[asterisk-dev] Packetization
Test User
minal123 at turk.net
Tue Feb 28 05:51:10 MST 2006
I have an issue about packetization/frame length customization for the RTP stream generation.
I needed to tune up the hardcoded packetization value of Asterisk (which is 20ms). Consequently I applied the patch located at http://bugs.digium.com/view.php?id=5162.
Patch worked and I am now able to customize packetization value per user/peer basis. However,
a.. packetization parameter setting has no effect if used at the [default] part of sip.conf, am I correct?
I performed several tests with SIP clients from Grandstream. Please note the below;
a.. models are HT & GXP with the latest firmware releases
b.. Asterisk version is 1.2.4
c.. choice of codec is G.729
d.. configured frame length is 50ms at the clients and Asterisk
e.. clients register to Asterisk with reinvite=no option for RTP proxying.
Grandstream clients fail to send RTP stream with 50ms customized frame rate. Instead, they send with the default frame length which is 20ms. However, Asterisk sends with 50ms, as configured. As a result, RTP flows with asymmetric bandwidth values, that is; 37Kbps from the client and 20.5Kbps from the Asterisk.
I trunked two Asterisk systems using SIP and noticed that RTP flows perfectly, with 50ms/50ms frame lengths to/from any Asterisk.
a.. When I checked the signaling with the packet sniffer, I noticed that Grandstream implements ptime attribute in SDP and explicitly declares frame length in the INVITE msg. However, 200 OK msg from Asterisk does not include the confirmation of the ptime value, resulting in Grandstream's decision to default to 20ms for the frame length. Is this a logical explanation?
ptime attribute of SDP is defined in RFC2327 and is obviously an optional attribute. What do you think about implementing this option? Is it possible that interop problems may arise due to the above phenomenon?
Regards,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-dev/attachments/20060228/9eca10de/attachment.htm
More information about the asterisk-dev
mailing list