A new method for an efficient optimization of MOS transistor models

Two methods that are especially useful for an accurate optimization of complex transistor models are presented. The first method focuses on sample reduction before the optimization process. An algorithm is described which allows a reduction of the number of samples by a factor of 10 to 20 in an efficient way. The second method is a trust-region-type optimization algorithm which is especially designed for this application. A substantial part of it is a new algorithm for the calculation of the step length. Industrial examples demonstrating the quality of the algorithms are given.<
>