PhD Thesis on HPA OFDM and PAPR

Resource Allocation and Reuse for Inter-Cell Interference Mitigation in OFDMA-

based Communication Networks

Von der Fakultät für Elektrotechnik und Informationstechnik der

Rheinisch-Westfälischen Technischen Hochschule Aachen zur

Erlangung des akademischen Grades einer Doktorin der

Ingenieurwissenschaften genehmigte Dissertation

vorgelegt von

Diplom-Informatikerin

Zheng Xie

aus Shanghai, V.R. China

Berichter: Universitätsprofessor Dr. Ing. Bernhard Walke

Universitätsprofessor Dr. rer. nat. Rudolf Mathar

Tag der mündlichen Prüfung: 22. Dezember 2011

ABSTRACT

Abstract

The future wireless systems are envisaged to offer ubiquitous high data-rate coverage

in large areas. With the Orthogonal Frequency Division Multiple Access (OFDMA)

transmission technique, great benefits in handling Inter-Symbol Interference (ISI),

inter-carrier interference and providing high flexibility in resource allocation can be

reaped. Nevertheless, Co-Channel Interference (CCI) or so-called Inter-Cell

Interference (ICI) as a big obstacle is still remaining in OFDMA systems, which

encumbers to attain both, wide area coverage and high spectral efficiency in multi-

cellular communication networks.

It is known that effective reuse of resources in a cellular system can highly enhance

the system capacity. With a smaller Frequency Reuse Factor (FRF), more available

bandwidth can be obtained by each cell. So, in this sense employing the classical FRF

of 1 is preferable. However, with the usage of FRF-1, most User Terminals (UTs) are

seriously afflicted with heavy ICI, especially in the border areas of cells. And this

causes low cell coverage and inferior system capacity. Conventional method to figure

out this problem is by increasing the cell-cluster-order to avoid the reuse of the same

frequency bands in neighboring cells, which can mitigate the ICI efficiently,

nonetheless at the cost of a decrease in available bandwidth for each cell. This would

result in reduced cell capacity and lower system spectrum efficiency in general, and

would worsen in the case of unbalanced traffic distribution among cells. Thus, it is

desirable to combat the ICI by other means.

A promising method to solve the ICI problem is ICI coordination, which may

potentially attain significant performance improvements and has become very

important in next generation wireless communication networks. To take aim at

improving cell-edge performance while retaining system spectrum efficiency of

Reuse-1, several representative local ICI coordination approaches with static

frequency resource partitioning are introduced and studied at first in this monograph,

including the classical Fractional Frequency Reuse (FFR) scheme, the well-known

Soft Frequency Reuse (SFR) scheme and the newly emerged Incremental Frequency

Reuse (IFR) scheme. Based on thoroughly analysing the advantages and limitations of

these approaches, a novel ICI mitigation design called Enhanced Fractional Frequency

Reuse (EFFR) scheme and its two derivatives (the EFFR-Advanced scheme and

EFFR-Beyond scheme) are proposed for a better fulfillment of the goals, namely, to

enhance the mean system capacity while restraining the ICI at the cell edge. The EFFR

scheme designs a resource allocation and reuse mechanism combined with power

Abstract II

allocation and interference-aware reuse. Taking the inherent vulnerability of the Cell-

Edge Users (CEUs) into account, the EFFR scheme reserves resources for them with

two specific emphases: 1) using dedicated FRF-3 subchannels; 2) data transmission

with higher transmit power. Taking advantage of the location-specific predominance

of the Cell-Centre Users (CCUs), the EFFR scheme allows them to occupy resources

with FRF-1 and interference awareness.

The performance evaluations are done by means of both, analytical models and

stochastic even-driven simulation. The presented results show that the EFFR schemes

can efficiently mitigate the ICI in OFDMA-based cellular networks and outperform

the SFR scheme, the IFR scheme and static Reuse schemes under any propagation

condition. With the usage of the EFFR scheme, the medium is able to be more

effectively utilized; higher flexibility as well as more robustness can be attained; the

overall cell capacity is significantly improved; and the cell coverage can be

substantially enlarged.

TABLE OF CONTENTS

Table of Contents

ABSTRACT .................................................................................................................. I

TABLE OF CONTENTS .......................................................................................... III

1 INTRODUCTION ................................................................................................ 1

1.1 Motivation and Objectives ............................................................................ 1

1.2 Contribution of the Thesis ............................................................................. 2

1.3 Outline............................................................................................................. 4

2 OVERVIEW OF CELLULAR NETWORKS ................................................... 7

2.1 The Wireless Channel .................................................................................... 8

2.1.1 Path Loss ................................................................................................ 9

2.1.2 Shadowing ............................................................................................ 10

2.1.3 Multipath Effects .................................................................................. 10

2.2 The Cellular Concept ................................................................................... 10

2.2.1 Frequency Reuse .................................................................................. 12

2.2.2 Carrier-to-Interference-plus-Noise Ratio .............................................. 14

2.3 OFDMA ........................................................................................................ 15

2.4 Resource Allocation and Interference Mitigation ..................................... 16

3 RELATED TECHNOLOGIES ......................................................................... 21

3.1 Inter-Cell Interference Mitigation .............................................................. 21

3.1.1 ICI Cancelation ..................................................................................... 22

3.1.2 ICI Randomization ............................................................................... 22

3.1.3 ICI Coordination ................................................................................... 23

3.2 Frequency Reuse Techniques for ICI Mitigation in Cellular OFDMA

Networks ................................................................................................................. 25

3.2.1 Static Fractional Frequency Reuse ....................................................... 26

3.2.2 Soft Frequency Reuse ........................................................................... 29

3.2.3 Incremental Frequency Reuse .............................................................. 34

3.2.4 Summary and Conclusion ..................................................................... 35

4 ENHANCED FRACTIONAL FREQUENCY REUSE DESIGN ................... 37

Table of Contents IV

4.1 EFFR ............................................................................................................. 37

4.1.1 Design Requirements ........................................................................... 37

4.1.2 Concept of the EFFR Scheme .............................................................. 38

4.1.3 Distinctions between the EFFR and other ICI Mitigation Schemes ..... 45

4.1.4 Expected Benefits by Using the EFFR Scheme ................................... 46

4.2 EFFR-Advanced .......................................................................................... 48

4.3 EFFR-Beyond............................................................................................... 50

5 PERFORMANCE ANALYSIS ......................................................................... 53

5.1 Cell Coverage Analysis ................................................................................ 54

5.1.1 Carrier to Interference Calculation ....................................................... 54

5.1.2 Cell Coverage Comparison .................................................................. 65

5.2 Cell Capacity Analysis ................................................................................. 77

5.2.1 Mean Cell Capacity Computation ........................................................ 77

5.2.2 Cell Capacity Comparison ................................................................... 80

5.2.3 Area Spectral Efficiency ...................................................................... 83

5.3 Comparison between Analytical and Simulation Results......................... 85

5.4 Summary and Conclusion ........................................................................... 90

6 PERFORMANCE EVALUATION BY MEANS OF SIMULATION ........... 91

6.1 Simulation Scenario and Simulation Setup ............................................... 92

6.1.1 Cellular Scenario .................................................................................. 92

6.1.2 Performance Metrics ............................................................................ 92

6.1.3 Link Adaptation and Error Modeling ................................................... 93

6.1.4 Frame Structure and Overhead ............................................................. 94

6.1.5 Resource Allocation and Scheduling ................................................... 95

6.1.6 Simulation Parameters ......................................................................... 99

6.2 Simulation Results ..................................................................................... 101

6.2.1 Reference Scenario............................................................................. 101

6.2.2 Impact of Range Ratio on System Performance................................. 129

6.2.3 Impact of Power Ratio on System Performance................................. 137

6.3 Summary and Conclusion ......................................................................... 145

7 CONCLUSION AND OUTLOOK ................................................................. 147

7.1 Summary .................................................................................................... 147

7.2 Conclusion .................................................................................................. 148

Table of Contents V

7.3 Outlook ....................................................................................................... 150

A SIMULATION ENVIRONMENT – OPEN WIRELESS NETWORKS

SIMULATOR ........................................................................................................... 153

A.1 Overview ..................................................................................................... 153

A.1.1 FUN .................................................................................................... 155

A.1.2 Configurability of the OpenWNS ....................................................... 156

A.1.3 Simulation Progress ............................................................................ 156

A.1.4 Simulator and traffic model ................................................................ 157

A.2 WiMAX MAC Layer ................................................................................. 158

A.2.1 User Plane........................................................................................... 158

A.2.2 Medium Access Control ..................................................................... 159

B REFERENCE SCENARIO ............................................................................. 161

B.1 Additional Simulation Results for NLOS DL .......................................... 161

B.2 Additional Simulation Results for NLOS UL .......................................... 163

C IMPACT OF RANGE RATIO ON SYSTEM PERFORMANCE ............... 169

C.1 Additional Simulation Results for LOS DL ............................................. 169

C.2 Performance in LOS UL ............................................................................ 170

C.3 Performance in NLOS DL ......................................................................... 176

C.4 Performance in NLOS UL ......................................................................... 182

C.5 Conclusion .................................................................................................. 185

D IMPACT OF POWER RATIO ON SYSTEM PERFORMANCE .............. 187

D.1 Performance in LOS UL ............................................................................ 187

D.2 Performance in NLOS DL ......................................................................... 193

D.3 Performance in NLOS UL ......................................................................... 198

D.4 Conclusion .................................................................................................. 201

LIST OF ABBREVIATIONS .................................................................................. 203

LIST OF FIGURES ................................................................................................. 207

LIST OF TABLES ................................................................................................... 215

BIBLIOGRAPHY .................................................................................................... 217

CURRICULUM VITAE .......................................................................................... 221

CHAPTER 1

1 Introduction

Introduction

1.1 Motivation and Objectives ............................................................................ 1

1.2 Contribution of the Thesis ............................................................................. 2

1.3 Outline............................................................................................................. 4

1.1 Motivation and Objectives

The future wireless systems are envisaged to offer ubiquitous high data-rate coverage

in large areas. Hence, aggressive spectrum reuse (frequency reuse of 1 or close to 1)

becomes a key objective in most Fourth Generation (4G) cellular standardization

bodies and forums, for example, Worldwide Interoperability for Microwave Access

(WiMAX) IEEE 802.16m [1][2][3] and Third Generation Partnership Project Long

Term Evolution (3GPP-LTE) [4][24][5], to achieve high system capacity and simplify

radio network planning.

Both IEEE 802.16m and 3GPP-LTE are based on Orthogonal Frequency Division

Multiple Access (OFDMA) air-interface technology, which uses multi-channel

Orthogonal Frequency Division Multiplexing (OFDM) and provides subchannel

access in time and frequency domain. Decisions of using which timeslot, subchannel

and power level for communication are determined by intelligent Medium Access

Control (MAC) protocol to seek to maximize the Carrier-to-Interference-plus-Noise

Ratio (CINR) for every Mobile Station (MS). Equipped with OFDMA, great benefits

in handling Inter-Symbol Interference (ISI), inter-carrier interference and providing

high flexibility in radio resource allocation can be reaped. Even so, Co-Channel

Interference (CCI) or so-called Inter-Cell Interference (ICI) as a big obstacle with the

OFDMA is still remaining, which encumbers to attain both, wide area coverage and

high spectral efficiency in cellular systems.

It is known that dense reuse of available spectrum in a cellular system may highly

enhance the system capacity. However, the obvious pitfall of a dense reuse is strong

ICI which limits network as well as cell-edge throughput. To obtain the full potential

of OFDMA in a dense reuse environment, appropriate Radio Resource Management

(RRM) algorithms for ICI mitigation are necessary. Furthermore, since solutions with

1. Introduction 2

low system complexity and flexible spectrum usage are desirable, systems with

distributed RRM techniques have gained much attention recently.

The objective of this thesis is to improve cell-edge performance while retaining the

maximum system capacity and highest spectral efficiency. Based on a thoroughly

analyzing of several to date prevailing ICI avoidance techniques, a new design called

Enhanced Fractional Frequency Reuse (EFFR) scheme is put forward in this

monograph, which is combined with a power allocation and an interference-aware

reuse mechanism to achieve not only ICI limitation at cell edge but also a great

enhancement of overall cell capacity in OFDMA-based communication networks.

This thesis aims at providing an overview of contemporary ICI coordination

techniques and introducing the new EFFR scheme for a better ICI mitigation in

OFDMA-based systems in general, instead of addressing a specific system standard

implementation and evaluation. The objective is rather the analysis and discussion of

the interaction of system, environment and various scheme-related parameters and

how these determine the system performance in terms of cell capacity, cell coverage

percentage as well as spectral efficiency depending on the applied resource allocation

and reuse strategies. This thesis provides detailed performance evaluations of the

proposed EFFR scheme by means of both, mathematical analysis and stochastic event-

driven simulation, in which its performance is compared with those using the

conventional reuse methods and some well-known representative ICI mitigation

techniques. In order to reach a reliable evaluation, all schemes are simulated with

individual power masks, and using a scenario with surrounding cells up to the 2nd

-tier.

1.2 Contribution of the Thesis

In this monograph, a novel ICI coordination scheme for both Downlink (DL) and

Uplink (UL) of a cellular OFDMA network is presented. The algorithm is evaluated at

the example of an IEEE 802.16e network, but the basic idea is applicable to any kind

of OFDMA- or Frequency and Time Division Multiple Access (FDMA/TDMA)-

based networks, like 3GPP-LTE [25] and its successor LTE-Advanced. It also can be

applied to other network types, such as Wireless Local Area Networks (WLANs) or

mesh networks, possibly with technology specific adaptations. The following gives an

overview of the main contributions of this work.

With a deepgoing study of several well-know frequency reuse schemes (static

Fractional Frequency Reuse, Soft Frequency Reuse, and Incremental

Frequency Reuse), which aim at mitigating excessive ICI among adjacent cells

in OFDMA-based communication networks, a series of novel ICI mitigation

designs (EFFR, EFFR-Advanced, and EFFR-Beyond) combined with power

1.2 Contribution of the Thesis 3

allocation and an interference-aware mechanism is proposed for a better

fulfillment of the goals, namely, to enhance the mean system capacity while

restraining the ICI at the cell edge.

With the usage of the CINR calculation, the maximal cell radius and

reasonable boundary definitions for separating different user-type zones for

each reuse partitioning scheme are determined.

Through analytical evaluations, the CINR distributions and the upper bounds

of an OFDMA-based cellular system in terms of cell coverage, mean cell

capacity, as well as area spectral efficiency of these schemes under various

propagation conditions (Line-of-Sight (LOS), Non Line-of-Sight (NLOS),

combined LOS-NLOS) are illustrated and compared. The analytical evaluation

supports and validates the setup of the Reference Scenario, which is chosen for

the simulative performance evaluation.

The performances of various proportion combinations of different frequency

reuse factor OFDMA subchannels in the EFFR schemes are demonstrated and

compared addressing the tradeoff between the cell capacity maximization and

fairness among users within a cell.

The implementations of the mainly concerned frequency reuse schemes,

including Soft Frequency Reuse, Incremental Frequency Reuse, and EFFR, are

embedded in an even-driven system level simulator. Comprehensive

performance evaluations of these frequency reuse schemes and two static

Reuse schemes are carried out by means of computer simulations under LOS

and NLOS propagation conditions, separately. Moreover, in order to reach a

reliable evaluation, schemes are simulated with individual power mask, using

scenarios with surrounding interfering cells up to the 2nd

-tier.

This thesis presents the system performance of each investigated scheme

(including CINR level distribution, mean throughput, coverage percentage,

and spectral efficiency), not only from the overall cell perspective but also

from different types of users (cell-centre users, cell-edge users, and possibly

the most remote users) perspective.

The most important factors that influence the OFDMA multi-cellular system

behavior are identified. And the impacts of traffic load, range ratio defined for

different types of users, as well as power ratio of higher power level to lower

power level on the system performance using each focused frequency reuse

scheme are evaluated. In addition, the effect of different proportion

combinations of reuse-3 and reuse-1 bandwidth on the EFFR performance is

also investigated and compared with the performances of other schemes.

1. Introduction 4

1.3 Outline

This thesis is organized in seven chapters. A short overview of the chapters is given as

follows.

The succeeding Chapter 2 first deals with the characteristics of a radio channel with

special focus on the channel properties which directly affect the performance of an

OFDMA-based cellular system. Then, basic concepts of planning wireless cellular

networks and principles of the multiple access scheme—OFDMA are outlined. The

chapter ends with an introduction of a three-stage radio resource allocation mechanism

concerning ICI mitigation in OFDMA-based cellular networks.

After a brief overview of the fundamental concepts of wireless and cellular systems in

Chapter 2, Chapter 3 elaborates the state-of-the-art of ICI mitigation techniques that

are currently under consideration within standardization bodies or that have recently

been used for emerging cellular systems. This goes along with outlining the categories

of ICI mitigation techniques, the most well-known approaches to each category as well

as their advantages and limitations. Finally, three most representative ICI mitigation

techniques, namely, the static Fractional Frequency Reuse (FFR), the Soft Frequency

Reuse (SFR), and the Incremental Frequency Reuse (IFR), are studied in detail.

Based on a thoroughly analyzing of their advantages and limitations respectively,

Chapter 4 proposes a novel ICI mitigation design referred as Enhanced Fractional

Frequency Reuse (EFFR) scheme and its two derivatives, namely, the EFFR-

Advanced scheme and EFFR-Beyond scheme.

Chapter 5 addresses the performance evaluation of the aforementioned reuse

partitioning techniques (including SFR, EFFR, EFFR-Advanced, and EFFR-Beyond)

in OFDMA-based cellular radio networks by means of mathematical analysis in

Matlab. The evaluation comprises the maximal cell radius, reasonable boundary

definitions for separating different user-type zones, CINR level distribution, cell

coverage, mean cell capacity, as well as area spectral efficiency by using each reuse

partitioning scheme under various propagation conditions.

The focused SFR, IFR, EFFR frequency reuse partitioning schemes are also integrated

into the so-call WiMAC module, which is an implementation of the IEEE 802.16

standard in the Open Wireless Network Simulator (OpenWNS) described in Appendix

A [36]. Using the OpenWNS and based on the upper bounds of cell radii resulting

from Chapter 5, an in-depth and comprehensive performance evaluation by means of

stochastic event-driven simulation of an OFDMA-based cellular system is completed

in Chapter 6. Valuable performance evaluation results by applying the proposed EFFR

scheme are compared with those using the SFR, the IFR and two static Reuse schemes.

1.3 Outline 5

The evaluation consists of three parts. The first part presents the performance of all

schemes depending on increased traffic load, whereas the other two parts give the

simulation results affected by the range ratio and the power ratio, respectively.

Finally, Chapter 7 concludes the thesis with a summary of the main findings and the

major results presented in this monograph, as well as an outlook on possible future

work.

CHAPTER 2

2 Overview of Cellular Networks

Overview of Cellular Networks

2.1 The Wireless Channel .................................................................................... 8

2.2 The Cellular Concept ................................................................................... 10

2.3 OFDMA ........................................................................................................ 15

2.4 Resource Allocation and Interference Mitigation ..................................... 16

In the past several decades, as time goes on and the demand for high data rate services

continuously increases, mobile radio networks have been developed from the 1st

generation analog networks over the 2nd

generation digital cellular networks

represented by the Global System for Mobile communications (GSM) and the General

Packet Radio Service (GPRS) standards to the 3rd

generation cellular networks

including the Universal Mobile Telecommunication System (UMTS), the Universal

Terrestrial Radio Access - Frequency Division Duplex (UTRA-FDD), as well as the

IEEE 802.16e technologies; and are further advancing rapidly towards the 4th

generation networks. This also goes along with a transition from offering circuit

switched services to high data rate packet switched services for Internet access.

Recently, both Third Generation Partnership Project (3GPP) and IEEE 802.16 Task

Group m (TGm) are actively collecting ideas and working on technologies for the next

generation (4G) systems to attempt to realize further high-speed and high-quality

packet switched services for end users. These next generation technologies are handled

under the name International Mobile Telecommunications – advanced (IMT-

advanced), which is the dedicated successor of IMT-2000. Since the available radio

spectrum for cellular systems appears increasingly precious and scarce, caused by the

growth of high rate services for a variety of applications, the need for advanced

algorithms and concepts to enhance the spectral efficiency of wireless networks

becomes thereby more and more urgent.

Before discussing specific methods to increase spectral efficiency in wireless networks,

the fundamental concepts of wireless networks with particular focuses on the

OFDMA-based cellular systems will be firstly overviewed in this chapter. In Section

2. Overview of Cellular Networks 8

2.1, the main characteristics of wireless channels are outlined. Section 2.2 introduces

the basic concepts of cellular systems, followed by a review of the multiple access

scheme OFDMA in Section 2.3. Finally, Section 2.4 concludes the chapter with an

introduction of radio resource management considering interference coordination in

OFDMA-based cellular networks.

2.1 The Wireless Channel

For designing mobile radio systems, it is essential to understand and be familiar with

the propagation characteristics of radio waves. In this section, some notable features of

wireless propagation will be briefly described.

The wireless communication channel between any pair of transmitter and receiver is a

fast fading accompanied multipath propagation channel, which holds several inherent

unfavorable characteristics [7]:

In most cases, radio channels are noisy and prone to distortions;

Moreover, signals propagating on a radio channel are often attenuated quickly,

and fluctuate continuously over time and frequency.

Hence, the data transmission over a wireless channel becomes one of the biggest

challenges in the development of mobile communication systems.

Free-space propagation without obstacles on a direct Line-of-Sight (LOS) path is an

ideal case, but of little practical relevance in mobile communication. In reality, various

obstructions and reflective surfaces may exist on the propagation path, for example,

buildings, cars, trees, etc. When radio waves impinge to obstructions, three

phenomena may occur on them, namely, reflection, scattering and diffraction. Due to

these effects, the transmitted signals arrive at the receiver over multiple paths of

different lengths and are superimposed there. According to their phases, the multipath-

signal may be amplified or weakened. The latter is called fading. While reflection

effect results in large scale attenuation, scattering and diffraction lead to the signal

decay on a small scale. In terms of signal level fluctuation frequency, they are also

known as slow and fast fading, respectively. Hence, signal propagation can be

described as follows:

Tx Tx RxRx

P g gP

L

(2.1)

, where PTx and PRx represent the transmission and the received power, respectively;

gTx and gRx stand for the antenna gains at the receiver and the transmitter side; L

denotes the signal attenuation in a mobile radio channel, which is composed of a

2.1 The Wireless Channel 9

superposition of three components: mean path loss LP, slow fading LS, as well as fast

fading LF

.P S FL L L L (2.2)

The large-scale path loss LP describes the mean signal degradation determined by

macro-parameters, such as distance between transmitter and receiver, carrier frequency,

and terrain features, etc. The slow fading LS (also known as shadowing) describes the

local signal variation averaged over short to medium time periods around a mean value

of LP, which is caused by changes in the propagation environment, for example,

constructions or roads with rough surfaces, or other obstacles with relatively smaller

dimensions in the order of tens of meters. The fast fading LF, however, is the effect of

time and frequency dependent signal variation around a mean value of LS on a short

time scale arising from movement of terminals and elements of the environment,

which reflect the micro-aspect of the wireless channel.

2.1.1 Path Loss

In a flat terrain model without taking obstacles into consideration, the path loss at

distance d can be simplified expressed as [8]:

1

2

2.Tx Rx

P Tx Rx

h hL g g

d

(2.3)

In this case, d >> hTxּhRx is a frequency-independent term with hTx and hRx being the

height of the transmitter’s antenna and the receiver’s antenna, respectively. It can be

seen that the signal power decays much faster (~ 1/d4) than with the free-space

propagation (~ 1/d2).

In a real environment, besides distance and reflections, signal further loses energy due

to wave scattering caused by rough ground surfaces and slow fading resulting from

obstacles on the propagation path. Thereby, the distance-dependent path loss can be

generally given by

PL d (2.4)

, where Η is a constant of proportionality and γ is the propagation coefficient, which

strongly depends on the environment and needs to be determined by measurements.

Typical values for the γ range between 2 and 5 including free-space, urban and

suburban environment.


2.1.2 Shadowing

Shadowing is mainly caused by obstacles on the LOS path between transmitter and

receiver, which further reduces the received signal power in addition to the distance-

dependent path loss. If the receiver is moving, the signal degradation by shadowing

mostly occurs for a relatively long time. Hence, shadowing is also referred to as slow

fading.

2.1.3 Multipath Effects

Along with the ubiquity of reflections, scattering and diffraction of radio waves in the

realistic environments, multipath propagation becomes therefore a typical and inherent

characteristic for a mobile radio channel, and mainly leads to three effects:

Delay spread In many cases, transmitted signals via multiple propagation

paths arrives at the receiver with different delays, which results in the

widening of a channel’s impulse response. When the resulted time dispersion

of the received signals (delay spread) is in the order of a data symbol period or

longer, Inter-Symbol Interference (ISI) will be caused and distort the

transmitted signal.

Fast fading All component waves, including the possible direct and

indirect reflection, scattering or diffraction radio waves, reach at the receiver

having individual propagation delays and phase-shifts. They thereby

superimpose themselves there constructively or destructively, leading to

variations of the received signal power of as much as 20 – 30 dB within a

distance in the order of the wavelength [9]. This effect is referred to as fast

fading.

Doppler effect When there is a relative movement between the transmitter

and the receiver, the received frequency differs from the original frequency at

the source. Depending on the direction of incidence of a component wave, the

relative to the transmitter moving receiver experiences either a positive or

negative Doppler shift, which results in a widening of the frequency spectrum.

2.2 The Cellular Concept

In order to offer sophisticated mobile communications over a large area, wireless

cellular networks divide the covered area into cells, as shown in Figure 2.1. All

communications within each cell are served by one Base Station (BS) located in the

cell-center using a certain frequency channel group, indicated by Fi (i = 1, 2, …, n) in

2.2 The Cellular Concept 11

Figure 2.11. The same frequency resource is repeatedly available (reused) for other

cells at a certain distance to avoid excessive conflicts due to severe Co-Channel

Interference (CCI) among adjacent cells. Hence, the main advantage of using cellular

systems is that through reusing radio channels in distant cells, the network coverage

can be provided to areas of any size.

However, how to determine the size and the shape of a cell, as well as how to allocate

resources among cells are very important in radio network planning, as they may

largely influence the system performance. The size and the shape of each cell depend

on signal quality received within the covered cell-area, which is related to many

factors, such as the surrounding terrain, buildings, the height of transmission antennas,

the transmission power of the BS, the expected traffic demands and density, as well as

the atmospheric conditions, etc. Cells are generally represented as idealized regular

hexagons, but because of topographical and environmental conditions, this is only an

F3

F3F3

F3

F2

F2

F2

F2F1 F1

F1

F1

F1

Figure 2.1: Illustration of a cellular system with omni-directional antennas. Each cell is served by one

in the cell-center located BS using a certain frequency bandwidth Fi.

——————————————

1 Another way of interpretation of Figure 2.1 is that a BS serves three sectors, each by a different frequency

channel group Fi. Then, a sector corresponds to a cell.


approximation of what actually occurs [8]. Naturally, in a real world scenario, the cell

shapes are very irregular and overlap with each other by approximately 10 – 15%. This

enables MSs operating near the boundary of a cell to choose which BS they are

associated to.

2.2.1 Frequency Reuse

As mentioned above, frequency reuse is a key characteristic in cellular networks. The

whole available bandwidth for a system is divided into several narrower subbands,

each of which is assigned once to a cell of each cluster consisting of several adjacent

cells. The number of subbands should equal the size of cell-cluster, termed as

Frequency Reuse Factor (FRF). This way, all directly neighboring cells in the system

use different subbands to avoid heavy CCI among them; and the entire available

system bandwidth can be reused in all cell-clusters distributed over the network

covered area so that the utilization of valuable spectrum resources can be ensured to

some extent.

Thus, the next question is how to determine the value of FRF δ, which is another

essential parameter in radio network planning. With a bigger FRF value, the distance

between inter-interfered cells becomes larger. And consequently, the CCI can be

significantly reduced, and better cell/system coverage can be attained. However, on

the other hand, since the available system bandwidth must be shared by a cell-cluster

(i.e., among every δ adjacent cells), each cell within the cell-cluster is assigned a

smaller number of channels and therefore can carry less traffic limiting the number of

User Terminals (UTs) that can be served. This may lead to an unfavorable spectral

efficiency. When a smaller FRF value is used, more bandwidth is available per cell.

Since the same frequency resources are then reused within a short distance, the CCI in

the system is increased limiting the number of UTs that can be served. The question is

to answer what FRF value would be the best choice to gain the maximum cell capacity.

Figure 2.2 illustrates cellular systems using a FRF of 1, 3, and 7, respectively. The

focused cell in the middle of each system is surrounded by 3 tiers of cells. Using the

single frequency deployment (FRF = 1), as shown in Figure 2.2a, all 3 tiers—a total of

36—surrounding cells are co-channel cells and originate heavy CCI. Yet, each cell in

the system has the whole available bandwidth to utilize. When the FRF increases to 3

(see Figure 2.2b), there is no interfering cells from the 1st-tier, and only 6 co-channel

cells from the 2nd

- and 3rd

-tier exist, respectively. Correspondingly, each cell is

assigned 1/3 of the whole available bandwidth for serving its traffic. Figure 2.2c gives

a cellular network with a FRF of 7, where the number of co-channel interfering cells is

reduced to 6 in total from the 3rd

-tier. However, in this case, only 1/7 of the whole

system bandwidth is available for each cell, which fundamentally limits the cell

capacity. Hence, an optimal assignment of resources to cells with a tradeoff between

2.2 The Cellular Concept 13

maximum spectrum utilization efficiency and optimal cell coverage in terms of CCI

needs to be found during the network planning process, which is also the main target

of this monograph.

31

186

1

1716

5

7

8

15

10 9

23

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

20

F1

F1

F1

F1

F1F3

F2

F2F1

F2

F3

F1

F3

F1 F2

F3F2

F3

F2 F3

F1 F2

F2

F2

F2

F2

F3

F3

F3

F3

F3

F3

F1

F1

F1

F1

F2

(a) FRF of 1 cellular system (b) FRF of 3 cellular system

F7

F1

F2

F5

F4F7

F2

F5F3

F6

F6

F5

F4

F6 F7

F3F4

F2

F3 F5

F2 F3

F1

F7

F1

F5

F4

F1

F6

F1

F2

F1

F7

F3

F4

F6

F1

(c) FRF of 7 cellular system

Figure 2.2: Cellular networks using different size cell-clusters, in each of which the in the middle

located cell is surrounded by 3 tiers of neighboring cells. (a) When FRF =1, the cell in the centre is

interfered by in total 36 co-channel cells in the near neighborhood; (b) When FRF =3, the cell in the

centre is interfered by 12 co-channel neighboring cells; (c) When FRF =7, the cell in the centre is just

interfered by 6 co-channel cells located on the relatively farther 3rd –tier.


2.2.2 Carrier-to-Interference-plus-Noise Ratio

When proper modulation techniques and link adaptation are employed, the capacity of

cellular systems depends essentially on the mean quality of received radio signals,

which can be characterized by the Carrier-to-Interference-plus-Noise Ratio (CINR):

1

M

k

k

CCINR

I N

(2.5)

, where C is the received carrier signal power, Ik denotes the received CCI signal from

one of M co-channel cells, and N is the noise power. The received signal power C is

determined by the distance between the transmitter and the receiver, the transmission

power and the associated path loss caused by the environment. The most dominant

noise type is thermal noise originated from the hardware components in the system.

And the ever present background radiation originating from cosmic or terrestrial

sources adds up to the overall noise. As for the perceived interference, actually, the

interference Ik includes not only CCI, which is also known as Inter-Cell Interference

(ICI), but also intra-cell interference. Since ICI is dominating in an OFDMA scenario

compared to the intra-cell interference, and in this monograph only OFDMA-based

cellular systems are under consideration, the focus will be therefore only on the ICI in

the remainder of this work. ICI depends on the number of co-channel cells, the

distance between these cells, the transmission power and the terrain characteristics, in

which the first two factors depend on the choise of the FRF value.

Depending on whether noise or ICI is the dominating effect, the cellular system is

referred to as noise-limited or interference-limited. In an interference-limited system,

CINR can approximately be replaced by the Carrier-to-Interference Ratio (CIR). And

the ICI plays in this case a crucial role in terms of achievable system capacity and

coverage performance. In the previous Subsection 2.2.1, it has been explained that

with a smaller FRF severe ICI could happen in the system. To be precise, the closer to

the cell edge, the heavier the ICI is. Due to a large distance between the transmitter

and the receiver, the desired signal C is then relatively weak compared to the ICI

signal strength. Without doubt, through increasing the FRF value the ICI can be

effectively limited, however, at the expense of sacrificing the available bandwidth for

each cell. To resolve such a problem is the main purpose of this thesis, and in Chapter

3 an overview of various contemporary and forward looking ICI mitigation techniques

will be summarized.

In addition, in order to further increase data rate and spectral efficiency without

superinducing more spectrum resources in cellular networks, two traditional spatial

2.3 OFDMA 15

reuse mechanisms are often taken into account, namely, cell splitting and sectorization.

Moreover, several modern spatial antenna techniques such as Multiple-Input Multiple-

Output (MIMO) and Space Division Multiple Access (SDMA), and adaptive

beamforming are also gaining much attention lately to enhance cellular system

performance. Yet, in order to facilitate discussion, this work just focuses on the

frequency reuse techniques including sectorized cells. Hence, none of the above

mentioned spatial reuse concepts will be further discussed hereinafter. However, all

frequency reuse schemes investigated in this thesis can be employed combined with

these spatial reuse techniques to achieve a higher system capacity.

2.3 OFDMA

Orthogonal Frequency Division Multiple Access (OFDMA) has gained increasing

interest recently. Due to its inherent robustness against frequency selective fading and

its capacity for achieving high spectral efficiency, OFDMA has been considered as a

modulation and multiple access method for 4G wireless networks, such as IEEE

802.16m (WiMAX) and 3GPP-LTE. With OFDMA, transmissions can be multiplexed

in both time and frequency based on the underlying OFDM system, which essentially

(a)

(b)

Figure 2.3: OFDMA subchannel structure: (a) distributed subchannel; (b) adjacent subchannel [10].


corresponds to a combination of Frequency and Time Division Multiplexing (FDM

and TDM).

In an OFDMA system, the available frequency subcarriers are separated into distinct

sets (be referred to as subchannels) and consequently can be allocated to different

data-streams for different users. An OFDMA subchannel is a logical channel which is

mapped onto subcarriers within the spectrum. And according to different modes of

construction, two kinds of OFDMA subchannels can be distinguished concerning the

diversity within a subchannel: distributed and adjacent subchannels as shown in

Figure 2.3 [10]. In a distributed subchannel, the subcarriers that form a subchannel are

spread across the spectrum, which results in an averaging effect concerning both

interference and fading. Hence, this approach is suitable for the cases that no detailed

Channel State Information (CSI) is available or mobile terminals move too fast. In

contrast to that, an adjacent subchannel is formed by subcarriers which are located

next to each other within the frequency spectrum. This allows the implementation of

frequency selective scheduling schemes, which provides an efficient exploitation of

diversity in multi-user scenarios by allocating subchannels to the user with good

channel quality, and can thereby significantly increase the system capacity. However,

it requires more accurate knowledge of the channel state. With inaccurate channel

estimation, large delays because of retransmissions in the case of packet errors might

be caused. Hence, such a scheme is only feasible if the channel does not vary too

quickly, i.e., if the velocity of the mobile terminal is not too high.

In summary, compared to the other multiple access schemes, like TDMA, FDMA,

OFDM, etc., OFDMA offers the following advantages:

OFDMA allows for a highly flexible resource allocation in the time and

frequency domain, and is scalable for the different considered bandwidths.

Either consecutive or non-consecutive subcarriers/time positions are supported

by the OFDMA scheduler.

A report of a channel quality indicator enables frequency selective scheduling.

2.4 Resource Allocation and Interference Mitigation

As already indicated in Section 2.2, ICI mitigation is a big challenge issue in cellular

systems. Excessive ICI may lead to severe performance degradation or connection loss

especially in the border area of cells. In order to efficiently reduce the ICI whilst not

drastically reduce the utilization of the scarce frequency spectrum, suitable radio

resource management (RRM) is desirable. In general, the resource allocation

scheduling process for an efficient handling of the RRM consists of three progressive

2.4 Resource Allocation and Interference Mitigation 17

and complementary stages concerning the scales of assignable resources, namely:

resource distribution, reuse partitioning and local OFDMA scheduling.

Resource distribution At this stage, the entire available resources in an OFDMA

cellular system will be shared among cells according to some certain

principles.

Reuse partitioning Then, the allotted resources from the first stage for each cell

will be subdivided among different types of users within each cell. This

stage is only employed in those schemes, where different resource

allocation mechanisms are applied to different types of users (for

example, the Fractional Frequency Reuse (FFR) scheme and the Soft

Frequency Reuse (SFR) scheme presented in Chapter 3, as well as the

Enhanced Fractional Frequency Reuse (EFFR) series proposed in

Chapter 4).

Relative to the local OFDMA scheduling introduced in the following, the first two

stages—the resource distribution and the reuse partitioning— together can be referred

to as system-wide or global resource allocation.

With respect to the execution mode, the global resource allocation can be realized in

two ways: static or dynamic. In a static manner, the global resource allocation is

applied during the network planning process. The resource share assigned to each cell

and how to subdivide this share for different types of users are fixed and will not be

changed in the subsequent scheduling process. In a dynamic manner, on the contrary,

the global resource allocation reacts on cell-load and user-load variations, where the

amount of resources distributed to each cell varies with taking into account the

demands of all cells within a large area. A central coordinator, Radio Network

Controller (RNC), collects continuously state information from all its associated cells

(or BSs), and re-distributes available radio resources for these cells after processing

this information at regular intervals.

In general, resources shared among cells can be time slots, frequency ranges, space, or

code resources. In this work, special focus will be put on a resource distribution in the

frequency domain. Hence, subchannel sets are assigned to cells by a RNC.

Furthermore, dynamic resource allocation can be done on different time scales. While

the fast global resource allocation takes place on the frame level or slightly above, the

communication intervals with the slow global resource allocation operates on time-

scales in the order of half a minute or even a minute. In order not to lead to drastically

increased signaling overhead and high computational complexity, which are important

prerequisites for the implementation in real networks, the slow global resource

allocation appears preferable to the fast one.


Local OFDMA scheduling The OFDMA scheduling is a fast resource

allocation operating on the frame level, which is locally performed by

the BS in each cell taking into account the information—the resource

share resulting from the global resource allocation—delivered by the

RNC. In contrast to the global resource allocation, the actual channel

characteristics should be considered during local OFDMA scheduling.

Different scheduling algorithms (such as, Round Robin, Proportional

Fair, Max-Throughput, etc.) can be applied to different reuse partitions.

Further, the overall transmission power should be spread over the

subchannels that are used for the parallel data transmissions, performed

in combination with an adaptive Physical Layer (PHY) mode selection.

An efficient local OFDMA scheduling requires accurate CSI and

interference estimation, which should be updated quickly to assure its

reliability. Hence, it would be inefficient in terms of signaling overhead

and system complexity to apply joint scheduling in the RNC, where

global resource allocation is done in the context of the OFDMA

scheduling for the entire system comprising all BSs and UTs [10].

It can be seen that the global and local resource allocation is interrelated and

complements each other so that precious frequency resources can be more efficiently

utilized without inducing heavy signaling overhead and computational complexity.

In this thesis, local OFDMA scheduling with static frequency resource allocation are

taken into consideration in both analytical and simulative performance evaluations.

They are applied based on the assumption that in each investigated approach all cells

in a network are assigned the same amount of resources, namely the same amount of

OFDMA subchannels. However, even in case dynamic global resource allocation is

employed, in the subsequent treatise it should be noticed that with the usage of the

SFR or the Incremental Frequency Reuse (IFR) approach, the amount of resources

assigned for each cell within the cellular system is fixed and impossible to be changed,

whereas using the EFFR series, the amount of resources assigned for each cell can be

dynamically adjusted by the RNC regarding the state reports from all its BSs. The goal

of resource distribution adjustment in the EFFR series can for example be the

reduction of the ICI, or the maximization of the overall network capacity, or the

maximization of the cell edge throughput. So, from this point of view, the EFFR series

possess a higher degree of freedom than the SFR and the IFR schemes.

Although the local ICI coordination mechanisms currently studied by the 3GPP build

on markedly lower complexity heuristics, ICI coordination mechanisms with slow

inter-cell communication (i.e. with slow global resource allocation) may mitigate the

ICI more efficiently so that the valuable limited resources can be more effectively

utilized. Thus, the future promising RRM taking ICI mitigation into consideration

should be realized through faster but localized updates of the resource metrics coupled

2.4 Resource Allocation and Interference Mitigation 19

with a slower but more system-wide change to the resource distribution and reuse

partitions.

CHAPTER 3

3 Related Technologies

Related Technologies

3.1 Inter-Cell Interference Mitigation .............................................................. 21


Networks ................................................................................................................. 25

After a brief overview of the fundamental concepts of wireless and cellular systems in

Chapter 2, this chapter focuses on the related work and standardization activities of

Inter-Cell Interference (ICI) mitigation techniques. First, the categories of ICI

mitigation techniques, the most well-known approaches to each category as well as

their advantages and limitations are outlined in Section 3.1. Then, Section 3.2 studies

three most representative ICI mitigation approaches. Their performances are compared

with the novel schemes proposed in Chapter 4 in subsequent Chapters 5 and 6.

It is known that effective reuse of resources in a cellular system can highly enhance

the system capacity. As mentioned in Section 2.2, with a smaller frequency reuse

factor (FRF), more available bandwidth can be obtained by each cell. So, in this sense

the classical FRF-1 deployment is desirable. However, with the usage of FRF-1, the

most UTs are seriously afflicted with heavy ICI, especially near the cell edge. And that

causes low cell coverage and inferior system capacity. The conventional method to

figure out this problem is by increasing the cluster-order, which can mitigate the ICI

efficiently, nonetheless at the cost of a decrease on available bandwidth for each cell.

This could result in restricted data transmission rate and lower system spectrum

efficiency in general, and would worsen in the case of unbalanced traffic distribution

among cells.

3.1 Inter-Cell Interference Mitigation

To seek for improving cell-edge performance while retaining system spectrum

efficiency of Reuse-1, the investigation of ICI mitigation techniques has become a key

focus area in achieving dense spectrum reuse in next generation cellular systems such

3. Related Technologies 22

as 3GPP-LTE, LTE-Advanced, and WiMAX. Through reduction of the ICI, one of the

following goals is desired: i) maximization of the overall network capacity, ii)

maximization of the cell-edge throughput, iii) enhancements in terms of both the

overall network capacity and the cell-edge performance. Based on approaches used,

mitigation techniques can be generally classified into three major categories, namely,

ICI cancelation, ICI randomization, and ICI coordination techniques.

3.1.1 ICI Cancelation

ICI cancelation techniques have been investigated and deployed with varying degrees

of success in terrestrial mobile networks for more than 20 years. The basic principle of

ICI cancelation techniques is to regenerate the interfering signals and subsequently

subtract them from the desired signal. Good overviews of historical approaches can be

found in [12] and the references contained therein. It has been successfully applied to

both CDMA systems and TDMA systems, as well as proposed within the 3GPP-LTE

standardization [7].

Various ICI cancelation techniques have been proposed in literature and they are

mainly categorized into two classes, filter-based approaches and Multi-User Detection

(MUD) [11]. Filter-based approaches try to mitigate ICI by means of linear filters and

interference models. In contrast, MUD directly includes the interfering signals in the

decoding process. This is done by jointly decoding the signal of interest and the

interfering signals, or by decoding and subtracting the interfering signals from the

signal of interest. ICI-cancelation techniques can also be jointly used with MU -

Multiple-Input Multiple-Output (MIMO) technique if the mobile terminals are

equipped with multiple-receive antennas.

From an implementation standpoint, interference cancelation does not require any

modifications of the system standard, making it an attractive technique. Although ICI

cancelation promises significant gains and its algorithms are mature, it is considered

mostly as a technique for the UL due to processing complexity (scales exponentially

with number of mobiles served at the BS) [13]. Furthermore, it requires exchange of

information in real time between BSs about every msec to maximize the system gain.

Besides, ICI cancelation can potentially improve channel performance to that of

AWGN if accurate channel estimation is available.

3.1.2 ICI Randomization

In contrast to ICI cancelation, which tries to eliminate interfering signals in the

received signal, ICI randomization, which is also known as ICI whitening or ICI

averaging, aiming at making ICI appear like background noise, i.e., it averages the

3.1 Inter-Cell Interference Mitigation 23

interference across the data symbols of a data block or the whole frequency band. The

approaches include Frequency Hopping (FH) and Interleave Division Multiple Access

(IDMA). FH ensures User Equipments (UEs) to access a range of channels rather than

a narrow set in a specific pattern so that interference effect is averaged out for all UEs.

It is widely applied in CDMA systems and, combined with interleaving, is applied in

FDMA/TDMA systems like GSM [7]. Both FH and IDMA-based schemes have also

been proposed within the 3GPP-LTE standardization [14][15][16]. Nevertheless, these

methods randomize the interference into ―White Gaussian Noise‖, which cannot

reduce interference in nature. Thereby, these approaches can hardly achieve a

substantial performance improvement.

3.1.3 ICI Coordination

ICI coordination techniques, also known as ICI avoidance, in OFDMA networks in

general and in the 3GPP-LTE system in particular have recently gained much attention

from both the academia and the standardization communities. In order to reduce ICI in

cellular networks, ICI coordination is to apply restrictions to resource management in

a coordinated way among neighboring cells. Resources for coordination can be space

(with beamforming antennas), time, frequency, code resources, or transmit power. And

restrictions can be, for instance, in the form of restrictions to what time/frequency

resources are available to the resource manager, or/and restrictions on the transmit

power that can be applied to certain time/frequency resources.

