MULTIPATH-EFFECT INVESTIGATIONS FOR QUALITY IMPROVMENT OF BIHPOS

Transcription

1 MULTIPATH-EFFECT INVESTIGATIONS FOR QUALITY IMPROVMENT OF BIHPOS A. Blajbegovć 1,, M. Mulc 2 1 Unversty of Appled Scences Dresden, 2 Unversty of Sarajevo Abstract Multpath effects were nvestgated for the CERGOP statons, future BHPOS statons as well as for the permanent referent statons n the SAPOS-Sachsen Networ, usng WaSoft/Multpath Software. Correlaton between the accuracy of coordnates and multpath ndexes were nvestgated. Also, the rover ntalsaton tme dependency on multpath ndex of reference staton was nvestgated as well. 1. Introducton We often meet problems n processng observatons of some partcular ponts durng the processng of GPS networs. These problems are often connected to locaton of GPS pont, obstructers of horzon, multpath effects and sgnal nterference. Because of mult reflectons of GPS sgnal (even on the pont where we could not see surface of the reflectons), multpath effects appear. For the purpose of the nvestgaton of the correlaton between multpath effects and the accuracy of the coordnates determnaton, the multpath effects were analysed: n the networ of the ponts whch were observed on the terrtory of Bosna and Herzegovna n the frame of CEGRN 05 GPS campagn, several futures BHPOS statons as well as on the all permanent referent statons n the SAPOS-Sachsen Networ. For these nvestgatons, the robot arm and GEO ++ software could be used (Fg. 1.) or, the software WaSoft/Multpath. The procedure by GEO ++ can not be mplemented on approprate way on the CERGOP statons, because of the expensve, long-lastng and tme consumng observatons. The WaSoft/Multpath procedure enables mplementaton for the measurements whch were already accomplshed and, n that way, mnmze the costs. Fg. 1: Robot arm and statc referent staton (Böder et al., 2001) Of course, t would be desrable to compare results of these two procedures. In the frame of several dploma wors at Unversty of Appled Scences n Dresden (HTWD), ntalsaton tme of the rovers were nvestgated usng networ of permanent statons n dfferent areas of the states Sachsen Anhalt and Sachsen. The nvestgatons showed that tmes were not the same, and, further, n the neghbourhood of the statons wth bg multpath ndexes, the ntalsatons were dffcult to acheve and too much more tme. In SAPOS-Sachsen networ of permanent statons, the old Trmble GPS recevers (4000 SS) and Compact antennas were replaced by new Trmble recevers (NetR) and specfc Trmble geodetc Zephyr antennas. It was nvestgated f the new equpment offered great mprovements of the multpath nfluence.

2 Of course, the specfc crcumstances must not be neglected n the procedure of the multpath effects estmaton by WaSoft/Software (Wannger, 2001), as the error sources: dfferent state of onosphere, troposphere and eventually, dfferent dsposton of surfaces of the reflecton around the reference staton. 2. Phase observaton equatons wth short systematzaton of GPS errors and multpath effects theory GPS phase measurements could be descrbed by the equaton: λ Φ E where are: = δ I = ρ c + c ( δ + δ T δ ) + λ N + δ O = + δ S + E + δ A + ε speed of lght n vacum,, + δ B wth + δ M λ = wavelenght of L or L 1 2 Φ = phase shft of carry wave (cycle), ρ = real dstance between recever and satellte n meters, δ δ = dfference between satellte and recever cloc error. N = ambquty (cycles), δi = nfluence of onosphere, δt = nfluence of trosphere, δo = orbt error, δs = measuerment error caused by varaton of phase center n satellte antenne, δa = measuerment error caused by varaton of phase center n recever antenne, δb = measuerment error caused by sgnal curvng, δm = measuerment error caused by multpath effects, δif = measuerment error caused by nterference wth other sgnals and δd = sgnal delay n satellte and recever hardware. (2.-1) Obvously, term E contents 9 sources of errors, whch we try to elmnate by: choosng approprate method of data processng (lnear combnaton of phase measurements), usng precse ephemers, usng calbrated hardware (GPS recevers and antennas), measurng and modellng the troposphere parameters, ntroducng unnown parameters n the adjustment and the adopton of the mean values of the longlastng GPS measurements (for example for the multpath effects). The problems appeared especally wth multpath effects n RTK measurements, whch tae just few mnutes and because of that t remans as the systematc errors n the results of the GPS measurements. GPS errors could be systematzed n accordance wth ther mpact: The errors dependent on the dstance between GPS recevers (errors n the satellte orbts, errors due to sgnal propagaton through onosphere and troposphere) errors of the local character (sgnal curvature, multpath effects, antenna phase centre varatons and sgnal nterference. + δ IF + δ d,

