結合動態通道調配機制和天線波束形成技術以支援包含非對稱性傳輸的分時雙工分碼多工存取系統

(1)

國立交通大學

電信工程研究所

碩士論文

結合動態通道調配機制和天線波束形成技術

以支援包含非對稱性傳輸的

分時雙工分碼多工存取系統

Joint Dynamical Channel Assignments and

Antenna Beamforming for the TDD/CDMA

Systems with Asymmetric Traffic

研究生：陳奕丞

(2)

Joint Dynamical Channel Assignments and

Antenna Beamforming for the TDD/CDMA

Systems with Asymmetric Traffic

A THESIS

Presented to

The Academic Faculty

By

Yi-Cheng Chen

In Partial Fulfillment

of the Requirements for the Degree of Master in Communication Engineering

Department of Communication Engineering National Chiao-Tung University

June, 2004

(3)

摘要

本論文主要目的在於希望能夠藉由適當地運用時間與空間二個

不同維度上的無線資源，以期在分時雙工（Time Division Duplex,

TDD）分碼多工存取（Code Division Multiple Access, CDMA）系統

下充分地支援非對稱性的資料傳輸，並同時提高整體系統效能。隨著

非對稱性傳輸需求的日益增加，分時雙工分碼多工存取系統在未來的

無線網路中將扮演著一個重要的角色。然而，由於分時雙工分碼多工

存取系統中的細胞都使用相同的頻帶來上傳和下載資料，因此非對稱

性的資料傳輸將會導致兩個彼此相鄰但是傳輸方向相反的基地台間

產生極大的交錯時槽干擾（Cross-Slot Interference）

。許多的研究

顯示，交錯時槽干擾不但嚴重地影響系統效能，並造成龐大無線資源

的浪費。

為了解決交錯時槽干擾的問題並增進整體系統性能，我們致力於

研究關於各種動態通道調配（Dynamic Channel Assignment, DCA）

機制以及天線波束形成技術（Antenna Beamforming）的特點。在前

(4)

時雙工分碼多工存取系統的影響。在指向性天線的幫助下，三區域的

細胞系統將會由三個不同基地台的相鄰區域來形成一個虛擬細胞

（virtual cell）

。我們發現在這種細胞架構下，交錯時槽干擾將會

被限制在一個虛擬細胞中，因此，鏈接相對性動態通道調配機制能夠

專注在虛擬細胞中藉由使用者的無線鏈結品質來做時槽的配置，以達

到充分降低交錯時槽干擾的需求。許多結果都顯示鏈接相對性動態通

道調配機制能夠顯著的勝過其他動態通道調配機制，並能提供分時雙

工分碼多工存取系統一個更有效率的資源配置法。

然而，我們發現大部分的動態通道調配機制，包括鏈接相對性動

態通道調配，都無法有效的解決上傳時的交錯時槽干擾。為了更進一

步解決這種交錯時槽干擾，我們更進一步提出了一個結合智慧型天線

波束形成技術的交錯時槽干擾為主的動態通道調配機制（Cross-Slot

Interference-Based DCA）

。我們所提出的交錯時槽干擾為主的動態

通道調配機制，主要希望能夠降低下載時的交錯時槽干擾，並利用細

胞各自分散的方式調整上傳和下載的時槽數目，以其在個別細胞中充

分地支援非對稱性的資料傳輸。智慧型天線的波束形成技術，在這邊

將被用來對付上傳時嚴重的交錯時槽干擾。我們實驗的結果顯示所提

出的交錯時槽干擾為主的動態通道調配機制能夠充分地壓制交錯時

(5)

槽干擾的影響，進而使得分時雙工分碼多工存取系統能夠充分地滿足

不同細胞間對於非對稱資料傳輸的個別需求，並同時能達到更高更好

的系統效能。

(6)

Summary

The key idea of this thesis is to exploit two different dimensions of radio resources -time and space - to support the diverse asymmetric traffic services in the -time division duplex/code division multiple access (TDD/CDMA) systems. Since the requirement for the asymmetric data services is growing, the TDD/CDMA system has been con-sidered an important wireless network in the future. However, different asymmetric traffic loads among cells may cause the heavy cross-slot interference, which can seri-ously degrade the system performance.

To alleviate the impact of cross-slot interference, we investigate the different dynamic channel assignment (DCA) and advanced antenna techniques. At first, we propose a novel link-proportional dynamical channel assignment (LP-DCA) scheme combined with tri-sector directional antennas to alleviate the impact of cross-slot interference for the TDD/CDMA systems. With the help of directional antennas, the tri-sector cellular system can form a virtual cell, which is composed of three sectors from three different base stations. For this kind of cellular structure, the cross-slot interference is restricted within a virtual cell. Thus the proposed LP-DCA scheme can concentrate on combating the cross-slot interference within a virtual cell by assigning time slots to the users according to their radio link quality. Our numerical results show that LP-DCA outperforms than other DCA algorithms and can more flexibly allocate resource the TDD/CDMA systems with asymmetric traffic services.

Nevertheless, most DCA algorithms including LP-DCA can not effectively al-leviate the uplink base-to-base cross-slot interference. To further reduce this kind of base-to-base cross-slot interference, we propose a cross-slot interference-based dy-namic channel assignment algorithm incorporated with antenna beamforming tech-niques. The proposed cross-slot interference-based DCA algorithm aims to reduce downlink mobile-to-mobile cross-slot interference and distributedly assign downlink and uplink time slots to support asymmetric traffic services in each cell. As for

(7)

an-tenna beamforming techniques, it is used here are mainly to avoid the impact of heavy uplink base-to-base cross-slot interference. Our numerical results show that synergy of combining the cross-slot interference-based DCA algorithm and antenna beamforming can effectively suppress both the mobile-to-mobile and base-to-base cross-slot interference in both downlink and uplink, respectively, thereby enabling a TDD/CDMA system to flexibly provide various asymmetric traffic loads in different cells.

(8)

Acknowledgments

I would like to thank my family who always support me with endless love. I especially thank Dr. Li-Chun Wang who provides me lots of his insights into important research problems, encouragement, and support. This work could not have been done without his advice, guidance, and comments.

I am also pleased to acknowledge the support of the Wireless Network Labo-ratory (WN Lab.) at the Department of Communications in National Chiao-Tung University.

I am grateful to thank my laboratory mates Chiung-Jang, Chih-Wen, Ander-son, Wei-Cheng, Chang-Long, Jian-Hua, Kuan-Jin, Ssonic, Bose, Shu-Yi, Ya-Wen, Tristone, Jun, Tom, Hyper, Kuang-nan, Halliday, and our best foreign friend and labmate - Assane at WN Lab for sharing many ideas and much happiness with me.