The system architecture that follows from the ICI coordination scheme can be mainly

divided into four major classes with respect to the degree of distribution [7], i.e.,

global, distributed, decentrialized as well as local ICI coordination schemes. Global

ICI coordination schemes require an omniscient central device with full system

knowledge. The global device is capable of acquiring the global system state in zero

time, and it can distribute scheduling and other decisions instantly to all nodes (e.g.,

BSs) in the network. Distributed ICI coordination schemes rely on one or more central

components, which exchange information potentially with all nodes in the network.

The central components thereby collect state information and distribute information

relevant for coordination to the network nodes. Coordinated Multi-point

transmission/reception (CoMP) as standardized in 3GPP LTE-Advanced is an example

[49]. Decentralized ICI coordination schemes are distinguished from distributed

schemes by there being no central entity involved in the coordination process. The

coordination is performed solely by communication and information exchange among

the network nodes, which all have equal rights. Lastly, local ICI coordination schemes

are based purely on information that is available locally in every BS. No coordination-

related communication and information exchange among neighboring BSs takes place.


Local state information includes data received from the mobile terminals, such as for

example pilot measurements or information about the strongest interferers.

Information exchange among BSs or with a central coordinator opens the way to much

larger possible performance gains, however at the expense of increased system

complexity. Especially, a global ICI coordination scheme is actually unfeasible in a

real network, although it may potentially attain the highest performance improvements

among all ICI coordination schemes. Local ICI coordination schemes, however,

require no separate network equipment and provide the most flexibility with respect to

the network design. The European Telecommunications Standards Institute (ETSI)

Digital European Cordless Telecommunications (DECT) system is an example of this

[8]. Therefore, they are generally most desirable from an operator’s point of

perspective. Albeit using local ICI coordination schemes may not achieve the highest

potential performance gain, they are easy to implementable. ICI mitigation by means

of local ICI coordination techniques is the focus of this monograph and will be deeply

discussed in Section 3.2 and Chapter 4. And the analytical and simulative performance

evaluation will be given in Chapter 5 and Chapter 6, respectively.

3.1.3.1 ICI Coordination techniques in 3GPP-LTE

Recently, the 3GPP has been completing most of the technical specifications for the

generation 3.5 in the 3GPP-LTE system. The technical targets of LTE include peak

data rates in excess of 300 Mbps, delay and latencies of less than 10 ms and manifold

gains in spectrum efficiency. Unlike the previous generations, LTE uses OFDM and

OFDMA as the baseline for modulation and multiple access schemes, respectively. [13]

gives an overview of the contemporary and forward looking ICI coordination

techniques for the 3GPP-LTE system, and provides a summary comparison of the

applicability, complexity and performance gains of these techniques.

Viable approaches include the use of power control, static Fractional Frequency Reuse

(FFR), Adaptive Fractional Frequency Reuse (AFFR), spatial antenna techniques such

as MIMO, Space Division Multiple Access (SDMA) and adaptive beamforming, as

well as recent innovations in decoding algorithms. Current mature approaches that

should be considered as baseline features for initial Uplink (UL) and Downlink (DL)

LTE deployments include power control, some form of FFR, preferably with

adaptation, as well as MIMO. SDMA is also relatively mature and can provide

significant aggregate capacity improvements for a small increase in complexity.

Beamforming technologies such as opportunistic spectrum access and organized beam

hopping show significant theoretical promise; however, as practical technologies for a

deployed LTE system they are still unproven. Newer decoding approaches such as

sphere decoding and dirty paper coding show considerable promise in terms of the

gain they can provide; however, the processing complexity of these approaches will


Networks

25

likely limit their use to the UL for the near future. Network power control and network

MIMO are inter-base station techniques that seem viable as features for LTE ICI

coordination evolution in the near future due to the potential performance gains and

the feasibility of their implementation at the BS. [13] also indicates that no single

approach will in itself provide complete ICI mitigation for an LTE implementation.

However, through combining some of these approaches may provide a robust N = 1

reuse capability in a heavily loaded LTE deployment. In the short term a combination

such as fractional power control and adaptive fractional frequency reuse based on

scheduling in high CINR regions could form the basis of a robust LTE ICI

coordination strategy. Longer term gains in ICI coordination performance could

potentially be achieved through the use of inter-base station network based algorithms,

including network MIMO, opportunistic and/or organized beamforming, and

distributed power control, as well as coding strategies such as sphere decoding or dirty

paper coding.

3.1.3.2 ICI Coordination techniques in IEEE 802.16m

IEEE 802.16m (WiMAX) standard utilizes similar ICI coordination techniques to that

applied in the 3GPP-LTE standard to mitigate ICI, including power control, AFFR,

smart antenna schemes such as MIMO and beamforming, etc. [17]. The exact details

and the relative emphasis of each technique differ across the two standards. For

example, 3GPP-LTE uplink power control may be configured to give a similar uplink

CINR distribution as possible in 802.16m; however, it requires usage of the closed

loop power update mechanism instead of the pure open-loop approach taken by

802.16m. Despite differences in certain implementation details, all approaches in both

standards are effective in reducing ICI and substantially improving the cell-edge

performance set forth by both standards.

3.2 Frequency Reuse Techniques for ICI Mitigation in Cellular OFDMA Networks

As mentioned in the preceding subsection, ICI cancelation approaches suffer from

heavy complexity overhead and barely limited amount of strong ICI can be cancelled.

Similarly, ICI randomization technique can hardly achieve substantial performance

improvements as ICI randomization randomizes the interference into ―White Gaussian

Noise‖, which cannot reduce interference in nature. Among all three ICI mitigation

categories, only ICI coordination technique may potentially attain significant

performance improvements and has become very important to mitigate ICI in next

generation wireless communication networks. In the remainder of this monograph, out


of regard for feasibility in the reality, local ICI coordination approaches with static

frequency resource partitioning are discussed. In fact, the local ICI coordination

mechanisms currently studied by the 3GPP build on markedly lower complexity

heuristics. From a system design perspective, ICI coordination mechanisms without

(or with slow) inter-cell communication—building on some pre-configured (simple)

OFDMA resource block allocation rule—are particularly attractive.

To take aim at improving cell-edge performance while retaining system spectrum

efficiency of Reuse-1, several local ICI coordination solutions [7] [18]-[23] [25]-[29]

[30]-[32] have been proposed recently. Among them, the most representative

approaches are the static Fractional Frequency Reuse (FFR) [18]-[23], the Soft

Frequency Reuse (SFR) [25]-[29] and the Incremental Frequency Reuse (IFR) [30]-

[32]. All these methods concentrate on attaining high system spectrum efficiency with

small FRF and efficient reduction of ICI simultaneously.

3.2.1 Static Fractional Frequency Reuse

One type of ICI coordination techniques, known as Fractional Frequency Reuse

(FFR) or Reuse Partitioning (RUP), aims at effectively mitigating ICI by applying

various FRFs to UTs situated in different regions in each cell [18]-[23]. Normally, in a

multi-cellular environment, the closer those UTs located to the cell edge, the stronger

they are afflicted with the ICI. Thus, FFR scheme, as shown in Figure 3.1,

geographically divides each cell in a system into two or more concentric zones. UTs

close to their BS with the best received signal quality in the inner zone are allowed to

use frequency resources with the smallest FRF, whereas those far away from their BS

in the outer zone are assigned frequency resources with the largest FRF. In this way,

the ICI in the outer regions can be simply mitigated by allocating orthogonal

frequency resources. And UTs, who are less susceptible to ICI in the inner regions,

FRFlow

FRFmiddle

FRFhigh

Figure 3.1: Concentric zones in each cell in a system.


Networks

27

have more frequency resources than using static Reuse schemes with FRF > 1 so that

the spectrum efficiency can be enhanced.

The FFR approach with its plain design can be realized easily with low system

complexity. However, the design also leads to a loss in frequency selective gain and

lower spectral efficiency compared to the Reuse-1 scheme, since for serving UTs in

the regions with higher FRFs there is only a limited portion of the total frequency

Reuse 7 zone

Reuse 3 zone

Reuse 1 zone

P(f)

f

P(f)

f

P(f)

f

bandwidth B

B_reuse7 B_reuse3 B_reuse1

available subband

for one cell

available subband

for one cell

available subband

for one cell

(a) Exclusive FFR

Reuse 7 zone

Reuse 3 zone

Reuse 1 zone

P(f)

f

P(f)

f

P(f)

f

B_reuse7 = B_reuse3 = B_reuse1 =

bandwidth B

available subband for one cell

available subband

for one cell

available subband

for one cell

(b) Inclusive FFR

Figure 3.2: Frequency partitioning example for comparing exclusive and inclusive FFR scheme in a

cellular system based on FRF =7 for cell-edge users, FRF =3 for cell-middle users and FRF =1 for cell-center users.


spectrum available for scheduling.

When taking into account the way of spectrum sharing among different FRF-zones in

each cell, FFR schemes presented in [18]-[23] can be distinguished into two classes,

namely, exclusive and inclusive reuse partitioning (RUP). Figure 3.2 gives an example

to compare the spectrum partitioning of both FFR schemes. In exclusive FFR, as

illustrated in Figure 3.2a, the overall bandwidth B is split into two or more mutual

disjoint subsets (different-color blocks in the figure: B_reuse7, B_reuse3 and B_reuse1). Each

subset serves the requirements from a specific FRF-zone. Unlike exclusive FFR, the

whole spectrum band B in inclusive design is allowed to be used by all different FRF-

zones, which results in an overlapping reuse effect, as shown in Figure 3.2b. From the

figures, it should be noticed that with inclusive FFR the whole system bandwidth B is

available for all cells, whereas using exclusive FFR only a part of subbands can be

used in each cell.

In FFR, no power adaptation mechanism is considered for different FRF zones. All

transmissions in a system are applied with equal transmission power. In DL, each BS

evenly dispenses its total power over the available subbands of its cell. Since exclusive

FFR owns less bandwidth than inclusive FFR to utilize, packets in exclusive FFR can

be delivered with higher transmission power than in inclusive FFR.

Table 3.1 Comparison between exclusive and inclusive FFR schemes.

exclusive FFR inclusive FFR

Power allocation

for each user higher lower

ICI mitigation

capability better worse

Available bandwidth 31 1

63 2B B

(for the example in Figure 3.2a)

B

Need resource partition

among various FRF-zones yes no

Common advantage Low complexity

Common limitation Loss of frequency selective gain and lower spectral efficiency

compared to the Reuse-1systems


Networks

29

When comparing exclusive FFR and inclusive FFR approaches, no one has an

overwhelming superiority over the other one. For example, an exclusive FFR may reap

more benefits than an inclusive FFR in ICI mitigation due to higher power allocation

and exclusive reuse mechanism applied on the active users. However, on the other

hand, how to determine the amount of resources (a fraction of the total available

bandwidth B) devoted to each FRF-zone is a critical issue for the exclusive FFR

scheme. By adjusting the share for each FRF-zone, a tradeoff between increased

spectrum utilization and CINR improvement exists. This bandwidth division problem

can be overcome by the inclusive FFR scheme, since the whole spectrum B is

available for each FRF-zone and can be used by every cell in an overlapping way.

Moreover, due to its full utilization of the whole bandwidth B, higher spectrum

efficiency is expected to be achieved by the inclusive FFR than by the exclusive one.

However, the CINR performance of the inclusive FFR is inferior to that using the

exclusive FFR [7]. This arises mainly from two reasons: one is that UTs of a certain

FRF-zone in a cell are not only interfered by the transmissions from the same-type

FRF-zones in the neighboring cells, but also suffer from the additional ICI caused by

the transmissions from the other FRF-zones in the adjacent cells. This leads to higher

ICI compared to the exclusive FFR. Another reason is that relatively lower

transmission power used in the inclusive FFR causes relatively lower carrier signal

value. Both factors result in a disadvantage of the CINR performance of the inclusive

FFR, which may not compensate its advantage in the spectrum utilization and cannot

substantially improve the system capacity as well as the cell-edge performance.

3.2.2 Soft Frequency Reuse

Focus on the two limitations of the inclusive FFR mentioned above, a so-called Soft

Frequency Reuse (SFR) scheme was first introduced by Huawei in [25] in 2005. The

SFR scheme, which has been adopted in the 3GPP-LTE system [24]-[25], addresses

the challenge issue (namely, to improve cell-edge performance while retaining system

spectrum efficiency of Reuse-1) by increasing FRF and transmission power for cell-

edge users, so that the ICI from neighboring cells to these users can be alleviated, and

thereby their performance is improved.

The basic idea of the SFR scheme is applying FRF of 1 to Cell-Centre Users (CCUs)

and FRF of 3 to Cell-Edge Users (CEUs) as illustrated in Figure 3.3. One third but

only one third of the whole available bandwidth named Major Segment can be used by

CEUs. Yet on this Major Segment, packets are sent with higher transmission power.

To actualize bigger FRF for CEUs, Major Segments among directly adjoining cells

should be orthogonal. In opposite to CEUs, CCUs may access the entire frequency

resources, however, with lower transmission power to avoid yielding too much ICI to


the co-channel users in the neighboring cells. Besides, on the Major Segment, CEUs

always take priority over CCUs at resources access.

In DL, since BSs are responsible for all transmissions occurring on the whole available

bandwidth and the maximal transmission power of each BS is assumed as fixed, the

assigned transmission power on each subchannel for different types of users is

therefore related with four factors: the maximal transmission power of a BS PBS, the

total number of subchannels in a system S, the power ratio of high power level for

CEU-transmissions to lower power level for CCU-transmissions α, as well as the value

of FRF δ for CEUs (in this monograph, δ is set as 3, may also be set to other values)

_

_

~ ( , , , ).CCU DL

BS

CEU DL

P

P SP

(3.1)

The total transmission power used by a BS for DL streams can be calculated as a

summation of the total power for all CCU-transmissions and the total power for all

CEU-transmissions

_ _

1 1

_ _

( )

(1 )

BS CCU DL CEU DL

CCU DL CEU DL

S SP S P P

S P P

(3.2)

with

_ _CEU DL CCU DLP P (3.3)

P(f)

P(f)

P(f)

f

f

f

Cell A

Cell C

Cell B

major subchannel

normal subchannel

f5

f2

f11

f7

f4

f5

f12

Cell

C

Cell

B

Cell

B

Cell

B

Cell

A

Cell

C

Cell

C

F3

F3

F3

F4

F4

F4 F5

F5

F5 F6

F6

F6F2

F2

F2

F1

F1

F1

F1 .. F6 F1 + F2

F5 + F6F3 + F4

Figure 3.3: Concept of the SFR scheme in a cellular system based on FRF =3 for CEUs and FRF =1

for CCUs.


Networks

31

, where S/δ stands for the number of available subchannels for the CEU-transmissions.

As a result, the DL power allocation for each type of users can be represented as

_ 1 1

_ 1 1

(1 )

.

(1 )

BSCCU DL

BSCEU DL

PP

S

PP

S

(3.4)

When the power ratio α = 0, no power will be assigned to the CEU-traffic, and all S

subchannels will be used to transport CCU-traffic with a power value of BSP S for

each subchannel. The SFR system is thereby a Reuse-1 system, yet no CEU-traffic

will be served. In this case, high system capacity may be achieved, however at the

expense of abandoning all CEUs. When the power ratio α = 1, PCCU_DL will have a

same value as PCEU_DL of BSP S . Note that this value is equivalent to the PCCU_DL in α

= 0 case, but the SFR system is now a typical inclusive FFR system, in which CCUs

are served with FRF of 1 while CEU-traffic is served on the Major Segment with FRF

of δ. When the power ratio α → ∞, PCCU_DL and PCEU_DL will converge at 0 and 1( )BSP S , respectively. It means only Major Segments are at work, and the SFR

system is then equivalent to a Reuse-δ system, but using a scheduler always favoring

the CEU-traffic.

Unlike DL, in UL, the power allocation for each type of users is simply interrelated

with two factors: the maximal transmission power of a UT PUT and the power ratio α

of high power level for CEU-transmissions to lower power level for CCU-

transmissions

_

_

~ ( , ).CCU UL

UT

CEU UL

P

PP

(3.5)

Each CEU concentrates its full power on one subchannel to send data, whereas each

CCU makes use of its location advantage to deliver packets with lower power, but may

occupy at maximum α subchannels at the same time

1

_

_ _

.

UTCCU UL UT

CEU UL CCU UL UT

PP P

P P P

(3.6)


Hence, for UL, the power ratio α is only allowed to vary within an interval (0, ∞).

The SFR scheme intends to abate ICI from neighboring cells by increasing FRF and

transmission power for the CEUs, and thereby to improve their performance and

enhance the whole system capacity. However, recently, the performance gain of the

SFR has been questioned in [33], [34] and [30]. In [30] the authors have given us a

study result that with the usage of the SFR scheme the cell throughput is even inferior

to the Reuse-1 system when the loading factor is over 0.5. The reason mentioned in

[30] is that in a SFR system at most one third of the subchannels can be used to

transmit data with higher power while the remaining two third subchannels work with

lower power, which causes an overall throughput loss. In other words, the SFR

ameliorates performance of the CEUs at the expense of degrading the overall cell

capacity. Nevertheless, this allegation is only correct when the boundary between the

CCU-zone and the CEU-zone in a cell is set quite close to the cell borderline. In fact,

later in the subsequent Chapters 5 and 6, we will disclose in both analytical and

simulative ways that the SFR is able to perform better than the Reuse-1 when suitable

range definitions for dividing the CCUs and CEUs are chosen.

But for all that, taking a view of the SFR design, some intrinsic limitations are exposed.

For one thing, how to define the borderline to divide cell regions for CCUs and CEUs

is a key issue in the SFR scheme. With an inappropriate zone definition, especially in

case the area for CCUs is a little wide defined, system performance will be severely

deteriorated. This effect will be shown in the performance evaluation in Section 6.2.2.

Secondly, in general, there are more CEUs than CCUs in a cell, since the outer surface

area is much larger than the inner part (see Figure 3.4). But, in SFR scheme CEUs

have maximum one third of the entire bandwidth to utilize, which results in the

unfairness between CCUs and CEUs, and lower spectrum efficiency caused by idle

f4

P(f)

P(f)

P(f)

f

f

f

Cell A

Cell C

Cell B

occupied idle

Figure 3.4: In a SFR system, less available resources for CEUs while more for CCUs result in

unfairness between CCUs and CEUs, as well as lower spectrum reuse efficiency.


Networks

33

channels and a waste of resources. Next, as shown by the sketch in Figure 3.5, more

ICI could happen even in a low-traffic-load situation, while there are still subchannels

idle and underutilized in the system. This is because the resource allocation of all cells

via the SFR scheme starts always from the first subchannel up. Lastly, Although FRF

of 3 and higher transmission power are applied to the CEUs, they are still grievously

interfered by the co-user transmissions in the neighboring cells when much traffic is

there, as illustrated in Figure 3.6. The reason is that the SFR applies an inclusive RUP,

which means the reuse-factor-of-3 Segment is not exclusively used by the CEUs, but

simultaneously co-used or re-used by the CCUs in the neighboring cells.

P(f)

P(f)

P(f)

f

f

f

Cell A

Cell C

Cell B

occupied idle

Figure 3.5: More co-channel interferences even at low load traffic situation with the usage of the SFR scheme.

f5

f2

f11

f7

f4

f5

f12

Figure 3.6: CEUs are grievously interfered by re-users in the neighboring cells since SFR is a

development design based on inclusive reuse scheme.


3.2.3 Incremental Frequency Reuse

Aiming at the limitations of the SFR scheme mentioned above, Ki Tae Kim et al. came

up with a new design referred as Incremental Frequency Reuse (IFR) scheme [30],

which can reduce the ICI effectively in the case of a low-offered traffic, and thereby

betters the overall system capacity.

The only difference between the IFR design and the Reuse-1 is that adjacent cells in an

IFR system start dispensing resources to their users from different points of the

available bandwidth, whereas the Reuse-1 and the SFR allocate resources always from

the first subchannel. Figure 3.7 exemplifies the operation method of the IFR scheme

for a cellular system with 3 various types of neighboring cells. Cells of type-A occupy

resources from the first subchannel, whilst cells of type-B from the one-third point of

the whole bandwidth, and cells of type-C from the two-thirds point of the bandwidth.

They allocate consecutive subchannels successively along with increasing traffic load

until the entire bandwidth is used up. A detailed description of the IFR scheme and its

variations can be found in [30]-[32].

Through such a resource assignment mechanism, the IFR scheme can overcome part

of the aforementioned limitations by applying the SFR scheme, namely, the improper

zone-definition problem, the unfairness and low spectrum reuse efficiency problem

and the problem of more ICI even with low-load traffic in a system. The ICI generated

by adjacent cells can be avoided completely in low-traffic situations, since all

transmissions in these cells work on different subbands and frequency reuse of the 1st-

tier neighboring cells doesn’t occur when the loading factor is below 0.3. In this case,

the whole system performs just like a Reuse-3 system. In essence, by means of the IFR

Cell B

Cell C

Cell C

Cell A

Cell B

Cell C

Cell B

1

f

Cell A

Cell B

Cell C

Start point of subchannel allocation

2 3 4 5 6 7 8 9

7

f

8 9 1 2 3 4 5 6

4

f

5 6 7 8 9 1 2 3

Figure 3.7: Operation policy of the IFR scheme in a cellular system with 3 various types of

neighboring cells.


Networks

35

scheme, the system operates with increasing traffic load like sliding from a Reuse-3

system to a Reuse-1 system.

Although some limitations of the SFR scheme can be eliminated by using the IFR

scheme, it only performs better under low traffic load to a system. When the loading

factor is greater than 0.3 (with moderate traffic load), though the IFR surpasses the

Reuse-1 scheme, it is inferior to the SFR scheme. With the help of its static

configuration, the IFR scheme disperses the ICI over the whole bandwidth, but with

increasing traffic in the system, the CEUs are still interfered severely. In a full-load

situation, the IFR scheme even cannot perform better than the Reuse-1 scheme. That is

to say, the system capacity cannot be substantially improved by the IFR scheme.

3.2.4 Summary and Conclusion

In this section, three most well-known RRM techniques—the FFR scheme, the SFR

scheme and the IFR scheme—for ICI mitigation in cellular OFDMA networks are

outlined and discussed. All of these schemes have a common advantage that they don’t

need global coordination among neighboring cells and thereby can reap significant

complexity reduction benefits. Nevertheless, each of them also has its specific

advantages and limitations, and none of them can absolutely dominate over the others.

The exclusive FFR may reap more benefits than the inclusive FFR with respect to the

ICI mitigation for the CEUs due to higher power allocation and exclusive reuse

mechanism applied on the active users. However, on the other hand, how to determine

the amount of resources—a fraction of the total available bandwidth—devoted to each

FRF-zone is an open issue for the exclusive FFR scheme, but it is not a problem for

the inclusive FFR scheme. Moreover, because of its ability of full utilization of the

whole bandwidth, higher spectrum efficiency is expected to be achieved with the

inclusive FFR than with the exclusive one. In respect to the SFR scheme, which has

been adopted in the 3GPP-LTE systems, the cell capacity might be advanced with a

careful range definition for partitioning CCUs and CEUs, yet the cell coverage can

still not be guaranteed. And lastly, the IFR scheme, proposed by Ki Tae Kim et al. in

2008, can effectively reduce the ICI and surpasses the SFR and the Reuse-1 only when

the system is loaded with low offered traffic.

CHAPTER 4

4 Enhanced Fractional Frequency Reuse Design

Enhanced Fractional Frequency Reuse Design

4.1 EFFR ............................................................................................................. 37

4.2 EFFR-Advanced ........................................................................................... 48

4.3 EFFR-Beyond ............................................................................................... 50

In the previous Chapter 3, three well-known resource allocation and reuse

approaches—the static FFR, the SFR as well as the IFR schemes—are introduced to

alleviate ICI in multi-cellular OFDMA networks. Based on thoroughly analyzing their

advantages and limitations respectively, in this chapter a novel ICI mitigation design

referred as Enhanced Fractional Frequency Reuse (EFFR) scheme and its two

derivatives, the EFFR-Advanced scheme and EFFR-Beyond scheme, will be

contributed for a better fulfillment of the goals, namely, to enhance the mean system

capacity while restraining the ICI at cell edge. Moreover, since solutions with low

system complexity and flexible spectrum usage are desirable, systems with distributed

RRM will be taken into account.

4.1 EFFR

The discussion about advantages and limitations of the static FFR, the IFR and the

SFR schemes in the preceding chapter motivates to propose a new design named

Enhanced Fractional Frequency Reuse (EFFR) scheme, which attempts to retain the

advantages of these approaches while avoiding their limitations, and seeks for a further

progress in attaining both coverage and higher system capacity in any traffic-load

situation.

4.1.1 Design Requirements

The EFFR scheme is designed to meet the following requirements:

Support flexibility with non-uniform user or traffic distribution

4. Enhanced Fractional Frequency Reuse Design 38

Support adaptation to time varying traffic conditions

Exploit possibility for self-setting up preferable reuse combinations

No need for the resource coordination among different BSs in radio network

controller (RNC) in the fixed resource allocation method

Applicable for high FRF systems

Low system complexity

4.1.2 Concept of the EFFR Scheme

The objective of the proposed EFFR architecture is to improve system capacity while

bettering spectrum efficiency at the cell edge. This can be achieved by being based

upon effectual mitigation of unwanted ICI for CEUs, further maximizing of the

opportunities for the other users to choose suitable resources (time share and

frequency share, respectively) to reuse.

4.1.2.1 Reuse Partition

The underlying principle of the EFFR design is similar to the SFR scheme to reduce

CINR level for those UTs (CCUs) that already have more than adequate transmission

quality while offering greater protection to those UTs (CEUs) that require it, but based

on an exclusive RUP (see Figure 3.2a in Section 3.2.1). Figure 4.1 gives an example of

the EFFR scheme which like the SFR scheme defines 3 cell-types for directly

contiguous cells in an OFDMA cellular system. According to different cell-types, the

EFFR scheme divides the whole available bandwidth of each cell-type into 2

Segments, namely, the Primary Segment as indicated in the right part of Figure 4.1

with thick border and the Secondary Segment comprising all reuse subchannels of the

available spectrum not assigned to the Primary Segment. The Primary Segments

among different cell-types should be orthogonal that also leads to the orthogonality of

the Primary Segments among directly adjoining cells in the system, as show in

Figure 4.1. A part of the Primary Segment is exclusive reuse-3 subchannels, which are

preferentially used by the CEUs with higher transmission power. They can only be

reused by the users in the same type cells and consequently cannot be co-used by the

directly neighboring cells, thus the ICI among them can be decreased. The remaining

subchannels are all reuse-1 subchannels allowing to be used with lower transmission

power. Here it has to be noted that part of these reuse-1 subchannels belong to the

Primary Segment, the others constitute the Secondary Segment. The reuse-1

subchannels in the Primary Segment of a cell-type are at the same time part of the

Secondary Segments belonging to the other two cell-types. Each cell can occupy all

4.1 EFFR 39

subchannels of its Primary Segment at will, whereas maybe only a part of resources in

the Secondary Segment can be used by this cell in an interference-aware manner.

4.1.2.2 Scheduling

As we have clarified in Section 3.2.3, the major advantage of the IFR scheme is that

through using various resource allocation sequences for different cell types, it

disperses the ICI over the whole bandwidth and can thereby effectively reduce the ICI

in low and moderate traffic load situations. This mechanism is also adopted in the

EFFR scheme and works as follows: Each cell starts its spectrum allocation from its

Primary Segment up. When the whole Primary Segment is used up and still frequency

resources are desired, the available reuse-1 subchannels in the Secondary Segment will

be further assigned, yet with a cell-type-specific segment allocation sequence.

In order to make the operation principle intelligible, we give an example based on the

EFFR reuse partitioning illustrated in Figure 4.1:

- Frequency segment allocation sequence for type-A cells:

(F1+F2+F3) → (F5+F6) → (F8+F9)

- Frequency segment allocation sequence for type-B cells:

(F4+F5+F6) → (F8+F9) → (F2+F3)

f

Cell A

Cell B

Cell C

Primary Segment

f

f

reuse-3 subchannel for each type of cell

reuse-1 subchannel in the Primary Segment

Quality good for me,

try to occupy

idle subchannel

reusing secondary subchannel after CQI estimation

P(f)

P(f)

P(f)

f5

f2

f11

f7

f4

f5

f12

Cell A

Cell C

Cell B Cell B

Cell B

Cell CCell C

F2

F2

F2

F1 F3

F3

F3

F9F5

F4 F5

F5

F6

F6

F6 F7

F8

F8

F8

F9

F9

F2 + F3 + F5 + F6 + F8 + F9

= FSUM - F1 - F4 - F7

F1 F4 F7

Figure 4.1: Concept of the EFFR scheme in a cellular system based on exclusive partitioning of

reuse-3 subchannels and reuse-1 subchannels in the Primary Segment, as well as interference-aware

reuse on the Secondary Segment.


- Frequency segment allocation sequence for type-C cells:

(F7+F8+F9) → (F2+F3) → (F5+F6)

It can be seen that in this way the ICI generated by directly adjoining cells can be

completely avoided in low traffic-load situation, because frequency reuse of the 1st-

tier-neighboring cells does not occur in this case, and the whole system works just like

a Reuse-3 system. Under middle offered traffic situation, the ICI will be scattered over

the Primary Segment and the newly added reuse-1 spectrum segment. Note that the

newly added spectrum segments among directly neighboring cells are also disjoint,

which results in less ICI. And this appears preferable to the mechanisms used in the

Reuse-1 and SFR systems. Under high offered load spectrum of the Secondary

Segment will be used based on CQI estimation in adjacent cells.

The detailed scheduling algorithm for DL traffic is relatively simple and slightly

differs from that in UL:

Step 1. The exclusive reuse-3 subchannels are preferentially assigned to the

CEU-traffic using Proportional Fair (PF) scheduling strategy. If there are still

resources remaining after all CEU-connections are satisfied, they will be

assigned to those CCU-connections which CCUs locate relatively far away

from the BS (or with relatively poor CINR values). All traffic on the reuse-3

subchannels will be sent with higher transmission power, no matter for which

kind of users, whereas all reuse-1 subchannels deliver packets always using

lower transmission power.

Step 2. When the reuse-3 subchannels are exhausted, the remaining reuse-1

subchannels (including reuse-1 subchannels in the Primary Segment and

secondary reuse-1 subchannels) are then allocated to residual unsatisfied

connections also using PF strategy until

demands of all connections are met, or

the entire available frequency resources are occupied, or

the residual resources cannot be validly used.

A difference in resource occupation between the primary reuse-1 subchannels

and the secondary reuse-1 subchannels is that the primary reuse-1

subchannels can be arbitrarily scheduled, whereas the secondary reuse-1

subchannels can only be occupied in interference-aware manner.

The UL resource allocation of the EFFR scheme works as follows:

Step 1. This step is identical with that in DL.

4.1 EFFR 41

Step 2. When the reuse-3 subchannels are completely allocated, the

remaining reuse-1 subchannels in the Primary Segment are then scheduled to

residual unsatisfied users using maximum throughput strategy until demands

of all users are met or the entire Primary Segment is occupied. Applying

maximum throughput strategy means that UTs having superior link quality are

preferentially served. The aim of operating like this in UL is to achieve that

UTs occupying the primary reuse-1 subchannels should stay on their allocated

resources and thereby keep the interference produced by the primary reuse-1

subchannels relatively regular in view of the adjacent cells. Since the primary

reuse-1 subchannels are simultaneously the secondary subchannels that may

be used by the adjacent cells, regular utilization of the primary reuse-1

subchannels helps UTs in the adjacent cells to predict the channel quality more

accurately and more validly use the secondary reuse-1 subchannels with

adequate PHY mode.

Step 3. If still resources are requested, the available reuse-1 subchannels in

the Secondary Segment will be scheduled to adequate users by applying

interference-aware operation, using the PF strategy. Since resource occupation

on the Secondary Segment is actually to reuse the primary subchannels

belonging to the directly adjoining cells, and in order not to disturb the

primary transmissions, the reuse occupation on the Secondary Segment should

base on an appropriate channel quality estimation method and a predefined

CINR-threshold. In this way, not only the primary reuse-1 subchannels can be

most effectively used, but the secondary reuse-1 subchannels are also brought

into full play, so that an enhancement of the whole system capacity can be

expected.

4.1.2.3 Transmission Power Allocation

As set forth in Section 3.2.2, each BS in a system is assumed to have a constant

maximal transmission power, and responsible for the power allocation of all DL

transmissions distributed on its accessible bandwidth. Hence, the assigned

transmission power on different types of subchannels is related with five factors:

1_

_

~ ( , , , , ).reuse DL

BS

reuse DL

P

P S MP

(4.1)

, where PBS is the maximal transmission power of a BS. S represents the total number

of subchannels in a system, which is, however, in most cases not completely available

for each cell. M stands for the number of available dedicated higher FRF subchannels

in the Primary Segment. Note that according to various traffic- or UT-distributions in a


system M can be adjusted in the EFFR design, whereas it is fixed defined as S/#FRF in

a SFR system. α denotes the power ratio employed on a reuse-δ subchannel to a reuse-

1 subchannel. And lastly, δ indicates the value of FRF for CEUs (in this monograph, δ

is set as 3, can be set as any values with δ ≥1).

From Figure 4.1, it can be seen that EFFR divides the whole frequency bandwidth

equally into δ Primary Segments for a cluster of δ cells. In this way, the orthogonality

of Primary Segments among δ different types of cells in the neighborhood can be

realized. The Primary Segment of each cell consists of M (= 1) exclusive reuse-δ

subchannels and N (= 2) reuse-1 subchannels

SM N

(4.2)

, whereas ( 1) N (= 4) reuse-1 subchannels constitute the Secondary Segment. As a

result, there are a total of M N (= 7) subchannels (i.e., about 78% of the whole

system bandwith) available for each EFFR cell. Two trivial special cases should be

noted: When 0N is chosen, which means no reuse-1 subchannels defined and all

UTs only have M S reuse-δ subchannels to utilize, the EFFR system is therefore a

quasi Reuse-δ system using minimum throughput scheduling strategy, as the CEU-

connections with relatively poor CINR values are preferentially served. When

N S is chosen (i.e. 0M ), which means the full set of bandwidth is available, in

this case the EFFR system is equivalent to a Reuse-1 system. Please note that the

EFFR scheme works for any suitable number S of subchannels, e.g., S = 9, 12, 15, etc.,

namely, multiple of δ, see for example Table 6.2.

The BS in each cell allots its transmission power on the accessible M reuse-δ and

N reuse-1 subchannels, respectively

_ 1_ .BS reuse DL reuse DLP M P N P (4.3)

As

_ 1_reuse DL reuse DLP P (4.4)

, the power allocation on each type of subchannel can be calculated as

1_

_

.

BSreuse DL

BSreuse DL

PP

M N

PP

M N

(4.5)

Derived from Eq. (4.2), N can be represented by S, δ and M:

4.1 EFFR 43

.S

N M

(4.6)

Thus, Eq. (4.5) can be finally formulated as

1_

_

( )

.

( )

BSreuse DL

BSreuse DL

PP

M S

PP

M S

(4.7)

When having a special case in consideration, where α = δ (for example, α = δ = 3 is

chosen for assessing the analytical performance of both SFR and EFFR schemes in the

succeeding Chapter 5), the power allocation applied on each type of subchannel in an

EFFR system is then:

1_

_

.

3

BSreuse DL

BS BSreuse DL

PP

S

P PP

S S

(4.8)

Eq. (4.8) shows that in case α = δ, both Preuse-1_DL and Preuse-δ_DL are not related to the M

and δ anymore. Since using the EFFR scheme, any cell-type (e.g., cell-type A in

Figure 4.1) can use only part of all reuse-δ subchannels (1 ) in the system. That is to

say, it is not allowed to use the remaining ( 1) reuse-δ subchannels dedicated to

the other two cell-types (e.g., cell-type B and C in Figure 4.1). This way, the power

allotted to the available reuse-δ subchannels Preuse-δ_DL can be δ-multiple (tripled when

δ equals 3) without decreasing the transmission power Preuse-1_DL for the other available

reuse-1 subchannels. As a result, as long as α = δ, Preuse-1_DL can keep constant with the

same power level as in the Reuse-1 situation. Otherwise, Preuse-1_DL varies along with

Preuse-δ_DL in various (α, δ) cases.

Note that when using the same power ratio α in both EFFR and SFR schemes, the

maximal assignable power for both, reuse-1 and reuse-3 subchannels in EFFR (except

the case M = 0) may always be higher than in a SFR system, respectively. This is

because the SFR scheme possesses the full set of system bandwidth to utilize, whereas

the EFFR due to applying the exclusive higher reuse factor for the CEUs can only

access part of the whole bandwidth unless no subchannel (M = 0)is reserved for higher

FRF area users. This results in SFR always owning more reuse-1 subchannels and


reuse-3 subchannels than the EFFR scheme. On the other hand, in EFFR, besides

altering the power ratio α, through adjusting the number M of reuse- δ subchannels, the

available subchannels and the corresponding power allocation for different types of

users can also be regulated, which is impossible for the SFR scheme owing to its

inherent fixed design. Hence, from this point of view the EFFR scheme seems more

flexible than SFR to adapt to various wireless environments.

The power allocation for the UL subchannels in EFFR is the same as the UL SFR. The

power assigned for each type of subchannels is simply interrelated with two factors:

the maximal transmission power of a UT PUT and the power ratio α of the high power

level applied on each reuse-δ subchannel to the lower power level for the reuse-1

subchannel

1_

_

~ ( , ).reuse UL

UT

reuse UL

P

PP

(4.9)

UTs (no matter CCUs or CEUs), who use reuse-δ subchannels to transmit data, will

contribute its full power on one subchannel, whereas UTs using reuse-1 subchannels

take advantage of its location to deliver packets with lower power, yet may occupy at

maximum α subchannels simultaneously

1

1_

_ 1_

, (0, ).

UTreuse UL UT

reuse UL reuse UL UT

PP P

P P P

(4.10)

Although Eq. (4.10) is the same as Eq. (3.6) for the SFR UL power allocation, the total

system power consumption of the UL SFR is much larger than that for the UL EFFR

due to more bandwidth available in the SFR scheme, unless the number of exclusive

reuse-3 subchannels is set to M = 0 in the EFFR scheme.

4.1.2.4 Signal-to-Interference-and-Noise Ratio Estimation

UTs in a cell act on the Secondary Segment as guests, and occupying secondary

subchannels is actually reuse the primary subchannels belonging to the directly

adjoining cells, thus reuse on the Secondary Segment by each cell should conform to

two rules:

monitor before use and

resource reuse based on CINR estimation.

4.1 EFFR 45

Each cell listens on every secondary subchannel all the time. And before occupation, it

makes CINR evaluation according to the gathered Channel Quality Information (CQI)

and chooses resources with best estimation values for reuse. If all available secondary

resources are either occupied or not good enough to a link, this cell will give up

scheduling resources for this link. This way, resource wasting cases can be

efficaciously avoided. For instance, some resources maybe not reusable for a link but

can be reused by other links. And another merit gained thereby is that it will not

generate excessive ICI for the neighboring cells which would degrade their

performances. So, an upgrade of spectrum efficiency is expected by using the

interference-aware-reuse mechanism on the Secondary Segment.

On the other hand, all above elucidation is based on a precise CINR estimation. Since

an improper modulation and coding scheme (PHY mode) selection due to a bad CINR

estimation would lead to either higher packet loss rate or lower spectral efficiency, and

thereupon wastes precious resources. Hence, to have a reliable CINR estimation is a

crucial factor for maximizing system spectrum efficiency.

4.1.3 Distinctions between the EFFR and other ICI Mitigation Schemes

The EFFR scheme owns mainly the following salient features, which are typically

different to the SFR scheme and the IFR scheme:

Since users close to the cell edge are very susceptible to ICI, the reuse-3

subchannels in the Primary Segment of each cell are exclusively available for

the users (only CEUs in most cases, but can also include CCUs, when all

CEU-connections are satisfied.) of the respective cell. This means real reuse-3

is applied on some subchannels per cell, and not the whole bandwidth is

available in a cell.

In order to advance spectral efficiency, users which are allotted shares of the

reuse-3 subchannels, transmit packets with higher power, whether they are

CCUs or CEUs. In contrast, to reduce excessive ICI to the neighboring cells

and avoid unwanted power wasting, packets will be sent on a reuse-1

subchannel always with lower power.

Allocation of reuse-1 subchannels in the Secondary Segment is not blindly

carried out, but in an interference-aware way according to CINR estimation.

In the Primary Segment, unsatisfied users, whether they are CCUs or CEUs,

have the same chance to get resources in the Secondary Segment, if they can

find usable resource in accordance with CINR estimation.


4.1.4 Expected Benefits by Using the EFFR Scheme

Now we would like to summarize the benefits by using the EFFR schemes compared

to the aforementioned SFR and IFR schemes. Just as explained in the foregoing

chapter, the merit by using the IFR scheme can be reflected in low and moderate

traffic load situations. Its capability of scattering ICI over the whole available

bandwidth make it perform better than the Reuse-1 and the SFR schemes when low to

moderate traffic exists in a system. This mechanism is adopted in the EFFR scheme so

that the system performs like a Reuse-3 system at low traffic load, and the ICI can also

be significantly alleviated in middle-loaded situation. Besides, the same problem in a

SFR system can also be resolved by the ICI dispersion mechanism introduced by the

IFR. Hence, we have implemented this mechanism in both EFFR and SFR schemes in

our simulator (refer to Appendix A).

Nevertheless, with increasing traffic in the system, the ICI among neighboring cells

grows. And, in high- and over-load situations, the IFR scheme resembles a Reuse-1

scheme. Hence, under these circumstances, we only compare the EFFR and the SFR

designs. As we have mentioned in Section 3.2.2, the performance of a SFR system

depends strongly on the accurate division of the CCU- and CEU-zones. Either a too

large or too small definition of the CCU- or CEU-zone would lead to severe

deterioration of the system capacity.

In case the CCU-area is less than adequate assumed, CCU traffic requirements are

reduced thanks to the smaller CCU-zone definition. However, the available bandwidth

for the SFR CCUs by definition is two times larger than that for the SFR CEUs, which

is impossible to be changed as this is an intrinsic feature of the SFR design. Besides,

the SFR CCUs can co-use the Major Segment with the SFR CEUs, but the SFR CEUs

are not allowed to use the remaining bandwidth for the SFR CCUs. All these factors

cause the wasting of valuable resources at the SFR CCUs, whereas the available

bandwidth for each SFR CEU is decreased due to the enlarged CEU-zone, and thereby

the degradation of the whole system performance.

In case the CCU-area is wider assumed, more UTs are defined as CCUs and the

number of CEUs is decreased. In this case, a part of the SFR CCUs on the outside

might not be served with lower transmission power since they are too distant from the

BS to keep a required CINR level (in SFR, CCUs are only allowed to transmit with

lower power level). Thus, even if spectrum resources are supplied sufficiently for the

CCU-traffic, the resource wasting at the CCUs can still not be avoided.

In contrast, using the EFFR scheme, the number of reuse-3 and reuse-1 subchannels

can be adjusted according to the needs (traffic- or user-distribution in the system). In

addition, different from the SFR, both CCUs and CEUs in the EFFR scheme may

4.1 EFFR 47

occupy resources on the reuse-3 and reuse-1 subchannels with higher and lower power

level respectively, only complying with some rules, namely:

The EFFR CEUs always have a prior right to utilize the reuse-3 subchannels in

the Primary Segment; and

Valid resource occupation on the secondary reuse-1 subchannels should be

assured based on CINR estimation.

In this way, the simultaneity of resource wasting at the CCUs and starving situation at

the CEUs can be effectually inhibited. Besides, as the SFR scheme is designed based

on inclusive FFR, whilst the EFFR scheme is designed on exclusive FFR, the mean

CINR level experienced by both, the SFR CEUs and the SFR CCUs can be expected

to be lower than by the EFFR CEUs on the reuse-3 subchannel and the EFFR CCUs

on the reuse-1 subchannel, respectively. That means the EFFR scheme potentially can

offer wider cell coverage and higher spectral efficiency than the SFR scheme.

In sum, the limitations by using the SFR scheme—the improper zone-definition

problem, the low spectrum reuse efficiency problem, the problem of more ICI even

with low traffic load in a system, as well as the cell coverage problem—can be

conquered or to some extent mitigated through the EFFR scheme. The EFFR design

contributes a novel resource allocation and reuse mechanism with the help of CQI

estimation to effectively avoid resource wasting, so that the precious radio resources

can be more efficiently exploited. In addition, the EFFR is more flexible (reflected in

that not only the ratio of higher power level to lower power level but also the ratio of

the number of reuse-3 subchannels to reuse-1 subchannels can be adjusted) and its

system performance will not be so strongly influenced by the CCU- and CEU-zone

definition as in a SFR system. Hence, through such a resource assignment mechanism,

a wider cell coverage percentage,

a higher system capacity, as well as

a better spectrum reuse efficiency

can be expected.

Nevertheless, it should be noted that the following relevant factors still play paramount

roles in the realization of the EFFR design and could influence the system

performance severely: 1) the ratio of the number of reuse-3 subchannels M to reuse-1

subchannels N in the Primary Segment; 2) the ratio of high power level to low power

level; 3) range definition for partition of CCUs and CEUs; 4) CINR threshold for the

guest-reuse of the secondary resources; etc. In what follows, the best range definitions


for dividing different user-type zones with a certain assumed power ratio for the EFFR

scheme and SFR scheme is given in Chapter 5 (analytical performance evaluation).

