GPS RECEIVER PERFORMANCE TEST AT ROA

Transcription

1 GPS RECEIVER PERFORMANCE TEST AT ROA Hetor Esteban, Juan Palaio, Franiso Javier Galindo, and Jorge Garate Real Instituto y Observatorio de la Armada 00 San Fernando, Spain hesteban@roa.es Abstrat Real Instituto y Observatorio de la Armada (ROA) arries out daily interomparisons of time and frequeny using the GPS satellite system. For this type of links and for high-preision global omparisons, the use of geodesi GPS reeivers is highly reommended, beause it provides the solution using ode P/P (P3) and also arrier phase (PPP). ROA has been operating two different lass of these reeivers: PolaRx (developed by Septentrio, Belgium) sine 005 and GTR50 (developed by DICOM, Czeh Republi) sine 006. Although the last one was initially developed for time and frequeny omparisons using the GPS Common View C/A method, this reeiver an perform ode and phase measurements and, after the issue of the last software version, it has provided the possibility of solution in arrier phase. In this paper, we present the omparison results of three GTR50 reeivers, in two different setups: ommon lok and zero baseline. Also, omparison with PolaRx data are shown and analyzed. I. INTRODUCTION GPS Time and Frequeny transfer is not only the most useful tool for remote loks omparison, and the basis for time international timing laboratories ontributions to International Atomi Time, but also one of the most aurate tehniques in this field. In the ase of Preise Point Positioning (PPP), it s almost at the same level of the state-of-the-art tehnique, two-way satellite time and frequeny transfer (TWSTFT). This last tehnique has, however, the drawbak that it is more expensive than the ones based on GPS, beause it requires renting the use of a ommuniation satellite transponder and usually more time for ontrol operations. For all these reasons, it s interesting to develop and study new geodeti GPS reeivers, whih an perform P3 and PPP solutions in an appropriate and ontinuous way. Moreover, we are thinking of the new Global Navigation Satellite Systems (GNSS) that are under development and deployment and that will provide time and frequeny transfer in a similar way as GPS does. In the next setions, we will first present the geodeti GTR50 for time and frequeny transfer, and the results found during the omparison between three of them and one PolaRx, in four different setups: Common lok of three idential GTR50 s Common lok and antenna of three idential GTR50 s Common lok and antenna of two idential GTR50 s and one PolaRx Common lok of two idential GTR50 s and one PolaRx. 349

2 II. GTR50 RECEIVER This reeiver is basially a Linux PC in a 9-inh hassis together with one GPS reeiver (originally it was a Javad GD and urrently it is a Javad GGD-T) and one time-interval ounter. The Javad GPS reeiver uses its own quartz osillator, supports both ode and phase measurements, and generates one PPS output synhronized to GPS Time. The differene between this PPS and the PPS input referene is measured by the time-interval ounter, that, together with the reeiver iruits and the GPS reeiver, is loated in a thermostatted box (based on thermoeletri Peltier modules) to minimize their delay temperature dependene. The GTR50 was originally intended only for time and frequeny omparisons via GPS ommon views in C/A ode, but new software versions released this year (.5.3. and later versions), also made possible omparisons with the ionosphere-free ode ombination P3 and PPP tehniques. III. COMMON CLOCK OF THREE IDENTICAL GTR50 S In Figure, we an see three similar geodeti GTR50 reeivers (with denominations GTR, GTR, and GTR3), eah of whih has been onneted to its own NovAtel antenna (type GPS-70-GG, with pinwheel tehnology for multipath rejetion and stable phase enter), and was driven by the same highperformane esium frequeny standard (type 507A). PPS NovAtel GPS-70-GG GTR50 0 MHz 507A Figure. First setup of three idential GTR50 s in ommon-lok. The time and frequeny transfer between these reeivers has been omputed initially in ode, by the ommon view (CV) method, using the CGGTTS C/A and P3 daily files generated automatially by eah reeiver. In Figure, we an see the CV results from day of year (DOY) 8 to 35 of the pair GTR-GTR3, where we an observe a standard deviation of resulting data differenes of 60 ps and a orrelation with the external temperature that rounds to 0.04 ns/ºc (Figure 3). The same number is found 350

3 for the GTR-GTR3 pair, but it goes down to ns/ºc for the last possible ombination GTR-GTR. The effet of environmental onditions on antenna eletronis in GPS time reeivers is well doumented [], and an also be deteted learly in terms of Modified Allan variane. Figure. CV CGGTTS C/A GTR-GTR3 and external temperature. Slope ns/ºc Figure 3. Temperature sensitivity oeffiient for the data of Figure. 35

