Bi-directional gated recurrent unit neural network based nonlinear equalizer for coherent optical communication...

Area of Science:

• Optical communication systems
• Digital signal processing
• Machine learning applications

Background:

• Coherent optical communication systems face significant nonlinear distortions.
• These distortions degrade signal quality and limit transmission performance.
• Advanced equalization techniques are crucial for overcoming these challenges.

Purpose of the Study:

• To propose and experimentally validate a bi-directional gated recurrent unit neural network based nonlinear equalizer (bi-GRU NLE).
• To assess the performance of the bi-GRU NLE in mitigating nonlinear distortions in a high-speed optical system.
• To compare the efficiency of the bi-GRU NLE against a bi-directional long short-term memory neural network based nonlinear equalizer (bi-LSTM NLE).

Main Methods:

• Implementation of a bi-GRU NLE architecture.
• Experimental demonstration in a 120 Gb/s 64-quadrature amplitude modulation (64-QAM) coherent optical communication system.
• Performance evaluation over a 375 km transmission distance.

Main Results:

• The bi-GRU NLE significantly mitigates nonlinear distortions.
• Q-factors exceeding the hard-decision forward error correction (HD-FEC) limit of 8.52 dB were achieved with the bi-GRU NLE for launched optical powers between -3 dBm and 3 dBm.
• The bi-GRU NLE showed comparable performance to bi-LSTM NLE but with 20.2% fewer parameters, shorter training time, and 24.5% fewer multiplications per symbol.

Conclusions:

• The proposed bi-GRU NLE is a highly effective solution for mitigating nonlinear distortions in coherent optical communication systems.
• The bi-GRU NLE offers superior computational efficiency compared to bi-LSTM NLE without compromising performance.
• This technology enables enhanced performance and efficiency in high-speed optical transmission.