3 The errors dependent on the dstance between GPS recevers could be modelled usng DGPS technque and some of them could be elmnated, whle the errors of the local character could not be modelled. In the case of long-lastng observatons some of them (for example the multpath effect) could be effectvely reduced by adopton of average value. Multpath effects are very acute n the case when the staton s located on the roofs wth reflectons below the antenna horzon, and could be reduced only n code measurements. In the case of phase measurements, and n the case of the antennas whch are usually used n practce, the multpath effects could not be reduced for short observaton perods. The sgnal curvature errors appear n the case of obstructon of the antenna horzon and could be dentfed by analysng S/N rato (sgnal and nose rato). The possbltes to elmnate or reduce nfluence of the multpath effects are as follow: 1. Selecton of the antenna locatons. There s possblty to detect multpath effects from the test measurng at the locaton of the planned reference statons (for example BHPOS statons). 2. Improvement of the measurng technques of recevers. There were the mprovements n the case of the code measurements durng last couple of years, but for phase measurements the mprovements almost were not apparent. 3. Usng the multpath resstant antenna. Durng last couple of years n ths feld almost no mprovements appeared and perhaps t s not expected n the near future. 4. Calbratng reference statons on the nfluence multpath effects (Wannnger, 2002). 5. Usng the mult-antenna systems (Wannnger, 2002). Multpath effects due to reflecton from the ground (below the antenna horzon) and vertcal surfaces (buldngs) are shown n Fg. 2. A β s 2β β B β C h β D d d Fg. 2: Multpath reflectons In accordance to Fg. 2 dfferences n the paths of the drect and reflected waves could be estmated for the sgnals reflected from ground: s = 2h sn β (2.-2) and for the sgnals reflected from buldngs: s = 2d cos β. (2.-3) From (2.-2) and (2. 3) approxmated perods of multpath effects could be estmated as the functon of: antenna heght, dstance to the buldng, elevaton angle and the frequency of the satellte sgnal, as shown n Table 1.

4 Table 1: The estmaton of multpath perods Multpath from the buldngs β = 45 β = 10 β = 45 β = 10 d=10 m T L1 =3.1 mn T L2 =3.9 mn d=1 m T L1 =30.7 mn T L2 =38.7 mn d=10 m T L1 =12.5 mn T L2 =15.8 mn d=1 m T L1 =124.9 (65,8) mn T L2 =157.7 (83,0) mn Multpath from the ground h=2 m T L1 =15.3 mn T L2 =19.4 mn h=1 m T L1 =30.7 mn T L2 =38.7 mn h=2 m T L1 =11.0 mn T L2 =13.9 mn h=1 m T L1 =22.0 mn T L2 =27.8 mn In accordance to Table 1. multpath perods are changed from 3.1 to 83 mnutes. Obvously, n the case of short-lastng observatons (RTK) multpath effects have systematc character. 3. Multpath effects n B&H CEGRN 05 networ and at several ponts of the future permanent networ of reference statons BIHPOS Applyng dfferent procedure for data processng of CERGOP and BHPOS networs usng dfferent software t was shown that there were dffcultes for some statons (Čvrsnca, Bjelašnca, Aladnća Brdo), whch requred analyzng of measurements process and adequate nterventon. Therefore, the estmaton of the map and multpath ndexes were perform from the result of the measurng n CERGOP and BIHPOS GPS networs. Data processng by software WaSoft/Multpath V. 3.3 and results were shown n the smple-plot and vector format, Fg.: 3, 4, 5 and 6. In the Smple-Plot-Format azmuths denoted every 10, elevaton angles every 2 (degrees) az el srjv.map. Input fles, multpath ndex: 1 srjv171.map 12 2 srjv172.map 21 3 srjv173.map 15 4 srjv174.map 7 5 srjv175.map 5 Mean ndex 12 +/- 2.8 Fg. 3: Multpath-map and multpath-ndex for EPN SRJV permanent staton n Sarajevo Legend not any observatons small effects mddle effects strong effects Fg. 4: Multpath-map n vector format for the EPN SRJV staton.