4 For the same period, the CV P3 differenes results are higher (all the setups omputed values are reported in Setion VII) and again two pairs are affeted by temperature (GTR-GTR and GTR- GTR3), with a temperature sensitivity on the same order as for the C/A ode. Finally, we proessed the resulting daily RINEX files using two different sets of PPP software. First, we used GIPSY 5.0 provided by the Jet Propulsion Laboratory, California Institute of Tehnology, released last summer, whih inludes new models for the troposphere mapping funtion, Earth orientation models, and the apability to model seond-order ionosphere effets. Then we used the NRCan 087 software, provided by the Geodeti Survey Division (GSD) of Natural Resoures, Canada. Both analyses have shown that the best performane has values of standard deviation less than 00 ps. A omparison of three time transfer tehniques in terms of frequeny stability is shown in Figure 4. Figure 4. Frequeny stability of three different tehniques for time and frequeny transfer in a ommon-lok onfiguration. IV. COMMON CLOCK AND ANTENNA OF THREE IDENTICAL GTR50 For this setup, we have the same distribution of Figure, exept for using only one Choke Ring Sensor Systems antenna (type SEN CR) that is normally used as the antenna input of one GPS Networking Rak Mount Amplified GPS Splitter (type RMALDCBSX6) and that finally provides the GPS signal to eah reeiver (see Figure 8.). In this onfiguration, we avoid the noise ontribution of eah antenna and we an really hek the reeiver s noise. Although we an t reognize any sensitivity temperature in C/A results for any GTR50 ombination, we reognize a daily phase variation in two of three ombinations, whih must be more losely studied. In Figure 5, we an see the frequeny stability of results from DOY 36 to 4, where the C/A and P3 values have redued signifiantly, remaining onstant for PPP, with respet to the previous setup. Figure 35

5 6 shows the residuals of differene between GIPSY solution and the one obtained with a 5-day rinex bath using NRCan software in bakward smoothed mode. Figure 5. Frequeny stability of three different tehniques for time and frequeny transfer in a ommon-lok and antenna onfiguration. Figure 6. Differenes between GIPSY and NRCAN BWD solutions. 353

6 In this setion, we onlude, remarking the noise type predominant in the three different tehniques. While in C/A and P3 ode we only observe white phase noise, in PPP arrier and ode ombination (Figure 7), we reognize white phase noise for averaging times less than 000 s and fliker phase noise for higher tau values, aused by the typial additive thermal noise of eletroni devies []. Figure 7. Modified Allan deviation for PPP results in ommon-lok and in ommonlok and antenna onfigurations. V. COMMON CLOCK AND ANTENNA OF TWO IDENTICAL GTR50 S AND ONE POLARX For this setup, we use the onnetions diagram showed in Figure 8, with the same antenna signal system mentioned in the preeding setion, two GTR50 s, one PolaRx GPS reeiver, and a Symmetriom Ative Hydrogen Maser (type MHM 00) driven by a Symmetriom Auxiliary Output Generator (type AOG-0). The PolaRx GPS reeiver, the other type of geodeti reeiver available in our laboratory, is haraterized by two main features: its internal lok is driven by an external frequeny (0 MHz) and it aepts PPS input to synhronize its internal lok [3]. Currently, this reeiver onstitutes our seond IGS station, with the denomination of ROAP. In Figure 9, we an see the Time deviation of PPP data proessing for eah pair of reeivers, where the lower noise ombination orresponds with the GTR50 pair, aording to expetations onsidering that we are using twin or almost idential reeivers. 354

7 Symmetriom AOG-0 Septentrio PolaRx Symmetriom MHM 00 Figure 8. Setup of two idential GTR50 s and one PolaRx in ommon-lok and antenna. GTR-ROAP GTR-GTR3 GTR3-ROAP Figure 9. PPP results for the three different ombinations in ommon-lok and antenna. 355

8 VI. COMMON CLOCK OF TWO IDENTICAL GTR50 S AND ONE POLARX In the last setup, we have the same distribution as Figure 8, exept for the fat that ROAP uses the Sensor antenna and eah GTR50 its NovAtel antenna. This onfiguration is the same for GTR50 as it was in the first ommon-lok setup; therefore, the results are on the same order. The standard deviation of the differenes with ROAP is higher, but less than ns, during the 0-day interval. In this setion, instead of looking at the pair differenes, we have studied the short-term noise in one GTR50 reeiver and in ROAP (Figure 0). For the former, we proessed the data using IGS high-rate lok produts (30 s), and we have found worse behavior in GTR50 for averaging times less than 000 s, with predominantly white phase noise, whereas in ROAP, and for these averaging times, we have found predominantly fliker phase noise. White PM Fliker PM Figure 0. PPP solution, REF Clok (H-maser) IGST. 356

