Optimal Linear Precoders for MIMO Wireless Correlated Channels With Nonzero Mean in Space-Time Coded Systems

摘要

This paper proposes linear precoder designs exploiting statistical channel knowledge at the transmitter in a multiple-input multiple-output (MIMO) wireless system. The paper focuses on channel statistics, since obtaining real-time channel state information at the transmitter can be difficult due to channel dynamics. The considered channel statistics consist of the channel mean and transmit antenna correlation. The receiver is assumed to know the instantaneous channel precisely. The precoder operates along with a space-time block code (STBC) and aims to minimize the Chernoff bound on the pairwise error probability (PEP) between a pair of block codewords, averaged over channel fading statistics. Two PEP design criteria are studied--minimum distance and average distance. The optimal precoder with an orthogonal STBC is established, using a convex optimization framework. Different relaxations then extend the solution to systems with nonorthogonal STBCs. In both cases, the precoder is a function of both the channel mean and the transmit correlation. A linear precoder acts as a combination of a multimode beamformer and an input shaping matrix, matching each side to the channel and to the input signal structure, respectively. Both the optimal beam direction and the power of each mode, obtained via a dynamic water-filling process, depend on the signal-to-noise ratio (SNR). Asymptotic analyses of the results reveal that, for all STBCs, the precoder approaches a single-mode beamformer on the dominant right singular vector of the channel mean as the channel K factor increases. On the other hand, as the SNR increases, it approaches an equipower multiple-mode beamformer, matched to the eigenvectors of the transmit correlation. Design examples and numerical simulation results for both orthogonal and nonorthogonal STBC precoding solutions are provided, illustrating the precoding array gain.