Feedback compressor

[nck]

Hi all,
i'm trying to do a simple feedback compressor:



It's pretty basic, logically it must work, but it doesn't: if i check audio with scope after envelope follower it goes always 1, if i bypass the envelope follower it works acting as a clipper(ehm delayed phase distorted clipper considering the 10ms delay of the feedback module and the characteristic of the signal thru the RMS module).

the problem seems to be on Envelope follower: it don't work on feedback'd signals.
I tried to put feedback module also after the VCA but no things changed.

Any advice? there's a bug in sonicbirth? (there's a bug in my brain?)




nck

EDIT:

well, i solved. Putting a cleaner BEFORE envelope follower. It works now, but it's a workaround
now the question is again: there's a bug or i'm wrong with something?

[zmix]

Won't work, nck. The Feedback element adds 10ms delay, so you will end up with a displaced dynamic envelope..

[nck]

thanks:)
but my problem isn't that it sounds bad(10 ms of delayed envelope is not useful for peak limiting, i know ), my problem is the NAN message on the number display, and it doesn't seems to make any sense!
a peak follower module should be a modified IIR filter, and it should not be a problem if i put it in a feedback chain

[dscro]

My experience with the envelope follower was that . . . it's better to build your own.

I had luck with things like:

squaring the input signal,
low pass filtering,
adding a small constant to keep the signal above zero
taking the square root

I think I also found that using a natural log function at some point in this smoothed things out.

Thing is, it's tough to implement attack and release this way. Instead I opted for changing the cutoff of the low pass filter or changing the mathematical operation on the filtered signal (ie natural log).