Stereo Tool

Hello,
The Time- constants of the AGC should be dependend from the "bigness" of the level jump.
If the Level drops suddenly about 3dB a very slow release time is ok to prevent the AGC from a steady "up and down".
But if the input level drops suddenly i.e. 10dB, then the AGC should rise the Level very fast to quickly compensate this to prevent a "low-level-whole".
It should be adjustable the transconductance of the release-speed curve as well as the maximum added speed to the standard (slower) value.

It does rise faster in that case, but proportional to the amount of change - and that can currently not be configured. I'm not too sure if this is a very good idea though - normally the rise is made SLOWER (gate!) to avoid extreme spikes after a short pause.

I could imagine that you mean something else instead: Let the AGC move (much) slower if there's little change in level?

Hello,
Thank you for your fast reply...
I´m from Germany therefore i have some little Problems with expression in english...
Of course the gating is necessary to prevent rising of "nothing" or extremly low level parts.
You are right, if the change in level is small the release time should be much slower than in bigger changes.
There you can allow more dynamics within a song and nevertheless you can be shure not to have such "wholes" because of the talent´s driving errors.
From Orban i´m very disappointed in this point, perhaps you will know it:
Their AGC has a definable window. You can define release times for in the window and out of the window.
The Idea is good, but they wanted to make soft change over between the 2 adjusted release times and have therefore inserted a constant delay before changing to the fast (ot of window) release time.
If you now have an element driven too high into processor followed by a song with low-level intro you have such a whole because it takes time for fastening up the AGC to the adjusted speed.
Regards,
Robert

Sorry for the slow response. And I didn't know that

This is definitely something I want to try in the (near) future. Just finishing the composite clipper etc. first, then when version 7.20 is ready I'll take a break (probably just a day or 2 before I get bored again...) and after that I'm going to look into lots of smaller issues. This is one I think I should include (and maybe also in the multiband compressor.........)

Bear in mind that the sooner you look into being able to move to Composer XE 2013 Update 3 (Intel C++ Compiler XE version 13.1.1 and IPP 7.1 Update 1), the less code you may need to modify to make it work. This assumes that something in how you are coding things is the source of the performance degradation. If that is the case, then the more you do, the more you may have to redo.

Just saw this thread and wanted to add that some AGCs even have a small configurable freeze window where they do nothing at all.
Adding a long look-ahead for non-realtime uses would also greatly improve the quality, making it possible to prevent unnecessary gain changes and hide the necessary ones better. And lastly improving the measurement by some loudness based algorithm (not just k-frequency weightings) would be great and the possibility to show the measured dB-values for broadcasting purposes. But for comparable results compliance to current standards is needed, of course.

Wow. Turns out that there was a bug in the AGC that caused the rise speed to be constant (depending on how deep the level has been dropped) instead of proportional to the amount of change.

This might be enough to do what you are asking, please check the next beta (732-003, probably tonight).