On the Downlink Capacity of WCDMA Systems with Transmit Diversity

Transcription

1 On te Downlink Capacity of WCDMA Systems wit ransmit Diversity Vaibav Sing, Oya Yilmaz, Jialing Wang, Kartigeyan Reddy, and S. Ben Slimane Radio Communication Systems Department of Signals, Sensors, and Systems Royal Institute of ecnology (KH) Electrum 48, 64 4 KISA, Sweden Abstract ransmitter diversity at te base station as been included in WCDMA standard for tird generation mobile systems as a means of increasing te downlink system capacity. In previous works, it as been sown tat open loop transmitter diversity can significantly increase te downlink capacity. However, te performance in multipat environment is not known. In our study, we present te simulation results for downlink system capacity taking into account te effect of Doppler frequency and multipat fading. e results sow tat even in severe multipat wireless cannel, diversity sceme can greatly improve te system capacity wen no soft andover is used. I. INRODUCION e Space ime ransmit Diversity (SD) sceme adopted by 3GPP in te current WCDMA standard is based on a modification of te Alamouti space time coding sceme ]. However tis sceme is based on ortogonal design. In a WCDMA system, due to multipat propagation, te signals received will no longer be ortogonal. Previous works on tis topic ], 3], 4] ave mainly focused on te impact of multipat fading on link level performance. In 3], 4], system level results ave been presented for SD but tey do not consider te increase in interference caused by multipat propagation. In tis paper, we present results for te downlink capacity of a WCDMA system bot wit SD and witout SD. We focus on te SD performance in typical urban macro cellular environment, assuming eiter te IU Veicular-A or te Pedestrian-A multi-pat delay profile 5]. We consider only speec traffic over dedicated cannel. In Section II, te concept of WCDMA system is introduced and te transmitter diversity sceme SD is discussed. In Section III, te system model used in te simulation is described. In Section IV, te simulation results are presented and analyzed. In te last section, a conclusion is drawn. II. WCDMA CELLULAR SYSEM A. Cellular System Model We consider a multi-cellular WCDMA system wit exagonal cells and tree sectors per cell. Eac sector is using an SD encoding sceme according to Figure. In tis work we only consider R =. kbps speec service over dedicated cannels in te downlink. In order to model te intra-cell and inter-cell interference adequately, te link between eac mobile and eac base station is modeled explicitly. We do not model te pilot cannel transmitted by eac base-station. b,b Fig.. SD b,b r,r a(t) RAKE receiver Dec. b,b ˆb, ˆb WCDMA wit Alamouti SD encoding and decoding. B. Cannel Model In cellular systems, te signals are exposed to bot long term fading and sort term fading effects. In general, te long term effect is modeled by a deterministic pat loss and a stocastic part modeled as a lognormal process. In tis paper, we model te long term effect between antenna of te mt sector and mobile user i as ]: g mi = d µ mi G mi () were d µ mi is te distance dependent patloss, d mi is te distance from te mobile to te mt cell site, and µ is te patloss exponent. G mi represents te log-normal sadow fading, i.e., log (G mi ) is a zero mean Gaussian random variable wit standard deviation σ s db. As te transmit antennas of eac sector are co-located, tey are assumed to ave te same long term effect. e sort term effect is modeled as frequency selective fading cannel wit equivalent lowpass cannel impulse response given by (t, τ) = i δ(τ τ i ) () were N is te total number of pats, i is zero-mean complex Gaussian random variable wit power spectral density σi representing te gain of pat i, andτ i is te time delay of pat i. We furter assume a normalized cannel fading coefficients, i.e., i = σi =. (3)

