The National Institute of Standards and Technology (NIST) currently operates 35 public network time servers that are located at many different sites in the United States. The servers provide time messages over the public Internet in a number of different formats. The times are derived from the NIST atomic clock ensemble in Boulder and are directly traceable to UTC(NIST). The ensemble of time servers currently receives about 7x106u96 requests per day, and the number of requests is increasing by a compounded rate of about 5% per month. I have continued to work on methods of improving the performance of the ensemble of servers so as to be able to support this increasing demand without a corresponding increase in the number of servers. My current efforts involve studies of replacing the phase lock and frequency lock synchronization algorithms with methods based on Kalman filters, and I will report on my latest results. A method based on Kalman filters has an advantage in principle, since these kinds of algorithms support more sophisticated noise models than can be used by the existing synchronization procedures. However, it is not easy to realize this advantage, because the real noise is generally not stationary and is not well characterized in statistical terms. I have experimented with addressing this problem by adding periodic or quasi-periodic terms to the covariance matrix of the process, and I will show the improvement that I have realized using these methods. The algorithms have two advantages over many other implementations. The more sophisticated noise models mean that the same performance can be realized using a longer interval between requests. This is a significant advantage, since it directly improves the level of service that can be provided with a given number of systems. The ensemble is independent of the GPS satellites and requires no outside antennas. It is therefore more robust and more difficult to jam.
Proceedings Title: Proceedings of 2011 Precise Time and Time Interval Planning and Applications Meeting
Conference Dates: November 14-17, 2011
Conference Location: Long Beach, CA
Conference Title: 43rd Annual PTTI Meeting
Pub Type: Conferences
Automated Computer Time Service, Internet Time Service, Kalman Filter