PART 7 Test Cases Involving Mobile Links

PART 7 Test Cses Involving Mobile Links

7.- CHAPTER 7. Simultion Approch Editor: Simon R. Sunders Authors: Mrtin Döttling 2, Clude Oestges 3 Centre for Communiction Systems Reserch, University of Surrey, Guildford, Surrey, GU2 5XH, UK Tel.: +44-483-25986, Fx.: +44-483-25954, e-mil: S.Sunders@ee.surrey.c.uk 2 Siemens AG, System Engineering / UMTS Development, ICM MP TI 6, Hidenupltz, D-8667 Muenchen, Germny Tel. +49-89-722-5873, Fx: +49-89-722-2789, E-Mil: Mrtin.Doettling@mch.siemens.de 3 Université ctholique de Louvin (UCL), Microwve Lbortory (EMIC), Plce du Levnt 3, B-348 Louvin-l-Neuve, Belgium Tel.: +32--4785, Fx.: +32--47875, e-mil: Oestges@emic.ucl.c.be

7.-2 7. Simultion Approch The work described in prt ws initited with the following objectives: To demonstrte tht the mobile propgtion models described in Section 4 cn be pplied to rel system performnce simultions. This helps to verify the usefulness of the individul models nd demonstrtes the generl method of ppliction to potentil users of these models. To show tht simultion pproches other thn the conventionl brute-force Monte-Crlo methods commonly followed cn yield dvntges in terms of llowing estimtion of system nd link performnce with reltively little computtionl effort. The generl pproch followed in the simultions is described in chpters 7.-, 7.-2 nd 7.-3. The pproch is then followed for two prticulr test cses: nrrowbnd LEO 66 TDMA system (see Chpter 7.2); widebnd S-IMT-2 system bsed on the Deligo constelltion nd CDMA ir interfce(see Chpter 7.3). In ech of the test cses, results from selection of the Prt 4 models re presented. 7.. Approch A two-level simultion pproch is followed, split into: Link-Level simultion to determine the dependence of rw nd coded BER on stellite, environment nd user prmeters, vi n nlyticl expression for the reltionship between instntneous signl-to-interference-plus-noise rtio (SINR) nd rw BER (see Figure 7.-). Stellite Environment Chnnel Model User SINR to BER mpping Eb/No Pe Time series or CDF Sttistics of Rw BER DeInterleve FEC correction/detection cpbilities Sttistics of coded BER/FER Figure 7.-: Link level simultion

7.-3 The results from the link level simultion re then implemented s look-up tbles within the System-Level simultion, which is used with constelltion prmeter genertor nd pproprite criteri for vilbility to determine overll system vilbility results (see Figure 7.-2). Stellites Environment Genertion of Time Series Link Modelling Sttistics of coded BER Avilbility Evlution Link Avilbility Sttistics Figure 7.-2: System level simultion The key reson for this split pproch is to dopt the simplest method tht will yield meningful outputs. Conventionl Monte-Crlo bit-by-bit simultion is voided s being too time-consuming. Note tht this simultion pproch (nd probbly ny other method cpble of yielding the sme level of informtion) requires chnnel models cpble of generting representtive time series. Thus purely sttisticl models tht re only cpble of predicting cumultive ttenution sttistics, such s those recommended by [ITU-R, 68-6, 999], re inpproprite. This is prticulrly the cse when considering coded modultion schemes or time-dependent lgorithms such s power control. 7..2 Link Simultion Simultions re mndtory for predicting the performnce of lnd mobile stellite (LMS) communiction systems. For system vilbility clcultions verstile two-level simultion pproch is used, which seprtes link-level nd system-level simultion. The former ims t deriving sttistics of coded bit error rte (BER) from propgtion models, wheres the ltter uses the results of the link-level simultion to clculte the service vilbility sttistics for dedicted opertionl scenrios. The gol of this contribution is to pproximte the function of chnnel coding without using expensive nd time-consuming commercil system simultion pckges. This llows free distribution mong the COST 255 members nd forms n open, simple nd well-defined interfce for different propgtion models. Moreover, the system simultions within COST 255 re not

7.-4 intended s in-depth studies of dedicted systems, but to highlight nd illustrte the ppliction of stte-of-the-rt LMS propgtion modelling. Bsiclly two different strtegies exist for the clcultion of coded BER from propgtion dt: simultion nd nlyticl pproches. Simultion softwre genertes rndom bit errors nd implements the receiver function to correct these errors nd retrieve the informtion. Vrious commercil nd cdemic softwre pckges exist. In generl, considerble computtionl effort is necessry nd the ppliction nd interfcing with different propgtion models is not strightforwrd. Therefore the COST 255 link-level simultion uses nlyticl formuls, which require only limited computer resources nd llow the use of both, nrrowbnd nd widebnd chnnel models. After brief review of nlyticl bounds on coded BER for convolutionl codes, the COST 255 link-level simultion pproch is outlined. 7..2. Anlyticl Bounds on coded BER For convolutionl codes with known trnsfer function tight upper bound on the BER for soft decision Viterbi decoding is derived in [Prokis, 995]: BER t R d E b c, ()< d erfc 2k () N t β 7.- d= d free where β d is fctor obtined from the derivtive of the code trnsfer function, the so-clled code weight distribution [Prokis, 995]. d free is the free distnce of the code, R = k/n is the code rte nd E b /N the informtion bit energy per spectrl noise density. For lrge E b /N vlues the symptotic performnce is given by βd free Eb BERc, 2()< t erfc R dfree 2k () N t 7.-2 If the code trnsfer function is not known the convolutionl chnnel coding theorem my be used [Viterbi nd Omur, 979], [Prokis, 995] which yields the ensemble verge error rte of convolutionl code on discrete memoryless chnnel: k kkr ()/ t R 2 2 ( ) BERc, 3()< t R R t k( R ( t) R)/ R 2 () 7.-3 2 [ ] where K is the constrint length nd R is the cut-off rte defined by [Viterbi nd Omur, 979], [Prokis, 995]: R t R E b ( N t ) log e 7.-4 ()= + 2 Unfortuntely, equtions 7.-2 nd 7.-3 do not hold for smll vlues of E b /N. In this cse eqution 7.- hs to be used. Figure 7.-3 shows comprison of these bounds for R = 3/4, K = 7 convolutionl code, s used by Iridium nd in the LEO66 test cse of COST 255. The first ten terms of the code weight

7.-5 distribution re tken into ccount in eqution 7.- [Hccoun nd Bégin, 989], [Cin et l., 979]. coded BER - -2-3 -4-5 -6-7 -8 Eq. Eq. 2 Eq. 3 2 3 4 5 6 7 8 E b /N in db coded BER - -2-3 -4-5 -6-7 -8 Eq. Eq. 2 Eq. 3 2 3 4 5 6 7 8 E b /N in db ) b) Figure 7.-3: coded BER for: ) R = 3/4, K = 7 (LEO66) ;b) R = /3, K = 9 convolutionl code (S-IMT2). For E b /N vlues less thn 4.5 db the symptotic formul (see eqution 7.-2) devites significntly from the tight upper bound in eqution 7.- nd cn be regrded s too optimistic. The convolutionl chnnel coding theorem is only ccurte enough in smll E b /N rnge. The required E b /N for trget BER of -2 is 3. db. 7.-3b shows the sme comprison for the R = /3, K = 9 convolutionl code, which is proposed for future Stellite-IMT 2 nd used in the corresponding COST 255 test cse. This chnnel code hs considerbly higher coding gin. The difference between the symptotic curve nd eqution 7.- becomes importnt for E b /N less thn 3.5 db. The upper bound of eqution 7.-3 is too loose for prcticl pplictions. For BERs of -3 nd -6, the required E b /N is 2. db nd 3.9 db, respectively. 7..2.2 Link-Level Simultion The link-level simultion is bsed on time series of effective E s /N, which re provided by model of Lnd Mobile Stellite propgtion. The use of effective E s /N llows simple interfce to both, nrrow-bnd nd wide-bnd propgtion models. For ech time series of input dt the men vlues of bit energy per spectrl noise density (E b /N ) nd the men coded BER re clculted. Additionlly sttistics of E b /N nd coded BER cn be displyed nd stored. Definition nd Clcultion of Effective E s /N The E s /N vlues re ssumed to be provided with sufficient temporl nd sptil resolution nd to include ll effects of system noise, multi-user interference nd inter-symbol interference. Thus E s /N is relted to the signl-to-noise rtio by: E N s ()= t N B sys St Rs () t Nisi() t Nmi() t, 7.-5 + + R R N () s c

7.-6 where S(t) is the received signl power, N sys (t) the system noise, B N the noise bndwidth, N isi (t) the equivlent noise due to inter-symbol interference, N mi (t) the multiple ccess interference, R s the chnnel symbol rte nd R c the chip rte for CDMA systems. This generic definition llows to investigte time division multiple ccess (TDMA), frequency division multiple ccess (FDMA) nd code division multiple ccess (CDMA) systems s well s combintions of these ccess schemes. While for TDMA/FDMA system noise will limit E s /N, N mi is dominnt in CDMA systems. For TDMA/FDMA schemes one might dditionlly set R c = R s nd use N mi to represent co- nd neighbouring chnnel interference. Simple nd nrrowbnd models my ssume constnt vlues for the noise contributions in eqution 7.-5. However, this definition is open enough to enble the use of more dvnced models, tht use time-vrint noise due to multiple ccess interference nd inter-symbol interference, e.g. [Döttling et l., 999]. The clcultion of the signl nd noise terms in eqution 7.-5 from widebnd dt is bsed on the complex polrimetric trnsmission mtrix T i (t), which is provided by the chnnel model for ech stellite nd every time step t. Additionlly, the dely time τ i (t) nd direction of rrivl (ϑ i,ψ i ) re given for ll N(t) relevnt rys t time t. After weighting ech ry with the ntenn pttern C R of the mobile terminl the complex trnsmission fctor A i (t) is obtined [Döttling et l., 999]: ()= ( ) () A t C ϑ, ψ T t 7.-6 i R T i i i where the superscript T denotes the trnspose opertion. The superposition of ll contributions yields the power dely profile, which is equivlent to the stellite's time-vrint chnnel impulse response Nt () ( )= () ( ) h τ, t A t δ τ τ ( t), t. 7.-7 ch i i i= The received signl in the n-th symbol period T s cn be expressed s yr( τ, t)= h l tot n l T s, t l= 7.-8 = h (, t)+ h n lt s, t, n tot ([ ] ) l tot l= l n ([ ] ) where the first term represents the contribution of the wnted symbol n nd the second term the influence of ll other symbols [Miller et l., 993], [Prokis, 995]. The totl impulse response h tot is given by: ( )= ( ) ( ). 7.-9 h τ, t h τ h τ, t tot sys ch The system impulse response h sys of LMS systems is often of the rised-cosine filter type. By evlution of equtions 7.-7, 7.-8 nd 7.-9 the wnted signl power yields: 2 Nt () 2 St () = λ PG T TPR Ai() t hsys i(), t t τ 4π i ( ) 7.-

7.-7 nd the corresponding inter-symbol interference power is N () t = λ P G P 4π isi T T R 2 Nt () l= i= l n () ([ ] ) Ai t hsys n l T s τi(), t t. 2 7.- Although N isi is most often negligible for tody's systems, it my influence future widebnd systems. For CDMA systems multiple ccess interference is the limiting fctor for E s /N. For the downlink the MAI is function of the received power P s = S s + N isi,s from ech visible stellite s nd the number of equivlent chnnels C eq per stellite: N () t P t C mi St () = () s= s s eq. 7.-2 P s is clculted by the propgtion model for ll S(t) visible stellites. For synchronous orthogonl CDMA s = 2 (i.e. no MAI from the serving stellite), or else s =. C eq depends on the number of co-chnnel bems, bem overlp, chnnel utilistion, voice ctivity nd ntenn discrimintion. Comprehensive derivtions of C eq cn be found in [De Gudenzi nd Ginnettti, 998] nd [Monsen, 995]. In the uplink, the chnnel sttes of the interferers re not known nd resonble ssumptions hve to be mde for the men vlue of N mi. Clcultion of Effective E b /N nd coded BER The corresponding E b /N is clculted ccording to the modultion index M nd the code rte R = k/n: Eb N t ()= R log M 2 Es () N t. 7.-3 As point of reference, the rw bit error rte BER r for BPSK nd QPSK cn be clculted by [Prokis, 995]: BER t r Eb N t erfc. 7.-4 ()= () 2 In the COST 255 link-level softwre, the coded BER is clculted using equtions 7.- nd 7.-3. For ech chnnel time series corresponding time series of E b /N nd coded BER re generted, s well s the corresponding probbility density functions nd the cumultive distribution functions [Döttling nd Sunders, 999]. Additionlly, the men vlue of E b /N nd the men coded BER re clculted.

7.-8 7..3 System-Level Simultion This section detils the system-level simultion method in the context of Lnd Mobile Stellite systems, s discussed in Section 7.-. The inputs of the link-level simultion softwre re : time-series of E b /N vlues; system prmeter file; severl simultion prmeters. The vilbility/coverge estimtion is bsed on new point of view, which differentites temporl vilbility nd sptil coverge. 7..4. Time series simultion nd evlution of performnce Assuming tht the elevtion ngle is kept constnt, time series of fde levels hve to be generted on the bsis of propgtion model. The fde time series re then esily converted into effective E b /N time series. Introducing the resulting E b /N time series into the link model, sttistics of BER re obtined for the environment. We note T X Θ ( x ϑ ), the sttisticl distribution of ny of these prmeters (X = coded BER or effective E b /N ), conditioned to the elevtion ngle. The time series tht re generted hve to include ll effects tht induce vrition of the mobile chnnel, nmely the re surrounding the terminl s well s the zimuth ngle, which continuously chnge over time. The zimuth ngle reltive to the street xis depends on both the rndom loction of the receiver (rndom distribution) nd the time-vrying stellite position (smooth vrition). In order to simplify, it is ssumed here tht the zimuth ngle is only ssocited with the mobile position. Therefore its vrition should be tken into ccount within the propgtion modelling. Performnce is usully expressed in terms of specified BER which is not exceeded for given percentge of time. For exmple, the LEO 66 performnce for voice trnsmission given in the next section is defined by BER of -2 with globl vilbility of 95 %. In this study, distinction is mde between temporl vilbility (denoted s vilbility ) nd sptil coverge. The ultimte gol is to define the performnce for given re, such s lrge city nd its suburbs, by performnce criterion which is not exceeded during τ % of the time nd over γ % of the potentilly usble re, e.g. non built-up res. The performnce criterion is for exmple the verge rw or coded BER/FER, etc. When the elevtion ngle is kept constnt ( ϑ = ϑ ), the conditionl distribution of X (e.g. coded BER, effective E b /N ) is equivlent to distribution over the re for this prticulr vlue of ϑ. The probbility density function of X is indeed clculted on the bsis of time series. These re governed by the chrcteristics of the re nd of the terminl movements tht feed the propgtion model. On the other hnd, the vrying elevtion ngle is relted to the stellite movement (depending on the system), nd thus to time. From the previous considertions, the conditionl distribution TX Θ ( x ϑ = ϑ ) cn be seen s sptil distribution (i.e. over the re) t fixed time given by ϑ. Since the criterion X is usully monotonic function of the elevtion ngle, knowing TX Θ ( x ϑ = ϑ ) mkes it possible to obtin sptil distribution chieved during certin cumulted percentge of time τ. The ltter is given by:

7.-9 TX ( τ ) = TX Θ ( x ϑ > ϑ ) π 2 τ = p[ ϑ > ϑ = ] TΘ ( ϑ ) dϑ ϑ 7.-5 where T X ( τ ) stnds for the distribution over the re of performnce prmeter X, for percentge of time τ. The sttisticl distribution of elevtion ngle ϑ cn be esily clculted from orbit genertors for ny ltitude on the globe. For exmple, the distribution of the elevtion ngle for the LEO66 system cn be well pproximted by log-norml distribution. If n nlyticl distribution cnnot be found, the probbility density function of elevtion ngle hs to be numericlly defined. A system-level simultion progrm hs been developed to implement the formlism outlined bove. Its inputs re: Time series of E b /N, for rnge of elevtion ngles; Vector of elevtion ngle corresponding to the time series; Stellite system to be simulted; Ltitude (the user cn here choose between two ltitudes: London nd Rome, Itly); Prmeter X to be represented (X = coded BER or E b /N ). The progrm produces severl curves of re distribution of X, T X ( τ ), ech being relted to certin time percentge τ.