2 wo multipat delay profiles are considered in tis paper: e IU Pedestrian-A (Ped-A) and Veicular-A (Ve-A) power delay profiles.e power of te multi-pats in te power delay profile as a ci-square distribution wit two degrees of freedom wile te amplitude is Rayleig distributed wit a classical Doppler spectrum. e cannels of te different transmit antennas are assumed independent. C. Receiver Model e receiver (mobile unit side) consists of one omnidirectional antenna and a RAKE receiver aving a total number of fingers equals to te number of pats N. Considering a certain mobile unit witin te coverage area of sector m, say mobile u, te equivalent lowpass of te received signal can be written as r(t) = l= s kl (t τ il )i e (t)z(t) (4) il K m k= were K m is te total number of active mobile users witin sector m, i e (t) is te external interference coming from oter sectors, z(t) is complex Gaussian random process wit zeromean and power spectral density N, il is a complex process representing te fading coefficient of pat i, s kl (t) is te transmitted signal of user k from transmit antenna l, s kl (t) = φ k g m p m a k (t)b k (t), were eac data symbol is assumed to ave unit amplitude, i.e., b k (t) =and te spreading waveform is denoted by a k (t). Herep m is te total transmitted power of base station m, g m is te link gain between te mobile receiver and base station m, andφ k is te portion of power allocated to user k. e spreading bandwidt is W / c were c is te cip duration. Wen te period of te spreading waveforms is larger tan te information symbol duration, te correlation properties follow ρ ik (τ) = For a spreading waveform defined as a i (t) = n= a i (t τ)a k (t)b i (t τ) dt a i (t τ)a k (t) dt. (5) a i,n c (t n c ) (6) were a i,n = ± wit equal probability and (t) is a rectangular pulse of lengt and amplitude equals to one, te cross correlations (5) can be determined from te results of 6]. Using our notation and considering users witin te same cell, it can be sown tat wen i k, {, τ = varρ ik (τ)] = W, τ (7) and { R varρ ii (τ)] =, τ = W, τ (8) were = for syncronous waveforms and = /3 for asyncronous waveforms. Assuming perfect cannel state information and tat te delays {τ i } N are known, te RAKE receiver output samples corresponding to te transmitted symbols, b,b, are obtained as follows: l r l = a u (t) ir(t τ i )dt, l =, (9) (l ) were a u (t) is te spreading code of mobile user u. Assuming BPSK modulation, te two symbols, b and b, after SD combining are estimated as ˆb =sgn{y } =sgn{r r } () ˆb =sgn{y } =sgn{r r } () Considering te receiver of mobile station in sector m, te combined sample at te mobile station, y, can be represented as: ( y = n, n, ) φ g m p m ρ () were p= n, ( p, p, ) n, ( p, p,) ] φ g m p m ρ (τ n, τ p, ) K m n, ( p, p, ) n, ( p, p,) ] p= k= φ k g m p m ρ k (τ n, τ p, ) ( ) n, i,n n, i,n ( ) n, z,n n, z,n z l,n = l (l ) () z(t τ il )a (t) dt (3) is zero-mean complex Gaussian wit variance N, i l,n = l (l ) i e (t τ il )a (t) dt (4) is te external interference sample, and ρ km (τ) is as defined earlier.

3 e communication quality of interest for mobile user, related to te bit error probability, is given by ( ) = Ey k b k (t)]e y k b k (t)] vary k b k (t)] N ( = n, n, )] φ g m p m Rσ (5) were E is te expectation operator, R is te information data rate, and te factor comes from te fact tat te user power is split between te two transmit antennas. Wit te assumption of te terms in te combined sample being Gaussian and mutually independent, te variance of te interference becomes N σ ( = n, n, )] ] θ N g p W B g i p i N W i= (6) were B is te total number of interfering base stations, θ N is te total ortogonality factor caused by fading multipat cannels and is given by n,( p, p, ) n, ( p, p,) θ N = p= N m= ( m, m, ) Replacing (6) in (5) and simplifying, te bit-energy-tointerference-spectral-density ratio of mobile user can ten be rewritten as ( ) N ( = W n, n, )] φ g m p m R θ N g p B i= g (7) ip i N W For te case of one transmit antenna, te bit-energy-tointerference-spectral-density ratio of mobile user takes te following form: N ] ( ) = W n, φ g m p m R θ N g p B i= g (8) ip i N W were now te total ortogonality factor, denoted θ N,isgiven by θ N = p= n, p, N m= m, (9) Comparing te two scemes, it is observed tat te SD sceme as a better diversity order and as te capability to resolve te fading multipat components from te two transmit antennas. However, it is not clear wic of te two metods perform better since tey ave different total ortogonality factors. e strengt of tis ortogonality factor will affect te system performance and it will be interesting to see te interaction between te increase in diversity order and te increase in experienced interference wen SD is used in