A Novel L1 and L2C Combined Detection Scheme for Enhanced GPS Acquisition

Transcription

1 A Novel L and LC Combned Detecton Scheme for Enhanced GPS Acquton Cyrlle Gernot, Surendran Konavattam Shanmugam, Kyle O Keefe and Gerard Lachapelle Poton Locaton Navgaton (PLAN) Group, Department of Geomatc Engneerng, Unverty of Calgary, Alberta, Canada. BIOGRAPHY Cyrlle Gernot a PhD. tudent n the Department of Geomatc Engneerng at the Unverty Of Calgary Schulch School of Engneerng. He arrved n Calgary n early 006 to perform a x month nternhp n the PLAN Group under the upervon of Dr. Gerard Lachapelle. Th tranng perod concluded 5 year of general engneerng educaton wth year of preparaton chool and 3 year at Telecom INT and eventually ended up wth a propoton to tay n Calgary a a MSc. tudent. After a few month n the PLAN Group, he fnally tranferred from MSc. to PhD. at the begnnng of May 007. He expect to complete h PhD. n September 009. Surendran K. Shanmugam a PhD canddate n the Department of Geomatc Engneerng at the Unverty of Calgary. He receved h MSc n Electrcal Engneerng at the ame unverty n 004 followng the bachelor n Electronc and Communcaton Engneerng, at the Anna Unverty (Inda) n 00. H reearch nteret nclude hgh-entvty GPS, ground baed wrele locaton and generally n the area of communcaton theory and tattcal gnal proceng. Kyle O Keefe an Atant Profeor of Geomatc Engneerng at the Unverty of Calgary. He completed PhD and BSc. degree n the ame department n 004 and 000. He ha worked n potonng and navgaton reearch nce 996 and n atellte navgaton nce 998. H major reearch nteret have been n GNSS ytem mulaton and aement, pace applcaton of GNSS, carrer phae potonng, and local and ndoor potonng wth ground baed rangng ytem. Dr. Gérard Lachapelle a profeor n the Geomatc Engneerng at the Unverty of Calgary. He hold a CRC/CORE char n wrele locaton. He ha been nvolved wth GNSS reearch, development and applcaton for the pat 5 year and ha authored/coauthored numerou related publcaton, oftware and lcene. H prmary reearch nteret are n the area of potonng, locaton and navgaton. More nformaton avalable on the PLAN Group webte, ABSTRACT Degraded gnal envronment uch a ndoor and urban canyon typcally lmt the cope of tandard GPS recever operaton. For ntance, GPS gnal can be attenuated by 0 db or more under thee advere condton. In degraded gnal envronment, enhancng the proceng gan by 0 db or more requre a gnfcant ncreae n coherent ntegraton. However, the coherent ntegraton tme decvely lmted to le than 0 m n the legacy GPS L gnal due to the preence of navgaton data modulaton. Bede, the detecton of weak PRN gnal alo an ue owng to the lmted correlaton uppreon performance of the L C/A code. Hence, the legacy GPS currently beng modernzed to nclude mproved gnal at L and L5 frequence. In partcular, GPS LC gnal not only attractve due to t mmedate avalablty but alo for t role n cvlan GPS applcaton. The ntroducton of new gnal tructure necetate the development of new acquton algorthm. In th paper, we ntroduce a novel detecton cheme that collectvely utlze the legacy GPS L C/A and LC CM code gnal to enhance the acquton entvty. The propoed acquton cheme derve t major mpetu from the fact that the data, code and carrer are coherently related on the GPS L and L gnal for a gven PRN atellte. Secondly, t alo ue both the non-coherent and dfferental detecton output n t mplementaton for further entvty enhancement. A modfed degn of the developed acquton cheme ntroduced, whch readly allow for an effcent mplementaton. The acquton entvty performance of the propoed acquton cheme wa evaluated along wth the tradtonal non-coherent acquton cheme ung hardware mulated and lve GPS gnal. Prelmnary analy corroborate the acquton entvty ION GNNS 007, Fort Worth TX, 5-8 September 007 /

