Turbo-greedy coding for multiple antenna systems

Layered MIMO scheme with antenna grouping

“…Assuming residual interference plus noise, v i + z i , to be AWGN, we use in LTST decoder the following sub-optimum branch metric (8) where N 0 is the variance of the elements of the vector v i + z i .The variance of interference plus noise N 0 is needed in (8) for incorporation of the a priori information in the MAP decoding, and is computed as…”

“…To make up for this loss of performance, single antenna per layer schemes need to employ sophisticated co-antenna interference suppression techniques. We will show that while the complexity of most single antenna per layer signal separation schemes such as [7] and [8], is polynomial in n T , the signal separation in LTST is linear with n T .…”

“…Use of codes external to the STBC entail lower overall coding rates. Finally, a third approach is BLAST and its variations, where layered architectures with suboptimal suppression of the interference between layers are usually adopted [4]- [8], etc. BLAST schemes can achieve high data rates, but at the cost of rather complex processing for separation between layers.…”

“…Iterative co-antenna interference (interference between different transmit antennas) suppression schemes have been shown to improve the performance of BLAST schemes [6]- [8] compared to non-iterative interference rejection [4] [5]. The penalty is an increase in complexity.…”

“…Iterative MIMO schemes with good performance appearing in the literature often use either optimum maximum likelihood detection or near optimum spatial separation schemes such as based on minimum mean-square error (MMSE). For example, the turbo-greedy algorithm [8] achieves near capacity performance utilizing ML spatial processing. In another example [7], the decoder supplies soft estimates of the transmitted symbols for coantenna interference cancellation and for instantaneous MMSE residual interference suppression.…”