Type B Uncertainty Analysis of Gravity-Based Determinations of Triaxial-Accelerometer Properties by Simulation of Measurement Errors.

Abstract

We simulated the effects of gimbal-alignment errors and rotational step-size errors on measurements of the sensitivity matrix and intrinsic properties of a triaxial accelerometer. We restricted the study to measurements carried out on a two-axis calibration system using a previously described measurement and analysis protocol. As well as imperfections in the calibration system, we simulated imperfect orthogonality of the accelerometer axes and non-identical sensitivity of the individual accelerometers in an otherwise perfect triaxial accelerometer, but we left characterization of other accelerometer imperfections such as non-linearity for future study. Within this framework, sensitivity-matrix errors are caused by imperfections in the construction and installation of the accelerometer calibration system, but not by the accelerometer imperfections included in the simulations. We use the results of this study to assign type B uncertainties to the components of the sensitivity matrix and related intrinsic properties due to imperfections in the measurement system. For calibrations using a reasonably well manufactured and installed multi-axis rotation stage such as that studied in this paper, we estimated upper bounds to the standard uncertainties of the order of 1×10-5, 2×10-5, 5×10-5, and 2×10-4 for the intrinsic sensitivities, diagonal elements of the sensitivity matrix, off-diagonal elements of the sensitivity matrix, and zero-acceleration offsets, relative to a sensitivity-matrix element of 1, respectively, and 5×10-3 degrees for the intrinsic angles.