2 mprovement of the propoed acquton cheme over tradtonal non-coherent acquton. INTRODUCTION Degraded gnal envronment uch a ndoor and urban canyon typcally lmt the cope of GPS recever operaton. For ntance, GPS gnal can be attenuated by 0 db or more under thee advere condton. In degraded gnal envronment, enhancng the proceng gan by 0 db or more requre a gnfcant ncreae n coherent ntegraton tme. Hgh entvty GPS (HS GPS) recever utlze ether atance nformaton or mave correlator archtecture to accomplh enhanced entvty. For example, ated GPS (AGPS) typcally utlze Doppler and tmng atance to extend the coherent ntegraton beyond navgaton data bt duraton. Whle t doe not requre extenve hardware reource, t lmted by network avalablty and the relablty of the atance nformaton. On the other hand, tand-alone HS GPS ue extenve hardware reource to ncreae the dwell perod. However, they can tll be lmted n term of coherent ntegraton perod. Conequently, GPS now beng modernzed wth new gnal at L (for cvlan) and L5 (for avaton) frequence. The new gnal tructure are pecfcally degned to overcome the lmtaton faced by legacy GPS L C/A gnal. The preence of plot channel of crtcal mportance a t allow for long coherent ntegraton perod. More pecfcally, the GPS LC of great nteret owng to t mmedate avalablty and for t role n cvlan GPS applcaton. The new LC gnal ha data and plot component. The data channel cont of a moderate rangng code namely the CM code, whch 0,30 chp long and clocked at 5.5 khz. The data channel further modulated wth the 50 Hz navgaton data. On the other hand, the plot channel cont of long rangng code, the CL code, whch alo clocked at 5.5 khz. The long CL code 77,50 chp long and offer ubtantal correlaton uppreon (~ 44 db). However, LC lmted to a ngle b-phae carrer a t ha to hare the frequency wth the mltary P(Y) code. Therefore, the CM and CL code are tme multplexed at.03 MHz. The tranmt power of LC gnal.5 db lower than that of L C/A and further hared equally between the data and plot channel. The preence of data-le channel and mproved correlaton properte beneft the GPS LC gnal n term of gnal acquton and trackng. Acquton of the LC gnal normally carred n two tage. The frt tage nclude a coare acquton of frequency offet alongde etmaton of the CM code phae. The econd tage nclude fner frequency acquton alongde CL code phae etmaton. It hould be emphazed here that the acquton of CM code tll lmted by 0 m data ymbol tranton. On the other hand, for tronger gnal, L C/A ated L CM acquton more derable a t reduce the CM code earch pace. Lm et al. (006) propoed a fat acquton cheme of LC gnal through the ad of L to accelerate CM code phae and frequency offet etmaton. On the other hand, Pak (004) developed a FFT-baed acquton cheme to acqure L CM and CL code under weak gnal condton wherea Yang (005) nvetgated acquton technque on L CM alone, L CL alone and poble jont acquton of CM/CL. Currently, three atellte are tranmttng LC a well a the uual L C/A gnal, namely PRN, PRN 7 and PRN 3. Snce the two gnal are tranmtted from the ame atellte, mot of the atellte and recever error are now correlated. Moreover, accordng to the Interface Control Document (ICD), the two code are ynchronzed and even though the 0 m long data bt are dfferent on both channel, they are tll algned n term of tmng. Secondly, the relaton between the Doppler on L and the Doppler on L hould mply be the frequency rato between L and L. PROPOSED ACQUISITION METHOD The propoed acquton method degned to etmate both L C/A and L CM code delay and Doppler frequency whle keepng the frequency bn ze a mall a poble by ung only m coherent acquton. A llutrated n Fgure, the number of frequency bn to conder a functon of the coherent ntegraton tme ued durng the acquton proce. Fgure : effect of the coherent ntegraton tme on the frequency bn ze to conder In order to ue a 0 m coherent acquton, a mnmum of 300 frequency bn ha to be ued to cover the entre Doppler frequency doman where a only 5 are neceary when a m coherent ntegraton ued. Therefore, f no a pror nformaton on the Doppler avalable, the pace earch to conder greatly enlarged. ION GNNS 007, Fort Worth TX, 5-8 September 007 /

