Machine Learning Assisted Design of mmWave Wireless Transceiver Circuits - Thesis - Master of Science

Aug 5, 2026

Upload by Riley

This thesis presents machine learning–assisted design methodologies for mmWave wireless transceiver circuits, targeting both transmitter and receiver architectures operating around 28 GHz. It reviews supervised learning and reinforcement learning approaches for circuit synthesis, then details the design, performance specification analysis, and Ocean script–based implementation for a 28-GHz transmitter and receiver. The work further develops a transceiver-circuit design workflow driven by dataset construction, model training, and evaluation, culminating in conclusions and future directions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages52

Reading Time2 h 11 min

AccessFree to download

1 Introduction

1.1 Background

1.2 Motivation

1.3 Organization of This Thesis

2 Literature Review

2.1 Supervised Learning

2.2 Reinforcement Learning

2.3 Combination of NN and Optimization Algorithms

3 A 28-GHz Transmitter Design

3.1 Transmitter Architecture

3.2 Performance Specifications Analysis

3.3 Ocean Scripts Implementation

4 A 28-GHz Receiver Design

4.1 Receiver Architecture

4.2 Performance Specifications Analysis

4.3 Ocean Scripts Implementation

5 Machine Learning Based Transceiver Circuits Design

5.1 Problem Statement

5.2 Data Collection

5.3 Model Evaluation

6 Conclusion and Future Works

6.1 Conclusion

6.2 Future Works

arXiv :2407 .07458v1 [ ee ss . SY] 10 Jul 2024 UNIVERSITY OF CALIFORNIA, IRVINE Machine Learning Assisted Design of mmWave Wireless Transceiver Circuits THESIS submitted in partial satisfaction of the requirements for the degree of MASTER OF SCIENCE in Electrical and Computer Engineering by Xuzhe Zhao Thesis Committee: Assistant Professor Hamidreza Aghasi, Chair Assistant Professor Rahim Esfandyar-Pour Assistant Professor Yanning Shen © 2024 Xuzhe Zhao DEDICATION To my parents and my friends, for their unwavering support and encouragement. TABLE OF CONTENTS Page LIST OF FIGURES v LIST OF TABLES vi ACKNOWLEDGMENTS vii ABSTRACT OF THE THESIS viii 1 Introduction 1 1.1 Background .................................... 1 1.2 Motivation ..................................... 3 1.3 Organization of This Thesis ........................... 4 2 Literature Review 5 2.1 Supervised Learning ............................... 8 2.2 Reinforcement Learning ............................. 9 2.3 Combination of NN and Optimization Algorithms ............... 11 3 A 28-GHz Transmitter Design 13 3.1 Transmitter Architecture ............................. 14 3.2 Performance Specifications Analysis ....................... 15 3.2.1 Individual Performance .......................... 15 3.2.2 System Performance ........................... 18 3.3 Ocean Scripts Implementation .......................... 18 4 A 28-GHz Receiver Design 21 4.1 Receiver Architecture ............................... 22 4.2 Performance Specifications Analysis ....................... 24 4.2.1 Individual Performance .......................... 24 4.2.2 System Performance ........................... 27 4.3 Ocean Scripts Implementation .......................... 28 5 Machine Learning Based Transceiver Circuits Design 31 5.1 Problem Statement ................................ 32 5.2 Data Collection .................................. 32 5.3 Model Evaluation ................................. 33 6 Conclusion and Future Works 36 6.1 Conclusion ..................................... 36 6.2 Future Works ................................... 37 Bibliography 38 LIST OF FIGURES Page 2.1 Classification for prior automating techniques for circuit synthesis....... 6 2.2 Machine learning techniques for analog circuit design reviewed in this thesis. 7 2.3 General learning loop for RL [46] ......................... 9 2.4 Design flow for RL used in analog circuit design [29] .............. 10 2.5 An example of leveraging the knowledge from (a) simpler circuits to (b) more complex one [51] .................................. 12 3.1 A 28GHz direct conversion transmitter architecture involving VCO and PA. 14 3.2 LC cross-coupled VCO schematic......................... 15 3.3 VCO implementation with buffer using NCSU 45nm PDK in Cadence Virtuoso 15 3.4 Two-stage differential cascode PA schematic................... 16 3.5 PA implementation using NCSU 45nm PDK in Cadence Virtuoso ...... 16 3.6 Schematic of target system on transmitter side in Cadence Virtuoso...... 19 4.1 A 28GHz heterodyne architecture comprising LNA, mixer and cascode amplifier. 23 4.2 Cascode LNA schematic with an inductive degeneration............ 23 4.3 LNA implementation with differential configuration using NCSU 45nm PDKin Cadence Virtuoso ............................... 23 4.4 Double-balanced Gilbert cell mixer schematic.................. 24 4.5 Mixer implementation using NCSU 45nm PDK in Cadence Virtuoso .... 24 4.6 Differential cascode amplifier schematic..................... 25 4.7 Cascode Amplifier implementation using NCSU 45nm PDK in Cadence Virtuoso 25 4.8 Schematic of target system on receiver side in Cadence Virtuoso........ 29 5.1 Procedure for creating datasets for 28GHz transceiver circuits ........ 33 5.2 An end-to-end model training and evaluation pipeline............. 35 LIST OF TABLES Page

FAQs

What problem does the thesis address in mmWave transceiver circuit design?

It addresses how to use machine learning to assist the design of mmWave wireless transceiver circuits, with emphasis on transmitter and receiver architectures around 28 GHz.

What learning paradigms are reviewed and how do they relate to circuit synthesis?

The literature review covers supervised learning, reinforcement learning, and combinations of neural networks with optimization algorithms, positioning them for automated analog circuit synthesis.

How are the transmitter and receiver designs implemented and evaluated?

The thesis specifies transmitter and receiver architectures, analyzes individual and system-level performance, and uses Ocean scripts for implementation and analysis, then evaluates machine-learning-driven design models using collected data and an end-to-end training pipeline.

Aug 5, 2026

Upload by Riley