Auto Release mode in compressor


I am trying to understand how “auto release” mode works in dynamic range compressors.
I tried to read about that in the proffesional books about DSP, but they use too difficult mathematical concepts for such noob like I am.
Maybe someone has some nice link with some basic explanation?

As I’ve already researched it relies on using two release mechanisms at the same time, but not sure how it works.

Probably that question is for some DSP forums, but I see here there are a lot of you who are interested in DSP. So maybe anyone could help. For any help thanks in advance.

In some articles I’ve seen opinions that it works in modern compressors because you can programm them to see forward and then set release time based on next transient. So you can use such compressors only for mixing, but not for recording because of big latency.

But actually maybe it’s true in some compressors. But for example in Fabfilter Pro-C2 there is no latency with auto release on.
Also as I know the famous gear Fairchild has auto release mode.

So I claim that for auto release mode there is smarter solution than look ahead. But I can’t figure it out.

I’ve just researched little bit more, and as I understand I need two “resistors” for release. One of them should be short (about 30 ms) and the second should be longer (for example about 700 ms).
But the question is should they be used parallel and then combining results of both of them, or should they be used serial (one by one) - first short release and next one long release?

Of course I will experiment with various variants, but maybe anyone could give some advice in that matter?

My understanding is that the short one is a smoothing factor to stop crazy loud, crazy short transients from causing the overall detection to duck massively:

Modern DSP compressors/limiters are quite different to their old analogue cousins. Lookahead (in any way that guarantees no transients going over the threshold) is something that you can only really do in the digital domain. OTOH, a lot of the charm and punchyness of compressors comes from transients leaking through before it’s had a chance to react.

Yes I read that article before, but the only problem for me is how to combine those two (or more) release resistors. Should they be one by one (serial), or parallel and then combine the results?

  1. if it should be serial, then which one should be first;
  2. or if parallel then how to combine the results? just avarage value or what?

It’s not just two resistors but two separate envelope detectors or at least a network with two capacitors with different discharge resistors.

There are different ways to set it up in code - one way would be the main detector’s input is the maximum of the audio signal level and the other slower detector’s level. So if the input signal goes silent the fast detector decays down to the slow detector’s level rather than to silence.