State Space Modelling of Electromagnetic Responses --- a Practical Approach to Extract Parameters from Simulated OR Measured Data
Krishna Naishadham
As computing power and algorithmic advances have evolved rapidly in the recent past, it is now feasible to solve complex electromagnetic (EM) problems involving scattering, radar cross section, antenna design, microwave circuit design, artificial EM materials etc., using full-wave numerical methods. Several general-purpose commercial software packages are routinely used in industry in all these domains for EM analysis or design. However, the task of processing large sets of data output from these design studies and analyses is generally beyond the realm of commercial software packages, and the designer spends many hours writing problem-specific computer programs to extract the desired performance parameters. Some examples where auxiliary processing is needed for the extraction of EM parameters of interest include determination of coupling coefficients or the unloaded quality factor of a dielectric resonator, de-embedding feed lines from antenna currents, removal of discontinuity effects, and the extraction of equivalent circuit models. The same considerations as simulated data apply to the parametric analysis of measured data in the presence of noise. This paper presents a versatile data-driven spectral model derived from a state-space system representation of the computed or measured EM fields, from which all the parameters of interest can be extracted. An attractive feature of the state space method is its ability to identify a small number of the system transfer function poles uniquely associated with a specific scattering mechanism or modal response, thereby enabling its isolation from the total response for detailed study. For example, using SSM, specular reflection and creeping waves on a smooth convex surface can be analyzed and the diffraction at the edges can be isolated from the composite RCS of a large body. The desired field parameter is extracted or estimated from synthetic or measured data using a linear system of a relatively small model order that characterizes the specific modal response of interest. Illustrative examples will be presented to demonstrate the usefulness of the proposed approach for parametric extraction.
