Capacity improvement of the single mode air interface WCDMA FDD with relaying

Transcription

1 2004 Internatonal Workshop on Wreless Ad-Hoc Networks Capacty mprovement of the sngle mode ar nterface WCDMA FDD wth relayng H. Nourzadeh, S. Nourzadeh and R. Tafazoll Centre for Comnurcaton Systems Research (CCSRJ Unuersly of Surrey, Guldford, UK E-mal: h.nou surrey. ac. uk Abstract- A new way to model a CDMA-system appljng relalng s proposed n ths paper. Ths method makes t possble to compare drectly the performance ol relayng. The outage probablty, whch represents the abltr of the users to reach the base Staton, s chosen as crtera to compare the system wth and wjthout relajng, The model k based on the sngle mode ar nterface WCDMA FDD wth a two-hop relay. When relayng s appled, the smulaton results show that even by usng the sngle mode FDD the uplnk capacty s sgnfcantly mproved by 82%. Also a new relay node selecton strategy s propcsed and the results show how mportant t s to choose approprately the relay node. And fnally, dfferent scenaros of relayng are smulated to show when or when t s not better to apply rclajng. I. INTRODUCTION In a CDMA system, all users nterfere wth each others. Therefore the CDMA system s nterference Imted. The users, whch suffer from a strong shadow effect or because they are far away from the BS, need more power to reach the latter. In ths case, theses users may transmt at ther maxmum allowed power wthout satsfjmg ther Qualty of Servce. Therefore, ths would leave some mobles out of the system and also create too much nterference to the neghborng cells. One-way to counter ths problem and therefore to mprove the capacty would be to ncrease the numbers of BS. But ths soluton cannot be effcent as t ncreases sgnfcantly the network nfkastructure cost, Another way to mprove the capacty s a system applyng relayng [I]. Ths means usng other exstng termnals located between the orgnatng termnal and the BS for the purpose of retransmttng the orgnal packet. Wth ths process, the users at the boundares of the cells wll need less power to reach the relay staton than to reach the BS and therefore create less nterference to the neghborng cells, hence mprovng the capacty. In the last few years, there has been a great nterest n cellular networks applyng relayng. Frst, ODMA (Opportunty Drven Multple Access), whch s an nteilgent protocol that sts upon a rado sub-system that support relayng, was proposed. One node relays ts packets through others nodes by usng ODMA wth a separate unpared spectrum band and the node the closer to the BS sends the packet by usng the TD/CDMA-FDD mode [11. The Ad Hac GSM cellular system adds the relay capablty to a second generaton GSM network to enhance the system coverage [2]. The CAR system can efcently balance traffc toads between cells by usng fxed relay staton to relay trafc [3]. The system proposed n ths paper s based on a two-hop reiay cellular network. In ths work, the relay statons are the moble staton non-communcatng wth the BS. The postve pont of usng MS as reiay s wth a great number of users n the system, a MS has more chance to choose an approprate MS relay. Wth fxed relay staton, the users far away may have dffcultes to reach the fxed relay staton. If a MS reiays ts packet through another MS, t can consderably decrease the battery of the MS relay. One soluton to ths problem would be to employ cars or tax as relay staton, snce they have nfnte battery. In ths system, the ar nterface WCDMA FJID s used for the two hops. Wth the current technology ths s not apphable, because wreless termnals cannot transmt and receve n the same frequency band n the mode FDD. However some changes can be made at the hardware of the moble, lke usng twwduplexer, n order to apply t n the sngle mode FDD. Also t can be nterestng to see f the capacty s mproved even by usng a sngle mode, whch s the worse case, because each hop nterferes wth each other. Secton II descrbes the system model proposed. Secton III explans the dfferent schemes to model relayng. The results are shown n secton IV. And fnally, a concluson s gven n secton V. D. MULTI-HOP BASED CELLULAR NETWORKS A. System descrpton The Fgure 1 (a) and (b) depct the scenaro of the smulaton model. The Fgure l(a) shows the MS actve and non-actve n a sngle cell wthout relayng. The MS actve are sendng contnuously data nformaton to the BS. Wth relayng, dfferent M5 role are dentfed such as: l$ IEEE 265

