New frequency comparisons using GPS carrier-phase time transfer

In October 2002, experiments were conducted to assess the frequency accuracy available from GPS carrier-phase time transfer (GPSCPTT). In these experiments, two GPS receivers, NIST/spl I.bar/A and NIST/spl I.bar/B, were operated in close proximity with one clock, UTC(NIST), serving as the receiver clock for both. Because the same clock was used for both receivers, the resultant time series Clk(NIST/spl I.bar/A)-Clk(NIST/spl I.bar/B) is expected to have a slope (frequency difference) of zero. The experiment yielded mixed results. Data recorded 22-25 October 2002 yielded slopes in the 12-24 ps/d range with both positive and negative signs, indicating that if one averaged long enough, the frequency error obtained from GPSCPTT might average out to zero. However, data recorded 5-9 October 2002 yielded slopes in the 45-77 ps/d range, with all of the slopes having the same sign. Thus, it appears that some sort of systematic frequency error is arising in either the measurement or data-analysis process. Attempts have been made to determine the cause of this systematic error. Potential sources investigated include (a) unequal sampling rates at the two receivers, (b) the analysis technique of fixing satellite-clock corrections to predetermined values rather than estimating them, and (c) errors in the estimation of the tropospheric delay. None of these appear be the root cause of the problem. Future work will include the investigation of site- and receiver-specific effects such as temperature sensitivity and multipath.