7.- 7..4 References [Cin et l., 979] Cin, J.B., Clrk, G.C., Geist, J.M., "Punctured convolutionl codes of Rte (n-)/n nd Simplified Mximum Likelihood Decoding", IEEE Trns. on Informtion Theory, vol. IT-25, no., pp. 97-, 979. [De Gudenzi nd Ginnettti, 998] De Gudenzi, R., Ginnetti, F., "DS-CDMA Stellite Diversity Reception for Personl Stellite Communiction: Stellite-to-Mobile Link Performnce Anlysis", IEEE Trns. on Vehiculr Technology, vol. 47, no. 2, pp. 658-672, 998. [Döttling et l., 999] Döttling, M., Zwick, T., Wiesbeck, W., "Investigtion of Stellite Diversity nd Hndover Strtegies in Lnd Mobile Stellite Systems bsed on Ry Trcing Propgtion Model," Proc. Int. Mobile Stellite Conference. IMSC'99, Ottw, Cnd, 999, pp. 28-33, 999. [Döttling nd Sunders, 999] Döttling, M., Sunders, S., "Bit Error Rte Clcultion for Stellite Communiction Systems", Proc. COST Joint Interntionl Workshop COST 252, COST 253, COST 255, Toulouse, Frnce,. pp. 5-55, 999. [Hccoun nd Bégin, 989] Hccoun, D., Bégin, G., "High-Rte Punctured Convolutionl Codes for Viterbi nd Sequentil Decoding", IEEE Trns. on Communictions, vol. 37, no., pp. 3-25, 989. [ITU-R, 68-6, 999] Interntionl Telecommuniction Union,"ITU-R Recommendtion P.68-6: Propgtion dt required for the design of erth-spce lnd mobile telecommuniction systems", Genev, 999. [Miller et l., 993] Miller, M.J., Vucetic, B., Berry, L., "Stellite Communictions Mobile nd Fixed Services", Kluwer Acdemic Publishers, 993. [Monsen, 995] Monsen, P.,"Multiple-Access Cpcity in Mobile User Stellite Systems," IEEE Journl on Selected Ares in Communictions, vol. 3, no. 2, 995, pp. 222-23, 995. [Prokis, 995] Prokis, J.G., "Digitl Communictions", Third Edition, McGrw-Hill, Inc., New York, 995. [Viterbi nd Omur, 979] Viterbi, A.J., Omur, J.K., "Principles of Digitl Communictions nd Coding", McGrw-Hill, Inc., New York, 979.

7.2- CHAPTER 7.2 LEO 66 Mobile Link Editors: Jörg Hbeth, Simon Sunders 2 Authors: Mrtin Döttling 3, Jörg Hbeth, Clude Oestges 4, Fernndo Pérez-Fontán 5, Mryn Ángeles Vázquez-Cstro 6, Werner Wiesbeck 7 Philips Reserch Lbortories, 'Digitl Comm. nd Networking', Weisshusstr. 2, D-5266 Achen, Germny Tel.: +49-24-63-56, Fx: +49-24-63-59, e-mil: joerg.hbeth@philips.com 2 Centre for Communiction Systems Reserch, University of Surrey, Guildford, Surrey, GU2 5XH, UK Tel.: +44-483-25986, Fx.: +44-483-25954, e-mil: S.Sunders@ee.surrey.c.uk 3 Siemens AG, System Engineering / UMTS Development, ICM MP TI 6, Hidenupltz, D-8667 Muenchen, Germny Tel.: +49-89-722-5873, Fx: +49-89-722-2789, E-Mil: Mrtin.Doettling@mch.siemens.de 4 Université ctholique de Louvin (UCL), Microwve Lbortory (EMIC), Plce du Levnt 3, B-348 Louvin-l-Neuve, Belgium Tel.: +32--4785, Fx.: +32--47875, e-mil: Oestges@emic.ucl.c.be 5 Universidd de Vigo, Escuel Superior de Ingenieros de Telecomunicciones, Deprtmento de Tecnologís de ls Comunicciones, Cmpus Universitrio, 362 - Vigo, Spin Tel.: +34-986-8237, Fx.: +34-986-822, e-mil: fpfontn@tsc.uc3m.es 6 Universidd Crlos III de Mdrid, Escuel Politécnic Superior, Deprtmento de Tecnologís de ls Comunicciones, Avd. de l Universidd, 3, 289-Legnés, Mdrid, Spin Tel.: +34-9-624973, Fx.: +34-9-624943, e-mil: mryn@tsc.uc3m.es 7 Universität Krlsruhe, Institut für Höchstfrequenztechnik und Elektronik, Kiserstr. 2, D-7628 Krlsruhe, Germny, Tel.: +49-72-68-6257, Fx.: +49-72-69865, e-mil: Werner.Wiesbeck@etec.uni-krlsruhe.de

7.2-2 7.2 LEO 66 Mobile Link The simultion pproch outlined in Chpter 7. will now be pplied to first test cse system. This is Low Erth Orbit (LEO) stellite system using constelltion of 66 stellites t n ltitude of 78 km. For the determintion of the most importnt prmeters of the system, dt vilble from the IRIDIUM stellite system hve been used. The IRIDIUM system hs lredy been introduced in Chpter 6.4, in which its feeder link hs been studied. In this chpter we will del with the mobile link between the user nd stellite. Even though the system hs lredy been described we will briefly mention the min chrcteristics of the system once gin. After short description of the system, the most importnt system prmeters s well s the link budgets for the mobile user link will be presented in form of tbles. Most of the dt re tken from the originl FCC ppliction of Motorol Inc. [Motorol, 99]. The 66 stellites of the test cse system re eqully distributed on 6 orbitl plnes, ech contining stellites. The plnes re inclined t 86 with respect to the equtoril plne. The system is designed to gurntee miniml elevtion ngle of 8 bove the horizon with t lest one stellite visible to the user t ny time. Ech stellite genertes cluster of 48 cells on the Erth s surfce, formed by phsed rry L-bnd ntenns. Vrible ntenn gins nd utomtic power control re used to compenste for the different pth losses from innermost to outermost cells nd vrying trnsmission conditions. Frequency reuse within stellite footprint follows 2 cell cluster pttern. The constelltion of stellites is controlled by System Control Segment (SCS) which is, mong other things, responsible for mintining the stellites in proper orbit. Gtewy sttions on the ground provide the connection with the public switched telephone network (PSTN) nd control the cll processing within the system. The gtewys re connected to the stellites vi high gin prbolic trcking K-bnd ntenns. Ech gtewy comprises 5 different min entities besides the dtbses HLR (Home Loction Register), VLR (Visitor Loction Register) nd EIR (Equipment Identity Register) ([Armbruster nd Lurin, 996] nd [Hutcheson nd Lurin, 995]). As most of the cll processing in the IRIDIUM system is bsed on GSM (Globl System for Mobile Communictions) protocols nd procedures, the MSC (Mobile Switching Center) is the hert of gtewy. A MSC is connected to n Erth Terminl Controller (ETC). The ETC is nlogous to the Bse Site Subsystem of terrestril GSM system nd controls t lest three Erth Terminls (ETs). The ETs provide the physicl connection to the constelltion using their K-bnd ntenns. One ET is responsible for mintining the trffic-bering connection with the current stellite pssing overhed. The second ET is involved in building connection to the next rising stellite nd the third ET provides redundncy in cse of hrdwre or softwre filure nd eventully rin diversity. Within the gtewy, pging services re hndled by specil entity, clled the Messge Originting Controller (MOC). The Gtewy Mngement System (GMS) supports nd orgnises opertion nd mintennce in the other entities of the gtewy. Mobile users communicte with the constelltion vi hnd-held mobile voice terminls (hndhelds) or pgers clled Messge Termintion Controllers (MTD). The ntenn of subscriber unit is Qudrifilr Helix L-bnd ntenn. Iridium subscriber links between the stellite nd the mobile operte in the frequency rnge from 6 to 626.5 MHz, but the FCC hs only licensed opertion from 62.35 to 626.5 MHz in the USA. Modultion type is QPSK with symbol rte of 5 kbit/s on the mobile user link.

7.2-3 Figure 7.2- illustrtes the IRIDIUM burst structure. The multiple ccess is combintion of FDMA, TDMA nd TDD. The foreseen TDMA frme hs durtion of 9 ms nd contins four uplink nd downlink chnnels respectively s well s pging or ringing chnnel. The dt rte is 2.4 kbit/s for dt nd 2.4 or 4.8 kbit/s for voice trffic. TDMA-Frme = 9 ms Gurd Time Burst Time Pging- Signllingchnnel Uplink Uplink 2 Uplink 3 Uplink 4 Downlink Downlink 2 Downlink Downlink 3 4 2.32 ms 8.28 ms ms.24 ms.22 ms. ms Figure 7.2-: IRIDIUM burst structure [Wlke, 999] System cpcity cn be clculted s follows: The 5.5 MHz of vilble spectrum re divided into 24 crriers, ech contining four full duplex chnnels, which results in 496 chnnels in totl. Figure 7.2-2 shows the frequency pln of the IRIDIUM system s it should look fter the frequency ssignment by the FCC. The cell cluster mkes it possible tht ech frequency cn be reused four times within the 48 cells of stellite footprint. Consequently ech stellite would hve theoreticl cpcity of 984 trffic chnnels. Due to the stellite s restricted bttery power cpcity of 4 W, stellite cn only crry bout connections t the sme time [Wlke, 999]. 4.67 khz 5 kbit/s QPSK 66.5 khz 3.5 khz 626.5 M Figure 7.2-2: IRIDIUM Frequency Pln [Wlke, 999] Overll BER objectives re -2 for voice nd -6 for dt trnsmission. The vilbility objective for the user link is vilbility for 95 % of the time. 7.2. System Prmeters System prmeters re summrised in Tbles 7.2- (generl prmeters), 7.2-2 (link prmeters), 7.2-3 (stellite prmeters) nd 7.2-4 (terminl prmeters). Tbles 7.2-5 nd 7.2-6 show link budgets for the stellite-mobile down- nd uplink (in one specific cell ring - No. 5), tken from the originl IRIDIUM FCC ppliction [Motorol, 99].

7.2-4 Gtewy (Erth Sttion) Loction 4 58 North, 2 35 Est, 6 m High Orbit type LEO (circulr) Orbit ltitude 78 km No. of stellites 66 No. of orbitl plnes 6 Inclintion 86.4 Min. Elevtion ngle 8 Inter-stellite links (ISL) 4 per St. Tble 7.2-: Generl System Prmeters Up-link frequency bnd 62.35-626.5 MHz Down-link frequency bnd 62.35-626.5 MHz Net (uncoded) chnnel dt rte 2.4 kbit/s Modultion scheme QPSK Symbol Rte 5 kbit/s Polristion Right circulr System Frequency Response rised cosine, roll-off fctor.4 Error protection Convolutionl r=3/4, K=7 Bit error rte speech -2 Bit error rte dt -3 Link Mrgin 6 db Required Eb/N 3. db for BER -2 Avilbility 95 % Chnnel bndwidth 3.5 khz Multiple ccess method FDMA / TDMA Duplex scheme TDD Tble 7.2-2 Link Prmeters

7.2-5 Antenn type Plnr phsed rry Emitter burst power.5-.8 dbw (depending on cell position) Emitter net ntenn gin 8.2-22.3 dbi (depending on cell position) Emitter feeder loss.3 db Emitter EIRP 2.3-3.7 dbw (depending on cell position) Receiver net ntenn gin 8.2-22.3 dbi (depending on cell position) Receiver ntenn noise temperture 29. K Receiver feeder loss.8 db Receiver noise figure. db Receiver noise bndwidth 26. khz Receiver G/T (-9.2)-(-5.) dbi/k (depending on cell position) No. of cells per stellite 48 Frequency reuse 2 cell cluster No. of chnnels per stellite 2.4 kbit/s (power limited) Stellite trnsmission mode Regenertive / Switching Tble 7.2-3: Stellite Prmeters Antenn type Emitter burst power Emitter net ntenn gin Emitter feeder loss Emitter EIRP Receiver net ntenn gin Receiver ntenn noise temperture Receiver feeder loss Receiver noise figure Receiver noise bndwidth Receiver G/T Qudrifilr Helix.6-8.5 dbw (depending on cell position).-3. dbi (depending on cell position).7 db.9-.3 dbw (depending on cell position).-3. dbi (depending on cell position) 5. K. db.8 db 28. khz (-23.8)-(-2.7) dbi/k (depending on cell position) Tble 7.2-4: Terminl Prmeters

7.2-6 With Shdowing No Shdowing Azimuth Angle (deg) 3. 3. Ground Rnge (km) 377.9 377.9 Ndir Angle (deg) 56.3 56.3 Grzing Angle (deg) 2.4 2.4 Slnt Rnge (km) 643.3 643.3 HPA Burst Power (W) 7.2 2.3 dbw 8.6 3.7 XMTR Pek Ant Gin (dbi) 23. 23. Edge Loss (db).5.5 Scn Loss (db) 3. 3. Tper Loss (db).. Net XMTR ANT GAIN (dbi) 7.5 7.5 XMTR Feed / Ckt Loss (db).3.3 EIRP (dbwi) 24.8 9.9 Pth Loss (db) 6.9 6.9 Polristion Loss (db).5.5 Atmos Absorption Loss (db).3.3 Men Vegettion Loss (db) 8.2. TOT PROPAG. LOSS (db) 69.9 6.7 RCVR Ant Net Gin (dbi).2.2 RCV'D SIG LEVEL, C (dbw) -43.9-4.7 RCVR Antenn N-Temp (K) 5. 5. RCVR Feed/Ckt Loss (db).. LNA Noise Figure (db).8.8 RCV SYS NOISE TEMP, Ts (K) 298.9 298.9 G/Ts (dbi/k) -23.6-23.6 RCVR Noise BW, B (khz) 28. 28. Sensitivity = ktsb (dbw) -49.4-49.4 RCV'D C/N (db) 5.4 8.7 C/I (db) 8. 8. RCV'D C/(N+I) (db) 5.2 8.2 RCV'D C/(N+I) (db) 59.7 62.7 Required Eb/N (db) 3.2 3.2 Modem Impl Loss (db) 2. 2. REQUIRED C/(N+I) (db) 5.2 5.2 LINK MARGIN (db). 3. FLUX DENS (dbw/sq-m, 4 khz) -27.5-32.5 Tble 7.2-5: Stellite-Mobile Link Budget - Downlink [Motorol, 99]

