lol I -loo 45 90 135 180 -6A 45 90 135 1:O Phi (degrees) Phi (degrees) 1 0 45 90 135 180 2 -30' 0 45 90 135 180 Phi (degrees) Phi (degrees) Figure 8 H-plane gain patterns for a 2 x 2 array of wire dipoles ohtained by superpoxition of the symmetric field patterns. Solid lines: ymmrtn-aided FDTD rewlts. Dottcd lines: method-of-rnoments results (NEC-2). ( a ) 0.5 GHz. ( h ) 1.0 GHz. (c) 1.5 GHz. ( d ) 2.0 GHzsingle-element field patterns. The results are nearly identical for the two calculations: the real and imaginary parts of the single-element E, in the far field agree to at least three decimal places a t all frequencics. The conventional FDTD code requires 28.3 megabytes of memory and 12.687 s of CPU time for a single run, whereas the symmetry-based code requires only 6.9 megabytes and 3,001 s of CPU time for a single run. These single-run results are summarized in Tablc 2. For either conventional or symmetry-aided FDTD. four runs arc required to calculate all four single-element field patterns. to which any desired set of amplitudes and phases can be applied. without the use of symmetry, 50,748 s of CPU are required for such a calculatiun, whereas only 12.004 seconds arc required when symmetry is used, yielding a memory and run-time savings of 3.1.
IV. CONCLUSIONSThe computational resources-memory and CPU timerequired for FDTD simulations of finite phased arrays can be reduced if the array and its excitation possess one or more planes of symmetry. Each plane of symmetry can be replaced with a n electric or a magnetic wall. eliminating that part of the structure on the other side of the wall from consideration and reducing the meniory and run-time requirements by a factor of 2. For a planar array with two planes of symmetry, a factor of 4 reduction in the memory and run-time requirements is obtained. This technique can also be applied t o threc-dimensional arrays with one or more planes of symmetry. The FDTD analysis of a three-dimensional array having three planes of symmcty can be carried out with one-eighth TABLE 2 Performance Comparison Data for the FDTD ~ Number ot unknown\ 1,290,000 324,000 Mcmory required (MB) 28 4 6 9 CPU time per run (s) 12.687 3,001 the computational resources required for a conventional FDTD analysis. REFERENCES 1. ABSTRACT In order to deuelop new methods of nondestructive control of material by microwatws, we are te,yting ihe capabiliq of determining the transmission and reflection properties of a lossy material (i.e., the scattering parameters) in near-field situations. The results giuen by off-the-shelf sofnvare (Hewlett-Packard high fiequency structure simulator) are compared with other computed data obtained by a modal method and experimental daia @(,en by an automatic network analyzer. 0 1996 John Wiley & Sons, Inc.