Capacity Study for Fixed Multi Beam Antenna Systems in a Mixed Service WCDMA System M˚arten Ericson1

Afif Osseiran2

J´ozsef Barta3

Bo G¨oransson & Bo Hagerman2

Ericsson Research Corporate Unit SE-971 28 Lule˚a Sweden

Ericsson Research Corporate Unit SE-164 80 Stockholm Sweden

Ericsson Research Corporate Unit H-1037 Budapest, Laborc u. 1. Hungary

Ericsson Research Corporate Unit SE-164 80 Stockholm Sweden

Abstract- Antenna array at the base stations allows capacity enhancement in cellular WCDMA networks. In this contribution we compare the relative downlink capacity between a system employing standard three sector site configurations with a system employing four fixed beam antenna systems. Since different packet services are expected to contribute largely to the traffic demand within a few years after initial deployment, this comparison is made in a mixed traffic environment. By realistic computer simulations it is found that a substantial gain can be expected by the introduction of adaptive (smart) antenna base stations. In this contribution a fixed multi beam system is considered, which is a fairly low complex adaptive antenna implementation. The simulations show that generally, a threefold capacity increase can be gained by employing a 4 beam antenna system. For some scenarios even higher gains can be achieved.

I. I NTRODUCTION Advanced (or smart) antenna arrays at the base station is one of the techniques which can be used to help improve both uplink and downlink capacity (or coverage) in WCDMA systems [1, 2]. Due to the expected growth in the number of subscribers using different services in the coming years, capacity limitations may appear in certain cells (e.g. in dense urban environments). Further, macro cells in rural areas will benefit from the increased coverage of advanced antenna array systems. Since WCDMA systems are expected to support a variety of services with different bit rate and quality of service requirements, studies with different user scenarios and environments are important. Up to this date most reported studies of using adaptive antennas has focused on baseband algorithms [3, 4]. Our previous work [5] investigated the benefits of using an adaptive antenna system in a single cell WCDMA (UMTS-FDD) system. Simulation results of different up- and downlink algorithms based on typical user scenarios were presented. In [6] both link and system level results were presented for uniformly distributed speech users and [7] showed the interference distributions in both time and spatial domain for mixed traffic in a non-homogenous environment. This paper is a continuation of these earlier studies, showing results for downlink capacity in a fixed multi beam antenna Email: 1 [email protected] 2 {Afif.Osseiran,Bo.Goransson,Bo.Hagerman}@era.ericsson.se 3 [email protected]

system in mixed traffic scenarios. Detailed radio network simulation studies are performed using a spatial channel model according to the COST259 proposal [8]. Beside voice, the multiservice environment consists of streaming and high data rate www users as these applications are expected to be the dominant wireless data applications. Compared to previous system simulations [6], this paper employs detailed and realistic radio network algorithms. II. F IXED M ULTI B EAM A NTENNA S YSTEMS This contribution deals with fixed (multi) beam antenna systems. That is, a number of antenna beams with fixed pointing directions are used to cover the cell, see Fig. 1. This can be implemented either at RF level with e.g. a Butler matrix or digitally at base band. Using a fixed beam system, the uplink receiver gathers and combines coherently the signal from all possible beams [9, 10]. The transmit beam is usually selected from uplink measurements of the desired signal power (for a specific user) per beam, and selecting the beam having the largest signal power. III. S IMULATION E NVIRONMENT A. General Description The deployment model is a homogeneous hexagon cell pattern with wrap-around to eliminate border effects. In each iteration of the main loop the simulator time is increased by the duration of one frame and all radio network algorithms are executed. However, the power control is executed for each slot. Table I summarizes the system parameters used in the simulations. TABLE I S UMMARY OF SYSTEM SIMULATION PARAMETERS .

Parameter Number of sites Number of cells per site Frame duration Slots per frame Chip rate

Value 7 3 10 15 3.84

Unit ms Mchip/s

TABLE II

B. Propagation Model The propagation model is based on the COST259 model for a typical urban environment and can be written as in (1). G = Ga + Gd + Gslow + Gf ast

