Accuracy evaluation of the primary frequency standard NIST-7

We develop a comprehensive approach to the evaluation of the uncertainty of a primary frequency standard. Indirect, model-dependent means are employed for the evaluation of all known biases. This approach leads to a lower combined standard uncertainty (CSU) and leaves frequency measurements under altered conditions as a powerful search technique for unknown biases. We report the results of this evaluation technique applied to NIST-7, one of the United States' primary frequency standards. The best CSU value we have reported to the Bureau International des Poids et Mesures (BIPM) is 5 × 10 −15 , while a vigorous application of this evaluation process would suggest that a limiting Type B uncertainty of 3.5 × 10 −15 is possible. We discuss this difference and the implications of reporting optimistic values. We also discuss the history of reported evaluations of NIST-7 including those made before the full implementation of the techniques presented here.