CSI Compression and Feedback for Network MIMO

We consider the downlink of a multi-cell system where base stations and users are equipped with multiple antennas. Therein, a cluster of base stations employ joint transmission with linear zero-forcing beamforming based on compressed channel state information (CSI) feedback. For propagation we assume a realistic 3D channel model including antenna patterns. We discuss in detail a novel compression algorithm exploiting the spatial and frequency domain and also present a feedback data format. We evaluate the proposed compression algorithm in the presence of channel estimation error in terms of throughput and feedback rate and compare it with the perfect CSI case. The proposed algorithm and feedback scheme is shown to reduce the feedback rate to less than 1 % of uncompressed feedback and at the same time increasing the sum throughput significantly.