随机逼近用于中心设计的一种新方法

A NEW APPROACH FOR DESIGN CENTERING BASED ON STOCHASTIC APPROXIMATION

本文提出了一种新的中心设计方法。该方法将一维搜索法与随机逼近优化法相结合,利用二次模型对参数空间上电路的响应进行近似,采用相关抽样和二次模型对设计参数的不同中心值进行Monte carlo分析,估计电路的成品率。实例表明,该方法收敛较快,进行优化设计所需时间较少。

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.