Hi,
I'm trying to do RTP Analysis with Ethereal 0.10.10 for a captured video stream session. I'm getting too high numbers in the jitter calculations. For example, I'm getting mean jitter values of 1771310.65 ms (which of course can not be true).
The following are some lines out of the analysis screen (some columns removed):
Packet Sequence Delta (ms) Jitter (ms)
27 0 0 0
29 1 230 14.38
32 2 80 562508.48
34 3 20 1089850.45
35 4 0 1584234.79
37 5 130 2047711.99
38 6 50 2482226.87
40 7 20 2889586.44
42 8 81 3271482.23
43 9 130 3629506.46
Even if the jitter calculation mistakenly assumes a clock rate of 1Hz (as suggested by comment below from this forum), the conversion from ticks to seconds still results in too high jitter value (e.g., 1771310.65/90000 = 19 secs, assuming video clock rate is 90KHz).
Note that even the Receiver Reports (RR) have very high jitter values. One sample RR gives interarrival jitter = 11333786 in ticks. When converted to seconds assuming 90KHz it gives 11333786/90000 = 125.xx secs.
Could somebody please offer an explanation for this?
Thanks,
Cesar
"# ubject: Re: [Ethereal-users] RTP Stream Analysis
# From: "Lars Ruoff" <Lars.Ruoff@xxxxxxxxxx>
# Date: Tue, 9 Nov 2004 16:52:06 +0100
The revision history of all this is something like this:
- somewhere before 0.10.3 jitter was calculated wrongly, which was fixed for
0.10.3.
From then on, jitter calculation used a default clock rate of 8000Hz for
all payload types.
- 0.10.7 incorporates a fix which defines the RTP clock rate values for
known static payload types, and which defaults to a clock rate of 1Hz for
unknown payload type (which is the case in your example capture)
Thus, the output of 0.10.3 should be conforming with what i said about the
jitter definition when assuming a default clock rate of 8000Hz, while the
output of 0.10.7 is conforming when assuming a clock rate of 1Hz (gives
monstreous jitter!).
If someone has another idea on how unknown payloads should be handled, let
us know.
Lars"