Robust recursive inverse approximation-a singular value decomposition-based statistical IC parameter extraction

Recursive Inverse Approximation (RIA) is a new method of statistical device model parameter extraction. The standard extraction process is most often based on a combination of initial direct parameter extraction using analytical formulas with optimization-based refinements (often using global optimization algorithms). This, however, introduces “noise” into the extracted parameter values, due to the uncertainties related to the termination of the optimization process. Moreover, the entire extraction process usually takes a long time. The new RIA-based approach avoids these problems, allows reliable direct accuracy checking of statistical parameters, controllable parameter extraction, is very fast (once the RIA is constructed) and accurate. However, in some practical cases (especially for device-level modeling) the construction of RIA can be difficult or impossible using the current linear regression approach, due to the ill-conditioning of the RIA modeling problem and/or algorithm divergence during the model creation. In this paper, a Singular Value Decomposition (SVD) approach is proposed to implement a Robust RIA (RRIA) algorithm