(9)

viii

List of Tables

2.1 Comparison of UTRA FDD and TDD physical key parameters. . . 11

3.1 System Parameters . . . 26

3.2 Inter-cell Interference Analysis in uplink. . . 27

3.3 Inter-cell Interference Analysis in downlink. . . 28

3.4 Simulation Example of Cellular Traffic Load. . . 30

3.5 System Parameters for Simulation. . . 31

3.6 Multiple Services Parameters. . . 35

(13)

xii

List of Figures

1.1 Frame structure and cross-slot interference in the TDD/CDMA system. 4 2.1 Frequency and time utilization of TDD and FDD mode . . . 8

2.2 Frame and time slot structure in the TDD-CDMA systems . . . 10 3.1 A trisector cellular system with the virtual cell. . . 16 3.2 Example: Users’ location distribution of each group in a sector. . . . 20 3.3 The proposed virtual-cell based LP-DCA. . . 21 3.4 Ring separation inside a sector. . . 26 3.5 The impact of the base station to base station cross-slot interference

to the different degree of asymmetric traffic in the uplink. . . 31 3.6 The impact of the mobile to mobile cross-slot interference to the

dif-ferent degree of asymmetric traffic in the downlink. . . 33 3.7 The Outage Probability corresponds to the different kinds of traffic

asymmetry when there are 60 users in each sector. . . 37 3.8 The Outage Probability corresponds to the different kinds of traffic

asymmetry when there are 75 users in each sector. . . 38 3.9 The Outage Probability corresponds to the different kinds of traffic

(14)

4.3 Effect of mobile-to-mobile cross-slot interference on the normalized ra-dius of the inner region . . . 57 4.4 The cellular system with grouped cells, where cell A has a symmetric

load, cell B has more downlink traffic than uplink traffic, and cell B has more uplink traffic than downlink traffic. . . 60 4.5 Effect of traffic asymmetry on the overall outage performance with

both downlink and uplink users. . . 64 4.6 Effect of traffic asymmetry and mobile-to-mobile cross-slot interference

on the outage performance for the downlink users in the outer region. 65 4.7 Effect of traffic asymmetry and the mobile-to-mobile cross-slot

inter-ference on the outage performance of all the downlink users in the system. . . 66 4.8 Effect of traffic asymmetry and base-to-base cross-slot interference on

the outage performance for all the uplink users. . . 67

(15)

(16)

1

CHAPTER 1 Introduction

The increasing demands for the higher speed wireless internet applications impose many new challenges on spectrum and radio resource management in wireless net-works. One of key challenges in supporting the wireless internet services is to handle the traffic asymmetry between the uplink and the downlink. That is, some services may require more radio resources in the downlink transmission, while some services may require more uplink radio resources [1]. Hence, an intelligent radio resource al-location to support asymmetric services becomes an important topic in the future wireless networks.

1.1 Problem and Solution

The objective of this thesis is to efficiently utilize the two dimensions of radio re-source - time and space - to support the traffic asymmetry and enhance the system performance in the TDD-CDMA systems. However, asymmetric traffic will cause the cross-slot co-channel interference, thereby seriously degrading the system perfor-mance. To alleviate the impact of cross-slot interference and improve the system performance, we propose to incorporate the DCA algorithms with advanced antenna techniques.

(17)

1.2 Mobile Radio System

Code Division multiple access (CDMA) system is a promising radio access technique for the third-generation mobile communication systems due to its high flexibility and efficiency. In the CDMA systems, there are two different operation modes, namely frequency division duplex (FDD) and time division duplex (TDD). Comparing to the FDD-CDMA system with a pair of separated frequency bands used for downlink and uplink transmissions, the uplink and downlink transmissions in the TDD-CDMA systems multiplex the uplink and downlink time slots on the same frequency band. By exploiting the inherent time division component, time division duplex (TDD) mode is very suitable to provide asymmetric traffic services. [2, 3].

However, to support the asymmetric traffic in the TDD-CDMA system, the different asymmetric traffic conditions among cells may cause heavy cross-slot

inter-ference, which will seriously degrade the system performance [3–5]. Take the

TDD-CDMA systems specified in the Universal Mobile Telecommunications System as an example (UMTS) [6, 7]. A TDD frame has 15 time slots, where the first one is usu-ally used for signaling, and the others can be allocated for either the uplink or the downlink traffic channels as shown in Fig. 1.1. The boundary between the uplink and downlink time slots within a transmission frame is called the switching point. When two neighboring cells have different switching points due to distinct uplink-to-downlink traffic ratios, some time slots may be used for uplink-to-downlink transmissions in one cell, while being used for uplink transmissions in other cells. The opposite up-link and downup-link transmissions in some time slots for two neighboring cells is called the cross-slot interference in this paper. Note that in Fig. 1.1, there are two kinds of cross-slot interference: base-to-base cross-slot interference in the uplink and the mobile-to-mobile cross-slot interference in the downlink. Because the transmission power of a base station is much higher than that of a mobile terminal, the

(18)

base-to-base cross-slot interference is quite significant. Meanwhile, as a mobile terminal approaches to another mobile of an adjacent cell at the cell boundary, the mobile-to-mobile cross-slot interference can not be ignored. Both types of cross-slot interference will degrade the system performance seriously [8,9], since it is usually suggested that a time slot should be used for the same transmission direction either uplink or down-link for two neighboring cells. This constraint, however, obviously wastes time slots if traffic asymmetric ratio of two neighboring cells differs significantly. Apparently, this approach may lose the key advantages of the TDD systems in supporting asymmetric traffic services [3,10]. The key to relax this restriction is to find an effective approach to overcome the cross-slot interference in the TDD/CDMA system.

In the literature, there are two research directions to avoid the cross-slot in-terference. The first one is to apply the dynamic channel assignment (DCA) tech-niques [11–13]. In [11], the authors proposed an ordered DCA algorithm to reduce the probability to use the time slot that may have a higher chance of experiencing the cross-slot interference. When the traffic load or the traffic asymmetric ratio is high, this method may have difficulty in overcoming the cross-slot interference. The authors in [12] and [13] proposed the region-based and path gain division DCA. In these algorithms, the users close to the home base station are assigned to use the time slots even having the cross-slot interference, whereas the users near the cell boundary are assigned the clean time slots without the cross-slot interference. The performance of this system will highly depend on the way of separating the inner and the outer regions. In the time-varying traffic condition, it is usually hard to accurately separate two regions with one being robust to the cross-slot interference and the other being untolerant to the cross-slot interference.

Another research direction of overcoming cross-slot interference is to adopt advanced antenna techniques [14–16]. The authors in [14] and [15], focusing on the TDD-CDMA system and TDD-TDMA system, respectively, suggested to adopt

(19)

UL UL UL UL UL UL BS-A BS-B Desired transmission direction Cross-slot interference

Cell A

Cell B

(b) cross-slot interference scenario

(a) TDD/CDMA frame structure

UL

Cell A

Cell B

DL DL DL DL DL DL DL DL DL UL UL UL UL UL UL UL DL DL DL DL DL DL DL

mobile-to-mobile cross-slot Interference

base-to-base cross-slot interference

MS-A MS-B

(20)

the sectorized antennas combined with time slot allocation methods to suppress the inter-cell cross-slot interference. In [16], the authors applied the minimum variance distortionless response (MVDR) beamformer technique to eliminate the cross-slot in-terference in the uplink. However, the mobile-to-mobile cross-slot inin-terference in the downlink was not considered in [16].

In this thesis, aiming to alleviate the cross-slot interference, we propose a link-proportional dynamic channel assignment scheme (LP-DCA) with sectorized antennas in Chap. 3. With the assistant of directional antennas, we utilize the concept of virtual cell composed from three neighboring sectors with the same coverage area of a cell [14]. By taking the advantages of virtual cell, we propose a effective DCA algorithm to flexibly alleviate the co-channel interference, especially for the cross-slot interference. The key idea of LP-DCA scheme is to classify the cross-slot interference and allocate the radio resource according to the users’ received signal quality. The total users of a sector are sorted based on their received signal strength. We partition these sorted users into some different groups and allocate the time slots with the consideration of alleviating the cross-clot interference. Specifically, the sector with the largest downlink traffic load will allocate both the downlink and uplink groups in a ascending order from the left side of available time slots. The sector with largest uplink traffic load will allocate both the downlink and uplink groups in a ascending order from the right side of the available time slots. The sector with similar uplink and downlink traffic load will allocate the downlink groups in ascending order from left side of the available time slots and uplink groups in ascending order from right side of the available time slots. By properly allocating users, LP-DCA outperforms other DCA algorithms with high ability to alleviate the co-channel interference. Nevertheless, we find that most DCA algorithms including the LP-DCA can not effectively alleviate the base-to-base cross-slot interference [11].

