The performance of existing moments-based non-data-aided (NDA) optical signal-to-noise ratio (OSNR) estimation approaches degrades greatly for coherent optical systems employing multilevel constellations. We propose a novel NDA OSNR estimation algorithm, which provides enhanced performance for such systems. The proposed algorithm utilizes the empirical cumulative distribution function of the signal's amplitude to extract the information on the noise variance. Analytical and extensive simulation results show the feasibility and advantages of the algorithm. For the studied systems employing multilevel constellations such as 8-quadrature amplitude modulation (QAM), 16-QAM, 32-QAM, and 64-QAM, the proposed algorithm attains the derived Cramer–Rao lower bound. Furthermore, it achieves a lower mean square error with significantly lower complexity when compared to the conventional moments-based NDA estimation approach. Moreover, the impact of fiber nonlinearity is investigated with a five-channel Nyquist wavelength division multiplexing system, and the proposed algorithm outperforms the moments-based counterpart.
Read full abstract