The Square Kilometre Array Mid-Frequency Aperture Array (SKA MFAA) uses a high number of similar elements in order to meet astronomical requirements. The Log-Periodic Dipole Antenna (LPDA) is one of the candidate elements for implementation of the MFAA. A single LPDA antenna consists of four sub-elements differentially fed, so only two sub-elements are active for a single excitation. 6.5 million LPDA elements will be required for implementation of the MFAA. Therefore, determination of yield is important for the SKA MFAA as the elements are not modified after manufacturing.
In analysis of yield, this antenna presents a high dimensional problem with 185 and 93 system parameters for the full and half models, respectively. The Non-Linear Partial Least-Squares based Polynomial Chaos Expansion (NLPLS-based PCE) was recently shown as an excellent way of calculating yield for high dimensional electromagnetic structures. Klink et al. validated the use of NLPLS-based PCE for yield analysis of a 37-variable diplexer, requiring only 30 full 3D electromagnetic analysis frequency sweeps instead of thousands required by Monte Carlo (MC) analysis [4].In this paper, NLPLS-based PCE is implemented for estimation of yield and sensitivity of the SKA MFAA LPDA antenna. The yield specification for this problem requires that the reflection coefficient must be below -7 dB across the MFAA's band, i.e., 450 -1450 MHz. The LPDA antenna attained 64 percent yield requiring 80 frequency sweeps while MC analysis required 300 analysis points. Furthermore, for the first time, curvature and rotation of antenna elements are considered in determination of yield.