Date Available

12-17-2025

Year of Publication

2025

Document Type

Master's Thesis

Degree Name

Master of Electrical Engineering (MEE)

College

Engineering

Department/School/Program

Electrical and Computer Engineering

Faculty

Ishan Ghanshyambhai Thakkar

Faculty

Daniel Lau

Abstract

A recent breakthrough in silicon photonics includes the discovery and use of phase changing materials (PCMs). These materials can be programmed to store nonvolatile values, and when a stored value in a PCM cell is read, it changes the amplitude of the read signal, imprinting the value held into the PCM cell on the amplitude of the read signal. This thesis proposes a new approach to using PCM cells not only for photonic memory but also as a substrate to perform multiplications in the photonic domain. The proposed multiplier uses PCM cells to encode amplitude-analog weight values and differing lengths of the read pulse that encodes the time-analog input value. The multiplication result can then be seen as the optical energy packetized in the resulting read pulse, which can be evaluated by measuring the area occupied by the pulse on the amplitude versus time axis. This measurement of the output read pulse can be done using a photodetector connected to a time-integrating receiver. Furthermore, the proposed multiplier can then be employed to build a dot product engine that can be used to perform MAC operations and be used as a tensor accelerator architecture. The proposed PCM-based tensor core accelerator is compared with other optical PCM-based tensor core accelerators. Finally, this thesis presents a deep and complex design space for future research and future improvements on the multiplier.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.563

Download

Share

COinS