A Feasibility Study of Using Store-and-Forward Data Communication Networks to Transmit Digitized Speech

This paper focuses on some issues related to transmission of packetized voice over store-and-forward data communication networks. We begin with discussing the general environment and availability of various hardware, e.g. voice digitizers, needed for this purpose. The issue of error checking of voice packets is discussed next. In contrast to general belief, we show that it is infeasible not to do error checking for voice packets in an environment where voice and data coexist. Furthermore, through mathematical modeling, we demonstrate that there is no significant advantage in providing special, less restrictive, error checking procedures for packetized voice. We then study the capacity and delay requirements in such networks. Through analytical modeling we draw some basic guidelines on how to design a data network to be able to use it for voice transmission as well. It is shown that even with low data rate vocoder type digitizers, high bandwidth channels (i.e. significantly higher than 50 kbps) are required in order to meet the delay requirements of voice communication. However, because of the present day high prices of the required hardware, providing packetized voice transmission is yet an uneconomical venture. In the last part of this report two procedures to provide guaranteed services are outlined. Guaranteed throughput/delay service is not only required for packetized voice in a voice/data network, but is a feature which is desirable to offer in most data networks.