The effects on the EFFR system performances with varying ratios of M to N in the

Primary Segment will be discussed in both Chapter 5 and Chapter 6 (simulative

performance evaluation). And how the power ratio employed on reuse-3 subchannels

to reuse-1 subchannels as well as how the range definition for dividing CCU-zone and

CEU-zone will impact on the performances of the EFFR and the SFR schemes, will be

shown in Chapter 6, as well.

Furthermore, to the best of our knowledge, this is the first work to present simulation

results of the SFR scheme with varying range definitions for partitioning of CCUs and

CEUs.

4.2 EFFR-Advanced

In some circumstances (for example, the suburban C1 Metropol path loss Line-Of-

Sight (LOS) model from the IST-WINNER project [35]), it is unlikely to attain 100%

cell coverage even with the usage of pure (exclusive) Reuse-3. In fact, [6] has

indicated that with a size of FRF ≤ 4 no system can provide a sufficient CINR level at

the cell border with any cell radius under the WINNER LOS condition. Therefore, in

order to achieve a full cell coverage or further promote the performance of the most

distant users located near the cell border, an Enhanced Fractional Frequency Reuse –

Advanced (EFFR-A) scheme is proposed. Based on the EFFR, the EFFR-A scheme

further separates the CEUs into Cell-Middle Users (CMUs) and Cell-Remote Users

(CRUs). For the CRUs, the EFFR-A enlarges the co-channel distance and possibly

increases the transmission power on their subchannels. Figure 4.2 illustrates the EFFR-

A design in a multi-tier cellular system, in which FRF of 1 is applied on the CCUs

with lower power, whereas FRF of 3 is used on CMUs with moderate power, and FRF

of 9 on CRUs with higher power. In this way, the CRUs become more robust against

CCI, but at the expense of a decrease on available bandwidth for the CMUs.

4.2 EFFR-Advanced 49

Figure 4.3 gives an example of the available subchannels for the cell 1 in Figure 4.2.

Suppose that there are S (= 30) subchannels in total in an EFFR-A system. The

Primary Segment of each cell consists of M2 (= 1) exclusive reuse-δ2 (= 9) subchannels

for CRUs, M1 (= 3) exclusive reuse-δ1 (= 3) for CMUs and N (= 4) reuse-1

subchannels for CCUs

2 2 1

1

SM M N

(4.11)

with 1

2

2

0

0

SN

SM

(4.12)

, whereas 1( 1) N (= 8) reuse-1 subchannels constitute the Secondary Segment. As

a result, there are a total of 2 1 1M M N (= 16) subchannels (i.e., about 53% of the

whole system bandwith) available for each EFFR-A cell.

32

197

2

1817

6

1 8

9

16

11 10

34

12

14

13

5

15

21

37

33

31

27

25

22

24

28

30

34

36

20

2326

29

35

23

9

11

201

19

35

17

32

15

29

13

26

3

2

4

5

6 7

8

1012

14

16 18

21

22

2425

27

28

30

31

33 34

37

36

Frequency reuse 1 for CCUs

Frequency reuse 3 for CMUs

(cell-middle users)

Frequency reuse 9 for CRUs

(cell-remote users)

Figure 4.2: Frequency assignment pattern of the EFFR-A scheme in a cellular system with interfering cells up to the 3rd-tier, where the CCUs use reuse-1 subchannels with lower power, the CMUs use

reuse-3 subchannels with moderate power, and the CRUs use reuse-9 subchannels with higher power.


For assessing performance of an EFFR-A system, a significantly larger scenario up to

at least 3 tiers with 37 cells in total should be considered, which can be hardly carried

out through our system-level simulator OpenWNS (refer to Appendix A) within

reasonable time constraints. Hence, we have implemented the EFFR-A scheme and the

oncoming introduced EFFR-Beyond approach in Matlab to investigate their

performance numerically and present the results in Chapter 5.

4.3 EFFR-Beyond

The only difference between the Enhanced Fractional Frequency Reuse - Beyond

(EFFR-B) design and the original EFFR scheme is that the policy, which is applied to

the CRUs in the EFFR-A scheme, is executed to the CEUs in the EFFR scheme

(namely, FRF of 9 and possibly further higher power on CEUs). In fact, the EFFR-B

scheme is a variation of the EFFR scheme with further raising the FRF and probably

further increasing the transmission power for the CEUs. As shown in Figure 4.4, the

EFFR-B applies reuse-1 for CCUs with lower power, reuse-9 for the residual CEUs

with higher power.

f

P(f)

F1

F4 F5 F6

F7 F8 F9 F10

F1

F17 F18 F19 F20 F1 F27 F28 F29 F30

Cell 1

F2 F3 F1F11 F12 F13

F14F15 F16

F1F1F21 F22 F23

F24F25 F26


Frequency reuse 3 for CMUs

Frequency reuse 9 for CRUs

Not available for Cell 1

Figure 4.3: An example of the available subchannels for the cell 1 in Figure 4.2, consisting of 1

exclusive reuse-9 subchannels (M2 = 1), 3 exclusive reuse-3 subchannels (M1 = 3) and 12 reuse-1 subchannels (N = 4).

4.3 EFFR-Beyond 51

Figure 4.5 gives an example of the available subchannels for the cell 1 in Figure 4.4.

Likewise, S (= 30) subchannels in total are assumed in an EFFR-B system. The

Primary Segment of each cell consists of M2 (= 1) exclusive reuse-δ2 (= 9) subchannels

for CEUs and N (= 7) reuse-1 subchannels for CCUs

2 2

1

SM N

(4.13)

with 1

2

2

0

0

SN

SM

(4.14)

, whereas 1( 1) N (= 14) reuse-1 subchannels constitute the Secondary Segment. As

a result, there are a total of 2 1M N (= 22) subchannels (i.e., about 73% of the

whole system bandwith) available for each EFFR-B cell.

32

197

2

1817

6

1 8

9

16

11 10

34

12

14

13

5

15

21

37

33

31

27

25

22

24

28

30

34

36

20

2326

29

35

23

9

11

201

19

35

17

32

15

29

13

26

2

34

5

6 7

8

1012

14

16 18

21

22

25 24

27

28

30

31

33 34

36

37


Frequency reuse 9 for CEUs

Figure 4.4: Frequency assignment pattern of the EFFR-B scheme in a cellular system with interfering

cells up to the 3rd–tier, based on FRF of 1 for CCUs and FRF of 9 for the CEUs.


In comparison with the original EFFR scheme, communications between BSs and their

CEUs in an EFFR-B system can have higher channel quality. Nevertheless, with the

usage of the EFFR-B scheme, the total available bandwidth for each cell is reduced.

And the flexibility for adjusting available bandwidth proportion for different FRF

zones is also decreased. When comparing the EFFR-B scheme with the EFFR-A

scheme, both can gain the same cell coverage percentage. Each EFFR-B cell can even

own more available bandwidth than when using the EFFR-A scheme. However, the

possibility of allotting resources for different FRF-zones according to traffic

distributions or user distributions in the EFFR-B system is still greatly reduced.

Detailed analysis and discussion of the achievable system performance including cell

coverage, cell capacity as well as area spectral efficiency with the usage of all three

EFFR schemes under various circumstances will be presented in the succeeding

Chapter 5.

F23

F24 F25

f

P(f)

F1

F4 F5 F6 F7 F8 F9 F10

F1

F17 F18 F19 F20

Cell 1

F2 F3 F1F11 F12 F13

F14 F15 F16

F1

F27 F28 F29 F30

F1 F22

F26

F21

Frequency reuse 9 for CEUs

Not available for Cell 1


Figure 4.5: An example of the available subchannels for the cell 1 in Figure 4.4, consisting of 1

exclusive reuse-9 subchannels (M2 = 1) and 21 reuse-1 subchannels (N = 7).

CHAPTER 5

5 Performance Analysis

Performance Analysis

5.1 Cell Coverage Analysis ................................................................................ 54

5.2 Cell Capacity Analysis ................................................................................. 77

5.3 Comparison between Analytical and Simulation Results ......................... 85

5.4 Summary and Conclusion ........................................................................... 90

This chapter addresses the performance evaluation of reuse partitioning (RUP)

techniques in OFDMA-based cellular radio networks by means of mathematical

analysis. RUP techniques (including SFR, EFFR, EFFR-Advanced, and EFFR-

Beyond) and two static Reuse schemes have been implemented in Matlab to

investigate their performance numerically. Perfect channel knowledge is assumed in

all calculations.

One of the major purposes of the mathematical analysis presented in this chapter is to

give the dimension under different propagation conditions. In the first section, the

Carrier-to-Interference-plus-Noise Ratio (CINR) for all RUP schemes mentioned in

Chapters 3 and 4 is investigated. Based on the CINR calculation, the maximal cell

radius and the reasonable boundary definitions for dividing different user-type zones

for each RUP scheme are given. Furthermore, through numerical evaluations, the cell

coverage of all studied reuse techniques is estimated and compared.

The second section addresses the mean cell capacity computation for all studied

partitioning techniques, which is based on the results of the first section. To be

consistent to the simulation results which are presented later in this work, the same

PHY modes and packet sizes are used in the analysis. The mean cell capacity is

determined by the mapping of CINR levels onto usable PHY modes. In addition, the

area spectral efficiency for these approaches is also evaluated and presented at the end

of this section.

Another purpose of the mathematical analysis carried out per Matlab in this chapter is

to validate the simulation tool which has been developed within the scope of this work

(see Appendix A). Therefore, a comparison between analytical and simulation results

using a specific scenario is exhibited in the third section. However, the influence of

some factors such as interference aware reuse mechanism as well as resource element

5. Performance Analysis 54

size definition cannot be handled by the mathematical analysis, which causes some

discrepancies between analytical and simulation results. Hence, this chapter also

motivates the need for stochastic simulations for investigation of the RUP techniques

in OFDMA-based cellular radio networks.

In the remainder of this monograph, Reuse with a capital "R" is distinguished from

reuse with a lowercase "r". Reuse with a capital "R" indicates static Reuse schemes,

whereas reuse with a lowercase "r" is an adjunct, which is always used in combination

with subchannels (e.g., the reuse-1 or reuse-3 subchannels in the SFR and the EFFR

schemes). The reuse-1 and reuse-3 subchannels in the EFFR series are exclusive reuse

subchannels as in the static Reuse schemes, however, in the SFR the reuse-1 and

reuse-3 subchannels are inclusive reuse subchannels. Hence, the reuse-1 and reuse-3

subchannels in the SFR should not be confused with those in static Reuse schemes or

in the EFFR schemes.

5.1 Cell Coverage Analysis

5.1.1 Carrier to Interference Calculation

In this section, Carrier to Interference plus Noise Ratio (CINR) of a series of

frequency reuse designs is derived, which contains the Reuse-1 scheme, the Reuse-3

31

186

1

1716

5

7

8

15

10 9

33

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

D3

D2

D5 D

4

Target cell

Interfering cell

xD1

y

20

R/2

R

√3/2 R

Figure 5.1: Cellular system with interfering cells up to 3 tiers with 5 different co-channel distance: D1, D2, D3, D4 and D5.

5.1 Cell Coverage Analysis 55

scheme, the well-known SFR scheme, the proposed EFFR scheme, and two EFFR

derivatives, namely the EFFR-Advanced scheme and the EFFR-Beyond scheme.

For a theoretical analysis of a Reuse-1 cellular system, an investigation on a cell with

surrounding interfering cells up to 3 tiers is reasonable and convincing. Figure 5.1

instantiates a cellular system with 36 adjacent interfering cells. It can be seen that there

are 5 different co-channel distances Dj for 0 ≤ j ≤ 5 within 3 tiers of Reuse-1

interfering cells. Table 5.1 and Figure 5.2 further illustrate the relation between each

co-channel distance Dj and cell radius R, as well as the number of each type of Dj in

each tier.

In OFDMA-based communication networks, resource allocation is based on channel

quality of each subchannel. And with the RUP techniques, subchannel assignment

depends on the specific locations of users. Therefore, in what follows, we are

concerned about the received CINR value on each subchannel. In the considered

network, DL and UL channels are assumed to be perfectly separated either by a FDD

or by a fully synchronized TDD scheme. Thus, in DL, neighboring BSs cause

interference, while in UL, interference is generated by UTs in the neighboring cells.

In the following analysis, the CINR values for UL traffic received at BS and for DL

traffic at UTs are discussed separately. According to [8], generally the radio

propagation can be modeled as

2

0_ _

1

4

SCHRx SCH Tx SCH Tx Rx

cP P g g

f l l

(5.1)

, where PTx-SCH is the transmission power level for one subchannel and PRx-SCH the

received power on this subchannel, c0 the speed of light, l the distance between

transmitter and receiver, γ a propagation coefficient between 2 and 5, gTx and gRx the

antenna gains at the receiver and the transmitter side.

Table 5.1: Relations between Dj, R and cell type for different tiers.

Tier Dj Value Corresponding

cell type

Number of cells

for each type

1st D1 3 R Γ 6

2nd D2 3 R Δ 6

2nd D3 2 3 R Θ 6

3rd D4 21 R Λ 12

3rd D5 3 3 R Ξ 6


Assuming the BS of the target cell being situated at the grid origin in the Cartesian

coordinates as shown in Figure 5.1, the received carrier level on a subchannel CSCH at

position (x, y) is therefore:

2 2( , )

( )

SCHSCHC x y

x y

(5.2)

31

186

1

1716

5

7

8

15

10 9

23

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

xD1

y

20

31

186

1

1716

5

7

8

15

10 9

33

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

D3

D2

x

y

20

(a) (b)

31

186

1

1716

5

7

8

15

10 9

23

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

D5 D

4

x

y

20

(c)

Figure 5.2: 5 different interfering cell types distributed up to 3 tiers: (a) 6 Γ interfering cells with co-channel distance of D1 locate on the 1st-tier; (b) 6 Δ and 6 Θ interfering cells, with co-channel distance

of D2 and D3 respectively, located on the 2nd-tier; (c) 12 Λ and 6 Ξ interfering cells, with co-channel

distance of D4 and D5 respectively, locate on the 3rd-tier.


, where ζSCH varies depending on the transmission power PTx-SCH applied on a certain

subchannel. PTx-SCH has different values in accordance with the type of UTs, such as

Cell-Centre Users (CCUs), Cell-Edge Users (CEUs), Cell-Middle Users (CMUs) and

Cell-Remote Users (CRUs), corresponding to each scheme mentioned in Section 3.2.2

and Chapter 4. For all studied schemes, Eq. (5.2) for calculation of received carrier

level CSCH can be used in both UL and DL. However, the computation of the co-

channel interference (CCI) ISCH in UL is much more complicated than in DL,

especially for the RUP schemes.

5.1.1.1 Mean UL Interference Generated by Co-channel UTs in the Neighboring Cells

In UL, UTs of co-channel cells generate interference. These UTs are assumed

randomly distributed within the cell area. In order to model CCI more accurately, the

mean interference generated by a co-channel cell is calculated by assuming a planar

transmitter with the shape of a hexagonal cell. The transmit power is assumed equally

distributed all over the cell surface area [6]. With the help of the analytical methods

described in [6], the mean UL interference UL

SCHI yielded by one interfering cell can be

calculated by integrating the received power of a point (x, y) over the whole hexagonal

cell area.

2 2 21

( )UL UT

SCH SCH

cellArea

I x ycellArea

(5.3)

, where the surface area of a hexagonal cell is known as

233 .

2cellArea R (5.4)

In this way, the calculation of the mean CCI is decoupled with geographic locations of

the users.

According to each type of reuse distance Dj in 3 tiers of surrounding cells, there are 5

types of co-channel cells (type-Γ, -Δ, -Θ, -Λ and -Ξ in Table 5.1), which generate CCI

to the observed center cell. As indicated in Figure 5.1 and Figure 5.2, 5 corresponding

mean UL interference (_

UL

SCHI ,

_

UL

SCHI ,

_

UL

SCHI ,

_

UL

SCHI and

_

UL

SCHI ) of different

cell types can be attained by using Eq. (5.4). Here, we give the mean UL CCI

expression _

UL

SCHI for the interfering cell Nr. 20 of type- Λ on the 3

rd-tier as an

example, which depends on two factors ζSCH and R. To integrate over the cell area, it

is divided into three parts (see Figure 5.1), resulting in


53 5 3

2 22 2_

2 32

2 3 32 2 2

2 3

3 2 32 2 2

3 3 32

( , ) [ ( )3

32

( )

( ) ].

UTR y R

UL UT SCHSCH SCH

R y

R R

R R

R y R

Ry R

I R x y dxdy

R

x y dxdy

x y dxdy

(5.3a)

The mean UL CCI of other cell types can be calculated in a similar way.

In the following, we present the UL CCI equations for all six schemes respectively,

namely: the Reuse-1 scheme, the Reuse-3 scheme, the SFR scheme, the EFFR scheme,

EFFR-A and EFFR-B schemes.

A. Mean UL interference using Reuse-1

Using the Reuse-1 scheme, the UL traffic received by the BS in the target cell is

interfered by the simultaneous UL transmissions occurring in all 3 tiers of surrounding

cells.

_ 1 1 2 3

_ _ _

1 2

_ _

3

6 ( , ) 6 ( , ) 6 ( , )

12 ( , ) 6 ( , )

UL UL UL UL

SCH Reuse SCH stTier SCH ndTier SCH rdTier

UL UL UL

SCH SCH SCH

st tier nd tier

UL UL

SCH SCH

rd tier

I I I I

I R I R I R

I R I R

(5.5)

B. Mean UL interference using Reuse-3

In comparison with the Reuse-1, using a cluster-order of 3, the BS in the target cell is

just interfered by the UL transmissions occurring in the 6 type-Δ cells on the 2nd

-tier

and 6 type-Ξ cells on the 3rd

-tier, see Figure 5.2b and Figure 5.2c or Figure 5.3.

Thereby, the mean CCI can be calculated as

_ 3 _ _

2 3

6 ( , ) 6 ( , ).UL UL UL

SCH Reuse SCH SCH

on the nd tier on the rd tier

I I R I R

(5.6)


C. Mean UL interference using SFR

As explained in Section 3.2.2, the SFR applies an inclusive RUP, which means the

Major Segment (FRF of 3, see Figure 3.3) for the CEUs in a cell is simultaneously

reuse-1 subchannels used by CCUs in the neighboring cells. Thus, the UL traffic from

a CEU in the target cell is interfered by the CEUs transmissions in the 6 type-Δ cells

on the 2nd

-tier and 6 type-Ξ cells on the 3rd

-tier with higher transmission power, which

is like the Reuse-3 situation illustrated in Figure 5.3. Besides, it is also interfered by

the CCUs UL transmission in all the other cells with lower transmission power.

Accordingly, the mean CCI received by the BS together with the UL traffic from a

CEU can be calculated as

( )

_ _ _ _

,

6 ( , ) 6 ( , ) 12 ( , )

6

UL CEU BS UL CCU CCU UL CCU CCU UL CCU CCU

SCH SFR SCH SCH SCH

CCUs in type and cells use this subchannel to transmit with lower transmission power

SC

I I R I R I R

I

_ _ _ _( , ) ( , ) 6 ( , ) ( , ) .UL CEU UL CEU CCU UL CEU UL CEU CCU

H SCH SCH SCH

CEUs in type and cells use this subchannel to transmit with higher transmission power

R I R I R I R

(5.7)

, where RCCU

defines the range for the CCUs.

Likewise, the UL traffic from a CCU in the target cell (assuming the type-A cell in

Figure 3.3) is interfered by the CEUs transmissions in the 12 neighboring cells (e.g.,

type-B cells in Figure 3.3), which is distributed in half of the type-Γ cells on the 1st-tier

(see Figure 5.4a), half of the type-Θ cells on the 2nd

-tier (see Figure 5.4b) and half of

31

186

1

1716

5

7

8

15

10 9

23

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

x

y

20

D2

D5

Figure 5.3: Interfering cells up to the 3rd-tier for a Reuse-3 cellular system: the transmission in the

target cell is interfered by concurrent traffic in the 6 type-Δ cells on the 2nd-tier with co-channel

distance of D2 and 6 type-Ξ cells on the 3rd-tier co-channel distance of D5, respectively.


the type-Λ cells on the 3rd

-tier (see Figure 5.4c). In addition, it is also interfered by the

CCU UL transmissions of the remaining 24 cells (consisting of type-A and -C in

Figure 5.4). So, the mean CCI for UL traffic from a CCU can be calculated as

x

y

B

B

B

C

C

C

A D1

3

D3

D2

x

y

A

A

A

A

A

A

C

B

B

B A

C

C

(a) (b)

D5 D

4

x

y

A

AA

A

AA

A

B

B

C

C

C

C

C

C

B

B

B

B

(c)

Figure 5.4: UL traffic from a CCU in the target cell (cell-type A) is interfered by the CEUs transmissions in the 12 neighboring type-B cells, which consist of half number of the (a) 6 Γ

interfering cells with co-channel distance of D1 locate on the 1st-tier; half number of the (b) 6 Θ

interfering cells with co-channel distance of D3 located on the 2nd-tier; as well as half number of the (c) 12 Λ interfering cells with co-channel distance of D4 locate on the 3rd-tier. Besides, it is also

interfered by the CCU UL transmissions of the remaining 24 type-A and C cells.


( )

_ _ _

_ _

_ _

_ _

_

3 ( , ) ( , )

3 ( , ) ( , )

6 ( , ) ( , )

3 ( , ) 3 ( , )

6

UL CCU BS UL CEU UL CEU CCU

SCH SFR SCH SCH

UL CEU UL CEU CCU

SCH SCH

UL CEU UL CEU CCU

SCH SCH

UL CCU CCU UL CCU CCU

SCH SCH

SCH

I I R I R

I R I R

I R I R

I R I R

I

_

_

( , ) 6 ( , )

6 ( , ).


SCH

UL CCU CCU

SCH

R I R

I R

(5.8)

It can be seen that in SFR whether for CEU-transmission or CCU-transmission, they

are all interfered by the whole 3 tiers of surrounding cells, but with different power

allocation.

D. Mean UL interference using EFFR

As described in Chapter 4, the EFFR scheme is an exclusive RUP technique. In the

EFFR scheme, the CEUs exclusively use the reuse-3 subchannels with higher

transmission power, whereas the CCUs use the reuse-1 subchannels with lower

transmission power (see Figure 5.5). Therefore, for the EFFR scheme, the mean UL

interference generated by the CEUs in the neighboring cells can be calculated as

31

186

1

1716

5

7

8

15

10 9

33

11

13

12

4

14 36

32

30

26

24

21

23

27

29

33

35

19

2225

28

34

x

y

20

19

20

21

36

35

18

7

8

1

17

6

2

343332

1615

31

30

29

28

14 5

413

1227

26

25 24

3

10

23

11 9

22


Exclusive frequency reuse 3

for CEUs

Figure 5.5: Using the EFFR scheme, the CEUs exclusively use the reuse-3 subchannels with higher

transmission power, whereas the CCUs use the reuse-1 subchannels with lower transmission power.


( )

_ _ _

_ _

6 ( , ) ( , )

6 ( , ) ( , )

UL CEU BS UL CEU UL CEU CCU

SCH EFFR SCH SCH

UL CEU UL CEU CCU

SCH SCH

I I R I R

I R I R

(5.9)

, which is similar to the Eq. (5.6) for the Reuse-3 scheme, but only for CEUs. And the

mean UL interference caused by the CCUs in the 3 tiers of surrounding cells can be

summed up as

( )

_ _

1

_ _

2

_ _

3

6 ( , )

6 ( , ) 6 ( , )

12 ( , ) 6 ( , )

UL CCU BS UL CCU CCU

SCH EFFR SCH

st tier


SCH SCH

nd tier


SCH SCH

rd tier

I I R

I R I R

I R I R

(5.10)

, which is similar to the Eq. (5.5) for the Reuse-1 scheme, yet it is only suited for the

CCUs.

For the EFFR-A and EFFR-B schemes, the mean UL interference UL CCU

SCHI is the same

as using the EFFR scheme. So, in what follow, we complete the UL CCI formulas at

the BS for its receiving packets from other more distant users.

E. Mean UL interference using EFFR-A

In order to further ameliorate the performance of the most distant users in a cell, which

are located near the cell border, the EFFR-A scheme based on the EFFR design further

separates the CEUs into CMUs and CRUs. For the CRUs, the EFFR-A enlarges the

co-channel distance and possibly increases the transmission power on their

subchannels. As illustrated in Figure 4.2 in Section 4.2, in an EFFR-A cellular system,

FRF of 1 is applied on the CCUs with lower power, whereas FRF of 3 is used on

CMUs with moderate power, and FRF of 9 on CRUs with higher power. So, the mean

UL interference together with the UL traffic from a CMU can be calculated similarly

as that for the CEUs in the EFFR scheme, see Eq. (5.9)

( )

_ _ _

_ _

6 ( , ) ( , )

6 ( , ) ( , )

UL CMU BS UL CMU CMU UL CMU CCU

SCH EFFR A SCH SCH

UL CMU CMU UL CMU CCU

SCH SCH

I I R I R

I R I R

(5.11)


, where RCMU

gives the maximum range for the CMUs. And the mean UL interference

yielded by the CRUs in the 6 interfering Ξ cells (see Figure 4.2 and Figure 5.2c) can

be computed as

( )

_ _ _6 ( , ) ( , ) .UL CRU BS UL CRU UL CRU CMU

SCH EFFR A SCH SCHI I R I R

(5.12)

F. Mean UL interference using EFFR-B

The only difference between the EFFR-B design and the EFFR scheme is that the

policy, which is applied on the CRUs in the EFFR-A scheme, is executed to the CEUs

in the EFFR scheme. That means the EFFR-B applies reuse-1 for CCUs with lower

power, reuse-9 for the CEUs with higher power. Thereby, similar calculation can be

taken for the EFFR-B scheme for the CCI resulted by the CEUs in the certain reuse-9

co-channel cells as shown in Figure 4.4, which are actually the 6 type-Ξ cells (see

Figure 5.2c)

( )

_ _ _6 ( , ) ( , ) .UL CEU BS UL CEU UL CEU CCU

SCH EFFR B SCH SCHI I R I R

(5.13)

5.1.1.2 DL Interference Generated by BSs in the Neighboring Cells

In DL, the CCI is caused by the concurrent DL traffic from the BSs in the neighboring

cells. Since the locations (x0i, y0i) of all BSs in the surrounding interfering cells are

fixed, the CCI ( , )DL

SCHI x y received by a user located at (x, y) for DL traffic can be

derived by:

( , ) ,[ ( , )]

iBSDL SCHSCH

i i

I x yd x y

(5.14)

with 2 2

0 0( , ) ,i i id x y x x y y for 1, 2, ... 36 .

According to different schemes, each interfering BS arranges its transmission power

on each available subchannel depending on the type of its user served on this

subchannel. Eq. (5.14) is applicable for all schemes except for the subset definition ,

which clarifies the set of the interfering BSs in the neighborhood for each scheme. In

the following, we give the specific definition for each scheme respectively.

A. Reuse-1


With the Reuse-1, the DL CCI is generated by all 36 surrounding cells with equal

transmission power. Thereby,

1 1, 2, ... 36Reuse (5.15)

B. Reuse-3

As shown in Figure 5.3, using the Reuse-3 in a cellular system up to 3 tiers, the

transmission in the target cell is interfered by concurrent traffic in the 6 type-Δ cells on

the 2nd

-tier with co-channel distance of D2 and 6 type-Ξ cells on the 3rd

-tier with co-

channel distance of D5, respectively. So,

3 8,10, ...18 19, 22, ... 34 .Reuse Cell Cell (5.16)

C. SFR

Since the SFR applies an inclusive RUP, whether for a CEU or for a CCU, a DL

transmission is always interfered by the whole 3 tiers of surrounding cells, but with

different power allocation. Accordingly,

1, 2, ... 36SFR . (5.17)

D. EFFR

By applying an exclusive RUP, the DL traffic for the CEUs using EFFR receives CCI

as using the Reuse-3, whereas for the CCUs the situation is as using the Reuse-1

scheme, see Figure 5.5. Herewith,

8,10, ...18 19, 22, ... 34CEU

EFFR Cell Cell (5.18)

and 1, 2, ... 36CCU

EFFR (5.19)

E. EFFR-A & EFFR-B

Likewise, for EFFR-A scheme (see Figure 4.2 and Figure 5.2c)

19, 22, ... 34 ,CRU

EFFR A Cell (5.20)

8,10, ...18 19, 22, ... 34 ,CMU

EFFR A Cell Cell (5.21)

and 1, 2, ... 36 ;CCU

EFFR A (5.22)


for EFFR-B scheme (see Figure 4.4 and Figure 5.2c)

19, 22, ... 34 ,CEU

EFFR B Cell (5.23)

and 1, 2, ... 36CCU

EFFR B . (5.24)

5.1.1.3 CINR Derivation

Now, we can accomplish the CINR expression at position (x, y) for the Reuse-1 and

Reuse-3 schemes as follows:

2 2( , )( , ) , (0, ]

( , )

SCHSCH

SCH

C x yCINR x y for x y R

I x y N

(5.25)

, and for the SFR, EFFR, EFFR-B schemes:

2 2

2 2

( , ), (0, )

( , )( , )

( , ), [ , ]

( , )

CCUCCUSCH

CCU

SCH

SCH CEUCCUSCH

CEU

SCH

C x yfor x y R

I x y NCINR x y

C x yfor x y R R

I x y N

(5.26)

, as well as for the EFFR-A scheme:

2 2

2 2

2 2

( , ), (0, )

( , )

( , )( , ) , [ , ).

( , )

( , ), [ , ]

( , )

CCUCCUSCH

CCU

SCH

CMUCCU CMUSCH

SCH CMU

SCH

CRUCMUSCH

CRU

SCH

C x yfor x y R

I x y N

C x yCINR x y for x y R R

I x y N

C x yfor x y R R

I x y N

(5.27)

In can be noticed that the CINR Eqs. (5.25), (5.26) and (5.27) are not distinguished for

UL and DL traffic, but when substituting the CCI for each type of users in each

scheme, the corresponding UL and DL expressions described in Subsections 5.1.1.1

and 5.1.1.2 should be used.

5.1.2 Cell Coverage Comparison

By using the CINR calculation introduced in Section 5.1.1, the maximal cell radius for

each RUP scheme and the range definition for dividing different user-type zones of


each scheme can be determined. And accordingly, their respective cell coverage can

be estimated and compared.

Figure 5.1 gives the considered cellular scenario which consists of 37 hexagonal cells

with central BSs. For the evaluation, antenna gain is neglected at the receiver as well

as at the transmitter. 30 non-overlapping frequency subchannels with a bandwidth of

20 MHz are located at 5.47GHz. In order to determine the user-type zones for each

scheme, the CINR at the cell border and at the border of each zone is of interest, where

the most robust Modulation and Coding Scheme (MCS), i.e. BPSK ½ has to be used.

The minimum receiver requirement for BPSK½ is 6.4 dB which is taken from the

802.16 standard [42]. The maximum transmission power of BSs is restricted to 2000

mW or 33 dBm, and for UTs is 200mW or 23 dBm. Table 5.2 details the transmission

power on a subchannel used by each type of users for all studied schemes. Thermal

noise of -174 dBm/Hz and a noise figure of 5 dB for BS as well as 7 dB for UT are

assumed.

For the path loss, the suburban C1 Metropol path loss model from the IST –WINNER

project [35] is chosen, which has also been implemented in the simulation

environment OpenWNS described in Appendix A. The C1 Metropol is a composition

of two models, a LOS and a NLOS model. Eqs. (5.28) and (5.29) list their parameters,

respectively.

LOS: 41.9

51010 6.457 10

23.82.38

10 (5.28)

Table 5.2: Transmission power applied in studied schemes.

Scheme

PTx_SCH

in UL

[mW]

∑PTx

in UL

[mW]

PTx_SCH

in DL

[mW]

∑PTx

in DL

[mW]

Reuse 1 66.67 2000 66.67 2000

Reuse 3 200 2000 200 2000

SFR CCU CEU

3333 CCU CEU

2000 66.67 200 40 120

EFFR CCU CEU

2000 CCU CEU

2000 66.67 200 66.67 200

EFFR-A CCU CMU CRU

2000 CCU CMU CRU

2000 66.67 200 200 66.67 200 600

EFFR-B CCU CEU

2000 CCU CEU

2000 66.67 200 66.67 600


NLOS: 27.7

31010 1.698 10

40.24.02

10 (5.29)

The other main relevant parameters used for evaluation are shown in Table 5.3.

In the following, we will compare the coverage of all studied RUP schemes under

LOS, NLOS as well as combined LOS-NLOS condition, separately. First, the CINR

level at the cell border is evaluated with varying cell radius R, so that the maximum

cell radius Rmax for each scheme can be found. Then, with the determined Rmax, the

CINR for a user traversing the cell across the x-axis is given, with which the ranges for

partitioning CCUs and CEUs (or CMUs and CRUs) for all in Chapters 3 and 4

mentioned schemes as well as the coverage using each scheme will come out.

5.1.2.1 LOS Condition

Figure 5.6a plots the UL CINR perceived at the BS versus the cell radius R, while a

UT as a transmitter is located at the cell border. And Figure 5.6b represents the DL

CINR received at the cell border with varying cell radius. Both scenarios are under

LOS propagation. The maximum cell radius (CINR of 6.4 dB) is highlighted by stems.

In general, the CINR decreases with an increasing cell radius for both UL and DL

situations. The both UL and DL CINR using the EFFR-A or -B scheme are better than

using the other schemes with any cell radius. This is mainly due to the fact that the

UTs at the cell border using the EFFR-A or -B work with a large co-channel distance

(D5 = 3 3 R ). Comparing UL and DL, applying the EFFR-A or -B scheme the DL

CINR (see Figure 5.6b) is higher than their UL CINR (see Figure 5.6a). This is

because in DL the BS may use three times stronger transmission power (600 mW)

Table 5.3: Assumptions for evaluation

Parameter Value

System bandwidth 20 MHz

Center frequency 5470 MHz

Subcarriers (FFT size) 2048

OFDMA symbol duration 102.858 μs

Number of data subcarriers 1440

Number of subchannels 30

Number of interfering cells 36 (up to 3 tiers)

UT thermal noise density -174 dBm/Hz

Noise figure at [BS, UT] [5, 7] dB

Minimum CINR 6.4 dB


than the maximum transmission power of a UT (200 mW) on one subchannel, see

Table 5.2. However, with the other schemes, the UL CINR seems slightly better than

the DL CINR. Another phenomenon exposed in both Figure 5.6a und Figure 5.6b is

that except EFFR-A and EFFR-B schemes, the other schemes do not provide a

sufficient CINR level at the cell border for DL or over cell radius for UL under LOS

propagation. Using the EFFR-A or -B scheme, the maximal cell radius reaches 2800 m

for UL and 4424 m for DL, respectively.

As cell radius for UL and DL should be identical, we choose the minor maximal cell

radius 2800 m as cell radius R to evaluate the CINR distribution along with varying

distance between a UT and the BS. Figure 5.7 displays the CINR for a UT traversing

the cell across the x-axis for both UL (Figure 5.7a) and DL (Figure 5.7b) under LOS

propagation. One can see the BS position and the cell border. The range of coverage of

a scheme is marked by two stems, whose height indicates the minimum receiver

requirement (6.4 dB) for the PHY mode BPSK½. The CINR by using the EFFR series

is better than using the SFR scheme at any position for both UL and DL. In the figures

we can also find the positions of the ranges for partition CCUs and CEUs (or CMUs

and CRUs) for all mentioned RUP schemes. The CINR of CCUs by using SFR is

worse than with the Reuse-1 scheme, since the CEUs of some neighboring cells reuse

the same resources at the same time with higher transmission power. Though the

CINR of the CEUs using the SFR is better than using the Reuse-1, it is still worse than

using the Reuse-3 scheme.

(a) (b)

Figure 5.6: CINR versus the cell radius R using C1 LOS path loss model: (a) UL CINR perceived at

the central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell border.


For the Reuse-1, Reuse-3 and EFFR, their coverage for UL is quite similar to the DL

coverage (see Figure 5.7 and Table 5.6). On the contrary, the SFR can reach maximum

1232 m from the BS for DL (see Figure 5.7b), but 1372 m for UL (see Figure 5.7a).

This is because the transmission power on each subchannel for DL is smaller than for

UL (see Table 5.2). As a result for the example power allocation chosen for CCUs and

CEUs, respectively, SFR can cover 29% of the cell for UL, but 23.4% of the cell for

DL, see Table 5.6. Among all schemes, only the EFFR-A or -B scheme has the

capability of serving the whole cell under LOS propagation. For DL, they even can

provide a CINR higher than 6.4 dB at the cell border (see Figure 5.7b).

5.1.2.2 NLOS Condition

Figure 5.8 and Figure 5.9 show the comparable results under NLOS propagation,

where the path loss coefficient γ is nearly two times higher than under LOS

propagation, see Eq. (5.29).

Comparing Figure 5.8 and Figure 5.6, the CINR under NLOS propagation at small

radii is higher than with LOS, as the CCI is substantially reduced caused by the bigger

path loss coefficient γ. Besides the EFFR-A and -B scheme, the Reuse-3 and the EFFR

scheme also allow for cell radii of 270 m for UL (see Figure 5.8a) and 248 m for DL

(see Figure 5.8b) in a NLOS scenario. Nevertheless, applying EFFR-A or –B scheme

the larger maximal cell radii can be attained, namely, 298 m for UL and 391m for DL.

And the SFR scheme and the Reuse-1 scheme can still not provide a sufficient CINR

(a) (b)

Figure 5.7: CINR distribution, when a UT traverses the cell with a radius of 2800m under the C1 LOS propagation: (a) UL CINR received by the BS; (b) DL CINR received by the UT.


for both UL and DL. Similar to LOS results as shown in Figure 5.6, other than the

EFFR-A and –B scheme, the UL CINR over the cell radius R is higher than the DL

CINR with the other schemes. This is because in UL the receiver is located at the

center of the cell and not at the border, which reduces the received interference.

(a) (b)

Figure 5.8: CINR versus the cell radius R using C1 NLOS path loss model: (a) UL CINR perceived at the central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell

border.

(a) (b)

Figure 5.9: CINR distribution when a UT traverses the cell with a radius of 298m under the C1 NLOS

propagation: (a) UL CINR received by the BS; (b)DL CINR received by the UT.


Like Figure 5.7, Figure 5.9 shows CINR distribution for both UL and DL under NLOS

propagation for a UT traversing across the cell, where the smaller UL maximal cell

radius of the EFFR-A scheme (298 m) is chosen as a cell radius for evaluation.

Figure 5.9 exhibits similar features as which in Figure 5.7. Moreover, either LOS or

NLOS, the CINR using the static Reuse schemes and the SFR for DL decays always

more rapidly than for UL, which means the the DL CCI when using those schemes is

more severe than the UL CCI. EFFR covers the cell up to 270 m.

5.1.2.3 Combined LOS-NLOS Condition

In urban Manhattan-like scenarios, the source and the destination have direct LOS

connection along the streets. In contrast, the interferers are shadowed behind buildings,

thus a NLOS path results. The same effect occurs in wide-area scenarios when the BSs

are deployed with an antenna tilt. Then, with high probability, the UTs of a cell have a

LOS connection to the BS while the UTs of the surrounding interfering cells perceive

NLOS attenuation. In both deployments, the carrier signal is attenuated by LOS

propagation, whereas interfering signals are attenuated with NLOS path loss.

Figure 5.10 plots the UL and DL CINR versus the cell radius of the LOS-NLOS

scenario. The curves differ from that shown in Figure 5.6 and Figure 5.8. In

comparison with pure LOS and pure NLOS condition, even the Reuse-1 and the SFR

now provide sufficient CINR for both UL and DL. And all schemes reach their

(a) (b)

Figure 5.10: CINR versus the cell radius R using LOS-NLOS path loss model: (a) UL CINR perceived at the central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell

border.


maximum radii far beyond that reached in LOS and NLOS scenarios (see Table 5.4).

Besides, the CINR under combined LOS-NLOS propagation is much higher than with

the other two conditions. This is because the interference attenuates faster than the

carrier signal due to its higher path loss coefficient γ = 4.02. Due to the relative

increase of the carrier signal compared to the interference, the CINR increases which

results in a maximum of the CINR curve among all conditions.

In UL, except the Reuse-1, for all the other schemes the maximal radius up to 3920 m

can be reached (see Figure 5.10a). In DL, however, same as LOS and NLOS scenarios,

the EFFR-A and –B schemes gain the best CINR, which results in the maximal cell

Table 5.4: Maximum cell radius under LOS, NLOS and combined LOS-NLOS propagations

(PBS_max = 2 W, PUT_max = 200 mW)

path loss Reuse-1 SFR Reuse-3 /

EFFR

EFFR-A /

EFFR-B

LOS UL - - - 2800 m

DL - - - 4424 m

NLOS UL - - 262 m 298 m

DL - - 248 m 390 m

LOS-

NLOS

UL 2436 m 3920 m 3920 m 3920 m

DL 2436 m 3920 m 3920 m 6160 m

- no valid maximal cell radius

Table 5.5: Optimal range definition for different type of users for RUP schemes under LOS, NLOS and

combined LOS-NLOS propagations (PBS_max = 2 W, PUT_max = 200 mW, RLOS = 2800 m, RNLOS = 298 m,

RLOS-NLOS = 3920 m)

path loss SFR EFFR series EFFR-A

Range

(CCUs)

Range

Ratio

Range

(CCUs)

Range

Ratio

Range

(CMUs)

Range

Ratio

LOS UL 728 m 0.26 896 m 0.32 1764 m 0.63

DL 728 m 0.26 924 m 0.33 1764 m 0.63

NLOS UL 150 m 0.50 174 m 0.58 276 m 0.93

DL 154 m 0.52 176 m 0.59 272 m 0.91

LOS-

NLOS

UL 2464 m 0.63 2464 m 0.63 - -

DL 1988 m 0.51 2464 m 0.63 - -

- no need of range definition


radius of 6160 m (see Figure 5.10b).

For Reuse-1, Reuse-3 and EFFR schemes, the UL CINR over the cell Radius R is quite

similar to that of the DL CINR, whereas SFR offers a better UL CINR than the DL

CINR, and with EFFR-A and –B schemes their UL CINR is worse than the DL CINR.

This is because using the SFR the DL transmission power for a CEU on a subchannel

is minor than the UL transmission power, see Table 5.2. On the contrary, in the EFFR-

A or –B scheme, CEUs-traffic is served with higher transmission power in DL than in

UL. Therefore, the UL CINR of the EFFR-A and –B attenuates faster than their DL

CINR, which causes a minor maximum cell radius of 3920 m.

Again, we utilize the minor valid maximal cell radius 3920 m of all studied RUP

techniques for UL and DL (see Table 5.5), as cell radius R to evaluate the CINR

distribution along with varying distance between a UT and the BS. Like Figure 5.7

under LOS condition and Figure 5.9 under NLOS condition, Figure 5.11 presents the

CINR for a UT traversing the cell across the x-axis for both UL (Figure 5.11a) and DL

(Figure 5.11b) under combined LOS-NLOS propagation. It should be noted that in UL,

unlike the pure LOS and NLOS cases, all RUP techniques own an identical CINR

distribution. And the ranges for partition CCUs from CEUs for SFR, EFFR and EFFR-

B schemes (or from CMUs and CRUs for EFFR-A scheme) occur at the same place,

see Table 5.5. Moreover, in UL, not only the CINR of CCUs using SFR is not worse

than using the Reuse-1 anymore, but also its CINR of CEUs is not inferior to the

Reuse-3 scheme. All these phenomena expose that the level of CCI under combined

(a) (b)

Figure 5.11: CINR distribution when a UT traverses the cell with a radius of 3920m under the LOS- NLOS propagation: (a) UL CINR received by the BS; (b) DL CINR received by the UT.


LOS-NLOS condition is rather low and does not affect the CINR much. This also

causes that in a aystem with limited interference, there is no difference among the

EFFR series any more in UL, and in DL the EFFR-A scheme is useless within cells,

since the optimal range for dividing CMUs and CRUs overlaps with the cell edge at

3920 m (see Figure 5.11b), which means no need to partition CRUs from CMUs to

attain 100% cell coverage.

5.1.2.4 Conclusion

Table 5.6 summarizes the corresponding maximal reaches of all studied schemes and

their cell coverage percentage for both UL and DL. The table contains results for

different propagation scenarios using different cell radii taken from the preceding

sections.

It can be seen, that in a LOS scenario, by application of the EFFR scheme (or the

Reuse-3 scheme ) and the EFFR-A or -B scheme (with FRF of 9 for remote users) in

UL, the maximal reaches are extended from 1736 m to 2800 m, which means the

system is interference-limited. In an interference-limited system, the CINR value is

more influenced by the interference level than by the fixed noise level. Therefore,

increasing FRF can benefit the CINR at the cell border and extend the coverage range.

On the contrary, under the NLOS and LOS-NLOS propagations the system is noise-

limited, since the maximal reaches with different FRFs (FRF of 3 for EFFR scheme

Table 5.6: Maximal reach and coverage percentage of each studied schemes under LOS, NLOS and

combined LOS-NLOS propagations (PBS_max = 2 W, PUT_max = 200 mW, RLOS = 2800 m, RNLOS = 298 m,

RLOS-NLOS = 3920 m)

schemes

path loss

Reuse-1 SFR Reuse-3 / EFFR EFFR-A / EFFR-B

maximum

reach

coverage

percentage

maximum

reach

coverage

percentage

maximum

reach

coverage

percentage

maximum

reach

coverage

percentage

LOS UL 868 m 11.6% 1372 m 29% 1736 m 46.5% 2800 m 100%

DL 896 m 12.4% 1232 m 23.4% 1736 m 46.5% 3220 m 100%

NLOS UL 172 m 40.3% 238 m 77.1% 274 m 97.1% 298 m 100%

DL 174 m 41.2% 208 m 58.9% 270 m 96% 390 m 100%

LOS-

NLOS

UL 2464 m 47.8% 3920 m 100% 3920 m 100% 3920 m 100%

DL 2436 m 46.7% 3136 m 77.4% 3920 m 100% 6160 m 100%


and FRF of 9 for EFFR-A and –B schemes) are close: between 274 m and 298 m in

NLOS UL, and even no change in LOS-NLOS UL case (both are 3920 m).

