Passive macromodeling via mode-revealing transformation

Direct application of model extraction methods to tabulated admittance data often gives a model which can suffer in accuracy when applied with high-impedance terminations. The problem is relevant in situations where the admittance matrix has a large eigenvalue ratio since the small eigenvalues are likely to become corrupted. We show a modeling procedure which alleviates the accuracy problem by introducing a mode-revealing transformation which is derived from the admittance eigenvector matrix. The transformed admittance matrix is subjected to model extraction and passivity enforcement by standard techniques, leading to a model which captures the full modal information of the transformed matrix and hence that of the original matrix. Finally, the model is transformed back to the original domain.