To furthermore alleviate the cross-slot interference, we propose a cross-slot 5

(21)

interference-based dynamic channel assignment scheme combined with the MVDR beamformer. In the proposed scheme, the DCA is focused on reducing the mobile-to-mobile cross-slot interference, while the MVDR beamformer is aiming to suppress the base-to-base cross-slot interference. To alleviate the mobile-to-mobile cross-slot interference, the basic idea of the cross-slot interference-based DCA is to allocate time slots to users in a specific order. Specifically, for reducing the base-to-base cross-slot interference, both receiving and transmitting beamforming weights are designed according to the MVDR beamformer criterion and the fourier beamformer criterion, respectively. According to our numerical results, the cross-slot interference-based DCA can improve system performance in both downlink and uplink, while providing asymmetric traffic services with a great deal of flexibility.

1.3 Thesis Outline

The rest of this thesis is organized as follows. Chapter 2 reviews the document of TDD-CDMA wireless cellular system and the basic concept of dynamic channel as-signment issues. In chapter 3, we evaluate a novel link-proportional dynamic channel assignment scheme with the assistance of directional antennas. In chapter 4, we consider a high efficient time and space radio resource allocation algorithm to inte-grate the dynamic channel assignment and the smart antenna techniques. Chapter 5 provides the remarks of this thesis and the suggestions for future work.

(22)

7

CHAPTER 2 Background

In this chapter, we will investigate the evaluated TDD-CDMA system in this thesis. We list the document as follows.

2.1 Introduction to the CDMA systems

The code division multiple access (CDMA) technique has become one of the leading standards in digital cellular and personal communication systems because of its ad-vantages in high capacity and flexible radio resource utilization. The CDMA system can provide low to high multiple data rate services and allow users transmitting at the same time. To supply multiple services and efficient spectrum utilization, two transmission modes-the frequency division duplex (FDD) and time division duplex (TDD) mode-are considered in the CDMA systems. Figure 2.1 illustrates the fre-quency and time allocation in both FDD and TDD modes. We can find that there is a pair of frequency band needed for the downlink and the uplink in the FDD mode, while the TDD mode only apply a single frequency band and separate the down-link and updown-link transmission in time. Table 2.1 illustrates the key parameters in the TDD-CDMA and FDD-CDMA system [7, 17–19]. The FDD mode is suitable for the large coverage areas, and the TDD mode is well suited for small cells with limited user mobility to provide high capacity and asymmetric traffic services. We will focus on the TDD-CDMA systems in the following article.

(23)

Bandwidth Power density time frequency time time

Uplink Bandwidth Downlink Bandwidth

Power density

frequency

TDD

FDD

(24)

The physical frame length of these two modes is similar. Each frame of length 10 ms is divided into 15 time slots, each of which may be allocated to either the uplink or the downlink as depicted in Figure 2.2. With such flexibility, the TDD mode can be adapted to different environments and deployment scenarios [17,20]. In figure 2.2, each time slot includes 2560 chips and divided into two data filed, one midamble filed, and one guard period filed. The data fields contain the the desired data bit to the receiver. Within the midamble filed, training sequences are transmitted. The Guard period (GP) is used to cope with timing inaccuracies, power ramping, and also needed to avoid the corruption of transmission by counting the propagation delay [21].

There are some characteristics of the TDD-CDMA system listed below.

• Dynamic capacity between the uplink and the downlink: In the TDD mode, uplink

and downlink transmissions are divided in the time domain. It is possible to change the duplex switch point and move the capacity from the uplink to the downlink, or vice versa, if the system capacity requirement is asymmetric between uplink and downlink [22].

• Cross-slot interference: Since both the uplink and downlink transmissions share

the same frequency in TDD mode, the signals of the two transmission may inter-fere with each other. In order to alleviate the impact of these interinter-ferences, an efficient opposite-direction interference avoidance algorithm should be taken into the consideration [23].

• Discontinuous transmission: The mobile and the base station transmission are

discontinuous in TDD. To avoid the overlapping of uplink and downlink transmis-sions, a guard period is used in the end of each slot.

• Reciprocal channel: The fast fading depends on the frequency and, therefore, in

FDD systems the fast fading is uncorrelated between uplink and downlink. As the 9

(25)

10 ms

10/15 ms

Data filed 1 Midamble Data filed 2 GP

Time Slot for the downlink transmission Time Slot for the uplink transmission GP Guard Period

2560 chips

Frame

Time Slot

Figure 2.2: Frame and time slot structure in the TDD-CDMA systems

same frequency is used both for uplink and downlink in TDD, the TDD transceiver can estimate the fast fading which affect its transmission. The knowledge of the fast fading can be utilized in power control or the adaptive antenna techniques.

(26)

Table 2.1: Comparison of UTRA FDD and TDD physical key parameters.

UTRA TDD UTRA FDD

Multiple access

method

CDMA(inherent FDMA) CDMA(inherent FDMA)

Duplex method

TDD (suitable for asym-metric services, e.g., web browsing.)

FDD (suitable for symmet-ric services, e.g., voice.)

Channel spacing 5 MHz (nominal)

Carrier chip rate 3.84 Mcps

Timeslot structure 15 slots/frame

Frame length 10 ms

Multirate concept

Multicode, multislot and or-thogonal variable spreading factor (OVSF)

Multicode and OVSF

Modulation QPSK

Detection Coherent, based on

mi-damble

Coherent, based on pilot symbols

Intra-frequency han-dover

Hard handover Soft handover

Inter-frequency han-dover

Hard handover

Channel allocation DCA supported No DCA required

Intra-cell interference cancellation

Support for joint detection Support for advanced re-ceivers at base station

(27)

2.2 Channel Assignment Schemes

The tremendous growth of the wireless/mobile user population, coupled with the bandwidth requirements of multiple applications, requires efficient reuse of scare radio spectrum allocated to wireless/mobile communications, i.e. CDMA system. Efficient use of radio spectrum is also important from a cost-of-service point of view. The basic prohibiting factor in radio spectrum reuse in interference caused by the environment or other mobiles. Interference can be reduced by deploying efficient radio subsystems and by making use of channel assignment techniques [24].

In the radio and transmission subsystems, techniques such as deployment of time and space diversity systems, use of low-noise filters and efficient equalizers, and deployment of efficient modulation schemes can be used to suppress interference and to extract the desired signal. However, co-channel interference caused by frequency reuse is the most restraining factor on the overall system capacity in the wireless network, and the main idea behind channel assignment algorithms is to make use of radio propagation path loss characteristics in order to minimize the carrier-ro-interference ration (CIR) and hence increase the radio spectrum reuse efficiency [25, 26].

In the TDD-CDMA systems, the time dimension component enables effective strategy of interference avoidance. This characteristic is very useful in some scenarios. It allows the deployment of TDD for coordinated as well as for uncoordinated oper-ation, since interference in certain time slots can be efficiently alleviated. Otherwise, for transmission of discontinuous packet data with high peak rates changing very fast, it is extremely complicated to operate a system at its interference limit, since power control cannot converge. In TDD, the traffic can be shifted to certain time slots and then does not degrade the performance of circuit switched transmission [27]. To utilize these advantages of TDD systems, a quality-based efficient dynamic channel

(28)

assignment (DCA) algorithm could be used, which guarantees a robust and efficient operation [22].

