Towards large-scale calibrations: A statistical analysis on 100 digital 3-axis MEMS accelerometers

Given the growing development and production of low-cost digital MEMS sensors, e.g. accelerometers, gyroscopes, microphones, humidity, pressure and temperature sensors, large-scale measurements are nowadays a possible reality in many different fields, from industry 4.0 to environmental monitoring and smart cities. However, in most of cases, digital MEMS sensors still lack the required metrological traceability needed to provide traceable measurements. As a matter of fact, at present, a preliminary sensitivity value of these sensors is provided by the manufacturers by performing a simple adjustment, without a proper traceable calibration. This is basically due to the impossibility, nowadays, to guarantee large-scale calibration procedures at costs comparable to those of the sensors. For this purpose, it is first of all necessary to know their current technical performances, in terms of sensitivity and associated uncertainties, and then to define possible large-scale calibration methods. In this work, 100 nominally equal 3-axis MEMS digital accelerometers are calibrated with a recently-developed calibration setup at INRiM. Sensitivity values, together with their calibration expanded uncertainties, are compared to statistically analyze their dispersion and distribution within the considered sample. This is the first necessary step towards the development of large-scale calibration methods.