3 However, n order to keep the frequency bn a large a poble, the gnal power requred ha to be farly hgh. The propoed acquton tre to make ue of the fact that through the recent avalablty of LC gnal, the power avalable to cvlan uer ha been ncreaed. Indeed, nce the noe tattc at the output of a non-coherent detector and dfferental detector are dfferent, the detecton output can be potentally combned to mprove the detecton performance. For ntance, the non-coherent and dfferental detecton output from both L and LCM gnal can be combned. However, uch an mplementaton would requre four multplcaton followed by three addton to obtan the fnal detecton output. In contrat, the acquton technque preented n th paper utlze only two multpler and fve adder to obtan the ame detecton output (fgure ). Fgure : propoed acquton cheme The combned L/L acquton preented n th paper make ued of the three aumpton already mentoned durng the frt part:. the navgaton meage even though t dfferent on L and L tll ynchronzed that to ay bt tranton wll occur at the ame tme on both channel. the Doppler frequency on L related to the one on L by the rato L/L 3. the code delay on L C/A the ame than the one on L CM Snce t ha been oberved and confrmed by the Unted State Coat Guard Center that no navgaton meage tranmtted on LC yet and nce the ICD tpulate that the navgaton meage of L and the one of L hould be ynchronzed, the frt hypothe condered valdated. Dop L L Redual Doppler (Hz) L L = = 7.60 = Dop L MHz MHz L L Dop L Redual of L Doppler - (L/L)L Doppler Tme () PRN 9 PRN 5 PRN PRN 8 PRN PRN 3 PRN 30 PRN 3 PRN 4 PRN Regardng the econd hypothe, data collecton wa performed ung a NovAtel OEMv3 LC capable recever and the Doppler frequency tracked on L and L ha been recorded. However, nce only PRN and PRN 3 were n vew at th pont, the Doppler nformaton of all atellte, even non LC one, wa ued. Ung th data, the Doppler redual obtaned accordng to the followng equaton wa plotted n Fgure 3: Fgure 3: redual Doppler obtaned by dfferencng L Doppler tme the rato L/L and L Doppler over mnute of real data tracked by a NovAtel OEMv3 LC capable recever Snce the redual found are mnmal, the theoretcal relatonhp relatng L and L Doppler can then be condered verfed. ION GNNS 007, Fort Worth TX, 5-8 September 007 3/

4 Fnally, table llutrate the fact that the L C/A code and L CM code are ynchronzed. Indeed, ung an RF front end and collectng real data, t can be hown that the C/A and the CM code are eparated by an nteger number of mllecond. Th reult make ene conderng the fact that the CM code 0 m long wherea the L C/A code only m long. PRN C/A code delay (m) CM code delay (m) Table : code delay found on L C/A and L CM ung real data collectng through a NovAtel RF front end However, due to onopherc actvty, the tme delay between L and L gnal can vary and pobly nvaldate th lat aumpton. Th epecally mportant a the acquton propoed make ued of the fact that L C/A code and L CM code are ynchronzed. Indeed, a delay of 0.5 chp between the two code can lead to a 50 % lo compared to the ynchronzed cae when addng the reult (fgure 4). Fgure 5: vertcal TEC meaured over Januray 007 at 0 lattude and -80 longtude. TECU = 0 6 TEC. However, n order to repreent the actual TEC encounter by the gnal whle travelng through the onophere, the elevaton angle of the atellte ha to be condered. The TEC encountered then relaton of the vertcal TEC (VTEC) and the elevaton angle E of the atellte through a mappng factor M(E): TEC = M ( E) VTEC Fgure 4: effect of nter-code delay on the ummaton of the correlaton peak. C/A code n blue, CM code n red. M ( E) = co( E) h + RE E elevaton angle h onophere alttude R E earth radu In order to verfy that the hypothe tll tand, a tudy of the onopherc actvty durng January 007 wa performed. The data ued wa from the IGS webte and repreent the vertcal TEC at 0 lattude and -80 longtude every two hour over the month and the vertcal Total Electron Content (TEC) obtaned are hown n Fgure 5. Fgure 6: mappng factor to conder a a functon of the elevaton angle ION GNNS 007, Fort Worth TX, 5-8 September 007 4/

