Towards Automated Worst-Case Analysis of Circuits: Selecting Initial Values for Global Optimization

Abstract

Worst-case circuit analysis is a mandatory practice in hardware verification and validation. To this end, several methods, including extreme value analysis (EVA) and Monte Carlo analysis are commonly used, however, each has its own limitations. Numerical optimization-based methods have the potential to be generally usable, but have the tendency to get stuck in a local minimum, which can be mitigated using carefully chosen initial values. In this paper we propose methods for automated initial value selection for black-box circuit models. The methods are demonstrated to work on several standard test functions, which is a first step in building an automated worst-case circuit analysis tool.