Measurements and Modeling of the Satellite-to- Indoor Channel for Galileo

Transcription

1 Measurements and Modelng of the Satellte-to- Indoor Channel for Galleo F. Pérez-Fontán, B. Sanmartín, Unversty of Vgo. Span. A. Stengass, A. Lehner, J.Selva, German Aerospace Center, Germany E. Kubsta, Joanneum Research. Graz. Austra. B. Arbesser-Rastburg, European Space Agency BIOGRAPHIES Fernando Pérez-Fontán, was born n Vlagarcía de Arousa, Span. He receved a Telecommuncatons Engneerng degree and a PhD from the Polytechnc Unversty of Madrd n 1982 and He has been wth the Telecommuncatons Engneerng School of Vgo Unversty snce 1988, where he currently s a full professor at the Department of Sgnal Theory and Communcatons. Belén Sanmartín, was born n Chantada, Span. She obtaned her Telecommuncatons Engneerng degree from Vgo Unversty n Se has been nvolved n several GALILEO projects dong determnstc and statstcal modelng of the navgaton propagaton channel. Alexander Stengass was born n Mettmann, Germany n He receved hs Dpl. Ing. dploma n Electrcal Engneerng n 1997 (Unversty of Ulm, Germany). Snce then he has been a research scentst at the German Aerospace Centre DLR - Insttute of Communcatons and Navgaton. He has been nvolved n several projects n the area of satellte navgaton and locaton dependent moble servces. In 2002 he was promoted Dr.Ing. at the Unversty of Essen. alexander.stengass@dlr.de Andreas Lehner was born n Gmunden, Austra n He receved hs Dploma Engneer degree n Mechatroncs from the Unversty of Lnz n Snce then he has been a research scentst at the German Aerospace Centre DLR. andreas.lehner@dlr.de Jesús Selva receved the Communcatons Engneer degree from the Techncal Unversty of Valenca (Span) n 1994, and the Mathematcs degree from the UNED Unversty (Span) n He currently wors at the German Aerospace Centre (DLR) and s about to complete hs Ph. D. thess at the Unversty of Catalona (Span). Erwn Kubsta receved the Dpl. Ing. degree n Telecommuncatons and Electronc Engneerng from the Techncal Unversty of Graz, Austra, n He s wth Joanneum Research, Graz, where he has been nvolved wth and has lead a number of projects dealng wth moble and fxed satellte wave propagaton carred out under ESA and EU contracts. Bertram Arbesser-Rastburg was born n Austra. He receved the DpI.-Ing. degree n electrcal engneerng from the Techncal Unversty of Graz, Graz, Austra, n In 1988, he joned the European Space Agency. Snce 1989, he has been Head of the Wave Interacton and Propagaton Secton at the European Space Agency/European Space Technology Center (ESA/ESTEC), Noordwj, and The Netherlands. Hs research actvtes nclude propagaton effects relevant to fxed, moble, and navgaton satellte systems as well as drect and ndrect models of mcrowave nteracton wth the Earth s surface and the atmosphere for actve and passve remote-sensng nstruments. ABSTRACT In ths paper measurements and modelng results obtaned n the course of an European Space Agency, ESA, project ( ) enttled "Navgaton sgnal measurement campagn for crtcal envronments", carred out by JOANNEUM RESEARCH (Austra) wth subcontracts to DLR (Germany) and the Unversty of Vgo (Span), are presented. Durng the project, measurements were performed usng a hgh resoluton channel sounder on board a helcopter, the objectve beng the characterzaton of the satellte-to-ndoor channel at frequences close to those assgned to the future Galleo navgaton system. The helcopter was flown around a buldng where a recever was located. Dfferent flghts were carred out to cover several elevaton angles. The recever was also placed n dfferent locatons at dfferent floors wthn the buldng so that dfferent penetraton condtons could be studed: close to the external walls, corrdors, nner rooms, et. al.. Typcal measured power-delay profles, PDPs, are presented and analyzed. In addton, the dynamcs of the channel were also studed and characterzed. To complement the study, a modelng actvty was carred

