Incorporating data from a primary frequency standard into a time scale

It is difficult to combine data from a primary frequency standard with the time-difference measurements that are usually used as input to most time scales because a primary standard usually operates only occasionally on an irregular schedule and because the fundamental output of a primary frequency standard is a frequency rather than a time, and there is often no natural way of inserting this kind of datum into the scale in a manner that is statistically robust. We will present a new time-scale algorithm that seeks to address these problems. We call this frequency-based algorithm AF1, by analogy with the time-based algorithm AT1 that has been used at NIST for many years. Unlike AT1 in which frequency is simply a parameter that specifies how the time of a clock evolves between measurements, however, the frequency of each clock is a fundamental parameter in AF1. This change in focus provides a natural way for incorporating data from a primary frequency standard into the ensemble. We will present the details of the algorithm and results using data from our primary frequency standard, NIST-7.