Numerical methods for the assessment of absorbed power density from 6G wireless equipment

Sammanfattning

Radio equipment needs to meet regulatory requirements related to radio frequency electromagnetic field (EMF) exposure of the human body. The basic restrictions limiting the local EMF exposure from a radiating antenna are defined in terms of absorbed power density (APD) for frequencies above 6 GHz. This thesis is focused on comparing the accuracy and efficiency of various numerical methods for the evaluation of APD: finite integration technique (FIT), finite element method (FEM), method of moments (MoM), multilevel fast multipole method (MLFMM); and two hybrid approaches: a combination of a Sommerfeld integral approximation and MoM, and a combination of FEM and MLFMM. Some of the computational methods use a volumetric mesh (FIT, FEM), while others use a surface mesh (MoM, MLFMM). For the comparisons, two commercial software are used: 3DS CST Studio Suite and Altair FEKO. The investigations are conducted for four types of antennas: dipole, patch, pyramidal horn with slot array, and cavity-fed dipole array. Moreover, arrays of patch antennas, with different number of elements, are also used in the study. The results presented in the thesis show that in most of the cases the difference in the peak value of the APD evaluated by the used software and methods is within 0.5 dB. The mesh density of the antenna and the phantom, and the phantom size are varied in order to find their optimal values such that the computational burden is reduced as much as possible while keeping a high accuracy in the APD results. The difference in the results calculated by the two formulas given by ICNIRP exposure guidelines: using the Poynting vector and using specific absorption rate (SAR), is found below 0.1 dB. In addition, the APD evaluation from equivalent surface currents on the phantom is checked and gives comparable values to the aforementioned formulas. According to the results from the study, the FIT performs best in terms of computational scalability, while the MLFMM is the fastest and most memory-efficient methods for simple single antennas. The implementation of a neural network (NN) is studied for the use case of decreasing the total simulation time when a large number of simulations need to be performed. Post-processing using NN allows accelerating the evaluation of the APD as a function of the antenna-phantom distance by conducting coarse mesh simulations but improving their accuracy.