Gradient approximation method for circuit optimisation

In the paper a new optimisation method, gradient approximation and descent control (GADC) is presented in two parts. The first part is an algorithm based on a data processing technique to approximate the performance gradients, which permits a fast convergence rate. The second is an iterative algorithm which allows an improvement in the approximation, and consequently a solution of the problem in reduced computation time. It is particularly suited to solve the `FIT´ problems, to applications related to numerous complex circuit models, and to multiple-objective problems having double-sided specifications (e.g. limited satisfaction intervals). Examples given include model parameter fitting with a very competitive overall computation time and satisfying very tight specifications