Polar coded probabilistic amplitude shaping for non-terrestrial networks in 5G

The dominant Gaussian noise effect of the non-terrestrial network-to-user equipment (NTN-UE) channel presents an opportunity for polar coded probabilistic amplitude shaping (PAS) application. However, the 5G NTN serving a mobile device poses energy efficiency and short block-length support challenges. Given this, we propose a sphere-shaped polar coded PAS with path metric (PM) inheritance supported multi-level coding and multi-stage decoding (MLC-MSD) architecture. In the proposed design, a modified de-mapping and code construction approach addresses the inherent inaccuracy in Polar code construction for the most significant bit (MSB) level in the polar PAS scenario. It utilizes the widened inter-path PM gaps at MSB decoding to improve the average log-likelihood ratio (LLR) prediction as compared to the regular mutual information (MI) based calculation; this improves the Gaussian approximation (GA) design. In addition, we propose a sphere-class aided polar decoding that utilizes the sphere-shaped symbol sequence energy pattern during the CRC-aided successive cancellation list decoding (CASCL). The frame error rate (FER) results show up to 1.4 dB gain as compared to the regular polar PAS architecture. Furthermore, energy efficiency results show that the polar PAS architecture leads to a significant energy gap between the sphere shaping and the legacy distribution matching approach.