[dB],

(1)

where Ga is the antenna gain, Gd is the distance dependent gain, Gslow is the log-normal (slow fading) gain and Gf ast is the fast fading gain. The COST259 also models the spatial behavior of the channel. The channel depends on the distance between the transmitter and receiver. Also, the downlink code orthogonality is calculated individually for each link based on the channel properties. Detailed description on the implemented COST259 model can be found in [8]. C. Radio Network Algorithms Compared to previous system simulations [6], the models used in this study are more realistic radio network algorithms working on a frame level basis. When a new user is created by the traffic generation and is assigned an initial position by the mobility model, the gain is calculated according to (1). Cell selection and re-selection occurs for mobiles without a channel (i.e. for packet users that do not currently have a channel) in every radio frame. The cell with the highest gain is selected. Link admission control (AC) is turned off in order not to influence the capacity results. The soft handover algorithm is evaluated in every frame. It compares the path gain of the connected cell with the path gain for neighboring cells. If the difference is within a threshold, the neighboring cell is added as a new soft handover leg (connection). On the other hand, for existing multiple leg connections, if the threshold is outside the valid range, the leg will be deleted. The power control consists of the inner loop, outer loop and initial power settings. For each link there are upper and lower power limits. The inner loop power control and the fast fading act on slot level. After the slot loop, the C/I for each block is converted to block error probability (BLEP). Each block is then classified as erroneous or not, which gives the block error rate (BLER) estimates. The BLER estimates are used by the outer loop algorithm in order to decide if the SIR target should be increased or decreased. Some of the relevant radio network parameters are given in Table II. D. Traffic Models WCDMA networks must simultaneously fulfill the requirements of several traffic types which differ significantly. Beside speech, the dominant part of the carried data in radio access networks is expected to be www traffic. Audio/video streaming applications are also likely to be important in WCDMA networks. These traffic types are modeled in the simulations. The speech model used in our simulations is a Poisson distributed model. Based on modem-pool measurements a fourlayer model is used for simulating the behavior of www users.

S UMMARY OF R ADIO N ETWORK A LGORITHM PARAMETERS . N OTE THAT THE BASE STATION MAXIMUM POWER IS 20 W FOR THREE SECTOR SYSTEMS AND 80 W FOR FOUR FIXED BEAM SYSTEMS IN THIS STUDY.

Parameter Common Channel Power Base Station Maximum power Admission Control Limit Inner Loop Power Control step Outer Loop Power Control Max. of links in Active Set SHO Add threshold SHO Delete threshold

Value 4 20/80 Infinity 1 yes 3 2 4

Unit W W dB dB dB

A detailed description of the model and the performed measurements can be found in [11]. For simplicity, the traffic generated by users of streaming applications is modeled as a continuous stream of 64 kbps data with retransmission of erroneous frames. Table III summarizes the parameters of the user scenarios investigated. TABLE III S UMMARY OF

USER SCENARIO PARAMETERS .

Parameter Max. data rate [kbps] Spreading factor Max. DL power [W] BLER target RLC Retransmission

Speech 12.2 128 1 0.01 No

Stream 64 16 4 0.1 Yes

High 384 8 10 0.1 Yes

IV. S IMULATION C ASES The capacity of the fixed beam simulations are compared to the three sector case. The fixed beam case also employs a three sector system, but each sector is equipped with a four fixed beam antenna. The sector antenna and the 4 fixed multi beam antenna are shown in Fig. 1. Except for the antennas, the only difference between the sector and the fixed beam cases is the maximum base station transmit power, which is 20 W for the three sector case and 80 W for the fixed beam case. The pilot power setting is important and affects the system capacity. The power of the common channels is modeled so that the mobiles experience the same common channel power for both cases in order to achieve a fair comparison. Two different types of traffic are used for each case: speech only and a mixed traffic scenario consisting of 60% speech users, 20% streaming users and 20% high data rate users. V. R ESULTS In this section the capacity results for the different cases are shown. The results focus on the relative capacity gain and not the absolute figures. For purpose of simplicity and clarity, the capacity is defined here as kbps/cell at a certain mean BLEP percentage, called accepted quality. Table IV shows the accepted quality figures for each service. The capacity for each

Sector and Multi Beam Antennas

Speech 95 percentile of the BLEP comparison between fixed beam and sector case. Cell radius 1000m.

25 Sector Antenna Multi Beam Antenna

0.1 Fixed Beam case Three Sector case

20

0.09

0.08

15

0.07 95 percentile of the BLEP

Gain [dB]

10

5

0

0.06

0.05

0.04

0.03

−5

0.02

−10 0.01

−15

0

50

100

150

200 Degrees

250

300

350

Fig. 1. The sector and 4 elements multi beam antenna diagram. TABLE IV ACCEPTED QUALITY

0

400

0

200

400

600 800 Served speech traffic kbps/cell

1000

1200

1400

Fig. 3. The 95 percentile of the Speech BLEP for sector and fixed beam case in a speech only traffic scenario.

FOR THE DIFFERENT SERVICES .

