Empirical Assessment and Modelling of RFI Impact on Aviation GPS/SBAS Receiver Performance

Transcription

1 Emprcal Assessment and Modellng of RFI Impact on Avaton GPS/SBAS Recever Performance M. Scaramuzza, P. Truffer, M. Troller, H. Wpf (retred), skygude, Swss Ar avgaton Servces Ltd. H. Lebundgut, REGA M. Bertsch, S. Räm (retred), Swss Ar Force BIOGRAPHY Maurzo F. Scaramuzza receved hs Dploma n Geomatcs n 1995 at the Swss Federal Insttute of Technology (ETH) Zürch. He joned n 1995 the Insttute of Geodesy and Photogrammetry at the ETH Zurch, where he receved n 1998 the doctorate n techncal scences n the feld of satellte based flght approaches and landngs. In 1999 he joned skygude and bult up and led the GSS team. Snce 006 he s head of the expert group on Communcaton, avgaton and Survellance. ABSTRACT Rado frequency nterference (RFI) on L-band affectng GSS recevers mght lead to unacceptable performance degradatons when operatng wth GSS only. Flghts usng GSS based procedures and operated under Instrument Meteorologcal Condtons (IMC), are partcularly concerned about such a threat. Ths paper focuses on the mpact of RFI on GPS/SBAS recevers. Based on a recorded RFI event, the behavor of the recevers estmated protecton level s analyzed. In order to understand the behavor of the HPL, the correspondng algorthm s analyzed n depth. It s shown, that RFI mpacts only few HPL nput parameters. The sze of HPL s only dctated by the ncreased pseudorange nose as well as the number of tracked satellte sgnals. The latter s drectly related to the recevers antenna pattern descrbng the expected Carrer to ose Rato (C/o). Hence, the vehcles atttude plays a relevant role, too. Dfferent smulatons are presented whch descrbe the HPL behavor n a more general way. to -0dB and more, resultng n loss of sgnal trackng of one or more satelltes. Even though a poston can stll be determned under these crcumstances t has to be expected, that the recevers performance s degraded. The mpact s determned by assessng the recorded on-board Horzontal Protecton Level (HPL), the Horzontal Fgure of Mert (HFOM), the number of satelltes used for postonng and the Horzontal Dluton of Precson (HDOP). METHOD FOR DETECTIO OF POTETIAL RFI The method for RFI detecton conssts n nstallng mn quck access recorders (Fgure 1) on board of two dozen helcopters operated by the Swss Ar Force and by Rega, the man Swss Helcopter Emergency and Medcal Servce (Fgure ). Data s recorded from daly operatons. By ths way, large parts of Swtzerland are randomly covered. More than 60 parameters avalable on the helcopter s ARIC 49 buses are constantly recorded. The data s retreved from the onboard GPS/SBAS sensor and consst amongst others of the poston and range doman ncludng, carrer to nose rato (C/o), recever status, performance values etc. Addtonal data stems from the Atttude and Headng Reference System (AHRS) and the Flght Management System (FMS). ITRODUCTIO In [1] a new method was presented on how the RFI stuaton can be assessed over a large regon and durng a long perod. Ths was developed as part of the project called 'Helcopter Recordng Random Flghts' (HRRF) wthn the Swss-wde Implementaton Programme for SESAR-orented objectves (CHIPS). In the course of the project several potental RFI events were dentfed so far. In multple cases the overall degradaton of the GPS C/o reaches values down Fgure 1: Installed Avonca mn Quck Access Recorder. Reprnt IO GSS+ Conference 016, Portland, Oregon

