Bin Shi

• Doctor of Philosophy
• PostDoc at Eindhoven University of Technology

Semiconductor optical amplifiers (SOAs) are widely used as active elements in optical switching networks for their features of loss compensation and fast reconfiguration time. To optimize the SOA chain operation, it is essential to know the gain response of each SOA. However, the performance of the individual SOA in an on-chip network is challengin...

Metro-access networks exploiting wavelength division multiplexing (WDM) to cope with the ever-growing bandwidth demands are sensitive to cost and need to be fast-configurable to meet the requirements of many new network services. Optical add-drop multiplexers (OADMs) are a key component in enabling fast dynamic wavelength allocation and optimizatio...

We demonstrate a flexible metro-access network exploiting SOA-based OADM nodes and digital subcarrier multiplexing with power loading. Results show that at least 4 nodes can be supported for 40-Gb/s transmission with bandwidth allocation on demand.

We fabricated and assessed a wideband (C-/L- band) nanosecond 1×2 electro-optic switch in 3-µm thick silicon. Results show 2.5dB lowest insertion loss, 16dB averaged extinction ration, 2.5dB polarization dependent loss and 6-ns switching time.

As the global internet protocol (IP) traffic volume growth puts more pressure on network connectivity, bandwidth and latency requirements, crucial network elements such as switches need continuous improvement. To this end, we report a monolithically integrated, ultra-compact 8×8 optical space switch based on semiconductor optical amplifier (SOA) ga...

We experimentally demonstrate the emulation of scaling of the semiconductor optical amplifier (SOA) based integrated all-optical neural network in terms of number of input channels and layer cascade, with chromatic input at the neuron and monochromatic output conversion, obtained by exploiting cross-gain-modulation effect. We propose a noise model...

We demonstrate the use of a wavelength converter (WC), based on cross-gain modulation in a semiconductor optical amplifier (SOA), as nonlinear function co-integrated within an all-optical neuron realized with SOA and WDM technology. We investigate the impact of fully monolithically integrated linear and nonlinear functions on the all-optical neuron...

Convolutional neural network (CNN) is one of the best neural network structures for solving classification problems. The convolutional processing of the network dominates processing time and computing power. Parallel computing for convolutional processing is essential to accelerate the computing speed of the neural network. In this paper, we introd...

The computing industry is rapidly moving from a programming to a learning area, with the reign of the von Neumann architecture starting to fade, after many years of dominance. The new computing paradigms of non-von Neumann architectures have started leading to the development of emerging artificial neural network (ANN)-based analog electronic artif...

We demonstrate the first optical processing of up to 9-bit/symbol multi-level modulated channels on a complete all-optical SOA-based neuron, with an error of 0.08. A higher number of modulation levels and inputs can improve accuracy.

We experimentally evaluate an InP 1×8 photonic integrated wavelength selective switch based on SOA optical gates. Results show error-free operation with a limited penalty at 10 and 40 Gb/s NRZ-OOK, on-chip gain >7 dB and cross-talk <−45 dB.

We demonstrate 4×35 Gbps error-free WDM data routing in a lossless compact 8×8 InP optical space switch with 2 dB worst-case penalty. 10 dB IPDR within 1.5 dB power penalty is measured at 12.5 Gbps.

We experimentally verify the noise modeling of SOA-based all-optical neural networks on the OSNR degradation of the neuron: The photonic neural network can scale up to 16×16 neurons per layer, resulting in NRMSE <0.10.

We emulate and experimentally validate the scaling of SOA-based all-optical deep neural networks by accurately modeling the OSNR degradation in a chain of SOAs: The photonic neural network can scale up to 16 layers when using only 4 neurons/layer without notable accuracy degradation.

We propose a novel photonic accelerator architecture based on a broadcast-and-weight approach for a deep neural network through a photonic integrated cross-connect. The single neuron and the complete neural network operation are numerically simulated. The weight calibration and weighted addition are reproduced and demonstrated to behave as in the e...

An 8×8 InP cross-connect chip for optical switching within ROADMs is employed for demonstrating optical feed-forward neural networks for analog data processing. An all-optical approach is also explored for deeper optical neuromorphic computing on chip.

We demonstrate a monolithically integrated SOA-based photonic neuron, including both the weighted addition and a wavelength converter with tunable laser as nonlinear function, allowing for lossless computation of 8 Giga operation/s with an 89% accuracy.

Photonic WDM switches and photonic integrated WSS are designed as building blocks to realize novel modular metro node architectures. Advances in compact photonic integrated InP switches using the InP generic technology will be discussed.

Photonic neuromorphic computing is raising a growing interest as it promises to provide massive parallelism and low power consumption. In this paper, we demonstrate for the first time a feed-forward neural network via an 8x8 Indium Phosphide cross-connect chip, where up to 8 on-chip weighted addition circuits are co-integrated, based on semiconduct...

We successfully demonstrate classification of three classes of Iris flowers by implementing a trained neural network on an SOA-based InP cross-connect chip. Classification accuracy of 91.6% is achieved after a fine optimization procedure.