5 Accordng to Fgure 6, the wort cae poble regardng the mappng factor to conder occur when the elevaton angle of the atellte 0. At th pont, the VTEC ha to be multpled by three n order to etmate the TEC encounter by the gnal along t path through the onophere. Then ung th nformaton the tme delay between L and L can be deduced ung equaton 3 of the followng: t t t L L R.34 0 = = c f L L t R.34 0 = = c f = t L L L t L = TEC TEC TEC f () () f f f (3) () Ionopherc tme delay on L () Ionopherc tme delay on L (3) Tme delay between L and L due to the onophere Therefore ung equaton (3), the lo experenced compared to the deal cae can be computed: Lo (%) = Delay( chp) 00 A llutrated n Fgure 7, n order to have 00% lo compared to the deal cae, t neceary to encounter a TEC of 800 TECU along the gnal path that a VTEC of 933 TECU. A a comparon, the trong onopherc torm of July 6, 000 had only a magntude about 85 TECU and wa already caung trouble to track the gnal (Walter et al, 004). Under normal onopherc condton uch a the one recorded durng January 007, the wort cae poble wa for the gnal to encounter 05 TECU along t path whch would reult le than 4 % lo. Conderng the cae of the onopherc torm wth 85 TECU mentoned earler, the wort TEC pobly encounter along the gnal path 555 TECU whch correpond to 0 % lo. Anyway, under uch harh condton, mot recever wll have a hard tme to track any gnal due to onopherc cntllaton uually occurrng durng the torm. Therefore, the L C/A code and L CM code can be condered ynchronzed wthout engenderng too much lo durng the acquton proce. THEORY Frt of all, the gnal are aumed to be ampled at an nteger multple of the chp rate: T = T c where T c the chp duraton, T the amplng tme and N the number of ample per chp of the C/A code, the C/A code contanng N c = 03 chp. Then begnnng wth the receved gnal already down converted to peudo-baeband : jφ ( k ) y = C d( k) c( k) e + w jφ ( k ) y = C d c e + w d c = cm( k) cl( k) p beng an even number N pn ( p + ) N () for L k < ( p + ) N () for L k < ( p + ) N where C the carrer power and aumed to be contant over the obervaton tme, c (k) repreent the tranmtted C/A code whch ha a perod of m or the tme-multplexed LC code, d (k) the data bt of the navgaton meage on L and lat 0 m and w (k) the complex addtve whte Gauan noe ample. Φ (k) the redual carrer phae wth frequency and phae offet F and θ and defned a follow: φ = π F k + θ beng or and tandng for L or L. Fgure 7: gnal lo experenced compared to the deal cae a a functon of the TEC on the gnal path A mentoned earler n th paper, the CM code repeat telf every 0 m and the CL code every.5 econd. Note that the CM or CL code chp duraton equal to the C/A code chp duraton and o one chp of the C/A code ION GNNS 007, Fort Worth TX, 5-8 September 007 5/

6 equal to one chp of CM code or one of CL code. However, due to the long perod of the CL code, the LC code conder for acquton formed a follow: ~ cm( k) c = 0 p beng an even number pn ( p + ) N k < ( p + ) N k < ( p + ) N The baeband archtecture of the propoed C/A-CM combned acquton technque ha been llutrated n Fgure. In Fgure, the upper and the lower halve pertan to the GPS L C/A and L CM detecton module. The ame redual carrer utlzed to compenate for the redual frequency offet. The redual carrer frequency caled (by a factor 0.779) for the LC detecton module. The receved ample after redual frequency compenaton are correlated wth the local C/A and modfed CM code code repectvely. The modfed CM code obtaned from the orgnal CM code wth every alternatve ample beng zero padded. The correlated ample are coherently ntegrated over the pre-detecton ntegraton tme (PIT) (.e. T PIT = NcTc). Then t can be hown that for a unque code offet l and frequency offet F, the correlaton/ntegraton output can be expreed a: nnc N j(π Ferr, k +Ω ) x = C d c e k= ( n ) Nc N j(π Fk +θ r ) + w c ( k l) e (5) c ( k l) where F err,, = F F. Ω = θ - θ r a random phae offet where θ r the arbtrary phae offet