Automation of targeted integrated circuit design tasks

Loading...
Thumbnail Image
Authors
Fiumara, Daniel
Issue Date
2025-12
Type
Electronic thesis
Thesis
Language
en_US
Keywords
Computer and systems engineering
Research Projects
Organizational Units
Journal Issue
Alternative Title
Abstract
As traditional device scaling delivers diminishing returns on system-level performance, circuitdesigners are forced to turn to other strategies to continue to deliver gains in efficiency and capability in modern systems. Industry trends indicate a paradigm shift from relying on transistor density to an emphasis on faster design iteration and vertically integrated system- level co-design. As a result, it is crucial that circuit designers further leverage software tools to automate, optimize, and accelerate electromagnetic-analysis-driven circuit design tasks. This work explores methods of automating and enhancing circuit design workflows for high frequency die-to-die interconnect design and other electromagnetic-simulation-heavy design tasks. Primarily, we utilize machine learning to reduce the time an expert designer spends waiting for results from time-consuming simulations. We achieve this by approximating the electromagnetic simulator with a fast and accurate prediction model, a technique ubiquitous in other engineering disciplines that is gaining traction, but not yet mainstream in integrated circuit design flows. This technique can enable faster design cycle times and more robust design-space exploration, allowing viable circuits to be designed in minutes rather than days. We also present comprehensive methods of large-scale dataset generation to train such prediction models.
Description
December2025
School of Engineering
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY
Terms of Use
Journal
Volume
Issue
PubMed ID
DOI
ISSN
EISSN
Collections