Problems in estimating some timing uncertainties of commercial frequency and time standards

Abstract

An end user cannot define the different timing errors of current cesium or GPS clocks properly. Conventional satellite-based techniques are able to detect long-term drifts only larger than 2 ns/day, and momentary receiver time deviations of the order of 100 ns can he expected with no practical improvement due to averaging unless extended to one day which precludes a medium-term analysis of local cesium performance. Bias-type errors, reliably found after a three-month test run and seriously hampering synchronization, with less costly user equipment show a typical offset of 200 ns. The generally applied test periods of 20 to 50 days, particularly for drift analysis, seem short since slow 10 ns to 20 ns fluctuations of the timing difference, lasting 20 days, are found. Many laboratories overlook environmental factors during long-term stability tests and thus unreliable results are produced. For a careful analysis of current commercial cesium clocks several colocated high-quality references are needed.