Probability-Aware Stokes Space Blind Polarization Demultiplexing for Probabilistically Shaped Signals

A probability-aware blind polarization demultiplexing (PolDemux) algorithm using symbols of different radii in the Stokes space is proposed for polarization division multiplexed (PDM) coherent optical communication systems with probabilistically shaped (PS) signals. By leveraging the larger radius symbols which correspond to a non-trivial proportion of the PS signal, a more accurate estimation of the normal vector is obtained for PolDemux by least squares plane fitting with these higher signal-to-noise ratio (SNR) symbols. To properly discriminate symbols of different radii, the optimal
${s_0}$
boundaries are derived using the maximum
a posteriori
criterion. Monte-Carlo simulation of a PDM PS-64QAM system with the Maxwell-Boltzmann distribution under various constellation entropies (H) is conducted to evaluate the performance of three PolDemux methods based on the smallest, the second smallest, and the third smallest radius symbols, namely Pt
₁
, Pt
₂
, and Pt
₃
, respectively. The numerical results verify that the method with larger radius symbols outperforms the one with smaller radius symbols in terms of both the normal vector estimation accuracy and the required optical SNR at a targeted normalized generalized mutual information (NGMI), which is more significant for larger entropies as the proportion differences among Pt
₁
/Pt
₂
/Pt
₃
symbols increase. Higher tolerance to polarization mode dispersion and polarization dependent loss is also achieved by the PolDemux method with larger radius symbols. The fiber transmission results show that, compared with Pt
₁
, the achievable transmission distances at the NGMI of 0.9 are extended by 115--206 km and 62--151 km with Pt
₃
and Pt
₂
in systems with H of 4.33--5.17, respectively.