2 Fgure 3: GPS antenna pattern on top of a helcopter fn. The red arrow ndcates the drecton towards an arbtrary GPS satellte. The helcopter s algned wth the topocentrc coordnate system. Fgure : EC635 of the Swss Ar Force (top) and EC145 of Rega (bottom). The GPS/SBAS antennas are mounted on the top of the fn n front of the strobe lght (Courtesy VBS and Rega). Based on measured carrer to nose rato C/o meas, GPS satellte azmuth and elevaton angles as well as roll, ptch and yaw angles of the helcopters, the expected carrer to nose rato C/o exp s derved dependng on the ncdence angle of the GPS sgnal at the antenna. Dfferent flters are appled n order to elmnate unwanted sgnal effects lke C/o degradaton for satelltes n proxmty of the rado horzon. Ths yelds to an antenna pattern for the C/o exp whch s depcted n Fgure 3. It s clearly vsble, that these values vary n a range of more than 15dB. Ths s manly caused by the close envronment of the recever s antenna,.e. the helcopters fuselage etc. For each epoch and for each satellte the C/o exp s subtracted from the measured carrer to nose rato C/o meas whch leads to a normalzed carrer to nose rato for each tracked satellte. Fnally these dfferences are averaged. The resultng value s called average of normalzed C/o an. It s ndependent from effects caused by the helcopters atttude and fuselage dsturbances. Further ths value drectly ndcates by whch amount n db an RFI affects all carrer to nose values. Fgure 4 shows an event over a tme perod of 140 seconds. On the top of the fgure all C/o are degraded smultaneously followed by loss of sgnal trackng of three satelltes. After about 30 seconds of severe RFI the sgnals begn to recover as well as the trackng of the three lost satelltes s restored. The C/o an s depcted on the bottom part of Fgure 4 ndcatng the mpact caused by the RFI,.e. n ths case down to -0dB. Reprnt IO GSS+ Conference 016, Portland, Oregon

3 of M. It appears that the RFI doesn t nfluence the HPL durng the ntal phase of the event. The HFOM shows a smlar behavor lke the HPL. Obvously, the HFOM values are smaller than those of HPL. Fnally, the HDOP values alter wth the changng number of tracked satelltes. Fgure 4: RFI event C/o meas degradaton on all satelltes (top) and C/o an (bottom). EMPIRICAL ASSESSMET OF RFI IMPACT O GPS/SBAS RECEIVER PERFORMACE Further dscusson s focusng on the performance assessment of GPS/SBAS recevers under nfluence of RFI. The man performance parameter beng assessed s the protecton level whch s the contrbutor to the ntegrty. Emphass s gven to the horzontal poston doman whereas the vertcal poston doman can be treated analogously. The protecton level behavor s based on the RFI event shown n Fgure 4. An alternatve llustraton for the HPL behavor s shown n Fgure 6. The x-axs represents C/o an whereas the HPL s on the y-axs. Each colored dot ndcates one epoch. The color tself references to the number of tracked satelltes at the correspondng epoch. The numbers related to the dots show the elapsed tme n seconds snce begnnng of the RFI event. The represented perod covers a tme span of 50 seconds. The begnnng of the RFI event s on the rght bottom sde. In a frst phase, durng the frst 15 seconds, the dots shft towards left wthout alterng the HPL. Wth loss of the frst two satellte sgnal trackngs the HPL starts to ncrease margnally. In a second phase, after roughly 30 seconds, the HPL scales up remarkably and comes back n a thrd phase. Fnally, durng the fourth phase, t returns n a horzontal shft back to the startng pont. Ths behavor s represented n a smplfed manner by the four gray arrows whereas the second and thrd arrows mght be tlted. Fgure 5 shows the correspondng number of tracked satelltes, the horzontal protecton level (HPL), the horzontal fgure of mert (HFOM) and the horzontal dluton of precson (HDOP) durng the perod of RFI. Fgure 6: Trajectory n C/o an HPL space durng a RFI event. Colored dot ndcates one epoch where the color references to the number of tracked satelltes. The numbers related to the dots show the elapsed tme n seconds snce begnnng of the RFI event. PROTECTIO LEVEL ALGORITHM Fgure 5: C/o an, number of tracked satelltes, HPL, HFOM and HDOP durng the RFI event. Loss of satellte trackng occurs only when the sgnal degradaton reaches at least -0dB compared to unaffected sgnals. Satellte trackng s resumed wth dmnshng RFI mpact. The HPL remans constant at M as long as all satelltes are tracked. Only wth loss of sgnal trackng, the HPL starts to ncrease up to a maxmum value For better understandng of the HPL behavor durng an RFI event, t s necessary to take a look at the correspondng algorthm, whch s descrbed n []. The HPL for SBAS s calculated as follows: HPL SBAS = K, d, (1) H proc maj Reprnt IO GSS+ Conference 016, Portland, Oregon

