Committee: 

Dr. Peterson, Advisor     

Dr. Cressler, Chair

Dr. Klein

Abstract:

The objective of the proposed research is to apply theoretical analysis and Machine Learning techniques to develop a class of solutions we call the BCOD for a variety of mm-wave design tasks, including Impedance Transforming Baluns, Power Combiners, Out-phasing circuits, and Doherty networks. The BCOD structure is an extension of a Marchand balun with a shorter electrical length and lumped capacitors. In this thesis, we analyze the network requirements of the mm-Wave tasks, apply mathematical constraints, solve for theoretical solutions on the BCOD structure, and calculate optimum electrical parameters to resolve the mm-Wave design tasks. To implement theoretical microwave structures in physical realizations, researchers often employ time-consuming techniques based on trial-and-error to tune EM parameters. We attempt to break this prolonged process by learning from data and employing Machine Learning techniques. We propose to develop a database of transmission-line and coupled-line implementations, build Machine Learning models that learn Electronic-EM relationships, and utilize optimizers to directly transform from electrical parameters to optimum physical dimensions with lowest loss. The goal is to fully automate a design flow that instantaneously generates a full EM design when given a specification of relevant mm-Wave tasks.