Op weg naar dual-mode MEMS-resonante accelerometers voor breedband- en ruisarme sensortoepassingen

Micro-electro-mechanical systems (MEMS) resonant accelerometers are widely regarded for their high quality (Q) factors, frequency stability, and integration potential, making them ideal for precision sensing in applications such as navigation, structural health monitoring, and metrology. However, achieving both high resolution and wide bandwidth within a single device remains challenging, due to intrinsic trade-offs between transduction gain, noise performance, and nonlinear dynamic behavior. This thesis addresses these limitations by developing and experimentally validating a dual-mode transduction framework that co-optimizes signal resolution and bandwidth, while also introducing targeted improvements to each transduction mode. To support this dual-mode approach, the performance of each mode—capacitive actuation and capacitive detection (CACD) and capacitive actuation and piezoresistive detection (CAPD)—is first explored individually. For CACD, a nonlinear compensation method based on electrostatic spring softening is employed to compensate Duffing nonlinearity. This enables dynamic range expansion and SNR improvement without sacrificing bandwidth. Experimental results demonstrate an increase in SNR from 55.9~dB to 68.7~dB and a reduction in noise floor to 0.88~$\mu$g/$\sqrt{\mathrm{Hz}}$, validating the effectiveness of this approach. For CAPD, a thermally boosted sensing scheme is introduced, using localized Joule heating to enhance the transduction sensitivity without the need for external amplification. This results in a noise floor of 0.9~$\mu$g/$\sqrt{\mathrm{Hz}}$ and a bandwidth exceeding 1~kHz, offering a significant improvement in the bandwidth-to-noise floor ratio. Building on these advancements, a dual-mode MEMS SRA is developed by integrating CACD and CAPD transduction modes within a single device. Each mode contributes distinct signal and noise characteristics, and their co-location allows for noise-aware signal fusion. A key innovation lies in exploiting the tunable thermal bias in the CAPD mode to control inter-modal noise correlation. A second-order Kalman filter is then used to perform adaptive signal fusion based on this correlation, achieving significant reductions in noise floor and bias drift, along with improved frequency stability. The resulting system achieves a -3~dB bandwidth greater than 1~kHz, a minimum noise floor of 0.49~$\mu$g/$\sqrt{\mathrm{Hz}}$, and a figure of merit (FOM) of 2347—among the highest reported for MEMS resonant accelerometers. In summary, this thesis presents a high-performance MEMS sensing platform that combines dual-mode transduction and signal fusion with tunable noise correlation. Together, these techniques provide a scalable and robust framework for next-generation inertial sensors capable of overcoming traditional trade-offs between resolution and bandwidth. The proposed system opens new opportunities for intelligent, self-adaptive MEMS devices in demanding environments, offering a path forward in both fundamental sensor design and application-specific performance optimization.