JLT22 – fmsyn

SOA-REAM Assisted Synaptic Receptor for Weighted-Sum Detection of Multiple Inputs (Top-Scored Paper)

JLT, vol. 41, no. 4, pp. 1258-1264, Feb. 2023


Margareta Vania Stephanie, Florian Honz, Nemanja Vokić, Winfried Boxleitner, Michael Waltl, Tibor Grasser, and Bernhard Schrenk


Neuromorphic photonics is a promising research field due to its potential to tackle the limitations arising from the bottleneck of the von-Neumann computation architecture. Inspired by the characteristics and behavior of the biological brain, photonic neural networks are touted as a solution for solving complex problems that require GHz operation at low latency and low power consumption. An essential building block of such a neural network is a low-complexity multiply-accumulate operation, for which efficient functional implementations in the optical domain are sought for. Towards this direction, we present a synaptic receptor that functionally integrates weighting and signal detection. This optical multiply-accumulate operation is accomplished through a monolithic integrated semiconductor optical amplifier and reflective electro-absorption modulator, which together serve as a colorless frequency demodulator and detector of frequency-coded signals. Moreover, we show that two spike trains can be simultaneously processed with alternating signs and detected as a weighted sum. The performance of the proposed synaptic receptors is further validated through a low bit error ratio for signal rates of up to 10 Gb/s.

Preprint of publication: