Numerical-experimental comparison of low-g and high-g tests on a polysilicon MEMS accelerometer

Shock tests are commonly executed on packaged MEMS to investigate and certify the reliability of the device when subject to strong accelerations or severe shocks. It is customary to interpret the results according to a pass/not pass approach, disregarding any analysis of the actual sensor behavior during the test. As a result, it is very difficult to learn the real sensitivity of the device to shock and vibrations. In order to understand and predict (design for reliability) this sensitivity, both in-situ monitoring and a detailed knowledge of the sensor dynamics are mandatory. In this work in-situ measurement results obtained during low-g and high-g experiments (g being the gravity acceleration), carried out on a polysilicon inertial accelerometer mounted on a PCB-board, are compared with numerical simulations. Tests are carried out in the interval 50 to 5500 g. In each test not only the shock waveform and the device response are recorded, but also the MEMS performance is checked. The shock tests are performed on the device attached on top of a falling cylinder in a drop testing machine, where the shock impulse is sensed by a quartz accelerometer mounted next to the tested device. The accelerometer output signal is used as an input for the numerical analysis of the MEMS. A three dimensional, finite element model of the packaged sensor is adopted to simulate its mechanical behavior in the transient regime. Therefore, a comparison with the measured output voltage of the tested MEMS devices allows to single out the effects of mechanical nonlinearities. Work presented in this paper is to our knowledge the first one comparing in-situ measurements of the MEMS response during drop tests with numerical simulations.