2 - MO (Moble Orgnator/Outage), whch have a weak lnk wth the BS. - MI (Moble Intermedate), whch can reach the BS. MS non-actve, whch are moble not sendng data to the BS. Wth the same locaton for all the users, the MOs look for a MS non-actve to relay ther nformaton, see Fg. 1 (b). In ths model, one MS relay can relay only one MO. Wth ths scenaro, at each nstant the performance of the system wth and wthout relayng can drectly be compared for the same poston for all users. relay-bs need to be calculated. As the MIS and the MSs relay send ther sgnal to the BS they have the same nterferers, and ther SIR can be expressed as: where ' L,<jq and I$o represent the ntracell nterference at the BS respectvely fom the MI, MS relay and MO from the same cell. If the MO U sends ts sgnal to the MS relay r wthn the cell of the BS, ts SIR can be expressed as: (a) Wthout relayng (b) Wth relayng where 'r,m1 3 Ir,Ms,+ and represent the nuacell nterference at the MS relay r respectvely from the MI, MS relay and MO n the cell of the BS. B. System model Fg. 1. Smulaton model A MS s n outage f ts SIR (Sgnal-toInterference rato) s below a certan threshold. At any tme, the SIR of all MS actve can be calculated. If G denote the gan, P denote the transmtter power and the uplnk measured ntracell and ntercell nterference at the BS are denoted by and I;nleFcell respectvely, the average SIR of user U s: where I;ntruce,/ GVq7, wth N' represents the number I js of users that belong to - the BS ; and AA '. A Y G t ercell =zxgjkek, wth N" the number of users that m# k-1 belong to the BS n, and M the number of cell n the system. n, represents the background nose. If the receved SIR of a MS s below ts SIR threshold, ths MS s consdered n outage. Ths means that the sgnal of the MS s not strong enough to overcome the nterference receved by the BS. In order to compare the two systems, wth and wthout relayng, the SIR of the lnks MI-BS, MO-MS relay and MS III. MULTI-HOP RELAYING SCHEMES In ths secton, three dfferent ssues on a two-hop cellular network are dscussed. In the part A, mprovng the lnk between the termnah by applyng power control s presented. The MS relay selecton, based on the SIR and the dstance, s dscussed n the part E. Fnally n part C, possble methods to reduce the number of relayed MS are dscussed. A. Power contro! between MS and MS relay Untl nowadays, few have consdered n the lterature to adapt the transmt power between two MSs, communcatng wth each other. Here, t s consdered the case where MO s capable of adjustng ts transmt power accordng to ts receved SIR at the MS relay. A perfect dstrbuted power control has been modeled for the lnks between MS and BS and also between MS and MS relay. The power control objectve between MS and MS relay s smlar than tradtonal power control algorthms used n CDMA-based cellular systems. It conssts of balancng all the receved SIR around the SIR target value, [4]. Power control s performed n subsequent teratons and the transmtter power P:*+') of each user U s adjusted at teraton n+l accordng to p;n+u = r pl'"' Y? (4) where y:n) s the measured SIR of the lnk at teraton n and r s the target SIR for the lnk. 266

3 B. MS relay selecton strategy The choce of the MS relay s an mportant ssue n mult-hop cellular networks. In [SI the crtera to determne the routng s based on mnmzng nterference. In ths paper, a new algorthm based on the maxmum SIR overall and on the dstance between users s ntroduced. Wthn a certan range, each MO wll choose the MS non-actve Wth the maxmum SIR overall as a MS relay. below the SIR-reference. Wth ths process, the SIR of the MOs not relayed can be ncreased and theses MO can acheve the target when relayng s done for oniy the MOs wth a SIR below the SII-reference. C. 1. Relay MS n the overlap area The fgure 2 shows the CDF (Cumulatve Dstrbuton Functon) of the dstance between me MO and the BS. The curve shows that most of the MOs are far away from the BS. 70% of the MOs are dstant from above 1600m. The MO cannot counter the loss along the path and reach the BS even wth the maxmum power. Not only the sgnal receved at the BS s weak but also they create a lot of ntercell nterference by usng ther maxmum power. Fg. 2. CDF of dstance between MO and 3s (m) a) Unly MO By relayng only the MOs far away from the BS, not only the ntercell nterference wll decrease but also the SIR of MO not n the overlap area wll ncrease and some of them may not be n outage wthout the need of relayng (see Fg. 3). b) AIE MS actve There are some MS actve n the overlap area, whch are not n outage, but because of a hgh propagaton loss, they transmt wth a great power. Therefore, they create also a lot of nterference to the neghborng cells. These MS actve, even f they are not n outage, can be relayed n order to use less transmt power and therefore decreasng the ntercell nterference. A new scheme s proposed where all the MS actve (n outage and not n outage) n the overlap area are relayed. C.2. Reluy MS wth SIR < SIR-ref A fnal scheme analyzed n ths paper, proposes to relay only the MOs, whch have the lowest receved SIR. For ths, a new threshoid SIR-reference s defned that s lower than the SIR