough to drive the sidechain? – Michael Hansen Buur – 2016-02-01T07:56:47.240