Sectoring of a Locally Centralized Communication S

Indoor Environment

Stefan Pettersson

Radio Communication Systems Laboratory

Dept. of Signals, Sensors and Systems

Royal Institute of Technology

SE-100 44 Stockholm, Sweden

PH: +46-60-14 85 81, FAX: +46-60-14 88 30

E-mail: stefan@ite.mh.se

Abstract – Future indoor communication systems will support different services with varying quality demands. Centralized radio resource management (RRM) is suitable for these types of systems. With central control the existing users can be protected and given priority over new allocations. The drawback is generally considered to be increased complexity. By dividing the coverage area into smaller segments using multiple controllers, the computational complexity can be reduced. The price for reduced complexity is lost capacity compared with a single controller system. In this paper we introduce sector antennas in a locally centralized radio communication system covering one building floor and investigate their ability to increase the capacity with maintained low complexity. The results show that the capacity can be almost as high for a sectorized system with two central units as for a system using one central unit covering the floor with omni-directional antennas. The results also indicate how well the studied wireless system can coexist with similar systems located in the vicinity. Sectoring reduces the transmitter powers and therefore improves the performance of closely located wireless networks.

I.Introduction

In the future, a variety of different services will be provided through wireless networks. The services now available on the Internet, such as WWW-browsing, electronic mail, file transfer, audio and video services will be offered to mobile users and not only to those connected to a fixed network. The demand for wireless access to these services comes mainly from the success of cellular telephony. The freedom that users have gained from cellular telephony is irreversible and will continue to grow. In office environments, even more services that uses wireless technologies will emerge. Laptop computers will have high-speed wireless access to the fixed network in the building. We want to put the computers where the people are and not where the fixed network is located. Different wireless services require different Quality of Service (QoS), such as delay, data rates and error rate. Differences in QoS must be handled in multi-service communication systems and centralized systems have been suggested as a solution. One drawback with centralized RRM is the computational complexity. In [1], a locally centralized (bunched) RRM concept is presented. The system complexity can be reduced by covering the area with multiple central controllers. The price for this reduction is lost capacity, especially in an indoor environment [2]. This paper investigates the use of sector antennas to improve the system capacity of the locally centralized communication system in an indoor office scenario. Several indoor networks must also be able to coexist close to each other. A wireless network located in a tall building can easily be heard a long distance. In a business complex, many companies may use networks operating in the same frequency band. In this paper, we will investigate two systems located on the same building floor to study the possibility of coexistence. We will look at the difference in performance between sector and omni-directional antennas.

Sector antennas have shown to perform well in indoor scenarios [3], [4], [5]. The systems in [3] and [4] have architectures with only one base station and do not consider co-channel interference. The sector antennas in [5] are constrained to transmit on one sector at a time. Our paper studies a dense indoor architecture with the ability to transmit on every sector simultaneously on different channels.

II.System Description

The locally centralized RRM concept from [1] consists of a Central Unit (CU) connected to a cluster of Remote Antenna Units (RAUs) as shown in Figure 1.

Figure 1. Remote Antenna Units connected to a Central Unit.

The resource management is performed centralized internally and in a decentralized manner externally.

To perform efficient RRM within the system, the CU uses central knowledge about the system that has been gathered from measurements done by either the terminals on beacon channels or the RAUs on terminal transmissions. The main idea is to make a feasibility check prior to an allocation of a new resource request.The allocation is performed only if the quality of the already allocated users can be maintained. The CU knows the link gains between the user terminals and the RAUs. The power control is performed centrally so the CU also knows the transmitter powers. The impact of the new allocation will have on other users can then be calculated in advance, prior to the allocation.

A more detailed explanation of the RRM concept can be found in [1] and [2].

III. Models and Performance Measures

A simple indoor office environment is used for the performance evaluation [4]. A building floor consists of offices and corridors. The offices are 10 by 10 by 3meters and the corridors are five meters wide. An RAU is placed in every second office as can be seen in Figure 2. The RAUs are placed in the ceiling and all terminals on 1.5 meters above floor level.

