Forward Link Capacity of 3G Wideband CDMA System with Mixed Traffic Sources

Transcription

1 Forward Link Capacity of 3G Wideband CDMA System with Mixed Traffic Sources Wan Choi* and Jin Young Kim** * Research and Development Center, KT Freetel, Korea **School of Electronics Engineering, Kwangwoon Universig, Korea ABSTRACT In this paper, we analyze and simulate forward link capacity of a CDMA system with mixed multirate sources in a multipath fading channel. The outage probability of the forward link is derived for a CDMA system with mixed multirate sources. The forward link capacity is analyzed in terms of the number of multipath, the number of rake fingers in a mobile station, closed loop power control, and impact of soft handoff. The results in this paper can be applied to overall system design of a CDMA system with multimedia services. I. INTRODUCTION These days, the majority of Internet services are provided via wire line, it is expected that Intemet service via wireless medium will dramatically increase in the near future. Existing CDMA (code division multiple access) cellular systems are designed to support for voice and lowspeed data services, but not for high rate data service [l, 21. Recently evolved and implemented CDMA standard, IS- 95B can support up to kbps data service by using multiple code [3]. In the IS-95B system, up to eight codes can be assigned to a single data service user for higher rate transmission, and a fundamental code channel must remain assigned to the data service user for entire call duration even though the data service user has nothing to transmitheceive. However, the IS-95B system is not sufficient so as to cover emerging data services in terms of spectrum and coverage efficiency [4]. In this paper, the forward link capacity of the CDMA system with mixed multirate sources is analyzed and simulated in a multipath fading channel. By introducing a forward link power factor, the forward link Erlang capacity and outage probability is obtained in a closed form. The rest of this paper is organized as follows: Data traffic and interference models are described in section 11. Forward link outage probability and Erlang capacity are derived in section 111. Numerical results are presented in section IV, and conclusions are drawn in section V. 11. SYSTEM MODEL II. 1. Data Trafpc Model In the source level, data traffic is commonly modeled as an ON/OFF source [5]. This model basically consists of two major sets of parameters: 1) distributions of ON/OFF periods, and 2) distribution of packet amvals during an ONperiod. In this paper, for simplicity of analysis, we assume continuous transmission during a burst or ON-period. Recent studies on intranet and Intemet traffic indicate that though the ON/OFF source model is appropriate for data packet, the distributions of the ON/OFF sources could have infinite variance as opposed to the finite variance assumption in the traditional traffic model [6]. A heavy-tailed distribution with infinite variance called Pareto distribution is well known to match well with the actual data traffic measurements in the application, the source, and the aggregate levels. Recent traffic measurements show that holding time of a data service user (i.e., data call duration) can also be modeled by a heavy tail distribution (such as Pareto distribution), and the arrival of data call remains Poisson (i.e., the inter-arrival time (idle time) can be modeled by an exponential distribution). The probability density of a Pareto distribution is given by f(0 apa, = - (t + py+ where expectation of the Pareto distribution is given by Iqt]=---. P a-1 The Pareto distribution has two parameters, a and,8. The a represents heaviness of the tail of distribution. When the a is closer to 1, the distribution tail becomes heavier, and the traffic becomes more bursty. Once a suitable value for the a is selected according to data characteristic, the,8 can then be set based on mean of the distribution. If the a is between 1 and 2, the variance of distribution becomes infinity. And, the variance becomes finite if the a is 2 or above. The data traffic can be characterized by an activity factor that is defined as a duty cycle. The activity factor, /01/$ IEEE 2620 VTC O I

