This paper introduces a novel approach to the design of non-linear circuits. A deterministic gradient-based optimization algorithm (Quasi-Newton method) and a stochastic algorithm (Evolution Strategy method), both coupled to the harmonic-balance analysis technique, are utilized. The deterministic algorithm, when it reaches a solution, uses less CPU time than the stochastic method. On the other hand, the stochastic algorithm generates solutions which are nearer to the global minimum and can be used for problems which are not solvable by means of the deterministic method (high non-linearity and problems with discrete variables).
展开▼
