Design of Two-Dimensional Non-uniform Constellations with Better Performance

Modern communication systems adopt non-uniform constellations (NUCs) to approach Shannon limit. Generally, the design process of NUCs is based on maximizing bit-interleaved coded modulation (BICM) capacity for given channel conditions. However, the traditional Monte Carlo methods for computing BICM capacity cannot ensure the accuracy of NUC optimization with low complexity. This paper proposes a new design method including the derivation of a computable expression about BICM capacity and a new optimization scheme with simplified calculation of gradient in the algorithm. In this way, we obtain a set of 2D-NUCs with low complexity and better performance than those adopted by ATSC 3.0.