Recently, the antenna techniques with extra spatial diversity have received increasing interest for improving the performance of wireless radio systems [28]. Dif-ferent antenna techniques will provide difDif-ferent kinds of assistance to the system environment. The one of the major contribution is the co-channel interference can-cellation according to the space separation between users. In the TDD systems, the antennas techniques can furthermore integrate with time-based channel assignment scheme. In this work, though the integration of the dynamic channel assignment and the antenna techniques, we demonstrate that the system performance can be improved significantly with the efficient radio resource allocation.

(29)

14

CHAPTER 3 A Novel Link Proportional Dynamic

Channel Assignment for a Virtual-cell

Based TDD/CDMA System with

Asymmetric Traffic

The increasing demands for the higher speed wireless internet applications impose many new challenges on spectrum and radio resource management in wireless net-works. One of key challenges in supporting the wireless Internet services is to handle the traffic asymmetry between the uplink and the downlink. Time division duplex (TDD) is an efficient approach to resolve the traffic asymmetry issue, which can dy-namically allocate the resource (time slots) to the uplink and the downlink. However, asymmetric traffic will cause the cross-slot co-channel interference, thereby seriously degrading the system performance. This paper proposes a novel link-proportional dy-namical channel assignment (LP-DCA) scheme combined with tri-sector directional antennas to alleviate the impact of cross-slot interference for the TDD code division multiple access (CDMA) systems. With the help of directional antennas, the tri-sector cellular system will form a virtual cell, which is composed of three sectors from three

(30)

concentrate on combating the cross-slot interference within a virtual cell by assigning time slots to the users according to their radio link quality. Our numerical results show that LP-DCA combined with tri-sector cellular structure can significantly reduce the cross-slot interference, thereby improving the outage and throughput performances for the TDD CDMA system under asymmetric traffic.

3.1 System Model

In this paper, a radio resource unit (RU) is defined as the combination of spreading code, time slot, and frequency [29]. As [30], we consider a system with maximum 32 codes (RU) simultaneously in each time slot on a single frequency of 5 MHz band-width. Hence, there are 448 RU allocated to provide the services for users. Different users with different services may require difference amount of RU. Next, we will in-troduce the cellular system model, and the propagation model used in this paper. We consider a TDD/CDMA hexagonal cellular system with directional antennas em-ployed at the base station. Mobiles are assumed to be uniformly distributed over the cells.

3.1.1 Virtual Cell Concept

By taking the advantage of directional antennas, a virtual cell can be established [31]. Figure 3.1 illustrates a trisector cellular system with the virtual cell, where a virtual cell is defined as the same coverage area of a cell but is composed of three sectors from the three neighboring base stations. As shown in Fig. 3.1, sectors SA1, SB2, and SC3

forms a virtual cell. By employing simple sector antennas at base stations, it is clear that the inter-cell interference can be restricted within a virtual cell coverage area. The similar concept was also proposed in [32] to reduce interference in sectorized

(31)

Cell A Cell B Cell C 1 A S 1 B S 2 A S 3 B S 3 A S 3 C S 2 B S 2 C S 1 C S Virtual cell

Figure 3.1: A trisector cellular system with the virtual cell.

FDD/CDMA systems. The intra-cell interference between sectors of a cell can be ignored. Here we assume that the antenna gain is the same over the whole sector. Taking advantages of this additional orthogonality from the direction separation of sector antennas, we propose a virtual-cell based interference avoidance algorithm to support asymmetric traffic services in TDD/CDMA systems.

3.1.2 Propagation Model

In our propagation model, we consider propagation loss, lognormal distributed shad-owing effect and the minimum link budget between two wireless entities (a mobile or a base station) i and j as Gi(γ, α) described as

Gi,j = G(ri,j, αi,j) = min (

GiGjh2ih2j

d4

i,j

· 10αi,j10 , M CL_i→j) (3.1)

where Gi and Gj are the antenna gain of communication entities i and j, hi and hj

(32)

and j, αi,j is the log-normal shadowing component with a standard deviation of σi,j.

Otherwise, we introduce minimum coupling loss (MCL) as the minimum distance loss including antenna gain measured between antenna connectors [33]. The value of σ and MCL are different with the distinct environment conditions. Here, we set shadowing deviation between the mobile to its served base station σs = 6 dB,

shadowing deviation between the mobile to the adjacent base station σn = 8 dB,

shadowing deviation between the base station to the adjacent base station σb = 3

dB, shadowing deviation between mobile to mobile σm = 10 dB, MCL between base

station and mobile equal to 10−5.3_{, and MCL between base station and mobile equal}

to 10−4 _[34].

3.2 The Proposed Link-Proportional Dynamical

Chan-nel Scheme

Based on the virtual cell concept, we propose a novel efficient virtual-cell based dy-namical channel assignment scheme. The objective of the proposed scheme is to minimize the overall system interference. With the advantage of directional anten-nas, we can easily get the desired information of some specific neighboring sectors and assume the slot synchronization between adjacent sectors is achievable. In our proposed scheme, the number of time slots allocated for the downlink or the uplink in a frame depends on the ratio between downlink traffic load to uplink traffic load of each cell. The proposed scheme can support diverse asymmetric services.

Assume that there exists Nsactive users in sector s. Define Ks= {1, 2, 3, . . . , Ns}

as the index set of all active users. And π : Ks→ Ks is the permutation of the index

set with respect to the link gains of all connections in index set so that

∀n, m ∈ π(Ks) and n < m, Gh,π(n) ≤ Gh,π(m) .

(33)

where Gh,i is the link gain from user i to his home serving base station defined in 4.1.

Algorithm: LP-DCA

Step 1: Traffic Load Information Updating

The sector s measures the aggregated request rates of the downlink and the uplink, denoted by R(d)_s and R(u)_s , respectively. As long as the measured R(d)_s and R(u)_s of sector s are updated, these two values will be signaled to other two sectors in the same virtual cell.

Step 2: Time-slot Number Determination

The number of time slots for downlink and uplink, denoted by T(d) _{and T}(u)_,

can be obtained by T(d) _{= round}

µ T · R(d)s R(d)s +R (u) s ¶

and T(u) _{= T − T}(d)_{, where}

T = T(d)_{+ T}(u) _{is the total number of time slots in each frame.}

Step 3: User Grouping Define ΩRU =

P

k∈KsRUk the total requested radio units. For the downlink time

slots, the users in Ksset can be partitioned into T(d)groups according to a partition

vector, ωd=

©

ωd,0, ωd,1, ωd,2, . . . , ωd,T(d)

ª

, which can be derived by

ωd,0 = 0. ωd,1 = arg min k    π(k) X m=1 RUm ¯ ¯ ¯ ¯ ¯ ¯ π(k) X m=1 RUm ≥ ΩRU T(d) · R(d)_s R(d)_s + R(u)_s   . ωd,x = arg min k    π(k) X m=ωd,(x−1)+1 RUm ¯ ¯ ¯ ¯ ¯ ¯ π(k) X m=ωd,(x−1)+1 RUm ≥ ΩRU T(d) · R(d)_s R(d)_s + R(u)_s   , x = 2, . . . , T (d)_. _(3.2)

(34)

Similarly, the user grouping for uplink time slot can be directly obtained by (3.2). We give an example about the distribution of users for each group in Fig. 3.2, when T(d) _{or T}(u) _{equal to 5.}

Step 4: Time Slot Allocation

The sector s determines the time slot assignment order in terms of the traffic load information from the other sectors. Three possible orders with corresponding status are shown in Fig. 3.3.

