Hi all.
I need to graphically represent jitter effects over a voip RTP stream
(ethereal dump)
So I thought of histogram of the interarrival times.
Shall I calculate these interarrival times considering consecutive
"received" packets or the arrival times of consecutive "sent" packets (by
considering sequence numbers, so negative values of the interarrival delay
are possible in the case of out of order received packets).
All the interarrival histograms I see on the net starts from 0 so I have two
possibilities:
I must ignore sequence number (but this way the calculations are somehow
wrong in case of out of orders)
I must calculate interarrival times of consecutive sent packets but then
discarding negative values.
As you can see, I'm not considering the RTP/RTCP moving average formula of
jitter.
That interarrival histogram (almost normal or biexponential, centered on
the sending timing, but with lots of elements near the origin due to back to
back received packets and a long tail of very late packets) approximates a
probability density function (it should be normalized). What can I say about
% of packet discards, given a fixed playout delay (a fixed jitter buffer).
If I had the graph of one way delay I could say that the area to the right
of a certain value is the probability that a packet arrives too late (so i
get the % of discards) but I have only this interarrival distribution. What
simplifications/approximations can led us to some conclusion?
I hope someone can help
Thanks in advance
Rocco
