| Transmit channel side information (CSIT) is channel information available to the transmitter. In multiple-input multiple-output (MIMO) wireless, CSIT can significantly increase the transmission rate and enhance reliability. The time-varying wireless channel, however, often induces partial CSIT, posing challenges to signal design and system performance analysis.;This thesis focuses on exploiting partial CSIT in single-user MIMO wireless systems, assuming perfect channel knowledge at the receiver. The thesis approaches this problem in three steps: building a dynamic CSIT model, deriving the capacity gains with CSIT, and designing optimal precoding schemes to exploit the CSIT. The results are applicable to practical MIMO wireless systems.;Due to inherent delays in CSIT acquisition, CSIT modeling must account for channel temporal variation. A dynamic CSIT model is accordingly constructed. The CSIT consists of a channel estimate and its error covariance, built on an initial channel measurement, the delay, and the channel statistics. Depending on the channel temporal correlation, dynamic CSIT ranges from perfect channel estimate at zero delay to the channel mean and covariance as the delay grows.;Dynamic CSIT increases the capacity multiplicatively at low signal-to-noise ratios (SNRs) for all multi-input systems, and additively at high SNRs for systems with more transmit than receive antennas. The optimal input at low SNRs is typically single-mode beamforming, while at high SNRs, it may also drop modes, depending on the CSIT. Furthermore, a convex-optimization program is developed to find the MIMO capacity with dynamic CSIT. Using this program, a simple, analytical capacity lower-bound, based on the Jensen-optimal input, is shown to be tight in many cases.;Linear precoders can optimally exploit CSIT. A linear precoder functions as a multi-mode beamformer, spatially directing signal and allocating power based on the CSIT. A precoder is designed to exploit dynamic CSIT in systems with space-time block code. The design uses a dynamic water-filling algorithm, with evolving beam direction and power at each water-filling iteration. The precoder achieves significant and robust SNR gains. Another precoder is designed for the CSIT as channel amplitudes and phase-shift distribution. This CSIT simplifies precoding to single-mode beamforming, with per-antenna power allocation dependent on the phase-shift distribution. |
