Research On Combination Method Of International Time Comparison Data Of Timekeeping System

High-precision time comparison technology is an indispensable part for the generation of coordinated universal time(UTC),and it is also a basic means for users that have time synchronization requirements to trace to the national standard time.UTC(NTSC)is one of the important timekeeping systems participating in UTC calculation,and its international comparison link must ensure continuous,stable and reliable operation.The UTC(NTSC)reference system has multiple independent two-way satellite time and frequency transfer(TWSTFT)and GNSS time comparison links.However,a single method is still used in the UTC comparison,and when the link is switched or fails,the reliability is difficult to be guaranteed.Therefore,how to integrate the existing redundant time comparison data to improve the stability and reliability of the international time comparison link is the current research focus.In addition,the multi-GNSS time transfer combination to improve the performance of single-constellation time comparison,as well as the combining GNSS time transfer and TWSTFT to improve the diurnal variations in TWSTFT is a hot topic in the current time and frequency area.This paper uses the existing multiple time comparison technologies of the time reference UTC(NTSC)system to carry out the research of time comparison technology combination through different data fusion algorithms.We study the time comparison principle and error term correction,the establishment of combination model and parameter estimation,and the performance evaluation of combination results,etc.,and use practical examples to verify.The main works and contributions are as follows:(1)The basic principles and performance evaluation methods of time comparison technologies used for fusion processing are introduced.Firstly,the basic principle and error source of TWSTFT,GNSS common view and precise point positioning(PPP)time transfer are sorted out,and the error correction methods corresponding to each method are given.The time link calibration methods,i.e.,direct calibration and indirect calibration,are analyzed.The indirect calibration based on GNSS travelling calibrator and link bridging calibration process and the corresponding uncertainty are discussed.The calculation and verification are carried out by using the actual data.The results show that the Eurasian TWSTFT link is successfully calibrated by GNSS travelling calibrator,and the actual calibration uncertainty is within the current calibration uncertainty range of 1.5 ns.The calibrated PPP link can calibrate GPS CV link through link bridging,and the calibration uncertainty is about 3.0 ns.(2)The combination of BDS CV and TWSTFT based on Vondrak-Cepek combinated filtering is studied.In order to improve the reliability of time transfer links and reduce the diurnal variations in the TWSTFT results,based on the fact that there is no diurnal in BDS CV link,the Vondrak-Cepek combined filtering is applied to combine BDS CV results and SATRE TWSTFT or SDR TWSTFT results over different baseline lengths.Two indexes,time deviation and amplitude spectrum,are used to evaluate the gain factors in reducing the diurnals of the fusion results,and the agreement of the two types of links was revealing by taking GPS PPP time link as a reference.The results show that the diurnals of the combined results are greatly improved after Vondrak-Cepek combined filtering.The amplitude of the spectrum component after combination for 24 hours is improved by at least 84% compared to that before combination.The time deviation of the combined results with an averageing time of 1d is significantly higher than that of SATRE and SDR TWSTFT.The average gain factor for long-baseline NTSC-PTB is 1.83(1 is zero gain),and the average gain factor for short-baseline NTSC-NIM is 1.64;The standard deviation of the differnce between the combined solutions and GPS PPP link(DCD)is also significantly reduced.The standard deviation of the DCD results is reduced from0.95 ns before combination to 0.35 ns on the NTSC-PTB SATRE TWSTFT link,and that is reduced from 1.02 ns to 0.61 ns on the NTSC-NIM SATRE TWSTFT link.The link noise is significantly improved after combination.(3)Based on the “prediction-correction” idea of Kalman filter method,a time transfer combination method is presented in this paper.The advantages of TWSTFT are its high accuracy,characterized by calibration uncertainty,and long-term stability,but studies have shown that TWSTFT is disturbed by diurnal variations.GPS PPP is characterized by its short-term stability and high resolution.However,GPS PPP solutions contain data boundary discontinuities that are reflected in their relatively lower calibration uncertainty.Therefore,the derivative of GPS PPP results is used as the average rate of TWSTFT results,together with TWSTFT data as observation measurements.The final combined solution is then obtained by selecting the optimal parameters using a Kalman filter.The normality of the Kalman filtering observation residuals is tested and analyzed from the quality control aspect.Consequently,the diurnal variations that exist in TWSTFT practically disappear in the combined solution on different baselines,and the short-term stability of the combination is significantly improved.On different baselines,the time deviation gain factor of the combination solution against SATRE TWSTFT is approximately 5 to 8,and that against SDR TWSTFT is approximately 4 to 6 for averaging times less than 32 h.The standard deviation of the double clock differences between the combination link and the reference link is kept within the calibration uncertainty on different baselines,which ensures the accuracy of the combined solution and improves the reliability of time links.(4)Utilizing the characteristics of parallel computing and high fault tolerance of the federated filter algorithm,a multi-GNSS common view time transfer fusion algorithm based on fault-tolerant federated Kalman filter is proposed.With the continuous construction of GNSS systems around the world,GNSS observation data are fully redundant,and multi-GNSS fusion time comparison has become one of the current hot spots.First,the single-constellation CV performance of GPS,Galileo,GLONASS and BDS-2 system is analyzed on long and short baselines.We show that Galileo provides performance that is better than,or similar to GPS,and is better than that of GLONASS and BDS-2 under the current constellation state.Then,taking the TAIPPP link as a reference,the bias between each single-constellation solution and TAIPPP was obtained.After all single-constellation solutions had been aligned to the TAIPPP link,the key parameters of the main filter and sub-filter of federal Kalman filter are assigned by using the characteristics of each link.At the same time,the measurement noise coefficient is introduced into the state equation of Kalman filter to adjust the measurement noise dynamically.When the subsystem fails,the fault is detected and isolated in real time,and finally a better combined solution is obtained.The combined solution is compared with the results of Galileo-only CV,standard-deviation-weighted solutions,and GPS PPP solutions.A large number of examples on different baseline lengths show that compared with other methods,the fault-tolerant federated Kalman combined solution has obvious advantages in reducing the link noise and improving the stability and robustness of the time transfer link.