I. Introduction
A physics-based time-domain macromodel is very challenging to obtain analytically for distributed high-speed modules, used in microwave and high-frequency packaging applications. However, accurate admittance (Y) or scattering (S) parameter data describing such structures over the frequency range of interest is available through full-wave simulation. Convolution-based techniques such as in [1] and [2] can be used to perform transient analysis using the band limited frequency domain data. However, such an approach would require a large number of frequency data points, which are CPU expensive to compute using full-wave simulation [3]. Furthermore, the convolution integral has to be computed at each time step, which is CPU expensive and has a high memory requirement [4]. This has led to the development of automated techniques for the generation of closed-form time-domain macromodels based on the frequency-domain response [5]–[18]. In general, the goal of these methods is to compute an approximate transfer function in descriptor system (DS) format or pole/residue format, which matches the frequency domain data. These can then be directly stamped into the modified nodal analysis equations [19] for SPICE type simulation. An alternative approach is to use the pole/residue representation in order to obtain the transient response using recursive convolution [20].