2 out n two drectons. For one, the measured envronments were studed usng a determnstc, ray-tracng based propagaton tool to gan further nsght on the propagaton phenomena. In a second step, a statstcal channel model was derved from the measurements. correspondng to a reflected/dffracted ray. The maxmum snapshot power usually has a 3 db varaton n consecutve snapshots as can be seen n Fgure 2. The statstcal model reflects the dfferent tme-spreadng condtons found n the measurements, whch are manly caused by reflectons and dffractons wthn the buldng tself and those produced by reflectons/dffractons on external buldngs. The statstcal model also addresses how the arrvng contrbutons reach the recever n groups or clusters of rays. INTRODUCTION Ths paper presents an analyss of expermental data for the satellte-to-ndoor channel obtaned n a measurement campagn carred out by the authors usng a helcopter to smulate the navgaton satellte and a hgh resoluton wdeband channel sounder. In addton, the paper descrbes the man ssues relatve to a determnstc wdeband channel model and a statstcal model also developed from these measurements. In the last step the measurements were also used to valdate the determnstc model. MEASUREMENT ANALYSIS Fgure 1. Impulse responses n the 30 elevaton crcle, scenaro P4 (fourth floor), wth delays relatve to the maxmum power epoch. For the plot to the rght, we selected a power threshold of 15 db below the maxmum power n all snapshots, and we plotted only those snapshots n whch the maxmum power s above ths threshold. In ths secton we present an analyss of the results of the satellte-to-ndoor channel measurements of Frst, we chose a crteron n order to separate the measured snapshots n two sets: (a) snapshots wthout lne-of-sght (LOS) ray and (b) snapshots wth LOS ray. Then, we proceeded to calculate the probablty of the reflected/dffracted rays fallng below a gven threshold for the non-los set, and performed a statstcal analyss of the LOS set. Several ssues, le the varaton wth the elevaton angle and the presence of rays from outsde the buldng, are dscussed along the analyss. Throughout ths paper we employ the term delta to refer to the resultng sgnal from one ray. Ths s the usual term n the sgnal processng feld (a Drac delta). From the vsblty calculaton, we selected the snapshots n whch there s a drect sgnal, and we referred the delays to the maxmum power epoch. Fgure 1 (left) shows the mpulse responses n scenaro P4 (Table 1) for the 30 elevaton crcular flght. The snapshots n whch there s a drect sgnal can be easly detected by loong at the maxmum power level. In Fgure 1 (rght), we have retaned only the snapshots wth suffcent maxmum power level. In addton t s notceable that there s a sgnfcant power at negatve delays n the snapshots n whch there s not a drect sgnal. Ths s because the delay reference has been taen at an epoch Fgure 2. Maxmum snapshot power n some snapshots of the 30 elevaton crcular flght. The red box to the left was zoomed to the rght. In the snapshots wthout LOS ray, t would be convenent to estmate the probablty of a ray fallng below a gven threshold relatve to the LOS level. Ths probablty could be calculated f, for each snapshot, the dstance between transmtter and recever, the antenna gans at the propagaton angles, and all the addtonal losses were nown. Unfortunately, t s not possble to determne the antenna

