Performance Evaluation of Global Differential GPS (GDGPS) for ...

Performance Evaluation of Global DifferentialGPS (GDGPS) for Single Frequency C/A CodeReceiversSundar Raman, SiRF Technology, Inc.Lionel Garin, SiRF Technology, Inc.BIOGRAPHYSundar Raman holds a Ph.D. in Electrical and ComputerEngineering from the University of Iowa. Since 2001, hehas been with SiRF Technology, Inc. where his focus isin the development of advanced signal processingalgorithms for High Sensitivity GPS applications. He isinvolved in the development of cross correlation andmultipath mitigation algorithms and very high sensitivitytracking. Prior to SiRF, he was with the cellular handsetgroup within Philips Semiconductors developingadvanced algorithms for IS-95, IS-2000 and WCDMAbased handsets.Lionel Garin is Director of Systems Architecture andTechnology, SiRF Technology, Inc. He has over 25 yearsof experience in GPS and Telecommunication fields.Since 1998, his focus is on non traditional approaches tolocation technology, A-GPS, Distributed Server/ClientArchitecture, Indoor High Sensitivity operations and nonGNSS location technologies. He holds severalfundamental patents on this topic. Prior to SiRF, heworked at Ashtech/Magellan (now Thales Navigation) ashead of Systems Engineering for Surveying GPS andGLONASS receivers. He holds an MSEE equivalentdegree in Digital Communication, Systems Control andEstimation Theory for ENST Paris, France and a B.S. inFundamental Physics from Pierre and Marie CurieUniversity Paris, France.ABSTRACTDifferential GPS is a technique that improves theaccuracy of GPS. In its most common form, localdifferential GPS (LDGPS), high quality referencereceivers are located at known, surveyed locations. Thesereference stations estimate the slowly varying errorcomponents of each satellite range measurement, formcorrections and broadcast these corrections to local userswithin a range of 150 km (typical).In this paper, we will analyze the performance one type ofdifferential GPS referred to as Global Differential GPS(GDGPS) [1]. GDGPS, which is provided by NASA’s JetPropulsion Laboratory (JPL), differs from LDGPS in thatthe corrections are valid globally. The rationale for usingGDGPS over LDGPS is that the corrections are availablefrom a single source and are uniformly valid. This avoidsthe degradation of the quality of corrections with distancefrom reference receivers inherent in LDGPS.GDGPS is primarily geared towards dual frequencyreceivers. In contrast, here, we will apply the correctionsto a single frequency C/A code receiver. We willcharacterize the improvement in horizontal positionaccuracy obtained by using the GDGPS. In GDGPS,global corrections to the broadcast ephemeris and satelliteclock corrections are computed using a global network ofGPS reference sites, namely NASA’s Global GPSNetwork (GGN) which is operated by JPL. Forionospheric delay corrections, we will use their databaseof global grid points of ionosphere electron density. Forpurposes of evaluation, these corrections were providedoffline by JPL. They were then included in themeasurements and post processed to evaluate theimprovement in performance.INTRODUCTIONDifferential GPS is a well established technique toimprove the accuracy of GPS [2]. The improvement inaccuracy arises because certain sources of GPS errorsvary slowly with time and are strongly correlated overdistance. For instance, error components due to incorrectephemeris data, satellite clock, ionosphere andtroposphere can be accurately estimated and cancelledusing a reference receiver at a known location.However, even the nominally correlated errors lose thatcorrelation if they are significantly delayed (temporallydecorrelated) or are applied to a receiver significantly

separated from the reference station(geographically/spatially decorrelated). The performanceof local DGPS (LDGPS) receivers degrade with thedistance from the reference receivers.There are many DGPS techniques and applications. Theseinclude the LDGPS which uses a single reference stationto develop a scalar correction to the code phasemeasurement for each satellite. Another alternative is theWide Area Differential GPS (WADGPS) where anetwork of reference stations is used to form a vectorcorrection for each satellite. The vector consists ofindividual corrections for the satellite clock, threecomponents of the satellite ephemeris and parameters ofan ionosphere delay model. The vector correction is validover much greater geographical areas as compared toLDGPS corrections.In this paper we will consider one form of WADGPSwhich is based on JPL's patented Global Differential GPS(GDGPS) architecture. The reader is referred to the linksprovided in reference [3] for a list of papers detailing theGDGPS architecture. GDGPS takes advantage of thenetwork of reference receivers provided by NASA’sGlobal GPS Network (GGN), which is operated andmaintained by JPL. GPS observables from remote sitesare compressed, packetized and transmitted over theinternet to the processing centers. At the processingcenters the global data is analyzed to produce precise GPSorbits and clocks. These are formatted as corrections tothe GPS broadcast ephemerides, encoded, and areprovided over a variety of communications channels toauthorized users [1,3,4].The fundamental tenet of GDGPS architecture is a statespaceapproach, where the orbits of the GPS satellites areprecisely modeled, and the primary estimated parametersare the satellite state and clock [5]. The advantage ofGDGPS over LDGPS is that the corrections are globallyuniform and seamless. The corrections are provided asvector corrections independent of the user rather than asscalar pseudorange corrections. Users anywhere on theground, in the air, or in near space can have access to theworld's most precise differential corrections.Robustness is achieved through redundancy. Multipleprocessing centers are operated in parallel and thecorrections are transmitted through redundantcommunications channels including the internet.GDGPS is primarily geared towards the dual frequencyusers. Various performance studies have been carried outfor dual frequency receivers [3]. However, relativelylittle analysis has been done on the improvement ofGDGPS for single frequency C/A code receivers. In lightof the above advantages, it is worthwhile to quantify theimprovements for this class of receivers using GDGPSarchitecture.For purposes of evaluation, GDGPS corrections wereprovided offline by JPL. They were then included in themeasurements and post processed to evaluate theimprovement in performance.The measurement errors that were corrected usingGDGPS corrections were1) Ionosphere Delay2) Satellite State (orbit + clock)Below we describe in detail the above types of correctionsthat are provided by JPL using their GDGPS system.SATELLITE STATE CORRECTIONSSatellite State corrections include corrections to satelliteorbit and satellite clock.The satellite orbit corrections are provided in the form ofmore accurate satellite orbit vectors. These vectors areprovided for all satellites at an interval of 1 minute. Theseaccurate orbit vectors are differenced from the broadcastsatellite position vectors at the corresponding times toobtain the errors in satellite orbits.The satellite clock corrections are provided as broadcastclock corrections and the estimated clock corrections. Theestimated value includes the corrections and is providedin km at an interval of 1 second. A “sigma” value whichprovides a formal error associated with the correction isalso provided.For evaluation purposes, the satellite orbit errors areconverted to a pseudorange correction by projecting themonto the line of sight (LOS) from user to the satellite.They are then combined with the satellite clockcorrections.IONOSPHERIC DELAY CORRECTIONSThe GPS signals are delayed in proportion to the numberof free electrons encountered. This delay is alsodependent on the frequency. The ionosphere is usuallyreasonably well-behaved and stable in the temperatezones; near the equator or magnetic poles it can fluctuateconsiderably. Users need to correct the raw pseudorangesfor this ionospheric delay.The method below for removing ionospheric delay fromsingle frequency GPS data is based on a database ofglobal grid maps of ionosphere electron density. Thesemaps are created in real time using input data from the