In DL, however, besides an increased value of FRF, the power for the cell remote

users (CRUs) in the EFFR-A and –B schemes can be set triple higher than for the cell

edge users (CEUs) in the EFFR scheme, see Table 5.2. Thus, increasing the carrier

signal power and reducing interference benefits both, interference- and noise-limited

systems.

For calculation of the cell coverage percentage, Eq. (5.30)

2*

2

2*

2

', ' (0, ]

33

2

' 6 ( ' ), ' ( , ]

33

2

Rfor R R

R

Cell Coverage PercentageR S S

for R R R

R

(5.30)

is used with

*

2' arccos( ) ''

RS R

R ; * 2 *2'S R R R and * 3

2R R (5.31)

, where 'R is the maximum reach by using a certain scheme in a hexagonal cellular

system with a cell radius of R, *R is the internally tangent circle radius of the

hexagonal cell, and 'S S denotes the shadowed area illustrated for example in

Figure 5.12.

R’

R

R* = √3/2 R

R*

Figure 5.12: Cell coverage percentage computation when the maximum reach R’ by using a certain

scheme is larger than the internally tangent circle radius R* of the hexagonal cell.


According to the results presented in Figure 5.6, Figure 5.8, and Figure 5.10, the cell

radii for CINR distribution analysis under different propagation condition (refer to

Figure 5.7, Figure 5.9, and Figure 5.11) are determined, namely, RLOS = 2800 m, RNLOS

= 298 m, and RLOS-NLOS = 3920 m, respectively. The curves in Figure 5.6 indicate that

except for the EFFR-A and EFFR-B schemes, no cell radius is found for the other

schemes (including the EFFR) under LOS propagation that a sufficient CINR level can

be provided at the cell border. And the results in Figure 5.7 and in Table 5.6 show that

the maximal reach by using the EFFR under LOS with a cell radius of 2800 m is 1736

m. Hence, the cell coverage percentage by using the EFFR scheme for example can be

calculated as follows:

2 2

2 2

' 173646,5%.

3 33 3 2800

2 2

EFFR LOS

RCell Coverage Percentage

R

Figure 5.13 visualizes the results of the cell coverage percentage given in Table 5.6 in

DL. It can be seen that the EFFR series always provide better cell coverage than the

SFR and the Reuse-1 schemes. And with the EFFR scheme, high cell coverage of

more than 96% can also be offered under NLOS condition though with a smaller

radius, see Figure 5.8. But only the EFFR-A and EFFR-B schemes can achieve 100%

cell coverage in every situation. In addition, under LOS condition (or in an

interference-limited system), the enhancement on the coverage by using the RUP

Figure 5.13: DL cell coverage percentage of each all studied schemes under LOS, NLOS and combined LOS-NLOS propagations as given in Table 5.6.

5.2 Cell Capacity Analysis 77

technique is more remarkable than under NLOS and LOS-NLOS propagations (or in

noise-limited systems).

5.2 Cell Capacity Analysis

Based on the CINR analysis in the previous section, this section provides an analysis

of the mean cell capacity for the two static Reuse schemes and all aforementioned

RUP techniques.

5.2.1 Mean Cell Capacity Computation

If perfect link adaptation is assumed, the subchannel data throughput (see Table 5.7) at

a certain position ThrSCH(x, y) can be derived by the perceived CINR for each

Modulation and Coding Scheme (MCS) in the scenario as described in Table 5.3. We

use the seven different PHY modes and their corresponding CINR measures from the

air interfaces of standard IEEE 802.16e-2004 [42].

In the next step, the average subchannel throughput ThrSCH-avg can be obtained by

integrating the ThrSCH(x, y) over the entire cell and dividing it by the cell area [43]. In

the end, we calculate the mean cell capacity CAP by multiplying the ThrSCH-avg by the

number of available subchannels. As the number of available subchannels is different

for each scheme, and the EFFR series calculates the ThrSCH-avg differently for different

zone-type of users, in what follows we give the calculations for all studied schemes

separately.

A. SFR Scheme and Reuse Schemes

Table 5.7: PHY modes and corresponding subchannel throughput

Modulation Coding rate Min. receiver CINR

[dB]

PHY Throughput per

subchannel [Mb/s]

BPSK 1/2 6.4 0.233

QPSK 1/2 9.4 0.467

QPSK 3/4 11.2 0.7

16QAM 1/2 16.4 0.933

16QAM 3/4 18.2 1.4

64QAM 2/3 22.7 1.867

64QAM 3/4 24.4 2.1


The average subchannel throughput for static Reuse schemes and the SFR scheme is

represented as

_

( , )

.

SCH

cellAreaSCH avg

Thr x y dxdy

ThrcellArea

(5.32)

The whole bandwidth, which means all 30 subchannels (see Table 5.3) in the system,

are available for the SFR and the Reuse-1 scheme, whereas just one third of the

bandwidth 10 subchannels can be used by the Reuse-3 scheme.

_

1 1

_

3 3

_

30

30

10

SFR SFR

SCH avg

reuse reuse

SCH avg

reuse reuse

SCH avg

CAP Thr

CAP Thr

CAP Thr

(5.33)

B. EFFR Scheme

Using the EFFR series, the subchannel allocation depends on the certain zone of a user.

So, according to different zone-types of the users we give their average subchannel

throughput separately. For the CEUs

_ _

_ _

1

( , ) ( , ) ,

EFFR

SCH avg CEU

EFFR EFFR

SCH CEU SCH CEU

cellArea CCUArea

ThrcellArea CCUArea

Thr x y dxdy Thr x y dxdy

(5.34)

and for the CCUs

_

_ _

( , )

.

EFFR

SCH CCU

EFFR CCUAreaSCH avg CCU

Thr x y dxdy

ThrCCUArea

(5.35)

Finally, the mean cell capacity can be calculated as

_ _ _ _3EFFR EFFR EFFR

SCH avg CEU SCH avg CCUCAP M Thr N Thr (5.36)

, where M denotes the number of available subchannels for the CEUs and 3N for the

CCUs. In addition, they are subjected to the constraints that 0 < N < 10 and

10M N which is the number of subchannels for the Primary Segment, see Eq.

(4.2) for S = 30 as assumed in Table 5.3. N should not be zero, since in this case there

are only reuse-3 subchannels available in an EFFR system, which always favor the

CEU-connections. This leads to a high probability of starvation of CCU-connections,

when not all CEU-connections can be satisfied in full-load or overload situations.


0N means M = 0 which is also unsuitable set, as in this case the full set of

subchannels S = 30 is available for each cell, and actually the EFFR system is now a

Reuse-1 system.

C. EFFR-A Scheme

Calculation of _ _

EFFR A

SCH avg CCUThr should be as same as that in EFFR scheme. The average

subchannel throughput for the CMUs and CRUs are similar to the _ _

EFFR

SCH avg CEUThr ,

however, not same

_ _

_ _

1

( , ) ( , ) ,

EFFR A

SCH avg CMU

EFFR A EFFR A

SCH CMU SCH CMU

CMUArea CCUArea

ThrCMUArea CCUArea


(5.37)

and

_ _

_ _

1

( , ) ( , ) .

EFFR A

SCH avg CRU

EFFR A EFFR A

SCH CRU SCH CRU

cellArea CMUArea

ThrcellArea CMUArea


(5.38)

The mean cell capacity for the EFFR-A scheme is

2 _ _

1 _ _ _ _3

EFFR A EFFR A

SCH avg CRU

EFFR A EFFR A

SCH avg CMU SCH avg CCU

CAP M Thr

M Thr N Thr

(5.39)

, where M1 and M2 are the available reuse-3 subchannels for CMUs and available

reuse-9 subchannels for CRUs, respectively. Likewise, they are subject to the

constraints

2

2 1

0 10

0 3

3 10

N

M

M M N

(5.40)

With M2 = 0, the EFFR-A scheme is just the EFFR scheme. And the EFFR-A scheme

is equal to the EFFR-B scheme, if M2 = 3.

D. EFFR-B Scheme


_ _

EFFR B

SCH avg CEUThr and _ _

EFFR B

SCH avg CCUThr can be calculated just as same as which in the

EFFR scheme. But the mean cell capacity should be reckoned up as

2 _ _ _ _3EFFR B EFFR B EFFR B

SCH avg CEU SCH avg CCUCAP M Thr N Thr (5.41)

, where M2 and N must be conformed to the constraints N > 0 and 3∙M2 + N = 10.

5.2.2 Cell Capacity Comparison

As the performance of EFFR series depends strongly on the N and M combination, the

numerical results for the mean cell capacity versus the number of reuse-1 subchannels

N will be displayed for all studied schemes under LOS, NLOS and combined LOS-

NLOS conditions, assuming S = 30 subchannels in 20 MHz, see Table 5.3.

5.2.2.1 LOS Condition

Under LOS propagation, the SFR scheme outperforms the Reuse-1 and Reuse-3

schemes for both UL (see Figure 5.14a) and DL (see Figure 5.14b). But the

improvement is limited. Using the EFFR series, they never perform worse than using

the SFR and static Reuse schemes with any value of N. With an increasing number of

subchannels for the CCUs N, the enhancement becomes more and more remarkable,

however, at the cost of sacrificing resources for the other users in a cell. This is

(a) (b)

Figure 5.14: Mean cell capacity under the C1 LOS propagation, having the same environment as in

Figure 5.7: (a) mean UL cell capacity; (b) mean DL cell capacity.


because CCUs are close to the BS, so they always can get high quality of CINR, and

thereby use high grade PHY mode to transmit. Hence, a tradeoff between capacity

maximization and fairness should be made. In addition, Figure 5.14 shows that all

EFFR schemes reach similar gains. Nevertheless, together with the results from the

Figure 5.7 in Subsection 5.1.2.1 and Figure 5.13 in Subsection 5.1.2.4, only the EFFR-

A and EFFR-B schemes can provide 100% coverage. And the EFFR-A with M2 = 1

always performs slightly better than the EFFR. As a consequence, the EFFR-A scheme

with M2 = 1 is the best solution for CCI mitigation among all studied schemes in a

cellular LOS scenario.

5.2.2.2 NLOS Condition

Figure 5.15 displays the mean reachable cell capacities of all studied schemes under

NLOS propagation. In UL as shown in Figure 5.15a, the SFR surpasses the Reuse-1

and Reuse-3 schemes more significantly than in DL (see Figure 5.15b). Nevertheless,

this doesn’t mean that the SFR performs better in UL, since the total system

transmission power using SFR is much higher than with the other schemes as

exhibited in Table 5.2. Compared to the performance under LOS condition, the EFFR

series can outperform the other schemes, when the number of exclusively assigned

reuse-1 subchannels is N ≥ 5 in UL and N ≥ 4 in DL, respectively. This is due to the

fact that the RUP techniques are used for Inter-Cell Interference (ICI) mitigation. As

explained in Subsection 5.1.2.4, the NLOS scenario presents actually a noise-limited

(a) (b)

Figure 5.15: Mean cell capacity under the C1 NLOS propagation, having the same environment as in

Figure 5.9: (a) mean UL cell capacity; (b) mean DL cell capacity.


system, which means the benefit from the interference mitigation will not much effect

on the CINR value. Thus, the interference advantage by using the EFFR series can

only compensate the loss of the available bandwidth compared to the SFR scheme,

when more than 60% of the whole frequency bandwidth (N ≥ 4) is used. In both UL

and DL, the EFFR performs slight better than the EFFR-A and EFFR-B schemes.

However, in consideration of the coverage as shown in Figure 5.9, the EFFR cannot

provide 100% coverage, but the EFFR-A and EFFR-B do. Furthermore, the EFFR-A

with M2 = 1 is the second best among the EFFR series. So, with a comprehensive

consideration of cell coverage and mean reachable cell capacity, the EFFR-A with N ≥

5 and M2 = 1 combination is still the best way to alleviate CCI in a cellular NLOS

scenario.

5.2.2.3 Combined LOS-NLOS Condition

In a combined LOS-NLOS scenario, it can be seen in Figure 5.16 that similar to the

NLOS condition the EFFR series cannot provide visible enhancement on the system

capacity until the number of exclusively reserved reuse-1 subchannels is N > 7 in UL

and N > 4 in DL respectively, in comparison with the SFR scheme. Nonetheless,

similar to the presented results in Subsection 5.1.2.4, the SFR and the Reuse-1

schemes can never cover the whole cell area under all propagation conditions.

Accordingly, the main effect with the usage of the exclusive RUP techniques (EFFR

series) in a noise-limited system is to enlarge the cell coverage while maintaining the

(a) (b)

Figure 5.16: Mean cell capacity under the combined LOS-NLOS propagation, having the same

environment as in Figure 5.11: (a) mean UL cell capacity; (b) mean DL cell capacity.


cell capacity of the SFR scheme. Note that in an interference-limited system for

example under the C1 LOS condition, both cell coverage and cell capacity can be

significantly enhanced by using the EFFR series, whereas the inclusive RUP technique

SFR scheme can just slightly increase the cell capacity and limitedly optimize the cell

coverage.

5.2.3 Area Spectral Efficiency

Table 5.8 and Table 5.9 give the cell capacity and the area spectral efficiency for the

two static Reuse schemes, the SFR scheme as well as the EFFR scheme with various

M to N combinations. The cell area depends on the propagation conditions. According

to the cell radii, namely RLOS = 2800 m, RNLOS = 298 m, RLOS-NLOS = 3920 m calculated

in Section 5.1.2, cell areas are therefore 20.3689 km2 under LOS condition, 0.2307

Table 5.8: Mean cell capacity under LOS, NLOS and combined LOS-NLOS propagations (RLOS = 2800

m, RNLOS = 298 m, RLOS-NLOS = 3920 m)

path loss

schemes

LOS NLOS LOS-NLOS UL mean cell

capacity

[Mbps]

DL mean cell

capacity

[Mbps]

UL mean cell

capacity

[Mbps]

DL mean cell

capacity

[Mbps]

UL mean cell

capacity

[Mbps]

DL mean cell

capacity

[Mbps]

Reuse-1 2.3287 2.5308 7.0151 7.0148 4.1781 4.1781

Reuse-3 2.8541 2.9119 8.5446 8.5442 6.4045 6.4045

SFR 4.0058 3.3130 13.8567 11.8236 13.2830 9.6509

EFFR

M:N

= 9:1 3.3085 3.2281 6.4693 6.3898 4.5948 4.5948

M:N

= 8:2 5.0853 4.9308 8.8045 8.6888 6.0225 6.0225

M:N

= 7:3 6.8620 6.6335 11.1397 10.9879 7.4503 7.4503

M:N

= 6:4 8.6387 8.3362 13.4748 13.2869 8.8781 8.8781

M:N

= 5:5 10.4155 10.0389 15.8100 15.5859 10.3059 10.3059

M:N

= 4:6 12.1922 11.7416 18.1452 17.8849 11.7337 11.7337

M:N

= 3:7 13.9690 13.4430 20.4804 20.1839 13.1615 13.1615

M:N

= 2:8 15.7457 15.1470 22.8156 22.4829 14.5893 14.5893

M:N

= 1:9 17.5225 16.8497 25.1508 24.7819 16.0171 16.0171

- M+N =10 subchannels are available for each Reuse-3 cell, refer to Eq. (5.33);

- M+N*δ subchannels are available for each EFFR cell, refer to Eq. (5.36).


km2 under NLOS condition and 39.9231 km

2 under LOS-NLOS condition,

respectively, see Eq. (5.4).

Generally, under NLOS condition, all schemes can reap their maximum cell capacity

and area spectral efficiency, nevertheless with a small cell area. The inclusive RUP

scheme—SFR—can attain a better cell capacity and better area spectral efficiency

under all propagation conditions compared to the static Reuse schemes, but, it still

cannot compete with the EFFR scheme. Under LOS condition, starting from M:N =

8:2 the EFFR scheme can surpass all the other schemes in terms of both cell capacity

as well as the area spectral efficiency. Under NLOS condition, with relatively big N

values (N > 4 in UL and N > 3 in DL, respectively), the EFFR can also gain the best

performance among all schemes. Only under LOS-NLOS condition, with very big N

values in UL (N > 7, which means only a few subchannels can be used by the CEUs in

UL), however, N > 4 in DL, the EFFR can outperform the others. All these again

imply that the exclusive RUP techniques are very helpful, especially for the

interference-limited systems (e.g., under C1 LOS condition), to enhance the cell

coverage, the cell capacity and the area spectral efficiency.

Table 5.9: Area spectral efficiency under LOS, NLOS and combined LOS-NLOS propagations (RLOS =

2800 m, RNLOS = 298 m, RLOS-NLOS = 3920 m)

LOS NLOS LOS-NLOS

UL area

spectral

efficiency

[bps/Hz∙km2]

DL area

spectral

efficiency

[bps/Hz∙km2]

UL area

spectral

efficiency

[bps/Hz∙km2]

DL area

spectral

efficiency

[bps/Hz∙km2]

UL area

spectral

efficiency

[bps/Hz∙km2]

DL area

spectral

efficiency

[bps/Hz∙km2]

Reuse-1 0.00572 0.00621 1.52027 1.52020 0.00523 0.00523

Reuse-3 0.00701 0.00715 1.85173 1.85164 0.00802 0.00802

SFR 0.00983 0.00813 3.00293 2.56233 0.01664 0.01209

EFFR

M:N = 9:1 0.00812 0.00792 1.40198 1.38475 0.00575 0.00575

M:N = 8:2 0.01248 0.01210 1.90805 1.88298 0.00754 0.00754

M:N = 7:3 0.01684 0.01628 2.41412 2.38122 0.00933 0.00933

M:N = 6:4 0.02121 0.02046 2.92017 2.87945 0.01112 0.01112

M:N = 5:5 0.02557 0.02464 3.42624 3.37767 0.01291 0.01291

M:N = 4:6 0.02993 0.02882 3.93231 3.87590 0.01470 0.01470

M:N = 3:7 0.03429 0.03300 4.43838 4.37412 0.01648 0.01648

M:N = 2:8 0.03865 0.03718 4.94444 4.87234 0.01827 0.01827

M:N = 1:9 0.04301 0.04136 5.45051 5.37057 0.02006 0.02006

5.3 Comparison between Analytical and Simulation Results 85

5.3 Comparison between Analytical and Simulation Results

The static Reuse, SFR and EFFR schemes are also implemented in a system-level

simulator OpenWNS (described in Appendix A). And comprehensive performance

evaluations of these frequency reuse schemes by means of computer simulations using

the OpenWNS are presented in Chapter 6 and Appendixes B, C, D. In this section, the

analytical results using a Matlab simulator are compared to the simulation results

performed in the simulation tool OpenWNS, by applying the SFR scheme the EFFR

scheme as well as the Reuse-1 and Reuse-3 schemes using certain scenarios. The

discrepancies between analytical and simulation results are explained and justified.

The simulation results presented in this section can also be found in the succeeding

Chapter 6.

Due to the power constraint of UTs, in UL each CCU using RUP techniques and each

UT using the Reuse-1 scheme can maximally occupy 3 subchannels at the same time,

whilst each CEU using RUP techniques and each UT using the Reuse-3 scheme can at

maximum use 1 subchannel to send packets. In DL, on the contrary, the whole

available bandwidth can be used by the BS in OFDMA-based communication

networks with any scheme. This also results in the equal total system transmission

power of all schemes in DL, but not in UL. Besides, in DL all schemes are assumed to

use the Proportional Fair (PF) scheduling strategy to assign available resources to

traffic links, whereas in EFFR UL, different scheduling strategies are used to allocate

resources of reuse-1 subchannels in the Primary Segment and those in the Secondary

Segment. Hence, in the following, only LOS DL and NLOS DL cases are taken into

account, since the DL implementations in the simulation tool OpenWNS (described in

Appendix A) are more approximate to the conditions assumed in the analytical model

in the Matlab simulator.

In both analytical model and simulation tool, 25 UTs are uniformly distributed within

each hexagonal cell. The total system transmission power is kept constant at 2000mW

except for the SFR UL (see Table 5.2). And for both SFR and EFFR schemes, the

power ratio of high power level to low power level is set as 3. The specific power

allocation for each type of users in each scheme presented in Table 5.2, the main

relevant parameters given in Table 5.3 as well as the switching thresholds for the PHY

modes in Table 5.7 are adopted in this subsection. Furthermore, 20 ms superframe is

assumed, in which using the TDD frame structure 87 and 96 OFDMA symbols in total

are used for up- and down-stream data transmission, respectively. Another difference

from the evaluations in the preceding sections is that scenarios with surrounding

interfering cells just up to the 2nd

-tier are investigated owing to the capability

constraints of hardware used for the simulations. This is also why the two EFFR-

derivatives are not discussed, but only the EFFR scheme is studied in this section.


In what follows, firstly, by using the analytical CINR calculation, the maximal cell

radii and the range definitions for dividing different types of users for RUP techniques

under both LOS and NLOS propagations are dimensioned. Then, with the ascertained

cell radii and range ratios, the analytical and simulation results in terms of cell

coverage percentage, overall cell capacity and spectral efficiency are compared and

(a) received at cell border using C1 LOS (b) received at cell border using C1 NLOS

Figure 5.17: DL CINR received at cell border versus cell radius R: (a) using C1 LOS path loss model; (b) using C1 NLOS path loss model.

(a) under LOS propagation (b) under NLOS propagation

Figure 5.18: DL CINR distribution received by a UT: (a) when the UT traverses the cell with a radius

of 1000m under C1 LOS condition; (b) when the UT traverses the cell with a radius of 220m under C1 NLOS condition.


clarified.

Figure 5.17 on the next page gives the DL CINR at the cell border in LOS and NLOS

scenarios, respectively. It can be seen none of the investigated schemes can provide a

sufficient CINR level of 6.4 dB at the cell border under LOS condition, whereas under

NLOS propagation the Reuse-3 and the EFFR scheme with a maximum cell radius of

220 m can provide sufficient CINR level at the cell border.

Hence, for the next step to determine range ratios for dividing different user-type

zones for RUP schemes, the cell radii of 1000 m (which is assumed in many

publications) and 220 m are chosen for the LOS and NLOS scenarios, respectively.

Figure 5.18 on the next page plots the DL CINR level received by a UT, when the UT

traverses the focused cell across the x-axis. The BS position and the cell borders are

indicated in both figures, and the range of coverage of a scheme is marked by two

stems, whose height indicates the minimum receiver requirement (6.4 dB) for the PHY

mode BPSK½.

Figure 5.18a shows that under LOS propagation the BS can reach the most remote

CCUs at a distance of 288 m by applying the SFR scheme (as visible from the

discontinuity of the CINR curve) with a relatively lower transmission power of 40 mW

compared to the EFFR scheme (see Table 5.2), and 368m by applying the EFFR

scheme with a power of 66.67 mW, respectively. In addition, in a SFR system the

most remote CEUs with a distance of 496 m far away from the BS can receive valid

DL packets. This implies the UTs located over half of the cell radius away from the

BS cannot be served any more. With the usage of the EFFR scheme, by contrast, the

most remote CEUs with a distance of 720 m far away can be reached by the central BS.

In the following performance comparison for the subsequent LOS evaluation, a range

ratio of 0.4 (which means the maximal distance between the BS and the CCUs is 400

m) is chosen to partition the CCUs and the CEUs for both SFR and EFFR scheme.

Although the range ratio of 0.4 is a little large, and with a range ratio of 0.3 seems

more suitable for the SFR, yet an extremely unfairness thereby occurs. In a SFR

system, a range ratio of 0.3 means among 25 users merely 3 users as CCUs can have

67% of the resource to utilize, whilst the remaining 22 users are just allowed to use the

33% bandwidth. Therefore, with a range ratio of 0.4 a trade off is made between

fairness and adequacy. Under NLOS propagation, as shown in Figure 5.18b, the most

remote CCUs at a distance of 120 m away from the BS can be reached by applying the

SFR scheme, whereas a region with a radius of 137.6 m for CCUs can be covered by

the central BS in an EFFR system. Thus, a range ratio of 0.6 is chosen for the

subsequent NLOS evaluation. In other words, the CCU-Zone radius is set to 132 m in

the SFR and EFFR systems. Besides, Figure 5.18b shows that the SFR is still not able

to provide a sufficient CINR level at the cell border under NLOS condition. On the

contrary, the EFFR scheme and the Reuse-3 can.


The concrete overall DL cell coverage percentage performance resulting from both,

analytical models implemented in the Matlab simulator and the simulation

environment OpenWNS, is displayed in Figure 5.19. It can be seen that the

theoretically derived overall cell coverage percentages of all studied schemes coincide

well with experimental results for both LOS and NLOS cases. The slightly

discrepancies are due to the user distribution differences in the simulation scenarios

and in the analytical scenarios. Although evenly distributed users are assumed in both

analytical scenario and simulation scenario, yet the specific location of each user in

both tools is generated randomly and cannot be identical. Under both conditions, the

EFFR scheme and the Reuse-3 can much enhance the cell coverage performance

compared to the Reuse-1 and the SFR scheme. Under LOS propagation the cell

coverage can be tripled by applying the EFFR scheme or the Reuse-3 scheme in

comparison with that using the Reuse-1 and the SFR scheme. Under NLOS

propagation the whole cell can be covered through the EFFR scheme or the Reuse-3

scheme. They are able to improve the cell coverage percentage by around 45%

compared to a Reuse-1 system, and by around 35% compared to a SFR system.

The mean achievable DL cell capacity and the average DL cell spectral efficiency of

all investigated schemes obtained via simulations and from the analysis are presented

in Figure 5.20 and Figure 5.21, respectively. To compare the analytical and simulation

results, MAC and PHY overhead is subtracted in the system capacity calculation.

Figure 5.19: DL cell coverage percentage of all studied schemes under LOS and NLOS propagations.


It can be seen that the simulation results are always lower than the analytical results

under both LOS and NLOS conditions. There are mainly two reasons for this. First, 3

OFDMA symbols as one resource element is assumed in the simulations. And the

scheduler implemented in the simulator allocates resources to every data link by

means of resource elements. That means in some instances a resource element might

not be completely occupied by a link in case this link is satisfied. However, the

(a) under LOS condition (b) under NLOS condition

Figure 5.20: Mean DL cell capacity of all studied schemes: (a) under LOS condition; (b) under NLOS

condition.

(a) under LOS condition (b) under NLOS condition

Figure 5.21: Average DL cell spectral efficiency of all studied schemes: (a) under LOS condition; (b)

under NLOS condition.


remaining free capacity within this resource element cannot be utilized by other links

either. This results in lower efficiency of radio resource usage and leads to lower than

optimal throughput. Secondly, due to the PF scheduling strategy applied in simulations

the unsatisfied links, which in the majority of cases can only use lower PHY modes,

yet often have opportunities to transmit. This causes an overall reduction of capacity

and lower spectral efficiency.

The power (ratio 1:3 used in this studies, see Table 5.2) applied for CCUs and CEUs,

respectively, could be optimized analytically to achieve the highest spectrum

efficiency. For this purpose, an iterative process should be applied to find for LOS and

NLOS conditions, separately, the optimum range ratio. This has not been done in this

work, since analytical and simulation results have some deviations making it

questionable that the value of an analytically (numerically) calculated ―optimum‖ ratio

would be.

5.4 Summary and Conclusion

In this chapter, theoretical analysis of a series of RUP approaches for mitigating ICI in

OFDMA-based cellular environments is carried out. With CINR calculations, the

range definition for dividing different user-type zones for each RUP scheme can be

determined individually. Furthermore, through numerical evaluations, the cell

coverage and the mean cell capacity of all studied reuse techniques under different

propagation conditions are exhibited. The results show that the EFFR series can

outperform the SFR and the static Reuse schemes under any propagation mode, where

significant coverage gains and cell capacity improvements can be achieved by

applying the novel EFFR schemes with adequate resource allocations.

CHAPTER 6

6 Performance Evaluation by Means of Simulation

Performance Evaluation by Means of Simulation

6.1 Simulation Scenario and Simulation Setup ............................................... 92

6.2 Simulation Results ..................................................................................... 101

6.3 Summary and Conclusion ......................................................................... 145

This chapter addresses the detailed performance evaluation of RUP techniques in

OFDMA-based cellular radio networks by means of stochastic event-driven simulation.

The aforementioned EFFR scheme, the SFR scheme and the IFR scheme are

integrated into the so-called WiMAC module, which is an implementation of the IEEE

802.16 standard [42] in the OpenWNS as described in Appendix A. The OpenWNS is

an event-driven simulator comprising emulation of a protocol stack that is embedded

in a simulation environment that cares for stochastic processes representing the real

radio propagation and data communications world. Influences of traffic characteristics,

varying interference and the behavior of protocol related algorithms can be jointly

analyzed in the OpenWNS.

The first section gives a brief description of the simulation environment developed

within the scope of this work and presents the simulation parameters which are the

same in all performance evaluation studies in this chapter.

The comprehensive performance evaluation is presented in the second section, where

the proposed EFFR scheme is compared with the SFR scheme, the IFR scheme as well

as two static Reuse schemes. In this chapter, the focus of interest is not only on the

improvement of the overall system’s cell capacity but also on the cell edge

performance gain as well as the weakest user performance enhancement that can be

achieved by using the exclusive RUP EFFR scheme. Furthermore, relevant factors

which play important roles in the realization of the RUP designs and significantly

influence the system performance are studied and the respective results are compared.

6. Performance Evaluation by Means of Simulation 92

6.1 Simulation Scenario and Simulation Setup

6.1.1 Cellular Scenario

A cellular layout according to Figure 5.1 is used as the underlying geographic scenario,

yet with surrounding cells up to the 2nd

-tier, only. The scenario comprises 19

hexagonal cells, each with a central BS and 25 User Terminals (UTs), which are

randomly positioned per cell, uniformly distributed over the cell area. According to the

analytical results for dimensioning purpose presented in Section 5.3, the cell radius R

is set to 1000 m for LOS propagation scenarios and 220 m for NLOS propagation

scenarios, respectively. The distance 3D R between BSs is therefore 1732 m and

381 m for LOS and NLOS, respectively.

Measurements are only performed in OpenWNS for the central (grey) cell at the

corresponding BS and UTs. Stations in the two tiers of surrounding cells are not

evaluated. They only serve to generate interference to the central cell. Nevertheless,

the same event driven stochastic simulation, with identical average traffic load, and

with the same degree of detail, is conducted at all 19 BSs and 475 UTs. All BSs are

assumed to operate synchronously in TDD. In order to avoid fatal BS-to-BS and SSs-

to-SSs interference, UL and DL subframes are assumed to occur simultaneously in the

19 cells of the network studied. Mobility of the UTs is not considered in this work.

6.1.2 Performance Metrics

In order to estimate the performance of the scenario the following metrics are defined.

All schemes studied in this work are used to mitigate excessive Inter-Cell Interference

(ICI) generated by the neighboring cells, which may severely degrade the performance

of the UTs which are located near the cell edge. Therefore, the performance metrics of

interest will be given not only in terms of the overall cell but also regarding the CEUs

and possibly the weakest users (which are close to the cell borders) separately; and in

both low offered traffic situation as well as in full load situation, respectively.

Carrier Signal strength during packet reception measured at the destination

station’s receiver in dBm. Transmit as well as receive antenna gain is

taken into account. The carrier signal is averaged during packet reception.

Interference Measured at the destination station’s receiver in dBm. All concurrent

transmissions in co-channel cells (inter-cell interference) and all

concurrent transmissions of own cell (intra-cell interference) are

accounted for. The interference level is averaged during packet reception.

It also includes the thermal noise and the noise figure of the receiver.

6.1 Simulation Scenario and Simulation Setup 93

CINR Calculated as the ratio between carrier and interference signal strength in

dB.

Throughput The amount of user data of all MAC packets successfully arriving at the

WiMAX MAC SAP during a fixed time window. The throughput is

measured at the destination station in kilobit per second (kbps). Separate

values are measured for packets travelling from / to every UT in UL / DL

direction.

Cell coverage percentage Measured as the successfully served UTs’

percentage. Packets travelling from / to those UTs in UL / DL direction

can be received with the satisfied CINR condition, which is here 6.4 dB

for the robust PHY mode BPSK1/2.

Spectral efficiency Measures the achieved aggregate throughput normalized to

the available system bandwidth. This metric is usually given in Bits per

second per Hz (bps/Hz), and allows a convenient comparison of radio

technologies independent of the available bandwidth.

All schemes studied in this work are used to mitigate excessive ICI generated by the

neighboring cells, which may severely degrade the performance of the UTs which are

located near the cell edge. Therefore, the performance metrics of interest will be given

not only in terms of the overall cell but also regarding the CEU-traffic and possibly the

weakest-user-traffic (which are very close to the cell borders), separately. In addition

to that, the system performance from low to full traffic load will be concerned and

evaluated.

6.1.3 Link Adaptation and Error Modeling

The link adaptation is performed in the WiMAC scheduling strategy. For each packet,

according to the estimated CINR a Modulation and Coding Scheme (MCS) is chosen.

Table 6.1: Switching thresholds and PHY data rates per subchannel for Modulation and Coding Schemes (PHY modes)

Modulation Code rate Min. CINR

[dB]

PHY data rate per

subchannel [Mbps]

BPSK 1/2 6.4 0.233

QPSK 1/2 9.4 0.467

QPSK 3/4 11.2 0.7

16QAM 1/2 16.4 0.933

16QAM 3/4 18.2 1.4

64QAM 2/3 22.7 1.867

64QAM 3/4 24.4 2.1


Table 6.1 depicts the CINR threshold of each MCS with a maximum BER of 10-6.

These values are taken from the IEEE 802.16 WiMAX standard [42]. The estimated

CINR determines the MCS with the next lower threshold. In order to avoid

transmission errors, conservative CINR thresholds are chosen. Thus, as long as the

CINR is estimated accurately, transmission errors are kept even below 10-6 in our

simulation studies.

6.1.4 Frame Structure and Overhead

The IEEE 802.16m basic frame structure introduced in [41] is adopted as the MAC

frame used in the simulation studies, see Figure 6.1. Each 20 ms superframe is divided

into four equally-sized 5 ms radio frames. According to the standard, 2048 data

carriers are available in 20MHz channel bandwidth. Each OFDMA symbol is 102.857

μs long. Using the TDD frame structure, TTG and RTG with a total length of 165.714

μs are inserted between the DL and UL switching points in each frame, resulting in a

whole number of 47 OFDMA symbols in each 5 ms frame.

Every superframe contains a Superframe Header (SFH) and a Bandwidth (BW)

request and initial ranging phase, which are located in the first frame. The SFH

includes the preamble (1 OFDMA symbol), the FCH as well as the DL and UL MAPs

(2 OFDMA symbols). The BW-request and initial ranging phase needs 2 OFDMA

symbols. As a result, 24 OFDMA symbols but 18 OFDMA symbols can be used for

Figure 6.1: IEEE 802.16m basic frame structure for 20 MHz channel bandwidth [41].


DL and UL traffic in the first frame, respectively. Other than the first frame, each of

the remainder 3 frames is equally divided between DL and UL data phases, in

consideration of the symmetric traffic load assumed. That means 24 OFDMA symbols

and 23 OFDMA symbols are applied for DL and UL data transmission, respectively.

The OpenWNS-based implementation is assumed to transmit FCH, DL and UL MAPs

without errors. This idealization allows evaluating the RUP enabled network even

when the control signaling would have been the system’s bottleneck. In the

simulations, the lengths of UL and DL MAPs are fixed and they are always

transmitted using the most robust MCS BPSK ½.

6.1.5 Resource Allocation and Scheduling

6.1.5.1 Available Bandwidth

An OFDMA system with a 20 MHz system bandwidth is investigated in this work.

The frequency channel is subdivided into 2048 subcarriers, which are grouped into 30

subchannels. The whole 20 MHz bandwidth can be used by the Reuse-1 scheme, the

IFR scheme and the SFR scheme. On the contrary, each cell applying the Reuse-3

scheme and the EFFR scheme can access just part of the system bandwidth. Only one

third of the whole bandwidth can be utilized by the Reuse-3 cell. And the EFFR

scheme with an increasing number of dedicated reuse 1 subchannels N can own more

Table 6.2: Available bandwidth for the cells using the EFFR scheme with different M to N combination

compared to the SFR, IFR and two static Reuse schemes.

Scheme

Nr. of

reuse 1

subchannels

Nr. of

reuse 3

subchannels

Total

available

subchannels

Available

bandwidth

percentage

EFFR

(M:N)

9:1 3 9 12 40.0%

8:2 6 8 14 46.7%

7:3 9 7 16 53.3%

6:4 12 6 18 60.0%

5:5 15 5 20 66.7%

4:6 18 4 22 73.3%

3:7 21 3 24 80.0%

2:8 24 2 26 86.7%

1:9 27 1 28 93.3%

SFR 20 10 30 100%

Reuse-3 0 10 10 33.3%

IFR 30 0 30 100%

Reuse-1 30 0 30 100%


and more available resources, though at the cost of losing the precious reuse 3

subchannels for the CEUs, see Table 6.2.

6.1.5.2 Power Allocation

In view of a constant system power assumed, the maximal transmission power for the

BS and UTs is fixed to 2 W and 200 mW, respectively. Different cell-specific power

masks in response to diverse spectrum usage are applied to all studied approaches as

given in Figure 6.2. A power mask describes the fraction of the maximum transmit

power which the BS or a UT uses depending on the part of the spectrum.

For the Reuse-1 scheme and the IFR scheme, since every cell may use the whole

system bandwidth, a uniform transmission power allocation is thereby used, see

Figure 6.2a. The same fraction of the overall transmission power is assigned to all

parallel transmissions over the whole bandwidth.

For the SFR scheme (Figure 6.2b) and the EFFR scheme (Figure 6.2c), like mentioned

in Sections 3.2.2 and 4.1.2.3, two different power levels are applied to the CCUs and

the CEUs. In Section 5.3 and subsequent Sections 6.2.1 and 6.2.2, the high power level

is set to be triple to the low power level. A discussion of the performance

corresponding to diversified power ratios between CCUs and CEUs is presented in

Section 6.2.3. Note that the both power levels for CCUs and CEUs in the EFFR

scheme are higher than in the SFR scheme for DL traffic (see Table 5.2), whereas in

UL they remain identical (see Table 5.2). This is because the whole bandwidth is

available for the SFR, whereas the EFFR uses only part of the total system bandwidth,

see Table 6.2. Thus, in DL, the BS can offer higher power to each subchannel for the

EFFR scenario than for the SFR scenario. Different from the DL transmissions, in UL,

each CEU sends packet on a subchannel with its full power, and each CCU employs

one third of its power on a subchannel to avoid generating excessive ICI for the

neighboring cells. As a consequence, each CCU can maximally occupy three

P(f)

f

Cell A

P(f)

f

Cell B

P(f)

f

Cell C

P(f)

fP(f)

fP(f)

f

P(f)

f

P(f)

f

P(f)

fP(f)

fP(f)

f

P(f)

f

(a) Reuse-1 & IFR (b) SFR (c) EFFR (d) Reuse-3

Figure 6.2: Different cell-specific power masks over system bandwidth for all studied approaches

including the Reuse-1 scheme, the IFR scheme, the SFR scheme, the EFFR scheme and the Reuse-3

scheme.


subchannels simultaneously (depending on the power ratio of the high power level to

the low power level), whilst each CEU can maximally use one subchannel in UL phase.

As the BS holds sufficient power to dominate the whole available bandwidth, such

connection constraints do not occur in DL.

With the Reuse-3 scheme, the transmission power over the available bandwidth is

tripled to that in the Reuse-1 system, but just on one third of the spectrum

(Figure 6.2d). Each UT may use maximally one subchannel to transmit data in a

Reuse-3 scenario, whilst a Reuse-1 UT can possess at maximum three subchannels

concurrently.

6.1.5.3 Scheduling

The resource assignment and OFDMA scheduling process is optimized for RUP

schemes in OpenWNS to handle not only full load situations but also low and

moderate traffic load situations. Both the SFR scheme and the EFFR scheme resort to

the ICI-dispersion mechanism of the IFR scheme to avoid unnecessary ICI in low and

mitigate ICI in moderate traffic load case by applying specific frequency allocation

patterns (sequences) among neighboring cells. The resource allocation carried out by

each cell among directly neighboring cells start from different subchannels up.

The scheduling of data packets in OFDMA systems is required to be able to exploit the

multi-user diversity and provide fairness at the same time. Hence, in each cell, the BS

scheduler allocates resources for both DL and UL directions using the Proportional

Fair (PF) scheduling strategy. PF is a compromise-based scheduling algorithm, which

aims at maintaining a balance between maximization of system throughput and

fairness among UTs, which allows all users at least a minimal level of service. This is

done by assigning each data flow a data rate that is inversely proportional to its

anticipated resource consumption. A consequence of this is that UTs that can apply a

high PHY mode will be assigned less radio resources than other UTs that are heavier

interfered and therefore must use lower PHY modes to carry the same amount of

traffic. The scheduling process is outlined in Figure 6.3. In each allocation step, a

single connection is selected based on the PF strategy. Then, the scheduler reserves

resources for this connection taking into account that these resource blocks should be

free (not have been allotted to other connections yet), and at the same time the

transmitter of this connection still has enough remaining power which can be spent on

these resource blocks. When suitable resources are found, the PHY mode selection

will be performed according to the allocated power used for this connection and

expected CINR at the receiver. After that, the resource occupation state of the

upcoming superframe and the power allocation for this connection are updated in the

database at the BS. The scheduling algorithm successively allocates resources to data

connections until all resources are assigned, or no more MAC PDUs need to be


transmitted, or transmission power is exhausted (only in UL, since each UT has

limited transmission power). Detailed resource allocation for each of the studied

schemes is described in Chapters 3 and 4.

The scheduling process is accomplished before starting the upcoming superframe. And

its duration is not counted in the simulation time.

Connection Selection

start of scheduling cycle

Resource Selection

Power Allocation

PHY Mode Selection

Database Adaptation

TRUE

FALSE

TRUE

TRUE

X

FALSE

FALSE

end of scheduling cycle

Resource

available?

MPDUs

pending?

Power

available?

Figure 6.3: Resource scheduling process


In order to eliminate other influencing factors when evaluating the performance of

RUP techniques neither Segmentation-and-Reassembly (SAR) nor Automatic Repeat

Request (ARQ) protocols are used.

6.1.6 Simulation Parameters

The overall system configuration is chosen according to band class index 2 as defined

by the WiMAX profile [44]. Hence, an IEEE 802.16m system operating in TDD mode

is considered. A bandwidth of 20MHz with a mid frequency of 5.47 GHz is used. The

superframe length is set to 20 ms. Thermal noise of -174dBm/MHz is considered and a

noise figure of 5 dB for BS and 7 dB for UT are added separately to that.

All simulation studies are performed with a fixed packet size of 96 Bytes. The Inter-

Arrival Time (IAT) of packets is controlled so that the desired offered traffic is

achieved. It follows a negative exponential distribution. For all UTs a symmetric

Table 6.3: Simulation parameters and values assumed

Parameter Value Comment

System bandwidth 20 MHz

Center Frequency 5470 MHz CEPT band B

Subcarriers (FFT size) 2048

Total data carriers 1440 Available for data transmission

Data carriers per subchannel 48 Available for data transmission

OFDMA symbol duration 102.857 μs

Number of subchannels 30

Superframe length 20 ms 194 OFDMA symbols

Number of interfering cells 18 Up to the 2nd-tier

Cell radius LOS path loss 1000 m

NLOS path loss 220 m BPSK ½ at the cell border

UTs per cell 25

Range for weakest users 176 m – 220 m Only defined and evaluated in NLOS

scenarios

Path loss

model LOS path loss 23.8log(d)+41.9 WINNER LOS C1 Metropolitan suburban,

including antenna gain 11.8 dB NLOS path loss 40.2log(d)+27.7

Thermal noise density -174 dBm/MHz

Noise figure at [BS, UT] [5, 7] dB

Transmission power [BS, UT] [2000, 200] mW Fixed for both BS and UT

Traffic model Symmetric Negative exponential IAT

Packet size 96 Byte Fixed

Link adaptation Adaptive CINR threshold in Table 6.1

SAR None

ARQ None


traffic load in DL and UL is assumed. All BSs and UTs in the scenario are equipped

with omni-directional antennas.

In this chapter, two path loss propagation conditions (namely, the WINNER LOS and

NLOS suburban C1 Metropol path loss model [35], see Eqs. (5.28) and (5.29)) will be

considered in the following performance evaluations. The corresponding radii used for

LOS is 1000 m and 220 m for NLOS, respectively. The proposed EFFR scheme is

evaluated with four M to N combinations (8:2 | 7:3 | 6:4 | 5:5) under LOS condition

and (7:3 | 6:4 | 5:5 | 4:6) under NLOS condition, respectively.

According to the results presented in Section 5.3, the maximum valid reach by using

the Reuse-1 scheme is approximately 0.4 of the cell radius of 1000m under LOS

condition, whilst 0.6 of the cell radius of 220m under NLOS. Thus, the range ratio r/R

to divide UTs into CCUs and CEUs for the SFR scheme and the EFFR scheme is set

as 0.4 for LOS and 0.6 for NLOS in the subsequent scenarios in Sections 6.2.1 and

6.2.2, where r is the maximal distance between the BS and the CCUs and R is the cell

radius. That is to say, among 25 UTs, 5 of them are CCUs and the remaining 20 UTs

are CEUs in LOS scenarios, whereas there are 12 CCUs and 13 CEUs in NLOS

scenarios. As indicated in Section 5.3, high cell capacity and 100% coverage can be

achieved by exploiting the EFFR scheme under NLOS condition. Hence, deeper

insight into the performance of the weakest users, located close to the cell edges

between 176 m and 220 m away from the BS (i.e., R'/R [0.8, 1]), will be gained.

