PUBLICATIONS. Radio Science. Time and frequency transfer system using GNSS receiver RESEARCH ARTICLE /2014RS005460

Transcription

1 PUBLICATIONS RESEARCH ARTICLE Secial Section: Asia-Pacific Radio Science Conference Key Points: The GPS time link accuracy can achieve within 2 ns in long-baseline link 10,000 km The system has less than 25 ns and uncertainty for time and frequency Proosed system rovides unbroken time and frequency link tracing back to SI Corresondence to: J.-L. Wang, darrenwang@cht.com.tw Citation: Wang, J.-L., S.-Y. Huang, and C.-S. Liao (2014), Time and frequency transfer system using GNSS receiver, Radio Sci., 49, , doi: / 2014RS Received 2 APR 2014 Acceted 29 AUG 2014 Acceted article online 6 SEP 2014 Published online 5 DEC 2014 Time and frequency transfer system using GNSS receiver Jia-Lun Wang 1,2, Shi-Yu Huang 2, and Chia-Shu Liao 1 1 National Time and Frequency Standard Laboratory, Telecommunication Laboratories, Chunghwa Telecom Co., Ltd., Yangmei City, Taiwan, 2 Deartment of Electrical Engineering, National Tsing Hua University, Hsinchu, Taiwan Abstract Global Positioning System (GPS) time and frequency transfer is one of the most useful ways for the comarison of remote clocks, and the comarison results are very imortant for the calculation of International Atomic Time and UTC (Coordinated Universal Time). For the timing laboratories, it is necessary to calibrate and eriodically evaluate their time transfer system to ensure the accuracy and long-term stability of their time and frequency comarison results. Once the calibration is achieved, it can be used as a standard for traceable time and frequency measurements. In this aer, we demonstrate Global Navigation Satellite System receiver calibration camaign between the National Time and Frequency Standard Laboratory of Telecommunication Laboratories in Taiwan and the Measurement Standards Laboratory in New Zealand. Two calibration strategies, receiver calibration and the link calibration, are adoted in this work. The receiver calibration is used for evaluating the erformance of the roosed system in domestic traceability network in Taiwan. The link calibration is used for minimizing the total uncertainty budget in calculating UTC. Exerimental results indicate that the exanded time and frequency uncertainty of the roosed system (with a coverage factor of k = 2) are less than 25 ns and , resectively, after 1 day of averaging. The accuracy of GPS time link is reorted to be better than 2 ns in long-baseline link (10,000 km) in Asia-Pacific Zone. 1. Introduction The Global Positioning System (GPS) has been designed for the urose of recise ositioning and navigation for the military and civilian users since the beginning of the 1980s. During the ast several decades it has roven to be a reliable source to disseminate time and synchronise clocks over long distances with high recision and accuracy. The GPS has also roven to be a reliable time transfer system due to a major imrovement in accuracy, recision, and coverage for the time metrology [Cerretto et al., 2010; Defraigne et al., 2007; Lahaye et al., 2010]. In time community, there are various methods used for time and frequency transfer. Basically, the GPS time and frequency transfer is one of the most effective tools for the comarison of the clocks between time laboratories contributing for the realization of International Atomic Time (TAI) and coordinated universal time (UTC). Moreover, the comarison results of these time links need to be calibrated and evaluated eriodically to ensure their calibration accuracy and long-term stability. Presently, there are two basic methods of calibration for time links, which are receiver calibration and the link calibration. A number of studies have shown that TWSTFT (Two-Way Satellite Time and Frequency Transfer) adoting the link calibration can be done with a Tye B uncertainty (ub) of 1 ns [Piesteretal., 2008], and GPS time transfer using receiver-calibration strategy has the ubof5ns[petit et al., 2001]. However for imroving the ub, the link calibration is needed. The conventional receiver-calibration suffers from the higher ub while the link calibration [Esteban et al., 2011; Jiang et al., 2011, 