7.2-7 With Shdowing No Shdowing Azimuth Angle (deg) 3. 3. Ground Rnge (km) 377.9 377.9 Ndir Angle (deg) 56.3 56.3 Grzing Angle (deg) 2.4 2.4 Slnt Rnge (km) 643.3 643.3 HPA Burst Power (W) 6.9 2.2 DBW 8.4 3.5 NET XMTR ANT GAIN (dbi).2.2 XMTR Feed / Ckt Loss (db).7.7 EIRP (dbwi) 8.9 4. Pth Loss (db) 6.9 6.9 Polristion Loss (db).5.5 Atmos Absorption Loss (db).3.3 Men Vegettion Loss (db) 8.2. TOT PROPAG. LOSS (db) 69.9 6.7 RCVR Pek Ant Gin (dbi) 23. 23. Edge Loss (db).5.5 Scn Loss (db) 3. 3. Tper Loss (db).. NET RCVR ANT GAIN (dbi) 7.5 7.5 RCV'D SIG LEVEL, C (dbw) -43.6-4.2 RCVR Antenn N-Temp (K) 29. 29. RCVR Feed/Ckt Loss (db).8.8 LNA Noise Figure (db).. RCV SYS NOISE TEMP, Ts (K) 552.6 552.6 G/Ts (dbi/k) -9.9-9.9 RCVR Noise BW, B (khz) 26. 26. Sensitivity = ktsb (dbw) -5.2-5.2 RCV'D C/N (db) 6.6 9.9 C/I (db) 8. 8. RCV'D C/(N+I) (db) 6.3 9.3 RCV'D C/(N+I) (db) 57.3 6.3 Required Eb/N (db) 4.3 4.3 Modem Impl Loss (db) 2. 2. REQUIRED C/(N+I) (db) 6.3 6.3 LINK MARGIN (db) -.6 3. FLUX DENS (dbw/sq-m, 4 khz) -42.5-44.9 Tble 7.2-6: Mobile-Stellite Link Budget - Uplink [Motorol, 99]

7.2-8 7.2.2 Simultions in Non-Urbn Ares bsed on 3D Ry Trcing Chnnel Model In this section results of simultions in non-urbn res, bsed on three-dimensionl (3D) ry opticl chnnel model, re shown. Two different system-level simultion pproches re used to post-process the chnnel dt. The first uses the COST 255 system-level softwre described in Chpter 7.. First chnnel chrcteristics for single stellite t fixed elevtion ngles re clculted nd the elevtion sttistics for different stellite constelltions nd different ltitudes re considered subsequently. This llows cler seprtion of chnnel chrcteristion nd stellite constelltion effects: one series of chnnel predictions cn be pplied to vrious constelltions nd loctions on Erth. However, this pproch does not llow investigtion of stellite hndover or stellite diversity. Hence second series of system-level simultions is performed tht is bsed on chnnel simultions using n orbit genertor. Although those simultions re more site- nd constelltion-specific, they include vrious other effects such s: Multiple concurrently visible stellites; Chnging stellite elevtion with time; Shdowing correltion; Time-vrint multiple ccess interference (MAI) in CDMA systems; Correltion between stellite diversity gin nd MAI in the downlink; Dependency of hndover nd diversity gin on signlling dely. First brief outline of the widebnd chnnel model nd the post-processing steps re given to mke the reder fmilir with the modelling pproch. The opertionl scenrio (i.e. the dt used to simulte the terrestril environment) nd specific LEO 66 test cse system prmeters re described next. Then, link-level nd system-level results re shown nd discussed. 7.2.2. The Propgtion Model The recently developed chnnel model Döttling et l., 999] used for the following investigtion is bsed on detiled nd deterministic description of the environment. In contrst to previously published model [Döttling et l., 998], the new model performs loclly three-dimensionl (3D) ry trcing nd therefore includes further rys nd voids pproximtions which re used in the former, prtly two-dimensionl, model. A comprehensive comprison of the two different models shows the improved performnce of the 3D model, especilly in producing continuous time series with relistic short term fding nd polristion effects [Döttling et l., 999]. The terrestril prt of the propgtion problem is simulted using digitl mps of topogrphy nd lnd use (cf. Figure 7.2-3), single rodside objects (like buildings nd trees) in vector formt nd mobile pth with rbitrry trjectory nd speed. The stellites' view ngles nd reltive velocity vectors re either user-specified or clculted by n orbit genertor. The ry trcing is bsed on n dpted version of the sweep line lgorithm [Brtuschk et l., 997], [Agelet et l., 997]. It ccounts for 3D wedge nd corner diffrction using Uniform Geometricl Theory of Diffrction (UTD), reflection from ground nd rodside objects, rough surfce scttering from terrin, s well s for trnsmission loss in vegettion lyers [Döttling et l., 999].

7.2-9 For ech stellite nd every time step t the complex polrimetric trnsmission mtrix T i (t), dely time τ i (t) nd direction of rrivl (ϑ i,ψ i ) re clculted for ll N(t) relevnt rys. After weighting ech ry with the ntenn pttern C R of the mobile terminl (MT) the complex trnsmission fctor A i (t) is obtined. The superposition of ll contributions yields the power dely profile, which is equivlent to the stellite's time-vrint chnnel impulse response: Nt () ( )= () ( ) h τ, t A t δ τ τ ( t), t ch i i i= 7.2-7.2.2.2 Signlling Dely For LMS systems propgtion time hs to be considered. Especilly the efficiency of system control commnds is impired by the signlling dely, which cn be pproximted by [Döttling et l., 999b]: ts()= t nst tst( h, εgw, ε mt() t )+ tpr, 7.2-2 where n st is the number of signl trips between gtewy nd MT required to complete the signlling, t st the corresponding propgtion time nd t pr n dditionl dely due to signl processing in the control instnces. The impct of the slnt rnge on t st is clculted bsed on orbit height h nd the stellite elevtions s seen from the gtewy (ε gw ) nd from the mobile terminl (ε mt ). 7.2.2.3 Power Control (PC) The system's power control is simulted by user-defined number of power control steps, the corresponding thresholds nd power correction steps. Additionlly, the trget signl-to-noise rtio (SNR), the dynmic rnge, s well s the power control updte rte re specified by the user. The power-controlled received power P s (t) is clculted by: ()= ( ) () P t γ t t P, t, 7.2-3 s s s nom where γ is the power control fctor nd P s,nom the received power for nominl trnsmitted power [Döttling et l., 999b]. 7.2.2.4 Hndover Modelling This study considers lso the fesibility nd benefit of stellite hndover (HO) bsed on SNR mesurements, since this type of hndover requires the highest signlling effort nd lods the system resources notbly. To mitigte shdowing, fst hndover scheme is necessry. It is bsed on constnt monitoring of the SNR vlues of ll stellites. If the SNR of the ctive stellite flls below certin mrgin SNR HO with respect to the best stellite, HO is initited. To prevent system congestion with signlling, mximum HO rte cn be specified. Bsed on these prmeters, time series of bit energy to spectrl noise density (E b /N ) is clculted, which includes the effect of hndover nd signlling dely t s. The performnce of the hndover scheme is evluted by compring the corresponding cumultive distribution function (CDF) of E b /N. Further detils cn be found in [Döttling et l., 999b]. 7.2.2.5 Stellite Diversity Modelling Another wy to increse system vilbility is stellite diversity. The SNR vlues of ll stellites re l i d F b h lli di i i i h i lli i

7.2- estblished nd power-controlled. The ctul number of stellites in the ctive set is determined by SNR dd nd SNR drop. If the SNR level of n inctive stellite is less thn SNR dd below the best stellite, request for dding this stellite to the ctive set is issued. In similr wy, request for dropping stellite from the ctive set is trnsmitted if its SNR vlue is more thn SNR drop below the best stellite. Both ctions occur with signlling dely. The superposition of the ctive stellites' signls is performed in the combiner. In this context only mximum rtio combining nd two-brnch diversity re considered. A comprison between different combining schemes nd two- nd three-brnch diversity is given in [Döttling et l., 999b], [Döttling et l., 999c]. In diversity systems the trde-off between diversity gin nd system loding due to the L(t) chnnels per connection hs to be considered. The PDF nd CDF of E b /N fter combining llows evlution of the diversity gin, while the cost in terms of system cpcity is determined by the men number of ctive chnnels. 7.2.2.6 Polristion Diversity Modelling Mesurement cmpigns [Agius et l., 999] nd simultions [Döttling et l., 999c] hve recently emphsised the potentil of polristion diversity t the mobile terminl in certin propgtion conditions. This diversity technique does not ffect system cpcity dversely, since only one trffic chnnel is required. In this investigtion polristion diversity with mximum rtio combining is considered. Throughout the simultions right hnd circulr polristion (RHCP) is used t the trnsmitter. The primry receiver ntenn is co-polrised. The second MT ntenn is ssumed to hve n identicl ntenn pttern nd to be mtched to the left hnd circulr polristion (LHCP). 7.2.2.7 The Opertionl Scenrio The simultions re bsed on topogrphicl nd lnd use dt of 2km x 2km re in the Rhine Vlley ner Krlsruhe, Germny. The z-xis in Figure 7.2-3) shows topogrphicl height, the lnd use clsses re shown in different greys. The lndscpe shows typicl non-urbn mixture of terrin nd lnd use elements. The mobile trjectory is depicted s white rrow. The grid resolution is 5 m. Rodside trees nd buildings re generted stochsticlly, with vrying sttistics for density, height nd loction ccording to the lnd use clss of the mobile position [IMST, DLR, IHE, 998]. By this wy syntheticlly nd utomticlly generted fisheye imge comprble to those in [Akturn nd Vogel, 997] cn be obtined. To illustrte this input dt, Figure 7.2-3 depicts snpshot of the LEO 66 simultions, showing the mximum elevtion cused by topogrphy nd lnd use, s well s the silhouettes of rodside trees nd buildings s seen from the mobile ntenn. The bsciss gives the zimuth ngle normlised to the momentry heding of the MT nd the ordinte is the elevtion ngle.

7.2- open/griculture medow deciduous forest coniferous forest dense urbn 2km urbn industril street wter mobile pth height in m 3 2km Figure 7.2-3: Topogrphy, lnd use nd mobile pth in the Rhine Vlley ner Krlsruhe, Germny. Tble 7.2-7 shows simultion prmeters tht re common to ll simultions. The simultions bsed on stellite elevtion sttistics use points nd 7 elevtion ngles from 5 to 85. prmeter signl dynmic rnge trnsmitted polristion simultions using stellite elevtion sttistics 3dB RHCP simultions using n orbit genertor smpling time ms 9ms simultion time min mobile speed m/s, 5m/s, 3m/s 3m/s totl pth length m, 5m, 3m 8 m Chnnel impulse responses per simultions 24 24 797 Tble 7.2-7: Generl Simultion Prmeters For ech elevtion ngle, 2 different zimuth ngles re simulted to perform sufficient verging over zimuth, yielding totl number of over 2 chnnel impulse responses. The smpling intervl is ms to resolve the chnnel coherence time nd to include ll relevnt chnges in the propgtion environment. Chrcteristics of pths used for simultions bsed on stellite elevtion sttistics re summrised in Tble 7.2-8. Pth simultes pedestrin user wlking t m/s. Mobile Speed [m/s] Totl Pth Length [m] Pth Pth 2 5 5 Pth 3 3 3 Tble 7.2-8: Chrcteristics of pths used for simultions using stellite elevtion sttistics

7.2-2 The orbit genertor simultions use 6655 simultion points. The number of visible stellites is timend constelltion-dependent. For the LEO 66 test cse 24 797 chnnel impulse responses re clculted. The smpling time is 9 ms, since these investigtions re intended to include the effect of the timevrying stellite positions nd to verge over lrger re of mobile loctions. 7.2.2.8 System simultion Prmeters The bseline simultions use isotropic MT ntenns nd no power control. Additionl simultions re performed to investigte the impct of the MT ntenn pttern, the efficiency of power control (PC), stellite hndover, s well s stellite nd polristion diversity. The min system prmeters relevnt for simultions re listed in Tble 7.2-9. prmeter LEO 66 simultions signlling dely n st = 2 power control 3-bit closed-loop power control ±db dynmic rnge 9ms updte intervl trget E b /N of 9.dB stellite hndover stellite diversity polristion diversity 6dB hysteresis 5ms processing time 9ms updte intervl 2-brnch mximum rtio combining 6dB hysteresis 9ms updte intervl mximum rtio combining Tble 7.2-9: Min LE 66 Simultion Prmeters 7.2.2.9 Simultions using Stellite Elevtion Sttistics In this section results re presented, tht use the chnnel model outlined bove in combintion with the COST 255 system- nd link-level softwre. Three different mobile pths re simulted, (see Tble 7.2-8). The MT ntenn pttern is ssumed to be either isotropic or tht of typicl low-directivity hndheld ntenn. The ltter ntenn pttern suppresses predominntly ground-reflected signls, the discrimintion of elevtion ngles greter thn 5 is lwys below 3 db. The LEO 66 stellite elevtion sttistics re tken from Chpter 7..3. Minly four different types of grphs re shown: the smples of E b /N, to show the results of the chnnel model, the cumultive distribution function (CDF) of E b /N for different elevtion ngles, the complementry cumultive distribution function (CCDF) of coded BER for different elevtion ngles (link-level simultion results), the verge coded BER versus percentge of re nd time using the LEO 66 stellite elevtion distributions for ltitudes of Rome nd London (system-level simultions results).

7.2-3 The chnnel model results for stellites t 25, 45 nd 65 elevtion re depicted in Figures 7.2-4 nd 7.2-5 pth (v = m/s) nd pth 3 (v = 3m/s), respectively. A thousnd smples correspond to ech of the twelve different zimuth ngles, which re seprted by 3. It is obvious tht the shdowing probbility s well s the short term fding depth decreses with incresing elevtion ngle. E b /N in db 3 25 2 5 5-5 25 45 65-2 3 4 5 6 7 8 9 2 smple no. Figure 7.2-4: E b /N smples for 25, 45 nd 65 elevtion, pth, smples correspond to one zimuth ngle 3 25 2 5 5 25 45 65-5 - 2 3 4 5 6 7 8 9 2 smple no. Eb /N in db Figure 7.2-5: E b /N smples for 25, 45 nd 65 elevtion, pth 3 The corresponding E b /N sttistics re shown in Figures 7.2-6) nd 7.2-9). Figures 7.2-6b) nd 7.2-9b) show how the distributions of E b /N trnslte into distributions of coded BER. The legends for these nd subsequent plots re given in Figure 7.2-7. Tble 7.2- summrises the probbilities tht the trget BER of -2 is exceeded t elevtion ngles of 5, 25, 45 nd 65 for the different simultions.

7.2-4 probbility (E b /N < bsciss).8.6.4.2 - -5 5 5 2 25 3 E b /N in db probbility (BER > bsciss).7.6.5.4.3.2.. -6-5 -4-3 -2 - BER ) b) Figure 7.2-6: CDF of E b /N () nd CCDF of BER (b) for pth, isotropic ntenn 5 5 2 25 3 35 4 45 5 55 6 65 7 75 8 85 % 99.98% 97.63% 87.3% 69.9% 5.58% 35.97% 24.8% 5.9%.33% 6.67% 4.29% 2.77%.79%.6%.76% ) b) Figure 7.2-7: Legends for subsequent plots: ) elevtion ngle in CDFs of E b /N nd CCDFs of BER b) Time percentges for system-level simultion results t the ltitude of Rome Elevtion ngle Pth Pth 2 Pth 3 Pth ntenn Pth PC 5 48.2% 48.2% 35.9% 55.5% 28.2% 25 36.2% 2.% 2.3% 37.% 6.% 45 7.8% 2.4% 4.6% 8.2% 6.% 65.2%.2% 4.3%.5% 2.8% Tble 7.2-: Probbility of BER> -2 for different simultion pths nd elevtion ngles System-level results re given in Figures 7.2-8 nd 7.2-9b. Figure 7.2-8 shows the impct of different ltitudes on system performnce for pth. In Figure 7.2-8 the stellite elevtion distributions for the ltitudes of London nd Rome re used. At the trget BER nd % time percentge, the difference in percentge of re between both plots is only.3% (97.3% for Rome nd 97.6 % for London).

