<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
shadowym wrote:
<blockquote cite="mid00c001c75053$a29d1660$0301a8c0@pma" type="cite">
<pre wrap="">I gotta take issue with your comments that a HWEC is just software running
on a DSP. In the case of Octasic, it's an ASIC. How it does EC is VERY
different because.....it's done completely in hardware, not firmware loaded
into memory and run on a specialized CPU! Yes, the ASIC does contain an DSP
but it is customized for EC. You cannot think of it as a CPU.
</pre>
</blockquote>
You've obviously missed the fact that the Octasic chips have loadable
firmware.<br>
<br>
Without being privy to any internal Octasic information, all I can
guess is that their ASIC is a customised DSP core with perhaps more
on-chip fast memory for FIR coefficient storage, and perhaps custom
instructions/custom logic blocks specifically designed to improve the
performance for the type of mathematical operations required for echo
cancellation.<br>
<br>
Nobody in their right mind is going to do this entirely in custom
circuitry - if you find a bug in your algorithm, or a way to improve
things, what then - spend a few million on getting your chip rebatched
and tell your users get their soldering irons out?<br>
<br>
Anyway, I'm going to shut up now before I get carried away and start a
flame war :-)<br>
<br>
Cheers,<br>
Nic.<br>
<blockquote cite="mid00c001c75053$a29d1660$0301a8c0@pma" type="cite">
<pre wrap="">
-----Original Message-----
From: Nic Bellamy [<a class="moz-txt-link-freetext" href="mailto:nicb-lists@vadacom.co.nz">mailto:nicb-lists@vadacom.co.nz</a>]
Sent: Tuesday, February 13, 2007 5:43 PM
To: Asterisk Users Mailing List - Non-Commercial Discussion
Subject: Re: [asterisk-users] The High Performance Echo Canceller (HPEC)
shadowym wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Interesting,
Is this just a more advanced software echo canceller or software with
hardware hooks or software with hardware assisted processing?
</pre>
</blockquote>
<pre wrap=""><!---->A more advanced software canceller (there's no magical thing that makes
"hardware" echo cancellers better, it's still software, but it's running on
a DSP so it has more grunt available to it).
It's licensed from Adaptive Digital Technologies - G.168 compliant, and
supports up to 1024 taps (128ms) of tail coverage. Comes as a binary blob,
but such is life.
</pre>
<blockquote type="cite">
<pre wrap="">How would it compare to a true hardware echo canceller like the one
Sangoma uses. Besides the extra CPU cycles required.
</pre>
</blockquote>
<pre wrap=""><!---->Quite comparable - not sure if Octasic (as used by Sangoma and the latest
Digium cards) or ADT would win in a shootout, but they're both in the same
quality class.
The main issue is going to be CPU usage - getting this going at 1024 taps on
a full T1/E1 span would likely require two fast CPUs with the interrupts
distributed evenly between them... and even then, *shrug*
Cheers,
Nic.
</pre>
<blockquote type="cite">
<pre wrap="">-----Original Message-----
From: Nic Bellamy [<a class="moz-txt-link-freetext" href="mailto:nicb-lists@vadacom.co.nz">mailto:nicb-lists@vadacom.co.nz</a>]
Sent: Tuesday, February 13, 2007 12:41 PM
To: Asterisk Users Mailing List - Non-Commercial Discussion
Subject: Re: [asterisk-users] The High Performance Echo Canceller
(HPEC)
Larry Shields wrote:
</pre>
<blockquote type="cite">
<pre wrap="">I recently read about the following new technologies from Digium.
Has anyone tried the new HPEC or knows when it will be available?
</pre>
</blockquote>
<pre wrap="">It's out now, and I've tried it - the difference between HPEC and MG2
from trunk is stunning - in situations with bad echo where MG2 can
take ten or more seconds to converge to a reasonable degree, HPEC does
it in perhaps 300ms - converging on my intake of breath before I say
"hello", and absolutely no echo after that unless I purposefully go
out of my way to screw it up (whistling/blowing into the handpiece for
instance - even then, the malfunction is minimal).
You can now buy it from the Digium website (US$10 per channel), or if
you have an in-warranty Digium card, email through the serial numbers
to Digium support and they'll give you a key (this is what I did).
You'll need Zaptel 1.2.13 to make it go.
It does take quite a bit of CPU though - perhaps 70% more compared to
MG2-trunk for the same number of taps from my rough measurements.
Cheers,
Nic.
--
Nic Bellamy,
Head Of Engineering, Vadacom Ltd - <a class="moz-txt-link-freetext" href="http://www.vadacom.co.nz/">http://www.vadacom.co.nz/</a>
_______________________________________________
--Bandwidth and Colocation provided by Easynews.com --
asterisk-users mailing list
To UNSUBSCRIBE or update options visit:
<a class="moz-txt-link-freetext" href="http://lists.digium.com/mailman/listinfo/asterisk-users">http://lists.digium.com/mailman/listinfo/asterisk-users</a>
</pre>
</blockquote>
<pre wrap=""><!---->
--
Nic Bellamy,
Head Of Engineering, Vadacom Ltd - <a class="moz-txt-link-freetext" href="http://www.vadacom.co.nz/">http://www.vadacom.co.nz/</a>
_______________________________________________
--Bandwidth and Colocation provided by Easynews.com --
asterisk-users mailing list
To UNSUBSCRIBE or update options visit:
<a class="moz-txt-link-freetext" href="http://lists.digium.com/mailman/listinfo/asterisk-users">http://lists.digium.com/mailman/listinfo/asterisk-users</a>
</pre>
</blockquote>
<br>
<br>
<pre class="moz-signature" cols="72">--
Nic Bellamy,
Head Of Engineering, Vadacom Ltd - <a class="moz-txt-link-freetext" href="http://www.vadacom.co.nz/">http://www.vadacom.co.nz/</a>
</pre>
</body>
</html>