CELLULAR SYSTEM CAPACITY and PERFORMANCE IMPROVEMENT wth SDMA Flpe Alves, Henrque Rbero, José Fernandes 3 Escola Superor de Tecnologa, Insttuto Poltécnco de Castelo Branco, 6 Castelo Branco, Portugal Phone:+35-7-339355 Fax:+35-7-339399 e-mal: falves@est.pcb.pt Insttuto das Comuncações de Portugal, 99-7 Lsboa, Portugal 3 Insttuto de Telecomuncações, Unversdade de Avero, 38 Avero, Portugal Abstract The need to fnd solutons that would support the ncreasng number of users has been stmulatng many research groups to explore new technques to ncrease the system capacty of the moble communcaton systems. In ths paper, the mpact of SDMA (Space Dvson Multple Access) technque n system performance and capacty s evaluated. Carrer-to Interference (C/I) level for FDMA / TDMA and for CDMA based type systems s analyzed. Spectrum effcency mprovement and ts mpact on system capacty (maxmum users per cell) due to SDMA s also shown n smulaton results. I. INTRODUCTION Wreless cellular communcaton system s wtnessng a rapd growth n marets, technology and range of servces. The ncreasng number of users and, the hgh densty n urban areas are forced to explore new technques to mprove the spectrum effcency. At the begnnng of the nd generaton moble communcaton system, -drectonal antennas were used at the cell ste. The result s an neffcent use of transmtted power, and C/I level to guarantee a good system performance s dffcult to control. The sectorzaton has been ntroduced at the base statons but ths technque requres a frequency re-assgnment plan and the ncrease of the system capacty s lmted. Adaptve antennas allow to explore the spatal doman more effcently tang advantage of the spatal separaton between the users to transmt the power through drectve beams to them. The nterference level s reduced and the system performance and capacty are mproved. Moreover, ths technque brngs another advantage that s the ablty to allocate dynamcally the resources accordng to the spatal dstrbuton of the traffc In secton II, the mpact of sectorzaton and SDMA on the FDMA / TDMA systems s studed and secton III analyzes the CDMA system performance mprovement usng sectored antennas and adaptve arrays. Fnally, secton IV presents the man conclusons. II. TECHNIQUES TO IMPROVE THE FDMA/TDMA SYSTEM PERFORMANCE A. Sectorzaton The sectorzaton conssts n the cell splttng n two or more sectors and s used to reduce the number of co-channel cells and the co-channel nterference level. As C/I ncreases, the cluster sze N and frequency re-use dstance can be reduced and the capacty of the system ncreases. If each cell s sectorzed n 3 sectors of º each, there are 3 cochannel cells for cluster sze of N=3, and co-channel cells for N=4 and N=7. If the sectorzaton s made wth 6 sectors of 6º each, co-channel cells exsts for N=3, and only one co-channel cell s present on the system for N=4 and N=7. The performance mprovement due to sectorzaton technque s evaluated by comparng the C/I level obtaned wth the reference values for the -drectonal antennas. The propagaton model that was used n the smulatons s gven by []: ( d) = P d () P r s the receved power at dstance d, P s a constant that nclude the transmtted power P t and the gans of the emtter and recever, and n s the path loss coeffcent. For n=4, ths model have 3dB more than Hata-Oumura model when the base statons have 5m heght. C/I s gven by []: C / I = I () P = For downln, P r s power receved from base staton of the current cell, P I s the power receved from base staton n I co-channel cells. Substtutng equaton () n (), the C/I level s gven by the followng relatonshp []: R C / I = I (3) D = For downln, R s the dstance between the desred user and the central base staton, and D s dstance between ths user and the co-channel base staton. Fgure shows the reference scenaro used n the smulatons. I He s now wth European Commsson, DG Informaton Socety Unt E4
D 5 D 4 R D 6 D 3 Fg. Scenaro for smulatons on FDMA / TDMA systems In moble communcatons, downln and upln must be balanced, and [3] shows that smulaton platform made wth the reference scenaro n the Fg. and equaton (3) to calculate the values of C/I s balanced. Identcal results can be obtaned f the sectorzaton s appled at the moble termnal. However, f t s appled at both base staton and moble termnal, for º, the number of co-channel cells can vary between or and are dependently of the angle of arrval from the moble. The results for downln smulatons are presented bellow n Fg.. In ths scenaro, 38 unformly dstrbuted postons on the cell are randomly generated for the moble. Cluster sze s N=7, path loss n=4, and m s used as the radus of the cells. Each C/I value s calculated wth equaton (3) measurng the dstances R and D for each poston of the moble termnal n the cell. ob(c/i < abcssa),8,6,4, D 5 5 5 3 35 4 45 5 55 6 65 7 75 8 D -drectonal sect. º at base