target. The dea s to relay only the MOs wth a SIR Fg. 3. Relayng only MO n the overlap area Iv. SIMULATION RESULTS A. Smulaton Envronment Cell plannng s an mportant ssue when relayng needs to be mplemented n cellular networks. The overlap area between neghborng cells should be kept to the mnmum so that the system experences mnmum nterference. Therefore a lnk budget for the vehcular envronment, as n 161, s performed n order to fnd the most accurate value for the radus of the cell. The system modeled ncludes 19 hexagonal cells of radus 2000m, wth BS at the center of the cells employng omn drectonal antenna. The MS follow the moblty and the propagaton model as proposed n [7] for the vehcular envronment and for the hard handover case. A new auto correlated shadowng model n 2 dmensons s modeled between the MS and the BS, as n [SI. There are no approprate path loss models that descrbe the channel between MS snce a moble termnal s usually postoned around head. However, the Lee model, whch takes account the recevers antenna heght, s chosen as model 191. The model s smulated at system level and the smulaton has been run for teratons. The smulaton resuit s based on outage probablty, whch represents the number of MS n outage over the total number of MS actve for each cell. At the end of the smulaton, the outage probablty calculated after each tme wth and wthout relayng for the seven central cells are averaged and compared. B. Results I) PC between MS-MS relay Fgure 4 shows the outage probablty wth and wthout relayng for dfferent number of actve users n each cell. The results show that the outage probablty decreases when relayng s appled. For nstance, at 10% of outage probablty, the number of actve users supported n the cell wthout relayng s 28 but wth relayng (No W) t s 32, whch shows an mprovement of 14%. In ths case, the MO 267

4 transmt to the MS relay wth the same fxed power, whch s the maxmum power allowed. Relayng wth power control between MS shows a sgnfcant mprovement of 32%. For 10% outage probablty, the number of users actve wth relayng and wth power control supported s 37. Wth power control between the MS, the MO wl transmt wth the mnmum power needed to reach ts MS relay. not be able to fnd a MS relay n that range and wll select a MS relay far away whch wll lead to hgh transmsson power. A good compromse has been found to be approxmately 1000m. All MO wll be able to fnd a MS relay n ths range, whch s half of the cell radus. For 10 % outage probablty, the number of actve users supported n a cell wth the relay selecton scheme based on the overall hghest SIR and the dstance between the MO and the MS relay s 5 1 (wth maxmum dstance=looom), whch leads to a sgnfcant mprovement of 82%. Fg. 4. Outage probablty vs. Number of users actve per cell wth power control between MSs 2) Interference study Smulatons have been performed to study the ntracell and ntercell nterference wth and wthout relayng. The nterference rato (ntercelllntrace11 nterference) calculated wthout relayng s approxmately 1.7, whch s n agreement wth [IO] for the hard handover case. The ntracell nterference s ncreased from db to db when relayng s appled. Wth relayng, the MS relay, whch relays the MO, can acheve the target and then send a stronger sgnal to the BS than the MO wthout relayng, whch ncrease the ntracell.nterference. On the other hand, snce smaller dstances need to be covered when relayng s appled, the ntercell nterference s decreased kom db to -123 db. 3) MS relay selecton schemes A new selecton scheme based on hghest SIR overall and the dstance between the MO and the MS relay has been tested. In ths algorthm, the MO fnds a MS relay wthn a certan range and selects the one wth the hghest SIR for the twe hop relay. Fgure 5 shows the outage probablty wthout relayng, wth relayng based on the hghest SIR overall wthout range and wth dfferent allowed ranges (maxmum dstance) of 500m, looom and 1500m. Ths fgure shows a sgnfcant mprovement when the relay selecton scheme s based on the maxmum SIR and dstance. When the maxmum dstance to fnd a MS relay s set to 1500m, the MO can easly fnd a MS relay n ths range but the dstance s stll too large n order for the system to operate effcently. Wth the maxmum dstance set to 500m, the entre MO may Fg, 5. Outage probablty vs. Number of users actve per cell wth dfferent range to fnd a MS relay 4) Relayng only MO n the overlap area Fgure 6 shows the outage probablty n terms of the number of actve users wthout relayng, wth relayng all MS actve n the overlap area, wth relayng only MO that are n the overlap area, wth relayng only MO wth SIRCSIR-reference and wth relayng all MO. The results show that the performance s better when relayng s performed for all the MO, and not only the MO n the overlap area. For nstance for 10% outage