Performance of Limited Feedback Strategies with Massive MIMO

Massive MIMO is considered an important technique for improving cellular system capacity through the use of high gain adaptive beamforming and high-order spatial multiplexing. For downlink Frequency-Division Duplex (FDD) cellular systems, limited feedback strategies can provide the information needed for computing the transmit weights for Single User or Multi-User MIMO. However, for MU-MIMO transmission accurate channel knowledge is critical for obtaining the best performance. In this paper, we explore the performance characteristics of several limited feedback schemes with 2D cross-polarized arrays for evolutions beyond the current 3GPP LTE Rel-12 standard. We characterize the performance of product codebook-based feedback along with several scalar quantization schemes, all schemes having different levels of feedback overhead requirements. This is compared against a high performing Rel-12 compatible vertical sectorization scheme at the low end and also against the performance with ideal knowledge of the downlink covariance matrix at the high-end. We consider a 64 element transmit array with 16 transceiver units as a case study on the effects of various restrictions on the feedback scheme, such as a DFT beam structure, block diagonal structure, and Kronecker azimuth/elevation structure.