I seriously appreciate this info, all of the contributors above, thanks!

hi Sharke, just wondering/not sure if that tutorial is correct, or I'm confused/not understanding.
Untill now always thought it was as simple as..
Attack : time it takes for the compression to step in after the detected transient/peak, let's say beater hitting the kickdrum head being the peak.
(higher attack value let's the kickdrum through uncompressed for a longer time).
Release : amount of time the compression (and make up gain etc) is applied.
So higher Release value compresses the audio for a longer period of time, I was thinking always.
 
To my ears it's most important (and difficult) to get attack/release to fit or enhance the tempo/drive of the music, especially on a mix bus.
When I tweak them, attack/release slows the music down, or speeds it up to my ears.
 
 
 
 

If you want to get technical, Attack refers to how long it takes to achieve maximum attenuation. The actual attack phase begins as soon as the threshold is crossed. Some modern digital compressors and gates do have the equivalent of a reverb's predelay, which inserts a delay before the envelope generator kicks in, but that's not the norm.
 

It's worth reading Roey Izhaki's "Mixing Audio" (I picked up my copy dirt cheap as an e-book on sale) because he goes into great detail about how compressors actually work under the hood and you start to realize that the explanations given in most mixing tutorials are, while helpful for learning how to use compressors, aren't really the full story with regards to what's actually going on. It gets quite technical but it's very interesting.

AIUI, the attack and release times are the times it takes for the compressor to apply, or restore (respectively) approx. 2/3rds of the gain reduction.
 
i.e.
 
attack time is the time it takes for the compressor to apply 2/3rds of the targeted amount of gain reduction.
 
release time is the time is takes for the compressor to restore 2/3rds of the reduced gain to the compressed signal.
 
'2/3rds' may vary from compressor to compressor ofc, but it's a reasonable guideline.
 
Note that attack/release times are not the amount of time before the compressor takes action, neither are they the amount of time taken for the compressor to fully compress or release the signal.
 
All times are relative to the signal crossing the threshold.
 

plus, for the SOFTWARE compressors,
don't forget about linear phase..
and look ahead....
 
plus, every GOOD compressor has a little of the designer in it,
and sometimes the numbers are more like 'guidelines'
 
LOL
 
this is why i still like to use hardware compressors first.
 

The bit I'd like to understand more is what the compressor 'sees' as the signal.  You don't want it chasing individual peaks and troughs of the 'KHz audio', so presumably there's some kind of averaging or envelope-generating circuitry which drives the compress/release part?

That's exactly what's in there: an envelope generator, very much like the one on your synthesizer.