created by the local ocllator. The correlaton/ntegraton output ampled every T PIT econd wth ample ndex beng n. On ubttuton and further mplfcaton, (5) can be reduced to: x ( l, + n n) = C d R ψ w ( ) (6) where d (n) the ntegraton output of d(k) over the predetecton ntegraton tme and R l, the auto-correlaton functon of the C/A code f = and the auto-correlaton functon of the modfed CM code f = over the predetecton ntegraton tme. ψ (n) the frequency ambguty functon and expreed a: A hown n fgure, the propoed acquton cheme ue the fact that both the CM and C/A code meaure the ame quantte (.e. code and frequency offet). The ntermedate detecton output at pont,, 3 and 4 are gven by: z = x + x z = x x z = x + x 3 4 z = x ( n ) x ( n ) The fnal detecton output then gven by: N ( 3 4 n n= 0 S, F) = z z + z z ( ) (7) From (7), the ndvdual gnal component can be derved: and = CRl,( n) nc ( π Ferr, ( n) N c N 3 + CR, ( ) ( ) l n n ψ ( n ) ψ (8) = C Rl, nc ( π Ferr, N c N 4 + CR,( ) ( ) l n n ψ ( n ) ψ (9) From equaton (8) and (9), t evdent that the fnal detecton output combne the ndvdual non-coherent and dfferental detecton from L C/A and L CM code gnal. TEST METHODOLOGY In order to tet the performance of the propoed acquton technque, two et of experment were conducted, one ung mulated data obtaned through an RF mulator and one ung actual data drectly collectng through a GPS L/L NovAtel Antenna (Fgure 8). The tet performed ung real data were lmted due to the lack of atellte and were conducted prmarly to valdate the mulated reult. The varable attenuator ha only been repreented n fgure 8 n order to llutrate the fact that both mulated and real reult where obtan wth dfferent C/N 0. ) ) ψ nc( π F err, N c N ) e j( π Ferr, Nc N + φ ) Where nc(x) the nc functon and Φ(n) the phae at the begnnng of the pre-detecton ntegraton. ION GNNS 007, Fort Worth TX, 5-8 September 007 6/

7 d = [ E( T / H ) E( T / H ] var( T / H 0 ) 0 The reult obtaned where then compared to the output SNR of non-coherent and coherent acquton on L C/A gnal. Moreover, the effect of the coherent ntegraton on the combned acquton wa tuded a well a the poblty to apply dfferent weght on each channel. RESULTS USING REAL DATA Durng the real data collecton, only PRN 3 and PRN were vble and tracked at 48.8 db-hz and 35.6 db-hz on L by a NovAtel OEM4 recever (Table 3). Fgure 8: tet-bed ued whle ung mulated or real data Intermedate frequency Samplng frequency L 70.4 MHz 0 MHz L 70. MHz 0 MHz Table: parameter of the RF frond end A hown n Table, even though the IF frequence were dfferent for L and L, the amplng frequence were dentcal. Therefore mot dffculte pertanng to reamplng problem were avoded. The parameter ued on the Sprent RF mulator durng the frt et of tet are decrbed n Table. The gnal power and navgaton meage tranmtted by each channel are the one defned n the Interface Control Document ICD-GPS-00C. The onophere wa condered durng the tet wth a VTEC of 30 TECU. Under thee condton the wort TEC encounter along the gnal path could be up to 90 TECU. L C/A LC Iono VTEC Sgnal power (dbw) 30 TECU Navgaton CNAV NAV meage type wth FEC Table : mulator parameter ued durng the tet PRN 3 Frequency L C/A L C/A Lock Ye Ye SNR 35.6 db-hz 48.8 db-hz Doppler -03 Hz 0 Hz Table 3: creenhot of the trackng parameter of PRN and 3 followed through a NovAtel OEM4 recever Gven the avalablty of thee two LC PRN wth trong SNR dfference, the analy wa conducted on both of them. PRN 3 The frt attempt nvolved acquton of the atellte ung m coherent ntegraton and 0 m ummaton over the whole Doppler frequency range that to ay Hz to Hz around the central frequency ung a frequency bn ze of Hz correpondng to the larget poble gven the coherent ntegraton tme. Moreover, the CM code delay beng unknown, each m of the CM code ha to be condered (Fgure 9). The peak due to the preence of the CM code combned wth L C/A code wa found between mllecond 5 and mllecond 6 of the code. Th reult wa expected gven that the CM code delay found when ung drectly the 0 m of the code to create a coherent acquton wa m. Moreover, the combned acquton outperformed the conventonal L C/A non-coherent acquton by 4 db. However, a expected the performance wa tll nferor compared to the optmal L C/A coherent acquton once t ynchronzed wth the data bt tranton. The performance of the propoed combned L/L acquton technque wa evaluated through the SNR obtaned through the deflecton coeffcent d. ION GNNS 007, Fort Worth TX, 5-8 September 007 7/