3 gans for all the propagaton angles, gven that they depend on the helcopter atttude. Besdes, there may be a sgnfcant obstructon of the drect sgnal, even f t s present. The thrd data column of Table 1 contans the rato n db from the average snapshot maxmum ray power to the flght maxmum ray power. We can see that t can reach hgh values (up to 14 db), and ths maes t mpossble to calculate the ln budget from the measurements wth LOS propagaton. For ths reason, we cannot employ the LOS power level at each snapshot as a reference. Instead, we propose to employ the maxmum ray power of the whole flght (crcular or crossover). For a gven level L n db, let Prob L denote the probablty that a delta n a snapshot wthout drect sgnal has a level close to the maxmum ray power (a dfference of less than L db). The frst and second columns of Table 1 contan the estmatons of Prob 15 and Prob 25 for all flghts and scenaros. For a gven flght, let the ndex run through the snapshots wth enough maxmum level, (15 db threshold below the maxmum flght ray power), and let P be the maxmum power n the -th snapshot. The thrd data column n Table 1 contans S = log max{ P }/ mean{ P } ( ) Ths statstc s a measure of the rregularty of the scenaro plus the varatons due to other factors, (LOS ray obstructon, gan varaton wth the helcopter atttude, etc). Whenever there s no drect sgnal, Prob 25 (or Prob 15 ) taes on a hgh value. Ths s because the delta n the -th snapshot wth power P has been produced by a reflecton. Therefore, ths power level s smlar to any other level n the same snapshot, and Prob 25 s hgh. Also, for these flghts, S taes on the smallest values (about 6 db), because max { P } s not set by any drect sgnal. It s dffcult to explan the remanng values due to the varous factors nvolved: helcopter atttude varatons, presence of rays from outsde the buldng, and naccurate LOS reference. N a 1 ϕ 1 τ 1 a, >1 ϕ, >1 τ, >1 No. of rays. PDF obtaned from measurements. Drect ray ampltude. (Input parameter n the tral generator.) Drect ray phase. Unform n [0, 2π]. Drect ray delay. It s taen equal to zero. Reflecton ray ampltude. PDF obtaned from measurements. Reflecton ray phase. Unform n [0, 2π]. Reflecton ray delay. PDF obtaned from measurements. The frst step n calculatng random trals of the ndoor channel s the generaton of the number of rays n the mpulse response. For ths tas we requre the cumulatve dstrbuton (cdf) functon of ths parameter. Fgure 3 shows the estmated cdf n the 30 elevaton crcular flght for the recever on the fourth floor. It s vsble that there s a jump at the abscssas 29 and 30. Ths seems due to the fact that the number of rays was lmted to 30. We can see that ths effect was not produced by the lmted dynamc range (35 db) by loong at Fgure 4. In ths plot, we fnd the number of snapshots wth a gven number of rays nsde the dynamc range. We would expect a smlar number of snapshots for 29 and 30 rays and for a number of rays close to these values, but there s a sharp ncrease. The best soluton s to dstrbute the snapshots wth 29 and 30 rays among the abscssas correspondng to a number of rays greater than 28, assgnng a smlar amount of snapshots to each abscssa, (see Fgure 5). In Fgure 6, we have recalculated the cdf n Fgure 3 by assgnng the mean number of snapshots n the last 10 abscssas to the abscssas greater than 28, so that the number of snapshots wth more than 28 rays s equal to the number of snapshots wth 29 and 30 rays n the uncorrected data. The resultng maxmum number of rays s 45. The tral generator employs ths cdf. STATISTICAL MODEL FOR LOS PATHS In ths secton, we propose a statstcal channel model for the lne-of-sght (LOS) paths n whch the mpulse response s the sum of a drect sgnal plus some delayed, attenuated and phase shfted replcas. The man parameters of the channel mpulse response (delays, ampltudes, number of rays) are consdered as random wth a probablty densty functon that s estmated from the measured data. The mpulse response model s where N j h( t) = a e d( t t = 1 ϕ, ) Fgure 3. Estmated cumulatve dstrbuton functons (cdfs) for the 30 crcular flght on the 4 th floor. Fgures 7-9 show the estmated cdfs n all three floors of the

