Design and Experimental Assessment of Low-Noise Piezoelectric Microelectromechanical Systems Vibration Sensors

The ubiquity of vibration sensors and accelerometers, as well as advances in microfabrication technologies, have led to the development of implantable devices for biomedical applications. This work describes a piezoelectric microelectromechanical systems accelerometer designed for potential use in auditory prostheses. The design includes an aluminum nitride bimorph beam with a silicon proof mass. Analytic models of the device sensitivity and noise are presented. These lead to a minimum detectable acceleration cost function for the sensor that can be used to optimize sensor designs more effectively than typical sensitivity maximizing or electrical noise minimizing approaches. A fabricated device with a $1~\mu \text{m}$ thick, $100~\mu \text{m}$ long, and $700~\mu \text{m}$ wide beam and a $400~\mu \text{m}$ thick, $63~\mu \text{m}$ long, and $740~\mu \text{m}$ wide proof mass is tested experimentally. Results indicate accurate modeling of the system sensitivity up to the first resonant frequency (1420 Hz). The low-frequency sensitivity of the device is 1.3 mV/g, and the input referred noise is 36.3 nV/ $\sqrt {\mathrm{Hz}}$ at 100 Hz and 11.8 nV/ $\sqrt {\mathrm{Hz}}$ at 1 kHz. The resulting minimum detectable acceleration at 100 Hz and 1 kHz is $28~\mu \text{g}/\sqrt {\mathrm{Hz}}$ and $9.1~\mu \text{g}/\sqrt {\mathrm{Hz}}$ , respectively. A brief explanation of the use of the validated cost function for sensor design is provided, as well as an example comparing the piezoelectric sensor design to another from the literature. It is concluded that a traditional single-resonance design cannot compete with the performance of acoustic sensors; therefore, novel device designs must be considered for implantable auditory prosthesis applications.