Statistical Multiplexing of Speech in Hybrid Integrated Networks

摘要

This report describes a technique for achieving TASI- or DSI-like operation in hybrid integrated networks. End-to-end voice circuits are established separately for individual talkspurts along routes that are determined at dial-up and fixed for the duration of a conversation. The packet handling capability of the hybrid system is used for high-priority internode data communication regarding the onset and termination of speaker activity. Subjects discussed include trade-offs between voice delay, cutout fraction and bandwidth efficiency that can be built into the nodal switching strategies. Analysis and simulation results are presented for cutout fraction versus voice delay in a talkspurt switching node with elastic buffering; the amount of buffer space needed as functions of TASI advantage and number of voice trunks per link; the probability of talkspurt buffering and average talkspurt delay. The report concludes with brief discussions of system vulnerabilities with respect to control errors and the compatibility of the concept with variable voice bit rate flow control schemes. (Author)