Table 6.3 presents all relevant simulation parameters together with the values assumed.

In addition, neither shadowing nor fast fading is considered in the following.

All performed simulations compromise 5 runs, each with different random seed,

resulting in a different random distribution of the stations per run. The results of the 5

runs are averaged, in order to have a more accurate evaluation of the system, with a

higher statistical relevance. The simulation time for each run is 2 s, which in reality

can be accomplished in almost 400 hours (more than two weeks), and needs 2 gigabyte

memory to store the data, since the used scenarios are huge, each consisting of 19 cells

including 19 BSs and 475 UTs. All simulation results presented in this work took more

than six months to be finished and consume large memories. Owing to the time

constraint and the capability constraints of hardware, we have not been able to run

simulation experiments with more than 5 seeds.

6.2 Simulation Results

6.2.1 Reference Scenario

The system configuration studied in this section is named ―Reference Scenario‖ which

is taken into account in the following to compare the performance of OFDMA-based

communication networks using EFFR scheme and other frequency reuse techniques

such as SFR, IFR, Reuse-1 as well as Reuse-3. Through an investigation of the five

frequency reuse techniques under different propagation conditions, a thorough and

comprehensive assessment of its strengths and weaknesses can be made.

Results of the Reference Scenario will be compared to that from other scenarios, see

Sections 6.2.2 and 6.2.3 that apply different range ratios and power ratios, respectively.

In what follows, the overall system performance and detailed observation on the

performance of different types of UTs are presented. Through individually dissecting

the performance of the CCUs and the CEUs, the reasons for the overall performance

difference among all frequency reuse techniques can be found. Moreover, the fairness

among individual types of UTs can also be disclosed.

6.2.1.1 Performance in LOS DL

Figure 6.4 displays the mean DL carrier signal strength perceived by the CCUs and the

CEUs, respectively. As expected, using the EFFR scheme, the mean CCU carrier

strength is very close to that using the Reuse-1 scheme, whereas the mean CEU carrier

strength is similar to that using the Reuse-3 scheme. Since the overall BS transmission

power is fixed, and using the SFR scheme the BS transmission power is shared among

all 30 subchannels, the power assigned to each subchannel for the CCUs is thereby

lower than for the subchannels with the Reuse-1 scheme, and the power allotted to

each subchannel for the CEUs is lower than for the subchannels with the Reuse-3

scheme. This leads to the mean DL CCU carrier strength of the SFR scheme being 2

dB lower than with the Reuse-1 scheme and the EFFR scheme, and the mean DL CEU

carrier strength being 2 dB lower than using Reuse-3 or EFFR schemes.

The main idea of the RUP techniques is through mitigating ICI from neighboring cells

to enhance the cell edge performance and the overall system capacity in a cellular

system. Figure 6.5 exhibits the mean DL interference level perceived by the CCUs and

the CEUs of the five investigated frequency reuse schemes. Compared to the Reuse-1

scheme, the ICI can be reduced by both, SFR and EFFR. For both CCUs (Figure 6.5a)

and CEUs (Figure 6.5b), the interference by using the EFFR scheme can be reduced

more effectively than using the SFR scheme, and for CEUs tends closely to the Reuse-

3 scheme.


In Figure 6.5a, it can be seen that, in general, the mean interference suffered by the

CCUs in the EFFR system is lower than in the Reuse-1 and the IFR scheme. This is

because the CCUs of the EFFR scheme are only interfered by the CCU transmissions

in the neighboring cells that are close to their BSs, but relatively far from the CCUs in

(a) Mean DL carrier perceived by CCUs (b) Mean DL carrier perceived by CEUs

Figure 6.4: Mean DL carrier signal strength perceived by different types of UTs versus offered traffic

per user under LOS condition. 25 UTs are uniformly distributed in each cell with a cell radius of 1000 m. The range ratio r/R of 0.4 for partitioning CCUs and CEUs is assumed. And for both SFR and

EFFR schemes, the power ratio of high power level to low power level is set as 3. (a) Mean DL carrier

signal strength perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs.

(a) Mean DL interference perceived by CCUs (b) Mean DL interference perceived by CEUs

Figure 6.5: Mean DL interference level perceived by different types of UTs versus offered traffic per user under LOS condition, having the same environment as in Figure 6.4: (a) mean DL interference

level perceived by CCUs; (b) mean DL interference level perceived by CEUs.

6.2 Simulation Results 103

the focused cell. In contrast, the CCUs using the Reuse-1 scheme or the IFR scheme

are very probably interfered by the CEUs in the neighboring cells, which are relatively

closer to the CCUs in the focused cell. As visible, the more exclusive reuse-1

subchannels N in the Primary Segment of the EFFR scheme are preassigned to the

CCUs, the more interference reduction can be gained at the CCUs, owing to the

adoption of the interference-disperse mechanism from the IFR scheme. At low to

moderate and with large N, even under high traffic load, the EFFR CCUs obtain even

less ICI than the Reuse-3 CCUs. This results from two reasons. One is that the

resource allocation is just like using the FRF of 3 for the CCUs at low offered traffic

due to adaptation of IFR mechanism. The other is that the transmission power for each

EFFR reuse-1 subchannel is only one third of the transmission power used on each

subchannel applying the Reuse-3 scheme. In this way, the interference perceived by

the UTs using the EFFR reuse-1 subchannels is correspondingly much lower.

However, with increasing traffic load, the resource assignment for the EFFR CCUs

varies gradually from reuse-3 to reuse-1, so that the interference advantages owing to

the lower power does not pay off the interference exacerbation caused by the increased

co-channel transmissions. On the contrary, the EFFR CEUs apply exactly the

exclusive reuse-3 mechanism. Thus, as shown in Figure 6.5b, they perform almost

identical, yet slightly inferior to the Reuse-3 scheme, since the reuse-3 mechanism in

the EFFR scheme is carried out among CEUs, whilst in the Reuse-3 among all UTs.

Regarding the SFR scheme, since it applies FRF of one to the CCUs and FRF of three

to the CEUs, the ICI can be alleviated for its CCUs at low traffic load (≤ 50 kbit/s in

Figure 6.5a) compared to the Reuse-1 scheme due to the adoption of the IFR

mechanism and lower transmission power. Nevertheless, the ICI at its CEUs (refer to

Figure 6.5b) is still much higher than with the Reuse-3 scheme and the EFFR scheme

at moderate to high load (> 150 kbit/s). This inherent disadvantage is mainly due to its

inclusive RUP design as explained in Section 5.1, where the CEUs suffer not only the

same quantities of ICI as the Reuse-3 CEUs, but also the extra ICI from the CCUs’

transmissions in the remaining Reuse-1 cells.

With SFR, its low CCU carrier signal strength (refer to Figure 6.4a) and the high CCU

interference level (refer to Figure 6.5a) lead to the worst CCU CINR level in DL

among all studied reuse schemes, as shown in Figure 6.6b. Note that the mean CCU

CINR level of the SFR is even lower than with the Reuse-1 scheme, but still above 6.4

dB, the required CINR threshold. In contrast to the SFR, the mean CCU CINR level of

the EFFR scheme drops off from the Reuse-3 CCU CINR level to the Reuse-1 CCU

CINR level, with increasing traffic load and a large value of N is recommended. As for

the mean CEU CINR level (see Figure 6.6c), although the SFR scheme outperforms

the Reuse-1 and the IFR schemes, its CEU CINR is barely able to reach 3.7 dB at high

load which is far below the required CINR threshold of 6.4 dB. Among all schemes,


only the CEUs using the EFFR scheme or the Reuse-3 scheme can achieve the

required DL CINR level of 6.4 dB. Figure 6.6a gives the mean overall DL CINR

levels. Both the EFFR and the SFR schemes can significantly increase the mean

overall CINR compared to the IFR and the Reuse-1 schemes. The mean overall CINR

attained by using the EFFR scheme is very close to and sometimes even higher than in

a Reuse-3 system. In contrast, although the mean CINR can be substantially increased

(a) Mean overall DL CINR

(b) Mean DL CINR perceived by CCUs (c) Mean DL CINR perceived by CEUs

Figure 6.6: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs as a function of offered traffic per user under LOS condition, having the same environment as in Figure 6.4: (a) mean overall DL CINR values; (b) the corresponding mean DL

CINR values perceived by CCUs; and (c) the corresponding mean DL CINR values perceived by

CEUs.


through the SFR scheme compared to the Reuse-1 scheme and the IFR scheme, it is

still worse than using the EFFR or the Reuse-3 scheme due to its inclusive reuse

mechanism. Applying SFR, the interferences experienced on both reuse-1 and reuse-3

subchannels are higher than by using the EFFR and the Reuse-3 scheme. Besides, it

can be seen that the curve for the IFR scheme confirms the explanation given in

Section 3.2.3. The IFR scheme disperses the interference over the whole bandwidth, so

that the mean CINR can be strongly enhanced at low load (e.g., 50 kbit/s). But with

increasing traffic load, the CINR degrades rapidly, and no improvement can achieved

in full load and overload situations.

Nevertheless, the profits in CINR do not imply an enhancement of capacity for certain.

Figure 6.7 presents the corresponding mean overall DL MAC throughput, as well as

the mean DL CCU throughput and the CEU throughput of the five investigated

frequency reuse schemes, respectively. It can be seen that the mean CCU throughput

of all schemes (see Figure 6.7b) does not correspond to the respective CINR levels

displayed in Figure 6.6b. Although the Reuse-3 CCUs can get the best DL CINR level,

their throughput stays lowest. Contrarily, the SFR performs much better than the

Reuse-3 scheme despite of its worst CCU CINR level. The reason is that the Reuse-3

has at most one third of whole bandwidth (10 subchannels here) for all UTs to utilize,

whereas the SFR scheme possesses 20 subchannels alone for CCUs. This means for

the Reuse-3 CCUs the benefits in CINR cannot compensate the loss in the bandwidth.

On the other hand, although the SFR CCUs outperform the Reuse-3 CCUs in

throughput when the offered traffic per user exceeds 50 kbit/s, yet the mean SFR CEU

throughput (see Figure 6.7c) deteriorates rapidly from that point. And in the end it can

only reach one fourth of the mean Reuse-3 CEU throughput. The essential reason is

the unfairness of frequency resource allocation. In an SFR system with a range ratio

r/R of 0.4 used for the LOS scenario, only 5 of 25 UTs are CCUs, the other 20 are all

CEUs. This means that two third of the available bandwidth (20 of 30 subchannels) are

exclusively used by 20% (5 of 25) UTs, whereas the remaining 80% have just a subset

of 33% available bandwidth to utilize. From results in Figure 6.6c, it can be seen that:

although higher power share on the SFR CEUs may improve their CINR level, and

thereby to enhance their throughput compared to the Reuse-1 CEUs, the unfairness of

the frequency resource distribution severely restricts and degrades the CEUs’

performance. So, the overall capacity of the SFR scheme is based on total unfairness,

sacrificing throughput of the CEUs favor of CCUs (see Figure 6.7a).

The results in Figure 6.7a show that the Reuse-3 scheme outperforms both the Reuse-1

and the IFR scheme, although only one third of the whole bandwidth is available per

cell with the Reuse-3 scheme. Combined with the CINR results in Figure 6.6, it

implies that the gain in CINR by using the Reuse-3 can compensate the loss in its

bandwidth. However, it cannot beat the EFFR scheme. The lack of the frequency

resources by using the Reuse-3 scheme limits its capacity, although it attains similar

CINR values as the EFFR scheme, see Figure 6.6a. With the IFR scheme, the cell


throughput can be considerably improved above that of Reuse-1 at low (up to 50

kbit/s) and moderate (from 50 kbit/s to 150 kbit/s) traffic loads. However, as explained

before, it cannot surpass the Reuse-1 scheme in full load and overload situations (≥

150 kbit/s). The EFFR inherits the merit of the IFR scheme, so that it behaves as a

Reuse-3 scheme at low traffic load (up to 50 kbit/s). The SFR scheme adopts the IFR

mechanism either, and performs much better than the Reuse-1 scheme. However, its

performance is still inferior to that gained with the EFFR scheme. This is because even

if the Major Segment is sufficient for serving all UTs (which means the exclusive

(a) Mean overall DL cell throughput

(b) Mean CCU DL throughput (c) Mean CEU DL throughput

Figure 6.7: Mean DL MAC throughput under LOS condition as a function of offered traffic per user

in the same environment as in Figure 6.4: (a) mean overall DL cell throughput; (b) the corresponding

mean CCU DL throughput; and (c) the corresponding mean CEU DL throughput.


reuse-3 occurs in the SFR system in this case), the CCUs are still allowed to send

packets with lower transmission power, which leads to the poorer cell capacity than

using the EFFR and no clear advantage compared to the Reuse-3 scheme.

Here it should be noted that with increased traffic load the saturation point of a system

is reached, once the throughput does not increase linearly any more. For example in

Figure 6.7, the saturation point by applying the Reuse-3 is at about 50 kbit/s load per

user (see Figure 6.7b and c), which cumulates to a system throughput of around 955

kbit/s (see Figure 6.7a). When the Reuse-1 scheme is used, the system is saturated at

app. 700 kbit/s with 150 kbit/s offered traffic per user (see Figure 6.7a). Other than

with these static Reuse schemes, the saturation points by using the ICI-coordination

designs (introduced in Chapters 3 and 4 including the IFR, the SFR and the EFFR) in

this chapter (and in Appendix B, C and D) are not in compliance with this rule. This is

because, as mentioned in Section 6.1.5.3, in OpenWNS the resource assignment and

OFDMA scheduling process is optimized in RUP schemes to handle not only full load

situations but also low and moderate traffic load situations. Both the SFR scheme and

the EFFR scheme adopt the ICI-dispersion mechanism of the IFR scheme to avoid

unnecessary ICI in low and mitigate ICI in moderate traffic load case by applying

specific frequency allocation patterns (sequences) among adjacent cells. The resource

allocation carried out by each cell among adjacent cells starts from different

subchannels up. As a result, the mean throughput under the IFR, the SFR or the EFFR

may also increase over-proportionally with increased offered traffic, before their

saturation points are reached. For example, as elucidated in Section 3.2.3, using the

IFR scheme, the system operates with increasing traffic load like sliding from a Reuse-

3 system to a Reuse-1 system. The results in Figure 6.7a, b and c show that the

throughput curves of the IFR are completely consistent with the curves of the Reuse-3

scheme until the traffic load increases to 50 kbit/s. This implies that up to 50 kbit/s

offered traffic, only 1/3 of the whole available bandwidth is used in each cell (with

loading factor ≤ 0.3, as explained in Section 3.2.3). With a further increased traffic

load, the IFR disperses the ICI over the whole bandwidth and can surpass the Reuse-1

scheme (see Figure 6.7a), but its CEUs are still interfered severly (see Figure 6.7c).

And in a full-load situation at about 150 kbit/s load per user (see Figure 6.7a, b and c),

the IFR system cannot perform better than the Reuse-1 scheme. Hence, this point is

actually the saturation point by applying the IFR with a system throughput of around

700 kbit/s.

Under static Reuse schemes all UTs always have the same saturation point, see

Figure 6.7b, c, where both, CEUs and CCUs are saturated at 50 kbit/s load per user.

An overall system saturation point results at about 955 kbit/s at 50 kbit/s offered traffic

per user, see Figure 6.7a. This is because UTs are served the same way, i. e. CEUs and

CCUs in static Reuse schemes are not handled, differently. With ICI-coordination

RUP schemes (including SFR and the EFFR), CEUs and CCUs are handled,

differently. In SFR, CEUs are only allowed to access the Major Segment (1/3 of the


system bandwidth), whereas the whole bandwidth is available for CCUs. Moreover,

transmission power and scheduling of CEUs and CCUs are also different in the SFR,

namely, CEUs are served by the Major Segment, preferentially, with high power. In

EFFR, the system bandwidth for both, CEUs and CCUs is same. But, on the reuse-3

subchannels, CEUs have higher priority for occupying subchannels than CCUs. In

addition, secondary reuse-1 subchannels under EFFR are scheduled based on CQI-

estimation. All this together leads to CEUs and CCUs having different saturation

points under ICI-coordination RUP schemes. For example in Figure 6.7b and c, CCUs

under SFR are saturated at 350 kbit/s load per user, whereas CEUs are saturated at 50

kbit/s load; and CCUs under EFFR with given M to N combination, say 6:4, are

saturated at 400 kbit/s load per user, whereas CEUs are saturated at 50 kbit/s load.

Once all the available bandwidth of a system is used up with increasing traffic load,

the system reaches its saturation point, which for the investigated ICI-coordination

schemes in this chapter is the point, where all UTs (including both CCUs and CEUs)

are saturated. Thus, as shown in Figure 6.7a, the SFR system reaches its saturation

point at about 1.53 Mbit/s with 350 kbit/s offered traffic per user. The saturation point

of the EFFR system with M:N = 8:2 is at about 1.68 Mbit/s with 200 kbit/s load per

user; of the EFFR with M:N = 7:3 at about 2.06 Mbit/s with 300 kbit/s load per user; of

the EFFR with M:N = 6:4 at about 2.46 Mbit/s with 400 kbit/s load per user; as well as

of the EFFR with M:N = 5:5 at about 2.86 Mbit/s with ≥ 500 kbit/s load per user,

respectively.

Among all investigated schemes, the EFFR scheme outperforms all the other schemes

in almost every situation, regardless of which M to N combination is used. Under the

LOS propagation, the EFFR with any M:N combination can achieve a cell capacity

Figure 6.8: Cell capacity and the corresponding mean user throughput of all studied schemes in LOS

DL, having the same environment as in Figure 6.4.


above 1.6 Mbit/s in DL, see Figure 6.8. And the EFFR with M:N = 5:5 gains the best

performance of all schemes. It reaps an immense gain of app. 300% compared to the

Reuse-1 scheme, and an app. 200% gain over the Reuse-3 as well as gets an advantage

of app. 87% over the SFR scheme.

To assess the performance of schemes for ICI mitigation in a multi-cellular system, the

achievable cell throughput should be concerned combined with another important

(a) Mean overall DL cell coverage percentage

(b) Mean DL CCUs coverage percentage (c) Mean DL CEUs coverage percentage

Figure 6.9: Mean DL coverage percentage of five frequency reuse schemes versus offered traffic per

user under LOS condition, having the same environment as in Figure 6.4: (a) mean overall DL cell

coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs coverage percentage in DL traffic.


performance indicator, namely, the cell coverage percentage. Figure 6.9a exhibits the

mean DL cell coverage percentage versus the traffic load per user in each cell under

LOS condition. The results in Figure 6.9a show that the EFFR scheme can have a very

close performance to the Reuse-3 scheme, which is the best among all schemes in

terms of cell coverage. Yet there is still an around 35% coverage gap in DL. Although

by applying the EFFR scheme the full cell coverage cannot be achieved under the LOS

condition, it still significantly increases the cell coverage by about 45% for DL over

the Reuse-1 scheme and the IFR scheme. The SFR scheme, on the contrary, only

slightly improves the coverage performance at full load and overload situations (≥ 350

kbit/s).

Figure 6.9b and Figure 6.9c give the corresponding CCUs coverage percentage and

CEUs coverage percentage of the five investigated frequency reuse schemes in LOS

DL case, respectively. The utilization of the EFFR scheme allows all UTs including

both CCUs and CEUs to have a very close performance to the Reuse-3 scheme. By

using the Reuse-1 and IFR schemes, more than 90% of their CCUs can also be validly

served. However, the performances of their CEUs are extremely poor that none of

them can be reached by the BS when offered traffic over the saturation point at 150

kbit/s, as shown in Figure 6.9c. With the EFFR scheme in contrast, the mean CEUs

coverage percentage can be improved up to around 56%. As concerning the SFR

scheme, with the results presented in Figure 6.9c, it can be seen that the mean CEUs

coverage can also be ameliorated by about 18% in full load situation (350 kbit/s).

However, only 65% of its CCUs can be validly served in this case (see Figure 6.9b)

and even lower in overload situations (> 350 kbit/s), which is much inferior to the

Reuse-1 scheme. This is because in DL the power level assigned to the SFR CCUs is

lower than with the Reuse-1 scheme (refer to Figure 6.4b), which leads to the lower

mean CCU CINR values (refer to Figure 6.6b). The poor CCU coverage percentage

also implies that the assumed range ratio r/R of 0.4 is not adequate, in fact too big for

the SFR scheme under LOS propagation (refer to Figure 5.18a). On the other hand, the

range ratio of 0.4 means the number of CCUs is only 5. And the 5 CCUs (i.e., 20% of

the UTs) can use 67% of the resources (here 20 subchannels), whereas the other 20

CEUs have just the remaining 33% bandwidth to utilize. It is quite unfair for the CEUs

in this case. If the range for the CCUs is set further reduced, the unfairness for CEUs

would be further enlarged. As a consequence, no matter how much the traffic demands

of the CCUs is, the CEUs would be starved even in a very low offered traffic situation.

Deeper observation on the performance of the RUP techniques depending on the range

ratios is presented in Section 6.2.2.

Together taking the mean DL cell capacity into consideration (refer to Figure 6.7 and

Figure 6.8), the exclusive RUP technique EFFR outperforms all the other schemes and


can gain substantial improvements in terms of both the overall cell capacity as well as

the cell coverage under LOS path loss propagation.

Accomplishment of high system capacity depends on two decisive factors: one is the

CINR level at the receiver; the other is the available bandwidth in the cell. In order to

enhance the system capacity, both should be concerned. For instance, though the best

CINR level can be achieved in a Reuse-3 system among all studied schemes (refer to

Figure 6.6), its mean cell throughput is inferior to the EFFR scheme due to the lack of

the available bandwidth (refer to Table 6.2). On the contrary, a SFR cell owns a

bandwidth three times larger than a cell using the Reuse-3 scheme. However, its

achievable suboptimal CINR level results in that its mean throughput performance

cannot outperform the Reuse-3 system in the majority of cases (refer to Figure 6.7a).

In this context, the spectral efficiency is another important metric to estimate the

system performance, which decouples the available bandwidth from the capacity. The

average DL cell spectral efficiency depending on the traffic load for LOS path loss is

indicated in Figure 6.10. It can be seen that the EFFR schemes are able to provide the

most efficient usage of the radio resources at their saturation points (200 kbit/s load for

EFFR 8:2, 300 kbit/s load for EFFR 7:3, 400 kbit/s load for EFFR 6:4, and 500 kbit/s

load for EFFR 5:5, respectively), which are higher than the DL cell spectral efficiency

by using the Reuse-3 scheme with a traffic load of 50 kbit/s (its saturation point, refer

to Figure 6.7a). In contrast, the SFR scheme enables an improvement of the cell

spectral efficiency over the Reuse-1 to a certain but minor extent at its saturation point

with a traffic load of 350 kbit/s. Here, it should be noted that the cell spectral

Figure 6.10: Average DL cell spectral efficiency of five frequency reuse schemes as a function of

offered traffic per user under LOS condition, having the same environment as in Figure 6.4. It can be seen that the simulation results generated by OpenWNS match the analytical results from the Matlab

simulator quite well, refer to Figure 5.21a.


efficiency by using the Reuse-3 scheme is three times larger than its system spectral

efficiency, since only 1/3 of the system bandwidth is required by each Reuse-3 cell.

For the Reuse-1, IFR and SFR schemes, however, the whole system bandwidth is

available for each cell. Hence, the cell spectral efficiency of these schemes shown in

the figure is actually also their system spectral efficiency.

6.2.1.2 Performance in LOS UL

In UL, except the SFR scheme, the carrier strength and the mean interference level of

different types of UTs using all the other schemes are similar to that in DL (refer to

Figure 6.4 and Figure 6.5), as displayed in Figure 6.11 and Figure 6.12.

Compared to Figure 6.4, the SFR scheme gains a better carrier strength performance

for both CCUs and CEUs in UL than in DL. Similar to using the EFFR scheme, the

mean SFR CCU carrier strength is very close to that using the Reuse-1 scheme, whilst

its mean CEU carrier strength is almost identical to that using the Reuse-3 scheme.

This is because in UL the both power levels for CCUs and CEUs using the SFR

scheme are higher than in DL, as explained in Section 6.1.5.2, and are the same as

using the EFFR scheme.

Figure 6.12 exhibits the mean UL interference level perceived by CCUs and CEUs of

the five investigated frequency reuse schemes. Higher transmission power can enhance

carrier strength, yet also leads to more ICI. In comparison with the results presented in


Figure 6.11: Mean UL carrier signal strength perceived at the central BS versus offered traffic per

user under LOS condition, having the same environment as in Figure 6.4: (a) mean CCU UL carrier signal strength perceived at the central BS; (b) mean CEU UL carrier signal strength perceived at the

central BS.


Figure 6.5, owing to higher transmission power available for each type of SFR UTs,

the ICI perceived by SFR UTs is also increased in UL. With the same power allocation

for each type of UTs in both RUP schemes, the mean CEU inference level can be

significantly reduced by using the EFFR scheme compared to the Reuse-1 scheme,

whereas using the SFR scheme the inference level is even inferior to that using the

Reuse-1 scheme by about 2 dB for both CCUs and CEUs.

Figure 6.13 presents the corresponding mean overall UL CINR level and the mean UL

CINR levels perceived by different types of user versus offered traffic per user,

respectively. For the SFR scheme, increased carrier strength and increased interference

level result in similar UL CINR values to DL CINR values for overall and both types

of UTs (refer to Figure 6.6). The other schemes have also similar results in UL and DL,

owing to similar carrier and inference strength in both transmission directions.

Figure 6.14 displays the mean overall UL MAC throughput as well as the

corresponding mean CCU and CEU UL throughput as a function of offered traffic per

user under LOS condition. It can be seen that all investigated schemes have similar

behavior in UL as in DL (refer to Figure 6.7). Yet their UL MAC throughput is

slightly lower than the DL MAC throughput. This is because as described in Section

6.1.4 for each superframe the UL phase is shorter than the DL phase for 9 OFDMA

symbols in total. Thus, the lower UL throughput arises from the relatively less

resource assignment. Like in DL, the EFFR scheme can also provide a considerable

improvement on the overall cell throughput in UL and outperforms all the other

(a) Mean CCU UL interference (b) Mean CEU UL interference

Figure 6.12: Mean UL interference level perceived at the central BS versus offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean CCU UL interference level perceived at the central BS; (b) mean CEU UL interference level perceived at the central BS.


schemes in almost every situation, regardless of with which M to N combination. The

saturation point of the EFFR system with M:N = 8:2 is reached at about 1.5 Mbit/s

with 150 kbit/s load per user; of the EFFR with M:N = 7:3 at about 1.9 Mbit/s with 250

kbit/s load per user; as well as of the EFFR with M:N = 6:4 at about 2 Mbit/s; and of

the EFFR with M:N = 5:5 at about 2.1 Mbit/s with 300 kbit/s load per user,

respectively. In contrast, the SFR system reaches its saturation point about 1.16 Mbit/s

with 300 kbit/s offered traffic per user. And the saturation point by applying the

(a) Mean overall UL CINR

(b) Mean CCU UL CINR (c) Mean CEU UL CINR

Figure 6.13: Mean UL CINR values perceived at the central BS as a function of offered traffic per

user under LOS condition, having the same environment as in Figure 6.4: (a) mean overall UL CINR

values perceived at the central BS; (b) mean CCU UL CINR perceived at the central BS; as well as (c) Mean CEU UL CINR perceived at the central BS.


Reuse-3 is at about 50 kbit/s load per user, which cumulates to a system throughput of

around 1.18 Mbit/s. Under the LOS propagation, the EFFR with whatever combination

can achieve a cell capacity above 1.5Mbit/s in UL, see Figure 6.15. And again, the

EFFR with M to N combination of 5:5 gains the best performance in system

throughput of all schemes. It reaps a noticeable cell capacity increase of app. 300%

compared to the Reuse-1 scheme, and an app. 74% gain over the Reuse-3 as well as

gets an advantage of app. 78% over the SFR scheme.

(a) Mean overall UL cell throughput

(b) Mean CCU UL throughput (c) Mean CEU UL throughput

Figure 6.14: Mean UL MAC throughput under LOS condition as a function of offered traffic per user

in the same environment as in Figure 6.4: (a) mean overall UL cell throughput; (b) the corresponding

mean CCU UL throughput; and (c) the corresponding mean CEU UL throughput.


As mentioned before, the benefits with applying the EFFR scheme are not only

reflected by throughput gains but also by the enhancement of the cell coverage

percentage. Figure 6.16 gives the mean UL coverage percentage of the five

investigated reuse schemes versus offered traffic per user under LOS condition. In

comparison with the DL performance presented in Figure 6.9, the overall UL cell

coverage percentage (see Figure 6.16a) by using the Reuse-3 and the EFFR schemes is

better than in DL by about 10% in full load and overload situations, which is actually

caused by the better UL coverage percentage of their CEUs (see Figure 6.16c). With

the SFR scheme, the overall UL cell coverage remains similar to the DL performance

when the offered traffic per user is smaller than 400 kbit/s. After this point in overload

situations, the UL performance becomes slightly better than its DL performance,

which results from its bettering UL CCUs coverage performance (see Figure 6.16b). In

contrast, the UL CCUs coverage percentage of the Reuse-1 and the IFR schemes

becomes worse by around 8%, so that their overall UL performance decreases slightly.

It is unchanged that in UL the EFFR scheme and the Reuse-3 scheme still attain the

best performance among all schemes in terms of not only the overall cell coverage but

also the CCUs and the CEUs coverage percentage. They improve the UL cell coverage

by more than 50% compared to the SFR scheme, as well as by around 60% compared

to the Reuse-1 and the IFR schemes in full- and overload situations.

Viewing the cell coverage performance with the UL mean cell capacity performance

together (refer to Figure 6.14 and Figure 6.15), it can be seen that the exclusive RUP

technique—EFFR scheme—has superiority over all the other schemes. Based on

maintaining the coverage percentage of both CCUs and CEUs close to the Reuse-3

scheme, it substantially improves the mean user throughput in UL LOS path loss

situation.

Figure 6.15: Cell capacity and the corresponding mean user throughput of all studied schemes in LOS UL, having the same environment as in Figure 6.4.


Lastly, the corresponding average UL cell spectral efficiency depending on the traffic

load for LOS path loss is shown in Figure 6.17. It can be seen that the Reuse-3 and the

EFFR 7:3 schemes are able to provide the most efficient usage of the radio resources

at their saturation points (50 kbit/s load for the Reuse-3 and 250 kbit/s load for the

EFFR 7:3, respectively) among all schemes. And the EFFR scheme with other M to N

combinations can achieve close but slight lower UL cell spectral efficiency at their

(much higher throughput) saturation points than the Reuse-3 scheme.

(a) Mean overall UL cell coverage percentage

(b) Mean UL CCUs coverage percentage (c) Mean UL CEUs coverage percentage

Figure 6.16: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per user under LOS condition, having the same environment as in Figure 6.4: (a) mean overall UL cell

coverage percentage; (b) the corresponding mean CCUs coverage percentage; and (c) the

corresponding mean CEUs coverage percentage in UL traffic.


All things considered, the EFFR scheme is the most attractive solution among all the

five investigated reuse schemes under LOS propagation. It not only reaps substantial

gains in the overall system performance, but also enhances the cell edge performance.

With the EFFR scheme, the medium is able to be more effectively utilized, the overall

cell capacity is significantly improved, and large cell coverage can be attained.

6.2.1.3 Performance in NLOS DL

In this subsection, a comprehensive performance evaluation under NLOS condition is

provided, which includes the estimation of the CINR value, the mean MAC

throughput, the mean cell coverage percentage as well as the average cell spectral

efficiency for both DL and UL traffic, respectively. Like in the previous subsection,

not only the overall system performance, but also detailed observation on the

performance of different types of UTs (the CCUs, the CEUs and weakest users) are

presented. Through an investigation of the five frequency-reuse techniques under LOS

and NLOS conditions, their performance behavior is compared, and different

characteristics of each scheme due to different propagation conditions are disclosed.

Unlike the LOS case, the cell radius of 220m is assumed, which is the maximum cell

radius for the Reuse-3 and the EFFR schemes under NLOS condition calculated in

Section 5.3. Furthermore, the proposed EFFR scheme with another four M to N

combinations (7:3 | 6:4 | 5:5 | 4:6) are evaluated. According to the analysis results

presented in Section 5.3, the maximum valid reach by using the Reuse-1 scheme is

0.6R under NLOS condition. Hence, the range ratio r/R of 0.6 to divide UTs into

Figure 6.17: Average UL cell spectral efficiency of five frequency reuse schemes as a function of

offered traffic per user under LOS condition, having the same environment as in Figure 6.4.


CCUs and CEUs for the SFR scheme and the EFFR scheme is chosen for NLOS

simulations. That is to say, among 25 UTs, 12 UTs are CCUs and the remaining 13

UTs are CEUs.

(a) Mean overall DL CINR (b) Mean DL CINR perceived by weakest users

(c) Mean DL CINR perceived by CCUs (d) Mean DL CINR perceived by CEUs

Figure 6.18: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by different types of UTs as a function of offered traffic per user under NLOS condition. 25 UTs are

uniformly distributed in each cell with a cell radius of 220 m. The range ratio r/R of 0.6 for

partitioning CCUs and CEUs is assumed. The weakest users are those UTs located between 176 m and

220 m away from the BS (i.e., R'/R [0.8, 1]). And for both SFR and EFFR schemes, the power ratio

of high power level to low power level is set as 3. (a) mean overall DL CINR values; (b) the

corresponding mean DL CINR values perceived by weakest users; (c) the corresponding mean DL

CINR values perceived by CCUs; as well as (d) the corresponding mean DL CINR values perceived by CEUs.


Figure 6.18a displays the mean overall CINR levels perceived by UTs for DL as a

function of offered traffic per user under NLOS condition. Generally, the mean CINR

in a NLOS scenario is much higher than in a LOS scenario (refer to Figure 6.6a) for all

schemes. Both the EFFR scheme and the SFR scheme can significantly increase the

mean CINR compared to the IFR scheme and the Reuse-1 scheme in both scenarios. In

the LOS case, the mean CINR attained by using the EFFR scheme is very close to and

sometimes even higher than in a Reuse-3 system. In the NLOS case, however, the

mean CINR of the EFFR scheme cannot catch up with the values of the Reuse-3

scheme any more. But, it is still much better than the other schemes. In contrast, in

both LOS and NLOS cases, the improving amplitude on the mean DL CINR with the

SFR is less than using the EFFR. This is because the SFR applies an inclusive reuse

mechanism. With the SFR design, the interferences experienced on both reuse-1

subchannels and reuse-3 subchannels are higher than by using the EFFR and the

Reuse-3.

Detailed observations on individual types of users are exhibited in Figure 6.18b, c and

d. The results in Figure 6.18b show that among all schemes, only the near cell border

UTs (weakest users), who apply the EFFR scheme or the Reuse-3 scheme, can achieve

valid DL CINR levels above 6.4 dB with offered traffic per user > 100 kbit/s.

Figure 6.19 presents the corresponding mean DL MAC throughput of all studied

schemes under NLOS propagation. Comparing the results under LOS condition (refer

to Figure 6.7a) and NLOS condition (Figure 6.19a), it can be seen that all studied

schemes can reap more cell throughput in the NLOS scenario than in the LOS scenario.

This is because all schemes can get higher mean CINR level under NLOS condition

than under LOS condition (refer to Figure 6.6a and Figure 6.18a). But, the RUP

techniques—the EFFR scheme and the SFR scheme—gain less improvement under

NLOS than under LOS, in comparison with the conventional methods. The reason is

that under LOS propagation the system is interference-limited whereas under NLOS

propagation the system is noise-limited, as explained in Chapter 5. The RUP

techniques investigated in this work are designed for dealing with excessive ICI in a

cellular network. Thus, they work more effectively in an interference-limited system

than in a noise-limited system. Despite that, in both path loss models, the EFFR

scheme can substantially enhance the mean overall cell throughput and offers the best

performance in every traffic load situation.

As shown in Figure 6.19a, although a high saturation point at about 4.2 Mbit/s with

400 kbit/s offered traffic per user can be reached by using the SFR scheme, its mean

throughput is just similar to the Reuse-1 when the traffic load is between 150 kbit/s

and 250 kbit/s, and cannot catch up with the Reuse-3 until the offered traffic increases

to 300 kbit/s. By contrast, the EFFR scheme performs much better than the SFR

scheme, and even exceeds the Reuse-3 scheme in full and overload situations. The

saturation point by applying the Reuse-3 is reached at about 3.6 Mbit/s with 150 kbit/s


load per user, whereas the saturation point of the EFFR system with M:N = 7:3 is

reached at about 3.6 Mbit/s with 150 kbit/s load per user; of the EFFR with M:N = 6:4

at about 3.97 Mbit/s with 200 kbit/s load per user; and of the EFFR with M:N = 5:5 at

about 4.36 Mbit/s with 250 kbit/s load per user; as well as of the EFFR with M:N = 4:6

at about 4.74 Mbit/s with 300 kbit/s load per user, respectively. With M to N

combination of 4:6 the EFFR scheme gains the best cell capacity performance among

all schemes (see Figure 6.20), and can reap a considerable increase of app. 133%

(a) Mean overall DL cell throughput (b) Mean weakest user DL throughput

(c) Mean CCU DL throughput (d) Mean CEU DL throughput

Figure 6.19: Mean DL MAC throughput under NLOS condition as a function of offered traffic per user in the same environment as in Figure 6.18: (a) mean overall DL cell throughput; (b) the

corresponding weakest user DL throughput; (c) the corresponding mean CCU DL throughput; as well

as (d) the corresponding mean CEU DL throughput.


compared to the Reuse-1 scheme, an app. 30% DL gain over the Reuse-3 as well as

gets an advantage of app. 12% over the SFR scheme.

By observing individual behavior of different types of UTs, the merits of the EFFR

can be further reflected. It can be seen that the EFFR significantly enhances the

throughput of the CEUs (see Figure 6.19d) and the weakest users (see Figure 6.19b)

without negatively affecting CCU performance (see Figure 6.19c). The SFR scheme,

however, cannot improve its CEUs performance much compared to the Reuse-1, even

though its CCU throughput is not bad. And its weakest users, which are close to the

cell borders, can even not be served by the BS at all. As for the IFR scheme, it can

only better the user performance when system is not highly loaded. But it performs

exactly like the Reuse-1 in overload situations.

In brief, the EFFR scheme has superiority over the other schemes and can gain

substantial improvements not only in overall cell capacity but also in cell edge

throughput.

Except the system throughput, the cell coverage percentage is another important

performance indicator used to estimate the effects of the ICI mitigation in a multi-

cellular system. Figure 6.21a exhibits the overall cell coverage percentage versus the

traffic load per user in each cell under NLOS condition in DL. The results show that

the EFFR scheme has a very close performance to the Reuse-3 scheme, and both are

much better than the other schemes. Unlike the LOS case (see Figure 6.9a), the

utilization of the EFFR scheme under NLOS condition enables the cell coverage of

approximately 100%, which means almost all UTs within the focused cell can be

served. In contrast, the SFR scheme can barely support 63% of the cell coverage, and

Figure 6.20: Cell capacity and the corresponding mean user throughput of all studied schemes in

NLOS DL, having the same environment as in Figure 6.18.


the Reuse-1 and the IFR scheme can even just cover 54% and 62% of a cell,

respectively, in full load situation.

Figure 6.21b, c and d show the corresponding mean coverage percentage of the most

remote UTs, the CCUs as well as the CEUs, respectively. Apparently, the EFFR

adopts the merit of the Reuse-3 scheme at ICI mitigation so that it considerably

enlarges the coverage percentage of the CEUs and the weakest users compared to the

(a) Mean overall DL cell coverage percentage (b) Mean DL weakest users coverage percentage

(c) Mean DL CCUs coverage percentage (d) Mean DL CEUs coverage percentage

Figure 6.21: Mean DL coverage percentage of five frequency reuse schemes versus offered traffic per

user under NLOS condition, having the same environment as in Figure 6.18: (a) mean overall DL cell

coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in

DL traffic.


Reuse-1 scheme. The SFR scheme can also significantly improve the mean CEUs

coverage, nevertheless, its cell border UTs can still not be covered just like using the

Reuse-1 and the IFR scheme, which confirms the analysis results presented in Section

5.3.

As a consequence, with the EFFR scheme the ICI can be successfully reduced, so that

both wide cell coverage and high data rate can be achieved in NLOS DL.

Figure 6.22 gives the average DL cell spectral efficiency depending on the traffic load

for the NLOS case, which is the third important metric for system performance

estimation. It can be seen that the EFFR schemes triples the cell spectral efficiency at

their saturation points (150 kbit/s load for EFFR 7:3, 200 kbit/s load for EFFR 6:4, 250

kbit/s load for EFFR 5:5, and 300 kbit/s load for EFFR 4:6, respectively, refer to

Figure 6.19a) compared to the Reuse-1 with a traffic load of 150 kbit/s (its saturation

point, refer to Figure 6.19a) and reaches the second best place, whereas the SFR just

slightly improves and the IFR even cannot better the cell spectral efficiency

performance. The Reuse-3 scheme can provide the best cell spectral efficiency

performance with a traffic load of 150 kbit/s (its saturation point, refer to Figure 6.19a).

However, as mentioned before, the system spectral efficiency of Reuse-3 can only

reach 1/3 of its cell spectral efficiency due to just 1/3 of the system bandwidth

available for each Reuse-3 cell, whereas for the SFR scheme, the cell spectral

efficiency presented in the figure is exactly its system spectral efficiency. Hence,

actually, the Reuse-3 is inferior to the SFR scheme in terms of system spectral

efficiency. The system spectral efficiency by using the EFFR scheme is also lower

than its cell spectral efficiency (53.3% of its cell spectral efficiency with EFFR 7:3,

Figure 6.22: Average DL cell spectral efficiency of five frequency reuse schemes as a function of

offered traffic per user under NLOS condition, having the same environment as in Figure 6.18.


60% with EFFR 6:4, 66.7% with EFFR 5:5, and 73.3% with EFFR 4:6, respectively)

due to limited available bandwidth for each cell, see Table 6.2. Even so, the system

spectral efficiency by using the EFFR 5:5 and EFFR 4:6 can still surpass that using all

the other schemes, including the SFR and the Reuse-3 schemes.

Taking all performance indicators into account, it can be noted that the EFFR schemes

(except M:N = 7:3) can offer the best system capacity, the best overall cell coverage

percentage and a moderate average cell spectral efficiency (but the best system

spectral effficency) among all investigated schemes, which is an attractive solution for

frequency reuse one deployment.

6.2.1.4 Performance in NLOS UL

Figure 6.23 presents the corresponding mean UL MAC throughput of all studied

schemes under NLOS propagation. It can be seen in Figure 6.23a that the SFR always

performs inferior to the Reuse-3 and the EFFR with any traffic load, although its

saturation point can be reached at about 3.06 Mbit/s with 200 kbit/s offered traffic per

user, which is much higher than the saturation points of the Reuse-1 and the IFR

schemes at about 1.42 Mbit/s with 100 kbit/s load per user and 2.22 Mbit/s with 150

kbit/s load per user, respectively. Detailed observation on the mean UL throughput of

diverse types of UTs can be found in Figure 6.23b, c and d. Like in DL (refer to

Figure 6.19), the EFFR significantly betters the CEUs’ including the weakest users

performance compared to the Reuse-1 and IFR scheme, and doubles the mean CEU

throughput and the weakest user throughput of the SFR scheme. In short, the EFFR

improves the user performance of all cell regions so as to make the whole system

throughput enhanced.

The benefits using the EFFR scheme are not only reflected in high data rate but also in

high cell user coverage percentage, as shown in Figure 6.24. From the results it can be

seen that like in DL (refer to Figure 6.21) among all schemes only the EFFR and the

Reuse-3 can attain 100% cell coverage percentage. The next is the SFR scheme with

around 81% cell coverage, as shown in Figure 6.24a. Furthermore, it can be noticed

that the weakest users coverage percentage (see Figure 6.24b) by using the SFR is

considerable increased from 0% to 46%, compared to Figure 6.21b. This leads to a

significantly bettered overall cell coverage percentage, nonetheless, at the expense of

sacrificing the overall cell throughput (refer to Figure 6.23a and Figure 6.19a).

Figure 6.25 gives the corresponding UL cell spectral efficiency of all five investigated

schemes under NLOS condition. Similar to the DL situation (refer to Figure 6.22), the

EFFR scheme achieves a moderate UL cell spectral efficiency, which is inferior to that

using the Reuse-3, though much better than with the other reuse schemes. And

different to the DL is that the more the resources invested for the EFFR CEUs the

higher cell spectral efficiency can be gained by the EFFR.


(a) Mean overall UL cell throughput (b) Mean weakest user UL throughput

(c) Mean CCU UL throughput (d) Mean CEU UL throughput

Figure 6.23: Mean UL MAC throughput under NLOS condition as a function of offered traffic per user in the same environment as in Figure 6.18: (a) mean overall UL cell throughput; (b) the

corresponding weakest user UL throughput; (c) the corresponding mean CCU UL throughput; as well

as (d) the corresponding mean CEU UL throughput.


(a) Mean overall UL cell coverage percentage (b) Mean UL weakest users coverage percentage

Figure 6.24: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per user under NLOS condition, having the same environment as in Figure 6.18: (a) mean overall UL cell

coverage percentage; (b) the corresponding weakest users coverage percentage.

Figure 6.25: Average UL cell spectral efficiency of five frequency reuse schemes as a function of

offered traffic per user under NLOS condition, having the same environment as in Figure 6.18.