4 reflected rays relatve powers. Some of the curves are below the probablty 1 value for postve power levels, whch s not consstent wth the normalzaton relatve to the maxmum power level n each snapshot. What happens s that the maxmum power level cannot be calculated from the data, because there s not a sngle snapshot wth drect ray. Therefore, t was necessary to extrapolate the maxmum power level from another flght that contaned snapshots wth LOS component. Specfcally, we chose the snapshots taen on the fourth floor wth drect component as a reference. In ths extrapolaton, t was also necessary to adjust the propagaton loss. In Fgure 10, we assess the qualty of the extrapolaton by calculatng the cdf of the 10 elevaton crcular flght wth and wthout extrapolaton from the 30 crcular flght on the fourth floor. We can see that the maxmum probablty varaton s about 7%. manly produced by drect rays, whereas the low power values are manly produced by reflected rays. Fgure 6. Estmated and corrected cumulatve dstrbuton functons (cdfs) for the 30 crcular flght on the 4 th floor. Fgure 4. Number of snapshots wth a gven number of rays nsde the dynamc range. Measurements on the fourth floor. Fgure 7. Estmated cdfs of the relatve ampltudes of the reflected rays on the ground floor. Fgure 5. Number of snapshots wth a gven number of rays nsde the dynamc range. The values for a number of rays greater than 29 have been extrapolated. Measurements on the fourth floor. Though at frst glance, the power cdfs mght seem Gaussan, we can chec that ths s not the case by loong at the pdfs n Fgure 11. The tals at low powers are longer than the tals at hgh powers. Ths s because the hgh power values are Fgure 8. Estmated cdfs of the relatve ampltudes of the reflected rays on the fourth floor. Fgures plot the cdfs of the reflected ray delays on all three floors. On the ground floor there s a hgher densty of reflected rays wth large delay for the 60 crcular flght than for the 10 and 30 crcular flghts. The measurements on the

5 fourth and sxth floor have a step n the cdfs due to the presence of rays from outsde the buldng (Fgure 13). We can see that there are not reflected rays from outsde the buldng for low elevatons (10 and 30 crcles). Fgure 9. Estmated cdfs of the relatve ampltudes of the reflected rays on the sxth floor. Fgure 22. cdf of the delays of the reflected rays on the ground floor. Fgure 13. cdf of the multpath delays on the fourth floor. Fgure 10. Estmated cdfs wthout and wth maxmum power extrapolaton. Fgure 14. cdf of the delays of the reflected rays on the sxth floor. Fgure 11. Estmated pdfs of the relatve ampltudes of the reflected rays on the fourth floor. Fgures show sequences of measured PDPs for the varous measurement paths. The estmated cdfs are the reference data for the tral

6 generators developed. Fgure 20 shows the bloc dagram of the tral generator. Fgure 18. Flght wth rays from outsde the buldng, (4 th floor, West to East flght). Fgure 15. Impulse responses on the sxth floor for the 10 crcular flght. The green rectangle mars up the multpath outsde the buldng. Fgure 19. Flght wth rays from outsde the buldng, (4 th floor, South to North flght). Fgure 16. Flght wth multpath outsde the buldng, (6 th floor, South to North flght). The nput parameters to the channel smulator select what measurement s gong to be used as a reference. They are, Floor FlghtType NTrals 0, 4, 6. (Avalable floors). 10 crcle, 30 crcle, 60 crcle, Flght over W-E, Flght over S-N Number of trals to generate. The output of the smulator s an array of structures, wth each component correspondng to one tral. The structure contans the followng felds Fgure 17. Flght wth multpath outsde the buldng, (4 th floor, South to North flght). ReflectedRaysDelays ReflectedRaysAmpltudes DrectRayAmpltude Vector of reflected ray delays. Vector of reflected ray ampltudes. Ampltude of the drect ray. It has been normalsed to one,.e., t contans only the phase.

