Analog Spiking Neural Network Based Phase Detector

• Published in: IEEE transactions on circuits and systems. I, Regular papers Vol. 69; no. 12; pp. 1 - 10
• Main Authors: Lehmann, Hendrik M., Hille, Julian, Grassmann, Cyprian, Knoll, Alois, Issakov, Vadim
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.12.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Analog circuits; BICMOS; hardware; Neural networks; neuromorphic; neuron; Phase detectors; phasedetector; Reference signals; Signal processing; SNN; software; Spiking; synapse; Time signals
• ISSN: 1549-8328; 1558-0806
• DOI: 10.1109/TCSI.2022.3204433

Spiking Neural Networks represent the third generation of biologically inspired systems for signal processing. They are associated with a particularly efficient and thus low-energy possibility of computing. However, this advantage can only be fully achieved if these networks utilize special neuromorphic circuits. In this work, an analog Spiking Neural Network Phase Detector is presented, from conceptual formulation to implementation in a <inline-formula> <tex-math notation="LaTeX">{130}~{\text{nm}}</tex-math> </inline-formula> BiCMOS process. The phase detector is capable of directly processing various continuous-time signals up to a frequency of <inline-formula> <tex-math notation="LaTeX">{200}~{\text{MHz}}</tex-math> </inline-formula>, while consuming just <inline-formula> <tex-math notation="LaTeX">{840}~{\mu\text{W}}</tex-math> </inline-formula>. The phase difference between the signal under test and the reference signal that shall be detected is adaptable. Experimental findings confirm the simulative investigations. The proposed method presented in the paper provides an entry-level approach to designing more complex analog spiking neural networks.