I'm still trying to calculate the bandwidth. I have a problem with the video
bandwidth values. The audio bandwidth is correct. G711a requires 64 Kbps Bit
rate and 87.2 Kbps Ethernet bandwidth (27% overhead) as reported here http://www.voip-info.org/wiki/view/Bandwidth+consumption.
For video I use H263p and H264. The picture format is CIF with a fps = 1 and
profile 4:2:0 so the total video bandwidth without compression is about
355*288*8*1 + 176*144*8*1*(2) = 1.25 Mbits.
I know that H264 codec can reduce the video bandwidth up to 50-60 kbps and
70-80 kbps Ethernet bandwidth. H263p is wore than H264 about 2 times so
requires 140-160 kbps Ethernet bandwidth.
I have analyzed with Wireshark 1.2.6 RTP->stream analysis the RTP flow. I
have taken in account the six column IP BW with Matlab and I have extract the
PDF function. The mean and max bandwidth values are very high respect how I'm
expecting. For H263p 343 kbps and 555 kbps...
I’m not sure what
you mean by profile 4:2:0
ITU H.263
X.4
Levels of performance capability
:
3) Level 30 − Support of CIF, QCIF and sub-QCIF
resolution decoding, capable of operation
with a bit rate
up to 6·(64 000) = 384 000 bits per second with a picture decoding rate up to
(30 000)/1001
pictures per second.
4) Level 40 − Support of CIF, QCIF and sub-QCIF
resolution decoding, capable of operation
with a bit rate
up to 32·(64 000) = 2 048 000 bits per second with a picture decoding rate up
to (30 000)/1001 pictures per second.
In the video CSV files of Wireshark I see a lot of incorrect timestamps while
in the audio CSV file there are not of these values.
See my previous answer:
The RTP Analysis is
designed with voice in mind, time stamps and marker bit is used differently in
Voice and video I think.
For video the marker bit is set for the last packet belonging to the same
Picture and the timestamp
will be the same for all packets making up one picture (if memory serves).