In Multi User Multiple Input Single Output (MU-MISO) systems one centralized transmitter with multiple antennas serves several decentralized single-antenna receivers. Precoding is an attractive way to remove multiuser interferences because it reduces cost and power consumption in the user equipment. When implementing precoding, however, the base station should know the Channel State Information (CSI). In Frequency Division Duplex (FDD), this information is sent from the receivers by means of a feedback channel, whose data rate is often severely limited. In this paper we investigate the utilization of the Karhunen-Loeve (KL) transform to compress the CSI sent through the feedback channel. We model the errors caused by channel estimation, truncation and quantization of the KL coefficients and feedback delay. This error modeling is the basic premise to design robust Tomlinson Harashima Precoding (THP) based on a sum Mean Square Error (MSE) criterion. Our results show that robust THP clearly outperforms conventional THP transmitting a minimum amount of information from the receivers.
展开▼