We study the efficiency and numerical accuracy of two digital backpropagation schemes for post-compensating SOA-induced nonlinear impairments in the context of coherent receivers for advanced modulated formats. While the classical Runge–Kutta numerical techniques provide almost ideal post-compensation when the receiver sampling time tends to zero, this accuracy diminishes quickly as we approach realistic sampling times. At rates near Nyquist, despite much reduced complexity, our proposed digital filter back propagation technique outperforms Runge–Kutta techniques in terms of root mean square (rms) residual distortion. We quantify rms residual distortion for both methods as sampling time varies. We also examine bit error performance for 16-QAM, as well as the impact of SOA saturation level. We examine robustness to imperfect channel estimation.
© 2011 IEEEPDF Article