Compressor effect

[billw58]

Anyone want to help me confirm the behaviour of the Compressor effect?

Generate Chirp: Sine; 440 Hz to 440 Hz; amplitude 0.1 to 1.0; linear; 30 seconds long
Effect > Compressor:
- Threshold: -20 dB
- Noise Floor: -80 dB
- Ratio: 10:1
- Attack Time: 0.1 secs
- Decay Time: 1.0 secs
- Make-up gain NOT checked
- Compress based on peaks IS checked

It looks to me like make-up gain is being applied even though the option is off. Also, there is a nasty bit between 0 and 0.5 seconds.

Undo and try again but this time set Compress based on peaks to UNchecked.

It appears that make-up gain is correctly not applied in this case. There's still the nasty bit in the first 0.5 seconds.

I think the Compressor effect might be broken.

-- Bill

[steve]

It's not broken, it's just a bit strange.

"Peak" compression is an "upward compression" which boosts low signal levels to compress the high level dynamics.
When "Peak" is not selected it bases the compression on rms levels and is the more familiar "downward compression" (lowers high level signals).

I think the "nasty bit" is due to a couple of factors - 1) this occurs near the noise floor level. 2) the compression is adjusted in "steps" rather than continuously.
The "steps" can be seen if you stretch the processed track vertically and zoom out horizontally.

You can see a full discussion about this effect here: http://n2.nabble.com/Audacity-s-built-i ... l#a3327277

It's a big improvement on the old version, it's fairly easy to use and with "real world" material the sound is quite good.

For a really good (conventional) compressor, try the "SC4" which is available here: http://plugin.org.uk/ladspa-swh/docs/ladspa-swh.html

[billw58]

It's not broken, it's just a bit strange.
"Peak" compression is an "upward compression" which boosts low signal levels to compress the high level dynamics.
When "Peak" is not selected it bases the compression on rms levels and is the more familiar "downward compression" (lowers high level signals).

I think that should be documented. I thought (based on the curve) that we were just talking about peak-detection vs. rms detection.

You can see a full discussion about this effect here: http://n2.nabble.com/Audacity-s-built-i ... l#a3327277

Ahhhh ... now I see it:
"The compression formula is at heart the same as before, except it is done
without log and exp functions (for performance reasons) and in peak mode it
maps a 1.0 input to a 1.0 output. By doing this I can avoid the need to do
a second (normalization) pass on the data. The previous code would map 1.0
input to something in the neighborhood of 0.25 (depending on the threshold
and compression levels), so you always had to do a normalization pass to
bring the peak levels back up to 1.0. I built a gain directly into the
compression formula to offset the amplitude loss inherent in the
compression. You will note that the compression graph in the UI also
reflects this change.

In the RMS case it isn't as predictable as the peak mode, but I picked 0.25
as a desired full-scale RMS level, and this gives a good result. There is
nothing magic about the 0.25 number, so feel free to suggest something
different. In the RMS case you may still need to do the normalization pass
to avoid clipping.

The quick way to see the difference is to generate a chirp with amplitude
ramped linearly from 0 to 1, and apply both versions of the Compressor
effect. "

So if peak detection is selected make-up gain happens automatically. If rms detection is selected the algorithm sets 0.25 (-12 dB) at the target value and gives the user the option of gain make-up or not - for speed, I suppose?

I could go on and on about the definition of a "compressor" versus "limiter" and how no-one seems to know how a real hardware compressor/limiter operates and how to link stereo channels and how Pro Tools got it right (see the RTAS "Limiter" effect in Pro Tools LE) ... but I will restrain myself.

For a really good (conventional) compressor, try the "SC4" which is available here: http://plugin.org.uk/ladspa-swh/docs/ladspa-swh.html

Yes, I've got that, and prefer it. Is it available for Windows, though? Not that it matters to me, but it would be nice if it could be included in all builds.

-- Bill

[steve]

Is it available for Windows, though?

Yes, it's in the "90 plug-in pack" (see http://audacityteam.org/download/plugins )