Performance and complexity comparison of MRC and PASTd-based statistical beamforming and eigencombining

Smart antennas may enhance performance by applying conventional algorithms such as maximal-ratio combining (MRC) or maximum average signal-to-noise-ratio beamforming, i.e., statistical beamforming (BF). However, MRC and BF yield advantages that offset their complexity only for extreme antenna correlation values, which seldom occur for space-limited base-station antenna arrays deployed in typical urban (TU) scenarios, with predominantly-small, random, azimuth spread (AS). Therefore, the principles of BF and MRC have been integrated to forge maximal-ratio eigencombining (MREC), which promises to reap the available array and diversity gains more effectively. Nonetheless, the relative performance and numerical complexity of MREC, BF, and MRC have not yet been investigated for channel estimated from received-signal-vector samples in a TU uplink scenario with realistic Laplacian base-station power azimuth spectrum and log-normally distributed AS with exponential temporal correlation. Therefore, herein, Yangpsilas effective and low-complexity deflation-based projection approximation subspace tracking (PASTd) algorithm is deployed to recursively update the channel eigenstructure required for MREC adapted to AS using the classical bias-variance tradeoff criterion (BVTC). Simulation results indicate that BVTC-based MREC can significantly outperform BF for much lower complexity than MRC.