7.2-5 verge coded BER - -2-3 -4-5 verge coded BER - -2-3 -4-5 -6 9 92 94 96 98 percentge of re -6 ) b) 9 92 94 96 98 percentge of re Figure 7.2-8: Averge coded BER for London () nd Rome (b), pth, isotropic ntenn probbility (E b /N < bsciss).8.6.4.2 - -5 5 5 2 25 3 E b /N in db verge coded BER - -2-3 -4-5 -6 ) b) 9 92 94 96 98 percentge of re Figure 7.2-9: CDF of E b /N (left) nd verge coded BER for Rome (right), pth 3, isotropic ntenn The corresponding grphs for pth 3 (see Figure 7.2-9b) t the ltitude of Rome show how vritions in chnnel sttistics yield different system performnces. Despite the notble difference in the E b /N sttistics versus elevtion ngle, the differences in vilbility re only round.5% for BER = -2 nd % of time. For lower time percentges the difference decreses even more (Tble 7.2-). Time Percentge Pth Pth 2 Pth 3 Pth ntenn Pth PC % 97.3% 97.8% 97.7% 97.% 98.6% 5.6% 98.2% 98.% 98.% 98.% 99.%.3% 99.3% 99.% 98.9% 99.2% 99.8%.2%.%.% 99.9%.%.% Tble 7.2-: Percentge of re tht the trget BER is chieved for the ltitude of Rome Since pth shows the worst propgtion conditions, it is used to investigte the impct of the MT ntenn pttern nd the efficiency of power control. Figure 7.2- shows the corresponding E b /N

7.2-6 distributions. The low-directivity hndheld ntenn pttern increses the brech of the trget BER by 7.3% t 5 elevtion, by.8% t 25 elevtion nd by.4% t 45 elevtion (Tble 7.2-). At the ltitude of Rome, this corresponds to decrese of covered re of.2% for % time percentge (Tble 7.2-). probbility (E b /N < bsciss).8.6.4.2 - -5 5 5 2 25 3 E b /N in db verge coded BER - -2-3 -4-5 -6 ) b) 9 92 94 96 98 percentge of re Figure 7.2-: CDFs of E b /N for pth including hndheld ntenn pttern (left) nd Averge coded BER for Rome (right) The efficiency of 3-bit closed-loop power control [De Gudenzi nd Ginnetti, 998] is illustrted in Figure 7.2-. The left side () shows the percentge of smples within n intervl of ± db round the trget E b /N of 9. db, versus elevtion ngle. Both curves (with nd without power control) show minimum round 25 elevtion. The increse of smples within this intervl is below % for ll elevtions. The benefits of PC seem to be more pronounced for medium elevtion ngles, since for low elevtion ngles the signlling dely is incresed nd the fding depth exceeds the PC dynmic rnge. For high elevtion ngles, there re few fding events, which need to be power-controlled. At 5 nd 25 elevtion, the power control llows dditionl 2% of the smples to chieve the trget BER. At 45 the increse is.8% nd t 65 still.6%. For the Rome elevtion sttistics, this corresponds (b) to increses in covered re of.3%,.8%, nd.3% t %, 5.6%, nd.3% of the time, respectively. Further investigtions of different PC schemes re given in [Döttling et l., 999b], [Döttling et l., 999c]. smples within ±db of trget E b /N in % 8 6 4 2 no PC with PC increse 2 3 4 5 6 7 8 9 elevtion in verge coded BER - -2-3 -4-5 -6 ) b) 9 92 94 96 98 percentge of re Figure 7.2-: Pth, isotropic ntenn, power control: () percentge of smples within ±db of trget E b /N nd (b) verge coded BER for Rome

7.2-7 7.2.2. Simultions using n Orbit Genertor The simultions bsed on the elevtion sttistics represent long-term men vlues, which do not include temporl nd locl effects. To investigte the impct of such momentry influences (such s street orienttion, stellite positions nd shdowing correltion) the simultions performed in this section use n orbit genertor to produce time series of stellite positions insted of the bove mentioned elevtion sttistics. The totl simultion time is min. Figure 7.2-2 shows the timevrint elevtion ngles for ll visible stellites. Since the highest stellite t the beginning of the simultions is no longer optiml towrds the end, stellite hndover (HO) is ssumed in the bseline simultions. Figure 7.2-3 compres the E b /N time series of this reference scenrio with dditionl nd polristion diversity (PD) t the mobile terminl, respectively. Notble improvement cn be seen especilly under shdowing conditions. Tble 7.2-2 shows tht in this specific test cse (note the high speed of the MT), polristion diversity is superior to power control, leding to 8.4% fewer smples tht exceed the trget BER of -2 with respect to the bseline scenrio (HO). Moreover, due to the signlling dely, power control leds to stronger fluctutions round the trget vlue of E b /N, including unnecessry high signl mplitudes (cf. Figure 7.2-4). 9 8 7 6 5 4 3 2 elevtion in 5 5 2 25 3 35 4 45 5 55 6 time in s Figure 7.2-2 Time series of stellite elevtions for LEO 66 simultions 3 25 2 Eb /N in db 5 5-5 - HO HO, PD 2 3 4 5 6 time in s Figure 7.2-3: E b /N time series for stellite hndover nd stellite hndover with polristion diversity (HO, PD Simultion HO HO, PC HO, PD SD BER > -2 3.9% 7.6% 2.5% 26.% Tble 7.2-2: Comprison of different strtegies to increse system performnce

7.2-8 Since the LEO66 stellite constelltion is not optimised for multiple stellite visibility, there is no significnt gin from stellite diversity (SD). Considering the decresed system cpcity (men number of occupied chnnels is.4 per connection), stellite diversity is not fvourble option for this constelltion. Figure7.2-4 compres the PDFs nd CDFs of E b /N for the different strtegies to increse system performnce. It hs to be noted, tht for idel mximum rtio combining, the polristion diversity gin is round db t E b /N = -7dB, greter thn 8dB t E b /N = db nd thn 7dB t E b /N = 3.dB. These results gree very well with observtions in experimentl cmpigns [Agius et l., 999]. The mjor dvntge from polristion diversity occurs just in the E b /N rnge where it is most useful. Further comprisons of different hndover nd diversity schemes re provided in [Bischl nd Werner, 997], [Döttling et l., 999b] nd [Döttling et l., 999c]..3.25 probbility density.2.5..5 HO HO, PC HO, PD SD 2 4 6 8 2 22 24 E b /N in db probbility (E b /N < bsciss).8.6.4.2 HO HO, PC HO, PD SD ) b) -5 - -5 5 5 2 25 E b /N in db Figure 7.2-4: PDF nd CDF of E b /N for different LEO 66 system simultions 7.2.2. Conclusions of the simultion bsed on 3D Ry Trcing Chnnel Model This section shows the ppliction of 3D ry trcing propgtion model to the LEO 66 test cse. Mjor dvntges of this pproch include the inherent considertion of shdowing correltion, polrimetric widebnd chnnel chrcteristion, the considertion of signlling dely nd the verstile post-processing fcilities to investigte power control, stellite hndover, stellite diversity nd polristion diversity. The model cn be used either to derive chnnel sttistics for single stellite scenrios t fixed elevtion ngles, or to simulte whole stellite constelltions (including effects of multiple visible stellites nd vrying stellite elevtion ngles). Due to its physicl bckground, the pproch is vlid in wide rnge of different frequency bnds nd opertionl scenrios. Arbitrry stellite constelltions cn be investigted. The results show the predominnt dependence of E b /N nd BER sttistics on the elevtion ngle. While the trget BER is exceeded for lmost hlf of the smples t 5 elevtion, the brech is below 8% for ll simultions t 45 nd decreses stedily with elevtion ngle. Three different mobile pths hve been investigted, showing tht for comprble opertionl scenrios, the mximum vribility of the system performnce prediction is in the rnge of.5% for % of time nd decresing with decresing time percentge. Comprison of the system-level results for the ltitudes of London nd Rome show only slight differences below %. Assuming n idel low-directivity hndheld ntenn pttern, the decrese in system performnce is negligible for LMS systems with high link mrgins. A closed-loop power control scheme increses system performnce t the trget BER by 3% to 8%. The bove results indicte tht similr or better improvement cn be chieved by polristion diversity t the mobile terminl

7.2-9 A comprison between the orbit genertor-bsed simultions nd the results obtined from elevtion sttistics show tht the temporry performnce (even for periods up to min) cn be significntly worse thn the long-term verge. 7.2.3 Simultions in built-up res bsed on physicl-sttisticl ry-trcing Severl simultions hve been performed using physicl-sttisticl pproch bsed on ry-trcing. Typicl urbn nd suburbn res re defined on the bsis of lognorml distribution (medin µ, stndrd devition σ) for the building height, nd constnt vlue w for the street width. Tble 7.2-3 summries the prmeters for ech environment. Type of re Building height µ [m] σ Street width w [m] Urbn (bsed on Westminster dt) Suburbn (bsed on Guildford dt) 2.6.44 5 7..25 5 Tble 7.2-3: Physicl prmeters for urbn nd suburbn types of re In order to use the system-level simultion pproch previously detiled, time series re generted for given elevtion ngles. The mobile terminl is ssumed to move down long stright streets, lined on both sides with buildings. The orienttion ngles of streets reltive to the stellite links cn be described by uniform zimuth distribution. The dtbse is built using rndom genertor whose inputs re the sttisticl distributions of physicl prmeters describing the re (building height, street width). The fde level reltive to the line-of-sight is estimted using ry-trcing method bsed on the UTD nd equivlent currents, s detiled by [Oestges et l. 998]. A trget E b /N vlue of 9. db is chosen s mentioned in Section 7.2- nd the coded BER is clculted s lredy described (see Chpter 7.). 7.2.3. Simultion results Figure 7.2-5 presents comprison between the distributions of E b /N s function of the sptil coverge in the two res nd for the two ltitudes (London nd Rome). The vrious curves should be relted with the following percentges of time (from left to right) :., 99.99, 98.88, 9.8, 77.5, 6.4, 43.63, 3.35, 2.5, 3.5, 8.93, 5.83, 3.79, 2.47,.6 nd.6 % of time if considering the ltitude of London;., 99.98, 97.63, 87.3, 69.9, 5.58, 35.96, 24.8, 5.9,.33, 6.66, 4.29, 2.77,.79,.6 nd.75 % of time if considering the ltitude of Rome.

7.2-2 ) b) Figure 7.2-5: Percentge of re over which E b /N is less thn ordinte, in suburbn () nd urbn (b) res for severl percentges of time Concerning the suburbn scenrio, the five first curves s well s the two lst curves re superimposed. The sme remrk is vlid for the three lst curves with regrd to the urbn cse. It is cler tht the curves remin the sme t both ltitudes; only the percentge of time relted to ech curve differs. The grphs of Figure 7.2-5 should be red s follows. For exmple, in the urbn cse t the ltitude of London, the men E b /N is less thn 5 db during : 99.98 % of time on 2.6% 87.3 % of time on 3.58 % 5.58 % of time on 28.2% of the considered re. 24.8 % of time on 35. %.33 % of time on 4.79 % Figure 7.2-6 llows to compre the performnce for both ltitudes (upper figure) s well s for both environments (lower figure). The upper figure shows wht is gined in terms of bsolute vilbility for the ltitude of Rome reltive to tht of London s function of the percentge of re over which E b /N is less thn 5 db. We see tht for both types of res, the vilbility cn be incresed by some 8 % t the ltitude of Rome with regrd to tht of London. The urbn curve is nturlly right-shifted, since the coverge is worse in n urbn environment. Concerning the improvement of performnce in terms of coverge, we see on the lower grph tht the results re quite the sme for both ltitudes, nd tht the bsolute gin vries between 5 % nd 3 %. This comprison represents mjor dvntge of the pproch, since it is bsed on physicl model, nd produces sttisticl results in terms of vilbility/coverge directly relted to physicl prmeters, such s building height nd street width.

7.2-2 Figure 7.2-6: Comprison - between vilbilities for both ltitudes s function of coverge (upper grph) nd between coverges for both res s function of vilbility (lower grph) The time series of fde lso led to the prediction of coded BER, s previously detiled. Figure 7.2-7 presents the coverge for number of vilbilities. The coverge is defined by the percentge of re over which the coded BER is less thn certin vlue. The vrious vilbilities depend on the ltitude, the percentges of time relted to ech curve remining the sme s those found for E b /N curves. ) b) Figure 7.2-7: Percentge of re over which the coded BER is less thn ordinte, in suburbn () nd urbn (b) res for the ltitudes of London nd Rome nd for severl percentges of time

7.2-22 For instnce, in urbn re t the ltitude of Rome, coded BER of -3 is not exceeded during: 99.98 % of time on 89.88 % 87.3 % of time on 9.86 % 5.58 % of time on 92.48 % of the considered re. 24.8 % of time on 92.92 %.33 % of time on 94.63 % It hs to be noticed tht in suburbn re, the six lst curves on the fr right re superimposed. Results my seem slightly optimistic, but this cn prtly be explined by the fct tht only three zimuth ngles (, 45 nd 9 degrees reltive to the street xis) were simulted in the time series. Since the xil cse ( degree) is ctully line-of-sight cse, its contribution in the verge process tends to overestimte the performnce. Agin, we find tht the suburbn scenrio is significntly less degrded thn the urbn one t the sme ltitude. The ltitude of Rome lso ppers to be more fvourble thn tht of London for the cse of the Iridium constelltion. Figure 7.2-8 illustrtes the improvement of performnce when compring results t both ltitudes or for both res. ) b) Figure 7.2-8: Comprison - between vilbilities for both ltitudes s function of coverge (upper grph) nd coverges for both res s function of vilbility (lower grph) The vilbility t the ltitude of Rome is up to 8 % higher thn t tht of London. Menwhile, we see on the lower grph tht for trget BER of -3, the coverge (bsolute vlue) is incresed by up to 6 % in the suburbn re reltive to the urbn scenrio. 7.2.3.2 Conclusion The simultion results demonstrte tht the system-level nlysis cn be successfully combined with physicl-sttisticl model bsed on ry-trcing pproch with regrd to nrrowbnd

7.2-23 mobile stellite system. Since the combined method quite directly links the sttisticl properties of the physicl re with the system performnce it is very dvntgeous. 7.2.4 Simultions in Urbn nd Suburbn Ares bsed on Widebnd Mrkov Chnnel (WMC) propgtion model The min focus of this section is on detiled study of some key fetures relted to the link-level performnce evlution of the LEO 66 Lnd Mobile Stellite system. The key chrcteristics of the performnce evlution to be investigted re those which cn be evluted directly from time series of chnnel mplitude generted with chnnel model. System level simultions results re lso presented. For performnce physicl-level evlution of 3 rd genertion networks single simultor pproch would be preferred but the complexity of such simultor - including everything from trnsmitted wveforms to multi-cell network nd world coverge - is fr too high. Therefore seprte link nd system level simultors re needed, even though it is not cler wht kind of formt is required for the link level simultion outputs from the point of view of the system level simultor. In this section, the interfce to connect link nd system level simultions is lso ddressed. A widebnd Mrkov chnnel propgtion model hs been used for genertion of chnnel time series. The model is dequtely described in Chpter 4.3 nd brief outline is given subsequently. Simultions hve been crried out for urbn nd suburbn res. The opertionl scenrio is described next. To stremline the reding, the foregoing nomenclture to specify different prmeters nd simultions results re in greement with similr topics described in other sections. Only the uplink is investigted here s it is considered to be the most criticl. For instnce, power mesurement is more complex since for the forwrd link, power is mesured continuously on pilot. It is worth clrifying tht simultions described in this section re lwys sttic simultions in the sense tht dynmic spects of the system producing rndomness in system prmeters re not tken into ccount. This mens for exmple tht simultions investigting the power control lgorithm, which is rdio resource mngement lgorithm consider sttic E b /N o trget. Consequently, dynmic spects such s net cpcity vribility due to hndover nd diversity, increse of other cell interference or compenstion of fst fding, lie beyond the scope of this section. Specific prmeters considered for this simultion re listed in Tble 7.2-4. Prmeter Elevtions 5º - 45º 45º - 65º 65º - 85º Power control LEO 66 simultions 3-bit closed-loop power control updte times (T pc ): from ms (T pc = T f ) to 2 ms (T pc = 2 T f ) 2 db dynmic rnge trget E b /N of 9. db Tble 7.2-4: Specific LE 66 Prmeters Used for Simultions bsed on WMC model