WCDMA systems. Assuming tat eac user requires a maximum tolerable bit error probability tat can be mapped into equivalent minimum E b /I value, denoted γ t. Hence, te signal quality of mobile user can be measured using te outage probability defined as ( ) ] O = Pr <γ t () were γ t is te minimum required SIR for good signal quality. III. SIMULAION MODEL e simulation is performed over a system wit 9 sites and eac site as 3 sectors. Wrap-around is used to eliminate te edge effects. e available system bandwidt is W =5MHz. e system offers speec service at a data rate of R =. kbps. e maximum base station power is limited to W4]. e maximum transmission power per link is W4]andte initial power is set to. W. e environment we assume is an urban environment. So we set te propagation constant µ to 4. e gain constant is cosen to be 8 db. e standard deviation for lognormal sadow fading is cosen as db. We use soft andover wit a andover margin of 3 db. e SIR balancing algoritm is used for power control and te power control is done in steps of db. Hence in transmitter diversity case, te power of eac antenna is canged in steps of db. In our model, te active mobile users are assumed to be uniformly distributed over te cells. e cell load is Poisson distributed. We consider two types of traffic models. Mobiles wit low mobility (3 km/r) Mobiles wit ig speed (6 km/r) e SINR tresold after despreading, γ t,issetto7 db corresponding to a BER of 3 (standard for voice communication systems) ]. All te simulation parameters considered in te paper are summarized in te following table: I ABLE I SIMULAION PARAMEERS Item Symbol Value System Bandwidt W 5 MHz Cip Rate R c 3.84 Mcps Information Bit Rate R. Kbps Gain Constant c 8 db Cell Radius r m Propagation Constant µ 4 Maximum ransmission Power p max 33 dbm 4] per link Initial ransmission Power p init dbm otal Base Station Power P tot W Noise Floor N 7 dbm 4] SIR resold γ t 7 db Lognormal Standard Deviation σ s db IV. SIMULAION RESULS e simulation is done for wit and witout diversity cases. e results obtained witout diversity are used as reference to evaluate te gain obtained wit SD. is is done for te two cannel profiles and te two different mobile speeds. e SIR of te user after despreading is averaged over te last five 3

4 frames after te power control algoritm as converged. If tis averaged SIR is less tan te target SIR of 7 db, te user is considered to be in outage. e outage probability is te ratio of te number of users in outage to te total number of users in te system. o reduce te statistical variance, we run five independent simulations and average te results obtained for tese simulations. e outage probability as a function of te cell load for te case of no D and te case of SD for low mobility and Ped-A cannel profile is sown in Figure. It is observed tat SD provides significant gain in WCDMA system capacity. For example, for a % outage probability, a 3% capacity gain is obtained wit SD. Ve A Model, Low Mobility (3km/) No SD SD Ped A Model, Low Mobility (3km/) data data Fig. 3. as a function of te cell load for Veicular-A model and low mobility (3 km/). A. System Capacity for Hig Mobility In order to find te effect of mobility on capacity, we also evaluate te system capacity for ig mobility. Results obtained for Ped-A cannel are sown in Figure 4. Figure 5 sows te capacity for Ve-A environment. e gain in Ped A Model, Hig Mobility (6km/) Fig.. as a function of te cell load for Pedestrian-A model (3 km/) and low mobility wit data corresponding to te case of No SD and data to te case wit SD. e results obtained wit Veicular-A model are sown in Figure 3. For tis case, te capacity gain for % outage is 3%. It is seen tat te gain obtained for Ve-A cannel is less tan tat for te Ped-A cannel. Since te number of pats in Ve- A model is iger, tere is already a iger degree of multipat diversity wic te RAKE receiver can take advantage of. is is te reason wy te gain wit SD is less for Ve-A environment tan for Ped-A environment. Using SD elps to smoot out te fading effect, wic results in a lower transmitted power. If a user is in a bad position, te transmitted power for tat user as to be increased in order to acieve te target SIR. is increases te interference for te oter users. For low mobility case, te fading will last for a longer time. By using transmitter diversity, we are able to negate te effect of fading and acieve improved system capacity. us for low mobility tere is more diversity gain and less interference averaging. No SD SD Fig. 4. as a function of te cell load for Pedestrian-A model and ig mobility (6 km/) system capacity for % outage is 8% for Ped-A cannel. For Ve-A cannel te gain obtained is very small. Compared to low mobility case, we ave muc lower gain wit SD in ig mobility case. Wen we ave ig mobility, a user does not stay in a bad position for long. us te interference is averaged out. Hence te interference in te system decreases 4