2 ~ m pd, can be obtained as [5] Tac/ive- holding-rime Pd = 3 Tholding -lime o n 1, (2c) E[TOfI+ EITonI where N is avergage number of sessions per a data call, Ton is the duration of transmitting data, T~~ is the time when there is no data transmission, and E[.] means the expectatation value. The To, depends on file size to be transmitted during the session and transmission rate. The ~4f/ is rather related to the human-computer interaction and server response time. The inter-cell interference power from adjacent BSs to the MS is given by K J I,, = 2 Ea(") 'P& k=ln=l K c p k 'Lk 9 k=l 'L&, (54 (5b) where Pk is total transmitted signal power from the kth BS. Then, the received E, IN, at the MS is given by [6] II. 2. Interference Model Path loss between a mobile station (MS) and the mth base station (BS) is given by = Dil. xm 9 (3b) where D, is the distance between the mth BS and the MS, 1 is path loss exponent (typically 3 to 4), and 5, is gaussian distributed random variable with zero mean and standard deviation (s.t.d.) cr, representing shadow fading. The X, denotes lognormally distributed random variable. Typically, the cr is between 6 to 10 db for the signals from adjacent BSs, and is between 2 to 2.5 db for the singals from home BS when closed-loop fast power control is employed. In our system model, the followings are assumed: 1) K multiple cells are uniformly loaded, 2) cell pattem is hexagonal, 3) data service users are classified into M classes in each cell, 4) each MS receives J multipath rays from each BS, 5) full orthogonality in forward link code channels is not guaranteed because of multipath, and 6) the receiver of MS has J'(0) rake receivers. Then, the intracell interference power on jth finger of the MS is given by I, = $a(.).(l-y.+l).po.lo n=l fl*j = (1 -a(j)). (1-ly. 4,). Po.Lo, (44 (4b) where a(') is power portion of the jth multipath so that y is a fraction of BS power assigned to traffic ia(1) = J=I channels, is relative power for the ith user, and Po is total transmited signal power from home BS. where W is spreading bandwidth, R is data rate, K J I,, = c xa(n) pk. L, 9 and I,, = i,(n).(1- vd). Po. L,. k=ln=l n=l ntj The transmitted signal power of a BS is actually a function of the number of ongoing calls served by the BS and thus, generally, it can be modeled as a random variable. However, in this paper, we assumed that all BSs are accomodating maximal ongoing calls and transmitting signals at full power level. This corresponds to the worst case condition PERFORMANCE ANALYSIS The background noise is assumed to be negligilbe compared to the total signal power. Hence, from (6a) to (6b), the received E, I Nt at the MS is given by where so(= p0. L,) denotes total received power at the MS from home BS. As shown in Fig. 1, in the worst case, i.e., the MS is at cell boundary, the ratio of inter-cell interference power to received power at the MS from the home BS is given by $ X k +(2)-1 5 X k +(2.633)-'? X k - k=l k=3 k=6, (8b) xo where D, is the distance from the kth BS to the MS, and /01/$ IEEE VTC'OI

3 xk is a lognormal random variable representing shadow fading from the kth BS. The can be approximated as a SO lognormal random variable with a mean db value my and a standard deviation of db value cry. Since I,V.+i << 1, the received E, I NI of (7) can be approximated as From (9), we can find that the received E, IN, at the MS is different according to data rate, R, and relative power for the ith user, bi. If the di, i = 1,..., M, denotes data service class, then the received E, IN, of voice user, d, -class data service user,...,and d, -class data service user are as follows. (9) For the worst case that all the MSs are at the cell boundary, now we introduce an average forward link power factor, 7, in order to take into account the fact that most of the MSs are not located at the cell boundary [8]. If all the MSs are uniformly distributed at each cell, and 4th power law and perfect power control are applied, the 17 is about 0.4. The average forward link traffic channel power is about 7 times the traffic channel power needed to serve the MS at the cell boundary. Then, outage probability is given by = ~r [. +] where p/') denotes activity factor of service class (.). IV. NUMERICAL RESULTS For the numerical examples, path loss exponent I = 4, target outage probability of 0.01, soft handoff fraction g = 0.3, and a fraction of BS power assigned to traffic channels - I,V = 0.8 are assumed. We consider two types of data wyl J ~ O ) #M).,(i) Yd, - c (12) services: a web service and a real time service. For the Web Rd, i=l.!e + (1 -,(i)) service, the average file size transmitted during a burst is SO assumed to be 13.9 Kbytes, and the mean OFF period, EITof], is 10.5 seconds. For the real time service, the data We assume that there are K, voice users, Kd, d, - activity is assumed to be unity because it requires class data service users,..., and Kd, d~ -class data continuous transmission. service users in a cell. We also assume that K,, Kd,,..., KdM are Poisson random variables with call arrival rates AV, Ad,,..., AdM, respectively and average call durations pv,,ud,,...,,udm, respectively. In (11) and (12), the relative power for the ith user being served is given by In Fig. 2, a forward link Erlang capacity of the cdma2000 (chip rate=l.2288mcps) is shown for multipath environment, standard deviation of lognormal shadow fading 8 db, forward link power control error 2.5 db, and power factor 0.4. We consider only three kinds of multirate traffic sources: 1) 9.6 kbps voice service with the required E,, /NI of 4 db and the voice activity factor 0.4, 2) 76.8 kbps real time data service with the required Eb IN, of 3 db, and 3) kbps web browsing service with the average required Eb / Nt of 3 db. The system can support any Erlang sets below the three dimensional surface. It is shown that the system can support more calls of 153 kbps web service than those of 76.8kbps real-time service because the capacity depends on not only the data rate of traffic sources but also the traffic pattern reflected in the activity factor. In Fig. 3, a forward link Erlang capacity is shown for the cdma2000 (chip rate=l.2288mcps) with multipath /01/$ IEEE 2622 VTC'O I