GDGPS network of GPS receivers. These maps wereprovided at 5 minute intervalsIn the GDGPS implementation, the ionosphere is modeledas a spherical layer 450 km above the earth. The verticaltotal electron content of the ionosphere as a function oflatitude and longitude is provided at 5 minute intervals.The data are presented on a 2 deg x 2 deg grid. The unitsare TECU. To obtain the correction for the GPS signal (atthe L1 frequency) transmitted vertically, interpolate thegrid to the desired latitude/longitude and multiply by thefactor 0.162 meters/TECU.We compare the position accuracy obtained with GDGPScorrections against the position accuracy without thesecorrections. Figures 1 – 4 show the position accuracywith and without GDPGS corrections for variouslocations. Figure 5 quantifies the average improvementover these locations.For a non-vertical GPS ray path, we first determine wherethe receiver/satellite line-of-sight intersects theionospheric layer (pierce points) and evaluate the verticalTEC (with appropriate interpolation). We then multiplythis by an obliquity (which is a function of elevationangle) to account for the signal path length through theionosphere as a function of elevation angle.TROPOSPHERIC CORRECTIONSFor tropospheric corrections, the model as detailed insection A.4.2.4 reference [6] was used.POSITION ACCURACY WITH AND WITHOUTGDGPS CORRECTIONSFigure 2: Plot of position accuracy with and withoutGDGPS corrections (PENC 47.79 deg. N, 19.28 deg. E)Below we quantify the improvement in horizontalposition accuracy using these GDGPS. It is assumed herethat the measurements correspond to a single frequency(L1) C/A code measurement. The test locations arechosen to span a range of latitudes to provide a betterunderstanding of ionosphere delay corrections. The testduration is 24 hours (January 12, 2005) at a samplinginterval of 30 seconds..Figure 3: Plot of position accuracy with and withoutGDGPS corrections (NTUS 1.345 deg. N, 103.67 deg. E)TEMPORAL CORRELATION OF GDGPSCORRECTIONSFigure 1: Plot of position accuracy with and withoutGDGPS corrections (CAGZ: 39.14 deg. N, 8.17 deg. E)The GDGPS satellite state corrections (orbit, clock) areprovided every minute and the ionosphere corrections areprovided every 5 minutes. It is desirable to quantify the

corrections will not be obvious to the user. The urbancanyon/indoor scenarios necessitate effective multipathcancellation in parallel with GDGPS.The application of GDGPS to Assisted GPS (A-GPS) hasseveral distinct advantages. These include the simplicityof the correction data collection and the datadissemination to the end user. Furthermore, thehorizontal accuracy can be brought less than 1 meter (onesigma) even when the reference network is not verydense.ACKNOWLEDGMENTSWe acknowledge the help of Dr. Yoaz Bar-Sever and histeam in JPL for providing GDGPS corrections and theirfeedback on the processed results.REFERENCES1) “An Internet-Based Global Differential GPS System,Initial Results”, Ronald J. Muellerschoen, Willy I.Bertiger, Michael Lough, Dave Stowers and Danan Dong,ION National Technical Meeting, Anaheim, CA, Jan,2000.2) “Global Positioning Systems; Theory and Applications,Volume II”, Edited by B. Parkinson, J.L. Spilker, Jr.3) “The Global Differential GPS (GDGPS) System”,http://www.gdgps.net4) “The development and demonstration of NASA’sGlobal Differential System”, Yoaz Bar-Sever, Ronald J.Muellerschoen, Angie Reichert, ESTC ConferenceProceedings, 2002.5) “Orbit Determination with NASA’s High AccuracyReal-Time Global Differential GPS System”, Ronald J.Muellerschoen, Angie Reichert, Da Kuang, MichaelHeflin, Willy Bertiger, Yoaz Bar-Sever, Proceedings ofION GPS-2001, Salt Lake City, UT, September 2001.6) Minimum Operation Performance Standards for GlobalPositioning System/Wide Area Augmentation System forAirborne Equipment, RTCA/DO-229C, November 28,2001.