Speech capacity comparison between Fixed Beam and Sector cases in a mixed traffic scenario. Cell radius 1000 m. 0.14

BLER Target 1% 10% 10%

Three Sector case Four Fixed Beam case

Accepted Quality 2% 16% 20%

0.12

service is found by running simulations with different traffic load. Thereafter the mean BLEP as a function of the load for each service can be determined. A. Speech only Traffic Scenario Fig. 2 shows the mean downlink BLEP as a function of the served speech traffic for both sector case and fixed beam case when the cell radius is 1000 m. Assuming a 2% mean BLEP as accepted quality, the relative capacity gain for the four fixed multi beam case is about 3.7 times compared to the three sector case. In order to validate the mean BLEP measurement, Fig. 3

DL mean BLEP of the speech user

Service Speech Streaming High Data Rate

0.1

0.08

0.06

0.04

0.02

0

0

20

40

60

80 100 120 Served speech traffic kbps/cell

140

160

180

200

Fig. 4. Speech capacity for sector and fixed beam case. Traffic scenario is mixed traffic and cell radius 1000 m.

Speech capacity comparison between Fixed Beam and Sector cases in a speech only user scenario. Cell radius 1000 m. 0.1

depicts the 95 percentile of the BLEP. The relative capacity gain is about 3.6 for the four fixed multi beam case compared to the three sector case when using 95 percentile of the BLEP. Thus, the 95 percentile of the BLEP shows very good resemblance with the mean BLEP.

Three Sector case Four Fixed Beam case 0.09

DL mean BLEP of the speech user

0.08

0.07

0.06

B. Mixed Traffic Scenario 0.05

0.04

0.03

0.02

0.01

0

0

200

400

600 800 Served speech traffic kbps/cell

1000

1200

1400

Fig. 2. Speech Capacity for sector and fixed beam case in a speech only traffic scenario.

This section presents some results for the traffic case with 60% speech user, 20% streaming user and 20% high data rate users with a cell radius of 1000 m. Fig. 4, 5 and 6 show the mean BLEP as a function of the served traffic for the speech, the high data rate and the streaming services respectively. The relative capacity gain at each service (assuming accepted quality level, see Table IV) is about 3.7, 2.9 and 3.0 times for the speech, high data rate and the streaming, respectively. Fig. 7 depicts the mean BLEP for all services and cases as a function of total served traffic for all services in kbps/cell. Assuming the accepted quality as in Table IV, the high data rate

Streaming capacity comparison between Fixed Beam and Sector cases in a mixed traffic scenario. Cell radius 1000 m.

HDR capacity comparison between Fixed Beam and Sector cases in a mixed traffic scenario. Cell radius 1000 m.

0.4

0.45

Three sector speech Fixed Beam speech Three Sector streaming Fixed Beam streaming Three Sector high data rate user Four Fixed Beam high data rate user

Three Sector case Four Fixed Beam case 0.35

0.3 0.35 0.25 DL mean BLEP

DL mean BLEP of the high data rate user

0.4

0.3

0.25

0.2

0.15

0.2

0.1

0.15

0.1

0.05

0 0

10

20

30

40 50 60 70 Served high data rate traffic kbps/cell

80

90

100

Fig. 5. High data rate capacity for sector and fixed beam case. Traffic scenario is mixed traffic and cell radius 1000 m.

0

100

200

300 400 Total Served Traffic kbps/cell

500

600

700

Fig. 7. Capacity of all services for sector and fixed beam case. Relative gain for the four fixed beam case compared to the three sector case for different cell radius. 7

Streaming capacity comparison between Fixed Beam and Sector cases in a mixed traffic scenario. Cell radius 1000 m.

speech streaming high data rate

0.26 Three Sector case Four Fixed Beam case

6.5

0.24

6

5.5 0.2

Relative Gain

DL mean BLEP of the streaming users

0.22

0.18

5

4.5

0.16

4 0.14

3.5 0.12

3 0.1

2.5 200 0.08

0

50

100

150 200 Served streaming traffic kbps/cell

250

300

400

600

800

1000 1200 Cell Radius [m]

1400

1600

1800

2000

350

Fig. 6. Streaming capacity for sector and fixed beam case. Traffic scenario is mixed traffic and cell radius 1000 m. user is clearly limiting the capacity for both sector and fixed beam case. One way of dealing with this problem is by utilizing radio resource management algorithms such as admission control. Policies how to employ the resource management in a mixed traffic environment can be found e.g. in [12, 13]. C. Capacity Gain as a Function of Cell Radius Fig. 8 shows the relative capacity gain for the four fixed beam case compared to the three sector case for different cell radius. For cell radius of 250, 500 and 1000 m, the systems are mostly interference limited, giving a gain of about 3 times. For cell radius of 2000 m, the system is more downlink power limited. Thus, the higher antenna gain for the fixed beam system also contribute to a fourfold capacity gain. D. Base Station Power Consumption In Fig. 9 the total mean power consumption of a base station is depicted for both the sector and the fixed beam case. One