2012] exhibits the otential benefits to reduce the ub which is the dominant art in the uncertainty of UTC-UTC(K) issued in BIPM (Bureau International des Poids et Mesures) Circular T. It is noteworthy that, for the link calibration method, measuring of any cable delay is unnecessary. When conducting the link calibration, we do not measure delays of any interconnecting cables between receiver, antenna, and inut reference ort. It is the reason that uncertainty can be reduced by avoiding error due to different tyes of cable delay measurement. Since 2011, the National Time and Frequency Standard Laboratory of Telecommunication Laboratories (TL) has suorted coordination of Asia-Pacific Metrology Programme (APMP) Technical Committee for Time and Frequency (TC-TF) roject and hosted GPS receiver calibration exercise granted by TC-Initiative roject in WANG ET AL American Geohysical Union. All Rights Reserved. 1171

2 Table 1. GPS Receivers Involved in This Calibration Exercise Laboratory Site Name Receiver Model Receiver Tye TL TWTF ASHTECH Z-XIIT Multichannel/Dual Frequency MSL MSL Tocon Euro-80 Multichannel/Dual Frequency MSL1 Setentrio PolaRx4 PRO Multichannel/Dual Frequency Traveling Receiver TR Setentrio PolaRx4 PRO Multichannel/Dual Frequency In this work, we demonstrate the roosed GNSS time and frequency transfer system based on two calibration concets. The first concet involves receiver calibration which is commonly used by time laboratories to calibrate their GPS facilities. When the calibration is finished, the system can be used as remote time and frequency calibration for recise online alications, which rovides unbroken time and frequency link tracing back to SI through TL in Taiwan. The other concet is the time link calibration which has great ossibility to decrease the ub. Our objective is to conduct the link calibration between TL s and MSL s (Measurement Standards Laboratory) GNSS receivers so as to measure UTC(TL)-UTC(MSL). Some GNSS receivers including the local, remote, and traveling receivers have been used for collecting data. The relevant arameters for the receivers used in the calibration exeriments are summarized in Table 1. At TL, we use an Ashtech Z-XIIT receiver to measure uncalibrated UTC(TL)-GPST, and this receiver is marked TWTF. At MSL, we use a Tocon Euro-80 receiver and a Setentrio PolaRx PRO 4 receiver to measure uncalibrated UTC(MSL)-GPST and UTC (MSL1)-GPST, and these two receivers are marked MSL and MSL1, resectively. Moreover, we use a Setentrio PolaRx PRO 4 receiver as the traveling receiver. For conducting the link calibration, traveling receiver (TR) is circulated between TL and MSL. At the remote site MSL, MSL and MSL1 should be comared with TR for some days by connecting both to a common clock. The total delay of the time link between MSL-TR and MSL1-TR can be calculated. The total delay of the time link between TWTF and TR is also calculated for estimating the link calibration value between TWTF-MSL and TWTF-MSL1. After the link calibration between TL and MSL is finished, the total delay of the time link between TWTF and MSL1 is obtained. This value is the calibration result which is used to directly measure UTC(TL)-UTC(MSL). Therefore, any systematic error between local and remote receivers, the internal and antenna cable delays of the traveling receiver can be canceled out by correcting this total delay. This method can otentially reduce the ub because the measurement of cable delay can result in different values deending on alied method. The structuring of the rest of the aer is organized as follows. First, we describe the calibration scheme for the GNSS receiver. Next we resent the exeriment setu and results, total uncertainty estimation for both strategies, and finally the conclusions. 