Optimizing Geometric Constellations for Phase Noise Channels Using Deep Learning
Amir Hossein Omidi, Xun Guan, Ming Zeng, Leslie Ann Rusch · 2022 Photonics North (PN) · 2022
We optimize 16 and 64QAM constellations and detectors via end-to-end learning for residual laser phase noise. At the bit-error-rate (BER) of 1e-3, the required signal-to-noise ratio (SNR) is shown to reduce about 0.9 and 1 dB for 16 and 64-QAM, respectively.