Low-Complexity Frequency-Domain Channel Equalizer for Transmitter IQ Impairments Mitigation

Transmitter in-phase and quadrature (IQ) impairments severely degrade the performance of high-baud-rate coherent optical communication systems. While widely linear multiple-input multiple-output (MIMO) time-domain equalizers (TDEs) can effectively mitigate these impairments, their computational complexity poses a significant challenge for hardware implementation. Shifting such equalization operations into the frequency domain offers significant complexity reduction benefits. However, it frequently suffers from unstable tap convergence caused by severe interference between the equalizer and carrier phase recovery loop. In this paper, we propose a unified frequency-domain equalizer (FDE) architecture with embedded dual-stage decision-directed carrier phase estimators to address this issue. Based on this architecture, we specifically investigate a real-valued4× 4FDE incorporating an intra-block symbol-by-symbol phase update mechanism, and a simplified widely linear4× 2FDE utilizing an inter-block parallel phase update strategy. The proposed architectures effectively decouple the transmitter IQ impairments from carrier phase noise and frequency offset without requiring additional overhead. Through comprehensive numerical simulations and transmission experiments using 60-GBaud DP-16QAM and 45-GBaud DP-64QAM signals, we demonstrate that both proposed FDEs achieve robust compensation performance highly consistent with the4× 4TDE. Specifically, under extreme transmitter impairments—including a 7 ps IQ skew, a 3.5 dB IQ gain imbalance, and substantial IQ phase imbalance—the maximum effective SNR penalties of the proposed FDEs are constrained to within 1.01 dB, 1.40 dB, and 2.27 dB for the 16QAM signals, and 0.40 dB, 1.75 dB, and 1.83 dB for the 64QAM signals, respectively. Moreover, compared to the4× 4TDE, the proposed4× 4and4× 2FDEs significantly reduce the computational complexity by 53.25% and 74.00% for the 16QAM scenario, and by 34.77% and 63.67% for the 64QAM scenario, respectively. These results confirm that the proposed embedded FDE architecture offers a highly hardware-efficient and robust digital signal processing solution for next-generation coherent optical networks.