8 ynchronzed wth the boundary of the ntegraton tme condered. Fgure 9: Reult for PRN 3. L C/A 0 m coherent ntegraton ynchronzed wth the navgaton meage SNR, L C/A non-coherent ntegraton of m wth 0 m ummaton SNR and propoed acquton SNR for m coherent ntegraton wth 0 m ummaton and coverng the 0 poblte of the CM code A expected, the propoed method actually performed the L C/A and LC jont etmaton of Doppler and code delay. Indeed, the output of the acquton correpond to the C/A code delay that can be found over m but alo provde whch mllecond of the CM code gave the dered combnaton. Therefore, by mply addng the C/A code delay over one mllecond to the mllecond of nteret of the CM code, the code delay of the CM code can be found. Then ung the knowledge that the CM code ynchronzed wth the data bt of L, t no longer neceary to perform a data bt ynchronzaton on L. Moreover, the propoed acquton alo output the Doppler frequency found on L (here 000 Hz a expected gven the frequency bn ze ued and the Doppler of the atellte 3) and ung the fact that there a drect relaton between the Doppler on L and the Doppler on L a hown durng the verfcaton of the aumpton, both L Doppler and L Doppler are actually output by the combned L/L technque. Secondly, a tudy of the effect of the coherent acquton tme wa conducted (Fgure 0). The two bet cae obtaned durng th tet were for m and 5 m coherent ntegraton tme. Thee outperformed all other duraton teted. Th wa to be expected gven the fact that the CM code delay wa m; that to ay that after th tme nce the begnnng of the RF data collected, a data bt tranton wa poble. Th tranton had actually happened and reduced the performance for, 4 and 0 m coherent ntegraton nce for each of thee cae, the tranton occurred n the mddle of the chunk of data condered to perform the ntegraton. However, a expected for uch a tuaton, 5 m ntegraton gve the bet SNR nce the tranton Fgure 0: effect of dfferent coherent ntegraton tme on the propoed acquton reult for PRN 3 Fnally, the effect of weghtng the L or L channel nvetgated and llutrated n Fgure. A expected the bet cae poble occur when both channel are weghted equally. Indeed, n any other weghtng cheme, one would not beneft from the full level of addtonal gnal power avalable. Fgure : effect of weghtng the channel for PRN 3 PRN Weght _ L = Weght _ L To confrm the prevou reult and n order to tet the propoed combned acquton under harher condton, the ame ere of tet were conducted on PRN whch wa tracked only at 35.6 db-hz by the NovAtel OEM4 recever ued. ION GNNS 007, Fort Worth TX, 5-8 September 007 8/

9 Frt of all, a normal LC CM acquton wa performed to determne the actual code delay of the code and compare t wth the code phae output by the combned acquton. The delay found wa of.7655 m and confrmed the reult obtaned on Fgure whch how that the correct CM code m wa the one between and 3. Fgure 3: effect of dfferent coherent ntegraton tme on the propoed acquton reult for PRN Fnally, Fgure 4 confrm that the bet choce poble to ue both gnal equally durng the acquton proce. Fgure 4: effect of weghtng the channel for PRN Fgure : Reult for PRN. L C/A 0 m coherent ntegraton ynchronzed wth the navgaton meage SNR, L C/A non-coherent ntegraton of m wth 0 m ummaton SNR and propoed acquton SNR for m coherent ntegraton wth 0 m ummaton and coverng the 0 poblte of the CM code Then followng the ame approach than for PRN 3, the effect of the coherent ntegraton tme wa nvetgated (Fgure 3). The bet ntegraton tme obtaned are and 4 m. Indeed, for both