7 DETERMINISTIC CHANNEL MODELING A determnstc model allows the user to descrbe n detal the propagaton envronment and ts boundary condtons. The man advantage of ths approach n comparson wth other approaches,.e., statstcal or emprcal, s that t s able to provde a detaled descrpton of the varous propagaton mechansms nvolved snce the dfferent multpath contrbutons reachng the receve antenna are dentfed and ther magntudes, phases, delays and drectons of arrval (DoA) are calculated. One further relevant advantage of ths modelng approach s that, once the model s approprately fne-tuned after comparson wth measurements, t s possble to use ths technque and repeatedly apply t to a large number of scenaros so that parameters may be extracted and fed to statstcal models whch are more convenent for carryng out system-level studes. Ths procedure was followed later n the development of a statstcal channel model. A ray-tracng approach was chosen as the soluton to the channel modelng problem of nterest,.e., characterzaton of the multpath structure of the receved sgnal n the satellte-to-ndoor channels. The wor performed also dealt wth the aeronautcal channel (wor presented on an accompanyng paper). Several of the developed software modules are common to the study of the two mentoned channels. Ray-tracng results are then fed to a PO (Physcal Optcs) electromagnetc (EM) model for the aeronautcal channel and to a UGTD (Unform Geometrcal Theory of Dffracton) electromagnetc model for the satellte-to-ndoor channel to complete the study. The man modelng steps covered n the smulator mplementaton are: - scenaro descrpton, - ray-tracng and polarzaton-tracng, - electromagnetc modelng, - generaton of results and - model valdaton and fne-tunng. Next, an outlne of the overall process s gven. The propagaton envronment can be descrbed n terms of flat facets of dfferent szes. Envronment data for a number of ctes, arports, sngle buldngs, arplanes, etc. can be found n standard CAD/vrtual realty fle formats such as DXF, IGS, VRML, etc. In other cases, CAD modelng tools (e.g. AutoCAD) can be used to generate, n a convenent way, computer descrptons of the modelng scenaros. The modelng of the envronment does not deal wth geometrcal ssues only. Materal characterzaton for each facet ncluded n the envronment fles must be characterzed from the electromagnetc pont of vew by means of ts constants: ε, µ, σ. Once a standard CAD fle s ready for a gven envronment, t may be translated nto an nternal format to the propagaton modelng tool usng very smple translaton routnes. After ths, ray-tracng and polarzaton-tracng technques may be used to fnd what possble multpath contrbutons are present at the receve antenna. The EM approach proposed for the modelng of the satellteto-ndoor paths s based on UGTD (Unform Geometrcal Theory of Dffracton) [1-4] technques appled to plates and wedges wth flat facets and straght edges. Ray-tracng s, hence, appled to a scenaro made up of flat plates and wedges. For such scenaros t s possble to dentfy three basc types of rays n addton to the drect ray (Dr): - reflected rays (R), - dffracted rays (D) and - transmtted rays (T). Further to these basc ray types, multple rays can also be traced: double, trple,... reflectons, dffractons and transmssons. Also multple combned (mxed) rays can be traced, e.g., RD, DR, RDR, TR, TRR,... Before proceedng on to the actual electromagnetc modelng stage, a polarzaton-tracng study has to be performed. For each nteracton, local coordnate systems must be dentfed to defne both the parallel ( ) and perpendcular ( ) vectors startng off from the transmt antenna polarzaton vector (crcularly polarzed),.e., ts?ˆ and φˆ components, and proceed through the varous nteractons up to the receve antenna. Assumng that a RHCP sgnal s transmtted, each contrbuton reachng the recever wll be depolarzed and wll, n general, consst of both RHCP and LHCP components (Fgure 21). Fgure 21. Polarzaton processng. The actual EM modelng stage s then carred out tang nto account the general representaton of a ray,.e., r r jφ ( p) js E = E p) e C A(?,?, s) e R, D, T ( 1 2

