Robust analysis of electromagnetic responses of resonant structures using early-time and low-frequency data.

摘要

Computational electromagnetic (CEM) methods and other numerical tools are a vital component of the design process used for many important applications, such as the development of antennas with the performance needed to meet requirements for modern wireless communications or the design of enclosures to harden electronic devices against electromagnetic interference (EMI). With CEM methods an engineer is provided the means to accurately simulate and study the EM behavior of a system or structure before constructing a physical prototype.;By fitting early-time and low-frequency numerical data a response can be simultaneously extrapolated in time and frequency. The accuracy of the extrapolation depends critically on the proper selection of several parameters. Selecting these parameters is very difficult in practice. In Chapter 1, a procedure is presented to address these limitations. Pole terms are incorporated into the representation of the response and are shown to accurately and efficiently model the effects of resonances. An optimization routine is developed that automates the selection of all the necessary parameters and provides confidence in the accuracy of the result. The advantages of the new procedure are demonstrated by extrapolating the wideband driving-point current of several antennas.;In Chapter 2, further improvements to the procedure in Chapter 1 are described, and three pole-estimation techniques are presented. It is shown that a response can be accurately extrapolated with poles estimated from either early-time data or low-frequency data. By comparing the two approaches, a time or frequency bias is discovered when estimating poles from early-time or low-frequency data, respectively. A procedure to combine the sets of poles determined from early time and low frequency into a single set is also presented and shown to reduce the amount of CEM data needed to successfully apply the extrapolation procedure. The relative performance of the three pole-estimation methods is studied, and the combined method is used to extrapolate the responses of a multi-band antenna and a cavity structure of interest to EMI applications.;In Chapter 3, a reliable and computationally-efficient procedure to extrapolate a response defined in a general spatial region is developed by extending the procedure of Chapter 2, which is applicable to point responses. The spatial variation of the response can be accurately modeled with spatially-dependent polynomial coefficients and pole residues, and it is shown that a single set of poles, common to each discrete spatial location, is sufficient to describe the resonant behavior over the entire spatial region. In the representation of the response poles are either physical poles, which correspond to structural resonances, or fitting poles, which are not associated with resonances but can improve accuracy of the representation. Identifying the physical poles of a response is valuable but often difficu however, a new procedure to automate this process is developed. The physical poles of a dipole are compared to complex natural resonances determined with the singularity expansion method (SEM), and the physical poles of a patch antenna are compared to the modes of a cavity model. The spatial variation of the residues of physical pole terms, referred to as modal residues, is found to provide valuable physical insight into the resonant behavior of a structure.;In Chapter 4, the use of modal residues for the analysis of antennas is explored. Modal residues, determined with the procedure of Chapter 3, are seen to be similar to the natural modes found with SEM. While applicable to many different types of resonant systems, the modal residues of patch antennas are determined to demonstrate the value of the approach. The extrapolation procedure is applied to data corresponding to the electric field between the patch and ground plane for several antennas. It is found that the spatial distribution of the modal residues illustrate the influential parameters of each resonance and can be used to identify the resonances that will be excited for a given probe location. Additionally, the spatial variation of modal residues is seen to be similar to the input resistance at resonant frequencies as a function of the probe location. Rectangular and non-rectangular patch shapes are considered to illustrate the generality of the approach. (Abstract shortened by UMI.)