A Novel Decentralized Time Slot Allocation Algorithm in Dynamic TDD System Young Sil Choi Email: choiys@mobile.snu.ac.kr Illsoo Sohn Email: sohnis@mobile.snu.ac.kr Kwang Bok Lee Email: klee@snu.ac.kr Abstract This paper proposes the decentralized time slot allocation (DTSA) algorithm in dynamic TDD (D- TDD) system to support traffic asymmetry. In D-TDD systems, additional interferences are introduced when one cell is uplink (downlink) and the other cell is downlink (uplink). These kinds of interferences result in severe performance degradation of SIR, which occurs especially to users in cell edge than users in cell center. To reduce additional interferences, DTSA algorithm is applied to D-TDD system. This algorithm assumes decentralized cellular network, so, it operates independently in each cell and has lower complexity than the other systems. Using DTSA algorithm, the outage probability of uplink and downlink at 0dB are decreased about 4.3% and 8.5% compared to conventional D-TDD system, respectively. Also DTSA algorithm increases the system capacity compared to S-TDD system and conventional D-TDD system. Simulation results show that D-TDD system with DTSA algorithm can support more asymmetric traffic than other systems, which increase system capacity. 1. Introduction In present cellular systems, frequency division duplex (FDD) has been used because it is suitable for voice services. As FDD systems use different frequency bandwidth in uplink and downlink, there are no additional interferences. However, future cellular systems are expected to provide various multimedia services which result in traffic asymmetry between uplink and downlink. To support traffic asymmetry, time division duplex (TDD) is considered as a good candidate. TDD systems have many advantages related to its ability to support traffic asymmetry and to use reciprocity of channel, etc. TDD systems are classified into static TDD (S-TDD) systems and This work was supported in part by the National Research Laboratory Program of Korea. <Cell 1> <Cell 2> Base 1 Base 2 Mobile a Mobile b BS to BS interference MS to MS interference Figure 1. Crossed slot interference scenario dynamic TDD (D-TDD) systems. S-TDD systems have the same ratio of allocated slots for uplink traffic and downlink traffic in every cell, while D-TDD systems adjust number of slots allocated to uplink and downlink dynamically based on traffic asymmetry of each cell. In FDD systems, traffic asymmetry can be supported by adjusting allocated bandwidth to both links, which is not feasible in practical systems. On the other hand, TDD systems can support traffic asymmetry by adjusting time allocation of uplink and downlink. Especially, D-TDD system can support traffic asymmetry more efficiently by allocating time slots of uplink and downlink dynamically [1][2]. However, D-TDD systems introduce additional interferences as shown in Fig.1 which don t exist in FDD systems and S-TDD systems [1]. It occurs when one cell is uplink (downlink) and another adjacent cell is downlink (uplink). If we adjust number of slots allocated to uplink and downlink dynamically based on traffic asymmetry, these kinds of interferences affect to the system performance severely [3]. We can t get advantages of D-TDD systems compared to FDD systems without solving the additional interference problem in D-TDD system. To reduce additional interferences and to support traffic asymmetry in D-TDD systems, hybrid FDD and TDD schemes have been considered. One of them 1-4244-0086-4/06/$20.00 (C) 2006 IEEE
Frequency Outer Threshold Asymmetry Boundary Outer /outer Downlink Cell 1 (1:2) r[m] time A B C D Frequency MS outer cell MS inner cell Outer /outer Downlink Cell 2 (1:1) Outer Threshold Asymmetry Boundary 1 Frame Figure 2. Frame structure of DTSA algorithm applies TDD to the small areas that require asymmetric traffic. These small areas such as offices or public buildings are referred to as hot spots. And FDD system is used in the rest of areas [4]. Though hybrid scheme has many benefits, their efficiency is decreased because they divide total bandwidth into two groups, one is for FDD and the other is for TDD. Practical FDD and TDD coexistence scheme is suggested in [5], which also divides total bandwidth into two groups. In FDD mode, one is used for FDD uplink and the other is used for FDD downlink. But, if downlink traffic