I have a test video file to you to use. It is located at <file://ftp.gcsd.harris.com/public>ftp.gcsd.harris.com/public, filename "live555_test_stream.ts". This is a blind FTP server, therefore it will only accept requests for exactly that filename. Also, I believe the policy is to remove the files nightly, so please let me know if you need me to upload the file again for your access. Hopefully you can reproduce what I've been seeing.

The problem with this file is that its PCR timestamps show that it is *extremely* VBR - far more than any reasonable stream should be. Because the streaming server uses these timestamps to figure out how to pace the outgoing stream (i.e., how much to delay after sending each network packet), extremely VBR timestamps messes up the streaming.

More specifically, here's what's happening with your test file:
As shown by its last timestamp, the stream is 37.78 seconds long, and contains 285866 188-byte Transport Packets. Therefore, on average, each Transport Packet should last about 132 ms. In practice, however, the stream's durations vary widely from this average.

E.g., here are the first few PCR timestamps
packet number   diff from last  PCR timestamp   diff from last  duration/packet
============    ==========      ============    ==========      =============
132 132 0.002257 0.002257 17 ms 197 65 0.002727 0.000470 7 ms 262 65 0.003187 0.000460 7 ms 264 2 0.024481 0.021294 10647 ms 302 38 0.024754 0.000273 7 ms 322 20 0.024896 0.000142 7 ms 326 4 0.048253 0.023357 5839 ms 454 128 0.049158 0.000905 7 ms

I hope you get the picture. The extremely large jumps in PCR timestamp are messing up our streaming server's estimate of how much time to delay after sending each network packet.

I don't know whether you really intended this stream to be so extremely VBR, but our server is currently not able to properly stream files like this.
--

Ross Finlayson
Live Networks, Inc.
http://www.live555.com/
_______________________________________________
live-devel mailing list
live-devel@lists.live555.com
http://lists.live555.com/mailman/listinfo/live-devel

Reply via email to