Fully blind adaptive digital backpropagation (DBP) based on a steepest descent algorithm is proposed. For homogeneous links, a two-parameter blind adaptive (TBA) scheme is presented, which assumes all DBP steps have the same nonlinear scaling factor and filter bandwidth parameter. For heterogeneous links, a multi-parameter blind adaptive (MBA) scheme is proposed to achieve the optimal performance by adapting chromatic dispersion, nonlinear scaling factor and bandwidth parameter of each DBP step. The principles and performance of the proposed schemes are presented for low-pass filter (LPF) assisted DBP, denoted as LDBP, in single carrier systems and cross-phase modulation (XPM) model based DBP in subcarrier-multiplexing (SCM) systems, denoted as SCM-DBP. In a 34.94 Gbaud dual-polarization (DP)-16QAM single channel transmission experiment with a homogeneous 2560 km link, we demonstrate that both TBA-LDBP and TBA-SCM-DBP can converge to optimal performance with various system configurations. In addition, for heterogeneous links, we show in simulations and experiments that MBA-LDBP and MBA-SCM-DBP achieve 0.2~1.4 dB improvement compared to their TBA counterparts.
14 September 2018 in Journal of Mechanical Science and Technology
We propose the design and implementation of frequency domain block least mean square and block recursive least square equalizers for adaptive zero-guard-interval (ZGI) coherent optical (CO) orthogonal frequency division multiplexing (OFDM) systems. We experimentally demonstrate the improvements in fiber nonlinear tolerance and channel estimation accuracy over conventional training symbol (TS)-based equalizers and data aided zero-forcing (ZF) equalizers, in single channel 112-Gb/s quadrature phase shift keying (QPSK), and 250-Gb/s 16 quadrature amplitude modulation (QAM) ZGI CO-OFDM systems. The transmission distance is increased by 20% and 7% for QPSK, and 53% and 35% for 16-QAM systems, compared with the TS-based and decision aided ZF adaptive equalizers, respectively. In addition, simulation results demonstrate the improved temporal tracking ability of the proposed algorithms over the previous ZF equalizer.