Feedback of quantized channel state information (CSI), called limited feedback, enables transmit beamforming in multiple-input-multiple-output (MIMO) wireless systems with a small amount of overhead. Due to its efficiency, beamforming with limited feedback has been adopted in several wireless communication standards. Prior work on limited feedback commonly adopts the block fading channel model where temporal correlation in wireless channels is neglected. This paper considers temporallycorrelated channels and designs single-user transmit beamforming with limited feedback. Analytical results concerning CSI feedback are derived by modeling quantized CSI as a first-order finite-state Markov chain. These results include the source bit rate generated by time-varying quantized CSI, the required bit rate for a CSI feedback channel, and the effect of feedback delay. In particular, based on the theory of Markov chain convergence rate, feedback delay is proved to reduce the throughput gain due to CSI feedback at least exponentially. Furthermore, an algorithm is proposed for CSI feedback compression in time. Combining the results in this work leads to a new method for designing limited feedback beamforming as demonstrated by a design example.
I. INTRODUCTIONFor a multiple-input-multiple-out (MIMO) communication system, transmit beamforming alleviates the negative effect of channel fading by exploiting spatial diversity, and thereby increases system throughput.Typically, transmit beamforming requires feedback of channel state information (CSI) from a receiver to a transmitter. Such CSI feedback can potentially incur excessive overhead due to the multiplicity of channel coefficients. This motivates designing highly-efficient CSI quantization algorithms based on communication measures such as information capacity. These algorithms for both beamforming and other MIMO techniques form an active field called limited feedback (see e.g. [1] and the references therein). This paper focuses on limited feedback beamforming systems over temporally correlated channels. It addresses a set of open issues including the information rate inherent in CSI, the required bit rate for a CSI feedback channel, the effect of feedback delay on throughput, and feedback CSI compression.