6.2.1.5 Conclusion

The EFFR scheme outperforms all the other schemes, as it can attain the best MAC

throughput and the best coverage percentage in both UL and DL under any

propagation mode; as well as the best cell spectral efficiency under LOS condition and

a second best cell spectral efficiency (but the best system spectral efficiency) under

NLOS condition in both UL and DL. Moreover, the enhancements by using the EFFR

scheme favor not only the CCUs but also the CEUs and eventually the weakest users,

who are more susceptible to the ICI, in every traffic situation. Compared to the

analytical results, see Figure 5.21b the cell spectral efficiency of EFFR schemes is in

between 0,3 and 0,4, SFR is at 0,2 and Reuse-3 close to 0,6. Simulation results of

analysis and very detailed simulation match quite well.

Under LOS path loss, the proposed EFFR scheme tends to increase the overall CINR,

and the cell capacity and the cell coverage can be remarkable upgraded compared to

static Reuse schemes. But an around 35% coverage gap in DL and an about 25%

coverage gap in UL still exist. In NLOS case, by contrast, the profit in the overall cell

capacity by applying the EFFR scheme seems not such great as in LOS case. Yet the

EFFR scheme enables the cell coverage of approximately 100% for both DL and UL,

which means almost all UTs within a cell can be validly served.

6.2.2 Impact of Range Ratio on System Performance

In the realization of RUP designs, the range definition for dividing CCU-zone and

CEU-zone plays a very important role and could influence the system performance

severely. In Chapter 5, based on CINR calculation the maximal cell radius and

adequate range definitions for seperating different user-type zones for each RUP

scheme are analyzed. In this section, a thorough simulation study on the SFR scheme

and the EFFR scheme with diverse range ratios r/R defined as the CCU-zone radius to

the cell radius is provided. Like the performance evaluation for the Reference Scenario

in Section 6.2.1, the results of the RUP techniques (namely, the SFR scheme and the

EFFR scheme) are compared with two static Reuse schemes and the IFR scheme.

However, in the following assessments it will be noted that the overall performance by

applying the static Reuse schemes and the IFR scheme stays constant with varying

range ratios. This is because no specific treatments for different types of UTs in terms

of resource scheduling and power allocation are carried out in these techniques. In the

concrete, all UTs (no matter CCUs or CEUs) in these schemes have equal

opportunities to access the medium and are allowed to use the same power on each

subchannel to transmit data for both DL and UL.

All simulation parameters used in Reference Scenario remain unchanged in this

section except that the performance estimation does not depend on increasing offered

traffic, but on various range ratios r/R from 0.1 to 0.8 with a constant offered traffic

500 kbps per user. Table 6.4 lists the number of different types of UTs separately with

mutable range ratio definitions when 25 UTs are uniformly distributed in each cell. It

can be noted that the larger the range ratio r/R, the more UTs belong to the CCUs and

the larger the CCU area is. And the range ratio r/R of 0.1 is a special case with zero

CCUs and all UTs are CEUs.

Table 6.4: The number of different types of UTs with diverse range ratio definitions when 25 UTs

are uniformly distributed in each cell.

Range ratio [r/R] # CCU # CEU

0.1 0 25

0.2 1 24

0.3 3 22

0.4 5 20

0.5 8 17

0.6 12 13

0.7 16 9

0.8 21 4



Figure 6.26 shows the impact of range ratios on mean DL CINR values under LOS

condition. For both CCUs and CEUs, the CINR of all investigated schemes reduces

with an increasing range ratio r/R. And it is certain that the CINR performance with



Figure 6.26: Mean overall DL CINR values and the corresponding mean DL CINR values perceived

by different types of UTs depending on range ratio r/R under LOS condition, which is defined as the

zone radius for the CCUs r to the cell radius R. 25 UTs are uniformly distributed in each cell with a

cell radius of 1000 m. Offered traffic per user of 500 kbps is assumed. And for both SFR and EFFR

schemes, the power ratio of high power level to low power level is set as 3. (a) Mean overall DL CINR

values; (b) the corresponding mean DL CINR values perceived by CCUs; and (c) the corresponding mean DL CINR values perceived by CEUs.


the Reuse-3 is always better than using the Reuse-1 and the IFR scheme due to the

larger distance of the interfering cells and the higher power allocation for each link.

The EFFR scheme takes advantage of the location-specific predominance of the CCUs

to allow them to occupy resources with FRF of 1, and applies the interference-aware-

reuse mechanism on the Secondary Segment (refer to Section 4.1.2.2). Thus, the EFFR

CCU CINR is very close to the Reuse-1 when the range ratio is r/R ≤ 0.4, and keeps

constant no longer decreasing after that, as shown in Figure 6.26b. Concerning the

inherent vulnerability of the CEUs, the EFFR scheme reserves resources for them

using dedicated FRF of 3 and higher transmission power, so that it achieves a decent

mean CEU CINR, which is similar to the Reuse-3 performance, as shown in

Figure 6.26c. In contrast to the EFFR scheme, the SFR owing to its inclusive reuse

mechanism gains the worst CCU CINR performance among all schemes and relatively

inferior CEU CINR performance after r/R > 0.2. From Figure 6.26b, it can be seen that

the EFFR can hold the mean CCU CINR values above the required CINR threshold of

6.4 dB for every range ratio, whereas for the SFR, r/R values not greater than 0.4 are

permitted. In terms of the CEU CINR performance (see Figure 6.26c), the EFFR

scheme can hold the valid CINR level until the range ratio r/R exceeds to 0.6 while for

the SFR it may not exceed 0.3. And for the Reuse-1 and the IFR scheme, valid CEU

CINR values can never be supplied.

As a result, as shown in Figure 6.26a, the mean overall CINR performance of the

EFFR and the Reuse-3 scheme is relatively stable and decent above the required CINR

threshold of 6.4 dB in every situation. This is not the case for the Reuse-1 and the IFR

scheme, where the mean overall CINR is always lower than the required CINR

threshold. The SFR scheme cannot hold an acceptable mean overall CINR level after

r/R > 0.4.

High CINR results in high data rate. Hence, Figure 6.27a shows a clear advantage by

using the EFFR scheme compared to the Reuse-1 or the IFR scheme. However, CINR

is merely one factor to influence the system capacity, the other is available bandwidth.

Lack of available bandwidth (see Table 6.2) could also degrade the throughput

performance. That explains why the Reuse-3 cannot catch up the EFFR throughput

performance (except the throughput by using EFFR 8:2 with r/R > 0.6), although they

have alike overall CINR value. The overall cell throughput of the RUP techniques

EFFR and SFR varies with the range ratio, whereas static and the IFR scheme are

independent of the r/R selection. When comparing the overall cell throughput of EFFR

and SFR, the performance of the SFR is obviously more sensitive to the range ratio

and undulates more heavily. The SFR with a range ratio r/R = 0.4 reaches its best

performance of about 1.95 Mbps. When r/R > 0.4 the SFR performance deteriorates

rapidly, so that from r/R of 0.5 it is inferior to the Reuse-3 performance and with r/R ≥

0.6 it is even worse than the Reuse-1 performance. In contrast, except the M to N of


8:2, the EFFR scheme with the other M to N combinations (i.e. with further decreasing

number of reuse-3 subchannels and increasing number of reuse-1 subchannels) can

surpass all the other schemes in DL LOS case. And the more reuse-1 subchannels N

are assigned for the CCUs, the greater the overall cell capacity gain is.

Detailed observation on the mean DL throughput performance of different types of

UTs can be found in Figure 6.27b and c for the CCUs and the CEUs, respectively.

Regarding the CCU performance, it is obvious that the RUP techniques outperform the

none-RUP techniques in the majority of cases. When comparing the RUP techniques,



Figure 6.27: Mean DL MAC throughput under LOS condition depending on range ratio r/R in the

same environment as in Figure 6.26: (a) mean overall DL cell throughput; (b) the corresponding mean CCU DL throughput; and (c) the corresponding mean CEU DL throughput.


it can be observed that the mean CCU performance of the SFR decays from r/R = 0.3

on more rapidly than using the EFFR scheme. Furthermore, with r/R > 0.5, the SFR

CCU performs the worst among all the schemes, whereas the EFFR scheme can still

outperform the Reuse-1 and the IFR, regardless of the M to N combination. Except M

to N of 8:2, the EFFR scheme with the other M to N combinations exceeds even all the

other schemes in every situation. According to the CEU performance displayed in

Figure 6.27c, the EFFR scheme and the Reuse-3 scheme are superior to the other

schemes. With the usage of the SFR, the CEU throughput performance can be

improved compared to the Reuse-1 with r/R < 0.6, yet it is still much inferior to the

EFFR and the Reuse-3. From r/R = 0.6 on, where the number of CEUs is 13 (refer to

Table 6.4), the mean CEU throughput is 0, which means 52% UTs cannot be served at

all. That also implies that the maximal cell coverage by applying the SFR with proper

range ratio selection is 48%. Another interesting phenomenon is that the EFFR

throughput performance for the CEUs even exceeds that of Reuse-3 when the range

ratio is r/R < 0.4, and the more the number of reuse-1 subchannels N is used in the

EFFR, the higher the mean CEU throughput can be achieved. This is because the

EFFR scheme owns a salient feature different to the SFR scheme that all unsatisfied

UTs, whether they are CCUs or CEUs, have the same chance to get reuse-1 resources,

if they can find usable resource in accordance with CINR estimation. In this way, even

when the range ratio is r/R = 0.1, which means all UTs are CEUs and no CCUs, the

reuse-1 subchannels are not wasted and can be used by suitable UTs who can receive

valid CINR values.

In conclusion, the EFFR (except with M:N = 8:2) is superior to all the other

investigated schemes under LOS condition. It is based on ensuring the CEU

performance similar to the Reuse-3 scheme, promotes CCU performance by peeling

off part resources from the reuse-3 resources to lunch in the reuse-1 utilization. As a

consequence, not only the overall cell capacity is noteworthy increased, but also the

CEU throughput performance can be substantially ameliorated.

The mean overall cell coverage percentage and the corresponding mean coverage

percentage of different types of UTs depending on range ratio r/R in LOS DL are

illustrated in Figure 6.28. It has been elucidated before that the SFR is an inclusive

RUP design whereas the EFFR uses the exclusive RUP technique. Hence, the mean

CCUs coverage percentage by using the SFR is inferior to the Reuse-1 when the range

ratio is r/R > 0.3 (see Figure 6.28b), and its CEUs perform much worse than with the

Reuse-3 scheme when the range ratio is r/R > 0.1 (see Figure 6.28b). In contrast, the

CCUs of the EFFR scheme can gain a very similar and even slightly better coverage

performance than with the Reuse-1 scheme, as well as its CEUs can have a slightly

worse but also very close performance to the Reuse-3 scheme. That results in an

overall better cell coverage performance by using the EFFR scheme than by using the

SFR scheme with any range ratio definition, as shown in Figure 6.28a. The curves in


the figures disclose that the EFFR can keep a similar overall cell coverage percentage

as the Reuse-3 scheme with an adequate range ratio definition (here for example when

the range ratio is ≤ 0.4). With further enlarging the CCU-zone, the CCUs coverage

performance deteriorates fastly like in a Reuse-1 system, which causes the gradually

attenuated overall cell coverage performance as shown in Figure 6.28a. Yet it is still

much better than the SFR scheme, the IFR scheme and the Reuse-1 scheme.



Figure 6.28: Mean DL coverage percentage of five frequency reuse schemes versus range ratio r/R under LOS condition, having the same environment as in Figure 6.26: (a) mean overall DL cell

coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs coverage percentage in DL traffic.


Together with the throughput performance displayed in Figure 6.27a, it can be

observed that the EFFR scheme (except with M:N = 8:2) with a range ratio r/R ≤ 0.4

can gain the best overall cell capacity and the second best cell coverage (but very close

to the best, namely the Reuse-3 scheme) among all investigated schemes under LOS

condition. The SFR scheme with a range ratio r/R from 0.2 to 0.4 can also improve the

cell throughput significantly compared to the Reuse-1 and Reuse-3 schemes. But

according to the SFR design, the CEUs are just allowed to utilize the Primary Segment,

which is only one third of the whole available bandwidth. Even if there may be still

free resources remaining in the other two third of the bandwidth resources for the

CCUs, the CEUs cannot not use it. So, with a range ratio r/R of 0.2 or 0.3 means only

1 or 3 CCUs use the 67% resources, whereas the other 24 or 22 CEUs use the

remaining 33% (refer to Table 6.4), which exposes an extremely unfairness of

resource distribution between the CCUs and the CEUs. With a range ratio r/R of 0.4,

which means the number of CCUs is 5 and the remainder 20 UTs are CEUs, the

unfairness of resource assignment is relatively reduced, however, its overall cell

coverage percentage is also decreased close to the Reuse-1 scheme. And after this

point, neither throughput performance nor cell coverage performance of the SFR can

be hold. Both performances downgrade so quickly that they are even worse than using

the static Reuse schemes. Therefore, with respect to the range ratio definition for

dividing CCU-zone and CEU-zone, the EFFR scheme can provide more flexibility,

more robustness and more fairness among UTs than the SFR scheme.

Figure 6.29 exhibits the average DL cell spectral efficiency of the five investigated

reuse schemes depending on range ratio r/R under LOS condition. The results show

Figure 6.29: Average DL cell spectral efficiency of five frequency reuse schemes depending on range

ratio r/R under LOS condition, having the same environment as in Figure 6.26.


that the EFFR scheme with M:N = 6:4 and M:N = 5:5 can gain a better performance

than all the other schemes when a range ratio r/R ≤ 0.7 is chosen.

Taking all evaluated performance into consideration (refer to Figure 6.27a,

Figure 6.28a and Figure 6.29), it can be concluded that except with M:N = 8:2 the

EFFR scheme outperforms all the other schemes with a range ratio r/R ≤ 0.4, where

the best overall cell capacity, the second best cell coverage (but very close to the best)

and the best cell spectral efficiency can be gained.

6.2.2.2 Conclusion

In this section, a comprehensive performance evaluation in terms of CINR, MAC

throughput, coverage percentage as well as cell spectral efficiency of all investigated

frequency reuse schemes depending on range ratio r/R in LOS DL is presented and

explained. The other corresponding simulation results for LOS UL and under NLOS

condition can be found in Appendix C. The results in LOS DL show that except with

the M to N combination of 8:2, the other EFFR schemes with a range ratio r/R ≤ 0.4

reap remarkable benefits in overall system capacity and cell spectral efficiency while

maintaining a close overall cell coverage percentage to the Reuse-3. Furthermore, with

respect to range ratio definition for dividing CCU-zone and CEU-zone in RUP designs,

the proposed EFFR scheme can provide more flexibility and robustness than the SFR

scheme. Among all M to N combinations, the EFFR 5:5 scheme with a range ratio r/R

≤ 0.4 is an optimal solution for ICI mitigation in an OFDMA-based cellular system,

which surpasses all the other schemes providing the best overall cell throughput, the

second best (but very close to the best) cell coverage percentage as well as the best cell

spectral efficiency.

6.2.3 Impact of Power Ratio on System Performance

Historically power control has been employed in cellular systems to minimize near-far

dynamic range effects by constraining the transmission power to be received with a

constant power level. Use of fractional power control in interference-limited

environments has been shown to provide improvements in aggregate throughput [45].

And [26] shows that the power masks have a significant impact on the SFR system’s

performance. In order to disclose and compare the impact of various power allocations

on the system performance using the SFR and the EFFR schemes, a deep simulation

study is provided in this section. Like in the previous sections, the results by applying

RUP techniques are compared to that using the static Reuse schemes and the IFR

scheme. Since the available power is distributed evenly over all resource blocks in

each static Reuse system cell and in each IFR cell (namely, for both UL and DL, a

transmission power of 67 mW is used on each subchannel in a Reuse-1 system and in

an IFR system, as well as 200 mW is used on each subchannel in a Reuse-3 system),

their performances play just a referenced role in the following evaluations and stay

constant with altering power ratio Phigh/Plow.

All simulation parameters used in the Reference Scenario (refer to Section 6.2.1)

remain unchanged in this section except that the performance estimation does not

depend on increasing offered traffic, but on various power ratios Phigh/Plow from 1 to 10

with a constant offered traffic 500 kbps per user. According to the assumptions of

constant system power and fixed maximal transmission power of BSs and UTs as well

as Eqs. (3.4), (3.6), (4.8) and (4.10), Table 6.5 and Table 6.6 give an overview of

power allocations for CCUs and CEUs using the SFR and the EFFR, respectively, on

each subchannel for both UL and DL. As explained in Section 6.1.5.2, the both power

Table 6.5: Transmission power applied for CCUs on each subchannel in the RUP schemes for both DL and UL with varying power ratios.

Phigh/Plow

=

PCEU /PCCU

PCCU in DL

[mW]

PCCU in UL

(for all EFFR

& SFR)

[mW] EFFR 8:2 EFFR 7:3 EFFR 6:4 EFFR 5:5 EFFR 4:6 SFR

1:1 143 125 111 100 91 67 200

2:1 91 87 83 80 77 50 100

3:1 67 67 67 67 67 40 67

4:1 53 54 56 57 59 33 50

5:1 43 45 48 50 53 29 40

6:1 37 39 42 44 48 25 33

7:1 32 34 37 40 43 22 29

8:1 29 31 33 36 40 20 25

9:1 26 28 30 33 37 18 22

10:1 23 25 28 31 34 17 20


levels for CCUs and CEUs in the EFFR scheme are always higher than in the SFR

scheme for DL traffic, whereas in UL they remain identical. Furthermore, it can be

noted that in DL the power levels for CCUs and CEUs using the EFFR scheme are

also different with different M to N combinations. This is because different M to N

combination results in different total available bandwidth—number of subchannels—

for each EFFR cell (refer to Table 6.2). And with the assumption of fixed BS

transmission power, the more reuse-1 subchannels N is reserved for the CCUs, the

relatively more power can be used for each reuse-3 subchannel and each reuse-1

subchannel when the power ratio is Phigh/Plow > 3.


Figure 6.30 shows the mean DL carrier signal strength perceived by different types of

UTs versus power ratio Phigh/Plow under LOS condition. In both SFR and EFFR

systems, since the transmission power applied by the BS on each subchannel decreases

for the CCU-traffic and increases for the CEU-traffic with increasing power ratio, the

DL carrier strength perceived by CCUs degrades, whilst it perceived by CEUs

increases. Besides, due to higher power portion applied for both CCUs and CEUs in

the EFFR scheme than using the SFR, the mean DL carrier strength perceived by UTs

in an EFFR system is always stronger than in a SFR system.

Although the trends of the DL carrier strength with increasing power ratio Phigh/Plow

are similar for both the SFR scheme and the EFFR scheme, the trends of their DL

interference level perceived by CCUs and CEUs run completely in opposite directions,

as shown in Figure 6.31. Increasing power ratio Phigh/Plow means downgrading

Table 6.6: Transmission power applied for CEUs on each subchannel in the RUP schemes for both DL

and UL with varying power ratios.

Phigh/Plow

=

PCEU/PCCU

PCEU in DL

[mW]

PCEU in UL

(for all EFFR

& SFR)

[mW] EFFR 8:2 EFFR 7:3 EFFR 6:4 EFFR 5:5 EFFR 4:6 SFR

1:1 143 125 111 100 91 67 200

2:1 182 174 167 160 154 100 200

3:1 200 200 200 200 200 120 200

4:1 211 216 222 229 235 133 200

5:1 217 227 238 250 263 143 200

6:1 222 235 250 267 286 150 200

7:1 226 241 259 280 304 156 200

8:1 229 246 267 291 320 160 200

9:1 231 250 273 300 333 163 200

10:1 233 253 278 308 345 167 200



Figure 6.30: Mean DL carrier signal strength perceived by different types of UTs versus power ratio

under LOS condition, which is defined as high power level to low power level used for the SFR and the EFFR schemes. 25 UTs are uniformly distributed in each cell with a cell radius of 1000 m. Offered

traffic per user of 500 kbps and the range ratio r/R of 0.4 for partitioning CCUs and CEUs are assumed.

(a) Mean DL carrier signal strength perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs.


Figure 6.31: Mean DL interference level perceived by different types of UTs versus power ratio of

high power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean DL interference level perceived by CCUs;

(b) mean DL interference level perceived by CEUs.


transmission power for the CCU-traffic and upgrading transmission power for the

CEU-traffic, which results in the reduced ICI from the neighboring cells for the CCUs

and increased ICI for the CEUs in an EFFR system. However, this is not the case in a

SFR system. Due to the inclusive reuse design in a SFR network, CEU-traffic in some

neighboring cells shares the same resources with the CCU-traffic in the focused cell,

which means part of the ICI perceived by the CCUs is generated by the CEU-traffic in

these neighboring cells. With increasing power ratio Phigh/Plow, increased transmission

power is applied for these CEU-communications, which results in the increasing DL

interference level perceived by the CCUs (see Figure 6.31a). In opposite to the CCU-

traffic, the CEU-traffic in the focused cell only shares resources with CCU-traffic in

the neighboring cells, which are applied with reduced transmission power along with

increasing power ratio Phigh/Plow. That is why the curve of the SFR scheme in

Figure 6.31b trends downward.

The mean DL carrier strength in Figure 6.30 and the mean DL interference level in

Figure 6.31 result in the mean DL CINR values perceived by CCUs and CEUs

separately, as displayed in Figure 6.32b and c. It can be seen that using the EFFR

scheme the mean DL CINR perceived by the CCUs is similar to the Reuse-1 CCU

performance, and with the M to N combination of 5:5 the EFFR scheme performs even

better than the Reuse-1. Moreover, the EFFR CCU CINR performance stays relatively

stable with increasing power ratio Phigh/Plow and always above the required CINR

threshold of 6.4 dB, whereas using the SFR scheme its DL CCU CINR performance

degrades with increasing power ratio and cannot maintain a valid mean CINR value

when the power ratio Phigh/Plow is greater than 4. For the CEU-traffic, the SFR scheme

performs even worse since its mean DL CINR never reaches the required CINR

threshold, although higher transmission power is used. In contrast, the DL CEU CINR

of the EFFR scheme keeps very close to that using the Reuse-3 scheme and always

valid. The CCU performance and the CEU performance lead to the mean overall DL

CINR performance of all investigated schemes presented in Figure 6.32a. The EFFR

and the Reuse-3 own similar overall DL CINR performance, which is much better than

with the other schemes. Applying the SFR, the overall DL CINR can also be

significantly enhanced compared to the Reuse-1 and the IFR scheme. However, when

the power ratio Phigh/Plow is larger than 3, its overall DL CINR cannot hold valid any

more. That also explains why the overall DL cell throughput using the SFR is better

than using the Reuse-1 and the IFR but worse than using the Reuse-3 scheme when the

power ratio Phigh/Plow is larger than 3 (see Figure 6.33a). When the power ratio

Phigh/Plow is smaller than 4, the SFR takes advantage of its plentiful available resources

and thereby gains decent mean DL cell capacity, which can compare favorably with

the EFFR scheme.

Figure 6.33b and c give insight into the corresponding mean DL throughput of

different types of UTs depending on power ratio Phigh/Plow. It can be noticed that when


Phigh/Plow is equal to 1, no CEUs can be served at all with the SFR scheme. This

implies that the improvement in the overall cell throughput at this point is purely in

consequence of the enhancement of its CCU throughput performance. And with

increasing power ratio Phigh/Plow, a tradeoff between a reduced overall cell capacity and

enhanced CEU performance in a SFR system occurs clearly. As a consequence, only

with a power ratio Phigh/Plow of 2 or 3, the SFR can outperform the static Reuse



Figure 6.32: Mean overall DL CINR values and the corresponding mean DL CINR values perceived

by different types of UTs depending on power ratio of high power level to low power level used for

the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean overall DL CINR values; (b) the corresponding mean DL CINR values perceived by CCUs;

and (c) the corresponding mean DL CINR values perceived by CEUs.


schemes and the IFR scheme in terms of the overall cell capacity, and at the same time

improves the CEU throughput performance.

Overall, the SFR scheme cannot rival the EFFR scheme. Both CINR and throughput

performance using the SFR are severely impacted by power ratio. On the contrary,

whether in CINR or in throughput, and whether for CCUs or for CEUs, the RUP EFFR



Figure 6.33: Mean DL MAC throughput under LOS condition depending on power ratio of high

power level to low power level used for the SFR and the EFFR schemes in the same environment as in Figure 6.30: (a) mean overall DL cell throughput; (b) the corresponding mean CCU DL throughput;

and (c) the corresponding mean CEU DL throughput.


scheme can always hold a relatively constant and better performance than the RUP

SFR scheme with varying power ratios.

Besides the throughput performance, the coverage percentage is another important

issue. Figure 6.34 displays the mean overall cell coverage percentage and the

corresponding mean coverage percentage of different types of UTs depending on



Figure 6.34: Mean DL coverage percentage of five frequency reuse schemes versus power ratio of

high power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean overall DL cell coverage percentage; (b) the

corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs

coverage percentage in DL traffic.


power ratio Phigh/Plow in LOS DL. It is obvious that the EFFR scheme performs much

better than the SFR in this respect. In comparison with the Reuse-1 and the IFR

schemes, the SFR can just slightly ameliorate the DL cell coverage when the power

ratio Phigh/Plow is equal to 2 or 3, which are the only two cases where the SFR

outperforms the static Reuse schemes and the IFR scheme in terms of the overall cell

capacity (refer to Figure 6.33a). With further increasing power ratio Phigh/Plow the

overall DL coverage percentage can be improved by using the SFR scheme, which

mostly arises from the promotion in the CEUs coverage percentage but at the sacrifice

of the CCUs coverage. The EFFR scheme by contrast can have an overall DL

coverage percentage and CEUs coverage percentage very close to that using the

Reuse-3 scheme, which is the best among all schemes. However, the curves of the

EFFR in Figure 6.34b reveal that with a power ratio Phigh/Plow > 4 for the EFFR with

an M to N combination of 8:2 as well as with a power ratio Phigh/Plow > 5 for the EFFR

with the other M to N combinations their CCUs coverage percentage starts performing

inferior to that using the Reuse-1 scheme. Therefore, a power ratio Phigh/Plow < 5

should be the optimal choice for the EFFR scheme under LOS condition.

Lastly, the average DL cell spectral efficiency depending on power ratio Phigh/Plow for

LOS path loss is indicated in Figure 6.35. It can be seen that in comparison with the

Reuse-1 and the IFR schemes the SFR can just slightly ameliorate the DL cell spectral

efficiency when the power ratio is Phigh/Plow ≥ 4, whereas the EFFR not only

substantially promotes the cell spectral efficiency which is similar to the Reuse-3

performance, but also significant increases the overall cell capacity which is noticeable

better than using the Reuse-3 (refer to Figure 6.33a).

Figure 6.35: Average DL cell spectral efficiency of five frequency reuse schemes depending on power ratio of high power level to low power level used for the SFR and the EFFR schemes under LOS

condition, having the same environment as in Figure 6.30.

6.3 Summary and Conclusion 145

Among all M to N combinations in the EFFR scheme, the M to N of 6:4 and 5:5

perform the best (see Figure 6.33a, Figure 6.34a and Figure 6.35). With a power ratio

Phigh/Plow ≤ 5, they reap the best performance in cell capacity and cell spectral

efficiency while maintaining a close overall cell coverage percentage to the Reuse-3

scheme.

6.2.3.2 Conclusion

By means of comprehensive and detailed performance evaluation presented in this

section, it can be concluded that the RUP EFFR design is more attractive than the RUP

SFR design under LOS condition (simulation results for LOS UL, NLOS DL and

NLOS UL are presented in Appendix D, where the preference of EFFR over the other

techniques is confirmed). The power ratio has a severe impact on the performance of

the SFR scheme, whereas using the EFFR scheme its performance including CINR,

throughput, cell coverage as well as cell spectral efficiency is superior and relative

stable under various power ratios. With an inappropriate power allocation, the

performance of the SFR will be strongly deteriorated. Furthermore, although the SFR

scheme outperforms the Reuse-1 and IFR schemes, it can never surpass the Reuse-3

scheme. With a careful selection of power ratios, the SFR may gain benefits in overall

cell throughput. But its CEU throughput performance, its cell coverage performance as

well as its cell spectral efficiency are still far more inferior to that using the Reuse-3

scheme. In this sense, the contribution of the SFR scheme for ICI mitigation and

bettering the CEUs performance is limited. In contrast, the EFFR scheme can provide

considerably better overall cell throughput than with the Reuse-3 scheme, whilst it can

hold its CEUs performance and cell coverage performance similar to the Reuse-3

scheme. In terms of the spectral efficiency, the EFFR scheme performs much better

than the Reuse-1 and the SFR scheme, and can supply a similar performance to (or

even better performance with a power ratio Phigh/Plow ≤ 5 using M:N = 6:4 or M:N =

5:5) than the Reuse-3. As a consequence, with the respect to the power allocation, the

proposed RUP EFFR design always performs better than the RUP SFR in every

situation, and can provide more flexibility and robustness than the SFR scheme. With

the EFFR scheme the medium is able to be more effectively utilized, and the

performance of all UTs including both CCUs and CEUs are advanced.

6.3 Summary and Conclusion

The simulation results presented confirm that exclusive RUP is more effective than

inclusive RUP in terms of ICI mitigation. Even with less available bandwidth, the

proposed EFFR scheme can provide considerable improvements, and beat the SFR

scheme, the IFR scheme as well as the static Reuse schemes. With the utilization of


the EFFR scheme, not only the overall cell performance but also the performance of

the CEUs can be substantially enhanced. In all three scenarios, the EFFR scheme can

always achieve the best overall cell throughput and the best cell coverage (similar to

the Reuse-3) among all investigated schemes under both LOS and NLOS conditions.

Under LOS propagation, the EFFR scheme can also gain the best cell spectral

efficiency, which is very similar to and in some cases even better than using the

Reuse-3 scheme, whereas under NLOS condition, it is inferior to the Reuse-3 and can

only attain the secondary best place in cell spectral efficiency, yet it still can gain the

best system spectral efficiency and surpass all the other schemes, including the Reuse-

3 and the SFR schemes.

In comparison to the EFFR scheme, the SFR cannot gain an absolute advantage over

the Reuse-3 scheme in overall cell throughput under any of the propagations. With the

SFR, the CEU performance, the cell coverage performance as well as the spectral

efficiency performance can also be improved to some extent compared to the Reuse-1

scheme. Nevertheless, they are always much inferior to those using the Reuse-3 and

the EFFR schemes. Moreover, the performance of the SFR scheme is strongly

influenced by range ratio definition and power allocation for different types of UTs.

With an inappropriate selection of range ratio or power allocation, its performance

deteriorates severely and could even become worse than with the Reuse-1 scheme.

Therefore, actually the well-known SFR scheme cannot outperform the static Reuse

schemes. As for the IFR, it never performs better than the Reuse-1 in full-and over-

load situations, which means no substantial improvement occurs.

In conclusion, with the EFFR scheme, the medium can be more effectively utilized,

more flexibility as well as more robustness can be achieved, the overall cell capacity is

substantially improved, and the cell coverage is enlarged.

CHAPTER 7

7 Conclusion and Outlook

Conclusion and Outlook

7.1 Summary ..................................................................................................... 147

7.2 Conclusion .................................................................................................. 148

7.3 Outlook ....................................................................................................... 150

The expected convergence of fixed and mobile Internet services, the emergence of

new applications and the growth of wireless subscribers lead to an ever increasing

demand for bandwidth in wireless access. With aggressive spectrum reuse (frequency

Reuse-1) deployments, more available bandwidth for each cell in the networks and

simplified radio network planning can be expected. Nevertheless, unfavorable cell-

edge performance and inferior utilization of the precious frequency resource are

caused due to severe Inter-Cell Interference (ICI). ICI coordination approaches,

designed without (or with slow) inter-cell communication, are promising methods to

solve the problem, whereby significant performance improvements can potentially be

attained without inducing excessive signaling overhead and heavy computational

complexity. In this monograph, a novel resource allocation and reuse technique,

named Enhance Fractional Frequency Reuse (EFFR) scheme, is proposed. Aiming at

mitigating excessive ICI among neighboring cells, the EFFR is combined with a

transmission power allocation and an interference-aware reuse mechanism to achieve

not only wide area coverage but also a great enhancement of overall system capacity

in Orthogonal Frequency Division Multiple Access (OFDMA) based cellular networks.

7.1 Summary

In this work, an extensive introduction of the state-of-the-art in ICI mitigation

techniques, the categories of these techniques, as well as advantages and limitations of

approaches in each category are presented. In particular, three most representative and

popular ICI mitigation approaches, namely the static Fractional Frequency Reuse

(FFR) scheme, the Soft Frequency Reuse (SFR) scheme and the Incremental

Frequency Reuse (IFR) scheme, are studied. Meanwhile, the notion of inclusive and

exclusive reuse partitioning (RUP) is also elaborated.

7. Conclusion and Outlook 148

Based on a thorough analysis of their advantages and limitations respectively, the

novel ICI mitigation design EFFR scheme and its two variations, EFFR-Advanced

scheme and EFFR-Beyond, are contributed for a better fulfillment of the goals, namely,

to enhance the mean system capacity while restraining the ICI at cell edge. The EFFR

scheme divides the whole available bandwidth into a Primary Segment and a

Secondary Segment. The exclusive reuse-3 subchannels in the Primary Segment will

be preferentially used by cell-edge users with higher transmission power, whereas the

remaining subchannels are all reuse-1 subchannels allowing to be used with lower

power. In addition, the resources in the Secondary Segment will be occupied by means

of Carrier-to-Interference-plus-Noise Ratio (CINR) estimation.

Analytical models are used to evaluate the CINR distributions and the upper bounds of

an OFDMA-based cellular system in terms of cell coverage, mean cell capacity, as

well as area spectral efficiency of the aforementioned RUP techniques (including SFR,

IFR, EFFR, EFFR-Advanced, and EFFR-Beyond) under various propagation

conditions (Line-of-Sight (LOS), Non Line-of-Sight (NLOS), combined LOS-NLOS).

With the usage of the CINR calculation, the maximal cell radius and reasonable

boundary definitions for dividing different user-type zones for each RUP scheme are

determined. The analytical evaluation supports and validates the setup of the

Reference Scenario, which is chosen for simulative performance evaluation.

The comprehensive performance evaluation by means of event-driven simulation is

conducted to compare the systems using the proposed EFFR scheme with those using

the SFR scheme, the IFR scheme, and two static Reuse schemes in terms of CINR

level distribution, mean throughput, coverage percentage, and spectral efficiency

under LOS and NLOS propagation conditions, separately. In order to reach a reliable

evaluation, schemes are simulated with individual power mask, using scenarios with

surrounding interfering cells up to the 2nd

-tier. Moreover, this thesis presents the

system performance of each investigated scheme not only from the overall-cell

perspective but also from the perspectives of different types of users (for instance,

Cell-Centre Users (CCUs), Cell-Edge Users (CEUs), and in some cases the most

remote users). Besides, the major parameters that influence the system performance,

such as traffic load, range ratio defined for different types of users, power ratio of

higher power level to lower power level, as well as various RUP for the EFFR scheme

are evaluated.

7.2 Conclusion

ICI coordination techniques may potentially attain significant performance

improvements and have recently gained much attention to mitigate ICI in next

generation wireless communication networks. All ICI coordination approaches

investigated in this thesis are aiming at a high system spectrum efficiency with

7.2 Conclusion 149

frequency reuse factor (FRF) of 1 and efficient reduction of ICI (especially near the

cell edge), simultaneously. Based on a thorough analysis of the static FFR, the SFR

and the IFR schemes, the novel EFFR approach proposes a resource allocation and

reuse mechanism which can provide a considerable improvement with the help of the

Channel Quality Information (CQI) estimation. Concerning the inherent vulnerability

of CEUs, the EFFR scheme reserves resources for them by two specific solutions to

provide better coverage: 1) assign dedicated FRF-3 subchannel; 2) provide higher

transmission power. Taking advantage of the location-specific predominance of CCUs,

the EFFR scheme allows them to occupy resources with FRF-1 and interference

awareness.

The analytical evaluation performed turns out that under LOS condition (or in an

interference-limited system), the enhancement on the coverage with the help of the

RUP techniques focused is more remarkable than under NLOS and LOS-NLOS

propagations (or in noise-limited systems). Besides, the EFFR designs are always able

to provide better cell coverage than with the SFR or the Reuse-1 scheme. Among all

EFFR variations, only the EFFR-Advanced and EFFR-Beyond schemes can reach

100% cell coverage in any case, whether under LOS or under NLOS path loss. With

respect to cell capacity and spectral efficiency, under NLOS condition, all schemes

can reap a higher cell capacity and area spectral efficiency, nevertheless within a

limited cell area. Moreover, although the SFR scheme can attain a better cell capacity

under all propagation conditions compared to the Reuse-3 scheme, its area spectral

efficiency is still much inferior to the Reuse-3. That means the SFR compared to the

Reuse-3 has no overwhelming superiority. In contrast, the results show that the EFFR

series can outperform the SFR and the static Reuse schemes under any propagation

mode, where significant coverage gains and cell capacity improvements can be

achieved by applying the novel EFFR schemes with adequate resource allocations.

The simulation results validate that exclusive RUP is more effective than inclusive

RUP in terms of ICI mitigation. Even with less available bandwidth, the proposed

EFFR scheme can provide considerable improvements, and beat the SFR scheme, the

IFR scheme as well as the static Reuse schemes. With the utilization of the EFFR

scheme, not only the overall cell performance but also the performance of the CEUs

can be substantially enhanced. In all three presented scenarios, the EFFR scheme can

always achieve the best overall cell throughput and the best cell coverage (similar to

the Reuse-3 scheme) among all investigated schemes under both LOS and NLOS

conditions. Under LOS propagation, the EFFR scheme can gain the best spectral

efficiency as well, which is very similar to using the Reuse-3 scheme, whereas under

NLOS condition, it is inferior to with the Reuse-3 and can only attain the secondary

best place in cell spectral efficiency, but still much better than with the SFR and the

Reuse-1 schemes.

7. Conclusion and Outlook 150

In comparison to the EFFR scheme, the SFR cannot gain an absolute advantage over

the Reuse-3 scheme in overall cell throughput under any of the propagations. With the

usage of the SFR, the CEU performance, the cell coverage performance as well as the

spectral efficiency performance can also be improved to some extent compared to the

Reuse-1 scheme. Nevertheless, they are always much inferior to those using the

Reuse-3 and the EFFR schemes. Moreover, the performance of the SFR scheme is

strongly influenced by range ratio definition and power allocation for different types

of users. With an inappropriate selection of range ratio or power allocation, its

performance deteriorates severely and could even get worse than with the Reuse-1

scheme. Therefore, actually the well-known SFR scheme cannot outperform the static

Reuse schemes. As for the IFR, it never performs better than the Reuse-1 in full-and

over-load situations, which means no substantial improvement occurs.

In conclusion, with the usage of the EFFR scheme, the medium can be more

effectively utilized, more flexibility as well as more robustness can be achieved, the

overall cell capacity is substantially improved, and the cell coverage is enlarged.

7.3 Outlook

This work is considered to be the basis for further research concerning radio resource

management (RRM) for ICI mitigation in OFDMA cellular systems. With regard to

that, the performance evaluation could be continued to investigate

- the impact of different CINR thresholds for the guest-reuse of the secondary

resources on the EFFR system performance,

- the system performance with the usage of the EFFR-Advanced and the EFFR-

Beyond scheme, and compare them with the original EFFR scheme,

- as well as the performance of all aforementioned schemes with non-uniform

distributed traffic (or user terminals) in each cell.

Moreover, the extension of approaches taking mobile user terminals into account

would be of great interest, which would allow studying the tradeoff between

frequency-selective scheduling (based on adjacent subchannels suitable for fixed

users) and interference-averaging operation (based on distributed subchannels proper

for mobile terminals). In addition, the research on local ICI coordination schemes in

collaboration with slow-dynamic global resource management (also be known as

adaptive fractional frequency reuse) to further optimize the utilization of limited

frequency resources, as addressed in Section 2.4, are also very desirable. ICI

mitigation in local (distributed) plus dynamic interference coordination among cells

(or BSs) with slow varying M to N ratio in the EFFR design may provide a feasible

optimal solution to attain both coverage and high spectral efficiency. Finally, future

work could also be directed to the combination of the frequency reuse schemes with

7.3 Outlook 151

the spatial reuse techniques to improve the network performance. An initial idea has

been mentioned at the end of Section 2.2. For example, frequency reuse approaches

would be particularly effective when combined with beamforming antennas, which

additionally allow the exploitation of spatial multiplexing and thus the transmission to

spatially separated terminals on the same frequency/time resources.

APPENDIX A

A Simulation Environment – Open Wireless Networks Simulator

Simulation Environment – Open Wireless Networks Simulator [36]*

A.1 Overview ..................................................................................................... 153

A.2 WiMAX MAC Layer ................................................................................. 158

A.1 Overview

The Open Wireless Network Simulator (OpenWNS), developed at ComNets, RWTH

Aachen University, is a sophisticated framework for event driven system level

simulation. The highly modular simulator for performance analysis of mobile radio

networks, written in C++, allows stochastic simulations based on prototypic protocol

implementations combined with a detailed channel model. Typically multi-cellular

scenarios with realistic interference, traffic models and propagation behavior are

simulated. The different ISO/OSI layers are configured as modules from the

application layer to the physical layer and radio channel due to the fact that the

modules can be exchanged according to the deployed approaches. Figure A.1 displays

the different modules of the OpenWNS and their corresponding ISO/OSI layers.

The OpenWNS provides modules for the following layers:

1. Multimedia load generator, e.g. voice, video, and web browsing

2. Transport Layer, e.g., TCP and UDP

3. Network Layer, IP

4. Radio Access Technology, including LLC, RLC, and MAC

5. Interference Calculation, including smart antennas and the link-to-system

interface

——————————————

* The configuration files and the compilable simulator code used to gain the results presented in this thesis

are available for download from [37].

Simulation Environment – Open Wireless Networks Simulator 154

To enable the fast and reliable development of new modules supporting the

performance evaluation of wireless networks, OpenWNS provides a set of support

libraries. The simulator finally consists of one or more modules being loaded at

runtime by the runtime environment. Each module has a specific task and roughly fits

into one of the five categories. The set of support libraries includes:

1. SPEETCL - SDL Performance Evaluation Tool Class Library — A library

being used by all simulators that are implemented in the Specification and

Description Language (SDL). SPEETCL functionality currently in use by

non-SDL simulators is the event scheduler, probes, random number

generation, distributions and basic data types for communication protocols.

2. RISE - Radio Interference Simulation Engine — A library supporting the

simulation of radio interference. Modules for interference calculation (like

OFDMAPhy) are based on this library. It provides the following models:

Stations (base, relay, mobile)

Transmitter and receiver

Figure A.1: The OpenWNS Modules and their corresponding OSI Layers.

A.1 Overview 155

Transmission behavior

Propagation including shadowing and fading

Interference calculation

Mobility

Antennas

3. libWNS - the OpenWNS support library — Among others, it features the

following functionality:

Module handling support

Logging system

Configuration facilities

Decorators for SPEETCL mechanisms including probes, probability

distributions and event queuing

Template based containers, extending those available in STL

Layer Development Kit (LDK) - The FUN based toolbox to model

different radio access technologies

The OpenWNS is an event-driven simulator, which means that the state of the whole

system is changed by discrete events like the transmission or reception of a data packet

or the beginning and ending of a frame. These events are organized by a superior

timing instance that keeps track of the simulation time and invokes all necessary

subroutines of the different modules at the time at which an event occurs. The

OpenWNS timing instance is called event scheduler.

A.1.1 FUN

A Functional Unit (FU) represents one functionality of a protocol within one ISO/OSI

layer. All the FUs of a layer form a Functional Unit Network (FUN).The most

fundamental requirement for FUs is the ability to handle data. A basic data unit that is

transmitted between FUs is denoted as a compound. FUs as part of a protocol stack

may receive compounds for processing before and after such a compound has been

transmitted over the air-interface. The first case is called outgoing data flow, while the

latter case is referred to as incoming data flow. Whenever a compound arrives at a FU,

the FU gains control over the compound. It then can realize different behaviors by

handling the compound accordingly. It may choose to mutate or drop the data unit,

buffer it, forward it to other FUs, or inject new compounds into the FUN.


The control information added by FUs is a so called command. A command can have

different characteristics for different purposes, like an information command or an

acknowledgment command for the ARQ. A FU is completely invisible to the FUs

above. Even underlying FUs do not need to have knowledge about the added

information by this FU. The only FU that is required to be able to handle this

command is the peer unit of this FU. The set containing all the commands of every FU

within a FUN is called command pool. The union of a data unit and a command pool

is denoted as a compound.

A.1.2 Configurability of the OpenWNS

The C++ programming language is utilized to implement the OpenWNS FUs. In

addition to C++, the Python programming language is applied in order to configure the

parameters. The Python code is applied to lightly change the settings of a simulation

campaign. The attributes of a Python class are passed on to the constructor of a C++

class. The idea behind using Python is the possibility of altering the behavior of the

OpenWNS without recompiling the C++ code.

A.1.3 Simulation Progress

The WNS-CORE is the only executable inside the OpenWNS; it serves as a module

loader to include other required modules. All modules are compiled as shared libraries

and accordingly loaded on demand during startup. Thereby, the two core libraries

SPEETCL and libWNS are included in every simulation. They provide various helper

functions and the fundamental classes for event-driven simulations with one global

timing. All simulators inside the OpenWNS are time-discrete and event-driven.

Simulations need the global time in order to control the chronological sequence of

events computed by the simulation.

A simulation possibly progresses as described in the following. Periodically, the load

generator of each station (UT/BS) generates an Internet Protocol (IP) data packet and

feeds it into the WiMAC protocol layer. Afterwards, the packet traverses several FUs

within the DLL FUN where several events are generated. When time for a certain

event is reached in the simulator the corresponding tasks will be executed immediately.

