2001 IEEE International Frequency Control Symposium and PDA Exhibition zyx COMPARING AND EVALUATING THE PERFOMANCE OF PRIMARY FREQUENCY STANDARDS: IMPACT OF DEAD TIME zy T. E. Parker National Institute of Standards and Technology Time and Frequency Division zyxw 325 Broadway Boulder, CO 80305, USA Abstract Comparing high-performance frequency standards often requires making comparisons over displacements in time andor space. The uncertainties introduced into comparisons made with dead time or with displacements in time are discussed. The estimates of uncertainty are based on the noise characteristics of an ensemble of cavity-tuned hydrogen masers used as a stable frequency reference. Specific examples are given. Key Words: Ensemble, frequency comparisons, hydrogen masers, uncertainties,. Introduction Comparisons with other standards, and comparisons with the same standard operated at different times, are necessary steps in evaluating the performance of a primary frequency standard (PFS). Thus, one often has to deal with both displacements in space and time. Standards in different laboratories can be widely separated around the world. This, in many ways, is a desirable feature since it minimizes common-mode environmental effects, but it also requires the use of long-distance, high-stability time or frequency transfer techniques. The issues of long distance comparisons are discussed in another paper in these proceedings [l]. In the present paper the techniques for handling displacements in time are discussed. Displacements in time necessitate a different set of tools than those for displacements in space. Specifically, comparisons over time require a stable (but not necessarily accurate) frequency reference. In the Time and Frequency Division of the National Institute of Standards and Technology (NIST) we use an ensemble of five cavity- tuned hydrogen masers and four high-performance, commercial cesium standards for this purpose. The ensemble uses a post-processed time-scale algorithm similar to our real-time time scale AT1. Details of this ensemble have been previously discussed [2], [3]. The post-processed scale as calculated is identified as TP171. It has a large frequency offset since it was started from the state of AT1 in April of 1997. When a nominal offset of 483~1O-l~ is removed to bring TP171 close to the rate of International Atomic Time (TAI) the new scale is referred to as ATlE. The name ATlES is used when a frequency offset and a linear slope are removed from TP171. The availability of a highly stable frequency reference has become particularly important at NIST with the development of the new frequency standards that use lasers for optical pumping andlor atom cooling. It has generally not been possible to operate these standards continuously because they are inherently complicated, and the lasers are not yet fully reliable. Consequently the standards are difficult to operate on demand, and evaluations are commonly interrupted by loss of laser lock. As a result it is difficult to have a long continuous run, or to have two standards operate at exactly the same time. However, with a sufficiently stable frequency reference significant dead time can zyxwv be tolerated with no major impact on the uncertainty of an evaluation. Also, when comparing two different standards it is not necessary that the runs overlap perfectly. The impact on the comparison uncertainty is generally significantly larger for time offsets than for dead time, but meaningful comparisons can still be made. The same approach can also be used to compare different runs on the same standard. This is very helpful in evaluating the run-to-run stability of a standard, which is important in determining whether a standard is operating properly. The stability characteristics of ATlE are reviewed here, followed by quantitative examples of the impact of dead time and measurement time offsets using ATlE as a reference. Finally a discussion of how ATlE is used to help evaluate the performance of several primary frequency standards is presented. Frequency Stability of ATlE The short-term stability of TP171 (ATlE) has been discussed in previous papers zyxw [2]-[4]. It is better than lxlO-” at one day, and often approaches 3 ~ 1 0 . ’ ~ at ten days. Currently, these stability levels can be determined only by internal measurements among the masers at NIST, but U.S. Government work not protected by U.S. copyright 57