Abstract
This paper describes a Matlab toolbox (MFT-NNLS) for passive macromodeling of large-scale problems defined by admittance, impedance or scattering data in the frequency domain. Vector Fitting (VF) is used for calculating an initial rational model having a common pole set. Passivity can afterwards be enforced by residue perturbation of the associated pole-residue model. Highly efficient implementations are utilized, including the so-called fast implementation of VF, and passivity enforcement with problem compacting via the QR-NNLS method with usage of a fast LU-based non-negative least squares solver. The final result is a rational model satisfying the physical conditions of realness, stability, passivity and reciprocity. The rational model is returned on alternative forms, including a pole-residue model and a state-space model with real-valued or complex-valued parameters. Optional parameters are available for tuning the behavior of the software to the given application, e.g., least squares weighting schemes and alternative methods for passivity assessment. Model export to leading electromagnetic transient programs is included. Several applications are shown, including large-scale problems from high-voltage power systems (high orders, many terminals) and high-speed electronics (many ports). The software consists of a collection of Matlab functions.