Design of Time-Domain Learned Volterra Equalisers for WDM Systems

We examine various design aspects of a learned time-domain multiple-input multiple-output (MIMO) Volterra-based equaliser and reveal their impact on the convergence and performance of the model. We show that appropriate parameter initialisation is vital for the model's convergence and scalability to a higher number of channels. This design optimisation enables the first demonstration of a 7 × 7 operation of the MIMO algorithm at one step per span, achieving 1.5 dB effective signal-to-noise ratio improvement over single-channel nonlinear equalisation, hence significantly enhancing the transmission performance in a wavelength-division multiplexing scenario.