8 where jφ ( p) E( p) e s the ncdent feld, module and phase, on nteracton pont p, C s a coeffcent representng the type of nteracton: reflecton, dffracton, transmsson,? 1? 2 A(? 1,? 2, s) = s an spreadng factor and (? + s)(? + s) js e 1 2 s a phase term due to the traveled propagaton dstance, s, from p to the recever. Ths step provdes nformaton on the ampltudes and phases for both polarzatons for the rays traced n prevous steps, whch already provded nformaton on delay and angle of arrval. Fnally, the results module carres out some smple calculatons such as receve antenna spatal flterng, etc. to produce the needed results,.e., a sequence of scatterng matrces made up of delta functons, two for each traced contrbutons (RHCP and LHCP or CPOL and XPOL) ncludng nformaton of ampltude, phase, delay and angle of arrval. The applcaton of UTD technques requres pror dentfcaton of all possble ray trajectores from the transmtter to the recever. Ths nvolves the applcaton of a number of tracng laws for the varous mechansms nvolved and the verfcaton of the exstence of such rays by checng weather they are obstructed or not. Once all possble ray trajectores have been dentfed polarzaton effects must be accounted for. Crcular polarzaton was used n ths study; ths polarzaton can be broen down nto ts two sphercal coordnate system components,?ˆ and φˆ. For each nteracton wth the propagaton envronment, reflecton, dffracton, transmsson, the ncomng feld may also be broen down nto ts parallel and perpendcular components for whch R, D and T coeffcents are avalable. Fnally, at the receve antenna, a projecton of the electrc feld onto ts local coordnate system s needed. These operatons whch are called here polarzaton tracng are also carred out before proceedng on to carry out the actual EM calculatons. The smulaton tool s further complemented wth a sequencng module capable of defnng transmtter / recever samplng ponts. For the satellte-to-ndoor channel, the tool s able to accept as an nput a sequence of transmtter ponts, helcopter postons representng a flght around the test buldng. The propagaton envronment was descrbed n a smplfed way for the applcaton of determnstc propagaton models. To acheve ths, only the most sgnfcant features,.e., the larger ones, were ncluded n the nput fles. The approach was to descrbe the propagaton envronment n terms of flat plates. These plates should, n prncple, be suffcently large so that the applcaton of the propagaton approach, UTD, s possble/vald. UTD technques were used on flat plates and straght edges (wedges). Ths approach s vald for propagaton scenaros nvolvng buldngs, whch can be descrbed n terms of large plates. Fgures 22 and 23 show two AutoCAD vews of the test buldng. In Fgure 23 the rays traced for a gven recever locatons are also shown. The use of avalable commercal products, more specfcally AutoCAD, was selected to generate the buldng envronments to be used n the model valdaton and fnetunng tass. Resultng propagaton scenaro fles n.dxf format were translated nto the nternal format used by the propagaton-modelng tool. Both, the.dxf and the nternal format, are ASCII text fles so that they can also be easly read usng text edtors. Fgure 22. Example of one test buldng floor n AutoCAD. Fgure 23. Example of the test buldng AutoCAD and traced rays. The three types of basc contrbutons sought, R, D and T, correspond to dfferent propagaton mechansms that can be descrbed n terms of the Snell s law for reflectons, the Keller's law for dffracton (Fgure 24) and the Snell's law for refractons (transmssons) (Fgure 25). Snell's law of reflecton states that θ = θ r, whle Snell's law

9 of refracton states that sn? e2µ 2 n2 = = sn? t e1µ 1 n1 where n 1 and n 2 are the refractve ndexes for both meda. The refractve ndex of a medum s the rato of the free space velocty, c, to the phase velocty n the medum,.e., n c e µ er r v e µ µ. = = = 0 0 Moreover, Snell s law states that the ncdent ray, the normal vector to the nterface at the refracton pont, and the refracted ray are n the same plane. In the context of outdoor propagaton, the nfluence of the transmtted rays s very low. When a ray s transmtted through an external buldng wall, t s rapdly attenuated due to the subsequent obstacles that t encounters after the transmsson. The buldng walls have a fnte wdth such that when a ray s transmtted through a wall. Fgure 25. Transmsson through walls [3]. Equatng both expressons, yelds? 