Dimensional optimisation of a MEMS BPSK to ASK converter in SOIMUMPs

Abstract

This work describes how the geometry of a relatively complex micro-electromechanical system is optimised, while taking in consideration process variability, to obtain a robust design. The designed and fabricated MEMS device is able to convert BPSK signaling to ASK. Optimisation of the geometric dimensions towards the design targets was achieved using particle swarm techniques followed by a screening stage which makes use of sequential quadratic programming implemented in MATLAB. This hybrid optimisation method required the development of a mathematical model that could replace computationally expensive finite element analysis. Finite element techniques were only used to validate the mathematical model. The screening step made use of MATLAB's nonlinear constrained optimisation and was instrumental in sieving out the final - broadly optimal - dimensions from the handful of solutions provided by the particle swarm technique. The fabricated device had critical dimensions that were far from those designed by as much as 25% and were at the tolerance limits for the process, however, the design specifications were still met. This is attributed to the screening step that picked a broadly optimal solution providing a robust design. The fabricated device was able to give ASK output with a modulation index of 0.8, an RMS voltage of 0.5 V at a data rate of 3 kHz when both the input BPSK signal and carrier have peaks of 9.2 v and carrier frequencies of 1 MHz.