5 Ve A Model, Hig Mobility (6km/) CDF of te ortogonality factor (θ) No D SD F(θ) Fig. 5. as a function of te cell load for Veicular-A model and ig mobility (6 km/).3.. Ped A No D Ped A SD Ve A No D Ve A SD θ Fig. 6. e cumulative distribution function of te ortogonality factor for te cannel models wit and witout SD leading to an increase in capacity. Since in ig mobility we already ave an interference averaging effect, te gain obtained by introducing diversity is lower as compared to te low mobility case. e cumulative distribution function of te ortogonality factor for te two cannel models is illustrated in Figure 6 for te case of WCDMA wit single transmit antenna and te case of WCDMA wit SD. It is observed tat te ortogonality factor increases wen te number of pats from te base-station to te mobile increase. It is also seen tat introducing transmitter diversity into te system increases te ortogonality factor and ence te intra-cell interference increases. In low mobility, te diversity gain is sufficient to compensate for te increased interference and additionally provide a capacity gain. V. CONCLUSIONS e results ave sown tat transmitter diversity provides downlink capacity improvement. e gain obtained is muc iger for low mobility case as compared to ig mobility case. e gain obtained decreases as te number of pats increases. e effect of using SD in combination wit various receiver diversity combining scemes as been studied on te link level in 3]. Future work can look into te system level gain obtained wen using bot transmit and receiver diversity in WCDMA systems. 3] Engstrom, B.,Ericson, M., WCDMA System Level Performance Wit Fast Fading And Non-Ideal Power Control, Proceedings IEEE Veicular ecnology ConferenceSept ] ecnical Specification Group erminals; erminal Conformance Specification; Radio ransmission And Reception (FDD) (Release 5); ecnical Specification: 3GPP S 34. V5.3. (4-4) 5] 3rd Generation Partnersip Project. ecnical Specification Radio Access Network Pysical Layer Procedures. 3GPP S 5.4 V3... October 999 6] Pursley MB., Performance evaluation for pase-coded spread spectrum multiple access communication Part I, IEEE rans. on Commun., Vol. 5, No. 8, pp , ] 3GPP, ecnical Specification 5., Pysical Cannels and Mapping of ransport Cannels onto Pysical Cannels July, 999 8] Parkvall, S., Karlsson, M., Samuelsson, M., Hedlund, L., Goransson, B., ransmit Diversity In WCDMA: Link And System Level Results, Veicular ecnology Conference Proceedings, 9] Canales, M., Valdovinos, A., Gallego, J.R., Gutierrez, F., Performance Analysis Of Diversity ransmission Modes In URA FDD Under ime-varying Multipat Cannels, Personal, Indoor and Mobile Radio Communications,. e 3t IEEE International Symposium on Volume: 3, 5-8 Sept. ] Hamalainen, S., Holma, H., oskala, A., Laukkanenz, M., Analysis of CDMA Downlink Capacity Enancements,Personal, Indoor and Mobile Radio Communications, 997. e 8t IEEE International Symposium on Volume:, -4 Sept. 997 ] Jalloul, L., Roani, K., CDMA Forward Link Capacity And Coverage In A Multipat Fading Cannel,Veicular ecnology Conference, 997 IEEE 47t Volume: 3, 4-7 May 997 ] Selection Procedures For e Coice Of Radio ransmission ecnologies Of e UMS. UMS 3.3 version 3..; ecnical Report; April, 998 3] Bjerke, B. A., Proakis, J. G., Zvonar, Z., Antenna Diversity Combining Scemes in Fading Multipat Cannels, IEEE Wireless Communications pp 97-6, January 4 REFERENCES ] Alamouti, S.M., A Simple ransmit Diversity ecnique For Wireless Communications, IEEE J. Select. Areas, Vol 6, No. 8, pp , October 998. ] Cai, M., Zang, X., Zou, N., Slimane, B., On e Capacity Of CDMA Wit Successive Interference Cancellation,EPMCC 3,Glasgow, Scotland, April 3. 5