Abstract
The development of MEMS has passed several decades. Compared with the traditional one, the manufacturing of MEMS is similar to the semiconductor processing that make the device become smaller and more accurate. However, there are some uncertainty issues in the manufacturing process, which will influence the accuracy and reliability of the device. This article will take MEMS capacitive type acceleration sensor as an example to discuss the effect of these errors. As expected the influence can be quantified by some simple electrical measurement. Since silicon forms a crystal structure, the Young's modulus is different in each direction. In general, it will simplify the problem by substituting anisotropic material parameters to the isotropic one during the mechanical analysis. This article will make use of commercial finite element software ANSYS® to calculate the natural frequencies of the devices which uses these two kinds of material properties respectively. In this analysis, three-dimensional finite element models of tri-axial MEMS acceleration sensors are built to verify that under what condition replacing the anisotropic material parameters by the isotropic one is a reasonable approximation. In the latter half of this paper, formulas for estimating the effective mass and the distance between the electrodes by measuring natural frequency, pull-in voltage, capacitance values and the voltage under certain conditions is developed. In the manufacturing process of MEMS devices, there may be some difference between manufactured devices and design due to the process uncertainty. By using the measured mass that is obtained from the proposed formula instead of the ideal design value, the calculation of spring stiffness will be more accurate so that the influences of residual stress due to the packing process can have a well assessment. In addition, the difference between the etching result and the design value can be confirmed by measuring distance between the electrodes, so that the geometric shape of the manufactured accelerometer can also be estimated. All the theory is verified by an ANSYS® model of single-axis MEMS acceleration sensor. During the simulation, modal analysis is used to determine the resonant frequency; and by adding the simulation of electromechanical coupling, the pull-in voltage can be found. Finally, the calculation result is compared with the theoretical values to prove the feasibility of these formulas.