A novel simulations scheduler for automated circuit sizing algorithms

Catalin Visan, Dan Curavale, Georgian Nicolae, Mihai Boldeanu, Horia Cucu, Andi Buzo

Abstract: This study introduces an innovative operating corner management strategy for automated circuit sizing, significantly reducing the number of simulations required by sizing algorithms. The proposed method relies on a simulation scheduler that orders the operating corners, based on their level of difficulty, and applies a selective evaluation strategy. Solutions that fail to meet specifications in more challenging corners are quickly discarded, allocating more resources to promising candidates. This approach, underpinned by a novel metric based on the number of corners passed, demonstrates a reduction of up to 71% in simulation budget requirements compared to the other considered methods. We validate our work using two of the most promising evolutionary algorithms for circuit sizing, applied to two proprietary low-dropout regulators.