Figure 2. The building floor with 20 RAUs.The user terminals are considered static (no mobility)during an iteration and they are distributed evenly over the area with the exception that the probability is 0.85 of being located in the office and 0.15 in the corridor.The path loss model is defined as:L [dB] = 37 + 30Log 10(R) + X σ,

where R is the transmitter-receiver separation given in

meters and X σ is the shadow fading with log-normal distribution having a standard deviation of 12 dB. We also assume perfect knowledge of the link gain matrix,i.e. all the path losses between the base stations and mobiles are known.

The relative load is defined as the fraction of requested channels in a cell covered by one RAU and we define capacity as the load that results in 2% assignment failure rate. The assignment failure rate is the fraction of users that did not get a feasible channel (blocking) or got a

channel that had to low quality (outage) to meet the specified SIR-target.

Another measure is the CDF of the transmitter powers ,where CDF is the Cumulative Distribution Function.From the distribution, we can for example see the power control range that is needed to maintain the capacity.Computational complexity is represented by floating point operations (flops) per allocation attempt.

We will use an ideal sector pattern for our study with sector antennas, i.e. non-overlapping as in Figure 3. This is to separate the effect that sectoring has on the system behavior and performance from special antenna patterns.

Figure 3. Non-overlapping antenna pattern.The antenna gain is normalized to zero dB for the main lobe and the front-to-back ratio is set to 15 dB. In for example a four-sector system, we will have four possible links from a mobile to a site with sector antennas. In our simulations, these links are independent with different shadow fading and treated as if transmitted from different RAUs located at the same position. The difference is that the load is related to the whole site, i.e.there is an equal number of users in systems both with and without sectoring at a specific load.

The channels are divided between the sectors into exclusive channel groups at a site. This means that if the number of sectors is doubled, the number of available channels in every sector is reduced by a factor of two.

IV. Simulation Description

As a basis for the performance evaluations lie snapshot simulations where only the downlink is considered. For every simulation, the minimum number of users is set to 30000 and the number of channels available at every RAU is 48. At a load of 0.5, on average 48⋅0.5=24 users are requesting a channel on every RAU. On a floor with 20 RAUs we have 20⋅24=480 users. To meet the minimum 30000 users that are required for every simulation, 7 iterations are used for the load 0.5. All iterations are considered uncorrelated, i.e. the users are regenerated and distributed over the building with new link gains calculated. There is no correlation between the different links in the system and we consider the path loss from one RAU to one user to be the same for all possible channels.

The users are assigned the strongest RAU and they are requesting one channel each. The channel requests are processed one by one until there are no requests left and the channel selection is made randomly . One by one, the available channels are tested for feasibility at the

V.Simulation Results

Two different scenarios have been investigated with computer simulations. One where one bunch is covering the building floor, i.e. one central unit is used and responsible for the channel allocations requested by all users on the floor. Another scenario has two bunches covering half the floor each. The ten leftmost RAUs are connected to one CU and the remaining ten RAUs are connected to another CU. The two central units are considered independent without any communication between them. By comparing the two cases with either one or two central units, we will see if the complexity of the central system can be reduced by splitting the covered area into smaller regions controlled by multiple CUs. We can also draw conclusions on how the performance is affected when two separate systems operate in a nearby area. This will very likely happen in future business environments with different companies using wireless indoor networks working in the same frequency spectrum.

In Figure 4, the performance for a system with one CU covering the whole building floor is shown. The capacity increases from 0.50 to 0.65 when a two-sector antenna is used compared with an omni-directional antenna (Nsec=1). The capacity is defined as the load that can be carried with less than 2 percent assignment failure.

With increased number of sectors the capacity is decreased in this case, which is without external interference. The trunking loss is larger than the gain from less internal co-channel interference.