9 Figure. CV CGGTTS C/A GTR-GTR3 in ommon-lok onfiguration. We also arried out the test showed in Figure in order to study the behavior of GTR50 after a reboot or shutdown. For this sheme, we used a pair of reeivers, one ontinuously running and the other to be swithed off and powered on. We ran week, swithing off the last reeiver for hour on different days, for a total of four shutdowns. We only observed anomalous behavior hour after the first stop, with a maximum differene of about ns from median. Finally, the standard deviation after the last day was 0.8 ns, a similar result to the one obtained in the first setup. VII. RESULTS To separate the noise ontribution of the three reeivers under test (Table ), with the assumption that these units have similar performane and the behavior is unorrelated [4], we make use of the so-alled three-ornered method for data in Table. We begin with these three ombinations: 3 3 = + = + 3 =

10 Table. Test results of four different setups in ommon-lok (CC) and in ommonantenna (CA). Setup -/POL Δ - -3 Δ -3 3-/POL Δ 3- CC 3 GTR CC & CA 3 GTR CC & CA GTR POL CC GTR POL C/A P PPP C/A P PPP P PPP P PPP Operating in this equation system, we an easily determine the individual stabilities: 3 = = = ( + ) ( + ) ( + ) And applying the law of propagation of unertainty, we an obtain the assoiated unertainties: u u u u N ( y) = u ( x ) i= f x i ( ) = ( u( )) + ( u( )) + ( u( )) ( ) = ( u( )) + ( u( )) + ( u( )) ( ) = ( u( )) + ( u( )) + ( u( )) i where =, 3 3 = and 3 3 = are the sensitivity oeffiients. 358

11 Table. Noise ontribution of eah individual reeiver under test. Setup Δ 3 Δ 3 /POL Δ CC 3 GTR CC & CA 3 GTR CC & CA GTR POL CC GTR POL C/A P PPP C/A P PPP P PPP P PPP VI. CONCLUSIONS In this paper, we have studied the behavior and noise ontribution of three GTR50 GPS reeivers, exploring all different tehniques for time and frequeny transfer in all possible onfigurations. Initially, we found a low level noise between all GTR50 pair ombinations, with a eiling noise lower than 50 ps for zero and short baselines for CV C/A mode. In P3, we have observed a high level of noise, less than 400 ps in the most unfavorable situation. We have also found an external temperature influene for both tehniques in ommon-lok onfiguration, that an give rise to the value of 0.04 ns/ºc, and it appears in a pair of reeivers in C/A ode, whih doesn t imply that it has to happen with P3 ode. In referene to PPP proessing, NRCan 087 and GIPSY 5.0 software have provided similar results, onfirming this tehnique as the most useful tool for time transfer, bringing less noise even for short and zero baselines. Taking the Allan variane results for ommon-lok, we will need averaging times higher than 3 hours to have a frational frequeny stability less than and averaging times higher than day for it to be in parts in 0-6. In regard to the GTR50 versus PolaRx omparison, we deteted dominant white PM noise in PPP proessing for averaging times less than 900 s. This an be due to not using an external frequeny referene in the GPS reeiver or to the intrinsi noise of the time-interval ounter module. Finally, we would also like to emphasize the very stable situation after any shutdown or restart, where we have not deteted time jumps and malfuntions just after the restart or during the following days of operation. 359

12 ACKNOWLEDGMENTS The authors aknowledge Jet Propulsion Laboratory (JPL), California Institute of Tehnology, for providing the Preise Point Positioning (PPP) GIPSY software, and the Geodeti Survey Division, Natural Resoures Canada (NRCan), for providing the PPP software. Speial thanks to François Lahaye (NRCan) for kind support and helpful advie provided on the usage of PPP software. REFERENCES [] W. Lewandowski, J. Azoubib, and W. J. Klepzynski, 999, GPS: Primary Tool for Time Transfer, Proeedings of the IEEE, 87, [] R. Costa, D. Orgiazzi, V. Pettiti, I. Sesia, and P. Tavella, 004, Performane omparison and stability haraterization of timing and geodeti GPS reeivers at IEN, in Proeedings of the 8th European Frequeny and Time Forum (EFTF), 5-7 April 004, Guildford, UK. [3] G. Petit, P. Defraigne, B. Warrington, and P. Uhrih, 006, Calibration of Dual Frequeny GPS Reeivers for TAI, in Proeedings of the 0th European Frequeny and Time Forum (EFTF), 7-30 Marh 006, Braunshweig, Germany, pp [4] C. R. Ekstrom and P. A. Koppang, 00, Degrees of freedom and three-ornered hats, in Proeedings of the 33rd Annual Preise Time and Time Interval (PTTI) Systems and Appliations Meeting, 7-9 November 00, Long Beah, California, USA (U.S. Naval Observatory, Washington, D.C.), pp