Abstract
MEMS gyroscopes, modeled as spring-mass-damper systems, are highly susceptible to interference near their resonant frequencies. This study compares the impact of mechanical vibrations and high-frequency sound waves on three ADXRS620 gyroscopes. Experimental results show that error outputs increase linearly with interference intensity, reaching up to 88.95 °/s. While the resulting error waveforms are nearly identical for both sound and vibration, mechanical vibration requires significantly less power to produce the same level of interference. To mitigate these effects, an orthogonal demodulation filtering algorithm was developed, achieving up to a 98.88% reduction in vibration-induced errors.
Introduction: The Vulnerability of MEMS Sensors
Microelectromechanical system (MEMS) gyroscopes are critical components in modern technology, ranging from consumer electronics like smartphones and VR headsets to high-stakes applications in drones and automotive safety. These sensors operate as "spring-mass-damper" systems that rely on a specific resonant frequency to detect angular velocity.
The primary weakness of these sensors is resonance interference. When external sources—such as loud acoustic signals or mechanical motors—emit frequencies close to the gyroscope's internal resonance, the sensor can fail to distinguish between actual rotation and external noise. This results in "ghost" signals or total system failure. This research directly compares acoustic and mechanical interference and proposes an algorithmic solution to maintain sensor reliability in noisy environments.
Experimental Methodology
The researchers established two distinct platforms to evaluate interference:
- Acoustic Interference Platform: A signal generator and high-frequency ribbon tweeter were used to subject the gyroscopes to sound waves (2–40 kHz) at volumes up to 125 dB.
- Vibration Interference Platform: A mechanical vibration exciter replaced the tweeter, allowing for precise physical shaking at frequencies between 0 and 18 kHz.
The study tested three commercial ADXRS620 gyroscopes and one custom-developed gyroscope for algorithm verification. To combat the interference, the team implemented an IQ two-channel orthogonal demodulation filtering algorithm. This technique splits the sensing signal into In-phase and Quadrature channels, allowing the system to mathematically "subtract" the interference by comparing its phase to the gyroscope's internal driving frequency.
Comparative Analysis of Interference
The study yielded several key insights into how these sensors respond to external stimuli:
The "Mirror Image" Effect and Resonance Sensitivity
A significant finding was that the error waveforms produced by sound and vibration are nearly identical. This indicates that the physical mechanism of disturbance within the MEMS structure is the same, regardless of whether the energy is transmitted through the air or a physical mount. However, each sensor exhibited a unique "sweet spot" for interference (e.g., 15.33 kHz vs. 13.80 kHz), demonstrating that even identical parts from the same manufacturer possess distinct vulnerabilities.
Efficiency and Linearity
While the effects appear identical, mechanical vibration is far more "efficient" at inducing errors. Due to the rapid decay of sound waves in air, it required a high Sound Pressure Level (SPL) of 90 dB to match the error (approx. 3.4 °/s) caused by a minor 0.050 g mechanical vibration. Furthermore, the error output was found to increase linearly with interference intensity, suggesting that these errors are predictable and modelable.
Key Results and Algorithm Performance
- Maximum Error: Under intense interference, gyroscopes reported false rotations as high as 88.95 °/s, a magnitude sufficient to cause catastrophic failures in autonomous systems.
- Filtering Efficacy: The IQ orthogonal demodulation filter proved highly effective. In tests on the self-developed gyroscope, the algorithm reduced vibration-induced errors by 98.88% when the interference was 4.7 Hz above the driving frequency and 96.31% when it was 5.3 Hz below.
Conclusion and Future Directions
This research highlights the critical need for robust filtering in MEMS applications. While the proposed algorithm is highly effective, future work should address the challenges of implementing such filters on "black box" commercial sensors where internal signals are inaccessible. Additionally, further study is needed to determine the thresholds at which resonance interference causes permanent mechanical damage or "stiction" within the MEMS components.
Summary updated