A result of the reduced internal co-channel interference can be seen in Figure 5. The CDF of the transmitter powers in the downlink are displayed for the load 0.60. We can clearly see the effect the sector antennas have on the transmitter power distribution. In average, the number of co-channel users is reduced if the number of sectors is increased. The used power control is SIR-based and since there is no channel reuse between the sectors at one site, the power range is dramatically reduced. Without sector antennas, 30 percent of the users are transmitting with minimum power. The same figure reaches almost 90 percent with a twelve-sector system.The necessary dynamic range of the power control is made significantly smaller with the use of sector antennas in our centralized communication system.

The reduced transmitter powers indicate that the performance can be improved with sectoring if there are external interference present. Uncontrolled external interference will arise if the area is covered with multiple controllers to lower the computational complexity or if several independent networks are operating close to each other. The result from [2], shows that the complexity can be reduced with multiple controllers but the price is lost capacity. In Figure 6, we see that the complexity is reduced with sector antennas. This is not so surprising considering that the number of co-channel users is reduced and therefore the feasibility check is less complex to perform. The reduced interference speeds up the process in finding a channel that can meet the SIR targets.

Looking at Figure 4 and Figure 6, we find that the complexity is slightly less for two-sector antennas with the assignment failure maintained at 2 percent compared with omni-directional antennas. Increasing the number

Figure 4. Capacity with one central unit controlling the building floor. Nsec equals the number of sectors. Figure 5. CDF of TX-power with one central unit.

of sectors, the complexity drops further. With six sectors, the capacity is larger than the non-sectored system and the number of flops necessary for an allocation is close to ten times lower. Sectoring of the centralized system reduces the allocation efforts with fewer co-channel interferer.

To lower the complexity even more, multiple CUs can be used to cover the same area. More CUs mean fewer RAUs per CU. This will reduce the number of co-channel users within the coverage area of one CU and therefore the computational complexity as well. The drawback is of course the generated and uncontrolled interference from areas covered by other CUs. In Figure 7, this phenomenon is shown. When two central units cover the building floor, the capacity drops from 0.50 to 0.12 without sector antennas. The capacity loss is five times for this method of complexity reduction.

With sector antennas, the capacity is greatly improved. A six-sector system improves the capacity from 0.12 to 0.46. This is close to the capacity that a single controller system can achieve with omni-directional antennas,shown previously in Figure 4. We also see a maximum in performance for six sectors. The capacity drops for twelve sectors due to lack of channels. The number of channels is only four per sector in this case and the risk of blocking is too high. Figure 8 shows the distribution of SIR for different number of sectors at the load 0.40.More sectors improve the overall SIR situation,primarily due to less external interference generated from uncontrolled sources outside the coverage area of the CU.

The feasibility check relies on measurements on beacon channels transmitted from RAUs connected to the CU,i.e. no external interference is taken into account in the decision whether the channel is feasible or not. To handle the unavoidable collisions that occur, the power control is run for five iterations after all allocation requests are handled and the limit for success is set to a SIR of 9 dB. In Figure 8, we see that for twelve sectors,the probability of having a SIR less than 9 dB is below 2⋅10-3 for a user. The assignment failure for twelve sectors at the load 0.40 is more than 2⋅10-2. The

difference in performance is the result of blocking, i.e.lack of channels to assign.

The computational complexity for the system with two central units is shown in Figure 9. We can see that the number of flops per allocation attempt is about the same for different number of sectors. The reason is that the smaller coverage area makes the number of co-channel users smaller and a feasible channel is often found on the first attempt when the load is not too high.

If the assignment failure is kept constant at 2 percent, we can compare the use of sector antennas with omni-directional antennas. In Table 1, the scenario with a single central unit is presented. The capacity gains and the complexity reductions are calculated from the simulations shown in Figure 4, 6,7 and 9. The capacity for the single CU case is 0.50 and the number of flops needed per allocation attempt is 2940 at that load. With a two-sector antenna, the capacity increases to 0.65 and the complexity drops to 1610. The capacity gain and complexity reduction is 1.30 and 1.8 in this comparison.Table 2 presents comparisons between the single CU

Figure 6. Complexity with one central unit.Figure 7. Capacity with two central units controlling half the building floor each.