2. Calibration Scheme Two tyes of calibration scheme will be introduced in this section: (1) receiver calibration and (2) the link calibration Receiver Calibration The receiver calibration is the conventional ways used by BIPM to calibrate GPS facility located in time laboratories. Nowadays, many national laboratories contributing to TAI are still using this tye of calibrations. In BIPM, there are many geodetic receivers to be used for receiver calibration, for examle, Z12-T, PolaRx2, and GTR50. Figure 1a illustrates the definition delays used in Z12-T geodetic receiver. The calibration of this system is divided into six different arts from the antenna to the laboratory reference and listed as follows: X S is the antenna delay. X C is the antenna cable delay. X D is the short cable and slitter delay. X R is the receiver internal delay. X O is the internal reference offset. X P is the 1PPS (one ulse er second) offset from 1PPS-in to the laboratory reference. The simlest way to imlement receiver calibration is to calibrate GNSS time station in common clock and near-zero baseline setu with resect to a reference facility or traveling receiver. However, X D is not always WANG ET AL American Geohysical Union. All Rights Reserved. 1172

3 Figure 1. The definition delays used in (a) Z12-T geodetic receiver and (b) the most general setu of GNSS receivers. used in this rocedure and only four quantities, X S + X R, X C, X O, and X P, are considered. In Figure 1b, where unknown antenna delay X S is absorbed by hardware delay of receiver X R. When conducting a new calibration rocedure at each time, the values of X P and X C need to be measured again. Then the values of internal calibration (X R + X S )inp1andp2gpscodeforthesystem under calibration are comuted by common clock differences. The common clock differences value can be obtained by common-view time transfer technique which uses standardized data collected in CCTF (The Consultative Committee for Time and Frequency) CGGTTS (CCTF Grou on GNSS Time Transfer Standards) format. [Defraigne and Petit, 200] ΔINTDðP1Þ ¼ REFGPSðP1Þ TR REFGPSðP1Þ LR (1) ΔINTDðP2Þ ¼ REFGPSðP2Þ TR REFGPSðP2Þ LR (2) INTDðP1=P2Þ NEW ¼ INTDðP1=P2Þ OLD ΔINTDðP1=P2Þ () where REFGPS(P1/P2) means the time difference values between the reference clock in the laboratory and GPS system time in P1 and P2 code. The notation TR and LR refer to traveling receiver and local receiver, resectively. Once we get the internal calibration values, the new receiver internal values can be corrected according to equation () Link Calibration The link calibration is the concet that measuring of any cable delay is unnecessary. When conducting this calibration, there are at least three receivers involved. Two receivers, L1 and L2, are located in different laboratories and one TR is circulated between the laboratories. After measuring the total delay of the time link between TR and L1 and TR and L2, the link calibration value is obtained. This value can be used to calculate the time scale difference or the calibration oint difference of two reference clock between laboratories. The total delay of GNSS receiver (DX) is the total electronic delay between the hase center of the antenna and the laboratory calibration oint. It is the sum of all the delays of the equiment and the cables on the ath of satellite signal from the antenna to the calibration oint, as shown in Figure 2a. The total delay of the time link here is defined as the difference of the total delay between two GNSS receivers connected to a common clock signal as illustrated in Figure 2b, where the TR1/2 refers to the traveling receiver located at location 1/2, L1/2 is the local receiver at location 1/2, and DXR1/2 and DXL1/2 mean total delay of the traveling receiver and local receiver at location 1/2, resectively. WANG ET AL American Geohysical Union. All Rights Reserved. 117

4 Figure 2. The total delay of (a) GNSS receiver (DX) and (b) the time link. The total delay of the time link is also called common clock difference (CCD) in the following sections. The CCD1 and CCD2 reresent the total delay of the time link between TR and L1 and TR and L2, resectively, and are deicted as follows: CCD1 ¼ DXL1 DXR1 (4) CCD2 ¼ DXL2 DXR2 (5) Then we can get the link calibration value (LinkCAL) from the above equations: The total delay of GNSS receiver DX is LinkCAL L1;L2 ¼ CCD2 CCD1 ¼ ðdxl2 DXL1Þ þ ðdxr1 DXR2Þ (6) DX ¼ REFD INTD CABD (7) where the REFD means the delay between UTC oint at Lab(i) and calibration oint, INTD refers to all the hardware delay including the receiver internal delay, antenna delay and equiment through this ath, and CABD