Status 1: The R(d)_s is maximum among the virtual cell

The time slot x is allocated to the users in downlink group x,¡π(ωd,(x−1)+ 1), . . . ,

π(ωd,(x))

¢

, where the allocation of time slot x is in the ascending order from slot 1 to T(d)_{. For uplink, the time slot x is then allocated to the users in uplink}

group (x − T(d)_), ¡_π(ω

d,(x−T(d)₋₁₎+ 1), . . . , π(ω_d,(x−T(d)₎)

¢

, where the allocation of time slot x is also in the ascending order from slot (T(d)_{+ 1) to time slot T .}

Status 2: The R(u)_s is maximum among the virtual cell

The time slot x is allocated to the users in downlink group (T(d) _{− x + 1),}

where the allocation of time slot x is in the descending order from slot T(d) _to

1. For uplink, the time slot x is then allocated to the users in uplink group (T − x + 1), where the allocation of time slot x is also in the descending order from slot T to (T(d) _{+ 1).}

Status 3: Otherwise

The time slot x is allocated to the users in downlink group x,¡π(ωd,(x−1)+ 1), . . . ,

π(ωd,(x))

¢

, where the allocation of time slot x is in the ascending order from slot 1 to T(d)_{. For uplink, the time slot x is then allocated to the users in}

uplink group (T − x + 1), where the allocation of time slot x is also in the descending order from slot T to (T(d)_{+ 1).}

(35)

BS Group 1 ( G1 ) Group 2 ( G2 ) Group 3 ( G3 ) Group 5 ( G5 ) Group 4 ( G4 )

(36)

Same directional inter-cell interference Same directional inter-cell interference Cross-slot inter-cell interference Downlink Uplink

Status 1 (The sector with more downlink traffic load)

G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G1 G2 G3 G4 smaller link gain smaller link gain larger link gain larger link gain Downlink Uplink Uplink

Status 2 (The sector with more uplink traffic load)

smaller link gain G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G1 G2 G3 G4 smaller link gain larger link gain larger link gain Downlink Uplink smaller link gain

Status 3 (The sector with similar uplink and downlink traffic load)

G1 G2 G3 G4 G5 G6 G7 G7 G6 G5 G4 G3 G2 G1 smaller link gain larger link gain larger link gain

Figure 3.3: The proposed virtual-cell based LP-DCA.

(37)

The proposed scheme can alleviate the cross-slot interference as like region-based dynamic channel assignment. The users near the cell boundary are automati-cally allocated in the edge time slots of a frame in which the users have lower proba-bility to be impacted by the cross-slot interference. The proposed algorithm is more flexible than region-based method which only defines a proper region radius thresh-old. Otherwise, the proposed scheme can reduce the near-far effect by allocating the users with similar received signal strength to the same time slot. Since the codes are not perfectly orthogonal due to the multipath fading, the proposed method can reduce the intra-cell interference resulted from the near far effect. Furthermore, the proposed method can reduce the same directional inter-cell interference [35]. This phenomenon can be seen when the group of higher link gain in one sector is allocated in the time slot in which the other group of lower link gain in the neighboring sector is allocated at the same time. It has been shown that this way can effectively increase the system performance.

3.3 Interference and Capacity Analysis

In this section, we will introduce the interference and capacity analysis method used in this work.

3.3.1 Uplink

Then the intra-cell interference for user i in the uplink transmission is defined as

I_h,i(u) =

Nh

X

j=1,j6=i

Pm,j· G(dh,j, αh,j) (3.3)

where Nh is the total number of users in the home sector h, Pm,i be the transmit

(38)

The received same directional inter-cell interference from the mobiles the neigh-boring sector in the uplink transmission can be denoted as

I_m→b,i(u) = X k∈B(u) Nk X ik=1 Pm,ik · G(dik,h, αik,h) (3.4)

where B(u) _{is the set of neighboring sectors in the uplink transmission, N}

k is the

number of users in the sector k.

If the adjacent sector k within the same virtual cell is in the downlink trans-mission, then the base-to-base cross-slot interference from sector k is

I_b→b,i(u) = X k∈B(d) G(dk,h, αk,h) Nk X ik=1 Pb,ik (3.5)

where B(d) _{is the set of neighboring sectors in are downlink transmission, and P}

b,ik

is the transmission power form the neighboring sector k to the mobile ik. By taking

the advantages of virtual cell, we can ignore the intra-cell inter-sector interference. From (3.3), (3.4), and (3.5), we can obtain the bit energy-to-noise ratio for

user i as follows _µ Eb N0 ¶ i = W Ri · Pr

I_h,i(u)+ I_m→b,i(u) + I_b→b,i(u) + η (3.6) where Pr is the received power level after power control, Ri is the transmission data

rate of user i, and η is the white thermal noise power. We define

³

Eb

N0

´

t as the target bit energy-to-noise ratio. We can obtain the

achievable data rate of user i as ζi, where

ζi = W · Pr ³ Eb N0 ´ t· (I (u) total,i+ η) (3.7)

where I_total,i(u) = I_h,i(u)+ I_m→b,i(u) + I_b→b,i(u) , and then we can get the overall sector aggregate data rate as C = Nh X i=1 ζi. (3.8) 23

(39)

3.3.2 Downlink

Similarly as subsection A, the intra-cell interference of the downlink user i can denoted as I_h,i(d) = (1 − ρ) · Nh X j=1,j6=i Pb,j · G(dh,i, αh,i) (3.9)

where ρ is the orthogonal factor between the codes utilized in the same sector. The received same directional inter-cell interference from the neighboring sec-tor k in the downlink transmission can be denoted as

I_b→m,i(d) = X

k∈B(d)

Nk

X

ik=1

Pb,ik · G(dk,i, αk,i) (3.10)

And the mobile-to-mobile cross-slot interference is denoted as

I_m→m,i(d) = X k∈B(u) Nk X ik=1 Pb,ik · G(dik,i, αik,i) (3.11)

Then we can get the bit energy-to-noise ration and aggregate data rate of the sector in the downlink as like the uplink case.

3.4 Numerical Results

In this section, we will investigate the location-dependent interference analysis, the capacity evaluation of the proposed link-proportional DCA scheme, and the blocking probability between the different dynamic channel assignment schemes. In subsection A and subsection B, we analyze the impact of two different inter-cell interference to explain the proposed scheme design rule. In subsection C and subsection D, we eval-uate the system capacity corresponding to the different dynamic channel assignment scheme and the degree of traffic asymmetry. To further verify the advantage of the proposed link-proportional dynamic channel assignment scheme, we design a multiple

(40)

D. We evaluate the system blocking probability when each dynamic channel assign-ment scheme experiences different kinds of traffic load and traffic asymmetry. These numerical results will demonstrate that the proposed link-proportional dynamic chan-nel assignment scheme outperform the other dynamic chanchan-nel assignment algorithms in most aspects.

3.4.1 Average Uplink Location-dependent Interference

Anal-ysis

In subsection A and subsection B, we analyze the mean received interference asso-ciated with a specific range which will form a partial ring in the sector as shown in Fig. 3.4. We assume each user applies one RU and uniformly distributed in a specific area. In [31], the received interference is highly dependent on the location of the tar-get user and the interfering sources. Here, we focus on the analysis of the impact of inter-cell interference according to the group of users’ distributed range in the sector. We partition the sector into ten rings. Each ring has the equal and distinct distance range from the base station. Table 3.1 illustrates the system parameters used in this paper.

Since the location of the receiver (base station) is fixed in the uplink, the received inter-cell interference will highly depend on the locations of the interfering users. We assume the number of users is the same in each analysis. In [31], we have derived the mean received cross-slot interference and same directional inter-cell interference mobile according to a served user of the neighboring sector MSn in a specific location, denoted by Im(u)_b→b(rs, θs), and Im(u)m→b(rs, θs), respectively, where rs

and θs denote the distance and the angle of the user MSn according to its serving

base station BSn. As an extension, we can obtain the mean cross-slot interference

(41)

Table 3.1: System Parameters .

