Minimum variance method for treatment of measurements with low-frequency noise

A new method is proposed for comparing measurements obtained in different laboratories when the values are correlated by low-frequency noise of the measured device. A minimum variance estimate for the value of the artefact standard during each measurement is calculated from the other measurement values using the real noise characteristics of the standard. This method solves the problem of proper Type A uncertainty treatment as it can deal with any type of noise in a straightforward manner. Results applicable to white, 1/f and 1/f2 noise are presented as examples. The minimum variance method eliminates self-correlation caused by the use of a single reference value. When compared with the use of a single reference value, the method provides a more optimal and impartial analysis of the results of comparisons in which the fundamental noise limit of the transfer standard is approached. Measurements with Zener voltage standards confirm the predictions of the model for 1/f noise.