[Asterisk-Users] BAD/GOOD Echo Cancel

[David Yat Sin]

The echo canceller chip used in the Sangoma AFT "D" cards are fully
compliant with all the G168 specifications. 

David Yat Sin
Sangoma Technologies
(905) 474-1990 x119
(800) 388-2475 x119
Fax: (905) 474 9223
MSN: david.ys1 at hotmail.com
Email: davidy at sangoma.com
Website: www.sangoma.com
 
 
> >Date: Mon, 06 Feb 2006 22:09:34 +0800
> > From: Steve Underwood <steveu at coppice.org>
> > Subject: Re: [Asterisk-Users] BAD/GOOD Echo Cancel
> > To: Asterisk Users Mailing List - Non-Commercial Discussion
> > 	<asterisk-users at lists.digium.com>
> > Message-ID: <43E7589E.4070200 at coppice.org>
> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> >
> > James Harper wrote:
> >
> > >>virtually all software echo cancelers cannot get double echo removed
> > >>completly.  It can get the first one but not the second one.  There
> > >>
> > >>
> > >are
> > >
> > >
> > >>instances where you get a 2nd echo, so ...  Asterisk is no exception
> > >>from this afaik nothing software only based is.
> > >>
> > >>If you really want good echo cancelation a hardware solution is the
> > >>
> > >>
> > >way
> > >
> > >
> > >>to go.
> > >>
> > >>
> > >>
> > >
> > >Just an enquiring mind wanting to know, but how is a
> > hardware solution
> > >different to a software solution? The echo cancellers in the Digium
> > >hardware presumably just use the same sort of algorithms as
> > the software
> > >versions, so it is just that they are dedicated and perform
> > better, that
> > >they are closer to the source of the echo, or some other
> > thing that I've
> > >overlooked?
> > >
> > >
> > There isn't much difference, except for the amount of CPU
> > taken, and the
> > issue that software echo cancellation forces the device to use very
> > short buffers. He's talking rubbish. Hardware echo cancellation
> > certainly eases the timing constraints on the E1/T1 card to host
> > processor interface. A lot more buffering can occur if the
> > host does not
> > do echo cancelling. A 20ms buffer on a PCI card will
> > practically all the
> > quirky timing issues people see go away. However 20ms of
> > buffering would
> > badly hurt an echo canceller's convergence.
> >
> > Most hardware cancellers, are actually software cancellers.
> > The software
> > just runs in a DSP (often a customised one) instead of the host
> > processor. Some are a hybrid hardware/software design. Few are pure
> > hardware.
> >
> > There are no standard algorithms for echo cancellation, and
> > no standard
> > level of performance. Few cancellers which claim G.168 compliance
> > actually pass all the tests. If you look in the small print they
> > generally say which tests they do pass. Echo cancellers vary a lot in
> > performance, and making them truly robust and efficient is still a
> > research topic.
> >
> > Regards,
> > Steve
> > \
> >