Microstrip antennas have been one of the most rapidly developing and intensely studied subjects in the past three decades. Due to the richness of its configuration, a full wave solver designed specifically for these structures is demanded. Microstrip structures usually consist of a sandwich of two parallel conducing layers separated by a single thin dielectric substrate. Since the dielectric substrates and the conducting ground are usually much larger compared to the radiating elements, it is preferable to use integral equation method and put unknowns only on the radiating elements than discretizing the entire structure. Thus the integral equation methods with a layered medium dyadic Green's function would be a good candidate as a full-wave solver for this type of problems. In addition, since the major advantages of microstrip antennas are realized in applications that require moderate size arrays, a full-wave solver with a fast-algorithm that can handle large problems efficiently is mostly desired. Thin-stratified medium fast-multipole algorithm (TSM-FMA) [1] is such an integral equation algorithm with layered-medium Green's function that has been proposed for this purpose. It was able to reduce the computational complexity of microstrip structures to O(NlogN). However, the previous algorithm is limited to the analysis of planar structures only. This is a significant drawback because it fails to model many features and variations of microstrip antennas, for example, vertical coaxial feed, inverted-F antennas, stacked antennas and so on. To make it a truly "full-wave" solver, we need to modify the algorithm to its modeling capability while retaining its efficiency. In this paper, we would first review a newly developed dyadic Green's function for layered medium [2], which is derived using vector wave functions approach. This Green's function is not only more suitable for moment method (MOM) implementation, but also more convenient for incorporating 2D fast-multipole-method (FMA) to accelerate it. Since it is naturally decomposed into TE to z and TM to z parts, the z' variation could be characterized by the propagation factor of these two parts and the algorithm could handle non-planar structures without increasing the computational complexity.
展开▼
