Title:  Automatic Surrogate Model Synthesis and Debugging of Analog/RF Designs via Collaborative Stimulus Generation and Behavior Learning

Committee: 

Dr. Chatterjee, Advisor  

Dr. Shaolan Li, Chair

Dr. Davenport

Abstract: The objective of the proposed research is to develop methodologies, support algorithms, test generation and model generation infrastructures for validation and diagnosis of RF, analog, and mixed-signal circuits. In this work a systematic approach for test generation and machine learningassisted model generation algorithms are presented. We show by exhaustive simulations, accurate behavioral model across input stimuli for different circuit examples can be synthesized with very fast simulation time compared to the respective transistor-level implementation. A behavior prediction algorithm also demonstrated the ability to quickly predict accurate process variation circuit performances without actually simulating transistor-level designs. Finally, a diagnosis algorithm that can detect and localize circuit error occurring during the manufacturing or design phases is presented. Results on voltage regulator, amplifier, and RF receiver are presented.