[Ekiga-list] G729 codec

PawelCarqowski paulino90 at tenbit.pl
Sun Nov 4 15:51:16 UTC 2007


Jure,
comments in text
Jure Petrovic writes:
 > 
 > 
 > I thought here might be the problem because I noticed differences in RTP
 > payload sizes between received and sent packets when using Ethereal.
 > 
 > However, I never had problems about not being heard by the other party.
 > Does the other party hear anything at all? For example squeaky noise or
 > just silence? 

this is squiky noise.

 > 
 > I would also like to know if you have a balanced connection when using
 > this codec.. i.e. upload_rate = download_rate 
 > 
 > Note that the sizes, you are showing in the table seem to me as
 > unencoded frames. g729_codec.c takes unencoded frame PCM-16 (160 bytes)
 > and makes a encoded frame G729 (10 bytes). So your outgoing RTP packet
 > should have payload size 10 or 20 bytes.

these values are differences in "timestamp" field of adjacent RTP packets.

 > 
 > Regards, 
 > Jure


Here are the results:


1 - sjphone.correct.call.tcp
2 - myg729.ver0.1.tcp
3 - myg729.ver0.2.tcp


Trace	outbound	inbound		outbound	inbound		outbound	inbound
	RTP TS		RTP TS		RTP Payload	RTP Payload	RTP rate	RTP rate
	difference	difference	size		size		(kbps)		(kbps)
================================================================================================
1	160		480		20		60		8,083		8,083
2	160		480		20		60		7,969		8,621
3	80		480		10		60		7,963		8,159

My voip provider probably uses silence suppression, because, inbound
RTP packets do not flow constantly. Sometimes they appear, and then
dissappear. They always disappear at RTP packet that has "comfort
noise" payload type. I have silence suppression off in my ekiga, so
flow from ekiga goes without the breaks. So for computing inbound RTP
rate I took only some part of my trace. For trace 2 (myg729 ver0.1),
it is range of 19 inbound RTP packets. For trace 3 (myg729 ver0.2), it
is range of 17 inbound RTP packets. For trace 1 (sjphone trace), it is
range of 48 inbound RTP packets. For oubound rate computing I took the
whole trace (so this is more acurate).

have you got any ideas?

Regards,
Pawel



More information about the ekiga-list mailing list