7.2-24 7.2.4. The Propgtion Model The Widebnd sttisticl model [Fontán et l., 997] nd [Fontán et l. 998], used for the following investigtion is bsed on the 3-stte Mrkov Model described in Chpter 4.3. A bsic feture of the model is tht it is ble to generte time-series of ny chnnel prmeter whose study is required: signl envelope, phse, instntneous power dely profiles, Doppler spectr, etc. As second step, conventionl sttistics, e.g. Cumultive Distribution Functions, re computed fterwrds from the generted series. The model mkes the simplifying ssumption of the existence of three bsic rtes of chnge (very slow, slow nd fst) in the received signl, corresponding to the different behviours of its components. As for the slow vritions, they represent smll-scle chnges in the shdowing ttenution produced s the mobile trvels in the shdow of the sme obstcle or shdowing vritions behind single building or group of buildings. In the line-of-sight (LOS) stte slow vritions my be due to different resons: for exmple, non-uniform receiving ntenn ptterns nd/or chnges in mobile orienttion with respect to the stellite. Correltion distnce, describing how fst the log-normlly-distributed vritions re, is lso considered in the model. Sttes represent different gross shdowing conditions. Signl vritions due to stte chnges re considered s the very slow vritions of the direct signl (nd consequently of the overll signl). Multipth contributions give rise to very fst vritions which cn be broken down into two clsses: those relted directly to the direct ry (tht is, those echoes generted by the direct signl's illumintion of nerby sctterers) nd those due to speculr rys, if they exist. A few strong, fr echo events (possible speculr rys) hve been registered in the experimentl dt sets [Smith nd Brton, 992] nd [Jhn nd Lutz, 995]. Nrrow-Bnd prmeters re: Stte probbility mtrices: bsolute stte probbility mtrix [W] nd stte trnsition mtrix [P]. Loo distribution prmeters for the different sttes: S i (α i, φ i, MP). The Loo distribution is used to chrcterise both the direct signl (Log-Norml distribution: (α i, φ i ) nd multipth (MP) for ech stte, environment nd elevtion. Correltion length of direct signl vritions. In order to generte relistic signl series, long term vrition is introduced. Stte frme length or minimum stte length. Wide-Bnd prmeters re: Averge decy slope, S (db, µs) Totl multipth power, MP (db reltive to LOS) Distribution of times of rrivl of echoes. Negtive exponentil distribution is ssumed. Directions of rrivl of multipth echoes. An uniform zimuth ngle is ssumed. 7.2.4.2 Power Control: Theoreticl spects Power control is prticulrly importnt in multiple ccess stellite system where users propgtion loss cn vry over tens of decibels due to different distnces nd, to lesser extent, to shdowing effects.

7.2-25 Power control mechnisms include both open nd closed loop power control for the uplink nd reltively slow power control for the downlink. Open Loop Power Control The open loop power control djusts the initil ccess chnnel trnsmission power nd compenstes vritions in the pth loss. The utomtic gin control (AGC), gives rough estimte of the propgtion loss of the forwrd link power received by the mobile unit. The mobile sttion controls its trnsmit power ccording this estimte. Closed Loop Power Control Open loop cnnot compenste the independence of the Ryleigh fding in the uplink nd downlink nd therefore closed loop power control is implemented for both the forwrd nd the return link. This loop is driven in order to set the mesured SIR (over n dequte period of time) to trget vlue [Viterbi, 998]. Power control commnds re trnsmitted uncoded cusing high error rtio on the order of 5% [Öjnper nd Prsd, 998]. However, since the loop is of delt modultion type (i.e. power is djusted continuously up or down) this cn be tolerble. The dynmic rnge for the closed loop power control is on the order of 2 db. Downlink Slow Closed Loop Power Control The trget vlue is itself dptively modified by mens of slower outer control loop bsed on frme error rtio (FER) mesurements. Trnsmitted power in the downlink is periodiclly reduced, the mobile unit mesures the FER nd when predefined limit is exceeded, dditionl power is requested. The size of power control step defines the chnge introduced in the trnsmission power. A simple up/down djustment or severl power djustment levels re possible. Typicl step sizes re between.5 nd 2 db. It is worth noting tht power control djustment is reltive to the previous power setting, since bsolute power setting would require expensive circuitry [Fukusw nd Sto, 996]. 7.2.4.3 Power Control: Simultions Since no ssumptions re mde on the system fetures relted to power control (PC) of the stellite component of LEO 66, simultions were crried out for severl T pc (updte rte), nmely, T pc = T f, 5 T f, T f, 5 T f nd 2 T f. Simultions following Montecrlo pproch re ble to simulte PC control lgorithms t bit level nd mkes FER nd SIR mesurements fesible. However, the purpose here is to use only informtion generted with the chnnel model, mening tht the criterion to be used throughout the simultions is bsed on pth loss. Chnnel mplitude time series were generted with the previously mentioned 3-stte Mrkov model, properly normlised to yield received signl or Eb/N. Power correction steps of ±.75, ±. nd ±.9 db were used s in [De Gudenzi nd Ginnetti, 998]. The time resolution hs been set to ms, which is coincident with the frme length. Coherence times of the chnnel for mobile speeds of m/s, 5 m/s nd 3 m/s re 8 ms, 5 ms nd 2.5 ms respectively, mening tht time vribility for 5 m/s nd 3 m/s is undersmpled. However it is not expected tht PC commnds re ctivted t such high rte.

7.2-26 The principl objective of the simultions is to nlyse the performnce of power control, extrcting error chrcteristion for different updte rtes, environments, elevtions nd mobile speeds. This informtion is independent of the specific trget vlue for given system nd therefore normlised mplitude time series were used. Figures 7.2-9 nd 7.2-2 show two exmples illustrting the effect of signlling nd processing dely on PC performnce. Nmely, performnce for T pc of ms is presented for mobile speed of 5 m/s for two different elevtions nd environments. It cn be seen tht PC is not ble to trck the fst chnnel vritions not only for updte rtes longer thn T f but lso for medium-high mobile speeds. u t i l p m d e s i l m r o N - -2-3 SUBURBAN 65º-85º, MOBILE SPEED 5 m/s, SAMPLING TIME ms, Tpc = FRAME no PC PC -4 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6 Probbility Distribution Functions no PC PC LOGNORMAL (,.3).5.4.3.2. -8-6 -4-2 2 4 6 8 Normlised mplitude (db) Figure 7.2-9: Exmple of effect of processing, mesurement nd signlling dely on PC performnce (T pc = T f ) Moreover, the fitting of time series fter PC to lognorml probbility distribution functions is lso presented. As cn be found in the literture, PC error cn be modelled by lognorml probbility nd here n nlysis of such sttistics hs been done. Figures 7.2-2 nd 7.2-22 show the trends found in urbn nd suburbn environments of the lognorml prmeters (men nd stndrd devition) s function of mobile speed, elevtion nd updte rte. From Figure 7.2-2 it cn be observed tht the trend of stndrd error is different from tht obtined in the IMT 2 test cse. For higher elevtions the error increment does not decrese since both men nd stndrd devition do not decrese but even increse for the highest elevtions. The reson is the significnt difference of strongest component men vlues hndled by the Mrkov model for ech of the 3 sttes, which were obtined in different cmpigns for L nd S bnd. These vlues re shown in Tble 7.2-5. The men vrition shown in Figure 7.2-22 does not exhibit notble dependency with T pc but with elevtion nd mobile speed.

7.2-27 u t i l p m d e s i l m r o N - -2-3 URBAN 5º-45º, MOBILE SPEED 5 m/s, SAMPLING TIME ms, Tpc = FRAME no PC PC -4 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6 Probbility Distribution Functions no PC PC LOGNORMAL (,.8).5.4.3.2. -8-6 -4-2 2 4 6 8 Normlised mplitude (db) Figure 7.2-2: Exmple 2 of effect of processing, mesurement nd signlling dely on PC performnce (T pc = T f ) 5º - 45º 45º - 65º 65º - 85º Stte Stte 2 Stte 3 Stte Stte 2 Stte 3 Stte Stte 2 Stte 3 S Sub. -. -3.7-5. -.3-2. -3.8 -.4-2.5-4.2 Bnd Urb. -.3-8. -24. -.35-6.3-5.2 -.25-6.6 - L Sub.. -6.3-9. -.5-6.5-4. -.2-6. -.5 Bnd Urb..6-5.3-3..2-4.2-6.5 -.5-4.3-6.6 Tble 7.2-5: Men vlue of the strongest component for ech stte, environment nd elevtion In conclusion, it is very difficult for PC lgorithm to trck fst fding t high mobile speeds due to mesurement dely, signlling dely, etc. A possible mitigtion of this is the use of coding nd interleving (time diversity) which re more effective t fst mobile speeds but there is n dverse region round 3 m/s where neither power control nor the coding nd interleving re effective. Then, requirements for power control prmeters should be set ccording to these mobile speeds. It would be optimum to hve vrible power control rte ccording to the mobile speed.

7.2-28 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 5º - 45º 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 45º - 65º 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 65º - 85º 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 5º - 45º 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 45º - 65º 5 5 2.5.5 2 2.5 3 3.5 Tpc (in number of frmes, frme = ms) L o g N o r m l s t n d r d d e v i t i o n Elevtions : 65º - 85º Figure 7.2-2: Effects of elevtion, mobile speed nd PC updte time on PC error mesured in terms of the lognorml stndrd devition vribility. Suburbn (top) nd urbn (down), (o) m/s, (*) 5 m/s, (+) 3 m/s

7.2-29.2 Elevtions : 5º - 45º.2 Elevtions : 45º - 65º.2 Elevtions : 65º - 85º n e m -.2 n e m -.2 n e m -.2 l m r o N g o L -.4 -.6 l m r o N g o L -.4 -.6 l m r o N g o L -.4 -.6 -.8 -.8 -.8 - - - 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms).2 Elevtions : 5º - 45º.2 Elevtions : 45º - 65º.2 Elevtions : 65º - 85º n e m -.2 n e m -.2 n e m -.2 l m r o N g o L -.4 -.6 l m r o N g o L -.4 -.6 l m r o N g o L -.4 -.6 -.8 -.8 -.8 - - - 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) Figure 7.2-22: Effects of elevtion, mobile speed nd PC updte time on PC error mesured in terms of the lognorml men vribility. Suburbn (top) nd urbn (down), (o) m/s, (*) 5 m/s, (+) 3 m/s 7.2.4.4 Link nd system level interfce In this section new qusi-nlyticl link-level pproch is proposed. It is cler tht system performnce cn be quntified by mens of verge probbility lthough the preferred formt for link-level outputs is not cler from system-level point of view. For 3 rd genertion systems n ccurte wy of doing the interfce between link nd system level simultions includes the effects of rdio resource mngement lgorithms. In the study presented here, the only rdio resource mngement mechnism considered is PC. The impct on performnce of E b /N o dispersion produced by the PC error cn be esily ssessed on the bit error rte curves. The verge probbility of error cn be computed s follows:

7.2-3 BER = BER ()() ζ p ζ dζ 7.2-4 where ζ is the energy-to-noise rtio t given time instnt. It is rndom vrible depending on the instntneous vlues of number of system prmeters (for this study only PC is considered). P(ζ) is the probbility of error introduced for PC which follows lognorml distribution. This qusinlyticl pproch ws found [De Gudenzi nd Ginnetti, 998] to yield hrdly distinguishble results from error-counting techniques, nd is significntly less time-consuming. Prmeters of P(ζ) for different elevtions, mobile speeds nd elevtions were extrcted from simultions s ws described t length in the point devoted to PC simultion results. BER(ζ) is the considered BER for the system to be studied which for the LEO 66 test cse is convolutionl chnnel code of R = 3/4 nd K = 7. For BER clcultions n upper bound [Prokis, 995] nd [Döttling nd Sunders, 999] ws used. Figures 7.2-23 nd 7.2-24 show the impct of PC error on BER curves for T pc = 5 ms s function of environment, elevtion nd mobile speed. It is pprent tht mobile speed introduces high degrdtion on system performnce, which decreses with elevtion. Coded BER, (Convolutionl, R = 3/4, K = 7) ; SUBURBAN, Elevtions: 5º - 45º - Coded BER, (Convolutionl, R = 3/4, K = 7) ; SUBURBAN, Elevtions: 45º - 65º - BER -2 BER -2-3 2 3 4 5 6 7 8 9 Eb/N in db -3 2 3 4 5 6 7 8 9 Eb/N in db - -2-3 2 3 4 5 6 7 8 9 Figure 7.2-23: PC error impct (T pc = 5 T f ) on coded BER (suburbn) - (o) m/s, (*) 5 m/s, (+) 3 m/s

7.2-3 Coded BER, (Convolutionl, R = 3/4, K = 7) ; URBAN, Elevtions: 5º - 45º - Coded BER, (Convolutionl, R = 3/4, K = 7) ; URBAN, Elevtions: 45º - 65º - BER -2 BER -2-3 2 3 4 5 6 7 8 9 Eb/N in db -3 2 3 4 5 6 7 8 9 Eb/N in db - -2-3 2 3 4 5 6 7 8 9 Figure 7.2-24: PC error impct (T pc = 5 T f ) on coded BER (urbn). - (o) m/s, (*) 5 m/s, (+) 3 m/s 7.2.4.5 System Level Simultions In order to clculte sptil coverge nd temporl vilbility [Oestges, 999], time series obtined in the simultions described throughout the previous sections hve been used s inputs to the COST 255 link nd system level softwre described in Chpter 7.. The ultimte objective of this study is to obtin the system performnce nd vilbility for the scenrios nlysed so fr, including different ltitudes (Rome nd London), environments (urbn nd suburbn) nd mobile speeds. For doing tht, simultions crried out ssuming constnt elevtion ngles re verged with probbility density functions of elevtions. In this cse, constelltion consisting of 66 stellites in 6 orbitl plnes ws considered. Sptil coverge obtined for 99% of time is presented in Tble 7.2-6. It cn be observed tht sptil coverge for verge coded BER of -3 never lies bellow 95%. Mobile speed does not seem to introduce significnt loss of coverge. ROME LONDON A A2 B B2 A A2 B B2 m/s 99. 96.6 98.5 96. 99. 96. 97.5 94.5 5 m/s 98.5 96.5 98. 95. 99. 96. 96. 93. 3 m/s 98.5 96. 98. 95. 99. 96. 96. 93. Tble 7.2-6: Sptil Coverge for 99% of the time A: suburbn environment, verge coded BER > -3 ; A2: suburbn environment, verge coded BER > -6 B: urbn environment, verge coded BER > -3 ; B2: urbn environment, verge coded BER > -6

7.2-32 7.2.5 Acknowledgements The uthors would like to thnk Dr. A. Jhn nd H. Ernst, DLR, Germny for providing the orbit genertor nd the module to generte rodside obstcles described in Section 7.2.2, nd Prof. Dr. B. Wlke, the hed of the Institute of Communiction Networks t Achen University of Technology, where the work of Jörg Hbeth hs been crried out.

