Time domain statistical multiplexing (packet mode communication) is similar to time-division multiplexing (TDM), except that, rather than assigning a data stream to the same recurrent time slot in every TDM, each data stream is assigned time slots (of fixed length) or data frames (of variable lengths) that often appear to be scheduled in a randomized order, and experience varying delay (while the delay is fixed in TDM).
Statistical multiplexing allows the bandwidth to be divided arbitrarily among a variable number of channels (while the number of channels and the channel data rate are fixed in TDM).
Statistical multiplexing ensures that slots will not be wasted (whereas TDM can waste slots). The transmission capacity of the link will be shared by only those users who have packets.
Static TDM and other circuit switching is carried out at the physical layer in the OSI model and TCP/IP model, while statistical multiplexing is carried out at the data link layer and above.
In statistical multiplexing, each packet or frame contains a channel/data stream identification number, or (in the case of datagram communication) complete destination address information.
Examples of statistical multiplexing are:
In digital audio and video broadcasting, for example, a statistical multiplexer is a content aggregating device that allows broadcasters to provide the greatest number of audio or video services for a given bandwidth by sharing a pool of fixed bandwidth among multiple services or streams of varying bitrates. The multiplexer allocates to each service the bandwidth required for its real-time needs so that services with complex scenes receive more bandwidth than services with less complex ones. This bandwidth sharing technique produces the best video quality at the lowest possible aggregate bandwidth.