Development of Mode-matching Techniques to Efficiently Model Multi-mode Horns with Non-PEC Walls

Abstract

In the design and analysis of millimetre wave components imperfections of the guide wall material are frequently overlooked and PEC (Perfect Electric Conductor) behaviour is assumed to simplify the problem. These imperfections could be due to the large yet finite wall conductivity or the surface roughness associated with manufacturing processes. In reality, when non-PEC walls are considered, they have a non-negligible effect on mode attenuation and mode mixing, particularly at guide discontinuities An extension of the well-established Mode-Matching method to include non-PEC materials is investigated in this paper. Simply put, Mode-Matching is a method used to determine the scattering coefficients at guide discontinuities. Here the transverse fields on both sides of the step are “matched” such that we have conservation of complex power for incident modes. The additional boundary conditions imposed by the non-PEC walls are considered as perturbations to the PEC solutions. In uniform guides this opens further channels of mode mixing, as opposed to modes bijectively matching; and introduces mechanisms for attenuation which exist due to the necessary surface impedance on the guide walls. While at junctions, further mechanisms for attenuation exist due to the surface impedance on the overlap region of the guides. A volumetric finite element solver is used as a benchmark for the verification of the method. However, for larger components the simulation time required for the finite element solver becomes impractical. Hence, the Mode-Matching description provides a near perfect description of the effect of the non-PEC walls at a fraction of the computational cost when compared to the finite element solver. We model two types of manufactured horn antennae; a conical cylindrical and corrugated cylindrical horn each with 100+ segments. A PEC solution is used as a benchmark for simulations where the physical surface parameters are included. This highlights the effect of including loss in the guide on the losses in individual horn modes. With the increasing detail of horn geometries, the computational effort and simulation time required to effectively model them also increases. We also discuss methods of parallelisation and hardware acceleration of the Mode-Matching code to deal with the increasing computation demand on electromagnetic simulations. These methods are developed in OpenCL to ensure portability of the software across different hardware architectures.