increases, FDD uplink bandwidth operates to TDD mode. By applying this scheme, traffic asymmetry can be supported, but this system also has additional interference problem. To use limited resources more efficiently, many algorithms for D-TDD systems have been suggested. In [6], numerical results show the possibility that D- TDD systems can achieve better capacity than S-TDD systems. Accordingly, many schemes are proposed for D-TDD systems. For example, both directional antennas and time slot allocation algorithm are applied for avoiding additional interferences in D-TDD systems [7]. But they assume centralized cellular network which means that the information of antenna direction of adjacent cells at a point of time is shared in every cell, which require the signaling overhead between cells. This scheme suggests good ideas for avoiding additional interferences, but signaling overhead between cells increases system complexity. In this paper, we propose practical D-TDD system with decentralized time slot allocation (DTSA) algorithm. Proposed system assumes decentralized cellular network which means that there is no signaling overhead for exchanging any information between cells. Then, we investigate the signal to interference power ratio (SIR) performance and the time UL for MS outer cell UL for MS inner cell Outer uplink threshold Asymmety boundary capacity of the proposed system, and compare the performances of S-TDD system, conventional D-TDD system, and proposed D-TDD system with DTSA algorithm. In this paper, D-TDD systems without DTSA algorithm are referred to as conventional D- TDD systems. The remainder of this paper is organized as follows. The next section explains the additional interferences specifically, and the proposed DTSA algorithm in D-TDD system is introduced in section III. In Section IV, we present simulation results about SIR performance and capacity of three systems. Finally, we conclude this paper in Section V. 2. Additional interferences Fig. 1 shows the additional interference scenario. There are two types of additional interferences, mobile station to mobile station (MS-to-MS) interference and base station to base stations (BS-to-BS) interference. In MS-to-MS interference case, downlink signals received by mobile stations in cell 1 are interfered by uplink signals transmitted by mobile stations in cell 2, and larger power of MS-to-MS interference is introduced to the mobile stations in cell 1 as mobile stations in each cell is located closer. signals received by base station 2 are also interfered by downlink signals transmitted by base station 1, which is BS-to-BS interference. Although BS-to-BS interference power doesn t change according to the location of uplink mobile station, SIR becomes smaller as uplink mobile station and base station in cell 2 is located far. If uplink mobile stations are located in cell edge, they experience severe BS-to-BS interference in cell 2 or introduce severe MS-to-MS interference to mobile stations in cell 1. Thus we need to control the mobile stations in cell edge for reducing two kinds of additional interferences. DL Figure 3. DTSA algorithm
r[m] <Cell 1> Mobile 3. Decentralized time slot allocation algorithm In this paper, we propose practical D-TDD system with DTSA algorithm to reduce additional interferences by controlling the mobile stations located in cell edge. Fig. 2 and Fig. 3 show the frame structure and proposed DTSA algorithm, respectively. Proposed scheme assumes decentralized cellular network, so it is implemented independently in each cell. In addition, we assume that the starting point of frame is synchronized in every cell. We consider OFDM- TDMA system, and assume that traffic asymmetry of cell 1 and cell 2 are (1UL:2DL) and (1UL:1DL), respectively. The point at which uplink slots are changed to downlink slots is referred to as asymmetry boundary as shown in Fig. 2. Because asymmetry boundary of each cell is different in D-TDD systems, the slots in period C exist, and these slots introduce the additional interferences. The slots in period C and additional interferences introduced due to them are referred to as crossed slots and crossed slot interferences, respectively. In Fig. 3, we divided cell into two regions. The region inside r is defined as inner cell, and the region outside r is defined as outer cell. The mobile stations located in outer cell are main concern of DTSA algorithm. 