RF-MEMS reconfigurable self-similar antennas: Design, analysis and measurements. Synthesis using a neural network technique

The research in this dissertation relates several areas of Electrical Engineering. Initially, self-similar Sierpinski gasket antennas are studied. These antennas can radiate with the same characteristics at different frequency bands due to their structural property of self-similarity.; The integration of ohmic contact cantilever series RF-MEMS switches with fractal planar antennas is investigated. The on-demand achievement of additional resonances with similar radiation patterns is accomplished. The switch biasing network is designed and fabricated from a high-resistive material. The rigid Si or GaAs wafers used for the switch structure, impose restrictions on the antenna design that are taken into consideration.; A methodology for designing planar self-similar 1-iteration Sierpinski gasket antennas fabricated on silicon is developed. The capacitive effect of the disconnected triangular patches is analyzed and modeled along with the effect of the finite substrate, using the Method of Moments. The observed frequency scaling is attributed to the difference in the active antenna's area at its various modes with respect to the area of a bowtie. A 40 GHz broadband CPS to CPW transition is developed to resolve the feeding restrictions that the structure imposes. A 1-iteration Sierpinski gasket RF-MEMS reconfigurable antenna on silicon is designed, fabricated and measured and the results validate the simulations. To the best of our knowledge, this is the first successfully implemented, fully integrated RF-MEMS reconfigurable multiple frequency antenna to be reported.; A reconfigurable antenna is measured and used with neural network techniques in antenna simulation and synthesis, in order to facilitate the design process. The analysis is simplified with a trained and generalized MLP neural network. This network can also be used to predict the performance of similar antennas. Next, the state of each switch is determined by the desired S11 through another neural network that clusters all responses of the reconfigurable antenna. The neural network exports all switch combinations that result in an antenna with the desired response and the final choice is made. With the extension of the radiation pattern phase, this network constitutes a novel antenna design and synthesis scheme.