Cell radius R=500 m

Base station antenna hight hb = 15 m

Mobile antenna height hm =2 m

Target SINR Eb

N0 = 4 dB

Thermal noise η = -112 dBm

number of rings for analysis Nr = 10

number of users per analysis N = 16

kth _{ring’s distance range of the BS 50(k − 1) ∼ 50k(m)}

Ring 2 Ring 1 Ring 1 Ring 2 Ring 10 BSh BSn MSt MSn Ring 10 Ring 3 Same directional inter-cell interference Cross-slot interference Ring 3

u cross

Im

same

Im

d cross

Im

(42)

Table 3.2: Inter-cell Interference Analysis in uplink.

Ring 6 Ring 7 Ring 8 Ring 9 Ring 10

Ir(u),k_b→b 8.4 16.2 28.6 47.1 73.3

Ir_m→b(u),k 0.32 0.79 1.82 3.5 5.7

from a specific k-th ring of the adjacent sectors in the uplink as

Ir_b→b(u),k = N Ak ring · Z Z kthring inBsn Im(u)_b→b(r, θ) drdθ . (3.12)

And the mean received inter-cell interference from the uplink mobiles within ring k of adjacent sector is obtained as

Ir_m→b(u),k = N Ak ring · Z Z kthring inBsn Im(u)_m→b(r, θ) drdθ . (3.13)

Table 3.2 illustrates the mean received inter-cell interference in the uplink while the users of neighboring sector are in the different rings. Because the received inter-cell interference from ring 1 to ring 5 is much smaller than the received inter-cell interference from ring 6 to ring 10 , we don’t list them here. For ease of analysis, we normalize Ir_b→b(u),k and Ir_m→b(u),k to Pr.

3.4.2 Average Downlink Location-dependent Interference

Anal-ysis

In the downlink case, we show the mean received interference level of the specific users who locate inside a specific ring. We assume that the interfering users are uniformly distributed in the neighboring sectors. From [31], we can get the mean received cross-slot interference Im(d)m→m(rs, θs) and same directional inter-cell interference from

adjacent sector Im(d)_b→m(rs, θs) of the target user MSt, where rsand θsare the distance

and the angle of the target user MSt to its serving base station BSh, respectively.

(43)

Table 3.3: Inter-cell Interference Analysis in downlink.

Ring 6 Ring 7 Ring 8 Ring 9 Ring 10

Irm→m(d),k 0.034 0.115 0.4 1.83 28.9

Ir_b→m(d),k 0.3 0.65 1.14 1.9 2.9

Consequently, we can get the mean received cross-slot interference of all users in the ring k of neighboring sector as follows

Ir(d),k m→m = N Ak ring · Z Z kthring inBSh Im(d) m→m(r, θ) drdθ . (3.14) where Ak

ring is the area of the ring. And the mean received the same directional

inter-cell interference of all users in the kth _{ring as}

Ir_b→m(d),k = N Ak ring · Z Z kthring inBSh Im(d)_b→m(r, θ) drdθ . (3.15)

Table 3.3 illustrate the mean received inter-cell interference for the users located in the difference rings of the sector in the downlink.

From the above inter-cell interference analysis, it can be found that the cross-slot interference is much dominant in the outer ring of a sector in both the downlink and uplink cases. We must avoid to allocate to the users near the cell boundary with the time slots that will receive huge cross-slot interference. In the uplink case, because the transmission power of a base station and the antenna gain between base stations are quite high, the received cross-slot interference will be always larger than the same directional inter-cell interference. To alleviate the cross-slot interference in the uplink is an important design issue. In the downlink case, because the link gain between mobile and mobile is small, the received cross-slot interference will be serious just in some outer rings. Since the probability for two mobiles closed to each other in different base stations is small, the mobile-to-mobile cross-slot interference problem will be

(44)

less than base-to-base cross-slot interference in the uplink. To increase the system capacity effectively in the downlink, both the intra-cell interference and the inter-cell interference should be taken into consideration. The proposed link-proportional dynamical channel assignment is designed to alleviate both of the intra-cell and inter-cell interference, and then increase the system performance.

3.4.3 Uplink Capacity Analysis

In the analysis, single service is assumed and each user is with the same amount of RU requirement and the target bit energy-to-noise ratio. To evaluate the effects on capacity resulting from traffic asymmetry, we define a parameter Λ as the degree of the traffic asymmetry among three sectors in the virtual cell. First, We choose the sector with the most symmetric load between the downlink and the uplink as the referenced cell, denoted as IN D

IND = min i n¯_¯ ¯T_i(d) − T_i(u) ¯ ¯

¯ , i ∈ the sectors in the virtual cell o

; (3.16)

where T_i(d) and T_i(u) are the number of time slots allocated in the downlink and the uplink for the sector i.

We set the degree of traffic asymmetry for the indicator ΛIN D to be zero and

define the number of downlink time slots of the sector IND as a indication T_{IN D}(d) . Then we set the degree of traffic asymmetry for the sector i as

Λi = kTi(d) − TIN D(d) k (3.17)

With the definition of traffic asymmetric level, we design a system traffic con-dition to evaluate the system performance corresponding to different degrees of traffic asymmetry as shown in Table 3.4. Table 3.4 illustrates the setting of the traffic loads of the considered sectors. We define the traffic factor TF to represent the total system

(45)

Table 3.4: Simulation Example of Cellular Traffic Load.

Sector A Sector B Sector C

T_A(d) T_A(u) ΛA TB(d) T (u) B ΛB TC(d) T (u) C ΛC 7 7 0 7 7 0 7 7 0 7 7 0 8 6 1 6 8 1 7 7 0 9 5 2 5 9 2 7 7 0 10 4 3 4 10 3 7 7 0 11 3 4 3 11 4 traffic load as Traffic Factor (TF) =

the utilized resource unit (RU)

the total number of available resource unit (RU) (3.18) Table 3.5 lists the system parameters for the rest of numerical results. Figure 3.5 depicts the impact of base-to-base cross-slot interference with respect to the traffic asymmetry for different dynamic channel assignment schemes in the uplink. Is is shown that the proposed LP-DCA scheme has the better system performance than those of other methods. The proposed scheme is the most robust one to suppress with the base-to-base cross-slot interference.

(46)

Table 3.5: System Parameters for Simulation.

Traffic Factor FT=0.95

Orthogonal factor in downlink Forth=0.8

Radius factor for the inner-region Rregion = 0.62

(for region-based DCA)

0 0.5 1 1.5 2 2.5 3 3.5 4 460 480 500 520 540 560 580 600 620

The degree of traffic asymmetric

Random Ordered Region LP

Aggregate Data Rate (ADR) (kbps)

Figure 3.5: The impact of the base station to base station cross-slot interference to the different degree of asymmetric traffic in the uplink.

(47)

3.4.4 Downlink Capacity Analysis

Figure 3.6 illustrates the impact of the mobile-to-mobile cross-slot interference with respect to the traffic asymmetry for the different dynamic channel assignment schemes in the downlink. It can be found that the impact of the mobile-to-mobile cross-slot interference is not as severe as the base station to base station cross-slot interference. Since the degradation of the link gain is in several order of the distance between two mobiles, only few very closed mobiles may cause large interference. While the users are uniformly distributed, the probability that two mobiles approaches to each other in different sectors is quite small. However, the proposed scheme can still outperform the other algorithms due to the alleviation of near-far intra-cell interference and inter-cell interference from the downlink users in the neighboring sector. The LP-DCA scheme can reduce the impact of the cross-slot interference similar with the region-based method but is more flexible and efficient. It can expend less radio resource to compensate the near-far effect and the same directional inter-cell interference with a proper slot allocation mechanism. It can be also found that the radius factor of the region-based dynamic channel assignment algorithm should be adequate to the degree of traffic asymmetry. In this environment, region-based algorithm will have the higher performance when the degree of traffic asymmetry Λ between adjacent sectors equals to 2. The performance of the ordered dynamic channel assignment scheme is slightly better than that of random dynamic channel assignment scheme. When the traffic load is heavy, the ordered dynamic channel assignment scheme fails to combat with cross-slot interference and approaches the performance of random dynamic channel assignment scheme. To further demonstrate the performance of the proposed scheme, we adopt a multiple services scenario in the next section. Two important factors, users density of the sector and the traffic asymmetry, are considered in this scenario that will affect the performance of TDD/CDMA systems significantly.

