Optimization and Performance of Non-Uniform Rotated Constellations With Multi-RF Transmission Techniques

Non-Uniform Constellations (NUC) have been introduced in ATSC 3.0 (Advanced Television Systems Committee - Third Generation) as one of the main novelties to improve the performance compared to uniform Quadrature Amplitude Modulation (QAM) constellations. NUCs are optimized by means of signal geometrical shaping, considering the signal-to-noise ratio (SNR) and the channel model. ATSC 3.0 implements two types of NUC, depending on the number of real-valued dimensions in which they are optimized: 1D-NUC and 2D-NUC. However, the gain of NUCs becomes almost non-existent at high SNRs, especially when optimizing for fading channels. In that particular case, Rotated Constellations (RC) can be used to further improve the overall system performance. RCs may become especially effective when using multi-radio frequency (multi-RF) SNR averaging techniques such as Channel Bonding (CB) or Time-Frequency Slicing (TFS), where in-phase (I) and quadrature (Q) components are transmitted in different RF channels. 2D-NUCs can be rotated without increasing the demapping complexity, since a 2D-demapper is also needed. In this paper, we propose an optimization method designed for rotated 2D-NUCs, in which the rotation angle is considered as an additional variable, together with the symbol positions. The SNR gain obtained in fading channels is also provided for three different use cases: single-RF transmissions, CB with 2 RF channels as adopted in ATSC 3.0, and extension of multi-RF techniques to 4 RF channels.