is sum of all the delay of cables. Introducing DX into equation (6) LinkCAL L1;L2 ¼ ðrefd L2 INTD L2 CABD L2 REFD L1 þ INTD L1 þ CABD L1 Þ þ ðrefd R1 INTD R1 CABD R1 REFD R2 þ INTD R2 þ CABD R2 Þ (8) The internal delay and cable delay of TR cancel out in equation (8): LinkCAL L1;L2 ¼ ðrefd L2 INTD L2 CABD L2 REFD L1 þ INTD L1 þ CABD L1 Þþ ðrefd R1 REFD R2 Þ (9) The link calibration value is assumed valid until any change or event haens in any of two setus. It can thus be used to directly measure time scale difference between two laboratories by correcting this value. UTCðL1Þ UTCðL2Þ ¼ REFGPS0 L1 REFGPS0 L2 þ LinkCAL L1;L2 (10) The REFGPS0 refers to uncorrected time difference values between the laboratory reference clock and GPS system time. After correcting by LinkCAL value, any systematic error, internal and antenna cable delays induced by the receiver between two laboratories are removed. The only two values which are the difference of reference delay to the TR are needed to evaluate carefully for a GPS link calibration.. Exeriment Setu and Results The roosed system can rovide timing reference signals, such as 1PPS and 10 MHz clock signal, make time comarison between two sites, and generate time transfer data format er day automatically. It can be used for time synchronization and frequency syntonization alications. The roosed system integrates GNSS WANG ET AL American Geohysical Union. All Rights Reserved. 1174

5 Figure. The roosed system architecture used for remote time and frequency calibration. Figure 4. The diagram of common-view common clock calibration. WANG ET AL American Geohysical Union. All Rights Reserved. 1175

6 Figure 5. Common-view common clock result relative to reference system. dual-frequency receiver, automeasurement hardware, user interface, and ostrocessing software. In the following sections, the detailed descritions of the system architecture using two tyes of calibration method, receiver calibration and the link calibration for the urose of the remote time and frequency calibration alications in domestic traceability network in Taiwan as well as calculating UTC, resectively, are resented..1. Remote Time and Frequency Calibration Exeriment When conducting remote time and frequency calibration alications, the roosed system needs to make receiver calibration as described in the revious section. This is done in TL rior to being shied to the customer. A series of exeriments were erformed in order to verify the erformance of the measurement system. During the duration of the exeriments, the roosed system consists of the receiver facilities located at master site and remote site, resectively, as shown in Figure. The master site includes an Ashtech Figure 6. Time deviation and frequency stability results of short baseline 25 km. WANG ET AL American Geohysical Union. All Rights Reserved. 1176

7 Figure 7. Time deviation and frequency stability results of long baseline 10,000 km. Z-XIIT receiver, choke ring antenna, and a 0 m antenna cable. The slave site comrises Setentrio PolaRx4 PRO receiver, CRG2-DM-mini-GG choke ring antenna, SR620 time interval counter, and a 45 m low loss CFD400 antenna cable and antenna triod. The common-view common clock calibration is used to evaluate systematic error of the designed system as shown in Figure 4. The TWTF receiver is one of the time reference systems at TL. The traveling receiver should be calibrated at TL laboratories rior to being tested on the remote site. The calibration is done by the common clock method, where the traveling receiver and the reference system at TL are both measuring the same clock, a 1PPS signal from the UTC(TL) time scale. Figure 5 shows measurement results of common clock calibration, where the 1 day average delay relative to the system at TL was equal to 4.6 fs, and the eak-to-eak variation is about 4.21 ns used for uncertainty analysis. Table 2. The Performance of GPS Common-View Calibration Systems a Tye Laboratory Time Stability (1 day) Frequency Stability (1 day) Baseline (km) Link NIST 1.5 ns NIST < > NRC NPL 2.4 ns NPL < > USNO NICT.6 ns NICT < > NMIA TL 1.7 ns TL < > MSL a NIST: National Institute of Standards and Technology; NRC: National Research Council; NPL: National Physical Laboratory; USNO: United States Naval Observatory; NICT: National Institute of Information and Communications Technology; NMIA: National Measurement Institute of Australia. WANG ET AL American Geohysical Union. All Rights Reserved. 1177