0 =?, whch means that 2 the transmtted ray extng the wall on the other sde s parallel to the ncomng ray. As for the dffracton law or Keller's law, t states that a ray oblquely ncdent on the edge of an obstacle, wth an angle β, defnes a cone, wth angle β, of possble dffracted ponts as llustrated n Fgure 24. The dffracted rays propagate from the dffracton pont to the observaton pont satsfy sˆ eˆ = sˆ eˆ where ŝ s the unt vector n the ncdent ray, ŝ s the unt vector n the dffracted ray and ê s the vector along the edge. Fgure 24. Dffracton law. Keller's cone of dffracton. The followng assumptons have been made wth regard to transmssons through walls: (a) The wall medum s homogeneous and sotropc. (b) The two nterfaces are locally plane at the transmsson ponts. Ths s a common assumpton n the context of a GO (Geometrcal Optcs) approach. Applyng Snell's laws of reflecton and transmsson, t can be observed how the ncdence angle on the frst ar-wall nterface s equal to the ext angle on the second nterface,.e., wall-ar. For the frst nterface (Fgure 25), and for the second nterface, ß 0 sn(? 0 ) = ßsn(? 1) ß 2 sn(? 1 ) = ß 0 sn(? ). The ray tracng tool mplements the followng ray and ray combnaton types: - R, RR, RRR. - D, DD. - RD, DR. - RRD, DRR, RDR. In addton to the rays of the above types, t s possble to trace the same rays but ncludng any number of transmssons. Pror to ray-tracng, acceleraton technques may be appled [2]. Such technques are mportant when the scenaros under analyss are large,.e., ctes, and when the number of smulaton ponts s also large,.e., test routes sampled wth a spacng of the order of the wavelength. Acceleraton technques are also needed when the dynamc range of the

10 sgnal s suffcently large that hgher-order nteractons are sgnfcant n the overall receved sgnal as t s the case n terrestral moble communcatons. For navgaton systems the dynamc range s much smaller and only sngle and, at most, double nteractons are meanngful. In case no acceleraton technques are employed, a systematc scan of all planes for transmssons and reflectons and edges for dffractons s carred out. Ths s so for frstorder nteractons. For second order nteractons nested scans (nested "for" loops) are needed, and so on for hgher-order nteractons. To carry out the analyss and comparson between measurements and smulatons, a secton of the helcopter flghts around the Joanneum Research buldng was selected. Ths secton corresponds to flght angles for whch the helcopter was n lne-of-sght, LOS, or partal LOS condtons wth respect to the recever. The only obstacle was the starcase stuated between the wde panel wndow loong out to the outsde of the buldng, ths s llustrated n Fgure 26. ncludes these long-delayed echoes, for valdaton purposes the comparsons wll be lmted to ths 20 ns delay wndow. Fgure 27. llustrates the dstrbuton of tmes of arrval n the measured PDPs (power delay profles). In the smulatons the followng ray types were consdered: R, RR, D, DD, RD and DR. Fgure 27. Measured nsde-outsde echoes. Fgure 26. Vew of the measurement locaton nsde the Joanneum Research buldng. Fgure 28. Measured Power Delay Profle for flght pont 1. The smulatons carred out reproduce the measurement condtons,.e., the recever was located on the 4th floor of the Joanneum Research buldng, the transmtter was located on a helcopter and the ln elevaton angle was on the order of 30º elevaton. Twelve measurement ponts were avalable correspondng to the LOS secton of the flght as mentoned above. Thus twelve smulaton ponts were set n order to perform the valdaton of the model. One of the conclusons that can be drawn from frst observaton of the measured and smulated data s that the paths wth delays up to 20 ns correspond to nteractons occurrng drectly wthn the buldng. Paths wth longer delays are assumed to orgnate at scatterng ponts outsde the buldng, whch have not been consdered n the smulaton scenaro. Thus, even though the actual channel Fgure 29. Smulated Power Delay Profle for flght pont 1. It can be observed how frst and second order dffractons produced on the spral starcase contrbute an mportant share of the overall receved power found n the smulatons gven that they show sgnfcant db levels n comparson to some of

