Automatic gain control and acoustic echo cancellation

This is a general question about acoustic echo cancellation based on
NLMS.

If an AEC receives a speaker signal which has been processed by an
automatic gain control algorithm, will that have any adverse effects
on the AEC's performance?

If the change in speaker gain level is too sudden or far from being
smooth, doesn't that result in the AEC having to re-adapt?

Thank you.

John McDermick wrote:
> This is a general question about acoustic echo cancellation based on
> NLMS.
> 
> If an AEC receives a speaker signal which has been processed by an
> automatic gain control algorithm, will that have any adverse effects
> on the AEC's performance?
> 
> If the change in speaker gain level is too sudden or far from being
> smooth, doesn't that result in the AEC having to re-adapt?

AGC inside EC loop makes the system nonlinear -> system model is 
inadequate -> EC doesn't work.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

>
>
>John McDermick wrote:
>> This is a general question about acoustic echo cancellation based on
>> NLMS.
>> 
>> If an AEC receives a speaker signal which has been processed by an
>> automatic gain control algorithm, will that have any adverse effects
>> on the AEC's performance?
>> 
>> If the change in speaker gain level is too sudden or far from being
>> smooth, doesn't that result in the AEC having to re-adapt?
>
>AGC inside EC loop makes the system nonlinear -> system model is 
>inadequate -> EC doesn't work.

If you can link the AGC to the EC you can make it all come right, but if
they work independently, you're screwed

Steve

If the speaker signal path looks like this:


----- speaker signal  ----> [AGC] ----- gain corrected speaker signal
----> [AEC] -----> sout

does that constitute the AGC being inside the EC loop ?  I would say
no, but I might be
missing something?

Steve, the AGC works independently of the AEC.

The speaker signal is processed by the AGC and the output of the AGC
is
used as the speaker input signal to the AEC.

So you're saying I have a problem then?