Performance Improvement of Directly Modulated VCSEL-based BVT via Optical SSB Filter and Time-Delay Neural Network

In the rapidly evolving landscape of optical fiber communication, achieving high performances in short to long reach transmission systems is paramount. Particularly, for addressing the challenges posed by future networks, bandwidth/bi-trate variable transceiver (BVT) based on directly modulated (DM) vertical cavity surface-emitting lasers (VCSELs) represents an attractive and efficient solution. This paper presents an approach that combines optical single sideband (SSB) filtering at the transmitter and time delay neural network (TDNN) at the receiver to enhance the performance of VCSEL-based BVT. We investigated the effects of DM VCSELs’ chirp characteristics on signal integrity. By using its ability to represent complex time-varying signals, the proposed TDNN successfully compensates for the inherent chirp in DM VCSELs, improving signal quality and lowering bit error rates (BERs). Using an optical SSB filter simultaneously reduces the effects of chromatic dispersion. Our results demonstrate that the combination of a 20 GHz bandwidth SSB filter and a trained TDNN model with a correlation coefficient of 0.9875 provides capacities in the 80–53 Gb/s range, at the BER of 4.62•10^-3, for short-to-medium-reach transmission networks with fiber lengths up to 40 km.