of thee tme, loe are mnmal nce the data bt tranton occur at.7655 m. However, wherea 5 m howed farly good performance compared to what wa expected. Th could be explaned by the fact that even f the tranton occurred n the mddle of a 5 m chunk of data, three of the four of the coherent ntegraton are tll maxmal. RESULTS USING SIMULATED DATA Two tet were conducted under a controlled C/N 0 envronment, one at 45 db-hz and one at 35 db-hz. A L/L RF mulator wa ued to generate the gnal tranmtted by 9 atellte wth equal power that to ay the gnal-noe rato receved by the recever wa the ame for each atellte. The only dfference mulated on each atellte wa the effect of the onophere through the elevaton angle of the atellte condered and the navgaton meage tranmtted on L or L. Indeed, nce the between atellte and recever dtance wa dfferent for each atellte, the data bt tranton occurred at dfferent ntant for each of them. The 9 atellte were then ued to create tattc on the SNR output by the propoed acquton and compared to the reult of a 0 m coherent acquton on L ynchronzed wth the navgaton meage and a noncoherent acquton on 0 m wth m coherent ntegraton. Sgnal tracked at 45 db-hz on L The mean and tandard devaton of the SNR obtaned for a 0 m tet va the combned L/L acquton for dfferent coherent ntegraton tme are llutrated on Fgure 5. ION GNNS 007, Fort Worth TX, 5-8 September 007 9/

10 Fgure 5: mean and tandard devaton of conventonal coherent and non-coherent acquton on L compared to the propoed novel acquton for dfferent coherent ntegraton tme whle L gnal tracked at 35 db-hz The mean and tandard devaton of conventonal L C/A are alo repreented. For ntance, ung a m coherent ntegraton followed by 0 non-coherent ummaton yeld a mean SNR of 30. db and a tandard devaton of 0.6 db wherea a full 0 m coherent ntegraton gave a mean SNR of 34. db and a tandard devaton of 0.5 db. The propoed acquton outperform the performance of the conventonal non-coherent acquton on L whle keepng the frequency bn ze a mall a poble. However, ncreang the coherent ntegraton tme on the propoed method doe not brng any mprovement nce the mean SNR found tayed approxmately the ame wherea the tandard devaton ncreae. Th wa to be expected a the m coherent ntegraton tme le affected by the data bt tranton. Indeed, f a tranton occur whle ung only m coherent ntegraton, only one twenteth of the 0 ummaton wll be affected by t wherea t one tenth for m, one ffth for 4 m and o on. Fnally, one hould notce the trong correlaton between the obtaned tattc and the tet performed ung actual gnal. Indeed, durng the acquton of PRN 3 tracked at 48.8 db-hz ung real data, the SNR obtaned wa of 36.5 db wherea the gnal tracked at 45 db-hz durng the mulaton yeld a 3. db output SNR. Th a mean 4. db dfference n SNR for a 3.8 db-hz dfference n C/ N 0. Secondly, the effect of weghtng the gnal wa nvetgated and the tattcal reult are hown n Fgure 6. Fgure 6: mean and tandard devaton of conventonal coherent and non-coherent acquton on L compared to the propoed novel acquton for weght appled on the gnal whle L gnal tracked at 35 db-hz Once agan, the bet tuaton proved to be when both gnal are weghted equally. Sgnal tracked at 35 db-hz on L A mlar et of tattc wa obtaned ung gnal tracked at only 35 db-hz by the OEM4 NovAtel recever. Once agan the effect of the coherent ntegraton tme a well a the poblty of weghtng L or L gnal were nvetgated. On the other hand, the performance of the non-coherent and coherent acquton on L were computed. The reult where then compared between each other and the compared wth the real cae (PRN ). Fgure 7 and Fgure 8 llutrate the reult obtaned for the effect of the coherent tme and the weght repectvely. The propoed acquton proved to be able to perform better than conventonal L C/A non-coherent acquton (mean of 3 db mprovement). However, a wa expected, t tll reman below the performance of a full 0 m coherent acquton but th lat one would need a total of 300 frequency bn wherea the combned method only requre 5. Increang the coherent ntegraton tme ued doe not brng uffcent mprovement. Indeed, even f the mean SNR lghtly mproved, the tandard devaton ncreae a well. On the other hand, obtaned reult are content wth the tet performed ung real data for the PRN tracked at 35.6 db-hz; RF mulaton yeldng a SNR of. db wherea actual data reultng n.6 db. ION GNNS 007, Fort Worth TX, 5-8 September 007 0/

