March 9, 2021 /

INÜ Demonstrates Highest Spectral Efficiencies in Optical Fiber Communications

Machine Learning teaches us novel Pulse Shaping and Detection Techniques for mitigating Nonlinearities in Optical Fibers

Researchers Tim Uhlemann, Sebastian Cammerer, Alexander Span, Sebastian Dörner and Stephan ten Brink of the Institue of Telecommunications (Nachrichtenübertragung, INÜ) demonstrated a new method based on machine learning for achieving highest spectral efficiencies in coherent optical fiber communications. For this, a novel joint pulse shaping and signal detection at transmitter and receiver, respectively, is learned through an "autoencoder" – a concept originating from the field of artificial intelligence (AI). The new pulse shapes turn out to stay spectrally more compact while propagating along the nonlinear fiber medium compared to conventional pulse shaping techniques, paving the way for further enhancing data rates and signal reach in long-haul coherent optical fiber communications.

Some of the results have already been published in  T. Uhlemann, S. Cammerer, A. Span, S. Doerner and S. ten Brink, "Deep-learning Autoencoder for Coherent and Nonlinear Optical Communication," Photonic Networks; 21th ITG-Symposium, Online, 2020, pp. 1-8.

For further information, check out the corresponding press release of the University of Stuttgart (see links).

To the top of the page