On the design of layered space-time systems for autocoding

Hochwald et al.(see IEEE Trans. Inform. Theory, Nov. 2001) have recognized that arbitrarily reliable communication is possible in multiantenna systems with coding over only a single coherence interval. In particular, they showed that reliable communication is possible for all rates R/spl les/C/sub a/ with code words that extend over a single coherence interval when the number of transmit antennas and coherence interval (n,T)/spl rarr//spl infin/. They coined the names "autocoding" for this phenomenon and "autocapacity" for C/sub a/. They also proposed a signalling scheme based on random unitary matrices that achieves a significant fraction of this capacity. The main limitation, however, is that currently no decoder of reasonable complexity is known for this signalling scheme. We investigate the application of space-time layering to autocoding. We show that properly constructed layered systems can achieve the autocapacity with a reasonable complexity receiver composed of minimum mean-square error (MMSE) decision feedback multiuser detectors and single user decoders. In addition to this asymptotic result, we propose a specific layering approach, the threaded space-time layering, that combines generalized bit interleaved space-time coded modulation, iterative signal processing and pilot symbol assisted channel estimation. We show that this approach is well suited for practical systems with limited numbers of transmit antennas and small coherence intervals. Finally, we report simulation results that demonstrate the ability of the threaded approach to achieve significant fractions of the autocapacity with a realizable receiver. The simulation results also indicate significant performance gains over the Cayley differential space-time signalling scheme in certain scenarios.