7.2-33 7.2.6 References [Agelet et l., 997] Agelet, F., Perez-Fontán, F., Formell, A., Fst Ry Trcing For Microcellulr nd Indoor Environments, IEEE Trns. on Mgnetics, vol. 33, no. 2, pp. 484-487, 997. [Agius et l., 999] Agius, A.A., Lech, S.M., Sunders, S.R., Mesurement of the Polristion Stte of Stellite to Mobile Signls in Scttering Environments, Proc. Int. Mobile Stellite Conf. IMSC'99, Ottw, Cnd, pp.34-38, 999. [Akturn nd Vogel, 997] Akturn, R. nd Vogel, W.J., Pth Diversity for LEO Stellite-PCS in the Urbn Environment, IEEE Trns. Vehicul. Technol., vol. 45, no. 7, pp. 7-6, 997. [Armbruster nd Lurin, 996] Armbruster, P., Lurin, M., The IRIDIUM Network for Globl Personl Communictions, Telecommunictions Review, Vol. 6, No. 6, 996, pp. 674-686, 996. [Brtuschk et l., 997] Brtuschk, U., Mehlhorn, K., Näher, S., A Robust nd Efficient Implementtion of Sweep Line Algorithm for the Stright Line Segment Intersection Problem, Proc. Workshop on Algorithm Engineering WAE, Venice, Itly, pp. 24-36., 997 [Bischl nd Werner, 997] Bischl, H. nd Werner, M., Chnnel Adptive Stellite Diversity for Non-Geosttionry Mobile Stellite Systems, Proc. Fifth Int. Mobile Stellite Conf. IMSC97, pp. 25-3, 997. [De Gudenzi nd Ginnetti, 998] De Gudenzi, R., Ginnetti, F., DS-CDMA Stellite Diversity Reception for Personl Stellite Communiction: Stellite-to-Mobile Link Performnce Anlysis, IEEE Trns. Vehicul. Technol., vol. 47, no. 2, pp.658-672, 998. [Döttling et l., 998] Döttling, M., Jhn, A., Kunisch, J., Buonomo, S., A Verstile Chnnel Simultor for Lnd Mobile Stellite Applictions, Proc. IEEE Vehicul. Technol. Conf. VTC'98, Ottw, Cnd, pp. 23-27, 998. [Döttling et l., 999] Döttling, M., Zwick, T., Wiesbeck, W., "Ry Trcing Techniques nd their Applictions in Lnd Mobile Stellite Propgtion," Proc. Europen Wireless '99, München, Germny, 999, pp. 7-, 999. [Döttling et l., 999b] Döttling, M., Zwick, T., Wiesbeck, W., Investigtion of Stellite Diversity nd Hndover Strtegies in Lnd Mobile Stellite Systems bsed on Ry Trcing Propgtion Model, Proc. Int. Mobile Stellite Conf. IMSC'99, Ottw, Cnd, pp. 28-33, 999. [Döttling et l., 999c] Döttling, M., Didsclou, D., Wiesbeck, W., Widebnd Chnnel Modeling nd Diversity Techniques for Stellite-UMTS, Proc. IEEE Vehicul. Technol. Conf. VTC'99 Fll, Amsterdm, The Netherlnds, 999, pp. 277-2774, 999.

7.2-34 [Döttling nd Sunders, 999] Döttling M., Sunders, S. Bit Error Rte Clcultion for Stellite Communictions Systems, COST 255 Rdiowve Propgtion Modelling for New StCom Services t Ku-Bnd nd bove, Toulouse, My 999. [Fontán, et l. 997] Fontán F.P., Pered J., Sedes M.J., Cstro M.A.V., Buonomo S., Bptist P., Complex envelope three-stte Mrkov chin simultor for the LMS chnnel, Interntionl Journl of Stellite Communictions, Jnury 997, pp -5, 997. [Fontán et l. 998] Fontán F.P., Cstro M.A.V, Buonomo S., Bptist P., Arbesser-Rstburg B., S-Bnd LMS propgtion chnnel behviour for different environments, degrees of shdowing nd elevtion ngles, IEEE Trnsctions Brodcsting, Vol. 44, No., Mrch 998, pp 4-76, 998. [Fukusw nd Sto, 996] Fukusw A., Sto T., Tkizw Y., Kto T., Kwbe M., Fisher R. E., Widebnd CDMA System for Personl Rdio Communictions, IEEE Communictions Mgzine, October 996, pp. 6-23, 996. [Hutcheson nd Lurin, 995] Hutcheson, J., Lurin, M., Network Flexibility of the IRIDIUM Globl Mobile Stellite System, Proceedings of the Interntionl Mobile Stellite Conference, Ottw, Cnd, 995, pp. 53-57, 995. [IMST, DLR, IHE, 998] IMST, DLR, IHE, Lnd Mobile Stellite Propgtion Model for Non-Urbn Ares, Finl Report, Europen Spce Agency (ESA) Contrct No. AO/-3/96/NL/NB, Document Reference 43/729/6/2, 998. [Jhn nd Lutz, 995] Jhn A. nd Lutz E., Propgtion Dt nd chnnel model for LMS systems, Finl report. ESA Purchse Order 4742. DLR, 995. [Motorol, 99] Motorol Stellite Communictions, Inc., Iridium System Appliction Before the Federl Communictions Commission, Wshington, USA, December 99. [Oestges et l. 998] Oestges, C., Vsseur, H., Vnhoencker, D., Impct of edge diffrction on the performnce of lnd mobile stellite systems in urbn res, Proc. Europen Microwve Conference EuMC'98, October 998, Amsterdm, The Netherlnds, pp. 357-36. [Oestges, 999] Oestges C., 999, System-level Simultion for Lnd Mobile Stellite Services, COST 255 Rdiowve Propgtion Modelling for New StCom Services t Ku-Bnd nd bove, Toulouse, My 999. [Öjnper nd Prsd, 998] Öjnper, T., Prsd, R., Widebnd CDMA for third Genertion Mobile Communictions, Artech House, 998. [Prokis, 995] Prokis, J. G., Digitl Communictions, McGrw-Hill, 3 rd edition, New York, 995.

7.2-35 [Smith nd Brton, 992] Smith, H., Brton S.K., Grdiner J.G., Sforz M., Chrcteristion of the Lnd Mobile-Stellite (LMS) Chnnel t L nd S bnds: nrrowbnd mesurements, ESA AOPs 4433/4473, Brdford, 992. [Viterbi, 998] Viterbi, A. J., CDMA principles of Spred Spectrum, Addison-Wesley Wireless Communictions Series, 998 [Wlke, 999] Wlke, B., Mobile Rdio Networks: Networking nd Protocols, John Wiley, 999.

7.3- CHAPTER 7.3 Stellite IMT-2 Mobile Link Editor: Simon Sunders Authors: Frncisco Cercs 2, Mrtin Döttling 3, Clude Oestges 4, Fernndo Pérez-Fontán 5, Simon Sunders, M. Ángeles Vázquez-Cstro 6 nd Werner Wiesbeck 7 Centre for Communiction Systems Reserch, University of Surrey, Guildford, Surrey, GU2 5XH, UK Tel.: +44-483-25986, Fx.: +44-483-25954, e-mil: S.Sunders@ee.surrey.c.uk 2 Instituto Superior Tecnico de Lisbo, Dep. Engenhri Electronic e de Comp, Avenid Roviso Pis, 96 Lisbo CODEX, Portugl, Tel: +35--84-8485, Fx: +35--84-764, e-mil: Frncisco.Cercs@lx.it.pt 3 Siemens AG, System Engineering / UMTS Development, ICM MP TI 6, Hidenupltz, D-8667 Muenchen, Germny Tel.: +49-89-722-5873, Fx: +49-89-722-2789, e-mil: Mrtin.Doettling@mch.siemens.de 4 Université ctholique de Louvin (UCL), Microwve Lbortory (EMIC), Plce du Levnt 3, B-348 Louvin-l-Neuve, Belgium Tel.: +32--4785, Fx.: +32--47875, e-mil: Oestges@emic.ucl.c.be 5 Universidd de Vigo, Escuel Superior de Ingenieros de Telecomunicciones, Deprtmento de Tecnologís de ls Comunicciones, Cmpus Universitrio, 362 - Vigo, Spin Tel.: +34-986-8237, Fx.: +34-986-822, e-mil: fpfontn@tsc.uc3m.es 6 Universidd Crlos III de Mdrid, Escuel Politécnic Superior, Deprtmento de Tecnologís de ls Comunicciones, Avd. de l Universidd, 3, 289-Legnés, Mdrid, Spin Tel.: +34-9-624973, Fx.: +34-9-624943, e-mil: mryn@tsc.uc3m.es 7 Universität Krlsruhe, Institut für Höchstfrequenztechnik und Elektronik, Kiserstr. 2, D-7628 Krlsruhe, Germny, Tel : +49 72 68 6257 Fx : +49 72 69865 e mil: Werner Wiesbeck@etec uni krlsruhe de

7.3-2 7.3 Stellite IMT-2 Mobile Link S-IMT 2 is widebnd CDMA system bsed on the specifictions in [Dhlmnn et l., 998] nd [Cire et l., 999]. This system ws proposed by the Europen Spce Agency s cndidte for S-IMT-2, providing berer dt rtes up to 64 kbit/s. 7.3. Orbitl Constelltion The originl ESA proposl did not specify ny prticulr stellite constelltion. For the purposes of the test cse, the DELIGO stellite constelltion [Meenn et l., 995] with 64 stellites in eight orbitl plnes t 626 km ltitude is used, since it provides high probbilities for multiple stellite visibility nd llows investigtion of stellite diversity. Two prticulr loctions were chosen for simultion, nmely London nd Rome, with elevtion sttistics s specified in Figure 7.3-. 4.5 4 London Rome 3.5 Probbility density function 3 2.5 2.5.5 2 3 4 5 6 7 8 9 Elevtion ngle, [deg.] Figure 7.3-: Elevtion Sttistics for the Deligo constelltion in London nd Rome 7.3.2 Link Prmeters (Common to Stellite nd Mobile Terminl) A sttic link mrgin of 6 db is ssumed. The stellite EIRP is ssumed to compenste idelly for free spce propgtion loss, i.e. the line-of-sight signl level corresponds to the trget E b /N (bit energy to spectrl noise density) of 8 db. The bseline simultions use n isotropic mobile terminl (MT) ntenn nd 2-bit closed-loop power control. To infer power control efficiency under different conditions, simultions without power control re lso shown. Additionl simultions re performed to investigte stellite hndover, stellite diversity nd polristion diversity. The min S-IMT 2 simultion prmeters re listed in Tble 7.3-.

7.3-3 prmeter constelltion symbol rte chip rte modultion S-IMT 2 simultions 64 stellites, 8 orbitl plnes, 54 inclintion, 626 km 256 kbit/s 4.96 Mcps QPSK system frequency response Rised cosine, roll-off fctor.22 chnnel coding Convolutionl, R = /3, K = 9 required E b /N 2. db for BER = -3 link mrgin 6 db signlling dely n st = 2 power control stellite hndover stellite diversity polristion diversity 2-bit closed-loop power control ± db dynmic rnge ms updte intervl trget E b /N of 8 db 6 db hysteresis 5 ms processing time ms updte intervl 2-brnch mximum rtio combining 6 db hysteresis 5 ms processing time ms updte intervl Mximum rtio combining Tble 7.3-: Min S-IMT-2 Simultion Prmeters 7.3.3 Simultions in Non-Urbn Ares bsed on 3D Ry Trcing Chnnel Model The ry trcing model [Döttling et l., 999] nd post-processing steps [Döttling et l., 999b] used in this chpter re outlined in Chpter 7.2. The reder is lso referred to this chpter for description of the opertionl scenrio nd the generl simultion prmeters. 7.3.3. Simultions using Stellite Elevtion Sttistics The results presented in this section use the chnnel model outlined in Chpter 7.2 in combintion with the COST 255 system- nd link-level softwre (see Chpter 7.). The mobile pths re identicl to those described in Chpter 7.2. For CDMA system the efficiency of power control (PC) is importnt for link qulity nd system cpcity. The results in this section will compre the performnce of power control using the following criteri: the probbility tht the trget bit error rte (BER) is exceeded, the percentge of smples tht re kept within ± db of the trget E b /N, the percentge of smples tht exceed the trget E b /N by more thn 3 db, the percentge of time nd re tht the trget BER is chieved using the stellite elevtion sttistics for different ltitudes.

7.3-4 The chnnel simultions re performed t 7 different elevtion ngles from 5 to 85 using twelve different zimuth ngles t ech elevtion ngle to provide sufficient verging over zimuth. Figures 7.3-2 nd 7.3-3 show the power-controlled time series for 25, 45 nd 65 elevtion for pth (v = m/s) nd pth 3 (v = 3 m/s), respectively. A thousnd smples correspond to ech zimuth ngle. It is obvious from these plots tht the power control efficiency decreses with incresing user speed. The PC is no longer ble to trck the fst chnnel vritions nd leds to strong fluctutions round the trget vlue. The signl vritions re more pronounced for low elevtion ngles, where multipth effects rech their mximum. E b /N in db 2 5 5-5 - -5 25 45 65-2 2 3 4 5 6 7 8 9 2 smple no. Figure 7.3-2: E b /N smples for 25, 45 nd 65 elevtion, pth, power control E b /N in db 2 5 5-5 - -5-2 2 3 4 5 6 7 8 9 2 smple no. 25 45 65 Figure 7.3-3: E b /N smples for 25, 45 nd 65 elevtion, pth 3, power control The cumultive distribution functions (CDF) of E b /N for pth, without nd with power control, re depicted in Figure 7.3-4. The corresponding legends re given in Figure 7.3-5. Tble 7.3-2 shows the corresponding BER sttistics. For pth, the power control increses the number of smples with BER < -3 by 5.6% t 5 elevtion, by 2.7% t 25 elevtion, by.4% t 45 elevtion nd by 6.2% t 65 elevtion.

7.3-5 probbility (E b /N < bsciss).8.6.4.2-2 -5 - -5 5 5 2 E b /N in db probbility (E b /N < bsciss) ) b).8.6.4.2-2 -5 - -5 5 5 2 E b /N in db Figure 7.3-4: CDF of E b /N )for, pth, without () nd with power control (b) 5 5 2 25 3 35 4 45 5 55 6 65 7 75 8 85 % 96.97% 89.63% 76.57% 58.29% 44.7% 3.54% 2.23% 3.44% 7.73% 3.6% Figure 7.3-5: Legends for subsequent plots: left: elevtion ngle in CDFs of E b /N nd CCDFs of BER, right: time percentges for system-level simultion results t the ltitude of Rome Elevtion ngle [degree] Pth Pth PC Pth 2 Pth 2 PC Pth 3 Pth 3 PC 5 63.7% 48.% 67.9% 52.% 57.6% 4.5% 25 58.8% 37.% 43.9% 26.8% 46.5% 29.2% 45 29.9% 8.5% 33.4% 6.4% 35.6% 2.6% 65 8.7% 2.5% 6.4% 2.5% 9.% 7.9% Tble7.3-2: Probbility of BER> -3 for different simultion pths nd elevtion ngles The chnnel simultions show mximum short term fding depths round 25º elevtion, which led to minimum percentge of smples within n intervl of ± db round the trget E b /N t this ngle (see Figure 7.3-6). The cpbility of the power control to keep the E b /N within this intervl decreses with incresing mobile speed. Figure 7.3-6b) shows tht the power control leds to n incresed number of smples, which exceed the trget E b /N by 3 db nd thus introduce unnecessry power consumption nd interference. This effect ggrvtes with incresing user speed. At MT speed of 3 m/s, PC leds to n increse of smples with E b /N > db greter thn 3% for the elevtion ngles between nd 35. In generl, the low elevtion ngles re most ffected by this effect.

