Inverse Analog IC Sizing and Exploration through Diffusion Models and Structural Knowledge

In the field of analog integrated circuit sizing, the ability to rapidly and efficiently explore design spaces is crucial due to the demands of fast development cycles, evolving specifications, and increasingly complex circuits. It is also well established that leveraging slack in specifications through adjustments to transistor sizes can enhance yield. To address these challenges, we propose a novel approach that combines two state-of-the-art machine learning techniques to accurately generate optimal performance points, while also enabling the exploration of nearby sizing configurations that may result in more robust designs. Specifically, we integrate a diffusion model with an algorithm that analyzes the circuit and decomposes the problem into simpler subproblems. The proposed models are evaluated on a set of typical operational amplifiers.