Figure 8. CDF of SIR with two central units.

The conclusion from the tables is that the complexity can be reduced with the use of sector antennas. The capacity gain is also significant in a single controller scenario. As much as 30 percent higher load can be offered with lower computational complexity. A more efficient way of reducing complexity is to have multiple controllers covering the area.

VI.Summary & Conclusions

In this paper, we have investigated the use of sector antennas in a locally centralized indoor communication system. The system uses central knowledge about the current propagation situation to calculate the impact that a new allocation will have on already admitted users prior to the allocation. Local centralization is one way of handling complexity in future indoor multimedia communication systems with QoS guarantee.

The results from the investigation shows that sector antennas increase the capacity and lower the complexity in the studied communication system. Sectoring gives us an opportunity to lower the system complexity by covering the area with multiple central units. Since the use of sector antennas can reduce the transmitter power and lower the produced external interference, it can be used for reducing the system complexity. Without loosing too much in capacity, the number of floating point operations needed for allocations can be reduced with a factor of 16.

The results also show that sector antennas are suitable when several wireless networks operate close to each other. The load that can be offered increases from 0.12 to 0.46 with six-sector antennas when two systems operate on the same building floor compared with the use of omni-directional antennas.

VII.References

[1]M. Berg, S. Pettersson, and J. Zander, ”A Radio

Resource Management Concept for ‘Bunched’Personal Communication Systems”, Multiaccess, Mobility and Teletraffic for Personal Communications Workshop, MMT’97, Melbourne, Dec 1997.

[2]S. Pettersson, “A Comparison of Radio Resource

Management Strategies in Bunched Systems for Indoor Communication”, VTC’99, Houston, May 1999.

[3]G. Yang, K. Pahlavan, and T. J. Holt, “Sector

Antenna and DFE Modms for High Speed Indoor Radio Communications”, IEEE Trans. On Vehicular Technology, Vol 43, No. 4, Nov 1994.

[4] A. S. Macedo, and E. S. Sousa, “Antenna-Sector

Time-Division Multiple Access for Broadband Indoor Wireless Systems”, IEEE Journal on Selected Areas in Communication, pp. 937-952, Vol 16, No 6, Aug 1998.

[5] A. S. Mahmoud, D. D. Falconer, and S. A.

Mahmoud, “A Multiple Access Scheme for Wireless Access to a Broadband ATM LAN Based on Polling and Sectored Antennas”, Personal, Indoor and Mobile Radio Communications 1995, PIMRC'95, Vol 3, pp. 1047-1051, Sep 1995. [6]Universal Mobile Telecommunications System

(UMTS); Selection procedures for the choice of radio transmission technologies of the UMTS, UMTS Technical Report 30.03, 1997.

Figure 9. Computational complexity with two central units controlling the building floor.

Table 1. Complexity reduction and capacity gain for a single CU system compared with the omni-directional case.

Number of sectors Capacity gain Complexity

reduction

20.65/0.50 = 1.302940/1610 = 1.8

40.58/0.50 = 1.162940 / 600 = 4.9

60.54/0.50 = 1.082940 / 440 = 6.7

120.39/0.50 = 0.782940 / 300 = 9.8 Table 2. Complexity reduction and capacity gain for a double CU system compared with a single CU system using omni-directional antennas.

Number of sectors Capacity gain Complexity

reduction

10.12/0.50 = 0.242940 / 90 = 33 20.25/0.50 = 0.502940/120 = 25 40.38/0.50 = 0.762940/160 = 18 60.46/0.50 = 0.922940/180 = 16 120.37/0.50 = 0.742940/170 = 17