Accelerometer microsystems : from MEMS structure to closed-loop optimization

Abstract

MEMS accelerometer sensors, known for their small size and low cost, have been widely used in consumer electronics, industry, health, and defense. Due to their continuous improvement in noise performance and sensitivity, these sensors have also begun to be employed in seismic applications. State-of-the-art devices that meet these demanding requirements are typically fabricated using customized technologies by research groups and companies, which are not accessible to most MEMS designers. This work presents an optimized MEMS accelerometer fabricated using a publicly available commercial service. This design is reproducible by any MEMS designer and can be adapted for further improvements. The characterized accelerometer device achieves an equivalent input acceleration noise level of 0.89 µg/√Hz, categorizing it alongside advanced seismic-grade accelerometers. Additionally, the design incorporates symmetric biasing functionality for parameter tunability. The inclusion of spare driving electrodes prepares the device for self-calibration and closed-loop operation. Additionally, this thesis details the implementation of a low-noise capacitive readout circuit to interface with the MEMS accelerometer device. Various techniques are employed to reduce circuit noise, including amplitude modulation, digital demodulation, and domain separation. The readout circuit achieves 100kHz bandwidth and a noise level of 0.36 aF/√Hz, surpassing the widely used capacitance readout IC AD7746 from Analog Devices. When integrated with the MEMS sensor, the accelerometer system achieves an exceptional noise level of 2.45 µg/√Hz, positioning this work competitively in both academic and industrial fields. This level of noise performance enables its application in the seismic field. For instance, the system is capable of measuring earthquake signals of magnitude 3.5 at 10 km. Lastly, this thesis proposes a Sliding Mode Control (SMC)-based feedback control scheme with novel nonlinear switching functions. Compared to the widely used Sigma Delta Modulation (ΣΔM) technique, the proposed SMC-based control scheme is optimized based on accelerometer parameters. It achieves a transient time from overloading recovery that is ten times shorter, without compromising the steady-state Signal to Noise Ratio (SNR) performance.