Efficient methods for computing the frequency response of linear networks

Abstract

An efficient method for computing the frequency response of linear networks is presented. Observing that the changes in the system matrix are due entirely to the frequency-dependent elements, the response of networks at new frequencies can efficiently be computed via the method of large change sensitivity. If the response has to be computed at a number of frequencies, the number of arithmetic operations required can further be reduced by maintaining the upper Hessenberg form of the reduced system. This technique finds applications in the optimization of linear circuits and steady-state analysis of nonlinear networks using the harmonic balance method.