How to type test the coefficient of variation of an indication

Type tests are common methods to verify whether an instrument is suitable for a specific measuring task or not. In several countries, the type test is legally required and if failed the respective instrument cannot be used for legal measurements and, therefore, will not be bought by many customers. One important quantity is the coefficient of variation of the indication, which shall not exceed a given acceptance limit. In order to keep the amount of required measurements manageable, the coefficient of variation is estimated at one or a few test conditions from a limited number of measurements. The current test procedure requires that the measured coefficients of variation shall not exceed the respective acceptance limit at all test conditions. This paper demonstrates that the current test procedure has the unwanted tendency to reject instruments which should pass the test. Based on this realisation, the authors modified the test procedure to achieve proper results. The changes do not require further measurements, only the data evaluation has been optimised.