Dissertation Defense

Analog and Digital Assist Techniques to Improve Mixed-Signal Performances for Edge Devices and Microsystems

Li Xu
1050 Ford Motor Company Robotics BuildingMap

The Internet has changed the world tremendously. Huge amount of data is being shared through the network, and data is vital for us or artificial intelligence to make decisions. Internet-of-Things (IoTs) is enabling finer grain data collection and sharing, and edge devices are essential elements. These devices must incorporate sensing, computing, communication, and other abilities, while they are limited by stringent power budgets and small form factors. This makes circuit designs for edge devices challenging. Novel architectural and circuit techniques are required for significant power reduction while maintaining or even improving performances.

This dissertation investigates circuit techniques to combine digital and analog circuits for better trade-off between power and performances. It focuses on a 32kHz crystal oscillator design and a digital low-dropout regulator (DLDO) design, which play key roles in the power management scheme of the edge devices. A fully integrated CMOS mixed-signal system for microscopic robots is also introduced as a proof-of-concept to present the latest advancement on ultra-low power microscopic systems.

Chair: Professor Dennis Sylvester