probablty, the numbers of actve users n a cell wth relayng only MO n the overlap area s 33, but wth relayng all MO t's 51. There are some MO close to the BS that acheve communcaton wthout relayng because the ntercell nterference s decreased but not all of them snce the ntracell nterference s ncreased wth relayng. The radus of the crcle that defnes the overlap area s an mportant factor. The bgger the crcle s, the less relayng wll be performed and the outage probablty wth relayng only MO n the overlap area wll be equal to the outage probablty wthout relayng. But the smaller s the crcle, the more relayng wll be performed and the outage probablty wth only MO relayng n the overlap area wll be equal to the outage probablty wth relayng. Even f the performance s better wth relayng for all MO, t may be nterestng to model these schemes n a system where the sgnalng overhead s ncluded n the smulaton. Snce the overhead wll ncrease wth relayng, the capacty of 268

5 a system may not mprove f the overhead s too hgh, as shown n [ Therefore, the system desgner can select only some MO to relay. 5) Relayng all MS actve n the overlap Fgure 6 shows that the outage probablty when relayng all actve MS s hgher than the outage probablty when relayng all the MOs. The performance s worse because some actve MS whch are not n outage may use a hgher overall transmt power wth relayng than a drect lnk wthout relayng. Also one of the MS actve-ms relay or MS relay-bs lnks can fal whch leads to ths actve MS to go nto outage whereas t was not n outage wth a drect lnk. 6) Relayng only MO nh SIR < SIR-ref The results n Fg. 6. show that the best performance s obtaned when all the MO are relayed. It s nterestng to observe that the outage probablty when relayng s appled for the MO wth SIR below the reference SIR s less than when relayng s appled to the MO n the overlap area. Some MO that are close to the BS but exhbt very low SIR cannot acheve the target. But relayng MO wth a small SIR can ncrease the SIR of the MOs not relayed wth a SIR close to the target. Ths means the Sgnal-to-Interference s a better crteron than dstance. Fg. 7. CDF of the MOs not relayed, V. CONCLUSION The results showed that relayng could sgnfcantly mprove the capacty even by usng the sngle mode FDD for the twohop relay. A new scheme to choose the most approprate relay has been proposed. Ths latter emphaszed the mportance of the relay node selecton strategy. Dfferent aspects of relayng ntegrated to cellular systems were dscussed and ther advantages demonstrated. It was shown that a combnaton of these proposed technques such as applyng power control between two MS, a routng seiecton based on both SIR and dstance and fnally relayng a11 the MOs can provde a capacty gan up to 828. ACKNOWLEDGMENT The authors gratefully acknowledge the support of Vodafone (UK) to make possble to do ths research. Fg. 6. Dfferent schemes to relay only some MO. (The MS relay selecton strategy s based on the hghest SIK and the dstance maxmum equal to I000m) Fgure 7 shows the SIR of the MOs not relayed Wthout relayng and wth relayng only the MO wth a SIRcSIR-ref. The SIR target s equal to -21 db whereas the SIR reference s set at -23 db. The MOs not relayed have a SIR between -23dB and the SIR target -21dB. But Wth relayng all MO wth a SIR below the SIR reference (-23 a), the MO not relayed ncrease ther SIR snce the nterference decreases. The curve wth relayng shows that 60% of the MO not relayed has a SIR above the target (-21dB). Ths shows that most of the MOs establsh a Ik wth the BS wthout the need of relayng. REFERENCES ETSLTR v3.0.0 UMTS 30.06, Support forrelayng and ODMA December G. Aggelou and R. Tafamoll, On the relayng capablty of next generaton GSM cellular networks IEEE Personal Communcatons. vol. 8, pp?047,2001 H. Wu, C. Qao, S. De, 0. Tonguz CAR; Integrated Cellular and Ad ha: Relayng system? IEEE journal on selected areas n communcatons, October I. Zander, Dstrbuted cochannel nterference conuol n cellular rado systems 1EEETrans. Veh. Technol.,vol. 41, pp. 5742,1992. T. Rouse, I. Band S. McLaughln Capacty and power nvesdgaton of ODMA n UTRA TDD IEE 3G 2001, pp I. Laho, A. Wacker, and T. Novosad, Rudo Networkpkmnng Optmsrlon for VMTX Wley, p82. ETSl (UMTS); selecton pmcedures for (he cfuce of rado tmnrmsson [echnubgy ofthe UMTS (UMTS v32.0) H. Nourzadeh, X. Yang and R. Tafazoll Generaton of twodmensonal shadowng for dymmc system-level smulators submtted for IEEE Communcatons Mm. S.C Yang CDMA RF Engneerng, Artech House, Boston London, A. Vterb, A. Vlerb, K. Glhousen, and E. Zehav, Soft Handoff extends CDMA cell coverage and ncreases reverse lnk crtpacty, IEEE Joumal on Se. Areas n Comm., vol. 12, pp , October B. Holm, T. Frso, and T. Hasleaad, Improvng UTRA Capacty wth ODMA, IST Moble Communcatons Summt,