8 Figure 8. The link calibration scheme between TL and MSL. The short-baseline test was erformed between reference site at TL and remote site at Taiei, and the distance between two sites is about 25 km. The reference site used UTC(TL) time scale as its time reference, and the remote site used cesium oscillator as its time reference. Figure 6 shows time comarison results of short baseline by using GPS common view. The frequency stability σ y (τ) atan averaging eriod of 1 day is about The time deviation σ x (τ) at an averaging eriod of 1 day is about 2 ns. The long-baseline test was erformed between reference site at TL and remote site at MSL in New Zealand, and the distance between two sites is about 10,000 km. Figure 7 shows time comarison results of long baseline by using GPS common view. The frequency stability σ y (τ) atan averaging eriod of 1 day is about The time deviation σ x (τ) at an averaging eriod of 1 day is about 1.7 ns. Table 2 summarizes the erformance of GPS common-view calibration systems develoed by rimary timing laboratories such as National Institute of Standards and Technology (NIST), National Physical Laboratory (NPL), National Institute of Information and Communications Technology (NICT) and TL [Davis et al., 200; Gotoh, 2005; Lombardi et al., 2005; Lombardi and Novick, 2006]. The time comarisons were made by these calibration systems with different baseline distance. According to the exerimental results, the time and frequency stability are aroximately 4 ns and , resectively, at an averaging eriod of 1 day within 10,000 km baseline. The results of common-view measurements of the NICT-National Measurement Institute of Australia (NMIA) link are significantly large due to the very much long baseline. When the length of the baseline increases, the number of satellites in common view decreases and then both the erformance of time and frequency transfer may Table. Mean and SD Values of the Common Clock Difference Measurements at TL C/A ns P/ns PPP/ns CCD1 Mean SD Mean SD Mean SD TR-TWTF (before the tri) TR-TWTF (after the tri) Closure measurement WANG ET AL American Geohysical Union. All Rights Reserved. 1178

9 Table 4. Mean and SD Values of the Common Clock Difference Measurements at MSL C/A ns P/ns CCD2 Mean SD Mean SD TR-MSL TR-MSL very significantly degrade. In comarison to those of calibration systems, the exerimental results of the roosed system are in accord with the results of the revious studies which have tested erformance, desite the fact that these studies used different baseline distance of common-view measurements..2. Link Calibration Exeriment in Asia-Pacific Region In order to examine the link calibration exeriment in Asia-Pacific region, TL roosed GPS receiver calibration exercise in During the eriod of the calibration exercise, the GNSS receivers include local receivers located at TL and MSL, resectively, and TR. The setu of the TR is the same as in the slave site described in section.1. Then TR was circulated and comared to all the receivers between TL and MSL as illustrated in Figure 8. First, we started by measuring common clock difference (CCD1) for a few days at TL. Second, the TR was shied to MSL and made a CCD2 measurement. Finally, the TR equiment was shied back to TL to check the closure measurement results. The CCD results at TL are resented in Table. As shown in the third row, the mean values and standard deviation (SD) between TR and TWTF were calculated in C/A, P, and PPP code before the tri to MSL. And the fourth row in the Table lists the results of measured mean and SD values after the tri. The mean values are CGGTTS common-view data averaged at each standard eoch. This CGGTTS files were generated from receiver-indeendent exchange observation files utilizing the software released by BIPM. The closure measurement indicates around 1 ns deviation from zero when the TR was shied back to TL. It seems to be closely connected to the abnormal behavior