Re: [Ccrtp-devel] Jitter calculations are making me jittery

[Federico Montesino Pouzols]

Hi,

you are right, an RTCP packet could break the jitter calculation
startup. I've commited a few fixes for this in CVS. As the RTCP
initial time seems to have no use, I've replaced the initialTime stuff
with analogous initialDataTime members and methods which are not
updated when RTCP packets are received. I've also patched the
lastPacketTime issue.

This needs some testing however. I'll try to do it shortly. Any
additional feedback will be gladly welcome.

On Tue, Sep 13, 2005 at 08:46:32PM -0700, Silverman, Greg wrote:
> I am a newbie, and, when looking at how ccRTP computes jitter I am
> confused about the way it computes transit time.The RFC calls for 
> 
> transitTime = arrival_time_in_ts_units-packet_timestamp
> 
> However, ccRTP, in IncomingDataQueue::recordReception, is computing
> arrival_time_in_ts_units as the difference between stack startup time
> and last packet time. This seems wrong because the first packet could be
> an RTCP and the first RTP packet could arrive many seconds later, so,
> the initial jitter value could be huge. Also,
> SyncSourceLink::lastPacketTime is initialized to (0,0) and never
> updated.
> 
> Greg Silverman
> NGC
> San Jose, CA

> _______________________________________________
> Ccrtp-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/ccrtp-devel