This site has been archived. For information on the GN Project’s eduPERT initiative please visit https://archive.geant.org/projects/gn3/geant/services/edupert/Pages/Home.html

Packetization Delay

Packetization is an important contributor to delay, in particular in packet voice applications such as Voice over IP. Thanks to advances in compression technology, audio streams can be digitally compressed to very low rates for voice applications. For example, the Internet Low Bit Rate Codec (iLBC) can generate payload bit rates of 13.333 kb/s. So close to an entire second of encoded speech would fit in a single 1500-byte packet! If it were done this way, then the packetization delay would be close to one second, as the sending system would accumulate this much digitized speech to fill a packet.

In order to keep packetization delay manageable, packet voice systems must use a minimum sampling rate; typical minimum sample sizes are 20-50ms (iLBC supports 20ms or 30ms). The compressed payload per sample is then very small - 50 bytes for iLBC at 30ms samples, 38 bytes at 20ms samples. Since the header overhead for RTP,UDP, and IP amounts to 40 bytes per packet, these small sample sizes cause significant waste. On low-speed links, one should use header compression to make this workable.

References

  • RFC 2508, Compressing IP/UDP/RTP Headers for Low-Speed Serial Links, S. Casner, V. Jacobson, February 1999
  • RFC 3545, Enhanced Compressed RTP (CRTP) for Links with High Delay, Packet Loss and Reordering, T. Koren, S. Casner, J. Geevarghese, B. Thompson, P. Ruddy, July 2003
  • RFC 3951, Internet Low Bit Rate Codec (iLBC), S. Andersen A. Duric, H. Astrom, R. Hagen, W. Kleijn, J. Linden, December 2004

– Main.SimonLeinen - 07 Apr 2006

  • No labels