Abstract
The stability of the International Atomic Time, TAI is limited by different causes, according to the averaging time τ. For values of τ less than two months, the time comparisons methods have a predominant influence. For larger values of τ, up to a few years, TAI is mainly based on a set of about 100 industrial caesium clocks, and reflects their performances. For the longer term, TAI follows the data of the best laboratory caesium standards.
Although the current accuracy and long-term stability, of about 1 × 10-13 satisfy almost all the users, improvements are desirable, especially for the mean-term stability, which might be increased by a larger influence of the laboratory standards. However a correct statistical handling of the data is made difficult by some systematic differences in the behaviour of primary laboratory standards and industrial caesium clocks, especially by seasonal variations.
It was first considered that the primary standards should be free of seasonal frequency fluctuations. Presently, although no firm conclusion can be reached, we increasingly suspect these devices. The participation of numerous well-protected industrial caesium clocks remains essential in the formation of TAI.