The creation of a new packet after a certain time by the load generator, the periodic

beginning of a new super-frame or frame respectively, as well as the transmission to a

certain time are events. When the packet leaves the DLL FUN of WiMAC after its

specific calculations, it is forwarded to the PHY module that performs an event-based

transmission to a certain time which is already determined within the WiMAC FUN

above.

A.1 Overview 157

When the simulation time has progressed and the time to conduct the transmission is

reached, the transmission event will be executed. This triggers the RISE module to add

the transmission to the set of currently active transmissions. Until the transmission is

finished, all other packets transmitted at the same time on the same frequency band

will experience the interference generated by the transmission. At this time RISE

models the channels behavior, e.g. interference, shadowing, mobility and path loss and

makes some calculations based on it.

As soon as the transmission is over, the RISE module will initiate the incoming packet

at the PHY service access points of the corresponding destination station. Evidently,

RISE can be regarded as the channel. The PHY module at the destination station will

then receive the CINR values and feed the packet to higher protocol functional units

what is consequently the PHY module.

The PHY module again possibly delivers the packet to the WiMAC until it reaches the

application layer of the corresponding station. Thus, the transmission is finished and

successfully transmitted packets can be used for calculations, like delay and

throughput.

A.1.4 Simulator and traffic model

The simulation flow in the internal simulation time: from time to time, the load

generator of each station generates an IP data packet and feeds it into the WiMAX

protocol layer using its SAP. From there, the packet traverses the different functional

units of the WiMAX data link layer until it eventually reaches a queue inside the

scheduler. When a new frame is started, the scheduler is triggered and packets from

the queue are scheduled. When the packet is scheduled, it is forwarded to the PHY

module that programs an event based on the transmission’s start time determined by

the scheduler. When the simulation time has progressed to that start time, the event

will be generated. This triggers the RISE module to add the packet’s transmission to

the set of currently active transmissions in the scenario. Until the transmission is over,

all other packets transmitted at the same time on the same frequency band will

experience the interference generated by the transmission. This interference takes into

account the path loss and the antenna characteristics in form of the beam patterns.

Once the transmission is over, the RISE module will indicate the incoming packet at

the PHY access points of all stations. The PHY module at the destination station will

then retrieve the CINR values and indicate the packet to higher protocol functional

units. These, in turn, will then deliver it to the WiMAX SAP on top of the WiMAX

MAC layer where it is counted as throughput. The time elapsed since the packet has

passed the sender’s WiMAX SAP is counted as the packet’s delay [6].


A.2 WiMAX MAC Layer

Figure A.2 shows the structure of the BS FUN of the WiMAX MAC Layer (WiMAC).

The multimedia traffic generator, the transport and the network layer on top, as well as

the PHY layer including interference calculation below the WiMAC are not shown in

the figure. The WiMAC module can be separated into a Radio Link Control (RLC)

and a Medium Access Control (MAC) sublayer. The RLC sublayer is further

subdivided into a user plane, control plane and management plane [6].

A.2.1 User Plane

The user plane handles user data and performs the corresponding RLC layer

functionality such as classification of SDU to connections, QoS control, SAR, ARQ,

and flow control. The traffic generator requests data transmissions through the Data

Link Layer (DLL) SAP which directs SDUs to the upper convergence FU of the

Figure A.2: WiMAC Functional Unit Network representing the MAC layer at the BS [6].

A.2 WiMAX MAC Layer 159

station. There, the SDU enters the WiMAC. First, the header is prepended which

transforms an SDU into a PDU. The classifier FU identifies PDUs with the help of the

Connection Manager. The classifier marks the PDU with the CID that the connection

manager has reported. Depending on the CID the flow separator FU directs the PDU to

the corresponding buffer FU. The PDU stays in the buffer until the scheduler requests

a compound. Then, the respective PDU passes the SAR and the ARQ FUs.

Other FUs, e.g., the Synchronizer, or the ACK Switch are used for internal purposes,

such as flow control of compounds.

A.2.2 Medium Access Control

Access to the radio resource is controlled by the MAC. It offers transmission services

to the user plane as well as to the control plane. PDUs carrying user data and PDUs

carrying control information pass the CRC and the error modeling FUs. A CRC is

appended at the receiver to detect erroneous PDUs. Except for its higher priority, there

is no difference in the handling, transmission and reception of a control and a data

PDU.

The basic part of the WiMAC is the frame configuration framework, which is

composed of the frame builder and its corresponding FUs. Having passed the DL

scheduler, PDUs are sent through the respective timing node.

Incoming traffic takes the reverse path through the FUN. Payload PDUs go through

the error modeling FU which determines a Packet Error Ratio (PER) for each PDU

received. The CRC FU drops the PDU if a bit error has been detected. Finally, PDUs

carrying user data are forwarded to the user plane while PDUs carrying control

information are forwarded to the control plane.

APPENDIX B

B Reference Scenario

Reference Scenario

B.1 Additional Simulation Results for NLOS DL .......................................... 161

B.2 Additional Simulation Results for NLOS UL .......................................... 163

B.1 Additional Simulation Results for NLOS DL

Figure B.1 shows the mean DL carrier signal strength perceived by the CCUs, the

CEUs as well as the weakest users, respectively. It can be seen that with the Reuse-3

scheme all kinds of UTs can obtain the best carrier single strength among all schemes.

For the CCUs, the SFR scheme is inferior to the other schemes, whereas for the CEUs

and the weakest users the carrier strength by using the Reuse-1 and IFR schemes

remains the worst. On the contrary, like in the LOS case, using the EFFR scheme, the

mean CCU carrier strength is similar to that using the Reuse-1 scheme, the mean CEU

carrier strength is close to which using the Reuse-3 scheme. And the mean weakest

user carrier is even almost identical to the Reuse-3 scheme.

ICI mitigation in a cellular system is the main tasks and discussion emphasis in this

thesis. Figure B.2 illustrates the mean DL interference level perceived by different

types of UTs with increasing offered traffic per user. Among all five investigated

schemes, all kinds of UTs (whether CCUs or CEUs or weakest users) using the Reuse-

3 schemes can be minimum interfered.

Compared to the mean CCU DL interference under LOS condition (refer to

Figure 6.5a), the ICI can be reduced by using the SFR scheme. In contrast, it cannot be

decreased but even increased under NLOS condition compared to the Reuse-1 scheme

(see Figure B.2a) at moderate and full traffic loads. Nonetheless, for the CEUs (see

Figure B.2b) and the weakest users (see Figure B.2c) the interference can be reduced

by around 1.5 dB.

In both LOS and NLOS scenarios, the EFFR scheme can reduce the ICI more

effectively than the SFR scheme for the CEUs, and tends to close to the Reuse-3

scheme (see Figure 6.5b and Figure B.2c). Moreover, the ICI decreasing amplitude

with both the SFR scheme and the EFFR scheme under NLOS condition is greater

than under LOS condition. Compared to the Reuse-1 and the IFR schemes, the EFFR

Appendix B 162

scheme degrades the mean CEU DL interference by 5 dB under NLOS condition,

while by 1.5 dB under LOS condition. The SFR scheme decreases it by 1.5 dB under

NLOS condition while 0.5 dB under LOS condition.


(c) Mean DL carrier perceived by weakest users

Figure B.1: Mean DL carrier signal strength perceived by different types of UTs versus offered traffic

per user under NLOS condition. 25 UTs are uniformly distributed in each cell with a cell radius of 220 m. The range ratio r/R of 0.6 for partitioning CCUs and CEUs is assumed. The weakest users are those

UTs located between 176 m and 220 m away from the BS (i.e., R'/R [0.8, 1]). And for both SFR

and EFFR schemes, the power ratio of high power level to low power level is set as 3. (a) Mean DL

carrier signal strength perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs; (c) Mean DL carrier signal strength perceived by weakest users.

B.2 Additional Simulation Results for NLOS UL 163

B.2 Additional Simulation Results for NLOS UL

The mean UL carrier strength perceived at the BS depending on increasing offered

traffic per user under NLOS condition is presented in Figure B.3. Since transmission

power affects the carrier signal strength, and SFR UTs send packets with more power

on each subchannel in UL than in DL, the received carrier signal strength are therewith


(c) Mean DL interference perceived by weakest users

Figure B.2: Mean DL interference level perceived by different types of UTs versus offered traffic per user under NLOS condition, having the same environment as in Figure B.1: (a) mean DL interference

level perceived by CCUs; (b) mean DL interference level perceived by CEUs; and (c) mean DL

interference level perceived by weakest users.

Appendix B 164

increased by about 2 dB compared to DL (refer to Figure B.1). The mean carrier

strength of the other schemes keeps almost unchanged in UL.

Certainly, higher transmission power also generates more ICI among neighboring cells

in cellular systems. Thus, the mean CCU UL carrier strength using the SFR scheme, as

shown in Figure B.4a, is also increased by around 3 dB when comparing to DL

performance (refer to Figure B.2a). However, there is nearly no effect on the SFR

CEUs and its weakest users (see Figure B.4b and Figure B.4c). Another phenomenon

(a) Mean CCU UL carrier (b) Mean CEU UL carrier

(c) Mean weakest user UL carrier

Figure B.3: Mean UL carrier signal strength perceived at the central BS versus offered traffic per user

under NLOS condition, having the same environment as in Figure B.1: (a) mean CCU UL carrier

signal strength perceived at the central BS; (b) mean CEU UL carrier signal strength perceived at the central BS; and (c) mean weakest user UL carrier signal strength perceived at the central BS.


can be seen in the figures is that the UL interference levels of the other schemes are all

slightly decreased compared to the DL traffic, although the transmission power

distribution over their available bandwidth is unchanged. And in consequence, for the

CCUs, the EFFR has a close performance to the Reuse-1 and the IFR as expected,

whereas the SFR perceives a higher interference than all the other schemes; for the

CEUs and weakest users, the EFFR attains a close performance to the Reuse-3,

whereas the SFR scheme just like the IFR scheme cannot surpass the Reuse-1 scheme

in terms of the ICI mitigation.


(c) Mean weakest user UL interference

Figure B.4: Mean UL interference level perceived at the central BS versus offered traffic per user under NLOS condition, having the same environment as in Figure B.1: (a) mean CCU UL interference

level perceived at the central BS; (b) mean CEU UL interference level perceived at the central BS; and

(c) mean weakest user UL interference level perceived at the central BS.

Appendix B 166

The overall mean UL CINR values perceived by the BS is displayed in Figure B.5a.

Like the LOS case, except the EFFR scheme, the other schemes have similar mean

CINR in DL and UL under NLOS condition. However, in LOS case the mean CINR

(refer to Figure 6.6a and Figure 6.13a) reaped by using the EFFR scheme is very close

to and sometimes even higher than with the Reuse-3 scheme, whereas in NLOS case

the EFFR mean CINR for both DL and UL (see Figure 6.18a and Figure B.5a) cannot

(a) Mean overall UL CINR (b) Mean weakest user UL CINR

(c) Mean CCU UL CINR (d) Mean CEU UL CINR

Figure B.5: Mean UL CINR values perceived at the central BS as a function of offered traffic per user

under NLOS condition, having the same environment as in Figure B.1: (a) mean overall UL CINR values perceived at the central BS; (b) mean weakest user UL CINR perceived at the central BS; (c)

mean CCU UL CINR perceived at the central BS; as well as (d) mean CEU UL CINR perceived at the

central BS.


catch up with the Reuse-3 scheme any more. But it is still much better than with the

other schemes. For the SFR scheme, the weakest users of the SFR scheme (see

Figure B.5b) can even reach a mean UL CINR value of 6.7 dB, which is above the

require valid CINR threshold of 6.4 dB.

Figure B.6 presents the mean CCUs and CEUs coverage percentage of five frequency

reuse schemes versus offered traffic per user in NLOS UL. It can be noticed that the

mean CEUs coverage percentage (see Figure B.6b) by using the SFR is much higher in

UL than in DL by about 36% (refer to Figure 6.21d). This leads to a significantly

bettered overall cell coverage percentage, nonetheless, at the expense of sacrificing the

overall cell throughput (refer to Figure 6.23a and Figure 6.19a). Because of the

relatively higher power level for both CCUs and CEUs in UL SFR, its mean UL CINR

level of the CEUs (refer to Figure B.5c) is advanced while its CCU UL CINR level

(refer to Figure B.5d) is a little impaired. As a result, the performance of CEUs

including average throughput and coverage percentage by using the SFR scheme is

enhanced in UL (refer to Figure 6.23d and Figure 6.24d), but at the cost of reducing

the overall and the CCU UL throughput (refer to Figure 6.23a and c).

(a) Mean UL CCUs coverage percentage (b) Mean UL CEUs coverage percentage

Figure B.6: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per

user under NLOS condition, having the same environment as in Figure B.1: (a) the corresponding

mean CCUs coverage percentage; (b) the corresponding mean CEUs coverage percentage in UL traffic.

APPENDIX C

C Impact of Range Ratio on System Performance

Impact of Range Ratio on System Performance

C.1 Additional Simulation Results for LOS DL ............................................. 169

C.2 Performance in LOS UL ............................................................................ 170

C.3 Performance in NLOS DL ......................................................................... 176

C.4 Performance in NLOS UL ......................................................................... 182

C.5 Conclusion .................................................................................................. 185

C.1 Additional Simulation Results for LOS DL


Figure C.1: Mean DL carrier signal strength perceived by different types of UTs versus range ratio r/R under LOS condition, which is defined as the zone radius for the CCUs r to the cell radius R. 25

UTs are uniformly distributed in each cell with a cell radius of 1000 m. Offered traffic per user of 500

kbps is assumed. And for both SFR and EFFR schemes, the power ratio of high power level to low power level is set as 3. (a) Mean DL carrier signal strength perceived by CCUs; (b) Mean DL carrier

signal strength perceived by CEUs.

Appendix C 170

Figure C.1 gives the mean DL carrier signal strength perceived by different types of

UTs versus range ratio r/R under LOS condition. For both CCUs and CEUs, the DL

carrier strength of all investigated schemes degrades with the increasing range ratio

r/R. Since the larger the range ratio r/R, the more UTs belongs to the CCUs and the

larger the CCU area is. That means the average distance between the BS and the CCUs

as well as that between the BS and the CEUs become farther away leading to

decreased carrier signal strength. Due to the interference-aware-reuse mechanism on

the Secondary Segment used in the EFFR scheme (refer to Section 4.1.2.2), the carrier

strength of the EFFR scheme stays constantly at about -81 dBm when the range ratio

r/R is greater than 0.4, whereas perceived carrier strength of the other schemes still

keep decreasing with the increasing range ratio r/R, see Figure C.1a.

In opposite to the carrier strength, the mean DL interference level for both CCUs and

CEUs of all investigated schemes upgrades with the increasing range ratio r/R, as

shown in Figure C.2. Among all, the interference level by using the SFR is influenced

by the range ratio more heavily than with the other schemes.

C.2 Performance in LOS UL

The mean carrier strength and interference level of different types of UTs experienced

at the BS for UL data stream are exhibited in Figure C.3 and Figure C.4 respectively,

which result in the mean CCU and CEU UL CINR levels as well as the corresponding

mean overall UL CINR as presented in Figure C.5.


Figure C.2: Mean DL interference level perceived by different types of UTs versus range ratio r/R under LOS condition, having the same environment as in Figure C.1: (a) mean DL interference level

perceived by CCUs; (b) mean DL interference level perceived by CEUs.

C.2 Performance in LOS UL 171

In comparison with the mean DL CINR values (refer to Figure 6.26), it can be seen

that all schemes own similar CINR performances in UL and DL. The deviation of the

EFFR scheme with various M to N combinations arises from different resource

scheduling applied on the reuse-1 subchannels in the Primary Segment for the


Figure C.3: Mean UL carrier signal strength perceived at the central BS versus range ratio r/R under LOS condition, having the same environment as in Figure C.1: (a) Mean CCU UL carrier signal

strength perceived at the central BS; (b) Mean CEU UL carrier signal strength perceived at the central

BS.


Figure C.4: Mean UL interference level perceived at the central BS versus range ratio r/R under LOS condition, having the same environment as in Figure C.1: (a) mean CCU UL interference level

perceived at the central BS; (b) mean CEU UL interference level perceived at the central BS.

Appendix C 172

downstream and upstream transmissions. In DL, PF combined with interference-

aware-mechanism is applied on all reuse-1 subchannels, whereas in UL the same

resource scheduling is only used on the reuse-1 subchannels in the Secondary Segment.

In order to attain relatively stable and accurate CINR estimation for the reuse-1

subchannels in the Secondary Segment in UL, the maximum throughput strategy is

carried out on the reuse-1 subchannels in the Primary Segment.



Figure C.5: Mean UL CINR values perceived at the central BS depending on range ratio r/R under LOS condition, having the same environment as in Figure C.1: (a) mean overall UL CINR values

perceived at the central BS; (b) the corresponding mean CCU UL CINR values perceived at the central BS; and (c) the corresponding mean CEU DL CINR values perceived at the central BS.


Figure C.6a displays the average overall cell capacity for uplinks as a function of the

range ratio r/R under LOS propagation. Like in DL, the EFFR scheme can provide a

remarkable improvement on the overall cell capacity. And it outperforms all the other

schemes in every situation, regardless of with which M to N combination. The

performance of the SFR scheme is strongly influenced by the range ratio. With an

inappropriate choice of the range ratio, its performance will be severely deteriorated,

whereas using EFFR the cell capacities do not vary so drastic under different range



Figure C.6: Mean UL MAC throughput under LOS condition depending on range ratio r/R in the same environment as in Figure C.1: (a) mean overall UL cell throughput; (b) the corresponding mean

CCU UL throughput; and (c) the corresponding mean CEU UL throughput.

Appendix C 174

ratios. As a consequence, the proposed EFFR design can gain more robustness than the

SFR scheme.

Figure C.6b and c give the corresponding CCU and CEU mean throughput depending

on the range ratio r/R. When comparing to the DL throughput performance, all

investigated schemes have similar behavior in UL and DL. Yet their UL throughput is

somewhat lower than in DL. This results mainly from two reasons. One is because the

UL phases in total are 9 OFDMA symbols shorter than the DL phases in each



Figure C.7: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R

under LOS condition, having the same environment as in Figure C.1: (a) mean overall UL cell

coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs coverage percentage in UL traffic.


superframe. The other reason causing the difference between the UL and the DL is due

to the different constraint of transmission power for the BS and the UTs. In all

scenarios in this chapter, a constant total system transmission power is assumed, and

thereby each UT has a maximal transmission power of 200 mW as described in

Section 6.1.5.2. That means in the SFR or the EFFR design each CCU may maximal

use three reuse-1 subchannels simultaneously and each CEU is however allowed to

occupy maximal one reuse-3 subchannel. Likewise, UTs in the IFR and Reuse-1 can

be assigned maximal three subchannels concurrently, whereas using the Reuse-3

scheme each UT cannot utilize more than one subchannel at the same time to transmit

packets.

The similar CINR performance in DL and UL (refer to Figure 6.26 and Figure C.5) is

also reflected in the similar cell coverage performance in DL and UL. Figure C.7

illustrates the mean DL overall cell coverage and the corresponding coverage

percentage of CCUs and CEUs depending on range ratio r/R under LOS condition. It

can be seen that the overall UL cell coverage percentage applying the EFFR scheme is

larger than using the SFR, the IFR and the Reuse-1 schemes when the range ratio is

r/R < 0.8 (see Figure C.7a). Even if the range ratio r/R = 0.8 is chosen, the overall UL

cell capacity can be enhanced by the EFFR scheme and higher than with any of the

other schemes (refer to Figure C.6a), although its cell coverage performance is close to

the Reuse-1 scheme. Moreover, when the range ratio is r/R < 0.4, both cell coverage

and cell capacity of the EFFR scheme are the best among all schemes. In contrast, with

the SFR scheme, the cell coverage and the cell capacity can also be significantly

improved compared to the Reuse-1 with the range ratio r/R < 0.5. Nevertheless, when

a range ratio r/R > 0.5 is chosen, both its cell coverage and cell capacity deteriorate

Figure C.8: Average UL cell spectral efficiency of five frequency reuse schemes depending on range ratio r/R under LOS condition, having the same environment as in Figure C.1.

Appendix C 176

sharply and become the worst among all reuse schemes.

In the end, the corresponding mean UL cell spectral efficiency of all five frequency

reuse schemes depending on range ratios under LOS propagation is shown in

Figure C.8. The results indicates that the proposed EFFR scheme has substantial

superiority over the SFR, the IFR as well as the Reuse-1 schemes, and even slightly

better than the Reuse-3 scheme when the range ratio is r/R < 0.5.

Taking all evaluated performance into consideration (refer to Figure C.6a, Figure C.7a

and Figure C.8), it can be concluded that with a range ratio r/R < 0.5 the EFFR scheme

outperforms all the other schemes in LOS UL, and can gain not only considerable

profits in the overall cell capacity, but also the best cell coverage percentage and the

best cell spectral efficiency.

C.3 Performance in NLOS DL

The subsequent two subsections present and compare the performance of all the

investigated frequency reuse solutions for ICI mitigation under NLOS condition,

where the path loss coefficient γ is nearly two times higher than under LOS

propagation (see Eq. (5.29)). Different affects caused by range ratio definitions in

comparison with that under LOS condition will be disclosed. Again, 25 UTs are

evenly placed in each cell but with a cell radius of 220 m, which might be the valid

maximal cell radius for the EFFR and the Reuse-3 schemes under NLOS condition

derived from Section 5.3 and has been confirmed in Subsections 6.2.1.3 and 6.2.1.4.

Same as in the Reference Scenario, besides the overall cell performance and the

performance of difference types of UTs, the weakest user performance will also be

evaluated in NLOS case. And the weakest users are those UTs located from 176 m to

220 m away from the BS, which means 4 of 25 UTs belongs to the weakest users

(refer to Table 6.4).

Figure C.9 illustrates the mean DL CINR performance of all investigated frequency

reuse schemes versus range ratio r/R under NLOS condition. Different to the LOS DL

situation, the mean overall DL CINR values of all five schemes are above the required

CINR threshold of 6.4 dB as shown in Figure C.9a, although the overall CINR of the

SFR is inferior to the Reuse-1 and IFR schemes with range ratio r/R ≥ 0.7. Even so,

the SFR CEUs cannot get a valid CINR when a range ratio r/R > 0.6 is chosen, and the

mean CINR perceived by its weakest users even does not keep gut enough when the

range ratio is r/R > 0.4. In contrast, with the EFFR scheme, the overall CINR retains

stable above the required threshold with any range ratio assumption located between

the Reuse-3 and Reuse-1 curves. Simultaneously, it can provide sufficient CINR for

C.3 Performance in NLOS DL 177

all UTs whether CCUs or CEUs or weakest users. That is to say, the ubiquity of user

throughput covering the whole cell surface area can be achieved.



Figure C.9: Mean overall DL CINR values and the corresponding mean DL CINR values perceived

by different types of UTs depending on range ratio r/R under NLOS condition. 25 users are uniformly

distributed in each cell with a cell radius of 220 m. The weakest users are those users located between 176 m and 220 m away from the BS (i.e., R'/R = [0.8 1]). Offered traffic per user of 500 kbps is

assumed. And for both SFR and EFFR schemes, the power ratio of high power level to low power

level is set as 3. (a) Mean overall DL CINR values; (b) the corresponding mean DL CINR values perceived by weakest users; (c) the corresponding mean DL CINR values perceived by CCUs; as well

as (d) the corresponding mean DL CINR values perceived by CEUs.

Appendix C 178

The resulting mean DL MAC throughput under NLOS condition depending on range

ratio is presented in Figure C.10. In general, the overall cell capacity of all reuse

schemes under NLOS propagation (see Figure C.10a) is much higher than in LOS case

(refer to Figure 6.27a). But the enhancement with the EFFR under NLOS condition is

not as great as under the LOS in comparison with the static Reuse schemes. For

example, the EFFR scheme with an M to N combination of 5:5 provides am best

around 26% overall throughput benefits compared to the Reuse-3 and around 63%



Figure C.10: Mean DL MAC throughput under NLOS condition depending on range ratio r/R in the

same environment as in Figure C.9: (a) mean overall DL cell throughput; (b) the corresponding

weakest user DL throughput; (c) the corresponding mean CCU DL throughput; as well as (d) the corresponding mean CEU DL throughput.


compared to the Reuse-1, whereas it reaps an immense increase of approximately

131% over the Reuse-3 and approximately 255% over the Reuse-1. However, under

NLOS condition the UTs close to cell border (i.e. weakest users, see Figure C.10b) of

the EFFR can be served, whilst for the LOS DL this is not the case.

With a range ratio r/R ≤ 0.6, the EFFR scheme can offer a stable overall cell capacity.

And it outperforms the Reuse-3 regardless of with which M to N combination.

Moreover, it can be observed that the more numbers of the reuse-1 subchannels N

invested for the CCUs, the more profits in the overall system capacity can be achieved.

Nonetheless, with a further increase of the CCU-zone the cell capacity decreases,

which arises from the improvement of the CEUs (including the weakest users)

performance at the sacrifice of the interests of the CCUs, as shown in Figure C.10b, c

and d.

With respect to the SFR scheme, it only performs better than the Reuse-3 when the

range ratio is 0.4 ≤ r/R ≤ 0.6 in terms of overall cell capacity. And its performance is

very unstable with different range ratios. It can be noted that the range ratio has a great

impact on the performance, and with an inappropriate choice the performance of the

SFR can be severely damaged. The results show that with r/R ≤ 0.5, the CCUs perform

very well due to the sufficiency of the available bandwidth (20 subchannels).

Nevertheless, the improvement on the CEUs is very limited compared to the Reuse-1

scheme. This is because the SFR CEUs are still vulnerable and grievously interfered

by the concurrent CCU transmissions in the adjacent cells, even though higher power

is used on them. And when the range ratio is r/R ≤ 0.3, only few UTs are defined as

CCUs who can use the 20 inclusive reuse-1 subchannels, but most of the UTs as the

CEUs are just allowed to utilized the remaining 10 subchannels, which results in both

lower overall cell throughput and lower CEU throughput. Even when the range ratio is

r/R = 0.5, the overall system capacity of the SFR attains its best performance, its

weakest users cannot be served at all and its CEUs performance stays unimproved

compared to the Reuse-1. Further when the range ratio is r/R > 0.5, not only cell

capacity but also mean CCU throughput degrade rapidly, and the cell coverage cannot

be kept any more (the user near to the cell border have no throughput as shown in

Figure C.10b). This effect implies an inappropriate range ratio definition in SFR not

only results in a wasting of precious resources (including spectrum and power), but

also yields excessive ICI to the reuse-1 UTs in the neighboring cells. And when a

range ratio r/R ≥ 0.7 is chosen, the overall cell capacity becomes even worse than with

the Reuse-1 scheme.

In contrast, the EFFR bases on ensuring the performance of the CEUs like which with

the Reuse-3, promotes the performance of the CCUs by peeling off part resources

from the reuse-3 resources to lunch into the reuse-1 utilization. As a consequence, the

Appendix C 180

average cell capacity is enhanced due to the increase of available bandwidth while

retaining lower ICI at the cell edge.

The weakest user throughput performance displayed in Figure C.10b can indirectly

reflect the cell coverage performance of all investigated schemes, though Figure C.11

can give more explicit and more detailed coverage percentage depending on various

range ratio definitions. The results show that the Reuse-3 can provide 100% cell



Figure C.11: Mean DL coverage percentage of five frequency reuse schemes versus range ratio r/R

under NLOS condition, having the same environment as in Figure C.9: (a) mean overall DL cell

coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in

DL traffic.


coverage under NLOS condition, whilst that is not the case under LOS condition. And

with the proposed EFFR scheme, its mean CCU coverage percentage always coincides

with the Reuse-1 (see Figure C.11c), whereas its mean CEU (including the weakest

users) coverage percentage is very close to that using the Reuse-3 scheme (see

Figure C.11d and b). This leads to a consequence as shown in Figure C.11a that the

range ratio should not be larger than 0.6 so that an approximately 100% cell coverage

can be ensured.

As for the SFR, only with a range ratio r/R ≤ 0.2 all UTs can be successful served.

This is due to the fact that the range ratio r/R ≤ 0.2 means maximal 1 CCU defined in

each cell, which results in very few ICI generated for the neighboring simultaneous

CEU and CCU transmissions. In the same way, its CEUs perceive also very few ICI

from the concurrent CCU transmissions in the neighborhood, and have a similar

environment to the UTs using the Reuse-3. However, as mentioned before, the cell

coverage performance should be observed combined with the achievable cell

throughput performance. From the results in Figure C.10a, with a range ratio r/R ≤ 0.3,

the SFR throughput performance cannot exceed the Reuse-3 scheme. Only with a

range ratio definition of 0.4 ≤ r/R ≤ 0.6, the SFR can reap a better throughput

performance than both static Reuse schemes. Nevertheless, its cell coverage

percentage performance in these cases can only surpass the Reuse-1, but inferior to the

cell coverage performance with the Reuse-3 and the proposed EFFR scheme. This

denotes that the increase in system capacity comes at the expense of cell coverage

reduction due to increased ICI, which severely impacts CEU coverage percentage and

throughput performance. Besides, with an inappropriate range ratio definition, for

example r/R ≥ 0.7, neither its cell coverage performance nor its cell capacity

Figure C.12: Average DL cell spectral efficiency of five frequency reuse schemes depending on range ratio r/R under NLOS condition, having the same environment as in Figure C.9.

Appendix C 182

performance can even match the Reuse-1. On the contrary, the proposed EFFR scheme

never performs worse than the Reuse-1 and the IFR scheme with any range ratio

definition.

Lastly, the average DL cell spectral efficiency of all investigated frequency reuse

schemes depending on range ratio definitions is exhibited in Figure C.12. Different to

that under LOS propagation, the EFFR scheme can never be superior to the Reuse-3

scheme under NLOS condition. And EFFR with different M to N combinations

achieves close cell spectral efficiency performance. But, they still evidently gain the

advantage over the SFR, the IFR and the Reuse-1 scheme. Comparing the EFFR and

the Reuse-3 system, a tradeoff between a reduced cell spectral efficiency and the

achieved cell capacity improvement occurs. Furthermore, as mentioned in Section

6.2.1.3, the system spectral efficiency of Reuse-3 can just reach 1/3 of its cell spectral

efficiency, and the system spectral efficiency of the EFFR scheme is also lower than

its cell spectral efficiency (53.3% of its cell spectral efficiency with EFFR 7:3, 60%

with EFFR 6:4, 66.7% with EFFR 5:5, and 73.3% with EFFR 4:6, respectively).

Hence, actually, the system spectral efficiency by using the EFFR schemes (except

M:N = 7:3) can surpass that with the Reuse-3 scheme, and the EFFR 4:6 can even

outperform all the other schemes, including the SFR and the Reuse-3 shemes.

C.4 Performance in NLOS UL

In Figure C.13, the mean UL MAC throughput of different types of UTs depending on

range ratios is presented. The results in Figure C.13a show that the EFFR scheme can

also provide a remarkable improvement on the overall cell capacity in UL, and it

outperforms all the other schemes in every situation regardless of with which M to N

combination. A little different to the DL is that all performances of the EFFR with

various M to N combinations stay relatively closer with each other.

The SFR can also enhance the system throughput in comparison with the static Reuse

schemes and the IFR scheme until the range ratio r/R < 0.6 is chosen. Nevertheless, it

cannot compare favorably with the EFFR scheme except with the range ratio r/R = 0.5.

And when a range ratio r/R larger than 0.5 is chosen, its performance deteriorates so

drastically that it starts inferior to the Reuse-3 when r/R ≥ 0.6, and even worse than the

Reuse-1 when r/R > 0.7. Even so, when comparing the SFR performance in DL (refer

to Figure C.10), its CEU and weakest user performance in UL are better than in DL

due to application of higher transmission power level on them. And although they

cannot come up with which using the EFFR scheme in every situation, they apparently

surpass the CEU and weakest user performance with the Reuse-1 and the IFR scheme.

C.4 Performance in NLOS UL 183

Not only the throughput performance but also the coverage percentage performance of

the SFR CEUs and weakest users are in UL better than in DL (see Figure C.14b and d,

and compared to Figure C.11b and d). And that finally results in a better overall UL

cell coverage percentage as shown in Figure C.14a. But it is still worse than using the

EFFR scheme and the Reuse-3. Yet the gap is reduced in UL.



Figure C.13: Mean UL MAC throughput under NLOS condition depending on range ratio r/R in the same environment as in Figure C.9: (a) mean overall UL cell throughput; (b) the corresponding

weakest user UL throughput; (c) the corresponding mean CCU UL throughput; as well as (d) the

corresponding mean CEU UL throughput.

Appendix C 184

The EFFR overall cell coverage performance in UL is also slightly better than in DL.

It can even ensure a 100% cell coverage until r/R = 0.6. However, with further

enlarging the range ratio, some of its CCUs cannot transmit data to the BS successfully

(see Figure C.14c). As a consequence, considering combined with the overall cell

throughput, the range ratio r/R = 0.6 should be the maximum adequate value for the

EFFR scheme in both UL and DL.


(c) Mean UL CCUs coverage percentage (d) Mean UL CEUs coverage percentage

Figure C.14: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R under NLOS condition, having the same environment as in Figure C.9: (a) mean overall UL cell

coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding

mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in UL traffic.

C.5 Conclusion 185

Figure C.15 exhibits the corresponding average cell spectral efficiency of all

investigated reuse schemes under NLOS condition in UL. Like in DL, the Reuse-3 can

always reap the best cell spectral efficiency, and the EFFR gains the second best place.

The SFR scheme can just beat the Reuse-1 and the IFR scheme until r/R = 0.7.

C.5 Conclusion

In this appendix, additional performance estimation in terms of carrier signal,

interference level, CINR, MAC throughput, coverage percentage as well as cell

spectral efficiency of all investigated frequency reuse schemes depending on range

ratio under both LOS and NLOS condition is presented and elucidated. The results in

this appendix and in Section 6.2.2 show that with the EFFR scheme not only the

system performance but also the CEU performance is able to be substantially

enhanced for both DL and UL with any propagation condition. Furthermore, with

respect to range ratio definition for dividing CCU-zone and CEU-zone, the proposed

EFFR scheme can provide more flexibility and robustness than the SFR scheme.

Under LOS propagation, the EFFR scheme except for the M to N combination of 8:2

in DL but with any M to N combination in UL can provide the best system capacity,

the second best (but very close to the best) cell coverage percentage as well as the best

spectral efficiency with a range ratio r/R ≤ 0.4. Under NLOS propagation with a range

ratio r/R ≤ 0.6, the EFFR surpasses the static Reuse schemes and the IFR scheme in

DL cell capacity, and ensure more than 95% overall cell coverage, as well as provides

second best average DL cell spectral efficiency, whereas in UL it can even provide the

Figure C.15: Average UL cell spectral efficiency of five frequency reuse schemes depending on range

ratio r/R under NLOS condition, having the same environment as in Figure C.9.

Appendix C 186

best cell capacity among all schemes, 100% cell coverage as well as second best cell

spectral efficiency. Thus, it can be concluded that the EFFR scheme is the best

solution among all five investigated frequency reuses schemes for mitigating ICI, with

which the precious resources can be more effectively utilized, and a good tradeoff

among cell capacity, cell coverage and cell spectral efficiency is made.

APPENDIX D

D Impact of Power Ratio on System Performance

Impact of Power Ratio on System Performance

D.1 Performance in LOS UL ............................................................................ 187

D.2 Performance in NLOS DL ......................................................................... 193

D.3 Performance in NLOS UL ......................................................................... 198

D.4 Conclusion .................................................................................................. 201

D.1 Performance in LOS UL

Since a fixed UT maximal transmission power of 200 mW is assumed and each CEU

is allowed to send packets with its full power, the only difference between DL and UL

is therefore that each type of UTs in SFR and EFFR has the same power allocation

(refer to Table 6.5 and Table 6.6). Therewith the influence on the performance caused

by the ICI difference between inclusive reuse and exclusive reuse can be disclosed.

And the ICI mitigation efficiency by using the SFR scheme and the EFFR scheme can

be easily compared.

Figure D.1 exhibits the mean UL carrier strength perceived at the central BS versus

power ratio Phigh/Plow under LOS condition. As CCUs and CEUs in the SFR transmit

with the same power as they in the EFFR, a close carrier performance can be achieved

by the both RUP techniques. The performance difference among the SFR scheme and

the EFFR scheme with various M to N combinations is due to 2 factors. One is that a

slight difference in the resource assignment and scheduling process exists for the SFR

scheme and the EFFR scheme. In order to attain more precise channel quality

estimation and at the same time not lose fairness among UTs, in the EFFR the reuse-1

subchannels in the Primary Segment are allocated to UTs using the maximum

throughput strategy, while the other subchannels (including the remaining reuse-3

subchannels in the Primary Segment and the reuse-1 subchannels in the Secondary

Segment) are distributed to UTs using the PF scheduling strategy. In the SFR, however,

only PF strategy is used. Another factor causes difference in the carrier performance

between the SFR scheme and the EFFR scheme is that an interference-aware-reuse

mechanism is applied in the Secondary Segment in the EFFR design, whereas plain

reuse without CQI estimation is used in the SFR scheme.

Appendix D 188

Figure D.2 illustrates the corresponding mean UL interference level of different types

of UTs experienced at the BS. Here a clear ICI difference caused by using inclusive

reuse and exclusive reuse is exposed. For the CCUs, with reduced transmission power

the UL ICI in the EFFR decreases quickly, whereas that is not the case in the SFR.


Figure D.1: Mean UL carrier signal strength perceived at the central BS versus power ratio of high

power level to low power level used for the SFR and the EFFR schemes under LOS condition, having

the same environment as in Figure 6.30: (a) Mean CCU UL carrier signal strength perceived at the central BS; (b) Mean CEU UL carrier signal strength perceived at the central BS.


Figure D.2: Mean UL interference level perceived at the central BS versus power ratio of high power

level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean CCU UL interference level perceived at the central BS;

(b) mean CEU UL interference level perceived at the central BS.

D.1 Performance in LOS UL 189

Hence, with a power ratio Phigh/Plow starting from 2 the UL ICI using the EFFR is

much lower than using the SFR. This results in the a lot better CCU UL CINR value of

the EFFR compared to the SFR scheme, as shown in Figure D.3b. With a power ratio

Phigh/Plow > 4, the BS of the SFR cannot perceive a valid CINR level of 6.4 dB any

more. For the CEUs (see Figure D.2b), with reduced power for the concurrent CCU-



Figure D.3: Mean UL CINR values perceived at the central BS depending on power ratio of high


the same environment as in Figure 6.30: (a) mean overall UL CINR values perceived at the central BS; (b) the corresponding mean CCU UL CINR values perceived at the central BS; and (c) the

corresponding mean CEU DL CINR values perceived at the central BS.

Appendix D 190

transmissions in the neighboring cells the UL ICI in the SFR decreases, nevertheless, it

still much higher than using the EFFR scheme. This results in inferior CEU CINR

values of the SFR scheme, which is invalid until when the power ratio Phigh/Plow is

larger than 9, as shown in Figure D.3c. Taking a look at both Figure D.3b and c, all

kinds of UTs using the EFFR can always provide valid transmissions regardless with

which power ratio, whereas using the SFR design it is never the case that the BS can



Figure D.4: Mean UL MAC throughput under LOS condition depending on power ratio of high power

level to low power level used for the SFR and the EFFR schemes in the same environment as in

Figure 6.30: (a) mean overall UL cell throughput; (b) the corresponding mean CCU UL throughput; and (c) the corresponding mean CEU UL throughput.

D.1 Performance in LOS UL 191

receive valid packets from both CCUs and CEUs. As a consequence, with the EFFR

scheme the UL CINR performance is always above the required CINR threshold of 6.4

dB and very close to the Reuse-3 performance (see Figure D.3a), whereas using the

SFR scheme its UL CINR performance decreases with increasing power ratio and

cannot hold valid in most cases, although it is much better than the Reuse-1

performance.



Figure D.5: Mean UL coverage percentage of five frequency reuse schemes versus power ratio of high


the same environment as in Figure 6.30: (a) mean overall UL cell coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs

coverage percentage in UL traffic.

Appendix D 192

The UL CINR performance and the available bandwidth of each investigated scheme

determine its UL throughput performance, as shown in Figure D.4. In general, the

EFFR scheme can provide a remarkable improvement on the overall UL cell

throughput (see Figure D.4a). The SFR scheme, however, can only surpass the Reuse-

1 and the IFR scheme with any power ratio. But, it can never exceed the EFFR scheme,

which is unlike the DL, although they apply identical power allocation for their UTs

and the SFR even has more available bandwidth than the EFFR. When a power ratio

Phigh/Plow > 4 is chosen, the SFR performs even inferior to the Reuse-3, although the

SFR owns triple available bandwidth than the Reuse-3 scheme.

Detailed observation in terms of mean UL throughput of a CCU and a CEU are

presented in Figure D.4b and c. Figure D.4b shows that the SFR scheme performs

even worse than the Reuse-1 when a power ratio Phigh/Plow > 6 is chosen, which favors

the SFR CEUs so that they can come up to the EFFR CEU throughput performance as

shown in Figure D.4c. The EFFR scheme, by contrast, can always provide a fairly

good performance whether in overall or in different types of UTs respect.

The superiority by applying the EFFR scheme is not only reflected in the mean

throughput but also in the cell coverage percentage. Figure D.5 illustrates the mean UL

overall cell coverage and the corresponding coverage percentage of CCUs and CEUs

depending on power ratio Phigh/Plow under LOS condition. For the CCUs, unlike the

DL, the EFFR scheme gains a close performance to the Reuse-3 scheme of around

95%, and never performs inferior to the Reuse-1. The SFR, by comparison, can only

hold a better performance than the Reuse-1 with a power ratio Phigh/Plow < 3, which is

the outcome of excessive ICI from neighboring cells (refer to Figure D.2a) due to its

Figure D.6: Average UL cell spectral efficiency of five frequency reuse schemes depending on power

ratio of high power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30.

D.2 Performance in NLOS DL 193

inherent but adversely inclusive RUP design. For the CEUs (see Figure D.5c), though

the SFR gains quite well in advance of the Reuse-1 scheme, it is much inferior to the

EFFR performance, which is almost identical with the Reuse-3 performance. As a

consequence (see Figure D.5a), the EFFR and the Reuse-3 provide similar UL overall

cell coverage performance, and gain a dominated advantage compared to the other

schemes.

Finally, the mean UL cell spectral efficiency of all five frequency reuse schemes

depending on power ratio under LOS propagation is given in Figure D.6. It can be

seen that the EFFR scheme outperforms all the other schemes except the point that the

power ratio Phigh/Plow equals 1, regardless of with which M to N combination.

Making a survey of all performance evaluations in respect of power allocation under

LOS condition, it can be concluded that the EFFR scheme is more flexible and stable

than the SFR scheme, and can gain decent performance at the same time including cell

capacity, cell coverage percentage as well as cell spectral efficiency. Furthermore, it

provides substantial improvements not only on the overall performance but also on the

performance of different types of UTs.

D.2 Performance in NLOS DL

Like in Section 6.2.1 and in Appendix C, the performance of all the investigated

frequency reuse solutions depending on power ratio under NLOS condition is revealed

in the following two subsections. According to the analysis results presented in

Section 5.3, the maximum valid reach by using the Reuse-1 scheme is 0.6 R under

NLOS condition. Hence, the range ratio r/R of 0.6 to divide UTs into CCUs and CEUs

for the SFR scheme and the EFFR scheme is chosen for NLOS simulations. That is to

say, among 25 UTs, 12 UTs are CCUs and the remaining 13 UTs are CEUs.

Figure D.7 displays the mean DL CINR performance depending on power ratio

Phigh/Plow used for the SFR and the EFFR schemes under NLOS condition. It can be

seen that using the EFFR scheme the mean DL CINR perceived by the CCUs is close

to the Reuse-1 CCU performance (see Figure D.7c), whereas its DL CINR

performance perceived by CEUs and weakest users upgrades gradually close to the

Reuse-3 performance with increasing power ratio Phigh/Plow (see Figure D.7d). But,

since range ratio r/R of 0.6 is assumed, the number of the CCUs and the CEUs is

nearly equivalent, which leads to the mean EFFR overall DL CINR not as good as the

Reuse-3 performance, but much better than the Reuse-1 performance (see Figure D.7a).

Moreover, in comparison with the SFR, the EFFR CINR performance retains

relatively stable with increasing power ratio Phigh/Plow and always above the required

CINR threshold of 6.4 dB, no matter perceived by which type of UTs.

Appendix D 194

For the SFR scheme, although its mean overall DL CINR and its mean DL CCU CINR

are above the valid CINR threshold with any assumed power ratio, the most remote



Figure D.7: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by different types of UTs depending on power ratio of high power level to low power level used for

the SFR and the EFFR schemes under NLOS condition. 25 UTs are uniformly distributed in each cell

with a cell radius of 220 m. Offered traffic per user of 500 kbps is assumed. The range ratio r/R of 0.6 for partitioning CCUs and CEUs is assumed, and the weakest users are those UTs located between 176

m and 220 m away from the BS (i.e., R'/R [0.8, 1]). (a) Mean overall DL CINR values; (b) the

corresponding mean DL CINR values perceived by weakest users; (c) the corresponding mean DL

CINR values perceived by CCUs; as well as (d) the corresponding mean DL CINR values perceived by CEUs.


UTs cannot receive valid CINR values until power ratio Phigh/Plow ≥ 6, as shown in

Figure D.7b. But on the other hand, higher power ratio causes higher ICI and thereby

lower received CINR level for the CCUs. The CINR gain for CEUs cannot

compensate the CINR loss for CCUs, which results in lower mean overall CINR very

close to that using the Reuse-1 when the power ratio is Phigh/Plow ≥ 6.



