Re: [Linphone-developers] Voice quality on 3G and EDGE

[Simon Morlat]

Hi Matt,

Thank you for the clarification.
After reviewing the code, I see there is a risk that duplicates are not
seen by oRTP when they are not entered in the jitter buffer together.
That is if a packet goes out from the jitter buffer because it's time to
play it, and just after the same one enters the jitter buffer, oRTP
won't see it is a duplicate.
It only sees when while inserting a packet into the jitter buffer, a
similar one is already present.
So you're right, some duplicates are not discarded.

Simon


Le jeudi 09 septembre 2010 à 13:37 -0600, Matt Gilg a écrit :
> Hi Simon, thanks for the info.
> 
> 
> I suppose my emphasis on the terms was less than ideal, perhaps plain
> wrong.  The entire contents of the packets -- including the ethernet,
> udp, and rtp headers (seq, timestamp, etc) -- were copied.  So, along
> with a similar timestamp, we ended up with the same sequence numbers.
> (an entirely identical packet in fact)
> 
> 
> oRTP didn't drop the duplicates like I expected.
> 
> -Matt
> 
> On Thu, Sep 9, 2010 at 1:09 PM, Simon Morlat
> <address@hidden> wrote:
>         Le mercredi 08 septembre 2010 à 08:43 -0600, Matt Gilg a
>         écrit :
>         > Has anyone tested the auto-elimination of duplicates?
>         >
>         >
>         > I've recently done a test that wraps and forwards RTP
>         packets (vpn
>         > style) as well as sent direct, and I found that the
>         duplicate
>         > detection wasn't working.  Care was taken to be sure that
>         identical
>         > packets with the same timestamp showed up on the same
>         session on the
>         > receiving end.
>         
>         "same timestamp": yes there is no reason to discard packets
>         that have
>         same timestamp. This is very common with video packets to have
>         identical
>         timestamps for packets belonging to the same frame.
>         Only sequence number matters to know whether a packet is a
>         duplicate.
>         
>         Obviously if oRTP doesn't drop packets with similar sequence
>         number that
>         is a bug.
>         
>         Simon
>         
>         
>         >  The audio sounded, well, like the stack was processing
>         duplicate
>         > packets.  Audio sounded a bit like a CD skipping.
>         >
>         >
>         > Anyhow, for what it is worth, oRTP used together with
>         Mediastreamer2
>         > didn't seem to drop duplicates like I expected.
>         >
>         >
>         > --Matt
>         
>         
>         > _______________________________________________
>         > Linphone-developers mailing list
>         > address@hidden
>         > http://lists.nongnu.org/mailman/listinfo/linphone-developers
>         
>         
>         
>         _______________________________________________
>         
> _______________________________________________
> Linphone-developers mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/linphone-developers