SHF: Small: Re-thinking Polynomial Programming: Efficient Design and Optimization of Resilient Analog/RF Integrated Systems by Convexification

Duke University, Durham NC

Investigators

Abstract

This project deals with a novel design and optimization framework to improve performance of advanced analog and radio frequency (RF) integrated systems over a broad range of applications, from consumer electronics to medical instruments with potential impact on the semiconductor industry and national economy. In addition, given its interdisciplinary coverage, the project offers opportunities for training to both university students and industrial engineers, including curriculum development, student advising, outreach activities and workshop organization. It could improve the education infrastructure and generate high-quality researchers and practitioners in related fields. These education activities integrated with the proposed research tasks facilitate the transfer of the new design and optimization techniques to the technical community, and will potentially lead to a broader impact affecting the US semiconductor industry. Aggressive technology scaling, large-scale process variation, rapid introduction of new standards and increased number of autonomous applications have made it necessary to develop resilient analog and radio frequency (RF) integrated systems that can adapt to all variabilities related to process, environment and standard. However, the design and optimization of resilient analog/RF systems has been considered as a grand challenge due to their irregular performance functions, discrete design spaces and high system complexities. This project exploits a novel optimization framework to efficiently design and implement resilient analog/RF systems. The framework is based on recent results optimization theory, and is expected to find the optimal design both efficiently (i.e., with low computational cost) and robustly (e.g., with guaranteed global optimum). The project would apply this technique to optimize resilient analog/RF systems that are composed of tunable analog/RF circuits, on-chip sensors and on-chip controllers. Hence, the proposed framework offers a novel infrastructure for analog/RF design enabling radical improvements.