Research of Processing Errors on Modal Frequency of a Dual-Mass Decoupled Silicon Micro-Gyroscope

Di Hu, Bo Yang, Bo Dai2015 · DOI

Abstract

In this paper, the architecture, processing errors analysis and optimization method of a dual-mass silicon micro-gyroscope are presented. The dual-mass silicon micro-gyroscope consists of two identical single mass gyroscopes and a lever mechanism. Driving decoupling springs and sensing decoupling springs are key components to achieve motion decoupling. The processing error analysis indicates that the main influence of processing errors on modal frequency are width error of spring beams. The optimization method which increases the width of spring beams is adopted to reduce the effect of width errors on modal frequency. The structure optimization simulation results demonstrate that the frequency difference variation resulting from the processing errors decreased from 53.45% to 7.08% by increasing the width of spring beams.

Introduction

Silicon micro-gyroscopes (MEMS gyroscopes) are the unsung heroes of modern tech, found in everything from automotive stability systems to smartphone navigation. However, as we push toward "inertial-grade" performance—the high-precision level required for autonomous vehicles or aerospace—manufacturing tolerances become a major bottleneck.

A primary challenge is processing errors during Deep Reactive-Ion Etching (DRIE). In DRIE, the rate at which silicon is etched depends on the width of the groove; narrower gaps etch slower than wider ones. This creates a discrepancy: by the time the narrow gaps of the sensing combs are finished, the wider gaps of the U-shaped structural springs have been "overetched." This overetching thins the spring beams, changing their stiffness and, consequently, shifting the device's resonant frequencies. In a device that relies on precise frequency matching between "driving" and "sensing" modes, even a $0.1\mu m$ error can significantly degrade performance.

Device Architecture

The investigated device is a dual-mass decoupled silicon micro-gyroscope. It features two identical masses connected via a lever mechanism, designed to operate in anti-phase modes to reject common-mode noise. Key structural elements include driving and sensing decoupling springs, which ensure that motion in the drive axis does not interfere with the sense axis.

Architecture of the dual-mass decoupled silicon micro-gyroscope and its lumped parameter model.

Sensitivity Analysis and Processing Errors

The resonant frequency of the gyroscope is highly sensitive to the width of its spring beams. Mathematically, spring stiffness ($K$) is proportional to the cube of the beam width ($w^3$). Consequently, even a minor width error ($\Delta w$) of $0.1\mu m$ can lead to significant frequency shifts. Because stiffness is proportional to the cube of the width, the ratio of actual frequency to ideal frequency is $(1 + \Delta w/w)^{3/2}$, meaning the relative error is amplified by the physics of the beam itself.

Using Finite Element Analysis (FEA) in ANSYS, the researchers simulated the impact of width variations on four modal frequencies: in-phase driving, in-phase sensing, anti-phase sensing, and anti-phase driving. The analysis focused on different components, including base beams and decoupling springs.

Sensitivity of modal frequencies to beam width variations across different structural components.

As shown in Figure 3, the driving and sensing decoupling springs (subplots b and c) exhibit the most significant impact on frequency splits, whereas base beams (subplot a) have a relatively minor effect. When all springs are affected (subplot d), the frequency shifts are most pronounced.

Structural Optimization and Results

To mitigate the impact of overetching, the researchers proposed a geometric optimization strategy: increasing the nominal width of the spring beams. By increasing the baseline width from $10\mu m$ to $15\mu m$ (and adjusting lengths accordingly to maintain the target frequency), the relative error ($\Delta w/w$) caused by a fixed processing error decreases.

The simulation results confirmed the effectiveness of this approach:

  • Sensitivity Analysis: For a standard $10\mu m$ beam, a $0.1\mu m$ processing error caused a natural frequency shift of 45.11 Hz.
  • Error Reduction: The total frequency difference variation dropped from 53.45% with $10\mu m$ beams to just 7.08% with $15\mu m$ beams.
  • Mode Stability: Anti-phase driving frequency variation improved from 6.52% to 4.24%, and anti-phase sensing frequency variation improved from 6.07% to 4.67%.

Conclusion

This research highlights that increasing the nominal width of spring beams is a highly effective strategy for reducing the sensitivity of MEMS gyroscopes to DRIE processing errors. By making the structural components "beefier," the design becomes inherently more robust against manufacturing tolerances, paving the way for higher-performance inertial sensors. Future work will involve physical fabrication and testing to validate these simulation-based findings.

Summary updated

Abstract

In this paper, the architecture, processing errors analysis and optimization method of a dual-mass silicon micro-gyroscope are presented. The dual-mass silicon micro-gyroscope consists of two identical single mass gyroscopes and a lever mechanism. Driving decoupling springs and sensing decoupling springs are key components to achieve motion decoupling. The processing error analysis indicates that the main influence of processing errors on modal frequency are width error of spring beams. The optimization method which increases the width of spring beams is adopted to reduce the effect of width errors on modal frequency. The structure optimization simulation results demonstrate that the frequency difference variation resulting from the processing errors decreased from 53.45% to 7.08% by increasing the width of spring beams.
BibTeX
@misc{hu2015research,
  author = {Di Hu and Bo Yang and Bo Dai},
  title = {{Research of Processing Errors on Modal Frequency of a Dual-Mass Decoupled Silicon Micro-Gyroscope}},
  year = {2015},
  doi = {10.2991/jimet-15.2015.80},
}
Figure