Packet dropping analysis for reservation voice TDMA systems

We analyze the packet dropping performance in an ideal reservation time-division multiple-access (TDMA) multiplexing voice system with the focus on the probability distribution of the number of packets dropped in a particular talkspurt. Our analysis method is based on the discrete three-state Markov speech model which corresponds to the voice source equipped with a fast speech activity detector (SAD). The numerical results for a system of 14 slots/frame reveal that the probability of losing several packets in a talkspurt is much higher than expected (by geometric distribution) and thus not negligible.