Silicon Photonic Integrated Reservoir Computing Processor with Ultra-high Tunability for High-speed IM/DD Equalization

Abstract

Intensity modulation and direct detection (IM/DD) technology still dominates the optical fiber communication region for the sake of cost and energy efficiency. Reservoir Computing (RC), a special machine learning algorithm suitable for sequence models, has recently been applied to reduce the inter-symbol interference (ISI) caused by dispersion and Kerr nonlinearity in IM/DD systems. In this paper, we designed and numerically simulated a Photonic Integrated Reservoir Computing Processor (PIRCP) with two recurrent nodes using a standard silicon-on-insulator platform. The PIRCP exhibits ultra-high tunability of phase, intensity, delay time and detuning frequency of the optical carrier, which greatly facilitates parameter sweeping for the obtaining of the optimal processing performance. To validate the efficiency of our design, we implemented the PIRCP along with a Feed Forward Equalizer (FFE) in the receiver-end, and finally achieved sub HD-FEC performance for 112 Gbps/λ transmission over 60 km standard single-mode fiber (SSMF) with an ROP of -15 dBm, showing an improvement of 5 dBm compared with non-RC scheme.