Partial Channel State Information Unitary Precoding and Codebook Design for MIMO Broadcast Systems

In downlink multiuser systems, multiple-input multiple-output (MIMO) technology can be leveraged to increase capacity in fading channels. Knowledge of channel state information at the transmitter (CSIT) is important to obtaining this capacity gain. It is unrealistic to obtain perfect CSIT, but partial CSIT can often be made available by using some practical feedback method. Based on partial CSIT, codebook techniques have been shown to improve performance of MIMO multiuser systems. Unfortunately, the codebook design problem has not been fully investigated. In this paper, we consider a multiuser multiple antenna downlink model with unitary precoding using a codebook consisting of 2B unitary matrices. The appropriate precoding matrix is chosen from this codebook, which is known to both the basestation and the receivers. By defining and using a carefully chosen distance function, the minimum distance of the codebook is shown to relate to the average sum-rate of the unitary precoded system. From our analysis, a codebook design criterion is developed based on maximizing the minimum distance of the codebook. Then, a framework for constructing the codebook is presented to maximize this minimum distance. Monte Carlo simulation results are given to show the average sum-rate capacity performance of the proposed codebook design.