Equalization of a 10  Gbps IMDD signal by a small silicon photonics time delayed neural network

Abstract

A small 4-channel time-delayed complex perceptron is used as a silicon photonic neural network (PNN) device to compensate for chromatic dispersion in optical fiber links. The PNN device is experimentally tested with non-return-to-zero optical signals at 10 Gbps after propagation through up to 125 km optical fiber link. During the learning phase, a separation-loss function is optimized in order to maximally separate the transmitted levels of 0s from the 1s, which implies an optimization of the bit-error-rate. Testing of the PNN device shows that the excess losses introduced by the PNN device are compensated by the gain in the transmitted signal equalization for a link longer than 100 km. The measured data are reproduced by a model that accounts for the optical link and the PNN device. This allows simulating the network performances for higher data rates, where the device shows improvement with respect to the benchmark both in terms of performance and ease of use.

Chromatic dispersion compensation via an all-optical perceptron

“…Furthermore, since the loss function does not significantly increase when N op > N exp T , it is convenient to operate the PNN with all the taps in the open state and train the PS only as done in. 7 This will increase the PRX and, in turn, will reduce the training time without comprising the equalization.…”

“…The training is performed offline via the Particle Swarm Optimizer (PSO) 12 by minimizing a properly built analog loss function L extensively described in. 7 The training aims to create the maximum relative separation between the distributions associated with the expected optical levels (i.e. the maximum contrast between levels, thus an aperture of the eye diagram) and minimize the Bit Error Rate (BER).…”

“…5 We proposed a Silicon-integrated Photonic Neural Network (PNN) for channel equalization in IM-DD schemes. 6,7 Compared to other impairment mitigation methods, such as dispersion-compensating fibers 8 (DCFs), our solution is reconfigurable, has low production costs, and introduces minimum latency (being the PNN of the feed-forward type). The PNN also features reduced power consumption compared to thermally or electrically tunable chirped fiber Bragg gratings.…”

“…10 Here, we discuss an improved multi-channel version of the PNN of ref. 7 where both the phase and amplitude of the weights are learned during the training of the PNN. .…”

“…A first equalization demonstration of the PNN design of Fig. 1 has been provided in 7 with N = 4, B = 10 Gbps, L = 100 km, and ∆ t = 50 ps. The ∆ t value was chosen to have a sufficiently dense sampling of the information in a single-bit time slot (at least 2 samples per bit).…”