Statistical IC simulation based on independent wafer extracted process parameters and experimental designs

A statistical bipolar IC simulation methodology that employs functional Gummel-Poon model parameters controlled by an independent set of process parameters extracted at a wafer probe is discussed. Fractional factorial screening and Box-Behnken experimental designs are used with regression analysis to develop a response surface polynomial model for each IC parameter as a function of the process parameters. A Monte-Carlo algorithm operates on each polynomial model to define the mean and standard deviation of each IC parameter, minimizing CPU time. A comparison is made between the simulated and measured statistical DC parameters of an operational amplifier in order to assess quantitatively the effectiveness of this method. Close agreement is found