Validation of IPO: S-duct cavity

Iterative physical optics (IPO) calculations of radar cross section (RCS) have been made on a medium size S-duct cavity for the purpose of benchmarking the method. Reference data has been collected by measuring a physical model of the cavity. The general agreement between the two data sets and derived results like range profiles and ISAR images is good. The method reproduces features whose apparent range indicate that they must be generated though quite complex scattering processes. With the run parameters used (relaxation parameter arel=0.6 and facet size less than half a wavelength) the method appears to converge safely in the present case. Cancellation of contributions from hidden facets appears to be fast, but results corresponding to complex scattering paths do require many iterations. Range profiles and ISAR (inverse synthetic aperture radar) images show some noise-like intensity not present in the measured data. However, the intensity level is rather low. It is tentatively suggested that it is due to "cancellation hum". Work done on improving the IPO code and other software include better options for saving intermediate results, improved iteration speed, better storage of parameter settings, software for CAD and result conditioning. Several of the changes were made to meet immediate needs and would need additional work. One particular advantage of IPO is the possibility to perform calculations which would be too large to address with e.g, MoM (the Methods of Moments). Although MoM can be expected in general to produce more reliable results than IPO, objects such as the cavity studied in this report could not be easily examined with MoM as the memory requirements and computing time would be unreasonable. The results produced by IPO, are not only easily achieved, and they are also encouragingly good. IPO might also be applied to electromagnetic problems other than radar cross section with advantage. The upper frequency or object size (area) is not limited by the IPO method itself but by calculation time, and in the present implementation also by memory. Hardware and software developments including the use of efficient, compiled codes written in e.g. Fortran or C, could allow significant increases in speed and drastically raise the permissible number of facets. Unfortunately the time and in some cases memory demands grow with the fourth power of the frequency, for a fixed object size. If it would be possible to find algorithms permitting larger facets the practical object size limit would increase roughly in proportion to the facet size making such an advance valuable. One possible option for such development is discussed, but more work would be needed to investigate its feasibility. Another possibility is to hybridise IPO with other methods.