Robust computation and modelling of wide-band system responses using the Cauchy method

Model estimation can be used to rebuild a wide-band system response from a reduced number of data samples. A particular case is the computation of rational polynomials with order as reduced as possible, using the Cauchy method. This technique can be used to speed up the numerical computation of electromagnetic responses, or as a fitting technique to find models from measured samples. When only finite numerical precision is available, the usual Cauchy method (see Adve, R.S. et al., IEEE Trans. on Microwave Theory and Techniques, vol.45, p.837-45, 1997) is applicable only to low order systems. A Krylov subspace method (Lanczos) is used to solve this limitation, making possible the use of higher order models.