C-band 100-GBaud/λ PAM-4 transmission enabled by hardware-efficient nonlinear equalizer with weight-sharing pruning

Area of Science:

• Optical Communications
• Signal Processing
• Digital Signal Processing

Background:

• Intensity-modulation direct-detection (IM-DD) systems face linear and nonlinear impairments.
• Volterra nonlinear equalizers (VNLEs) mitigate these issues but suffer from high computational complexity.
• Reducing computational load is crucial for practical VNLE applications in high-speed optical systems.

Purpose of the Study:

• To propose a hardware-efficient Volterra equalizer with weight-sharing pruning (VE-WSP).
• To significantly reduce the computational complexity of VNLEs.
• To maintain or improve equalization performance in IM-DD systems.

Main Methods:

• Implemented a weight-sharing strategy using the k-means++ clustering algorithm.
• Applied pruning based on ranked cluster centroid weights.
• Experimentally verified the VE-WSP in a C-band 100-GBaud/λ PAM-4 transmission system over 1-km standard single-mode fiber (SSMF).

Main Results:

• The VE-WSP achieved the 20% soft-decision forward error correction (SD-FEC) threshold in the experimental setup.
• Reduced computational complexity by 51% for multiplications and 21% for additions compared to standard VNLE.
• Demonstrated effective mitigation of impairments without performance degradation.

Conclusions:

• The proposed VE-WSP offers a hardware-efficient solution for VNLEs in IM-DD systems.
• VE-WSP significantly reduces computational complexity while preserving equalization performance.
• This approach enables high-speed optical transmission, meeting stringent error correction thresholds.