(48)

0 0.5 1 1.5 2 2.5 3 3.5 4 800 900 1000 1100 1200 1300 1400

The degree of traffic asymmetric

Aggregate Data Rate (ADR) (kbps)

Random Ordered Region LP

Figure 3.6: The impact of the mobile to mobile cross-slot interference to the different degree of asymmetric traffic in the downlink.

(49)

3.4.5 Multiple Services

In this subsection, we evaluate the performance of the system outage probability of the LP-DCA scheme compared with other dynamic channel assignment schemes in the multiple services environment. Each service class has its requirements of radio resource and the bit energy-to-noise ratio, and different traffic asymmetry of different service classes is imposed on the system. The performance of outage probability significantly impacts on the transmission quality of each service class. In TDD mode, cross-slot co-channel interference is a dominant factor to cause severe performance degradation. A well designed DCA scheme to address the different traffic asymmetry among the virtual cell will effectively improve the performance of system outage probability. Three different classes of service are assumed in the system and the setting of the parameters for these service classes are shown in Table 3.6. Service class 1 is the traditional voice service with balanced traffic load in the downlink and uplink. Service class 2 has the heavy traffic load in the downlink and called downlink enhancement data service. And Service class 3 has the heavy traffic load in the uplink and called uplink enhancement data service. The data service has the lower target Eb

N0

than the voice service due to advanced error correction coding schemes. The system outage probability in the multiple service environment is defined by

Psys,o= X ∀i 3 X k=1 P r ½ Eb N0 < γk ¯ ¯ ¯ ¯ i ∈ class k} · φk, (3.19)

where γkis the bit energy-to-noise ratio requirement for class k, and φk= P r {i ∈ class k}

is the traffic ratio of class k.

The performance of system outage probability is mainly affected by two factors: the distribution of the traffic asymmetry of all sectors in a virtual cell and the traffic load of each sector. To evaluate the performance of outage probability with respect to traffic asymmetry, a scenario is adopted to vary the traffic ratios φ to adjust the

(50)

Table 3.6: Multiple Services Parameters.

Required Radio Resource Required Radio Resource

in the Downlink in the Uplink Target Eb

N0

Service Class 1 1 RU 1 RU 4 dB

Table 3.7: The Distribution of Traffic Load of three services in each sector.

Sector A Sector B Sector C

φ1 φ2 : φ3 φ1 φ2 : φ3 φ1 φ2 : φ3 Step 1 1/3 1 : 1 1/3 1 : 1 1/3 1 : 1 Step 2 1/3 1 : 1 1/3 3 : 2 1/3 2 : 3 Step 3 1/3 1 : 1 1/3 3 : 1 1/3 1 : 3 Step 4 1/3 1 : 1 1/3 9 : 1 1/3 1 : 9 Step 5 1/3 1 : 1 1/3 1 : 0 1/3 0 : 1 35

(51)

distribution of the traffic asymmetry with a fixed total number of users. In each sector, the number of voice users is fixed to be one-third (φ1 = 1/3) of the total number of

users. Initially, the traffic ratios for downlink and uplink enhancement service, φ2

and φ3, are set to be φ2 = φ3 = 1/3, too. In the following steps, we gradually change

some users in sector B from the downlink enhancement data service to the uplink enhancement data service for each time. Conversely, we gradually reduce some users in sector C from the uplink enhancement data service to the downlink enhancement data service at the same time. And the traffic ratios φ2 and φ3 of sector A are all

fixed 1/3. Table 3.7 illustrates the traffic load condition in the following results. Figure 3.7, 3.8, and 3.9 illustrate the system outage probability with respect to the traffic asymmetry with given total number of users by 60, 75, and 90. Here, the traffic asymmetry is defined as the difference of the traffic load in downlink between sector B and sector C that is set according to the above description. While the user density is light as shown in Fig. 3.7, the proper outage probabilities of each DCA schemes are below 6% in most traffic asymmetry condition. When the user density increases as shown in Fig. 3.8 and Fig. 3.9, the outage probabilities of LP-DCA are always the lowest one as the traffic asymmetry is increased, while the outage probabilities of random, ordered, and region-based DCA schemes are higher than that of LD-DCA by 5%, 4%, and 2.5% , respectively. It can be also found in Fig. 3.7, Fig. 3.8, and Fig. 3.9 that the increase of the additional outage probability of LP-DCA is limited below 6% as the traffic asymmetry is increased, while the increase of the additional outage probability of the random, ordered, and region-based DCA schemes are about 11.5%, 11%, and 9%, respectively. This is because that LP-DCA can minimize the total receive interference flexibly. The proposed LP-DCA scheme can attain the better system performance and alleviate the cross-slot interference due to the increasing traffic asymmetry. From the above numerical results, the proposed

(52)

1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 6 7

The corresponding steps with different traffic asymmetry in Table VII

Outage Probability (%)

Random Order Region LP

Figure 3.7: The Outage Probability corresponds to the different kinds of traffic asym-metry when there are 60 users in each sector.

and support asymmetric traffic services effectively.

(53)

1 1.5 2 2.5 3 3.5 4 4.5 5 0 2 4 6 8 10 12

Outage Probability (%)

Random Order Region LP

(54)

1 1.5 2 2.5 3 3.5 4 4.5 5 2 4 6 8 10 12 14 16

Outage Probability ( γ >abscissa) (%) Random Order Region LP

(55)

40

CHAPTER 4 Joint Cross-Slot Interference-Based

Dynamic Channel Assignment and

Antenna Beamforming for the

TDD/CDMA Systems with Asymmetric

Traffic

As the requirement for the asymmetric data services is growing, the time division duplex/code division multiple access (TDD/CDMA) system has been considered as an important wireless network in the future. However, different asymmetric traffic loads among cells may cause heavy cross-slot interference, which can seriously degrade the system performance. To alleviate the impact of the cross-slot interference and improve the system performance, we propose a cross-slot interference-based dynamic channel assignment algorithm incorporated with antenna beamforming techniques. The proposed cross-slot interference-based DCA algorithm aims to reduce downlink cross-slot interference and distributedly assign downlink and uplink time slots to support asymmetric traffic services in each cell. The antenna beamforming techniques adopted here are mainly to avoid the impact of heavy uplink cross-slot interference.

(56)

based DCA algorithm and antenna beamforming can effectively suppress the cross-slot interference in both downlink and uplink, thereby enabling a TDD/CDMA system to flexibly provide various asymmetric traffic loads in different cells and achieve high system performance.

4.1 System Model

In this section, we introduce the cellular system model and the propagation model. We consider a TDD/CDMA hexagonal cellular system and the mobiles are assumed to be uniformly distributed over the system. It is assumed that power control is conducted in both the downlink and the uplink transmission.