5 az el cvrs.map. Input fles, multpath ndex: 1 cvrs171.map 39 2 cvrs173.map 68 3 cvrs174.map 70 4 cvrs175.map 59 Mean ndex 59 +/-10.0 Fg. 5: Multpath-map and multpath-ndex of the staton Čvrsnca n the smple plot format Fg. 6: Multpath-map n vector format and photo of Čvrsnca CERGOP staton The short explanatons for the correct use of the WaSoft/Multpath software, detecton and localsaton of the multpath effects wll be mentons here. The results of the detecton and the localsaton could be presented n graphcal way n the smple-plot format and as the vectors, Fg. 3 and 4. Multpath-map for each staton s shown. The axs -X descrbes the azmuths and the axs-y descrbes elevaton angles. The algorthm of the estmaton uses those satellte sgnals as reference, wth the elevaton angles above 50, because they have nor, or have very small multpath effects. It s explanaton why Multpath-map s lmted to elevatons from 0 to 50. In the Smple-Plot-Format azmuths are mared at every 10, but the elevaton angle at every 2 (degrees). Four symbols descrbe Smple-Plot-Format, Fg. 5: (Empty spaces) there are not any observatons. (Dot) - small multpath effects: RMS values < 5 mm 1 - mddle multpath effects: RMS- values <15 mm 2 - strong multpath effects: RMS- values > 15 mm. The azmuth and the elevaton angle refer to the angle of the satellte sgnals comng down, and n general, not descrbng drecton to reflectng surfaces. Only n the case of the staton sufferng of the sgnal reflecton from the ground (below of the antenna horzon), drecton to the satellte s n the same drecton as the drecton to the reflectng surface (as t s the case for the most SAPOS statons). For the purpose of detectng multpath effects, t s necessary to have smultaneous observaton at 3 statons at least (better 5-6, the most at 10 statons) n the perod of at least 24 hours. For the redundant estmatons, the multple-days smultaneous observatons are preferable, on several GPS statons. Theoretcally spng, multpath effects nfluence on the L1 and L2 frequency could be maxmum 6 cm and because of that t s relevant for all precse measurements. For the dstances longer than 10 m t can not be set asde from other error sources (especally from the errors caused by onosphere). Because of that, the algorthm of the lnear combnatons free of onosphere

6 nfluence s appled (onosphere free soluton, L3 frequency.e. Lc). These ncreases nfluence of multpath effects for the factor about 3, referrng on the nfluence on the orgnal measurements on the L1 and L2 frequences. It s obvous that these nfluences can be estmated only from the two frequences recever observatons. The necessary assumptons for the successful detecton and localsaton of the multpath effects usng WaSoft/Multpath software are: Permanent, synchronsed measurements on the statons, as well as Broadcast-ephemers n RINEX format. Dstances between statons must not be longer than 100 m. The coordnates of antennas have to be nown wth accuracy of about 1 cm. Set of the observatons must not be shorter than 24. Otherwse, resultng detecton of multpath effects could be ncorrect. For the more redundant estmaton results the mult-set-of 24 hoursmeasurements are recommended. Mult-day estmatons could be merged later, usng software module MAPMERGE. Samplng rate can be 60 seconds, for the reason of shorter tme of the estmaton, but values of 15, 20, 30 seconds could be used. Elevaton mas for the date recordng should be 0. The analysng results of the multpath data are vald: only for the same locaton of the GPS-antenna, only for appled antenna type, only for appled recever type, ncorporated wth nner software and for the same antenna envronment as t happened to be durng the observaton tme. Every sngle change n the staton poston or type of GPS equpment could change nfluence of multpath effects to the phase measurements. Especally, the snow on the antenna or, the snow on the reflectng surfaces could produce change n multpath effect. For the purpose of the comparsons the multpath nfluences on the dfferent statons t was necessary to ntroduce a smple ndcator of the ntensty of the multpath reflectvty,.e. multpath ndex. Ths s actually, the general arthmetc mean of standard devatons of the resduals of double phase dfferences DD of the multpath effects. The realstc categorsatons of the ndexes of the multpath nfluences are not made by now. The frst categorsaton by Höntzsch (2005), made for the SAPOS-Sachsen networ (n accordance wth the authors opnon) s too harsh, Table 2. Table 2: Categorsaton of the multpath effects Höntzsch categorsaton Categorsaton authors Multpath ndex Multpath ndex 1 to 3 No nfluence 1 to 4 No nfluence 3 to 5 Small nfluence 5 to 9 Small nfluence 5 to 10 Mddle nfluence 10 to 18 Mddle nfluence 10 to 20 Strong nfluence 19 to 30 Strong nfluence > 20 Very strong nfluence >30 Very strong nfluence The bgest nflence multpath effect apeared for the proposed locaton of future BIHPOS n town Zenca, (Fg. 7). But, results for proposed Zenca staton was expected because of nd of smooth steel materal of the roof. Materal wll be probably changed before GPS staton would be establsh.

