FEATURE: PHOTONIC INTEGRATION
AS PRESSURE ON TELECOMMUNICATIONS NETWORKS CONTINUE TO GROW, PHOTONIC INTEGRATED CIRCUITS (PIC) ARE BECOMING MORE ADVANCED FOR OPTICAL COMMUNICATIONS TO SUSTAIN THE UNPRECEDENTED RISE IN TRAFFIC
ABIGAIL WILLIAMS
PIC IN THE MIX O
ne of the most interesting recent developments in this area is a research project carried out at Princeton University and NEC
Labs America on the development of a silicon photonic-electronic neural network that could enhance submarine transmission systems. As Chaoran Huang, visiting research
scholar at Princeton University and assistant professor at the Chinese University of Hong Kong, explained, the novel system uses silicon photonic technologies, which provide ‘a most highly scalable platorm for photonic integrated circuits because it offers high-quality photonic interconnects, while also delivering high- speed active components with competitive integration on the largest diameter wafers in production’. Based on this technology, the project team
developed a silicon photonic neural network chip to emulate the underlying neural network model with a so-called ‘broadcast-and-weight protocol’ demonstrated by the research group at the Princeton Lightwave Lab. Tis protocol uses the concept of wavelength division multiplexing (WDM) to enable
16 FiBRE SYSTEMS n Issue 35 n Spring 2022
scalable interconnections between photonic neurons. Neurons in this architecture produce optical signals with distinct wavelengths. Tese photonic neurons are multiplexed into a single waveguide and broadcast to all others. ‘Weights are applied to signals encoded
on multiple wavelengths using groups of tunable wavelength filters. Tuning a filter along its transmission edge alters the transmission of each signal through that filter, effectively multiplying the signal with a desired weight. Te resulting weighted signals are sent to a photodetector, which can receive multiple wavelengths in parallel and perform a summing operation on them,’ said Huang. Te generated photocurrent drives an
optical modulator, which translates a nonlinear conversion of electrical photocurrent into optical power. As a result, optical modulators act as the nonlinear activation function – that is, the ‘neuron’ – in the photonic neural network. ‘Tese photonic devices have unmatched
speed and bandwidth,’ said Huang. ‘As a result, algorithms running on the photonic neural networks (PNNs) could break performance limitations in electronics, and gain advantages
www.fibre-systems.com @fibresystemsmag
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40