4.1.1 Propagation Model

We define the propagation loss between two wireless entities (a mobile or a base station) i and j as

Gi,j = G(ri,j, αi,j) = min (

GiGjh2ih2j

r4

i,j

· 10αi,j10 , M CL_i→j) (4.1)

where Gi and Gj are the antenna gains of communication entities i and j; hi and

hj are the antenna heights, ri,j is the distance between i and j, and αi,j is the

log-normally distributed shadowing component with standard deviation σi,j .

Further-more, we introduce the minimum coupling loss (MCL) as the minimum propagation loss including antenna gain measured between antenna connectors [33]. We defined

MCLb→m as the MCL between the base station and the mobile and MCLm→m as

the MCL between two mobiles. In [34], the value of MCLb→m is defined as 10−5.3,

and MCLm→m is equal to 10−4.

(57)

4.1.2 Uplink SINR

Let Pm,i be the transmit power level of the target mobile i. Then the different kinds

of the interference received for mobile i at any time slot can be categorized as (1) Intra-cell interference I_h,i(u) = Nh X j=1,j6=i Pm,j· G(rh,j, αh,j) (4.2)

where Nh is the number of interfering mobiles observed in the uplink direction at the

home base station h, and G(rh,j, αh,j) is defined in (4.1).

(2) Inter-cell interference from the mobiles in the neighboring cell k

I_m→b,i(u) = X

k∈B(u)

Nk

X

ik=1

Pm,ik · G(rh,ik, αh,ik) (4.3)

where B(u) _{is the set of adjacent cells in the uplink transmission, and N}

kis the number

of interfering mobiles in cell k.

(3) Base-to-base cross-slot interference

I_b→b,i(u) = X k∈B(d) G(rh,k, αh,k) Nk X ik=1 Pb,ik (4.4)

where B(d) _{is the set of neighboring cells in the downlink transmission, and P}

b,ik is

the transmission power form the neighboring cell k to the mobile ik.

Combining all types of interference, the signal-to-interference and noise ratio (SINR) of user i can be written as

γi =

Pm,i· G(rh,i, αh,i)

I_h,i(u)+ I_m→b,i(u) + I_b→b,i(u) + η (4.5) where η is the white thermal noise power, and I_h,i(u), I_m→b,i(u) , and I_b→b,i(u) are defined in (4.2), (4.3), and (4.4).

(58)

For simplicity, let

I_b→b,i(u) = I_b→b(u) , for i = 1, 2, . . . , Nk0, k0 ∈ B(d), (4.6)

I_m→b,i(u) = I_m→b(u) , for i = 1, 2, . . . , Nk, k ∈ B(u), (4.7)

and

Ia= Im→b(u) + I

(u)

b→b+ η. (4.8)

Then the system adjust the power of user i iteratively to achieve its target SINR γt. We can obtain the power in (n + 1)-th step as

P_m,i(n+1) = γt· " _N h X j=1,j6=i P_m,j(n)Gh,j Gh,i + I (n) a Gh,i # . (4.9)

where γt is the target SINR for each user, and then

P(n+1) m = γt· ¡ F · P(n) m + I(n) ¢ , (4.10) where Pm = [Pm,1, Pm,2, . . . , Pm,Nh] T_, _{I = [}Ia(n) Gh,1, Ia(n) Gh,2, . . . , Ia(n) G_h,Nu,h], and F is a

nonnegative matrix, defined as [F]ij =    0, if j = i Gh,j Gh,i > 0, if j 6= i (4.11)

4.1.3 Downlink SINR

Three different kinds of received interference in the downlink transmission for mobile

i at any time slot are defined as

(1) Intra-cell interference I_h,i(d) = Nh X j=1,j6=i (1 − ρ) · Pb,j· G(rh,i, αh,i), (4.12)

where ρ is the orthogonal factor between the codes used in the downlink of the same cell.

結合動態通道調配機制和天線波束形成技術以支援包含非對稱性傳輸的分時雙工分碼多工存取系統

國 立 交 通 大 學

電 信 工 程 研 究 所

碩 士 論 文

結合動態通道調配機制和天線波束形成技術

以支援包含非對稱性傳輸的

分時雙工分碼多工存取系統

Joint Dynamical Channel Assignments and

Antenna Beamforming for the TDD/CDMA

Systems with Asymmetric Traffic

研 究 生： 陳奕丞

Joint Dynamical Channel Assignments and

Antenna Beamforming for the TDD/CDMA

Systems with Asymmetric Traffic

A THESIS

Presented to

The Academic Faculty

By

Yi-Cheng Chen

June, 2004

摘要

本論文主要目的在於希望能夠藉由適當地運用時間與空間二個

不同維度上的無線資源，以期在分時雙工（Time Division Duplex,

TDD）分碼多工存取（Code Division Multiple Access, CDMA）系統

下充分地支援非對稱性的資料傳輸，並同時提高整體系統效能。隨著

非對稱性傳輸需求的日益增加，分時雙工分碼多工存取系統在未來的

無線網路中將扮演著一個重要的角色。然而，由於分時雙工分碼多工

存取系統中的細胞都使用相同的頻帶來上傳和下載資料，因此非對稱

性的資料傳輸將會導致兩個彼此相鄰但是傳輸方向相反的基地台間

產生極大的交錯時槽干擾（Cross-Slot Interference）

。許多的研究

顯示，交錯時槽干擾不但嚴重地影響系統效能，並造成龐大無線資源

的浪費。

為了解決交錯時槽干擾的問題並增進整體系統性能，我們致力於

研究關於各種動態通道調配（Dynamic Channel Assignment, DCA）

機制以及天線波束形成技術（Antenna Beamforming）的特點。在前

時雙工分碼多工存取系統的影響。在指向性天線的幫助下，三區域的

細胞系統將會由三個不同基地台的相鄰區域來形成一個虛擬細胞

（virtual cell）

。我們發現在這種細胞架構下，交錯時槽干擾將會

被限制在一個虛擬細胞中，因此，鏈接相對性動態通道調配機制能夠

專注在虛擬細胞中藉由使用者的無線鏈結品質來做時槽的配置，以達

到充分降低交錯時槽干擾的需求。許多結果都顯示鏈接相對性動態通

道調配機制能夠顯著的勝過其他動態通道調配機制，並能提供分時雙

工分碼多工存取系統一個更有效率的資源配置法。

然而，我們發現大部分的動態通道調配機制，包括鏈接相對性動

態通道調配，都無法有效的解決上傳時的交錯時槽干擾。為了更進一

步解決這種交錯時槽干擾，我們更進一步提出了一個結合智慧型天線

波束形成技術的交錯時槽干擾為主的動態通道調配機制（Cross-Slot

Interference-Based DCA）

。我們所提出的交錯時槽干擾為主的動態

通道調配機制，主要希望能夠降低下載時的交錯時槽干擾，並利用細

胞各自分散的方式調整上傳和下載的時槽數目，以其在個別細胞中充

分地支援非對稱性的資料傳輸。智慧型天線的波束形成技術，在這邊

將被用來對付上傳時嚴重的交錯時槽干擾。我們實驗的結果顯示所提

出的交錯時槽干擾為主的動態通道調配機制能夠充分地壓制交錯時

槽干擾的影響，進而使得分時雙工分碼多工存取系統能夠充分地滿足

不同細胞間對於非對稱資料傳輸的個別需求，並同時能達到更高更好

的系統效能。

Summary

Acknowledgments

Contents

List of Tables

List of Figures

CHAPTER 1

Introduction

1.1

Problem and Solution

1.2

Mobile Radio System

Cell A

Cell B

(b) cross-slot interference scenario

Cell A

Cell B

1.3

Thesis Outline

CHAPTER 2

Background

2.1

國立交通大學

電信工程研究所

碩士論文

研究生：陳奕丞