[Asterisk-Dev] Dev Meeting this week.

[Steve Kann]

Stephan A. Edelman wrote:

>I don't think just software AGC is going to do it without significant
>loss of "resolution". The peak amplitude that I'm seeing on the recorded
>WAV files that appear in my e-mail is probably about 2-4% of maximum.
>Software amplifying this will result in significant noise and distortion
>because the S/N ratio is so low. (This still may be better than not
>being able to hear it at all when listening to your voicemail over
>PSTN).
>
>I have no idea if the zap gain can be adjusted on the fly with the
>current driver implementation, or if it requires additional ioctl's or
>other fixes to make this work. 
>
>I think tying a relatively fast acting software AGC to the hardware gain
>of the zap interface would be ideal. 
>  
>
Yep.

I have this implemented, in iaxclient already, doing software (digital) 
AGC in a fast-acting manner, and hardware (analog) AGC, adjusting mixer 
gains, in a slower, and much coarser manner.

Doing this right, though, requires having a good VAD implementation, 
because it's important to not adjust things when there's silence.

It's not perfect, but it works pretty well. However, it's somewhat 
computationally expensive.

-SteveK