Novi pristupi u modelovanju RF MEMS prekidača
MentorMarinković, Zlatica D.
Committee membersMarković, Vera
MetadataShow full item record
Modeling of RF MEMS switches includes modeling of electromagnetic (EM) and mechanical characteristics, which is standardly performed in commercial EM (full-wave) and mechanical simulators. Although these methods provide necessary accuracy, they are generally limited to one analysis for a certain structure, also very computationally and time demanding, especially in the procedures of optimizing the dimensions of the considered switch structure. RF MEMS switch models often lack direct relationships between the geometric parameters of the switch and its EM/mechanical characteristics, which would be used to optimize the characteristics of the switch or circuits that contain them. Precisely for these reasons, this dissertation presents a research whose goal is to develop new approaches for reliable and efficient modeling of the characteristics of RF MEMS switches. Modeling of RF MEMS switches was performed using artificial neural networks. New approaches were developed for modeling EM and... mechanical characteristics: scattering parameters, resonant frequencies, actuation voltages, as well as elements of the equivalent circuit depending on the geometric dimensions of RF MEMS switches. Application of the developed models for the analysis of the sensitivity of the characteristics of capacitive RF MEMS switches is presented, in order to observe the behavior of the switches, i.e. change of resonant frequency and actuation voltage with changes of bridge dimensions which are conditioned by dimensional deviations during switches fabrication. A new inverse modeling approach is presented, which significantly shortens the time required to optimize the characteristics of the switch, i.e. design of switches in accordance with the desired characteristics. The further developed models refer to the modeling of the equivalent circuit elements of an RF MEMS switch depending on the lateral dimensions of the switch bridge. Neural models have been developed for the optimization of circuit elements, i.e. calculation of the equivalent circuit elements for the given lateral dimensions of the switch. This new approach provides significant shortening of the time required to determine the equivalent circuit elements and characteristics of RF MEMS switches. Finally, hybrid inverse models have been developed, aimed for direct determination and optimization of the switch bridge dimensions and the values of the equivalent circuit elements for given characteristics of the switch.