Power-Controlled Reverse Channel Training in a Multiuser TDD-MIMO Spatial Multiplexing System With CSIR.

This paper considers the design of a power-controlled reverse channel training (RCT) scheme for spatial multiplexing (SM)-based data transmission along the dominant modes of the channel in a time-division duplex (TDD) multiple-input and multiple-output (MIMO) system, when channel knowledge is available at the receiver. A channel-dependent power-controlled RCT scheme is proposed, using which the transmitter estimates the beamforming (BF) vectors required for the forward-link SM data transmission. Tight approximate expressions for 1) the mean square error (MSE) in the estimate of the BF vectors, and 2) a capacity lower bound (CLB) for an SM system, are derived and used to optimize the parameters of the training sequence. Moreover, an extension of the channel-dependent training scheme and the data rate analysis to a multiuser scenario with M user terminals is presented. For the single-mode BF system, a closed-form expression for an upper bound on the average sum data rate is derived, which is shown to scale as ((Lc - LB, \tau)/Lc) \log \log M asymptotically in M, where Lc and LB, \tau are the channel coherence time and training duration, respectively. The significant performance gain offered by the proposed training sequence over the conventional constant-power orthogonal RCT sequence is demonstrated using Monte Carlo simulations. [ABSTRACT FROM PUBLISHER]