4 wth d maj beng the semmajor axs of the horzontal poston error ellpse (see Fgure 7). The varance σ s a sum of multple zero mean, Gaussan dstrbuted varances,.e. σ = σ + σ + σ + σ, (5), UIRE, tropo, flt, ar Fgure 7: Poston error ellpse wth the maxmum varance of the poston error descrbed through d maj (semmajor axs) and the correspondng varances d and d E n north and east drecton. The K factor s a multpler to nflate the horzontal poston error ellpse n order to acheve a hgh probablty of the true poston beng located wthn the enlarged ellpse. The value of K depends on the currently flown procedure, e.g. K = 6.18 for LAV or K = 6.0 for LAV/VAV, LP and LPV. The K value remans constant through the entre procedure. The semmajor axs d maj s derved as follows: de + d de d d maj + + d = () E wth the varances of the poston error d and d E n north and east drecton, respectvely the covarance d E d d = s = 1 d E = s E E = = 1 = 1 s, σ, σ s E,, σ. σ s the varance of the th satellte pseudorange error. represents the number of satelltes used for the poston estmaton. The varables s E and s are the projectons of σ n east and north drecton,.e. σ s converted from the range doman nto the poston doman. These varables are contaned n the projecton matrx S whch s derved from the least square adjustment for the poston estmaton S T 1 T ( G WG) G W (3) =, (4) where G s the geometry matrx dependng solely on the relatve postons of the satelltes wth respect to the recever antenna and W the weghtng matrx composed by the nverse of the varances σ. wth σ : varance of onospherc delay,,uire σ : varance of tropospherc errors,,tropo σ : varance of fast & long term correcton resduals,, flt σ : varance of arborne recever errors.,ar The varance of arborne recever errors s further decomposed nto: σ = σ + σ + σ, (6), ar, nose, mp, dvg wth σ : varance of nose error,,nose σ : varance of multpath error,,mp σ : varance of dvergence error,,dvg where Θ 10, multpath = e σ, (7) wth Θ beng the th satellte poston elevaton. Let s assume that the exposure duraton of a low to medum power RFI to a GPS/SBAS recever on-board of a flyng helcopter s below 60 seconds. The varaton of most nput parameters for the HPL s mnmal durng the exposure tme and can therefore set to be constant. Followng consderatons are lmted to the assumed exposure tme: The K factor s assumed to be constant durng an RFI event as the probablty of changng the procedure type durng ths perod s neglgble. The satellte geometry alteraton s very small and therefore the varaton of G s neglgble as long as tracked satelltes reman the same. In case of satellte sgnal trackng loss the number of satelltes decreases and alters the geometry matrx G and consequently the HPL too. The varance of onospherc delay can be assumed to be constant due to the fact, that t s derved from the transmtted SBAS messages. Severe onospherc condtons lke scntllatons or large spatal or temporal onospherc gradents are not ncluded n the SBAS onosphere model. Reprnt IO GSS+ Conference 016, Portland, Oregon

5 The varance of the tropospherc delay s derved through a model. The nput parameters such as temperature, pressure, satellte elevaton etc. alter mnmally durng short perods and hence the varance of the tropospherc delay too. Ths varance can therefore be assumed to be constant. respectvely EC145 (bottom). Although the dentcal antenna type s used, the patterns vary remarkably, especally between the azmuths of 10 and 10 at lower elevatons. The dfferent antenna envronment on both helcopters manly causes ths dfference (see fn tp n Fgure ). The varance of long term correcton resduals can be assumed constant too due to the low temporal gradent. The fast term correcton s bascally caused by rapdly changng satellte clock errors and