11 the reflected rays and ther delays ft n the 20 ns wndow mentoned above. In Fgures 28 and 29 one example of measured and smulated PDPs s gven. The mean excess delay and RMS delay spread are two useful parameters. These values are measures of the severty of a multpath envronment. For a gven set of multpath sgnal powers, P and correspondng delays, τ, the mean and RMS excess delays are calculated as: D = t p / p and 2 2. S = t p D One fnal parameter of nterest can be the Rce factor or carrer-to-multpath rato, defned for LOS condtons as K( db) = 10 log10( p ). Table 2 shows a comparson between measured and smulated results. A farly good agreement was found between the measured and smulated parameters. Smlarly, the measured and smulated PDPs also show good smlartes. Table 3 shows the lst of materals used n the smulatons. Table 2. Measured and smulated values of parameters D, S, K for the twelve measurement ponts. CONCLUSIONS In ths paper, we have presented the analyss of satellte-tondoor channel measurements. In the case of non lne-of-sght propagaton, we have analyzed the probablty of a reflected ray fallng below a gven threshold (15 db and 25 db) relatve to the LOS level. In the case of LOS propagaton, we have presented the cdfs of the number of rays, and of ther ampltudes and delays. The results show that t s unlely to have out-of-buldng rays at low elevatons. For the LOS case, we have developed a statstcal model n whch the rays n one mpulse response are modeled ndependently wth random ampltudes and delays. Fnally, we have presented a generator of random mpulse response trals that mplements ths model, and whose nput data are the cdfs estmated from the measurements. Through determnstc, ray-tracng based modelng, t has been confrmed that wall attenuaton s the man propagaton element, whch drastcally reduces n-buldng coverage. Ths coverage s lmted, except for very rare cases, to rooms wth external walls. For these external wall rooms and hallways t has been found that accountng for dffracted rays helps mprove the predctons based only on reflected rays. In ths way, a larger number of rays are traced that mach more closely the dstrbuton of rays and the power balance n the measurements. ACKNOWLEDGEMENTS The wor presented n ths paper was carred out under an European Space Agency, ESA, contract enttled "Navgaton sgnal measurement campagn for crtcal envronments", carred out by JOANNEUM RESEARCH (Austra) wth subcontracts to DLR (Germany) and the Unversty of Vgo (Span), are presented. REFERENCES Table 3. Propertes consdered for the varous materals n the smulaton scenaro. Materal ε r σ (S/m) Thcness Doors cm Internal Walls cm Parttons btwn floors cm External Walls cm Wndows cm Stars cm [1] R.G.Kouyoumjan and P.H.Patha A unform geometrcal theory of dffracton for an edge n a perfectly conductng surface. Proc. IEEE, Vol 62, pp , 1974 [2] F. Aguado, A. Formella; J.M.Hernando, F.Isas and F. Perez Fontan. Effcent ray-tracng acceleraton technques for rado propagaton modelng. IEEE Transactons on Vehcular Tech., Vol. 49, No. 6, Nov. 2000, pp [3] I. de Coster. Determnstc propagaton predcton for wreless communcaton systems PhD Thess. Katholee Unverstet Leuven, [4] G. A. J. van Dooren. A determnstc approach to the modelng of electromagnetc wave propagaton n urban envronments. PhD Thess, TUE, Endhoven.

12 Scenaro Ground floor (P5) Fourth floor (P4) Sxth floor (P1) Flght type Table 1. Analyss of the measurements wthout drect ray. Probablty of beng above the 15 db lmt ( Prob 15 ) Probablty of beng above the 25 db lmt ( Prob 25 ) Flght Max. Power to Flght Average Max. power rato (S, db) Presence of drect sgnal on the whole flght Multpath outsde the buldng 10 crcle Yes No 30 crcle Yes No 60 crcle No No Fly over S-N Yes No Fly over W-E No No People movng n lobby No No 10 crcle Yes No 30 crcle Yes No 60 crcle No Yes Fly over S-N Yes Yes Fly over W-E Yes Yes 10 crcle Yes Yes 30 crcle No No 60 crcle No No Fly over S-N Yes Yes Fly over W-E No No Cdfs obtaned from the measured data Delay, Ampltude, Phase Delta Delays Delay, Ampltude, Phase Number of replcas Complex ampltude s Delta Impulse response Phases Delay, Ampltude, Phase Fgure 20. Bloc dagram of the tral generator. Delta Complex ampltude of drect sgnal Delta Phase