[Live-devel] RTSP and assembling audio buffer from audio frames

[Ross Finlayson]

>I'm not really sure which side is responsible for the problem... 
>More like that I need your advice what to do and investigate next.
>
>I use live555 to receive video and audio streams using RTSP client. 
>The problem is: when I fill audio buffer with audio-frames according 
>to their timestamps

I hope you mean "presentation times" rather than "timestamps".  (You 
should not be dealing with RTP timestamps directly, but instead 
"presentation times", which are derived automatically from RTP 
timestamps.)

>, I get a small gaps between frames in the buffer. That causes 
>unacceptable artefacts while playing. From the other side, if I 
>simply concatenate incoming audio-frames, I got clear audio without 
>artefacts. The problem is protocol-independent, I can hear the same 
>artefacts on both TCP and UDP protocols.

First, you should make absolutely sure that you're not seeing network 
packet loss.  You can check this by running "openRTSP" with the "-Q" 
option.  (A small amount of packet loss would explain the 
presentation time gaps, but might not be noticeable when you 
concatenate the frames that you end up receiving.)

Second, you haven't said anything about your server.  It is the 
server that determines the frames' presentation times.  I hope you 
don't see this problem with presentation times when you use *our* 
server implementation (either "testOnDemandRTSPServer" or 
"live555MediaServer").  What server are you using?
-- 

Ross Finlayson
Live Networks, Inc.
http://www.live555.com/