7.3-6 smples within ±db of trget Eb/N in % 9 8 7 6 5 4 3 2 p p, PC p2 p2, PC p3 p3, PC 2 3 4 5 6 7 8 9 elevtion in increse of smples with E b /N > db in % ) b) 2 5 5 p p2 p3 2 3 4 5 6 7 8 9 elevtion in Figure 7.3-6: Efficiency of power control for pth (p), pth 2 (p2) nd pth 3 (p3): percentge of smples within ±db of trget E b /N () increse of smples tht exceed the trget E b /N by more thn 3dB (b) The system-level results of pth for the ltitude of Rome re given in Figure 7.3-7 without nd with power control. For the ltitude of Rome, power control increses the covered re by.2% to.% for wide rnge of time percentges (Tble 7.3-3). Due to the different stellite constelltion nd elevtion sttistics the dependence of vilbility on ltitudes is slightly greter thn in the LEO 66 test cse. The covered re in London is round.8% greter thn in Rome (t BER = -3 nd for % of the time). verge coded BER - -2-3 -4-5 -6 9 92 94 96 98 percentge of re verge coded BER - -2-3 -4-5 -6 ) b) 9 92 94 96 98 percentge of re Figure7.3-7: Averge coded BER for Rome, pth, without () nd with (b) power control Time percentge Pth Pth PC Pth 2 Pth 2 PC Pth 3 Pth 3 PC. % 97.% 97.4% 95.9% 96.9% 95.6% 95.8% 58.3 % 97.7% 98.3% 97.% 98.% 96.3% 96.9% 3.4 % 99.% 99.5% 99.5% 99.8% 98.% 98.8% 3.6 % 99.9% 99.9%.%.% 99.4 99.6% Tble 7.3-3: Percentge of re tht the trget BER is chieved for the ltitude of Rome

7.3-7 Figure 7.3-8 shows the CDF of E b /N nd system-level results for pth 3. A comprison of the results for different user pths highlights the importnt influence of the locl terrin effects on system performnce (cf. Figures 7.3-4) nd 7.3-7b). Tble 7.3-3 summrises the min results for the ltitude of Rome nd BER = -3. probbility (E b /N < bsciss).8.6.4.2-2 -5 - -5 5 5 2 E b /N in db verge coded BER - -2-3 -4-5 -6 ) b) 9 92 94 96 98 percentge of re Figure 7.3-8: CDF of E b /N (left) nd verge coded BER for Rome (right), pth 3, power control 7.3.3.2 Simultions using n Orbit Genertor The simultions in this section use time series of stellite positions provided by n orbit genertor insted of elevtion sttistics. Figure 7.3-9 shows the elevtion ngles for the 6 s simultion time. The S-IMT 2 stellite constelltion provides high probbility of multiple stellite visibility t high elevtion ngles. Thus the investigtion of stellite diversity is especilly interesting. The bseline scenrio uses the power control scheme outlined bove. elevtion in 9 8 7 6 5 4 3 2 5 5 2 25 3 35 4 45 5 55 6 time in s Figure 7.3-9: Time series of stellite elevtions for S-IMT 2 simultions The simultions distinguish between up- nd downlink, since in CDMA systems multiple ccess interference (MAI) is ssumed to be the limiting fctor on E b /N. For the uplink resonble ssumptions for the noise due to MAI hve been mde, since the chnnel sttes of the interfering users re not known. The number of equivlent chnnels (see lso Chpter 7.) for the uplink hs been chosen to yield the trget E b /N for line-of-sight conditions without power control. Thus the MAI noise power is constnt: N tot () t P C const = 73- up mi i LOS eq =

7.3-8 For the downlink, however, the ry trcer provides the chnnel impulse responses for ll visible stellites simultneously. The number of equivlent chnnels per stellite is djusted in wy tht, for ech stellite, line-of-sight conditions would correspond to the trget E b /N. No MAI from the serving stellite i is considered, since orthogonl nd synchronous CDMA is ssumed. This leds to time-vrint MAI noise power: N d mi S() t () () S, i t = PS t Ceq. 7.3-2 Thus the correltion between downlink MAI nd stellite diversity gin cn be modelled. s= s i Uplink Figure 7.3- depicts comprison of the chnnel time series using power control (PC), PC with 2- brnch stellite diversity (SD) nd mximum rtio combining, s well s combintion of PC, SD nd polristion diversity (PD) t the mobile terminl. The simultions show considerbly stellite diversity gin. In combintion with polristion diversity dditionl improvements re seen especilly between t = s nd t = 5 s. The CDF of E b /N in Figure 7.3- shows stellite diversity gin of 8.8 db t E b /N = 2 db. Even t high hndover rte of.3 Hz, stellite hndover would provide less gin (6.5 db). Polristion diversity provides 5.7 db gin nd combintion of both diversity schemes results in. db gin (see Figure 7.3-b). The men number of required chnnels for stellite diversity is.44 for 6 db hysteresis to dd nd drop stellites to/from the diversity connection. The use of stellite diversity decreses the probbility to exceed the trget BER by 3.5% (from 43.2% to.7%). The combintion of stellite nd polristion diversity would llow reching the trget vlue in 96.9% (Tble 7.3-4). E b /N in db 2 5 5-5 PC - PC, SD -5 PC, SD, PD -2 2 3 4 5 6 time in s Figure 7.3-: E b /N time series compring power control (PC), PC in combintion with stellite diversity (SD) nd the combintion of PC, SD with polristion diversity (PD), uplink

7.3-9 ) b) probbility density.5.4.3.2. 5 5 2 E b /N in db PC PC, HO PC, PD PC, SD PC, SD, PD probbility (E b /N < bsciss).8.6.4.2-2 -5 - -5 5 5 2 E b /N in db PC PC, HO PC, PD PC, SD PC, SD, PD Figure7.3-: PDF nd CDF of E b /N for different S-IMT 2 system simultions, uplink Downlink: Figure 7.3-2 show the sme investigtion for the downlink. Due to the time-vrint MAI noise power, the time series of E b /N differ from the uplink. This is clerly visible in Figure 7.3-2 where the PDFs show flttened shpe with respect to Figure 7.3-. The gin from stellite hndover nd polristion diversity is round 5.5 db t E b /N = 2 db. Stellite diversity offers 6.8 db nd the combintion of stellite nd polristion diversity yields gin of. db. Due to the considertion of time-vrint MAI, the stellite diversity gin is 2 db lower thn for the uplink simultions. This highlights the necessity of relistic MAI modelling for CDMA stellite systems. The use of stellite diversity decreses the probbility to exceed the trget BER by 36.2% (from 49.5% to 3.3%). The combintion of stellite nd polristion diversity leds to BER of less thn -3 in 95.6% (Tble 7.3-4). probbility density.5.4.3.2. 5 5 2 E b /N in db PC PC, HO PC, PD PC, SD PC, SD, PD probbility (E b /N < bsciss).8.6.4.2-2 -5 - -5 5 5 2 E b /N in db PC PC, HO PC, PD PC, SD PC, SD, PD ) b) Figure 7.3-2: PDF nd CDF of E b /N for different S-IMT 2 system simultions, downlink

7.3- Simultion gin t E b /N = 2dB uplink gin t E b /N = 8dB probbility BER > -3 gin t E b /N = 2dB Downlink Gin t E b /N = 8dB probbility BER > -3 PC 43.2% 49.5% PC, HO 6.5dB.8dB 5.6% 5.3dB.3dB 7.% PC, PD 5.7dB 2.3dB 25.6% 5.7dB 2.5dB 3.3% PC, SD 8.8dB 3.6dB.7% 6.8dB 2.dB 3.3% PC, SD, PD.dB 4.6dB 3.%.dB 4.8dB 4.4% Tble7.3-4: Comprison of different strtegies to increse system performnce 7.3.3.3 Conclusions of the simultion bsed on 3D Ry Trcing Chnnel Model A comprison of the results obtined in this section (cf. Tbles 7.3-2 nd 7.3-3) shows tht the CDMA system using 6 db sttic link mrgin, power control nd R = /3, K = 9 convolutionl code performs similrly to the TDMA/FDMA system using 6 db link mrgin nd R = 3/4, K = 7 convolutionl code (see Chpter 7.2). However, the S-IMT2 system performnce decreses with incresing MT speed due to the degrded performnce of the S-IMT 2 power control scheme. Tble 7.3-5 compres the covered re for % of time. Simultion Pth Pth 2 Pth 3 System LEO 66 S-IMT 2 LEO 66 S-IMT 2 LEO 66 S-IMT 2 Percentge of re 97.3% 97.4% 97.8% 96.9% 97.% 95.8% Tble 7.3-5: Comprison of system-level results of the LEO 66 (see chpter 7.2) nd S-IMT 2 test cse for % of time The impct of elevtion ngle nd MT speed on power control efficiency hs been investigted. The results show tht the lrgest short term fding depths re experienced t medium low elevtion ngles, cusing considerble problems for the power control to trck the signl vritions. These problems become worse with incresing MT velocity for ll elevtion ngles. For high speed pplictions, the delyed power control rection cuses unnecessry high signl mplitudes (in cses where the chnnel hs lredy re-entered line-of-sight conditions while the power control still remins ctive), which genertes unwnted interference power. The simultions bsed on n orbit genertor llow the investigtion of the performnce of different system designs for S-IMT 2. For both, uplink nd downlink, stellite diversity keeps the brech of the trget BER below 4%. The stellite diversity gin (with respect to the bseline scenrio using power control) is in the rnge of 7 db to 9 db t the trget E b /N vlue. Other simultions show lower gin [Döttling et l., 999c]. Thus it is nticipted tht the stellite diversity gin is sensitive to the opertionl scenrio, the stellite constelltion nd the probbility or E b /N vlue t which it is clculted. For exmple, t E b /N = 8 db the stellite diversity gin is only between 2 db nd 4 db. Polristion diversity provides considerble gin (5 db to 6 db t E b /N = 2 db), which is in good greement with mesurements [Agius et l., 999]. The combintion of stellite nd polristion diversity t the MT yields round db diversity gin nd coverge for more thn 95% of the time. A comprehensive comprison of the different strtegies to increse system performnce is given in Tbl 734

7.3-7.3.4 Simultions in built-up res bsed on physicl-sttisticl ry-trcing In this prgrph the physicl-sttisticl pproch bsed on ry-trcing is used to crry out simultions (see lso Chpter 7.2.3). Typicl urbn nd suburbn res re defined on the bsis of lognorml distribution (medin µ, stndrd devition σ) for the building height, nd constnt vlue w for the street width (see Tble 7.2.3.-). A more detiled description of the model cn be found in Chpter 7.2. A trget E b /N vlue of 8 db hs been selected (see Tble 7.3-). However, since S-IMT2 is widebnd CDMA system, simultions hve to tke into ccount the multiple ccess interference (MAI) s well s the power control tht is pplied to the received signl. The intersymbol interference is neglected becuse its effect seems to be insignificnt compred to MAI. The methods described in Chpter 7.2 re pplied to integrte MAI. On the other hnd, n open-loop power control is implemented, with mximl correction of db nd n updte time of ms (see Tble 7.3-). 7.3.4. Simultion results for pedestrin speed The distribution of coded BER s function of the sptil coverge re shown in Figure 7.3-3 in both urbn nd suburbn res nd for the two ltitudes (London nd Rome). A pedestrin speed of ms - is first ssumed. The vrious curves on the figure re ssocited with the following percentges of time (from left to right) :.,., 96.79, 89.35, 78.52, 68.25, 57.83, 46.4, 3.95, 6.5 nd 7.64 % of time while considering the ltitude of London;., 96.97, 89.63, 75.57, 58.29, 44.7, 3.54, 2.23, 3.44, 7.72 nd 3.6 % of time t the ltitude of Rome. For instnce, in n urbn environment t the ltitude of Rome, the men coded BER is less thn -3 during : 96.97 % of time on 96.3% 75.57 % of time on 96.53 % 44.7 % of time on 97.24 % of the considered re. 2.23 % of time on 97.35 % 7.72 % of time on 99.37 %

7.3-2 Suburbn scenrio Urbn scenrio 2 2 Averge coded BER 3 4 5 Averge coded BER 3 4 5 6 6 7 7 8 98 98.5 99 99.5 Percentge of re 8 85 9 95 Percentge of re ) b) Figure 7.3-3: Percentge of re over which the coded BER is less thn ordinte, in suburbn () nd urbn (b) res for severl percentges of time Although the fde mrgin is much lower for this test cse thn for the LEO 66 cse (see Chpter 7.2), the results re ctully better. This is due to both the power control, which in the cse of pedestrin speed ppers to be quite efficient, nd the stellite constelltion. The Deligo constelltion provides higher elevtion ngles thn Iridium. Figure 7.3-4 llows us to compre the performnce t both ltitudes (upper grph) s well s for both environments (lower grph). Difference of vilbility, [%] 3 25 2 5 5 suburbn urbn 96 96.5 97 97.5 98 98.5 99 99.5 Percentge of re over which coded BER < e 3 Coverge increse, [%] 4 3 2 London ltitude Rome ltitude 2 3 4 5 6 7 8 9 Percentge of time Figure 7.3-4: Comprison between vilbility for both ltitudes s function of coverge (upper grph) nd coverge for both res s function of vilbility (lower grph)

7.3-3 It cn be seen tht there is gin of time vilbility between the two simulted ltitudes of bout 25 %. This is much greter thn the vlue obtined for the LEO 66 test cse (see Chpter 7.2). In the lower grph, the coverge is incresed by only 3% when compring suburbn to urbn res. Finlly, Figure 7.3-5 highlights the effect of power control on the performnce. For ll simulted scenrios, the gin in performnce is plotted ginst the vilbility when compring the coverge between those results tht tke power control into ccount nd those tht do not. In suburbn res, the gin of % is quite constnt. Menwhile, the gin in urbn res decreses for high percentges of time, which could be explined by the deeper fdes encountered t low elevtion ngles in urbn res..6.4 suburbn London ltitude suburbn Rome ltitude urbn London ltitude urbn Rome ltitude.2 Difference of coverge, [%].8.6.4.2 2 3 4 5 6 7 8 9 Percentge of time Figure 7.3-5: Effect of power control in ll simulted cses : performnce gin (coverge increse) 7.3.4.2 Simultion results for high-speed scenrios A high-speed scenrio (5 ms - ) is lso investigted, ll other prmeters remining the sme s before. Figure 7.3-6 depicts the degrdtion of re performnce s function of the percentge of time. It is ctully quite limited (.5 to.5%), despite significnt signlling delys in the power control process. It hs however been noticed tht these results re very sensitive to the power control prmeters (updte rte, correction step). Considering mobile speed of 3 ms -, it cn be seen in Figure 7.3-7 tht the degrdtion is gin quite smll ( to.8 %).

