The behavior of modern integrated circuits is altered by numerous effects such as process variation, voltage or temperature shifts and aging (PVTA). To evaluate the impact on system level, extensive simulations must be carried out on transistor-level which is not applicable for larger systems. This paper presents methods for the generation of stochastic behavioral models (SBMs) of circuit components to enable the reliability analysis of individual performances on system level. Different modeling approaches are presented, based on either Response Surface Models for parametric or Kernel Density Estimation for non-normal performance distributions. Additionally, an approach to incorporate the correlation between individual system components is analyzed. The SBMs are evaluated on analog example circuits consisting of several components.

现代集成电路的行为会受到许多影响，例如工艺变化，电压或温度变化和老化（PVTA）的影响。为了评估对系统级的影响，必须在晶体管级上进行广泛的仿真，这不适用于大型系统。本文介绍了用于生成电路组件随机行为模型（SBM）的方法，以能够在系统级别上对单个性能进行可靠性分析。基于用于参数化的响应曲面模型或用于非正态性能分布的内核密度估计，提出了不同的建模方法。此外，分析了一种将各个系统组件之间的相关性纳入其中的方法。在由几个组件组成的模拟示例电路上评估SBM。