High-speed PAM8 silicon photonics IM-DD system with artificial neural network equalization

The intensity modulation direct detection (IM/DD) system based on silicon photonic devices stands out as a leading contender for the next generation of short-reach optical communication due to its cost-effectiveness, low power consumption, and compact physical footprint. Nonetheless, its direct representation of digital information through amplitude variations renders them acutely susceptible to transmission impairments. To improve the signal quality at the receiver, digital signal processing (DSP) based equalization plays a pivotal role due to its programmability, flexibility and stability. Among different kinds of equalization methods, neural network (NN)-based equalization algorithms have attracted considerable attention, surpassing traditional algorithms such as feed-forward equalization (FFE), decision feedback equalization (DFE) and Volterra series-based nonlinear equalization (VNLE) et al. This increased attention is attributed to their robust capability for modeling both linear and nonlinear systems. In this paper, by employing a novel NN-based equalization with eight saturation regions activation function, we successfully transmit a 60 GBaud 8-arypulse amplitude modulation (PAM8) signal with the bit error rate (BER) below high-definition forward error correction(HD-FEC) threshold of 3.8×10^-3 and a 70 GBaud PAM8 signal with BER below soft-decision forward error correction(SD-FEC) threshold 2×10^-2 using a 4-layer network architecture. Compared to the traditional activation function such as sigmoid and tanh, 1~3 orders of magnitude of BER can be decreased. The results show that the proposed innovative NN-based equalization has the potential to significantly enhance the performance of the next generation silicon photonics based short-range optical communication systems.