7.3-4.8.6 suburbn London ltitude suburbn Rome ltitude urbn London ltitude urbn Rome ltitude.4 Difference of coverge, [%].2.8.6.4.2 2 3 4 5 6 7 8 9 Percentge of time Figure 7.3-6: Degrdtion of performnce for high-speed (5 ms - ) scenrio reltive to pedestrin cse.8.6 suburbn London ltitude suburbn Rome ltitude urbn London ltitude urbn Rome ltitude.4 Difference of coverge, [%].2.8.6.4.2 2 3 4 5 6 7 8 9 Percentge of time Figure 7.3-7: Degrdtion of performnce for high-speed (3 ms - ) scenrio reltive to pedestrin cse 7.3.5 Simultions in Urbn nd Suburbn Ares bsed on Widebnd Mrkov Chnnel (WMC) propgtion model The min focus of this section lies on detiled study of some key fetures relted to the link-level performnce evlution of Stellite test-environment bsed on Widebnd Direct-Sequence Code Division Multiple Access (W-DS-CDMA) system. The key chrcteristics of the performnce evlution to be mde re those which cn be evluted directly from time series of chnnel mplitude generted with chnnel model. The generl discussion on pproch bsed on the Widebnd Mrkov Chnnel propgtion model nd Power Control is given in Chpter 7.2.

7.3-5 Throughout the simultions, multiple ccess interference (MAI) is considered to be constnt (note tht only the uplink is under study) with the number of equivlent chnnels being chosen to yield the trget E b /N o for line of sight conditions (without power control). By using such n ssumption nd neglecting therml noise the expression for the number of equivlent chnnels is given s: E CG b p Eb = n = F Gp + 7.3-3 N I intr + I inter + N N where F is the frction of intrcell interference cused by users operting in the sme cell, to the totl interference (I intr /(I intr +I inter ), C is the received signl strength nd G p is the processing gin or spreding fctor. When simulting dynmic spects, this cpcity would be rndom vrible tht depends on fctors such s multi-user detection (MUD), power control, voice ctivity, etc. The min S-IMT 2 technicl specifictions relevnt to this simultions re listed in Tble 7.3-6. Prmeter Elevtion (deg) Power control RAKE receiver (Multipth nd Stellite diversity) S-IMT 2 simultions 5-45 45-65 65-85 3-bit closed-loop power control updte times (T pc ): from ms (T pc = T f ) to 2 ms (T pc = 2 T f ) 2 db dynmic rnge trget E b /N of 8 db Equl Rtio Combining Tble 7.3-6: Min S-IMT-2 Simultion Prmeters 7.3.5. Power Control: Simultions Since no ssumptions re mde on the system fetures relted to power control (PC) of the stellite component of S-IMT 2, simultions were crried out for severl T pc (updte rte), nmely, T pc = T f, 5 T f, T f, 5 T f nd 2 T f. Simultions following Monte-crlo pproch re ble to simulte PC control lgorithms t bit level so tht FER nd SIR mesurements cn be mde. However, the purpose here is to use only informtion generted with the chnnel model, mening tht the criterion to be used throughout the simultions is bsed on pth loss. Chnnel mplitude time series were generted with the bove mentioned 3-stte Mrkov model, properly normlised to yield received signl or E b /(N o +I). Power correction steps of ±.75, ±. nd ±.9 db were used s proposed by [De Gudenzi nd Ginnetti, 998]. As pointed bove, MAI noise is considered to be constnt choosing the number of equivlent chnnels to yield the trget E b /N o for LOS conditions without power control. The time resolution hs been set to ms which is coincident with the frme length. Coherence times of the chnnel for mobile speeds of m/s, 5 m/s nd 3 m/s re 8 ms, 5 ms nd 2.5 ms

7.3-6 respectively mening tht time vribility for 5 m/s nd 3 m/s is undersmpled, however PC commnd is not expected to be ctivted t such high rte. The principl objective of the simultions is to nlyse the performnce of power control, extrcting error chrcteristion for different updte rtes, environments, elevtions nd mobile speeds. This informtion is independent of the specific trget vlue for given system nd therefore normlised mplitude time series were used. Figure 7.3-8 shows the effect of signlling nd processing dely on PC performnce for suburbn environment nd mobile speed of m/s. Performnces for T pc of ms, 5 ms nd 2 ms re presented. It cn be seen tht PC is no longer ble to trck the fst chnnel vritions for updte rtes longer thn frme, leding to very strong fluctutions round the trget vlue (time series re normlised for trget vlue of db). Figure 7.3-9 shows the effect of signlling nd processing dely on PC for the sme rnge of elevtions but for urbn environment nd mobile speed of 5 m/s. In this cse, even for T pc = T f, the PC error leds to very strong fluctutions round the trget vlue. Figures 7.3-2 nd 7.3-2 show more clerly the effect of the mobile speed hving set T pc to T f nd compring m/s nd 3 m/s. Figures 7.3-8, 7.3-9 nd 7.3-2 show lso the fitting of time series fter PC to lognorml probbility distribution functions. As cn be found in the literture, PC error cn be modelled by lognorml probbility distribution for terrestril CDMA systems such s IS-95 [Öjnper nd Prsd, 998] nd here n nlysis of such sttistics hs been done. Figures 7.3-22 nd 7.3-23 show the trends found in urbn nd suburbn environments of the lognorml prmeters (men nd stndrd devition) s function of mobile speed, elevtion nd updte rte. It is cler from Figure 7.3-22 tht stndrd devition of the error increses for low elevtions with T pc nd mobile speed. For higher elevtions the error increment is less dependent on Tpc nd mobile speeds. The men vrition shown in Figure 7.3-23 does not exhibit notble dependency on T pc but on elevtion, showing reduction with elevtion, except for suburbn environments t 65º-85º. A comprison of the lognorml curves fitted to PC error for urbn nd suburbn environments s function of elevtion is given in Figure 7.3-24. Even though the urbn environment shows worse chrcteristics, the difference with the suburbn environment is not significnt. In conclusion, it is very difficult for PC lgorithm to trck fst fding t high mobile speeds due to mesurement dely, signlling dely, etc. A possible mitigtion technique is the use of coding nd interleving (time diversity) which is more effective t high mobile speeds but there is n dverse region round 3 m/s in which neither power control nor coding nd interleving re effective. Then, requirements for power control prmeters should be set ccording to these mobile speeds. It would be optimum to hve vrible power control rte ccording to the mobile speed.

7.3-7 SUBURBAN 45º-65º, MOBILE SPEED m/s, SAMPLING TIME ms, Tpc = FRAME Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions no PC PC LOGNORMAL (,.4) -8-6 -4-2 2 4 6 8 Normlised mplitude (db) SUBURBAN 45º-65º, MOBILE SPEED m/s, SAMPLING TIME ms, Tpc = 5 FRAMES Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions no PC PC LOGNORMAL (-.2,.2) -8-6 -4-2 2 4 6 8 Normlised mplitude (db) SUBURBAN 45º-65º, MOBILE SPEED m/s, SAMPLING TIME ms, Tpc = 2 FRAMES Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. no PC PC LOGNORMAL (-.2,.4) -8-6 -4-2 2 4 6 8 Figure 7.3-8: Effect of processing, mesurement nd signlling dely on PC performnce Suburbn environment, m/s, from top down, T pc = T f, 5T f, 2T f.

7.3-8 URBAN 45º-65º, MOBILE SPEED 5 m/s, SAMPLING TIME ms, Tpc = FRAME Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions no PC PC LOGNORMAL (,.5) -8-6 -4-2 2 4 6 8 Normlised mplitude (db) URBAN 45º-65º, MOBILE SPEED 5 m/s, SAMPLING TIME ms, Tpc = 5 FRAMES Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions no PC PC LOGNORMAL (-.4,2.2) -8-6 -4-2 2 4 6 8 Normlised mplitude (db) URBAN 45º-65º, MOBILE SPEED 5 m/s, SAMPLING TIME ms, Tpc = 2 FRAMES Normlised mplitude (db) - -2-3 -4 no PC PC 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. no PC PC LOGNORMAL (-.4,3) -8-6 -4-2 2 4 6 8 Figure 7.3-9: Effect of processing, mesurement nd signlling dely on PC performnce. Urbn environment, 5 m/s, from top down, T pc = T f, 5T f,2t f.

7.3-9 d u t i l p m 5 SUBURBAN 5º-45º, MOBILE SPEED m/s, SAMPLING TIME ms, Tpc = FRAMES no PC PC d e s i l m r o N 5-5 - -5 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions -8-6 -4-2 2 4 6 8 Normlised mplitude (db) no PC PC LOGNORMAL (,.6) Figure 7.3-2: Suburbn environment, m/s. d u t i l p m 5 SUBURBAN 5º-45º, MOBILE SPEED 3 m/s, SAMPLING TIME ms, Tpc = FRAMES no PC PC d e s i l m r o N 5-5 - -5 5 5 2 25 3 35 4 45 5 time (s).9.8.7.6.5.4.3.2. Probbility Distribution Functions -8-6 -4-2 2 4 6 8 Normlised mplitude (db) no PC PC LOGNORMAL (,.9) Figure 7.3-2: Urbn environment, 5 m/s

7.3-2 3.5 Elevtions : 5º - 45º 3.5 Elevtions : 45º - 65º 3.5 Elevtions : 65º - 85º 3 3 3 2.5 2.5 2.5 LogNorml stndrd devition 2.5 LogNorml stndrd devition 2.5 LogNorml stndrd devition 2.5.5.5.5 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 3.5 Elevtions : 5º - 45º 3.5 Elevtions : 45º - 65º 3.5 3 3 3 2.5 2.5 2.5 LogNorml stndrd devition 2.5 LogNorml stndrd devition 2.5 2.5.5.5.5 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Figure 7.3-22: Effects of elevtion, mobile speed nd PC updte time on PC error mesured in terms of the lognorml stndrd devition vribility. Suburbn (upper) nd urbn (lower), (o) m/s, (*) 5 m/s, (+) 3 m/s

7.3-2 Elevtions : 5º - 45º Elevtions : 45º - 65º Elevtions : 65º - 85º -.2 -.2 -.2 LogNorml men -.4 -.6 LogNorml men -.4 -.6 LogNorml men -.4 -.6 -.8 -.8 -.8 - - - 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) Elevtions : 5º - 45º Elevtions : 45º - 65º -.2 -.2 -.2 LogNorml men -.4 -.6 LogNorml men -.4 -.6 -.4 -.6 -.8 -.8 -.8 - - - 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Tpc (in number of frmes, frme = ms) 5 5 2 Figure 7.3-23: Effects of elevtion, mobile speed nd PC updte time on PC error mesured in terms of the lognorml men vrition. Suburbn (upper) nd urbn (lower), (o) m/s, (*) 5 m/s, (+) 3 m/s

7.3-22.3 SUBURBAN, 3 m/s, Tpc = 2 frmes (2 ms).25.2 5º-45º, LN(-.8,3) 45º-65º, LN(-.5,.7) 65º-85º, LN(-,.8).5..5 - -8-6 -4-2 2 4 6 8 Normlised mplitude (db).3 URBAN, 3 m/s, Tpc = 2 frmes (2 ms).25.2 5º-45º, LN(-.3,3.7) 45º-65º, LN(-.2,2.5) 65º-85º, LN(-.,2.2).5..5 - -8-6 -4-2 2 4 6 8 Normlised mplitude (db) Figure 7.3-24: Effects of elevtion, nd environment on PC performnce. 7.3.5.2 Rke Receiver: Theoreticl spects A Rke receiver consists of correltors, ech receiving resolvble multipth signls (rriving more thn one chip prt from ech other). Theoreticlly, the Rke receiver hs receiver finger for ech multipth component. After despreding by correltors, the signls re weighted nd combined. Mximl rtio, equl gin nd selection diversity re the principl combining techniques. In the mximl rtio technique, the received signls re weighted ccording to their received SIR, while in equl gin combining, no weighting is performed. Selection diversity consists of selecting the signl with the highest SNR. It is importnt to tke into ccount smll nd lrge-scle chnges. Due to the mobile movement, the delys nd ttenution fctors ffecting the multipth will chnge (lrge-scle chnges) with the locl environment nd therefore Rke fingers hve to be rellocted. Smll-scle chnges produce chnges of less thn one chip. Time delys of ech multipth signl re trcked by code-trcking loop. A Rke receiver cn be used not only for multipth diversity, where multipth is resolved nd coherently dded, but lso for mcro-diversity or stellite diversity where different trnsmissions from severl sources re hndled by one receiver unit. 7.3.5.3 Rke Receiver: Simultions In this cse, simultions hve been crried out for two different scenrios: multipth diversity nd mcro or stellite diversity.

7.3-23 Multipth Diversity Even though LMS chnnel is not likely to produce multipth contributions, the /496 MHz (.2 µs), mesurement cmpigns crried out by ESA [Smith nd Brton, 992] nd [Jhn nd Lutz, 995] hve shown tht they exist t lest for round 2% of the time. Simulted Rke is not n idel Rke since unresolvble multipth ws tken into ccount. The simulted scheme is given in Figure 7.3-25. Equl gin combining ws used. PN code Tc Tc Tc PN code PN code PN code Wveform mtched filter Wveform mtched filter Wveform mtched filter Wveform mtched filter # #N f E Q U A L R A T I O C O M B I N I N G Figure 7.3-25: Simulted CDMA demodultor (non idel Rke receiver) for multipth diversity. Tble 7.3-7 shows the multipth diversity gins obtined for different urbn scenrios with Rke receiver of 2 fingers. From these results it is cler tht the multipth diversity gin is not relevnt for fde depths bove db. The reson is tht ll the contributions rriving t the receiver present correlted fding nd shdowing nd therefore the energy gthered from different fingers is correlted. Simultions with 3 nd 4 fingers gve rise to similr results. Fde Depth [db] m/s 5 m/s 3 m/s 5º-45º 45º-65º 65º-85º 5º-45º 45º-65º 65º-85º 5º-45º 45º-65º 65º-85º < - 8 3. 4 2. 2..5 > - 2.5 2.2.5. Tble 7.3-7: Multipth diversity gins for Urbn environment Figure 7.3-26 shows n exmple of cumultive distribution for n urbn environment nd mobile speed of 5 m/s, s function of elevtion.

7.3-24 Figure 7.3-26: Exmple of multipth diversity. Urbn environment, elevtion of 5º - 45º nd mobile speed of 3 m/s. Stellite Diversity Simultions crried out for stellite diversity ssume tht signls coming from different stellites re uncorrelted nd resolvble. In this wy, time series for different elevtions generted independently were tken s snpshots of generic constelltion. Figure 7.3-27 shows the Rke receiver scheme for this cse. Only 3 brnches with 2 fingers ech were considered. As for multipth diversity, equl gin combining technique ws used. CDMA DEMOD # CDMA DEMOD #2... Sort nd select Nf lrgest #... # N CDMA DEMOD #N s Figure 7.3-27: Simulted CDMA demodultors (non idel Rke receiver) for stellite diversity.

7.3-25 Figure 7.3-28 shows n exmple for urbn environment nd mobile speed of 3 m/s. Figure 7.3-28: Exmple of Stellite diversity. Urbn environment, mobile speed of 3 m/s Since the im of these simultions ws to find men vlues of stellite diversity gin independent of constelltion geometry nd trnsmitted power, only rough estimtions of the rnges of chievble vlues were obtined. Gins rnging from 4 to 7 db were found in n urbn re nd from 5 to db for suburbn re. 7.3.5.4 Link nd system level interfce In this section new semi-nlyticl link-level pproch is proposed. It is cler tht system performnce cn be quntified by mens of verge probbility lthough the preferred formt for link-level outputs is not cler from system-level point of view. For 3 rd genertion systems n ccurte wy of mking the interfce between link nd system level simultions must include the effects of rdio resource mngement lgorithms. In the study presented here, the only considered rdio resource mngement mechnism is PC. The impct on performnce of E b /N o dispersion produced by the PC error cn be esily ssessed from the bit error rte curves. Figure 7.3-29 shows the grphicl interprettion of this effect. As proposed by [De Gudenzi nd Ginnetti, 998] nd [Hämäläinen nd Slnin, 997] the verge probbility of error cn be computed s follows BER BER = ()() ζ p ζ dζ 7.3-4