A modified simulated diffusion algorithm for global optimization of model parameters

The conventional optimization technique such as the gradient-based Levenberg-Marquardt (L-M) method, which has been typically used in the commercially available parameter extraction tools, can easily be trapped in the local minima of the objective function, On the contrary, a method called simulated diffusion (SD), inspired by the simulated annealing, is known to be the most effective way to find the global minimum. In this paper, the parameter extraction algorithm based on the SD, which we have modified slightly to improve the efficiency, is presented. This is the first time that "both" dc and ac model parameters of a bipolar transistor or dc model parameters of NMOS transistors in different geometry have been extracted simultaneously via the global optimization methodology using the SD.<
>