Optimization of circuits by sampling in solid angle segments bordered by ellipsoids

A statistical Monte Carlo-based procedure for the optimization of the performance of a given electrical circuit is introduced. In the proposed strategy, the total cloud of points in the parameter space generated by a Monte Carlo analysis is approximated by sampling only in a few selected solid angles bordered by hyperellipsoids. As this saves a considerable amount of CPU time, additional search procedures could be introduced that avoid local optima even in cases where other statistical methods fail. The effectiveness of the algorithm is demonstrated by two examples: a bipolar amplifier and a noninverting NMOS superbuffer.<>