Piezoelectric MEMS Accelerometers for Sensing Ossicular Vibration

Abstract

The structures comprising the human auditory periphery each perform unique functions to direct incoming sound, provide frequency-dependent gain, match impedance between air and liquids, and decompose the signal into its component frequencies. However, conductive and sensorineural hearing loss can significantly inhibit these functions, necessitating intervention to acquire or restore hearing capability. It is important to address an individual's hearing loss to avoid communication difficulties, cognition decline or hindered development, and social isolation. Technological solutions such as hearing aids, cochlear implants, and active middle ear implants have been developed to mitigate hearing loss. Despite the many historical successes of these devices, the externally-placed components still have significant drawbacks that have motivated research and development of completely-implantable sensing options. Not only do they require specific care and removal for certain activities (sleeping, bathing, exercise), but also necessitate signal processing to address feedback, improve microphone directionality, and reduce wind noise. An implanted system could function continuously and would utilize the outer ear for sound localization and protection from wind noise. The performance of implantable sensors must achieve all desired outcomes to outweigh the primary disadvantage - invasive implantation surgery. Therefore, the ultimate goal of this work is to advance completely-implantable auditory prosthesis systems through improved sensing capabilities.

In this work, we design a piezoelectric MEMS accelerometer as an ossicular vibration sensor. Piezoelectric sensing offers output linearity, low-noise material options, and the ability to interface the sensor with low-power circuits. To design the sensor, we first derive an analytic model. The analytic solution provides the full frequency response of the piezoelectric beam to physical or electrical stimuli, but it does not readily allow for the observation of trends in the minimum detectable acceleration with dimension parameter changes. Hence, an assumed-mode model is used to obtain a closed-form solution of the minimum detectable acceleration of the sensor. With this model, we have enabled full design space analysis and can assess the influence of each design parameter on this key performance metric. Additionally, fabricated devices were tested experimentally to validate the analytic model.

In order to propose a design for this application, we map input sound pressure levels to acceleration values using ossicle vibration data from the literature. Based on hearing thresholds, we then set a minimum detectable acceleration design limit, which can be as low as 0.12 mm/s^2. Throughout the historical development of accelerometers for this purpose, the standard device design has a single resonance. However, with insights gained from this work, we conclude that a single-resonance accelerometer cannot meet the minimum detectable signal requirement across the entire frequency range in a small sensor die volume (less than 2.2 mm x 2.2 mm x 0.4 mm). Therefore, we propose a dual-resonant piezoelectric accelerometer design that incorporates two sensing elements, each with its own sensitivity and resonant frequency (i.e. functional bandwidth). This provides the necessary minimum detectable acceleration improvement over the low-frequency range, while the higher-resonance sensor maintains a wide bandwidth (8 kHz). The proposed sensor design can detect 20 phon equivalent acceleration levels from 100 Hz to 8 kHz. Modeled results also indicate that the proposed sensor design maintains a small die area (1.1 mm x 0.74 mm x 0.4 mm). Thus, the proposed design holds the potential to be the best-performing implantable ossicular vibration sensor in the literature.