Constellation Quantization in Constrained Backhaul Downlink Network MIMO

Abstract

In this paper we consider a downlink multi-cell scenario where a central processor is connected by a finite-throughput backhaul to base stations (BSs) employing beamforming and QAM constellations. Both the serving BS and auxiliary BSs transmit a signal that combines at the mobile terminal (MT) to provide a QAM symbol while achieving a diversity gain. In order to reduce backhaul occupation, auxiliary BSs receive from the central processor only a quantized version of the QAM symbols to be forwarded to MTs. We formalize the problem of maximizing the network spectral efficiency on air for all the MTs within an area illuminated by the cooperative BSs optimizing a) the QAM constellation size, b) the quantization rate dictated by the finite-throughput backhaul and c) the power allocated by each BS. Since the resulting optimization is a mixed integer programming problem, we investigate a suboptimal solution where the same power is allocated to each MT and obtain a simple iterative algorithm for the rate optimization. Numerical results show that for typical cellular scenarios the suboptimal approach yields a network spectral efficiency close to the theoretical limit of Slepian-Wolf encoding.