staton sect. º at base staton and moble sect. 6º at base staton Fg. C/I levels for sectorzaton n FDMA / TDMA system B. SDMA Adaptve antennas are systems n whch the radaton pattern can change dynamcally accordng to the traffc densty varatons. Maxms and nulls can be steered n specfcs drectons accordng to some rules that can be appled at the beamformer. The maxm are steered n drecton of the desred user and the null are steered to the drecton of the nterferng user. Thus, nterference level can be reduced drastcally and the system performance mproved. SFIR (Spatal Flterng for Interference Reducton) s a balanced system that has only one drectve beam to trac the user n all drectons. Addng beamformers at each cell stes, the re-use of one carrer or tme-slot s possble wth the same antenna array, and each beam can be multpled by SMG (Spatal Multplex Gan). The channels re-use s called SDMA. Wth SDMA, SMG users can share the same channel because they are dstngushed by ther spatal poston (angle of arrval). For downln, the co-channel nterference s occurs only when two or more beams ntersect the same user. One of these beams s the beam of the current cell and the other s from an nterference cell. The probablty of a co-channel beam wth beamwdth amng a moble n drecton θ s gven by: P θ = (4) () π If t s assumed that each co-channel beam am the same user wth ndependently occurrence, the probablty of moble unt s to be affected by co-channel beam, for SFIR s gven by [4]: P () = π As the SDMA s SMG replcatons of SFIR, ths probablty s now: P = SMG (6) () π SMG s number of beams per channel n each co-cell. Fg. 3 shows the CDFs of varous C/I, and the table n the fgure s the probablty of occurrence of each C/I level. ob. (C/I < Abcssa),8,6,4, ob. Of occurrences (n %): Nº nterf. Sfr SMG= SMG=4 SMG=8 4.7 8.33 6.67 33.33.7.69.78. 3..5.46 3.7 4.8.3 5..4 6.4 5 5 5 3 35 4 45 5 55 6 co-channel nterf. co-channel nterf. 3 co-channel nterf. 4 co-channel nterf. 5 co-channel nterf. 6 co-channel nterf. Fg. 3 C/I levels as functon of co-channel number for SDMA In ths smulaton, cluster sze s N=7, path loss factor n=4, beamwdth s 5 o and SMG s the same number n all cells. These results allow the evaluaton of the system performance but system capacty can be analyzed only wth spectrum effcency gan. The spectrum effcency measures the gan obtaned n users per MHz per m related to -drectonal systems, and s gven by the followng relatonshp [5]: (5)
N m G = (7) β h m h N s cluster sze, β s beamwdth, m s the number of physcal channels (tme-slots) and h s number of sectors per cell. The results of the smulaton for 3 and 6 sectors per cell, for and beams of 5º per channel are presented n Fg. 4: Users/MHz/m 8 6 4 sect.º, Beamwdth=5º, SMG= sect.º, Beamwdth=5º, SMG= sect. 6º sect. º -drectonal 5 5 5 3 P r represents the power receved at base staton from the test moble, P s ntra-cellular nterference, P j s ntercellular nterference and N s the and t s assumed that all cells are equally dstrbuted. No thermal nose s consdered here, and t s assumed a perfect power control. Usng the propagaton model descrbed n secton II, Fg. 5 llustrates how the nter-cellular nterference can be calculated, beng the C/I level gven by : C / I = (9) N 8 n N + d d ( ) ( ) d s dstance between the external user and the external base staton, d s dstance between the external user and the nternal base staton. d External moble termnals d Internal moble termnals Radus of cell (m Fg. 4 Spectrum effcency of sectorzaton and SDMA III. TECHNIQUES TO IMPROVE CDMA SYSTEM CAPACITY In a CDMA system, all users are transmttng n the same RF band, ndvdual users are dstngushed from each other va the use of orthogonal codes. IS-95 specfes that each user convey baseband nformaton at 9.6bps. The rate of the fnal spread sgnal s.88mcps, resultng n an RF bandwdth of approxmately.5mhz [6]. Ths bandwdth wll be 5Mhz for UMTS accommodatng 3.8Mcps [7]. The performance of a CDMA system s nterference lmted because de frequency reuse s N=. Ths means that the capacty of the system s lmted by the amount of nterference power present n the band. Capacty s defned as the total number of smultaneous users the system can support, and performance s evaluated under the condton of rado ln assgned to a partcular user and for a partcular type of servce. The qualty of the rado ln s related to the bt error rate (). Dfferent types of nterference levels can be produced n CDMA system and dstngushed as the follow: ntra-cellular nterference, defned as the nterference power produced by the users n the same cell (nternal moble termnals), ntercellular nterference, defned as the nterference power produced by the user n the other cells (external moble termnals). The scenaro used for up-ln smulatons s a mcro cellular envronment and s smlar to the used n [8]. C/I levels are expressed by: C / I = (8) N N 