11 Fgure 7: mean and tandard devaton of conventonal coherent and non-coherent acquton on L compared to the propoed novel acquton for dfferent coherent ntegraton tme whle L gnal tracked at 35 db-hz code phae and frequency offet. However, wherea uch combnaton would requre four multplcaton and four addton, the propoed algorthm need only two multplcaton and four addton to obtan the ame reult. In o dong, the proceng tme needed almot dvded by two nce multplcaton repreent the crtcal factor compared to addton n term of computatonal burden. All the aumpton requred for the propoed method have been checked and valdated before tetng. Concern were epecally drected toward the aumpton of ynchronzed L/L code due to the onophere effect. However, through a mple nvetgaton of onopherc normal condton, t ha been demontrated that the combned could only be affected by le than 4 % lo compared to the deal cae. In order to verfy the theory, everal tet were conducted ncludng tet ung real data through the ue of the three PRN already broadcatng LC and mulaton through the ue of an RF oftware mulator to create a controlled C/N 0 envronment and perform a tattcal tudy of the acquton method performance. In both cae, data collecton ncluded gnal tracked at 45 db-hz and 35 db-hz by a NovAtel OEM4 recever. Durng all of the tet, the combned acquton outperformed the capablty of common L C/A non-coherent acquton by to 3 db whle keepng a coherent acquton tme of only m. Under thee condton, the frequency bn ze kept at Hz that to ay that only 5 bn are requred to cover the entre Doppler pace compared to 300 bn n the cae of 0 m coherent ummaton. Th lat pont epecally mportant a t how that the acquton tme can be greatly reduced under challengng condton. ACKNOWLEDGMENTS Fgure 8: mean and tandard devaton of conventonal coherent and non-coherent acquton on L compared to the propoed novel acquton for weght appled on the gnal whle L gnal tracked at 35 db-hz Fnally, ung both gnal equally durng the acquton proce offer the bet performance, far above the cae where only one of the two gnal ued. Th pecfc behavor explaned by the fact that the noe of a noncoherent acquton and a dfferental acquton are correlated to ome extent wherea the noe on L and the noe on L are completely uncorrelated. CONCLUSION In th paper, a novel acquton combnng L and L cvl gnal ha been propoed. The output obtaned correpond to the ummaton of non-coherent and dfferental acquton performed on L and on L and allow performng a jont etmaton of L C/A and LC The frt author would lke to thank Florence Macch, PhD tudent n Geomatc Engneerng, for her contant help and upport, a well a the Informatc Crcle Of Reearch Excellence and the GEOIDE Network of Centre of Excellence for ther fnancal upport. REFERENCES Cho, D. J., C. S. Park, S. J. Lee (004) An Ated GPS Acquton Method ung L Cvl Sgnal n Weak Sgnal Envronment, Journal of Global Potonng Sytem, vol 3, no -:5-3 Lm, D.W., S.W. Moon, C. Park, S.J. Lee (006) L/LCS GPS Recever Implementaton wth Fat Acquton Scheme, Poton, Locaton, And Navgaton Sympoum, 006 IEEE/ION, 5-7 Aprl, pp Pak, M. L., L. Wnterntz, M. Moreau (004) FFT- Baed Acquton of GPS L Cvlan CM and CL ION GNNS 007, Fort Worth TX, 5-8 September 007 /

12 Sgnal, n Proceedng of the ION GNSS 004, Long Beach, September 004. Walter, T., S. Rajagopal, S. Datta-Barua, J. Blanch Protectng Agant Unampled Ionopherc Threat, n Proceedng of the Beacon Satellte Sympoum, October 004, Trete, Italy Yang, C. (005) Jont Acquton of CM and CL Code for GPS LC, n Proceedng of ION GNNS 005, Long Beach, Calforna. ION GNNS 007, Fort Worth TX, 5-8 September 007 /