1918 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 7, JULY 2010 Measurements and Error Sources in Time Transfer Using Asynchronous Fiber Network Sven-Christian Ebenhag, Per Olof Hedekvist, Per Jarlemark, Ragne Emardson, Kenneth Jaldehag, Carsten Rieck, and Peter Löthberg Abstract—We have performed time transfer experiments based on passive listening in fiber optical networks using Packet over synchronous optical networking (SONET)/synchronous digital hi- erarchy (SDH). The experiments have been performed with dif- ferent complexity and over different distances. For assessment of the results, we have used a GPS link based on carrier-phase observations. On a 560-km link, precision that is relative to the GPS link of < 1 ns has been obtained over several months. In this paper, we describe and quantify the different error sources influencing the fiber time transfer measurements. We show that the temperature dependence of the optical fiber is the major contribution to the error budget, and, thus, reducing this effect should be the best way of improving the results. Index Terms—Fiber link, fiber network, GPS carrier phase, synchronous optical networking (SONET)/synchronous digital hierarchy (SDH), time transfer, two-way. I. I NTRODUCTION A LTHOUGH GPS supports accurate synchronization, and possibility for traceability, of clocks and oscillators at different locations, it relies on transmission over weak radio sig- nals and, as such, lacks robustness. With increased utilization of accurate time, the need for a complementary time transfer tech- nique has been identified [1], where the synchronization can be performed even in cases where a GPS signal is unavailable. A novel technique for performing time transfer by passively monitoring the data bit stream in an optical fiber has been devel- oped during the recent years [2]–[6]. It is based on synchronous digital hierarchy (SDH) transmission in networks using packet over synchronous optical networking (SONET)/SDH STM-64, 10 Gb/s, between core IP routers. This time transfer method is capable of operating on both aerial and buried SMF-28 fiber (a standard single-mode fiber) and has been implemented in a system with corresponding lengths of dispersion compensation fibers (DCFs). Manuscript received December 22, 2008; revised June 18, 2009; accepted June 19, 2009. Date of publication October 6, 2009; date of current version June 9, 2010. This work was supported by The Swedish National Post and Telecom Agency (PTS). The Associate Editor coordinating the review process for this paper was Dr. George Xiao. S. C. Ebenhag and P. O. Hedekvist are with the SP Technical Re- search Institute of Sweden, 501 15 Borås, Sweden, and also with Chalmers University of Technology, 412 96 Göteborg, Sweden (e-mail: sven-christian. ebenhag@sp.se). P. Jarlemark, R. Emardson, K. Jaldehag, and C. Rieck are with the SP Technical Research Institute of Sweden, 501 15 Borås, Sweden. P. Löthberg is with STUPI LLC, Los Altos, CA 94022-2778 USA. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIM.2009.2028214 Fig. 1. Two types of fiber links. (Top) Single link between two clocks. (Bottom) How the clocks in node A and node C can be compared using the signals going in both directions through node B. The precision and the reliability of time transfer over the fiber link were both evaluated by a GPS link based on carrier- phase measurements. The clock difference was solved for in a Kalman filter as a stochastic process [7], and the positions of both stations were assumed to be known and held fixed in this solution. By comparing the results from the fiber link with the results from the GPS carrier-phase link, an RMS difference between the two methods of less than 1 ns is achieved over six months of measurements. These results are comparable with alternative fiber-based solutions using dedicated bandwidth for transmitting timing signals [8], [9]. Several experiments based on the presented technique are performed. There are two types of setups for the experiment. The first one is between two nodes, and the second is between three or more nodes. In the latter system, the path between nodes A and B is asynchronous to the path between B and C, which requires a slave clock in node B to bridge the paths (Fig. 1). For the purpose of comparison and evaluation of error sources between the fiber and the GPS link used for time transfer, electronic cabinets are equipped with temperature and humidity sensors. The fiber path between the nodes contains amplifier stations with a DCF and an erbium-doped fiber am- plifier. Since temperature influences the time transfer and the temperature in the nodes is under supervision, only the fiber 0018-9456/$26.00 © 2009 IEEE