This article is based on a paper "ANALYSIS OF HIGH EFFICIENCY REFLECTOR FEED ARRAY USING A GENERAL-PURPOSE SOFTWARE PACKAGE" . The spacing of elements in an antenna array is determined by the requirement to suppress grating lobe responses at the maximum operating frequency, and is typically around 0.6 wavelengths at this frequency. For wideband arrays this means that the element spacing may approach a quarter of a wavelength or less in the lower part of the operating band, implying strong electromagnetic coupling between the array elements. This mutual coupling has a significant effect on the element match and radiation pattern and must be modeled accurately for optimal array design. In large phased arrays, the "infinite array" approximation can be used for the majority of array elements and edge effects are reduced to some extent by the use of passive, "guard" elements around the periphery of the array. However, for arrays of moderate size, such as a focal-plane array (FPA) feed for a parabolic reflector antenna, neither of these approaches is so attractive. The use of symmetry can reduce the problem, but still the complete electromagnetic analysis of one quadrant or one sextant of the array is required. FPAs of the order of 100 dual-polarized elements are being considered for next-generation radio telescopes, so an initial goal is the analysis of an array of the order of 20 dual-polarized elements. The aim of this work is to show the capability of CST MICROWAVE STUDIO® (CST MWS), to perform a complete analysis and optimization of a 19-element FPA. The eventual aim is to use this approach for the practical design of a wideband radio telescope FPA. The array elements chosen here are dual-polarized “four-square” dipoles used to feed a 13.7m paraboloid with a f / D ratio of 0.41, operating in the frequency band of 1.4 to 1.7 GHz. The array excitations are obtained by sampling and phase conjugation of the focal field calculated by physical optics....
SINGLE ELEMENT: As a first step, a single element of the array has been built to optimize the position of the probe feed. The entire geometry setup as been parameterized in order to allow optimization and frequency scaling. The single elements have been excited by using a multi-pin waveguide port as shown in Figure 2. These ports allow the definition of arbitrary differential and common current sets and use the eigenmodes of the coax feeds as excitations.
Figure 3 depicts the return loss of a single element as a function the feeding position calculated by the time domain solver of CST MWS. Based on this parameter study, a distance of 6 mm between corner of the dipole elements and the probe has been chosen.
Array of 19 elements: The entire array consist of 19 dual-polarized dipole elements. In this simulation the amplitude and phase for each element was calculated using a physical optics approach. To determine the far field pattern of the array, CST MWS allows the simultaneous excitation of all 38 Ports with a user defined amplitude and phase. To determine the correct resulting far field pattern at 1.4 and 1.7 GHz, only two calculations are required. Alternatively each port might be calculated separately. This would allow the combination of arbitrary port amplitudes and phases in a post processing step. In order to reduce the calculation time for structure with many ports, CST MWS built-in network capabilities might be used. If network computing is selected the simulation task can be distributed in a computer network and each port simulation might run on a different computer. The entire array was discretized with about 5 million mesh nodes. The program automatically adds a 4-Layer PML as radiating boundary condition. The peak memory usage for the solver process was roughly 775 MByte. This included the memory overhead for the program itself as well as for the defined far field monitors. The calculation time for the simultaneous excitation of all ports is about 5h on a 3 GHz Intel XEON double processor computer.
Figure 5 and 6 show the farfield pattern of the array at 1.4 and 1.7 GHz if the previously calulated excitation pattern is used. The pattern is subsequently used as input for a physical optics program to calculated the farfield pattern for the entire dish. The result of this calculation can be seen in figure 7.
The article shows that CST MWS is perfectly suited to handle electrical large structures like FPAs. In this example, only 19 neighbouring elements have been calculated. The memory usage for this simulation was less then 1 GByte. One main advantage of the time domain solver of CST MWS is that the resource requirement only scales linearly with the number of mesh nodes (and therefore the problem size). This demonstrates that it is possible to handle a complete array with more than 100 radiating elements in CST MWS.
 Frank Demming-Janssen, John S. Kot and Christophe Granet (CSIRO ICT Centre), "Analysis of high efficiency reflector feed array using a general-purpose software package", Ninth Australian Symposium on Antennas, Sydney 16-17 February 2005
 Buxton, C.G., “Design of a Broadband Array Using the Four square Radiating Element”