Noncoherent trellis-coded quantization for massive MIMO limited feedback beamforming

Abstract

Abstract-Massive multiple-input multiple-output (MIMO) techniques are regarded as a potential breakthrough in cellular network design, with transmit precoding to multiple downlink users playing a central role in the projected capacity gains. Most prior work assumes that the channel estimates required for precoder design are obtained by exploiting reciprocity in time division duplexed (TDD) systems. However, most cellular networks today are frequency division duplexed (FDD), for which reciprocity based channel side information (CSI) is not available, hence explicit channel feedback from mobiles to base station is required for downlink precoding. Conventional "limited feedback" approaches, based on quantizing the channel coefficients based on a predefined vector quantization codebook, do not scale to massive MIMO because of computational complexity and lack of flexibility. We propose a novel noncoherent trelliscoded quantization (NTCQ) which uses the duality between noncoherent sequence detection and vector quantization for maximizing beamforming gain. The advantages of NTCQ include linear complexity in the number of transmit antennas, with vector quantization accomplished using an off-the-shelf Viterbi algorithm, and flexibility in adjusting the feedback overhead to support a variable number of transmit antennas without changing the structure of the quantizer.