3.1. MS-to-MS interference case <Cell 2> Figure 4. Crossed slot interference scenario in D-TDD system with DTSA algorithm Because mobile stations in outer cell introduce stronger MS-to-MS interference than mobile stations in inner cell, mobile stations in outer cell are allocated to the former slots of the frame which are far from the crossed slots as shown in Fig. 3. We limit the time slots allocated to the mobile stations in outer cell until outer uplink threshold, which prevent that the MS-to- Mobile Base 1 Base 2 BS to BS interference MS to MS interference Table 1. Simulation parameters Parameter Value Number of interfering cells 18 Cell radius 1000 m outer uplink radius, r [m] 500 m 3 (BS-to-BS), Path loss coefficient 4 (MS-to-MS and BS-to-MS) Standard deviation of Log normal fading 8 db Number of slots/frame 60 slots/frame Outer uplink threshold 10th slot Max. tx power of mobile station 200 mw Tx power of base station 10 W Asymmetry of Static TDD 1 UL : 2 DL MS interference is introduced to uplink mobile stations in outer cell. The outer uplink threshold and r are set to the same value in every cell. In Fig. 2, only time slots in period A are assigned for the uplink mobile stations in outer cell, and time slots in period B are assigned for the uplink mobile stations in inner cell. Slots in period A can t be allocated to the mobile station in outer cell if it exceeds the outer uplink threshold. Some uplink mobile stations in outer cell, which are not selected, are dropped. By applying DTSA algorithm, the MS-to-MS interference scenario is changed from Fig. 1 to Fig. 4. mobile stations in the adjacent cell which introduce MS-to- MS interference are limited to the mobile stations in inner cell, which increases the distance between mobile stations in each cell. Accordingly, MS-to-MS interference decreased in crossed slots compared to conventional D-TDD systems, and reduction of MSto-MS interference improves SIR performance of conventional D-TDD system. 3.2. BS-to-BS interference case The BS-to-BS interference scenario is also changed from Fig. 1 to Fig. 4 by applying DTSA algorithm. Although BS-to-BS interference power doesn t change according to the location of uplink mobile station and is the same with conventional D-TDD systems, SIR of the proposed system is increased. Because only uplink mobile stations which are located in inner cell are allocated to the crossed slots, the received uplink signal power in center base station is stronger than that in conventional D-TDD systems. Thus, the SIR performance of the proposed system is improved than that of conventional D-TDD systems by increasing the received signal power with the same interference power.
Figure 5. SIR performance Figure 7. Capacity comparison 4.2. Simulation results Figure 6. Downlink SIR performance 4. Simulation results 4.1. Simulation configurations SIR performance and capacity of S-TDD system, conventional D-TDD system, and D-TDD system with DTSA algorithm are evaluated. We assume hexagonal cells with radius of 1km, frequency reuse factor of one, and interfering cells of 2-tier. Each TDD frame is divided into 60 time slots, and asymmetry boundary of S-TDD system is set to (1UL:2DL). Outer uplink threshold and r is fixed to 10th slot and 500m, respectively. Log-normal fading with standard deviation of 8dB is considered. Users are uniformly distributed in every cell and the maximum transmit power of all mobile stations is 200mW and the transmit power of base station is fixed to 10W. We separate the types of traffic to voice traffic and data traffic, and each mobile station selects the type of traffic randomly. Simulation parameters are listed in Table 1. 4.2.1. SIR performance in uplink case. Fig. 5 and Fig. 6 show the uplink and downlink SIR performance, respectively. S-TDD system has better SIR performance than the other systems which are degraded by BS-to-BS interference. For example, when the probability below 0dB of SIR is defined as outage probability, outage probability of S-TDD system is 7.8% less than that of conventional D-TDD system. Outage probability of conventional D-TDD system is improved about 4.3% by applying DTSA algorithm because only uplink mobile stations in inner cell experience BS-to-BS interference. 