in TR and measurement uncertainty induced by time interval counter. Table 4 rovides the CCD results at MSL which shows the mean values and SD in C/A and P code. From Tables and 4, we can derive GPS link calibration values which are summarized in Table Uncertainty Estimation In estimating exanded uncertainty of the roosed system, both Tyes A and B uncertainties evaluation must be made as described in the International Organization for Standardization [IEEE Standards Coordinating Committee 27 on Time and Frequency, 1999]. The exanded uncertainty reresents an interval that contains true value with the required confidence level. It is calculated as combined uncertainty multilied by coverage factor and is defined in BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML [199], so that U ¼ k u c (11) where U is exanded uncertainty, k is coverage factor, and u c isthecombineduncertainty.hereweusecoverage factor k = 2 for 95% confidence level if we assume tye B uncertainty will be at least twice as big as tye A uncertainty and degrees of freedom of tye uncertainty as infinite. The exanded uncertainty estimation of these two calibration exeriments will be introduced in the following sections. Table 5. Calibration and Uncertainty Values for TL-MSL GPS Links GPS LINK GPSCAL CA/ns GPSCAL P/ns TWTF-MSL ± ± 4.26 TWTF-MSL ± ± 1.75 WANG ET AL American Geohysical Union. All Rights Reserved. 1179

10 Table 6. The Exanded Time Uncertainty of the Proosed System Tye Uncertainty Comonent Uncertainty (ns) Coverage Factor Probability Density Distribution Measurement Uncertainty (ns) B Equiment resolution at TL B Equiment resolution at Remote B Ehemeris error B Common-View Common Clock eak-to-eak variation B Antenna coordinate error B Environmental factor B Multiath B Delay error of Ionoshere B Delay error of Atmoshere B UTC-UTC(TL) 160 days measurement uncertainty normal 5.7 B Time interval counter A UTC(TL)-Cs between two sites normal 2.00 Exanded time uncertainty with a coverage factor of k = Remote Time and Frequency Calibration System Uncertainty For analysis of time uncertainty, we use time deviation σ x (τ) value at an averaging eriod of 1 day as Tye A time uncertainty. The short-baseline test result was chosen for conservative estimate of Tye A time uncertainty (U a = 2 ns), and 11 comonents were taken into consideration as Tye B evaluation, which otentially introduce systematic errors. The time uncertainties of Tyes A and B are summarized in Table 6. The combined Tye B uncertainty (U b ) equals 5.66 ns. The exanded time uncertainty U with a coverage ffiffi factor of k = 2, where U ¼ k U 2 2 a þ U b, is equal to ns. For analysis of frequency uncertainty, the frequency stability σ y (τ) was used as Tye A frequency uncertainty. The short-baseline test result was chosen for conservative estimate of Tye A frequency uncertainty (U a = ). Table 7 summarizes the frequency uncertainty comonents of Tyes A and B. The exanded frequency uncertainty with a coverage factor of k =2 is equal to Exeriment results indicate that uncertainty of time and frequency of the designed system can meet ITU-T G.811 standard [International Telecommunication Union, 1997] and requirements for the dominant wireless and ower systems as listed in Table Link Calibration Uncertainty Estimation To estimate the total uncertainty of GPS link calibration value, there are seven items to be considered in the following exression: U GPSLINK ¼ u 2 CCD1 þ u 2 CCD2 þ u 2 B1 þ u 2 B2 þ u 2 B þ u 2 B4 þ u 2 B5 (12) Table 7. The Exanded Frequency Uncertainty of the Proosed System Tye Uncertainty Comonent Uncertainty (ns) Coverage Factor Probability Density Distribution Measurement Uncertainty (ns) B Frequency drift tracing back to BIPM 5.5 E-15/2 1.59E-15 B Uncertainty tracing back to BIPM 1.6E-15 1 normal 1.60E-15 B Hydrogen maser standard affected by temerature 2.0 E-14/2 5.77E-15 B Stability of Hydrogen maser standard (1000s) 2.0E-15 1 normal 2.00E-15 B AOG hase controller affected by temerature 1.0E-14/2 2.89E-15 B Cable affected by temerature 5.0E-14/2 1.44E-14 B SDI distributor (1).0E-15/2 8.66E-16 B SDI distributor (2).0E-15/2 8.66E-16 B Equiment at TL site 4.6E-14/2 1.E-14 B Equiment at Remote site 1.2E-1/2.46E-14 A Frequency measurement between two sites.8e-14 1 normal.8e-14 Exanded frequency uncertainty with a coverage factor of k = E-1 WANG ET AL American Geohysical Union. All Rights Reserved. 1180

