Windowing technique in frequency-domain simulation

The verification of a number of mixed-technology and mixed-signal design (RF circuits, for instance) often requires frequency domain simulations of nonlinear systems in the presence of signals that are best characterized as having continuous frequency spectra. This paper describes a new frequency-domain simulation algorithm for the computation of the response of nonlinear systems to band-limited signals with continuous spectra. The algorithm is based on an orthogonal series expansion of the signals in the frequency domain, and does not rely on a linearization of the system equations. Signal spectra are obtained from the solution of a system of nonlinear algebraic equations, and windowing techniques are used to eliminate the Gibbs phenomenon and to improve the spectral accuracy of the results, fir this purpose, the performance of a number of windows is compared on the basis of a newly introduced characteristic function, the windowing error kernel. Numerical results obtained from the simulation of an optical amplifier that are representative of the algorithm´s performance are presented.