Fig. 8. Relative capacity gain for the four fixed beam case compared to the three sector case for different cell radius. interesting thing can be noted here: In the case of mixed traffic, according to Fig. 9, the 4 fixed beam solution can handle more than twice the traffic amount compared to the sector case, even if the output power was limited to 20 W. However, this result should be interpreted with care since the radio network algorithms may work differently in a fully loaded system compared to a system with moderate load. VI. C ONCLUSIONS In this contribution the capacity has been studied for a WCDMA system where each sector has been equipped with a four fixed multi beam antenna. The capacity has been compared to that provided by an ordinary three sector site system. This study has focused on downlink capacity since it is expected to be the limiting link due to the high asymmetry of the data traffic. Simulations have been conducted by a system simulator containing detailed models of the radio network algorithms. The propagation model is based on the COST259 proposal which includes both temporal effects such as fast and

Total DL Tx Power per base station comparison between four fixed beam and sectorized system. Cell radius 1000 m. 80 Three Sector case Four Fixed Beam case

[7]

70

Mean DL Tx Power per BS

60

50

[8] 40

30

[9]

20

10

[10] 0

0

100

200

300 400 Total served traffic [kbps/cell]

500

600

700

Fig. 9. Total mean base station power consumption for sector and fixed beam case. [11] log-normal fading as well as the spatial behavior of the channel. We have compared two scenarios, one with speech users only and one with a mix of services (speech, streaming and high data rate users). The simulations show that for an interference limited system, the 4 fixed multi beam concept will provide 3 times the capacity compared to a standard sector system for all investigated service types. For downlink power limited systems scenarios even higher gains can be found. It should also be noted that the capacity gain obtained with adaptive antennas can be traded for better coverage, or a combination of the two. R EFERENCES [1]

A. F. Naguib, A. J. Paulraj, and T. Kailath, “Capacity Imrovement with Base Station Antenna Arrays in Cellular CDMA,” IEEE Trans. on Vehicular Technology, vol. 43, pp. 691–698, August 1994. [2] J. C. Liberti and T. S. Rappaport, Smart Antennas for Wireless Communications: IS-95 and Third Generation CDMA Applications, Prentice Hall, 1999. [3] J. H. Winters, “Optimum Combining in Digital Mobile Radio with Cochannel Interference,” IEEE Trans. on Vehicular Technology, vol. 33, no. 3, pp. 144–155, August 1984. [4] B. D. V. Veen and K. M. Buckley, “Beamforming: A Versatile Approach to Spatial Filtering,” IEEE ASSP Magazine, pp. 4–24, April 1988. [5] B. G¨oransson, B. Hagerman, and J. Barta, “Adaptive Antennas in WCDMA Systems - Link Level Simulation Results Based on Typical User Scenarios,” in Proceedings IEEE Vehicular Technology Conference, Fall, vol. 1, pp. 157–164, Boston, MA, USA, 2000. [6] B. G¨oransson, B. Hagerman, S. Petersson, and J. Sorelius, “Advanced Antenna Systems for WCDMA: Link and System Level Results,” in International Symposium

[12]

[13]

on Personal, Indoor and Mobile Radio Communications (PIMRC), pp. 62–66, London, UK, 2000. J. Barta, M. Ericson, B. G¨oransson, and B. Hagerman, “Interference Distributions in Mixed Service WCDMA Systems - Opportunities for Advanced Antenna Systems,” in Proceedings IEEE Vehicular Technology Conference, Spring, Rhodes, Greece, 2001. L. Correia, ed., Wireless Flexible Personalized Communications - COST 259 Final Report, John Wiley & Sons, 2001. E. Tiirola and J. Ylitalo, “Performance Evaluation of Fixed-Beam Beamforming in WCDMA Downlink,” in Proceedings IEEE Vehicular Technology Conference, Spring, vol. 2, pp. 700–704, Tokyo, Japan, 2000. K. I. Pedersen, P. E. Mogensen, and J. B. Andersen, “Vector RAKE Receivers for Third Generation Systems - An Evaluation,” in Sixth Annual Workshop on Smart Antennas in Wireless Mobile Communications, Stanford University, CA, USA, July 1999. ´ o, A. Vid´acs, J. Barta, Z. Kenesi, and T. Eltet˝ “Measurement-Based WWW User Model for Radio Access Networks,” in International Workshop on Mobile Multimedia Communications (MoMuC), Tokyo, Japan, October 2000. A. Furusk¨ar, P. de Bruin, C. Johansson, and A. Simonsson, “Mixed Service Management with QoS Control for GERAN - The GSM/EDGE Radio Access Network,” in 3G Wireless 2001 San Fransisco, CA, USA, June 2001. D. Imbeni and M. Karlsson, “Quality of Service Management for Mixed Services in WCDMA,” in Proceedings IEEE Vehicular Technology Conference, Spring, Tokyo, Japan, June 2000.