4.2.2. SIR performance in downlink case. S-TDD system also has 1.6% less outage probability than that of conventional D-TDD system. Proposed system improves SIR performance of conventional D-TDD system by reducing the effect of MS-to-MS interference. Outage probability of the proposed system is about 8.5% less than that of conventional D- TDD system and 7.6% less than that of S-TDD system at 0dB. It is because that the interfering sources of MS-to-MS interference in conventional D-TDD system are limited to mobile stations in inner cell by applying DTSA algorithm. In addition, the received MS-to-MS interference power introduced in the proposed system is weak compared to the received interference power from base stations in S-TDD system because maximum transmitting power of mobile stations is 17dB smaller than that of base stations. For these reasons, proposed system shows better SIR performance than those in other systems. 4.2.3. Comparison of capacity. We also evaluate capacity. The number of slots allocated to uplink and downlink mobile stations per
frame is defined as total used slots, and the slots that have the SIR below 0dB is defined as outage slots. Capacity is defined as the number of supported slots which are calculated by subtracting the number of outage slots from the number of total used slots per frame. Fig.7 illustrates the capacity. When R down /R up = 2 which is the same with asymmetry boundary of S-TDD system, S-TDD system has the maximum capacity because it can utilize all the slots in a frame. Except R down /R up = 2, the capacity of S-TDD system decreases as traffic asymmetry increases. In contrast, D-TDD systems including proposed system have flexibility to adjust the asymmetry boundary while asymmetry boundary is fixed in S-TDD system. In other word, the number of total used slots of these systems is larger than that of S-TDD system. So, S-TDD system has smaller capacity than conventional D-TDD system in spite of having less outage slots. On the other hand, the number of total used slots of D-TDD system with DTSA is comparable with conventional D-TDD system, and D-TDD system with DTSA has less outage slots than that of conventional D-TDD system. Accordingly, D-TDD system with DTSA has the largest capacity among three systems as traffic asymmetry increases. IEEE Wireless Communications, vol. 4, pp. 51-56, Apr. 1997. [3] G. J. R. Povey, Frequency and time division duplex techniques for CDMA cellular radio, in Proc. IEEE Int. Symp. Spread Spectrum Techniques and Applications (ISSSTA), pp. 309-313, July 1994. [4] G. J. R. Povey, H. Holma, and A. Toskala, Hybrid FDD/TDD-CDMA for third generation cellular systems, IEE Colloquium on CDMA Techniques and Applications for Third Generation Mobile Systems, pp. 2/1-2/6, May 1997. [5] K. Osaki, D. Minamihira, H. Furukawa, and Y. Akaiwa, An FDD and TDD Coexistence Scheme for Imbalanced Traffic Compensation, IEEE Vehicular Technology Conference (VTC), June 2005. [6] W. S. Jeon, and D. G. Jeong, Comparison of time slot allocation strategies for CDMA/TDD systems, IEEE Journal on Selected Areas in Communications, vol. 18, pp. 1271-1278, July 2000. [7] W. Jeong, and M. Kavehrad, Cochannel interference reduction in dynamic-tdd fixed wireless applications, using time slot allocation algorithms, IEEE Transactions on Communications, vol. 50, pp. 1627-1636, Oct. 2002. 5. Conclusions A DTSA algorithm in D-TDD system is proposed in this paper. We compare the SIR performance and capacity of S-TDD system, conventional D-TDD system, and D-TDD system with DTSA algorithm. S- TDD shows the better SIR performance than conventional D-TDD system in both uplink and downlink. On the other hand, conventional D-TDD system has flexibility to adjust time slots allocation according to traffic asymmetry. The proposed scheme also has flexibility, and improves the SIR performance of conventional D-TDD system. For these reasons, the proposed system has the largest capacity among three systems as traffic asymmetry increases. Furthermore, the proposed scheme is suitable for practical systems because it operates independently in each cell without signaling overhead between cells. References [1] G. J. R. Povey, and M. Nakagawa, A Review of Time Division Duplex CDMA Techniques, in Proc. IEEE Int. Symp. Spread Spectrum Techniques and Applications (ISSSTA), vol. 2, pp. 630-633, Sept. 1998. [2] R. Esmailzadeh, M. Nakagawa, and E. A. Sourour, Time-division duplex CDMA communications,