Two-Layer J-Best Selection / Maximal-Ratio Combining in Rayleigh Fading

Two-layer selection / maximal-ratio combining (2L-S/MRC) at the receiver consists in dividing a large array into a number of subsets, the best antenna in the SNR sense being selected within each subset. The set of selected antennas is then combined in the second layer using maximal-ratio combining (MRC). This is closely related to the well-known generalized selection combining (GSC), except that the latter involves selecting the best L-c antennas out of an L-antenna array, and then applying MRC to those. In theory and all other parameters being equal, GSC should outperform 2L-S/MRC since the L-c selected antennas constitute the optimal set. However, it has been shown previously that constraint on selection imposed by the subsets in 2L-S/MRC has minimal impact on ideal performance and, more importantly, this constraint allows the switching circuit at the RF stage to be much simpler and thus to exhibit minimal insertion loss. Thus, in any practical system where the selection process is leveraged to reduce the number of RF chains and low-noise amplifiers (LNA) with respect to the array size L, 2L-S/MRC will significantly outperform an equivalent GSC receiver once insertion loss at the RF stage is factored in. In this paper, a new two-layer scheme is proposed where J-best selection is performed in each subset, e.g. selecting the best J antennas instead of just one (equivalent to the first step of GSC), prior to the MRC stage. This will improve performance, while maintaining much of the RF-stage complexity advantage, provided that the subset size is small. Performance of the proposed scheme is analyzed in uncorrelated Rayleigh fading.