11 Table 8. Requirements of Time and Frequency Accuracy for the Dominant Wireless and Power Systems a Alication Frequency Accuracy Time Accuracy GSM N/A CDMA μs(10 μs holdover) WiMAX (FDD mode) N/A WiMAX (TDD mode) μs(25 μs holdover) LTE (FDD mode) N/A LTE (TDD mode) μs intercell Smart Grid DME N/A 1 ms Smart grid PMU N/A <1 μs a PMU: Phasor Measurement Unit; LTE: Long-Term Evolution; WiMAX: Worldwide Interoerability for Microwave Access; CDMA: Code Division Multile Access; DME: Disturbance Monitoring Equiment; FDD: Frequency Division Dulexing; TDD: Time Division Dulexing. where u CCD1 and u CCD2 refer to the statistical uncertainty of common clock difference results at TL and MSL, resectively, the u B1 is taken into account for the closure measurement, the u B2 reresents the 1 PPS delay uncertainty from the local UTC to the inut of GPS receiver at TL, which including uncalibrated time interval counter (TIC) uncertainty (0.5 ns), a TIC jitter during measurement (0.05 ns) and the instability of the distribution equiment at local site (0.1 ns), the u B is the uncertainty of 1 PPS delay at MSL, the u B4 is one of the uncertainty items we account for the instability of the receivers system (0.5 ns) affected by environmental factors, and the u B5 reresents the uncertainty for signal roagation effects which are almost canceled based on near-zero baseline common-view common clock configuration and only multiath errors may result in small contribution to the uncertainty (0. ns for C/A and P). Brief examle for evaluating the total uncertainty of GPS link calibration value is shown in Table 9, and the results of the link calibration and its uncertainty values are listed in Table 5. The total uncertainty value for TWTF-MSL link is about ns, and the overall uncertainty value for TWTF-MSL1 link is better than 2 ns. Here we do not take the results of TWTF-MSL link into account since an abnormal variation was detected during the eriod of calibration at MSL. 5. Conclusions In this work, we demonstrate the GPS calibration camaign with long-baseline link beyond 10,000 km in Asia-Pacific region. The two strategies, receiver calibration and the link calibration, are used to calibrate GNSS receiver for recise time and frequency transfer. The evaluated uncertainty of the link calibration value indicates that the accuracy of GPS time link calibration can achieve better than 2 ns (C/A, P) in long-baseline link. It also shows that GPS receiver calibration using time link calibration rather than traditional receiver calibration method can greatly decrease the ub which is the dominant term of total uncertainty in calculating UTC. On the other hand, we used traditional receiver calibration to calibrate our roosed remote time and frequency calibration system in Taiwan and then evaluated its erformance. Exerimental results indicate that the exanded time and frequency uncertainty of the roosed system (with a coverage Table 9. The Total Uncertainty Values for GPS Link Calibration Between TL and MSL Receivers TWTF-MSL1 Method U GPS_link U CCD1 U CCD2 U B1 U B2 U B U B4 U B5 C/A P TWTF-MSL Method U GPS_link U CCD1 U CCD2 U B1 U B2 U B U B4 U B5 C/A WANG ET AL American Geohysical Union. All Rights Reserved. 1181

12 factor of k = 2) are ns and resectively, which meets timing synchronization requirements for the dominant wireless and ower system alications. Moreover, the roosed system rovides unbroken time and frequency link tracing back to SI through National Metrology Institute (NMI) TL in Taiwan. It is very useful for laboratories or industries who want to establish traceability of their own measurement standards and measuring instruments to the SI without moving their equiment to NMI. There are several exerimental studies that could be undertaken to further exlore these two methods in the near future. For the link calibration technique, TL is going to host calibration camaign again with APMP members, NICT and National Institute of Metrology (NIM) in Beijing, in order to obtain more reliable