A Comparison between Modeled and Measured ZSU 23-4 at 35 GHz

In recent years several powerful RCS modeling codes like Xpatch (US), RAPPORT (NL), Fermat (FR) or FACETS (UK) have been developed. Their medium and long term goal is to replace tower/turntable measurements on real targets. The advantages of applying RCS modeling codes are many: one does not need the real target, the logistics are much easier, above all, modifications of the target can be accommodated easily, be it the geometry, material properties, or camouflage measures. Thus, all different types of RCS reduction can be implemented and their effects studied. On the other hand, there are still several shortcomings: it is still not clear which degree of fidelity is necessary for the facet model that describes the target shape. Many different material parameters like dielectric constants and conductivities have to be taken into account. Geometrical imperfectness (angular deviations, surface roughness) has to be considered. The latter becomes the more important the closer the dimensions are to the wavelength. Due to all these effects the present situation is such that one cannot yet fully trust modeling results when applied to problems like “automatic target recognition” (ATR), especially the training of ATR algorithms. Rather, it is still of great importance to compare the modeling results to real high resolution measurements (e.g. tower / turntable) in order to verify their accuracy or to identify any shortcomings. For the present analysis, the Dutch RCS modeling code RAPPORT was used on a 600,000 facet model of the ZSU 23-4 air defence tank, provided by ARL /xxxARL/. The computations were performed simultaneously at TNO and at FGAN, the results of different aspect angle intervals combined afterwards. For comparison, an ISAR tower/turntable measurement of the same target, also performed by ARL was used.
The paper is organized as follows: First, the data that are used will be described in detail, especially the RCS modeling code RAPPORT. Next, several approaches to quantitatively characterize the fidelity of the modeling results will be presented. The main conclusion will be that there is a certain agreement between both data sets, but that even for the high number of facets, the modeling results are not yet fully satisfactory for their use in ATR algorithms.