A NEW APPROACH FOR DESIGN CENTERING BASED ON STOCHASTIC APPROXIMATION
随机逼近用于中心设计的一种新方法

A new optimization technique which combines one dimension exploration with stochastic approximation is presented. A quadratic model is used to approximate the response of a circuit over a parameter space. The circuit yield for different design center values is estimated using correlated sampling and quadratic model via Monte Carlo method. An example shows that the convergence of this approach is faster and required CPU-time is less.