Mismatched Decoding Metrics for Fiber Channels With Nonlinear Phase Noise

Aiming at reducing the complexity of the soft demapper for fiber channels dominated by nonlinear phase noise, we introduce the mismatched metrics $q^{\rm wg}$ and find optimal nonlinear correction factors for $q^{\rm wg} $ . The proposed mismatched metrics use simple arithmetic functions to replace the special functions and Fourier summations in the soft demapper. The numerical results show that the general mutual information (GMI) of $q^{\rm wg} $ can achieve the capacity of bit-interleaved coded modulation by using the good approximations to the probability density function of nonlinear fiber channels. Finally, we present a corrected metric $q^{\rm cr}$ to further improve the performance of $q^{\rm wg} $ in the middle nonlinear region. The numerical results show that the performance of $q^{\rm cr}$ outperforms $q^{\rm wg}$ and Gaussian channel metrics $q^{\rm gau}$ with pre- and post-nonlinear compensation.