[Asterisk-Users] Why echo occurs

[Andrew Kohlsmith]

On February 10, 2005 08:57 pm, Eric Bishop wrote:
> Can someone give me a simple rational explanation why a $5 analog
> handset  gives me no echo whatsoever on an analog PSTN line, but
> PSTN-VoIP devices such as the TDM400 and Sipuras do and thus require
> software-based echo cancellation. Surely a $5 analog handset does not
> have an "echo canceller".

The $5 handset isn't introducing signficant delay into the audio stream.
The $5 handset has just as much echo as a TDM400/Sipura or T100P, you just 
don't hear it as echo because you're hearing it "at the right time" and all 
you percieve it as is a sidetone.

> I have heard it said it is because of some slowdown of the signal. But
> where is this mysterious bottleneck?

It's in the digitization of the voice frames.  It's additionally in the 
transformation between codecs.  It's additionally in the time it takes to 
perform these steps and move the intermediary data around in memory multiple 
times.  It's additionally in the time it takes to get the data out the 
network card and across the internet and finally, it's additonally in all 
these reverse steps to get the data back out to a POTS interface like your 
friend's TDM400P.  :-)

> 1. It is not in the Asterisk box because IP to IP calls do not suffer
> this malady

IP to IP calls don't have a hybrid circuit to introduce reflected voice 
energy, which is the basic source of the echo problem.  The $5 handset has 
the same problem, but without the delay, all you hear is a comfortable 
"sidetone" (your own voice in the earpiece).

> 2. It is not from the Central Office to my premesis because my $5
> analogue handset works without echo. Also PRI ISDN works without echo.

Actually no it doesn't.  I have significant echo problems on my ISDN PRI, as 
do many others.  Those without ISDN PRI echo problems have good echo 
cancellation hardware sitting on the physical T1.

This isn't an asterisk-specific problem.  Every single piece of VOIP (or 
cellular, for that matter) hardware that interfaces to the PSTN has to deal 
with this in one form or another.  Some methods are just better than others.  
Googling will reveal a LOT of research into this problem.  It's by no means 
trivial.

-A.