Mathematical approach based on a "Design of Experiment" to simulate process variations

this paper describes a Design Of Experiments (DOE) based method used in computer-aided design to simulate the impact of process variations on circuit performances. The method is based on a DOE approach using simple first and second order polynomial models with multiple experiment maps. It is a technology & circuit-independent method which allows circuit designers to perform statistical analysis with a dramatically reduced number of simulations compared to traditional methods, and hence to estimate more realistic worst cases, resulting in a reduced design cycle time. Moreover, the simple polynomial models enable direct linking of performance sensitivity to process parameters. The method is demonstrated on a set of circuits. It showed very accurate results in linking linearity, gain and noise performances to process parameters, for both RF and analog circuit.