LAUSR

To obtain the beamforming gain of a frequency-division duplex massive multiple-input multiple-output system, extensive channel state information (CSI) should be accurately quantized and fed back to the base station. Trellis-coded quantization (TCQ) can reduce codebook size and codebook-searching complexity that are the main drawbacks of vector-quantization-based CSI feedback schemes. However, the beamforming gain of the conventional TCQ cannot be fully obtained in a spatially correlated channel that is inevitably involved in massive transmit arrays. In this letter, we propose a differential TCQ that exploits the channel correlation by using a compressive constellation and phase difference quantization. We present an algorithm for generating a phase difference constellation and a decoding process for the proposed differential TCQ. The proposed scheme can reduce the overhead of the codebook size problem by trellis decoding and improve the quantization performance at the same feedback rate compared to the conventional TCQ scheme.