Robust recursive inverse approximation and its application to parameter extraction of behavioral models

An algorithm, Robust Recursive Inverse Approximation (RRIA), for parameter extraction in the statistical IC modeling is presented. RRIA gives very accurate modeling parameter formulas based on observation data after the behavioral model is set up. It can extract the parameters simultaneously for multiple circuits which have same structure but with different parameter values. RRIA combines Newton-Raphson, least-squares and Singular Value Decomposition techniques. The theory of this algorithm is presented in detail. Its efficiency for different modeling levels is verified by two examples: the MOSFET device and the bandpass filter.