Design and optimisation of one-port RF MEMS resonators and related integrated circuits using ANN-based macromodelling approach

Abstract

The authors propose a modelling methodology which combines finite element method (FEM) and artificial neural networks (ANNs) techniques for the design and optimisation of integrated circuits (ICs) incorporating one-port radio frequency microelectromechanical systems (RF MEMS) resonating structures. The FEM is employed to predict the electro-mechanical performance of MEMS resonators. ANNs are utilised to model complex relationships between the physical parameters of the MEMS resonator and the corresponding electrical equivalent circuit parameters. These parameters are extracted directly from the FEM analysis and used for training the ANNs. It is shown that the relative error of the ANN model for a single MEMS device is less than 0.5% while the computational time is more than 40 times reduced compared to the FEM. To extend the capabilities of the proposed methodology, the developed ANN model is implemented into a commercially available circuit simulator, Agilent's Advanced Design System (ADS). This allows a seamless modelling process for the design and optimisation of MEMS resonators embedded ICs at both the device and circuit levels. As an example, a 30 MHz Pierce oscillator with the ANN model of a free-free beam MEMS resonator is optimised within the circuit simulator