Recurrent Neural Network Soft-Demapping for Nonlinear ISI in 800Gbit/s DWDM Coherent Optical Transmissions
Maximilian Schaedler, Fabio Pittalà, Georg Böcherer, Christian Bluemm, Maxim Kuschnerov, Stephan Pachnicke
Abstract
Optical transmission systems suffer from ISI induced by nonlinear components. As a countermeasure, a recurrent neural network soft-demapper is proposed and benchmarked against a combination of Volterra equalizer, noise whitening filter and soft-output Viterbi-algorithm. A substantial gain up to 1.52dB in 800G-DWDM-600km measurements is achieved.
Topics & Concepts
Computer scienceWavelength-division multiplexingArtificial neural networkViterbi algorithmTransmission (telecommunications)Nonlinear systemEqualization (audio)Noise (video)Optical filterElectronic engineeringFilter (signal processing)TelecommunicationsArtificial intelligenceDecoding methodsEngineeringOpticsPhysicsWavelengthImage (mathematics)Computer visionQuantum mechanicsOptical Network TechnologiesAdvanced Photonic Communication SystemsPhotonic and Optical Devices