s ndependent from users poston. Therefore ths varaton s ndependent from RFI and consdered constant n order to avod any non-rfi related nfluences on the HPL. The varance of arborne recever errors depends on multpath, dvergence and recever s nose. Accordng to equaton (7), multpath depends on the satellte elevaton, whch changes only mnmal durng a short perod. Hence, multpath s assumed to be constant. The dvergence s manly caused by the onosphere and s ndependent from RFI and consdered constant for further analyss. Fnally the nose s dependent among others on RFI and alters the HPL. To recap, a recever affected by RFI nfluences the HPL only through the 1. number of tracked satelltes, whch n turn has an mpact on the geometry matrx G and weght matrx W.. nose error σ,nose, whch has an mpact on the varance of pseudorange errors σ and consequently on weght matrx W. Fgure 8: Antenna pattern wth C/o exp for the EC635 (top) and EC145 (bottom). Although dentcal antennas are used, larger dfferences are observable n the azmuthal range of 10 to 10 at lower elevaton angles. MODELLIG OF RFI IMPACT O HPL Further dscusson focuses on the number of tracked satelltes and ts mpact on HPL. In a frst step, t s of nterest to estmate the mnmum requred carrer to nose rato C/o mn that enables GPS satellte sgnal trackng. Ths value vares dependng on dfferent factors such as the type of RFI or the behavor of the recever s Automatc Gan Control (AGC). Based on assessments on dfferent track losses recorded durng RFI events, a range of values between 5dB-Hz and 30dB-Hz was observed. These values were reproduced under laboratory condtons wth the use of GPS and RFI sgnal generators and appled to the same type of recever as used on-board of the helcopters. Based on these observatons a value of 8dB-Hz was chosen for C/o mn. The dfference between the C/o mn and C/o exp corresponds to the maxmum tolerable carrer to nose degradaton wthout sgnal trackng loss C / omar = C / o o (8) mn C / ote that the margn C/o mar depends on GPS sgnal angle of ncdence and vares for dfferent helcopter types. Fgure 8 depcts the C/o exp for the EC635 (top), exp The antenna pattern descrbng the margn C/o mar s derved wth equaton (8) and s depcted n Fgure 9 for the EC635. Fgure 9: Antenna pattern wth C/o mar for the EC635. The crcles wthn Fgure 9 correspond to the satellte postons durng the RFI event descrbed n Fgure 4. The red-bordered crcles ndcate the three satelltes from whch sgnal trackng was lost. Ths occurs at a C/o an of -18 to -0dB-Hz, whch perfectly concdes wth the C/o mar n Fgure 9. However, f the helcopter s atttude alters, then the satellte postons are changng as well. By consequence, dfferent satellte sgnals would be lost Reprnt IO GSS+ Conference 016, Portland, Oregon

6 resultng n a varable number of tracked satelltes. In turn, the HPL s altered too. The fndng s that the helcopter s atttude nfluences the HPL. Ths fndng can be consdered comprehensble for large roll angles varaton. Tough, the helcopter headng, whle keepng the roll angle at 0, nfluences the HPL too. Wthn Fgure 9, a headng change would shft all satellte postons by the same amount to the rght or to the left. A smulaton s carred out for the RFI where the helcopter headng s rotated 360 by 10 steps. Further the C/o an s decreased gradually. Frst, the number of tracked satelltes s estmated dependng on the C/o an and C/o mar. ext, the HPL s derved for all helcopter headngs. The requred error varances for the HPL calculatons are retreved from EGOS montor statons operated by skygude for the correspondng day. The varance of the recever s nose s kept constant despte an expected ncrease n course of the RFI event. The result s shown n Fgure 10. satellte postons fnally hold ths satellte probablty densty. The smulaton results are shown n Fgure 11 for ntal stuatons wth 1 (top), 10 (mddle) and 8 satelltes (bottom). The colors ndcate the probablty of occurrence wthn a bn of 0.05dB-Hz wdth and M heght. It results, that for an ntal stuaton