7 Fg. 7: The multpath-ndex (n Zenca=46) on the futures BHPOS staton n Zenca 3. 1 Correlaton between the accuracy of GPS coordnates estmaton and multpath-ndex For the purpose of these nvestgatons, GPS measurements are processed and multpath nfluences are estmated: n the GPS networ observed n Bosna and Herzegovna n the frame of CEGRN 2005 GPS campagn, on some ponts of the future BHPOS networ of permanent GPS statons, as well as n SAPOS-Sachsen networ (usng Trmble Total Control or Bernese software). The results are presented graphcally, Fg. 8 and BJEL CVRS ALBR Multpath-ndex KUDI JABL y = 4,2507x + 10,1 ZENI LIVN TURI LIVA PONI KABA OSJE LEOT VLAS BABJ JAJC DOBO BNLK TUZL D-standard devaton n mm Fg. 8: Multpath-ndex as a functon of standard devaton of the coordnates of CEGRN 05 B&H statons and planed BHPOS (Software TTC)

8 35 30 Mulpath-ndex y = 153,84x - 26, ,2400 0,2500 0,2600 0,2700 0,2800 0,2900 0,3000 3D-standard devaton n mm Fg.9: Multpath-ndex as a functon of standard devaton of the SAPOS coordnates (Bernese Software) To provde an overvew of the realstc correlaton between a multpath ndex and an accuracy of GPS coordnates determnatons, the correlaton coeffcents between tree-dmensonal (horzontal and vertcal accuraces) of SAPOS-Sachsen coordnates and multpath ndexes are estmated, Table 3. Tab.3: Correlaton coeffcent between multpath ndex and coordnate accuracy of SAPOS-Sachsen statons Correlaton coeffcent between multpath ndex and the nner accuracy Tree dmensonal coordnates r =0.24 horzontal coordnates r =0.21 heghts r =0.22 In spte of the fact that these measurements were carred out durng 48 hours (t s nown that for a longlastng observaton multpath effects get random propertes), the correlaton coeffcent s not equal to zero nor approxmately equal to zero. 4. Investgaton of multpath effects for dfferent types of GPS antennas, frmware antenna verson and recever type One of the objects of the nvestgaton was to establsh multpath effects dependency on appled types, nd and frmware of antennas. For ths purpose, the measurements were accomplshed at the same staton wth dfferent nd of antennas and for the each antenna nstallaton, for the perod of 48 hours. Data processng was made by WaSoft/Multpath V. 3.3 software and observaton at 5 neghbourng SAPOS statons whch have relatvely small multpath effects. Choe rng antenna Leca (actually Turborouge Margolne) and Topcon choe rng antenna CR-3 wth radom presented the same characterstc. For the both antennas multpath ndex, wth elevaton angle of 0 degree, was 21, see Fg. 10 and 11. Trmble geodetc Zephyr and Compact L1/L2 wth ground plan antennas showed (surprsngly) almost the same ndex: 48 and 47, Fg. 12 and 13. However, there was unexpected bad result for the antenna Trmble geodetc Zephyr, whch was constructed and (n accordance wth frmware specfcatons) has specally desgned absorbng layer wth bg electrc resstance for the absorpton the reflectng sgnals from the ground. These nvestgatons presented that ths antenna, n ths sense, does not dstngush tself from the earler verson of the Trmble antennas. Postve characterstc of ths type antenna s that, t recorded sgnals successfully from satelltes wth the elevaton angle even lower than 5 degrees.

9 az el A312.map M= 21. Fg. 10: Leca choe rng antenna wth recever GTX 1220 at the pont 135A (HTWD) multpath ndex was equal to az el A305.map M= 21. Fg. 11: Topcon choe rng antenna CR-3 wth recever GB multpath ndex equal to 21 on the pont 135A (HTWD)

10 Choe rng antenna presented multpath ndexes reduced by double (more exact 2.3 tmes smaller). They collect prmary and secondary waves (whch are out of phase) and on so, rapdly reduce reflecton of waves from ground, see Fgures: 14, 15, 16 and especally Fgures: 10, 11, 12 and az el A308.map M= 48. Fg. 12: Trmble geodetc zephyr antenna wth recever R7- multpath ndex equal to 48 on the pont 135A (HTWD) az el a327.map M= 47. Fg. 13: Trmble Compact L1/L2 antenna wth GP wth recever 4000SS - multpath ndex equal to 47 at the pont 135A (HTWD)

