Efficient Methodologies for Statistical Characterisation of Analogue Designs for Submicron CMOS Technologies

This paper presents an efficient methodology for statistical characterization of analog circuits, using Principal Component Analysis (PCA) to generate equations with uncorrelated variables for SPICE model parameters. The design is simulated with variables generated by Monte-Carlo, DOE, or sensitivity analysis, and these three methods are compared using an operational amplifier as a testbench.