Cell loss ratio and multiplexing gain of an ATM multiplexer for VBR voice sources

Asynchronous transfer mode (ATM) is designed to support various traffic classes. It is necessary to select a proper traffic class to support a specific application. Voice traffic in its inherent nature is time-sensitive and variable. Therefore, it is natural to consider voice as real-time VBR traffic. Human speech has a voice activity factor of about 42% on an average, which makes it possible to multiplex a lot of voice sources together. The study of cell loss is of great importance to the design of an ATM multiplexer whose input consists of voice sources. With a maximum delay allowed in the network, it is the behavior of cell loss, rather than the variation of network delay, that determines the quality of reconstructed voice at receivers. This paper analyzes the cell loss performance and multiplexing gain of an ATM multiplexer loaded with VBR voice sources, where multiple voice streams, utilizing compression and silence detection, are multiplexed into a fixed bandwidth circuit such as T1 or T3 ATU trunk. This paper shows that the cell loss ratio and multiplexing gain depend on the voice activity factor, the mean talk spurt length, the mean silence period, the link capacity, the compression ratio, and the buffer size of the ATM multiplexer