[SIPForum-discussion] Different types of G729 codec

[Yusuf Mayet]

On Wed, Aug 28, 2013 at 5:06 PM, Waqas Arshad <waqas.arshad000 at yahoo.com>wrote:

> Hi,
>
> Can someone please tell me the difference between following codecs in
> detail and their compatibility with each other?
>
> g729
> g729a
> g729b
> g729ab
> g729 annexa=yes
> g729 annexa=no
> g729 annexb=yes
> g729 annexb=no
>
> I'll highly appreciate the responses.
>
> Regards
> Waqas
>
>
> Hi,

I can refer you to document called "Everything you wish you didn't need to
know about VoIP" by Kristian Kielhofner.
An excerpt:

G729 and G729a are bitstream compatible. On the wire it should be
impossible to tell them apart and a G729 implementation should be able to
operate with a G729a implementation without issue. G729a was created as a
“reduced complexity” variant that creates an identical bitstream using less
CPU power at the expense of encoding (voice) quality. With that being said
there is one issue that seems to keep coming up...

Some implementations that support G729a like to put G729a in their rtpmap
with payload type 18:

a=rtpmap:18 G729a/8000

This would be fine if G729a were a different codec using a dynamic payload
type (96-127). Unfortunately that’s not the case here. In this case the
designers that implemented this SDP engine didn’t understand the specs.
This is broken. Cisco is the most well known vendor to make this mistake.
These implementations should be fixed to use G729 as the codec name with
payload type 18.

G729b is G729 with silence suppression support. In a nutshell silence
suppression stops transmitting RTP packets when the audio level drops below
a certain threshold. This saves on bandwidth but can cause audio clipping
and interop problems. G729b isn’t widely deployed but an SDP with G729
looks like this:

a=rtpmap:18 G729/8000.

a=fmtp:18 annexb=yes.

Notice the fmtp map line with the annexb=yes parameter?  That’s G729b.

Yusuf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sipforum.org/pipermail/discussion/attachments/20130830/6d138d58/attachment-0002.html>