[Scott Lowrey <slowrey@xxxxxxxxxxx>]

According to the original RFC, jitter should be measured by comparing 
the interarrival time at the receiver with the difference in sample 
times as measured by the sender.  The former is obtained by measuring 
the arrival time between packets at the receiver, the latter is the 
difference in timestamps within the packets.

Therefore, if done correctly, you only need to watch the receiving end 
of the stream.

I don't know if Ethereal *does* do this correctly, but there was a 
discussion a few months back on that subject:

http://www.ethereal.com/lists/ethereal-users/200411/msg00051.html


John Graves wrote:

> I understand what jitter is and it's effect on VoIP quality as well as 
> the working and configuration of jitter buffers.   My question is 
> simpler.  Ethereal is only looking at one side of a VoIP conversation. 
> I have been trying to understand what the jitter figure shown by 
> ethereal is and the nature of the calculation.  Then I can make a 
> reasonable determination on whether the measurement is meaningful in 
> the  configuration of jitter buffers or whether it is just a relative 
> figure more useful as an index.
>
> My original thought was that it was making a numeric comparison in the 
> packet to packet time variation, but it doesn't seem to be that simple.
>
> John G.
>
> Visser, Martin wrote:
>
> > Jitter is variation in delay. All networks, encoding  and decoding
> > processes add delay. If you only had delay (and no packet loss) you
> > would not need buffering as you would know always when the next voice
> > frame was going to arrive you could jjust play the voice sample. But
> > because of congestion, multiple network paths, and other variations
> > there will be always times when a voice frame arrives late. Because you
> > always want to have something play (not just silence) the designers of a
> > VoIP system always will have buffer at least as long maximum tolerable
> > delay variation or jitter, plus some breathing space. Unfortunately I
> > expect that jitter is subject to bell curves, etc where while average
> > jitter might by say 20ms, occasionaly you might get jitter of 50 ms and
> > very occasionally 100ms.  The VoIP designer needs to make a compromise
> > in choosing the tolerable jitter - between buffer size which increases
> > the minimum delay and voice sample loss.
> >
> > ----Original Message-----
> > From: ethereal-users-bounces@xxxxxxxxxxxx
> > [mailto:ethereal-users-bounces@xxxxxxxxxxxx] On Behalf Of John Graves
> > Sent: Tuesday, 12 April 2005 7:19 AM
> > To: Ethereal user support
> > Subject: Re: [Ethereal-users] Jitter Measurement
> >
> > Scott,
> >
> > If you highlight a RTP stream, and go Statistics > RTP > Stream Analysis
> >
> > It shows you both sides of an RTP conversation, and one of the
> > statistics is Jitter.  I was looking at that and going back to the
> > inter-packet times and trying to figure out how Jitter is measured.  
> > What I see in a lot of sessions is a growth and decline of Jitter across
> > the session, and usually one side is worse than the other.  I am trying
> > to understand if jitter is an important variable in diagnosing voice
> > that goes into the toilet or just another interesting set of numbers.
> > Is this an estimated amount of jitter or is this absolute?  Can I use
> > this as a sign that the voip traffic starts to degrade with only 10-15
> > ms of jitter.  Specs on the equipment say that it can tolerate up to 100
> > ms of jitter before the buffer blows.
> >
> >
> > John G.
> >
> > Scott Lowrey wrote:
> >
> >
> > > I don't know how Ethereal measures jitter.  (Does it? :)
> > >
> > > RFC 3550 explains the concept and the required formulas for RTP/RTCP.
> >
> >
> >
> > > In  a nutshell, jitter is a statistical measurement of the packet 
> > > inter-arrival time variation in a stream, and is expressed in units of
> >
> >
> >
> > > time.
> > >
> > > There are other definitions of jitter, most dealing with electronic 
> > > communications and circuitry.
> > >
> > > John Graves wrote:
> > >
> > >
> > > > Looking at the numbers for packets included in the RTP stream 
> > > > analysis, it is not clear to me how jitter is measured.  Can 
> > > > someone point me to an explanation of this or explain how it is 
> > > > determined?
> > >
> > > --
> > > *Scott Lowrey*
> > > Test Engineering Manager
> > > NexTone Communications <http://nextone.com> Gaithersburg, Maryland USA
> > >
> > > /1.240.912.1369/
> > >
> > > -----------------------------------------------------------------------
> > > -
> > >
> > > _______________________________________________
> > > Ethereal-users mailing list
> > > Ethereal-users@xxxxxxxxxxxx
> > > http://www.ethereal.com/mailman/listinfo/ethereal-users
> >
> >
> > --
> > John Graves
> >
> > Dynamic Devices Inc.       
> > 781-245-9100
> >
> > Your source for
> > Security, Wireless LANS & Wide Area Networking for converged Data, Voice
> > and Video
> >
> >
> > _______________________________________________
> > Ethereal-users mailing list
> > Ethereal-users@xxxxxxxxxxxx
> > http://www.ethereal.com/mailman/listinfo/ethereal-users
> >
> > _______________________________________________
> > Ethereal-users mailing list
> > Ethereal-users@xxxxxxxxxxxx
> > http://www.ethereal.com/mailman/listinfo/ethereal-users