8 P + P j=, Intra cellular j cellular Inter Fg. 5 Scenaro to calculate the nterference of external moble unts A. Sectorzaton and SDMA In ths secton, sectorzaton and SDMA are appled only n the central cell. For sectorzaton case the central cell s dvded on 3 sectors of º each. In CDMA systems, snce all moble unts share the same frequency band, the number of potentals nterferng unts s very hgh and the SDMA concept must be changed. The man beam s smply pontng n drecton of the desred user, and the nterference level s decreased due to the secondary lobes of the antenna array radaton pattern that attenuates the sgnals receved from the other drectons. In ths paper, the mplementaton of SDMA technque s acheved wth a very smple antenna array that has always the same radaton pattern presented n Fg. 6 wth maxmum amed to the desred moble and perfect tracng s assumed (N s the number of array elements). normalzed gan.9.8.7.6.5.4.3.. N=8 N=6 N=3 - -8-6 -4-4 6 8 angle (degree) Fg. 6 Radaton patterns of the array used n the smulatons Consderng the gan of the array, expresson (9) result n the followng relatonshp:
C () / I = N 8 n ( ) ( n N + d d G ) G s the normalzed gan of the array n the drecton of the nterferng moble unt. In Fg. 7 C/I levels can be observed for urban area (n=5) where the cells have radus of 5m. Fg. 8 Smulaton results for spreadng factor G p =4,E+,E-,E-,E-3,E-4,E-5 - -4 SDMA, N=3-6 -8 SDMA, N=6 - SDMA, N=8 - SECT. (º) -4 OMNI. -6-8 5 5 3 35 4,E-6 sect.,e-7 SDMA, N=6,E-8 5 5 3 35 4 45 Fg. 9 Smulaton results for spreadng factor G p =6,E+,E-,E-,E-3 Fg. 7 Smulaton results for sectorzaton and SDMA The system capacty s obtaned from a mnmum value of C/I, whch can be calculated by [6]: Eb C = G p () N I v where G p s the spreadng gan and v s voce actvty. Wdeband CDMA suggests QPSK modulaton and spreadng factor between 4 and 56 [9]. The for ths modulaton technque s gven by []: Eb = Q () N Substtutng equaton () n (), for voce actvty of %, the result s: C = Q G (3) p I For the C/I levels presented n Fg. 7, and for spreadng factor G p =4, G p =6 and G p =64, the Fg. 8, to Fg. show the as functon of number of users. Wth these results, t s possble to evaluate the system capacty as functon of the system performance.,e+,e-,e-,e-3,e-4 sect. SDMA, N=6 5 5 3 35 4 45,E-4,E-5,E-6,E-7,E-8 sect. SDMA, N=6 5 5 3 35 4 45 number of use rs per cell Fg. Smulaton results for spreadng factor G p =64 If the power control s not perfect, all of these results can change accordng to the Table I [8]. Power control error standard devaton (db) Capacty reducton (%). 3.5 8. 3.5 Table I Capacty reducton wth power control error IV. CONCLUSIONS For FDMA / TDMA systems, sectorzaton wth 3 or 6 sectors mprove C/I on 6dB and db respectvely, and f the moble unt s also sectorzed wth 3 sectors of º, the mprovement s.5db more than sectorzaton of º at base staton only. Wth SDMA, the capacty can be ncreased sgnfcantly by the channel re-use and n many cases the spectrum effcency can be two or four tmes hgher. In CDMA systems, the sectorzaton technque can ncrease the system capacty untl 5 users per cell related to the -drectonal case. SDMA technque can multply
ths capacty by a factor of 5. Ths factor depends on the beamwdth and the capacty ncreases as the secondary lobes decrease. The last smulatons show the capacty and performance mprovement due to the mplementaton of the spatal dvson technques only n the central cell. If these technques were appled n all cells, and wth narrow beamwdth and low secondary lobes, the capacty for a gven wll be much hgher. These results led us to conclude that the SDMA s very advantageous wll be used on 3 rd generaton moble communcaton systems. V. REFERENCES [] T. S. Rappaport, Moble Communcatons ncpales and actce, entce-hall, 996, pp. 37-4. [] W.C.Lee, Moble Communcatons Desgn Fundamentals, Wley Interscence, 993, pp. 4-44. [3] Flpe Alves, Henrque Rbero, José Fernandes, Impacto da técnca SDMA em Sstemas Celulares, Electrónca e Telecomuncações, Revsta do Dept. de Electrónca e Telecomuncações da Unversdade de Avero, Vol II, pp. 83-83, January. [4] P.Petrus, J.H. Reed, Capacty Enhancement Usng Adaptve Arrays n an AMPS System, n IEEE Transactons on Vehcular Technology, vol. 47, nº3, August 999. [5] M. Tangemann, et all, Report on the Benefts of Adaptve Antennas for Cells Archtectures, CEC RACE Delverable Document, ERA Technology, May 995. [6] S. C. Yang, CDMA RF System Engneerng, Artech House, 998, pp. 75-85. [7] R. asad, W. Mohr, W. Konhauser, Thrd Generaton Communcaton Systems, Artech House,, pp. 7-3. [8] T. Ojanpera, R. asad, Wdeband CDMA for Thrd Generaton Moble Communcatons, Artech House, 998, pp. -8. [9] R. asad, W. Mohr, W. Konhauser, Thrd Generaton Communcaton Systems, Boston, Artech House,, pp. 73-75. [] A. B. Carlson, Communcaton Systems, Mc Graw Hll, 3 rd edton, 986, pp. 54-549.