How to Use Calibration Errors to Determine Measurement Uncertainty

Abstract : This report illustrates two methodologies to establish measurement uncertainty for a family of digital pressure transducers using calibration data. The lumped method assumes calibration data at each input level are statistically independent and lumps all errors together into a single population without regard to pressure level. Descriptive statistics are used to define population parameters that become the basis for measurement uncertainty. The grouped method groups errors according to input level and establishes separate values of measurement uncertainty corresponding to each input level. Both methods produce comparable results and both offer advantages. The lumped method has the advantage of establishing a single measurement uncertainty for measurements made at any input level. The disadvantage is that measurements made at less than midscale are penalized with a higher measurement uncertainty. The grouped method's advantage overcomes the disadvantages of the lumped method by presenting measurement uncertainty as a percent of reading format. The disadvantage is additional complexity in computing measurement uncertainty as a function of input pressure level.