[Asterisk-Users] Bounty! For help with echo cancellation code.

[Rich Adamson]

> echo-asterisk at secondphone.com wrote:
> >>From the CLI and during a call I want to be able to:
> > 
> >   *** Pulse the outgoing line and record at least 50 ms of the incoming line.
> > 
> >       The pulse waveform must be specifiable as a series of amplitudes
> >       for each 1/8000 sec time slot.  It would be best of these values
> >       could be read from a file specified on the CLI command line.
> > 
> >       Timing should be synced between the pulse and the echo so that the
> >       delay from the pulse to the echo can be accurately determined.
> > 
> >       Echo cancellation should be disabled during this operation.
> > 
> >       This would operate similar to the echo-training code that operates
> >       at the initiation of a call except that this could be done at
> >       any time.
> > 
> >       The initial pulse and any echoes can be combined and saved in a
> >       single channel.
> > 
> >       Output should go to a file and should be in a simple format that
> >       a program such as Audacity can read, display and play. 
> >        
> > 
> >   *** Pulse the outgoing line and record at least 50 ms of the incoming line.
> > 
> >       Same as above EXCEPT echo cancellation would not be disabled during
> >       this test and the results of the echo cancellation operations should
> >       be recorded and saved in a separate channel.
> >       
> > 
> >   *** Change variables used to control echo cancellation.
> > 
> >       Only the code in mec2.h is of interest.
> >    
> >       I will help identify the variables and modify the mec2.h code as
> >       needed to accomplish this goal.
> > 
> >       There are a lot of parameters in mec2.h that may affect the quality
> >       of the echo cancellation.  I want to be able to adjust them 'on the
> >       fly' and be able to immediately hear the results.
> > 
> > 
> > I am open to alternative proposals which would accomplish the same goals.
> > 
> > Name your price.
> 
> How about being able to "see" the results real time?
> I use a package called SMAART from siasoft.com.
> It is a dual channel spectrum analyzer.
> Run the output line as your reference channel
> and the input line as your measurement channel.
> 
> You can get great info from the impulse response
> and transfer function.
> 
> You could also use this to compare different codecs.
> The impulse function will tell you how long it takes.
> The transfer function will tell you just how good a
> job it did at reconstruction the original audio.

At least one of us that have worked a fair amount with the echo
problem tends to believe the issue is "system" related as opposed
to pstn line issues. Off-list, we found that swapping motherboards
"does" have a very noticable impact, and processor speed does
not appear to be a consideration. (Kind of thinking the echo
(or feedback loop) is actually internal to the system.)

Would the SMAART package help if this is the case?

Rich