4 environment, and three rake receivers, i.e. J (0) = 3 with power portions for the three-tap rake receivers are (0.8, 0.1, 0.05), respectively. Other parameters assumed in this figure are the same as those of Fig. 3. It is shown that the capacity for the case with multipath components is reduced compared to the case without multipath components. Under the multipath environment, the orthogonality among code channels cannot be maintained, and this eventually leads to the capacity decrease. In Fig. 4, forward link Erlang capacity of the cdma2000 (chip rate=3.6864mcps) is shown for no multipath components, standard deviation of lognormal shadow fading 8 db, power control error of 2.5 db, and forward link power factor 0.4. Three kinds of multirate traffic sources are considered: 1) 9.6 kbps voice service with the required E, I NI of 4 db and the voice activity factor of 0.4, 2) kbps real time data service with the required E, IN, of 3 db, and 3) kbps web browsing service with the required Eb I N, of 3 db. The system can support any Erlang sets below the three-dimensional surface. The capacity is proportional to the spreading bandwidth but it is not linearly proportional because the outage probability is not a linear function of spreading bandwidth. In Fig. 5, a forward link Erlang capacity bound of the cdma2000 (chip rate=3.6864mcps) is shown for three-tap rake receivers, i.e. J (0) = 3 with power portions (0.75, 0.2, O.Ol), power control error of 2.0dB, and forward link power factor 0.4. Three kinds of multirate traffic sources are considered: 1) 9.6 kbps voice service with the required Eb IN, of 4 db and the voice activity factor of 0.4, 2) kbps real time data service with the required Eb I Nt of 5 db, and 3) kbps web browsing service with the required Eb IN, of 3 db. In Fig. 6, we show the impact of power control error on the forward link capacity of the cdma2000 (chip rate=3.6864mcps) under multipath environment [7,8]. The conditions assumed in this figure are the same as those of Fig. 8 except that there are only two kinks of multirate traffic sources: 9.6 kbps voice and kbps real time services. It is confirmed that the power control error has a negative influence on the forward link capacity. link than in the reverse link. The results in this paper can be applied to overall system design of a CDMA system with multimedia services. REFERENCE [l] N. B. Mandayam, J. M. Holtzman, and S. Barberis, Erlang capacity for an integrated voice/data wireless CDMA system with variable bit-rate sources, in Proc. of IEEE PIMRC 95, pp , Toronto, Canada, Sept [2] I. Koo, J. Ahn, J. Lee, and K. Kim, Analysis of Erlang capacity for the multimedia DS-CDMA systems, IEICE Trans. Fundamentals, vol. E82-A, no. 5, pp , May [3] D. Ayyagari and A. Ephremides, Cellular multicode CDMA capacity for integrated (voice and data) services, IEEE J. Select. Areas Commun., vol. 17, no. 5, pp , May [4] K. S. Gilhousen, I. M. Jacobs, R. Padovani, A. J. Viterbi, L. A. Weaver, Jr., and C. E. Wheatley, On the capacity of a cellular CDMA system, IEEE Trans. Eh. Technol., vol. 40, no. 2, pp , May [5] R. Jain and S. A. Routhier, Packet trains: Measurements and a new model for computer network traffic, IEEE J. Select. Areas Commun., vol. 4, pp [6] M. Cheng and L. F. Chang, Uplink system performance of high-speed IS-95 CDMA with mixed voice and bursty data traffic, in Proc. of IEEE PIMRC 98, pp ,Boston, MA,U.S.A., Sept [7] F. Kikuchi, H. Suda, and F. Adachi, Effect of fast transmit power control on forward link capacity of DS- CDMA cellular mobile radio, IEICE Trans. Commun., vol. E83-B, no. 1, pp.47-55, Jan [8] W. Choi and J. Y. Kim, Forward link capacity of a DSICDMA system with mixed multirate sources, to be published to the IEEE Trans. Eh. Technol. V. CONCLUSIONS We demonstrated that the capacity of a CDMA system with multirate sources could be effectively depicted in the multidimensional surface through the numerical examples. In the future mobile communication systems, it is expected that a forward link can be a limiting link because perspective services are likely to require higher data rate in the forward O /OI/$I I IEEE 2623 VTC O I

5 kbps data (Erlang) kbps data (Edang) Fig. 1. Forward link interference model when mobile station is at cell boundary. Fig. 4. Forward link Erlang capacity of cdma2000 (chip rate=3.6864mcps) with no multipath component kbps data (Erlang) kbps deta (Erlmg) Fig. 2. Forward link Etlang capacity of cdma2000 (chip rate=l.2288mcps) with no multipath component kbps data (Erlang) kbps data (Erlang) Fig. 5. Forward link Erlang capacity bound of cdma2000 (chip rate=3.6864mcps) with multipath components. I powercontrol error= 1.5 db.i powercontrol error = 2.0 db I.o "'X 0.8 i 0.6 w kbps data (Erlang) kbps data (Erlang) Fig. 3. Forward link Erlang capacity of cdma2000 (chip rate=l.2288mcps) with multipath components IS I kbpr voice (Erlang) Fig. 6. Forward link Erlang capacity bounds of cdma2000 (chip rate=l.2288mcps) with multipath components /01/$ IEEE VTC'O 1