11 Fg. 14: Reflected L1-vawes on flter (Flpov et. al., 1999). Fg.15: L2-vawes pass through flter and reflect tself from the bottom of the antenna, (Flpov et. al., 1999). Fg.16: Summng sgnals E 1 and E 2 resultng n reduced sgnal E (Hecht, 2001). Landesvermesungsamt Sachsen (Geodetc admnstraton of the state Sachsen) has replaced the GPS recevers Trmble 4000 SS and the antennas Compact L1/L2 wth ground plane by the recevers Trmble NetR and the antennas Trmble geodetc Zephyr. Investgaton of multpath ndexes wth the old equpment accomplshed n sprng 2005, one wee long observatons (Höntzsch 2005) was repeated wth new equpment n fall 2005, tree-day long observatons, usng the same statons combnatons and WaSoft/Multpath software. The average multpath ndex of all Sachsen-statons wth old equpment amounted to 11.53, but wth new one, However, ths research wll be contnued wth more observatons, but results presented by now ndcated that multpath ndex dd not reduced by mplementng the new equpment, more exactly, specal antennas. 5. Is t tme for rover ntalsatons usng permanent reference statons networ (HEPS Servs) functon of multpath ndex? In the frame of several dploma wors at HTWD the ntalsaton tme of the rovers were nvestgated usng permanent reference statons networ, n the dfferent areas of state Sachsen Anhalt and Sachsen. Dfferent types of the GPS recevers were used and the dfferent methods for the ntalsaton of rovers were appled, Table 4. Table 4: The average tme of ln to RS wth successful ntalzaton. Recever type Locaton of the measurements Number of the measurements Average tme of the ln to RS n seconds Leca SR 530 Lepzg out of the area S.- A. Sachsen-Anhalt Trmble 4700 Lepzg out of the area S.- A. Sachsen-Anhalt Trmble 5800 Lepzg out of the area S.- A. Sachsen-Anhalt

12 Extremely bg multpath ndex (equal to 31) was showed durng nvestgaton (for the perod of 7 days) at reference staton n Lepzg, Fg. 17. Mean multpath ndex n SAPOS-Sachsen networ amount to Probably, ntalzaton tme problems of rovers n the neghbourng reference staton Lepzg are assocated wth bg multpath effects on that staton. Fg. 17: Multpath map of SAPOS staton Lepzg and Zttau. Because of presented results, the authors would recommend nvestgaton of multpath effects before stablsaton of the planned permanent reference GNNS statons.

13 6. Concluson The conclusons are as follows: 1. Durng the GPS data processng n the CEGRN05 networ on the B&H terrtory, problems of the bg multpath effects on the several statons occurred (Bjelašnca, Čvrsnca, Aladnća brdo). 2. Results of WaSoft/Multpath v. 3.3 (.e. multpath effects) are realstc only f about 80% ponts whch are computed together, (4-6) have small multpath effects. 3. Multpath ndexes resultng from the analyse of the SAPOS networ, CEGRN 05 and future BHPOS networs present that correlaton exst between coordnates accuracy and multpath ndex, n spte of fact that we processed mult-day-observatons data. 4. In the SAPOS networ, the ntalsaton tme wth HEPS rover, n the neghbourhood of the statons wth bg multpath effects, s qute longer and sometmes ntalsaton s mpossble. 5. Prelmnary analyse of the replacement of Trmble Compact L1/L2 wth ground plane antennas by Zephyr geodetc antennas n SAPOS networ Sachsen dd not ndcate mprovement n multpath ndexes. 6. The nvestgaton should be contnued. 7. Acnowledgment A nd acnowledgment goes to Dpl.-Ing. Andreas Zeller for carryng out measurements n SAPOS- Sachsen Anhalt networ. 8. References Blajbegovć, A. (2005): Vorlesungssrpte Satelltengeodäse, HTWD Dresden. S Böder, V./Menge, F./Seeber, G. (2001): How to Deal wth Staton Dependent Errors New Developments of the Absolute Feld Calbraton of PCV and Phase-Multpath wth a precse Robot, Salt Lae Cty, Source: Date: Flppov, V./ Sutagn, I./Ashjaee, J (1999): Measured Characterstcs of Dual Depth Dual Frequency Choe Rng for Multpath Rejecton n GPS Recevers, Nashvlle Source: Date: Hecht, E. (2001): Opt; Oldenburg Verlag München,Wen, p Höntzsch, D. (2005): Untersuchung der Multpath-Effete mt WaSoft/Multpath Software. Dplomarbet an der HTWD, S Wannnger, L (2002) : Kalbrerung von Referenzstatonen, Hannover, Source: Date:

14 Wannnger, L. (2003): Carrer Phase Multpath Detecton and Localzaton n GPS Reference Staton Networs, Neuwed. Source: Date: Wübbena, G./Schmtz, M. (2002): Developments n Absolute Feld Calbraton of GPS Antennas and Absolute Ste Dependent Multpath, Ottawa. Source: Date: