Gaussian blur aided blind phase search algorithm in low SNR QAM transmission system

Abstract

Probabilistic constellation shaping (PCS) reduces the required SNR compared with the uniform distribution system to achieve the same mutual information (MI). However, severe amplified spontaneous emission (ASE) noise degenerates the performance of the conventional blind phase search (BPS) algorithm, showing an unpleasant MI penalty and reducing the shaping gain. In this paper, we reveal that the MI impairment comes from the misjudgment of phase noise estimation by the conventional BPS algorithm interfered by severe ASE noise. To cope with this problem, we propose a Gaussian blur-aided BPS (GBA-BPS) algorithm by introducing an unequal weight filter with an efficient solution of variance configuration into the conventional BPS. Details about why the GBA-BPS algorithm outperforms the conventional BPS algorithm are elaborated based on the spectrum analysis of the phase noise estimated by both schemes. Evaluated on a 50-GBaud PCS-64QAM system, the proposed GBA-BPS algorithm with a 31-tap filter effectively mitigates the misjudgment of phase noise estimation, achieving a maximum MI gain of 0.5 and 0.25 bits/symbol compared with conventional BPS when the SNR is 16 dB and 18 dB respectively. The proposed GBA-BPS is a significant improvement over conventional BPS and it is more suitable for carrier phase recovery in low SNR systems.