A quantitative comparative analysis of alternative implementation methods is presented with respect to network transmission capabilities, gateway design, and the presence or absence of silence detection algorithms. Performance measures presented include station, gateway, and network workloads. In presenting the results, the authors account for the effects of different network configurations and capabilities. In particular, they distinguish between multicast-capable and nonmulticast-capable networks. Results are given for conferences on a single network and across interconnected networks.>

A distributed control mechanism for managing a multi-party, packet switched voice conference connection has been previously presented [Weis88, Weis90, Zieg89]. The principal concept introduced was to view a conference connection as a logical ring of participants. Alternate methods for implementing voice conferencing on both multicast-capable and non-multicast-capable networks were discussed and analyzed. The mechanisms were shown applicable to both intranet and internet conferences. This paper extends those results to allow for multi-party, integrated voice and data conferences. The paper introduces the technique developed and then presents experimental benchmark performance measures with respect to station and gateway workloads.

Experimental implementations of packet voice communication systems over two types of PC-based local area networks are described. The first is a Proteon proNET token-passing ring network, and the second is an Ethernet network. System configuration, operation, and performance are described for both networks. Models of network performance for estimating the maximum allowable number of active voice stations without incurring intolerable packet loss are presented for each system. The models are defined for systems with and without silence detection. PC-related implementation issues are also discussed.>

A distributed control mechanism for managing a packet-switched voice conference connection is presented. The principal concept introduced is the idea of viewing a conference connection as a logical ring of participants. Alternative methods for implementing voice conferencing on both broadcast and point-to-point networks are introduced, analyzed, and compared. Tradeoffs between the two methods with respect to station workload and maximum number of conference participants are discussed. Experimental implementations on both a carrier-sense multi-access with collision detection (CSMA/CD) Ethernet and a token-ring ProNet are described. The mechanisms presented can be used as part of a packet-switched voice communications protocol that includes conferencing capabilities.>

The authors extend the logical ring control mechanism of C. Zielger et al. to the case of a multiparty conference across interconnected networks. In addition, the alternate implementation is expanded to allow for multiple-network connectivity. A comparative network between the alternate implementation methods with respect to station workload, gateway workload, network workload, and maximum number of conference participants is presented and discussed.>

This paper describes an implementation of a real-time, packetized voice communication system over a personal computer (PC) based token-passing ring local area network. System configuration and operation are presented. The system implements real-time, conversational transmission of voice packets between two stations on the network. An analysis is shown to calculate the maximum number of two way conversations that can be handled by the system without any reconstitution delay.

The paper concerns the study of waiting times at a packet-switched merger node for which the outgoing capacity exceeds or is equal to the sum of the capacities of the incoming channels. The case of constant service times, fixed-length packets, and general independent inputs is considered. A trivial upper bound on the waiting time is derived. Exact results on waiting time are obtained for certain specific network configurations.

In this paper, the performances of the Song Voice Adaptive Delta Modulator (SVADM) and the Continuously Variable Slope Delta Modulator (CVSD) in terms of dynamic range, sampling rate and the channel errors are compared. The use of the SVADM and the CVSD in a packet voice system, the algorithms for digital detection of silent periods and the performance of a packet voice system using the SVADM and the CVSD as source encoders are presented. The parameters employed for subjective evaluation of the packet voice system are packet size, silence detection algorithm, bit rate and packet loss rate.

Two networks consisting of single server queues, each with a constant service time, are considered. The external inputs to each network are assumed to follow some general probability distribution. Several interesting equivaleneies that exist between the two networks considered are derived. This leads to the introduction of an important concept in delay decomposition. It is shown that the waiting time experienced by a customer can be decomposed into two basic components called self delay and interference delay.