wth 1 satelltes the HPL remans constantly below 0.0M for carrer to nose degradatons down to -18dB-Hz. The HPL behavor s smlar for the ntal stuaton of only 8 satelltes, but wth the maxmum HPL remanng below 0.04M. In all cases, the HPL begns to ncrease rapdly for degradatons below -18dB-Hz and stops at -6dB-Hz. Common to all ntal stuatons s that a HPL of 0.1M s rare and requres a carrer to nose degradaton of more than -db-hz. The number of tracked satelltes at a C/o an of -19dB-Hz vares between 8 and 11 whle the HPL vares only between M and M. Despte the number of tracked satelltes vares by three unts, the mpact on the HPL s relatvely small. Ths can be explaned by the generally large number of tracked satelltes, whch keeps the geometry matrx stable. At a C/o an of -3.5dB-Hz the number of tracked satelltes vares between 5 and 8, agan wthn a range of three satelltes. ow, the HPL vares over a larger range between M and ca M. By further decreasng the C/o an the HPL begns to ncrease rapdly. Fgure 10: Smulaton of number of tracked satelltes (bottom) and HPL (top) dependng on helcopter s headng and C/o an. In a more general smulaton, the satellte postons are altered too dfferent stuatons are generated randomly. A constrant s to generate an ntal stuaton wth a constant number of tracked satelltes (.e. 1, 10 and 8) whle keepng the orbt parameters reasonable to reflect the GPS constellaton. Ths s acheved by dervng the satellte poston probablty densty relatve to Swtzerland durng a perod of several years. The randomly generated Fgure 11: Smulaton of HPL dependng on C/o an for rotatng helcopter headng, satellte poston stuatons and ntal stuatons wth 1 (top), 10 (mddle) and 8 satelltes (bottom). Reprnt IO GSS+ Conference 016, Portland, Oregon

7 ote, that the results of these smulatons gve an ndcaton regardng the avalablty of the procedure under a degraded envronment caused by RFI. However, there s no ndcaton on the ntegrty due to the unknown Total System Error (TSE). Further, HPL degradaton due to σ ar sn t taken nto account at ths stage. COCLUSIOS Ths paper has assessed the mpact of RFI on a GPS/SBAS recever n terms of HPL degradaton. Ths work s based on GPS and AHRS data recorded on-board of two dozen helcopters durng several thousands of flght hours. The HPL behavor for a GPS exposed to RFI s dscussed by means of an example. It s observed, that the HPL alters only slghtly despte degradatons on all tracked satelltes down to -18dB. The HPL manly begns to ncrease when sgnal trackng of satelltes are lost and s recovered when satellte trackng s resumed. For better understandng of ths behavor, the algorthm for the HPL calculaton s analyzed. It s shown, that only the number of tracked satelltes and the varance of nose error determnes the HPL under RFI crcumstances. Focusng on the number of tracked satelltes t s further shown, that the GPS recever antenna pattern has a drect mpact on the sze of HPL. Even more, t s shown through smulatons, that the helcopter s headng may remarkably vary the HPL value. More extensve smulatons have shown, that even under strong RFI nfluence the HPL rarely reaches values of 0.1M or more. On the other hand, poston solutons are not possble at C/o degradatons of more than 6dB. Ths fndng vares dependng on the recever, antenna and helcopter type. Further work wll assess the varance of the pseudorange error due to nose when exposed to RFI n order to generate smulatons that are more comprehensve. ACKOWLEDGMETS Ths study has been fnanced by skygude n the frame of the appled research and development programme to mplement future navgaton applcatons (CHIPS). REFERECES [1] M. Scaramuzza, H. Wpf, M. Troller, H. Lebundgut, S. Räm, and R. Wttwer, 015. GSS RFI Detecton: Fndng the eedle n the Haystack. IO GSS+ 015, Tampa, FL, USA, Proceedngs. [] RTCA SC-159, 006. Mnmum Operatonal Performance Standards for Global Postonng System / Wde Area Augmentaton System Arborne Equpment. DO-9D, December 13, 006. Reprnt IO GSS+ Conference 016, Portland, Oregon