and objective links results in Asia-Pacific region. For the receiver calibration, a more rigorous test for traceable caability of remote time and frequency will be erformed in fourth generation of mobile telecommunication facilities which have strictly demands of synchronization for reliable transmission of data and real-time service alications. Acknowledgments Data to suort this aer are from National Time and Frequency Standard Laboratory of Telecommunication Laboratories (TL) in Taiwan. The data are available and rovided by TL. A reader could get in touch with Jia-Lun Wang for requesting the data. His address is darrenwang@cht.com.tw. The authors would like to thank Tim-Armstrong of Measurement Standards Laboratory in Wellington, New Zealand, for his assistance in setting u TR at MSL and related affairs of shiment. We also thank APMP technical committee suorting the funding for hosting GPS receiver calibration exercise in Asia-Pacific region. References BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML (199), Guide to the exression of uncertainty in measurement, Geneva: International organization for standardization, 1st ed., Corrected and rerinted 1995, 101 ages, ISBN: Cerretto, G., A. Perucca, P. Tavella, A. Mozo, R. Píriz, and M. Romay (2010), Network time and frequency transfer with GNSS receivers located in time laboratories, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, 57(6), , doi: /tuffc Davis, J. A., M. Stevens, P. B. Whibberley, P. W. Stacey, and R. Hlavac (200), Commissioning and validation of a GPS common-view time transfer service at NPL, IEEE International Frequency Control Symosium (IFCS) and PDA Exhibition Jointly with the 17th Euroean Frequency and Time Forum (EFTF), doi: /freq Defraigne, P., and G. Petit (200), Time transfer to TAI using geodetic receivers, Metrologia, 40, Defraigne, P., P. Banerjee, and W. L. Lewandowski (2007), Time transfer through GPS, Indian J. Radio Sace Phys., 6, Esteban, H., J. Palacio, A. Bauch, D. Piester, G. Cerretto, and T. Feldmann (2011), EURAMET Project 1156, GPSCALEU: Results of two year GPS receiver calibration camaign, IEEE International Frequency Control Symosium (IFCS) and Euroean Frequency and Time Forum (EFTF), doi: /eftf Gotoh, T. (2005), Imrovement GPS time link in Asia with all in view, IEEE International Frequency Control Symosium and Exosition (IFCS), doi: /freq IEEE Standards Coordinating Committee 27 on Time and Frequency (1999), IEEE standard definitions of hysical quantities for fundamental frequency and time metrology Random instabilities, IEEE Standards , doi: /ieeestd International Telecommunication Union (1997), ITU-T. G.811 Timing characteristics of rimary reference clocks. Jiang, Z., G. Petit, F. Arias, W. Lewandowski, and L. Tisserand (2011), BIPM calibration scheme for UTC time links, IEEE International Frequency Control Symosium (IFCS) and Euroean Frequency and Time Forum (EFTF), doi: /fcs Jiang, Z., L. Tisserand, and G. Petit (2012), Tests of the BIPM ortable calibration station - METODE: Measurement of total delay, Proceedings of the 44th Annual Precise Time and Time Interval Systems and Alications Meeting, Reston, Virginia, November 2012, Lahaye, F., G. Cerretto, and P. Tavella (2010), GNSS geodetic techniques for time and frequency transfer alications, Adv. Sace Res., 47, , doi: /j.asr Lombardi, M. A., and A. N. Novick (2006), Remote time calibrations via the NIST time measurement and analysis service, Meas.: J. Meas. Sci., 1(4), Lombardi, M. A., A. N. Novick, J. M. Loez, J.-S. Boulanger, and R. Pelletier (2005), The inter-american metrology system (SIM) common-view GPS comarison network, IEEE International Frequency Control Symosium and Exosition (IFCS), doi: /freq Petit, G., Z. Jiang, P. Moussay, J. White, E. Powers, G. Dudle, and P. Uhrich (2001), Progresses in the calibration of geodetic like GPS receivers for accurate time comarisons, in Proceedings of the 15th Euroean Frequency and Time Forum, Piester, D., A. Bauch, L. Breakiron, D. Matskis, B. Blanzano, and O. Koudelka (2008), Time transfer with nanosecond accuracy for the realization of international Atomic Time, Metrologia, 45(2), WANG ET AL American Geohysical Union. All Rights Reserved. 1182