Shared time-division duplexing (STDD): impact of runlengths of dropped packets and fast-speech activity detection

Shared time-division duplexing (STDD) is a proposed high-quality low-delay multiple-access technique for microcellular systems. By sharing time slots in both directions of transmission, it provides a considerable increase in capacity over time-division systems with speech activity detection (SAD) with relatively few users per carrier frequency. We investigate the temporal robustness of time-division multiple access (TDMA) with TDD/SAD and STDD. We specifically consider the statistics of the runlength of dropped packets for each scheme and show that although STDD is superior to TDMA/TDD/SAD when both systems are operating at full capacity, both systems suffer from long periods of dropped packets. We introduce circular interleaving as a method of reducing the average runlength of dropped packets. Fast SAD (FSAD) is also considered. We obtain a novel model that describes (FSAD) speech for a pair of users by merging two existing models: one for (slow) SAD on-off speech for a pair of users and one describing FSAD speech of a single user. It is concluded based on this model that FSAD results in a considerable increase in the capacity over systems with slow SAD as well as in shorter average runlengths of dropped packets. A number of numerical results are included, especially for the short frame length of 2 ms. We note that FSAD yields a further capacity gain of about 20% beyond that of STDD with slow SAD.