LAUSR

Temperature-dominated drift is generally the main error source for high-performance micromachined resonant accelerometers (MRAs) due to inherent thermal stress effect of resonator structure and die-attach process. This paper describes the design and experimental evaluation of a temperature compensation scheme for MEMS resonant accelerometers that demonstrates excellent bias and scale factor stability against temperature variation. An on-chip temperature sensor fabricated by sputtering platinum film on glass substrate is proposed to accurately sense the temperature-induced frequency change of the resonator. A polynomial fitting-based post-compensation model is firstly used to suppress the temperature sensitivity of the MRA over dynamic temperature environment. The temperature drift test and compensation of four accelerometer prototypes in a range from −40 ∘C to 60 ∘C show that the stability of bias and scale factor has been improved greatly with navigation-grade performance. Temperature compensation results with three improved drift models based on polynomial fitting, convolutional neural network and support vector regression respectively are presented and compared to suppress the temperature drift hysteresis in consecutive temperature-varying tests. These experimental results indicate that this resonant accelerometer exhibits excellent temperature stability after compensation, which offers the promise for high-performance inertial navigation applications.