A comparative study of corrugated horn design by evolutionary techniques

Abstract

Corrugated horn antennas are frequently used as the feed elements in ground-based reflector antennas for satellite and deep space communications. A particular application is the multi-frequency feed horns for the reflector antennas of JPL/NASA Deep Space Network (DSN). In this application, it is desirable to design a horn that has a nearly perfect circularly symmetric pattern (i.e., identical E- and H-plane patterns), with zero or low cross- polarization, and at the same time achieves a specified beamwidth and a low return loss at the design frequency range. A parametric study of the corrugated horns, however, shows that the objective function relating the pattern shape, beamwidth and return loss is a nonlinear function of the corrugation dimensions and has many local optima. As a result one has to resort to global optimization techniques, such as Evolutionary Algorithms (EA) or Genetic Algorithms (GA), for a successful design of these antennas. Here an evolutionary programming (EP) algorithm is used to optimize the pattern of a corrugated circular horn subject to various constraints on return loss, antenna beamwidth, pattern circularity, and low cross-polarization. EP algorithms with different mutation operators including Gaussian, Cauchy and hybrid are applied and the results are compared. Also a hybrid EP-GA algorithm is used for comparison. The EP algorithm is inherently suited for parallelization on a Massively Parallel Computer (MPP), which increases the computation speed by orders of magnitude. Examples of design synthesis for a few corrugated horns are presented. The results show excellent and efficient optimization of the desired horn parameters.