Preliminary evaluation of time scales based on hydrogen masers

Two experimental time scales based on ensembles of hydrogen masers were generated and compared with a variety of other references both internal and external to NIST. The masers all had some type of active cavity control to reduce frequency drift due to cavity changes, The first experimental time scale, TA2M, was generated over the interval from MJD 49050 to 49190 (March 4, 1993-July 22, 1993) using measurements between the masers every 6 hours, TA2M was compared to the NIST AT1 time scale, primarily based on commercial cesium frequency standards, and via once-per-day GPS common-view measurements to the USNO unsteered master clock and to UTC. The linear frequency drift of TA2M relative to UTC was small compared to the uncertainty of 2/spl middot/10/sup -16//d in estimating linear frequency drift. The masers at NIST were then linked to two masers at USNO using GPS common-view time transfer to study the performance of a maser ensemble, TA2M1, at measurement times from a few days to a few months. Again we found that the frequency drift of TA2M1 relative to UTC was smaller than the estimation uncertainty of 0.6/spl middot/10/sup -16//d. The stability of the ensemble at 1 month appeared to be about 1-2 parts in 10/sup 15/. From this work we see that the use of autotuned hydrogen masers in time scales is very promising. A single such maser rivals the 1993 stability of UTC. The typical fractional frequency stability of the masers showed flicker frequency modulation at 4/spl middot/10/sup -15/ or less for measurement times of 10 d-20 d. Linear frequency drift was measured for the masers against the TA2M1 time scale at levels under 1/spl middot/10/sup -16//d with uncertainties of 0.3/spl middot/10/sup -16//d. Measuring frequency drift against the definition of the SI second at this level pushes the limits of current technology.