Figure D.8: Mean DL MAC throughput under NLOS condition depending on power ratio of high power level to low power level used for the SFR and the EFFR schemes in the same environment as in

Figure D.7: (a) mean overall DL cell throughput; (b) the corresponding weakest user DL throughput;

(c) the corresponding mean CCU DL throughput; as well as (d) the corresponding mean CEU DL throughput.

Appendix D 196

The key object of investigating RUP techniques is to improve cell-edge performance

while retaining system spectrum efficiency of reuse-1. Figure D.8 exhibits the mean

DL MAC throughput under NLOS condition depending on power ratio. From the

results it can be seen that the SFR can substantially advance the CEU performance



Figure D.9: Mean DL coverage percentage of five frequency reuse schemes versus power ratio of high power level to low power level used for the SFR and the EFFR schemes under NLOS condition,

having the same environment as in Figure D.7: (a) mean overall DL cell coverage percentage; (b) the

corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in DL traffic.


(see Figure D.8d) and the most remote use performance (see Figure D.8b) when the

power ratio is Phigh/Plow ≥ 5. And at the same time it can hold the overall cell

throughput performance similar to using the Reuse-3 and better than the Reuse-1

performance. The EFFR scheme, by comparison, performs better than the SFR. No

matter with which power ratio, the EFFR can considerably enhance both the CEUs

performance as well as the overall cell throughput. Furthermore, among all four M to

N combinations, the EFFR with M to N combination of 5:5 and 4:6 are better than with

the other two combinations, as they can maintain the mean CCU throughput higher

than using the Reuse-1 while the mean CEU performance close to the Reuse-3

performance, which leads to the overall cell throughput significantly higher than with

both static Reuse schemes in every situation.

Figure D.9 displays the mean overall cell coverage percentage and the corresponding

mean coverage percentage of different types of UTs depending on power ratio

Phigh/Plow in NLOS DL. Like under LOS condition, the EFFR scheme performs much

better than the SFR. Moreover, the CEU coverage performance of all investigated

schemes under NLOS propagation is better than under LOS condition.

The EFFR scheme by contrast can achieve a 100% CCUs coverage when the power

ratio Phigh/Plow does not exceed 7, whereas its mean CEU (including the weakest users)

coverage percentage is always very close to that using the Reuse-3 scheme, and almost

all CEUs can be successful served.

In the end, the average DL cell spectral efficiency of all investigated frequency reuse

schemes as a function of power ratio Phigh/Plow is illustrated in Figure D.10. Different

Figure D.10: Average DL cell spectral efficiency of five frequency reuse schemes depending on power ratio of high power level to low power level used for the SFR and the EFFR schemes under

NLOS condition, having the same environment as in Figure D.7.

Appendix D 198

to that under LOS propagation, the cell spectral efficiency performance of the EFFR

scheme is inferior to that with the Reuse-3 scheme under NLOS condition. Even

through, it is still much better than using the Reuse-1, and increases with increasing

power ratio, whereas using the SFR scheme the average DL cell spectral efficiency is

just slightly enhanced compared to the Reuse-1 and reduces with increasing power

ratio.

As a conclusion, in NLOS DL case, the utilization of both SFR and EFFR schemes

allows substantial performance improvements for wireless broadband systems in

multi-cellular deployments. Nevertheless, the SFR system performance is more

susceptible to the power ratio definition, whereas the EFFR design can provide more

robustness and flexible than the SFR scheme in this respect. Moreover, much more

enhancement can be gained with the application of the EFFR scheme than using the

SFR scheme in terms of overall cell capacity, CEU throughput, overall cell coverage

as well as cell spectral efficiency. In the concrete, the above presented results show

that the power ratio Phigh/Plow of 7 is the best choice for the SFR systems, where it

reaps an increase in overall cell throughput by about 33%, in mean CEU throughput by

150%, in overall cell coverage by about 56% and in cell spectral efficiency by around

33%. With the EFFR scheme, the M to N combination of 5:5 and 4:6 are better than

the other combinations, and any of the power ratio Phigh/Plow ≤ 7 is adequate to achieve

around 67% gains in overall cell throughput, more than 150% benefits in mean CEU

throughput, 100% cell coverage as well as approximately 167% profits in cell spectral

efficiency.

D.3 Performance in NLOS UL

Like in LOS UL scenario, CEUs in both SFR system and EFFR system send packets

with their full power of 200 mW. Hence, power allocation for both CEUs and CCUs in

an EFFR system is as same as that in a SFR system. Specific power allocation for

CCUs and CEUs with varying power ratios can be found in Table 6.5 and Table 6.6,

respectively.

In Figure D.11, the mean overall UL cell throughput and the corresponding mean UL

MAC throughput of different types of UTs as a function of power ratio are exhibited.

It can be seen that the EFFR scheme regardless of with which M to N combination

outperforms both static Reuse schemes in terms of overall cell throughput and mean

CCU throughput in every situation, whilst its mean CEU and weakest user throughput

keep close to the Reuse-3 throughput performance, which is the best among all

schemes.

D.3 Performance in NLOS UL 199

Figure D.12 displays the mean overall cell coverage percentage and the corresponding

mean coverage percentage of different types of UTs depending on power ratio

Phigh/Plow in NLOS UL. Different to the DL case, applying the SFR scheme all UTs

near the cell rand can successfully send packets to their BS when Phigh/Plow ≥ 8 (see

Figure D.12b). Nonetheless, in these situations the BS can receive valid packets from

only less than 60% of the CCUs (see Figure D.12c), which causes the overall UL cell

coverage less than 80%. The results in Figure D.12a show that the SFR reaches its best



Figure D.11: Mean UL MAC throughput under NLOS condition depending on range ratio r/R in the

same environment as in Figure D.7: (a) mean overall UL cell throughput; (b) the corresponding weakest user UL throughput; (c) the corresponding mean CCU UL throughput; as well as (d) the

corresponding mean CEU UL throughput.

Appendix D 200

UL cell coverage performance of around 84% when Phigh/Plow = 5. And in this case,

neither CCUs nor CEUs can attain full coverage percentage. The EFFR scheme by

contrast performs much better than the SFR. It can be seen that 100 % overall UL cell

coverage can be achieved by using the EFFR scheme with any power ratio selection.


(c) Mean UL CCUs coverage percentage (d) Mean UL CEUs coverage percentage

Figure D.12: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R under NLOS condition, having the same environment as in Figure D.7: (a) mean overall UL cell

coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding

mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in

UL traffic.

D.4 Conclusion 201

Finally, the average UL cell spectral efficiency of all investigated frequency reuse

schemes as a function of power ratio Phigh/Plow under NLOS condition is exhibited in

Figure D.13. Like in DL, the Reuse-3 can always reap the best cell spectral efficiency.

The EFFR gains the second best place and immense improve the performance

compared to the Reuse-1 and IFR scheme. In contrast, the SFR scheme also

outperforms the Reuse-1, yet its contribution cannot rival the EFFR scheme.

D.4 Conclusion

By means of comprehensive and detailed performance evaluation presented in this

section, it can be concluded that whether under LOS or NLOS condition the EFFR

design is more attractive than the SFR design. The power ratio has a severe impact on

the performance of the SFR scheme, whereas using the EFFR scheme its performance

including CINR, throughput, cell coverage as well as spectral efficiency is superior

and relative stable under various power ratios. With an inappropriate power allocation,

by comparison, the performance of the SFR will be strongly deteriorated. Furthermore,

under both LOS and NLOS propagation, although the SFR scheme outperforms the

Reuse-1 and IFR scheme, it can never surpass the Reuse-3 scheme. With a careful

selection of power ratios, the SFR may gain benefits in overall cell throughput. But its

CEU throughput performance, its cell coverage performance as well as its cell spectral

efficiency are still far more inferior to that using the Reuse-3 scheme. In this sense, the

contribution of the SFR scheme for ICI mitigation and bettering the CEUs

performance is limited. In contrast, the EFFR scheme can provide considerably better

Figure D.13: Average UL cell spectral efficiency of five frequency reuse schemes depending on range

ratio r/R under NLOS condition, having the same environment as in Figure D.7.

Appendix D 202

overall cell throughput than with the Reuse-3 scheme, whilst it can hold its CEUs

performance and cell coverage performance similar to the Reuse-3 scheme. In terms of

the cell spectral efficiency, the EFFR scheme can also supply a similar performance to

the Reuse-3 under LOS propagation, though cannot catch up with the Reuse-3 in

NLOS case. Even so, it is still much better the Reuse-1 and the SFR scheme. As a

consequence, with the respect to the power allocation, the proposed EFFR design

always performs better than the SFR in every situation, and can provide more

flexibility and robustness than the SFR scheme. With the EFFR scheme the medium is

able to be more effectively utilized, and the performance of all UTs including both

CCUs and CEUs are advanced.

LIST OF ABBREVIATIONS

List of Abbreviations

3GPP Third Generation Partnership Project

3GPP-LTE Third Generation Partnership Project Long Term Evolution

4G Fourth Generation

ACK Acknowledgment

AFFR Adaptive Fractional Frequency Reuse

ARQ Automatic Repeat Request

AWGN Additive White Gaussian Noise

BER Bit Error Ratio

BPSK Binary Phase Shift Keying

BS Base Station

BW Bandwidth

CCI Co-Channel Interference

CCU Cell-Centre User

CEU Cell-Edge User

CID Connection Identifier

CINR Carrier-to-Interference-plus-Noise Ratio

CIR Carrier-to-Interference Ratio

CMU Cell-Middle User

CoMP Coordinated Multi-Point transmission/reception

CQI Channel Quality Information

CRC Cyclic Redundancy Check

CRU Cell-Remote User

CSI Channel State Information

DECT Digital European Cordless Telecommunications

DL Downlink

DLL Data Link Layer

EFFR Enhanced Fractional Frequency Reuse

EFFR-A Enhanced Fractional Frequency Reuse - Advanced

EFFR-B Enhanced Fractional Frequency Reuse – Beyond

ETSI European Telecommunications Standards Institute

FDD Frequency Division Duplex

FDM Frequency Division Multiplexing

FDMA Frequency Division Multiple Access

FFR Fractional Frequency Reuse

FFT Fast Fourier Transform

FH Frequency Hopping

Abbreviations 204

FRF Frequency Reuse Factor

FU Functional Unit

FUN Functional Unit Network

GPRS Fractional General Packet Radio Service

GSM Global System for Mobile communications

IAT Inter-Arrival Time

ICI Inter-Cell Interference

IDMA Interleave Division Multiple Access

IEEE Institute of Electrical and Electronic

IFR Incremental Frequency Reuse

IMT International Mobile Telecommunications

IP Internet Protocol

ISI Inter-Symbol Interference

ISO International Organization for Standardization

LDK Layer Development Kit

LLC Logical Link Control

LOS Line-of-Sight

LTE Long Term Evolution

MAC Medium Access Control

MCS Modulation and Coding Scheme

MIMO Multiple-Input Multiple-Output

MS Mobile Station

MU Multi-User

MUD Multi-User Detection

NLOS Non Line-of-Sight

OFDM Orthogonal Frequency Division Multiplexing

OFDMA Orthogonal Frequency Division Multiple Access

OpenWNS Open Wireless Network Simulator

OSI Open Systems Interconnection

PDU Protocol Data Unit

PER Packet Error Rate

PF Proportional Fair

PHY Physical Layer

QAM Quadrature Amplitude Modulation

QoS Quality of Service

QPSK Quadrature Phase Shift Keying

RISE Radio Interference Simulation Engine

RLC Radio Link Control

RNC Radio Network Controller

RRM Radio Resource Management

RUP Reuse Partitioning

SAP Service Access Point

List of Abbreviations 205

SAR Segmentation and Reassembly

SDL Specification and Description Language

SDMA Space Division Multiple Access

SDU Service Data Unit

SFH Superframe Header

SFR Soft Frequency Reuse

SPEETCL SDL Performance Evaluation Tool Class Library

STL Standard Template Library

TCP Transmission Control Protocol

TDM Time Division Multiplexing

TDMA Time Division Multiple Access

TG Task Group

UDP User Datagram Protocol

UE User Equipment

UL Uplink

UMTS Universal Mobile Telecommunication System

UT User Terminal

UTRA Universal Terrestrial Radio Access

WiMAC WiMAX MAC Layer

WiMAX Worldwide Interoperability for Microwave Access

WINNER Wireless World Initiative New Radio

WLAN Wireless Local Area Networks

WNS Wireless Network Simulator

LIST OF FIGURES

List of Figures

Figure 2.1: Illustration of a cellular system with omni-directional antennas. Each cell is served by one in the

cell-center located BS using a certain frequency bandwidth Fi. .................................................................... 11 Figure 2.2: Cellular networks using different size cell-clusters, in each of which the in the middle located cell is surrounded by 3 tiers of neighboring cells. (a) When FRF =1, the cell in the centre is interfered by in

total 36 co-channel cells in the near neighborhood; (b) When FRF =3, the cell in the centre is interfered by

12 co-channel neighboring cells; (c) When FRF =7, the cell in the centre is just interfered by 6 co-channel cells located on the relatively farther 3rd –tier. .............................................................................................. 13 Figure 2.3: OFDMA subchannel structure: (a) distributed subchannel; (b) adjacent subchannel [10]. ........ 15 Figure 3.1: Concentric zones in each cell in a system. ................................................................................. 26 Figure 3.2: Frequency partitioning example for comparing exclusive and inclusive FFR scheme in a cellular

system based on FRF =7 for cell-edge users, FRF =3 for cell-middle users and FRF =1 for cell-center users.

...................................................................................................................................................................... 27 Figure 3.3: Concept of the SFR scheme in a cellular system based on FRF =3 for CEUs and FRF =1 for

CCUs. ........................................................................................................................................................... 30 Figure 3.4: In a SFR system, less available resources for CEUs while more for CCUs result in unfairness between CCUs and CEUs, as well as lower spectrum reuse efficiency. ........................................................ 32 Figure 3.5: More co-channel interferences even at low load traffic situation with the usage of the SFR

scheme. ......................................................................................................................................................... 33 Figure 3.6: CEUs are grievously interfered by re-users in the neighboring cells since SFR is a development

design based on inclusive reuse scheme. ...................................................................................................... 33 Figure 3.7: Operation policy of the IFR scheme in a cellular system with 3 various types of neighboring cells. ............................................................................................................................................................. 34 Figure 4.1: Concept of the EFFR scheme in a cellular system based on exclusive partitioning of reuse-3

subchannels and reuse-1 subchannels in the Primary Segment, as well as interference-aware reuse on the Secondary Segment. ..................................................................................................................................... 39 Figure 4.2: Frequency assignment pattern of the EFFR-A scheme in a cellular system with interfering cells

up to the 3rd-tier, where the CCUs use reuse-1 subchannels with lower power, the CMUs use reuse-3 subchannels with moderate power, and the CRUs use reuse-9 subchannels with higher power. ................... 49 Figure 4.3: An example of the available subchannels for the cell 1 in Figure 4.2, consisting of 1 exclusive

reuse-9 subchannels (M2 = 1), 3 exclusive reuse-3 subchannels (M1 = 3) and 12 reuse-1 subchannels (N = 4). ...................................................................................................................................................................... 50 Figure 4.4: Frequency assignment pattern of the EFFR-B scheme in a cellular system with interfering cells

up to the 3rd–tier, based on FRF of 1 for CCUs and FRF of 9 for the CEUs. ................................................ 51 Figure 4.5: An example of the available subchannels for the cell 1 in Figure 4.4, consisting of 1 exclusive

reuse-9 subchannels (M2 = 1) and 21 reuse-1 subchannels (N = 7). .............................................................. 52 Figure 5.1: Cellular system with interfering cells up to 3 tiers with 5 different co-channel distance: D1, D2,

D3, D4 and D5. ............................................................................................................................................... 54 Figure 5.2: 5 different interfering cell types distributed up to 3 tiers: (a) 6 Γ interfering cells with co-channel distance of D1 locate on the 1st-tier; (b) 6 Δ and 6 Θ interfering cells, with co-channel distance of D2

List of Figures 208

and D3 respectively, located on the 2nd-tier; (c) 12 Λ and 6 Ξ interfering cells, with co-channel distance of D4

and D5 respectively, locate on the 3rd-tier...................................................................................................... 56 Figure 5.3: Interfering cells up to the 3rd-tier for a Reuse-3 cellular system: the transmission in the target cell is interfered by concurrent traffic in the 6 type-Δ cells on the 2nd-tier with co-channel distance of D2 and

6 type-Ξ cells on the 3rd-tier co-channel distance of D5, respectively. ........................................................... 59 Figure 5.4: UL traffic from a CCU in the target cell (cell-type A) is interfered by the CEUs transmissions in the 12 neighboring type-B cells, which consist of half number of the (a) 6 Γ interfering cells with co-channel

distance of D1 locate on the 1st-tier; half number of the (b) 6 Θ interfering cells with co-channel distance of

D3 located on the 2nd-tier; as well as half number of the (c) 12 Λ interfering cells with co-channel distance of D4 locate on the 3rd-tier. Besides, it is also interfered by the CCU UL transmissions of the remaining 24

type-A and C cells. ....................................................................................................................................... 60 Figure 5.5: Using the EFFR scheme, the CEUs exclusively use the reuse-3 subchannels with higher transmission power, whereas the CCUs use the reuse-1 subchannels with lower transmission power. ......... 61 Figure 5.6: CINR versus the cell radius R using C1 LOS path loss model: (a) UL CINR perceived at the central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell border. 68 Figure 5.7: CINR distribution, when a UT traverses the cell with a radius of 2800m under the C1 LOS

propagation: (a) UL CINR received by the BS; (b) DL CINR received by the UT. ...................................... 69 Figure 5.8: CINR versus the cell radius R using C1 NLOS path loss model: (a) UL CINR perceived at the

central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell border. 70 Figure 5.9: CINR distribution when a UT traverses the cell with a radius of 298m under the C1 NLOS propagation: (a) UL CINR received by the BS; (b)DL CINR received by the UT. ....................................... 70 Figure 5.10: CINR versus the cell radius R using LOS-NLOS path loss model: (a) UL CINR perceived at

the central BS while a UT as a transmitter located at the cell border; (b) DL CINR received at the cell border. ...................................................................................................................................................................... 71 Figure 5.11: CINR distribution when a UT traverses the cell with a radius of 3920m under the LOS- NLOS

propagation: (a) UL CINR received by the BS; (b) DL CINR received by the UT. ...................................... 73 Figure 5.12: Cell coverage percentage computation when the maximum reach R’ by using a certain scheme

is larger than the internally tangent circle radius R* of the hexagonal cell.................................................... 75 Figure 5.13: DL cell coverage percentage of each all studied schemes under LOS, NLOS and combined LOS-NLOS propagations as given in Table 5.6. ........................................................................................... 76 Figure 5.14: Mean cell capacity under the C1 LOS propagation, having the same environment as in

Figure 5.7: (a) mean UL cell capacity; (b) mean DL cell capacity. ............................................................... 80 Figure 5.15: Mean cell capacity under the C1 NLOS propagation, having the same environment as in

Figure 5.9: (a) mean UL cell capacity; (b) mean DL cell capacity. ............................................................... 81 Figure 5.16: Mean cell capacity under the combined LOS-NLOS propagation, having the same environment as in Figure 5.11: (a) mean UL cell capacity; (b) mean DL cell capacity. .................................................... 82 Figure 5.17: DL CINR received at cell border versus cell radius R: (a) using C1 LOS path loss model; (b)

using C1 NLOS path loss model. .................................................................................................................. 86 Figure 5.18: DL CINR distribution received by a UT: (a) when the UT traverses the cell with a radius of

1000m under C1 LOS condition; (b) when the UT traverses the cell with a radius of 220m under C1 NLOS

condition. ...................................................................................................................................................... 86 Figure 5.19: DL cell coverage percentage of all studied schemes under LOS and NLOS propagations. ..... 88 Figure 5.20: Mean DL cell capacity of all studied schemes: (a) under LOS condition; (b) under NLOS

condition. ...................................................................................................................................................... 89 Figure 5.21: Average DL cell spectral efficiency of all studied schemes: (a) under LOS condition; (b) under

NLOS condition. ........................................................................................................................................... 89

List of Figures 209

Figure 6.1: IEEE 802.16m basic frame structure for 20 MHz channel bandwidth [41]................................ 94 Figure 6.2: Different cell-specific power masks over system bandwidth for all studied approaches including

the Reuse-1 scheme, the IFR scheme, the SFR scheme, the EFFR scheme and the Reuse-3 scheme. .......... 96 Figure 6.3: Resource scheduling process ..................................................................................................... 98 Figure 6.4: Mean DL carrier signal strength perceived by different types of UTs versus offered traffic per

user under LOS condition. 25 UTs are uniformly distributed in each cell with a cell radius of 1000 m. The range ratio r/R of 0.4 for partitioning CCUs and CEUs is assumed. And for both SFR and EFFR schemes,

the power ratio of high power level to low power level is set as 3. (a) Mean DL carrier signal strength

perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs............................................ 102 Figure 6.5: Mean DL interference level perceived by different types of UTs versus offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean DL interference level

perceived by CCUs; (b) mean DL interference level perceived by CEUs. .................................................. 102 Figure 6.6: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs as a function of offered traffic per user under LOS condition, having the same environment as in Figure 6.4: (a) mean overall DL CINR values; (b) the corresponding mean DL CINR

values perceived by CCUs; and (c) the corresponding mean DL CINR values perceived by CEUs. .......... 104 Figure 6.7: Mean DL MAC throughput under LOS condition as a function of offered traffic per user in the same environment as in Figure 6.4: (a) mean overall DL cell throughput; (b) the corresponding mean CCU

DL throughput; and (c) the corresponding mean CEU DL throughput. ...................................................... 106 Figure 6.8: Cell capacity and the corresponding mean user throughput of all studied schemes in LOS DL, having the same environment as in Figure 6.4. ........................................................................................... 108 Figure 6.9: Mean DL coverage percentage of five frequency reuse schemes versus offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean overall DL cell coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean

CEUs coverage percentage in DL traffic. ................................................................................................... 109 Figure 6.10: Average DL cell spectral efficiency of five frequency reuse schemes as a function of offered traffic per user under LOS condition, having the same environment as in Figure 6.4. It can be seen that the

simulation results generated by OpenWNS match the analytical results from the Matlab simulator quite well,

refer to Figure 5.21a. .................................................................................................................................. 111 Figure 6.11: Mean UL carrier signal strength perceived at the central BS versus offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean CCU UL carrier signal

strength perceived at the central BS; (b) mean CEU UL carrier signal strength perceived at the central BS. .................................................................................................................................................................... 112 Figure 6.12: Mean UL interference level perceived at the central BS versus offered traffic per user under

LOS condition, having the same environment as in Figure 6.4: (a) mean CCU UL interference level perceived at the central BS; (b) mean CEU UL interference level perceived at the central BS. .................. 113 Figure 6.13: Mean UL CINR values perceived at the central BS as a function of offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean overall UL CINR values

perceived at the central BS; (b) mean CCU UL CINR perceived at the central BS; as well as (c) Mean CEU

UL CINR perceived at the central BS. ........................................................................................................ 114 Figure 6.14: Mean UL MAC throughput under LOS condition as a function of offered traffic per user in the same environment as in Figure 6.4: (a) mean overall UL cell throughput; (b) the corresponding mean CCU

UL throughput; and (c) the corresponding mean CEU UL throughput. ...................................................... 115 Figure 6.15: Cell capacity and the corresponding mean user throughput of all studied schemes in LOS UL, having the same environment as in Figure 6.4. ........................................................................................... 116 Figure 6.16: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per user

under LOS condition, having the same environment as in Figure 6.4: (a) mean overall UL cell coverage

List of Figures 210

percentage; (b) the corresponding mean CCUs coverage percentage; and (c) the corresponding mean CEUs

coverage percentage in UL traffic. .............................................................................................................. 117 Figure 6.17: Average UL cell spectral efficiency of five frequency reuse schemes as a function of offered traffic per user under LOS condition, having the same environment as in Figure 6.4. ................................ 118 Figure 6.18: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs as a function of offered traffic per user under NLOS condition. 25 UTs are uniformly distributed in each cell with a cell radius of 220 m. The range ratio r/R of 0.6 for partitioning CCUs and

CEUs is assumed. The weakest users are those UTs located between 176 m and 220 m away from the BS

(i.e., R'/R [0.8, 1]). And for both SFR and EFFR schemes, the power ratio of high power level to low

power level is set as 3. (a) mean overall DL CINR values; (b) the corresponding mean DL CINR values

perceived by weakest users; (c) the corresponding mean DL CINR values perceived by CCUs; as well as (d)

the corresponding mean DL CINR values perceived by CEUs. .................................................................. 119 Figure 6.19: Mean DL MAC throughput under NLOS condition as a function of offered traffic per user in

the same environment as in Figure 6.18: (a) mean overall DL cell throughput; (b) the corresponding weakest

user DL throughput; (c) the corresponding mean CCU DL throughput; as well as (d) the corresponding mean CEU DL throughput. ......................................................................................................................... 121 Figure 6.20: Cell capacity and the corresponding mean user throughput of all studied schemes in NLOS DL,

having the same environment as in Figure 6.18. ......................................................................................... 122 Figure 6.21: Mean DL coverage percentage of five frequency reuse schemes versus offered traffic per user

under NLOS condition, having the same environment as in Figure 6.18: (a) mean overall DL cell coverage

percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in DL traffic. ..... 123 Figure 6.22: Average DL cell spectral efficiency of five frequency reuse schemes as a function of offered

traffic per user under NLOS condition, having the same environment as in Figure 6.18. ........................... 124 Figure 6.23: Mean UL MAC throughput under NLOS condition as a function of offered traffic per user in

the same environment as in Figure 6.18: (a) mean overall UL cell throughput; (b) the corresponding weakest

user UL throughput; (c) the corresponding mean CCU UL throughput; as well as (d) the corresponding mean CEU UL throughput. ......................................................................................................................... 126 Figure 6.24: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per user

under NLOS condition, having the same environment as in Figure 6.18: (a) mean overall UL cell coverage percentage; (b) the corresponding weakest users coverage percentage. ...................................................... 127 Figure 6.25: Average UL cell spectral efficiency of five frequency reuse schemes as a function of offered

traffic per user under NLOS condition, having the same environment as in Figure 6.18. ........................... 127 Figure 6.26: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs depending on range ratio r/R under LOS condition, which is defined as the zone

radius for the CCUs r to the cell radius R. 25 UTs are uniformly distributed in each cell with a cell radius of 1000 m. Offered traffic per user of 500 kbps is assumed. And for both SFR and EFFR schemes, the power

ratio of high power level to low power level is set as 3. (a) Mean overall DL CINR values; (b) the

corresponding mean DL CINR values perceived by CCUs; and (c) the corresponding mean DL CINR values

perceived by CEUs. .................................................................................................................................... 130 Figure 6.27: Mean DL MAC throughput under LOS condition depending on range ratio r/R in the same

environment as in Figure 6.26: (a) mean overall DL cell throughput; (b) the corresponding mean CCU DL throughput; and (c) the corresponding mean CEU DL throughput. ............................................................ 132 Figure 6.28: Mean DL coverage percentage of five frequency reuse schemes versus range ratio r/R under

LOS condition, having the same environment as in Figure 6.26: (a) mean overall DL cell coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean

CEUs coverage percentage in DL traffic. ................................................................................................... 134 Figure 6.29: Average DL cell spectral efficiency of five frequency reuse schemes depending on range ratio r/R under LOS condition, having the same environment as in Figure 6.26. ................................................ 135

List of Figures 211

Figure 6.30: Mean DL carrier signal strength perceived by different types of UTs versus power ratio under

LOS condition, which is defined as high power level to low power level used for the SFR and the EFFR

schemes. 25 UTs are uniformly distributed in each cell with a cell radius of 1000 m. Offered traffic per user of 500 kbps and the range ratio r/R of 0.4 for partitioning CCUs and CEUs are assumed. (a) Mean DL

carrier signal strength perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs. ....... 139 Figure 6.31: Mean DL interference level perceived by different types of UTs versus power ratio of high power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the

same environment as in Figure 6.30: (a) mean DL interference level perceived by CCUs; (b) mean DL

interference level perceived by CEUs. ........................................................................................................ 139 Figure 6.32: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs depending on power ratio of high power level to low power level used for the SFR

and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean overall DL CINR values; (b) the corresponding mean DL CINR values perceived by CCUs; and (c) the

corresponding mean DL CINR values perceived by CEUs. ........................................................................ 141 Figure 6.33: Mean DL MAC throughput under LOS condition depending on power ratio of high power

level to low power level used for the SFR and the EFFR schemes in the same environment as in Figure 6.30:

(a) mean overall DL cell throughput; (b) the corresponding mean CCU DL throughput; and (c) the corresponding mean CEU DL throughput................................................................................................... 142 Figure 6.34: Mean DL coverage percentage of five frequency reuse schemes versus power ratio of high

power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) mean overall DL cell coverage percentage; (b) the corresponding

mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs coverage percentage in DL

traffic. ......................................................................................................................................................... 143 Figure 6.35: Average DL cell spectral efficiency of five frequency reuse schemes depending on power ratio

of high power level to low power level used for the SFR and the EFFR schemes under LOS condition,

having the same environment as in Figure 6.30. ......................................................................................... 144 Figure A.1: The OpenWNS Modules and their corresponding OSI Layers. .............................................. 154 Figure A.2: WiMAC Functional Unit Network representing the MAC layer at the BS [6]. ....................... 158 Figure B.1: Mean DL carrier signal strength perceived by different types of UTs versus offered traffic per user under NLOS condition. 25 UTs are uniformly distributed in each cell with a cell radius of 220 m. The

range ratio r/R of 0.6 for partitioning CCUs and CEUs is assumed. The weakest users are those UTs located

between 176 m and 220 m away from the BS (i.e., R'/R [0.8, 1]). And for both SFR and EFFR schemes,

the power ratio of high power level to low power level is set as 3. (a) Mean DL carrier signal strength perceived by CCUs; (b) Mean DL carrier signal strength perceived by CEUs; (c) Mean DL carrier signal

strength perceived by weakest users. .......................................................................................................... 162 Figure B.2: Mean DL interference level perceived by different types of UTs versus offered traffic per user under NLOS condition, having the same environment as in Figure B.1: (a) mean DL interference level

perceived by CCUs; (b) mean DL interference level perceived by CEUs; and (c) mean DL interference level

perceived by weakest users. ........................................................................................................................ 163 Figure B.3: Mean UL carrier signal strength perceived at the central BS versus offered traffic per user under

NLOS condition, having the same environment as in Figure B.1: (a) mean CCU UL carrier signal strength

perceived at the central BS; (b) mean CEU UL carrier signal strength perceived at the central BS; and (c) mean weakest user UL carrier signal strength perceived at the central BS. ................................................ 164 Figure B.4: Mean UL interference level perceived at the central BS versus offered traffic per user under

NLOS condition, having the same environment as in Figure B.1: (a) mean CCU UL interference level perceived at the central BS; (b) mean CEU UL interference level perceived at the central BS; and (c) mean

weakest user UL interference level perceived at the central BS. ................................................................. 165 Figure B.5: Mean UL CINR values perceived at the central BS as a function of offered traffic per user under NLOS condition, having the same environment as in Figure B.1: (a) mean overall UL CINR values

List of Figures 212

perceived at the central BS; (b) mean weakest user UL CINR perceived at the central BS; (c) mean CCU UL

CINR perceived at the central BS; as well as (d) mean CEU UL CINR perceived at the central BS. ......... 166 Figure B.6: Mean UL coverage percentage of five frequency reuse schemes versus offered traffic per user under NLOS condition, having the same environment as in Figure B.1: (a) the corresponding mean CCUs

coverage percentage; (b) the corresponding mean CEUs coverage percentage in UL traffic. ..................... 167 Figure C.1: Mean DL carrier signal strength perceived by different types of UTs versus range ratio r/R under LOS condition, which is defined as the zone radius for the CCUs r to the cell radius R. 25 UTs are

uniformly distributed in each cell with a cell radius of 1000 m. Offered traffic per user of 500 kbps is

assumed. And for both SFR and EFFR schemes, the power ratio of high power level to low power level is set as 3. (a) Mean DL carrier signal strength perceived by CCUs; (b) Mean DL carrier signal strength

perceived by CEUs. .................................................................................................................................... 169 Figure C.2: Mean DL interference level perceived by different types of UTs versus range ratio r/R under LOS condition, having the same environment as in Figure C.1: (a) mean DL interference level perceived by

CCUs; (b) mean DL interference level perceived by CEUs. ....................................................................... 170 Figure C.3: Mean UL carrier signal strength perceived at the central BS versus range ratio r/R under LOS

condition, having the same environment as in Figure C.1: (a) Mean CCU UL carrier signal strength

perceived at the central BS; (b) Mean CEU UL carrier signal strength perceived at the central BS. .......... 171 Figure C.4: Mean UL interference level perceived at the central BS versus range ratio r/R under LOS

condition, having the same environment as in Figure C.1: (a) mean CCU UL interference level perceived at

the central BS; (b) mean CEU UL interference level perceived at the central BS. ...................................... 171 Figure C.5: Mean UL CINR values perceived at the central BS depending on range ratio r/R under LOS

condition, having the same environment as in Figure C.1: (a) mean overall UL CINR values perceived at the

central BS; (b) the corresponding mean CCU UL CINR values perceived at the central BS; and (c) the corresponding mean CEU DL CINR values perceived at the central BS. ................................................... 172 Figure C.6: Mean UL MAC throughput under LOS condition depending on range ratio r/R in the same

environment as in Figure C.1: (a) mean overall UL cell throughput; (b) the corresponding mean CCU UL

throughput; and (c) the corresponding mean CEU UL throughput. ............................................................ 173 Figure C.7: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R under

LOS condition, having the same environment as in Figure C.1: (a) mean overall UL cell coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean

CEUs coverage percentage in UL traffic. ................................................................................................... 174 Figure C.8: Average UL cell spectral efficiency of five frequency reuse schemes depending on range ratio r/R under LOS condition, having the same environment as in Figure C.1. ................................................. 175 Figure C.9: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by

different types of UTs depending on range ratio r/R under NLOS condition. 25 users are uniformly distributed in each cell with a cell radius of 220 m. The weakest users are those users located between 176 m

and 220 m away from the BS (i.e., R'/R = [0.8 1]). Offered traffic per user of 500 kbps is assumed. And for

both SFR and EFFR schemes, the power ratio of high power level to low power level is set as 3. (a) Mean overall DL CINR values; (b) the corresponding mean DL CINR values perceived by weakest users; (c) the

corresponding mean DL CINR values perceived by CCUs; as well as (d) the corresponding mean DL CINR

values perceived by CEUs. ......................................................................................................................... 177 Figure C.10: Mean DL MAC throughput under NLOS condition depending on range ratio r/R in the same

environment as in Figure C.9: (a) mean overall DL cell throughput; (b) the corresponding weakest user DL

throughput; (c) the corresponding mean CCU DL throughput; as well as (d) the corresponding mean CEU DL throughput. ........................................................................................................................................... 178 Figure C.11: Mean DL coverage percentage of five frequency reuse schemes versus range ratio r/R under

NLOS condition, having the same environment as in Figure C.9: (a) mean overall DL cell coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs

coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in DL traffic. ..... 180

List of Figures 213

Figure C.12: Average DL cell spectral efficiency of five frequency reuse schemes depending on range ratio

r/R under NLOS condition, having the same environment as in Figure C.9. .............................................. 181 Figure C.13: Mean UL MAC throughput under NLOS condition depending on range ratio r/R in the same environment as in Figure C.9: (a) mean overall UL cell throughput; (b) the corresponding weakest user UL

throughput; (c) the corresponding mean CCU UL throughput; as well as (d) the corresponding mean CEU

UL throughput. ........................................................................................................................................... 183 Figure C.14: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R under

NLOS condition, having the same environment as in Figure C.9: (a) mean overall UL cell coverage

percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in UL traffic. ..... 184 Figure C.15: Average UL cell spectral efficiency of five frequency reuse schemes depending on range ratio

r/R under NLOS condition, having the same environment as in Figure C.9. .............................................. 185 Figure D.1: Mean UL carrier signal strength perceived at the central BS versus power ratio of high power

level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same environment as in Figure 6.30: (a) Mean CCU UL carrier signal strength perceived at the central BS; (b)

Mean CEU UL carrier signal strength perceived at the central BS. ............................................................ 188 Figure D.2: Mean UL interference level perceived at the central BS versus power ratio of high power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same

environment as in Figure 6.30: (a) mean CCU UL interference level perceived at the central BS; (b) mean

CEU UL interference level perceived at the central BS. ............................................................................. 188 Figure D.3: Mean UL CINR values perceived at the central BS depending on power ratio of high power

level to low power level used for the SFR and the EFFR schemes under LOS condition, having the same

environment as in Figure 6.30: (a) mean overall UL CINR values perceived at the central BS; (b) the corresponding mean CCU UL CINR values perceived at the central BS; and (c) the corresponding mean

CEU DL CINR values perceived at the central BS. .................................................................................... 189 Figure D.4: Mean UL MAC throughput under LOS condition depending on power ratio of high power level

to low power level used for the SFR and the EFFR schemes in the same environment as in Figure 6.30: (a)

mean overall UL cell throughput; (b) the corresponding mean CCU UL throughput; and (c) the

corresponding mean CEU UL throughput................................................................................................... 190 Figure D.5: Mean UL coverage percentage of five frequency reuse schemes versus power ratio of high

power level to low power level used for the SFR and the EFFR schemes under LOS condition, having the

same environment as in Figure 6.30: (a) mean overall UL cell coverage percentage; (b) the corresponding mean CCUs coverage percentage; as well as (c) the corresponding mean CEUs coverage percentage in UL

traffic. ......................................................................................................................................................... 191 Figure D.6: Average UL cell spectral efficiency of five frequency reuse schemes depending on power ratio of high power level to low power level used for the SFR and the EFFR schemes under LOS condition,

having the same environment as in Figure 6.30. ......................................................................................... 192 Figure D.7: Mean overall DL CINR values and the corresponding mean DL CINR values perceived by different types of UTs depending on power ratio of high power level to low power level used for the SFR

and the EFFR schemes under NLOS condition. 25 UTs are uniformly distributed in each cell with a cell

radius of 220 m. Offered traffic per user of 500 kbps is assumed. The range ratio r/R of 0.6 for partitioning CCUs and CEUs is assumed, and the weakest users are those UTs located between 176 m and 220 m away

from the BS (i.e., R'/R [0.8, 1]). (a) Mean overall DL CINR values; (b) the corresponding mean DL

CINR values perceived by weakest users; (c) the corresponding mean DL CINR values perceived by CCUs;

as well as (d) the corresponding mean DL CINR values perceived by CEUs. ............................................ 194 Figure D.8: Mean DL MAC throughput under NLOS condition depending on power ratio of high power

level to low power level used for the SFR and the EFFR schemes in the same environment as in Figure D.7:

(a) mean overall DL cell throughput; (b) the corresponding weakest user DL throughput; (c) the corresponding mean CCU DL throughput; as well as (d) the corresponding mean CEU DL throughput. ... 195

List of Figures 214

Figure D.9: Mean DL coverage percentage of five frequency reuse schemes versus power ratio of high

power level to low power level used for the SFR and the EFFR schemes under NLOS condition, having the

same environment as in Figure D.7: (a) mean overall DL cell coverage percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs coverage percentage; as well as (d)

the corresponding mean CEUs coverage percentage in DL traffic. ............................................................. 196 Figure D.10: Average DL cell spectral efficiency of five frequency reuse schemes depending on power ratio of high power level to low power level used for the SFR and the EFFR schemes under NLOS condition,

having the same environment as in Figure D.7. .......................................................................................... 197 Figure D.11: Mean UL MAC throughput under NLOS condition depending on range ratio r/R in the same environment as in Figure D.7: (a) mean overall UL cell throughput; (b) the corresponding weakest user UL

throughput; (c) the corresponding mean CCU UL throughput; as well as (d) the corresponding mean CEU

UL throughput. ........................................................................................................................................... 199 Figure D.12: Mean UL coverage percentage of five frequency reuse schemes versus range ratio r/R under

NLOS condition, having the same environment as in Figure D.7: (a) mean overall UL cell coverage

percentage; (b) the corresponding weakest users coverage percentage; (c) the corresponding mean CCUs

coverage percentage; as well as (d) the corresponding mean CEUs coverage percentage in UL traffic. ..... 200 Figure D.13: Average UL cell spectral efficiency of five frequency reuse schemes depending on range ratio r/R under NLOS condition, having the same environment as in Figure D.7. .............................................. 201

LIST OF TABLES

List of Tables

Table 3.1 Comparison between exclusive and inclusive FFR schemes. ........................................................ 28 Table 5.1: Relations between Dj, R and cell type for different tiers. ............................................................. 55 Table 5.2: Transmission power applied in studied schemes. ......................................................................... 66 Table 5.3: Assumptions for evaluation ......................................................................................................... 67 Table 5.4: Maximum cell radius under LOS, NLOS and combined LOS-NLOS propagations (PBS_max = 2 W,

PUT_max = 200 mW) ........................................................................................................................................ 72 Table 5.5: Optimal range definition for different type of users for RUP schemes under LOS, NLOS and

combined LOS-NLOS propagations (PBS_max = 2 W, PUT_max = 200 mW, RLOS = 2800 m, RNLOS = 298 m, RLOS-

NLOS = 3920 m) ............................................................................................................................................... 72 Table 5.6: Maximal reach and coverage percentage of each studied schemes under LOS, NLOS and

combined LOS-NLOS propagations (PBS_max = 2 W, PUT_max = 200 mW, RLOS = 2800 m, RNLOS = 298 m, RLOS-

NLOS = 3920 m) ............................................................................................................................................... 74 Table 5.7: PHY modes and corresponding subchannel throughput ............................................................... 77 Table 5.8: Mean cell capacity under LOS, NLOS and combined LOS-NLOS propagations (RLOS = 2800 m,

RNLOS = 298 m, RLOS-NLOS = 3920 m) ............................................................................................................... 83 Table 5.9: Area spectral efficiency under LOS, NLOS and combined LOS-NLOS propagations (RLOS = 2800

m, RNLOS = 298 m, RLOS-NLOS = 3920 m) .......................................................................................................... 84 Table 6.1: Switching thresholds and PHY data rates per subchannel for Modulation and Coding Schemes

(PHY modes) ................................................................................................................................................ 93 Table 6.2: Available bandwidth for the cells using the EFFR scheme with different M to N combination

compared to the SFR, IFR and two static Reuse schemes. ............................................................................ 95 Table 6.3: Simulation parameters and values assumed ................................................................................. 99 Table 6.4: The number of different types of UTs with diverse range ratio definitions when 25 UTs are

uniformly distributed in each cell. .............................................................................................................. 129 Table 6.5: Transmission power applied for CCUs on each subchannel in the RUP schemes for both DL and

UL with varying power ratios. .................................................................................................................... 137 Table 6.6: Transmission power applied for CEUs on each subchannel in the RUP schemes for both DL and UL with varying power ratios. .................................................................................................................... 138

BIBLIOGRAPHY

Bibliography

[1] IEEE 802.16m. Evaluation Methodology (EMD), IEEE 802.16m-09-0004r5. 2009.

[2] IEEE 802.16m. System Requirements Document (SDD), IEEE 802.16m-09-0002r10. January 2010.

[3] IEEE P802.16/D4. Advanced Air Interface. February 2010.

[4] 3GPP TS 25.814. Physical layer aspects for evolved Universal Terrestrial Radio Access (UTRA)

(Release 7). 3rd Generation Partnership Project, June 2006.

[5] 3GPP TS 36.213 V8.2.0. Evolved Universal Terrestrial Radio Access (E-UTRA); Physical layer

procedures (Release 8). 3rd Generation Partnership Project, March 2008.

[6] C. Hoymann. IEEE 802.16 Metropolitan Area Network with SDMA Enhancement. PhD thesis,

Department of Communication Networks (ComNets), Faculty 6, RWTH Aachen University, July 2008.

[7] M. Necker. A Novel Algorithm for Distributed Dynamic Interference Coordination in Cellular

OFDMA Networks. PhD thesis, Institute of Communication Networks and Computer Engineering,

Faculty 5, University of Stuttgart, February 2009.

[8] B. Walke. Mobile Radio Networks: Networking, Protocols and Traffic Performance. John Wiley &

Sons, December 2001.

[9] H. L. Bertoni. Radio Propagation for Modern Wireless Systems. Prentice Hall PTR, 2000.

[10] M. Einhaus. Dynamic Resource Allocation in OFDMA Systems. PhD thesis, Department of

Communication Networks (ComNets), Faculty 6, RWTH Aachen University, February 2009.

[11] P. Hoeher, S. Badri-Hoeher, W. Xu, and C. Krakowski. Single-antenna cochannel interference

cancellation for TDMA cellular radio systems. IEEE Wireless Communications, 12(2):30–37, April

2005.

[12] J. Andrews, W. Choi, and R. Heath. Overcoming Interference in Spatial Multiplexing MIMO Cellular Networks. IEEE Wireless Communications Magazine, 14(6): 95-104, December 2007.

[13] G. Boudreau, J. Panicker, N. Guo, R. Chang, N. Wang, and S. Vrzic. Interference Coordination and

Cancellation for 4G Networks. IEEE Communications Magazine, 47(4):74-81, April 2009.

[14] 3GPP TSG RAN WG1 Ad Hoc on LTE, R1-050608. Inter-cell interference mitigation based on IDMA.

Technical document, RITT, Sophia Antipolis, France, June 2005.

[15] 3GPP TSG RAN WG1 #46, R1-062303. Frequency hopping in uplink localized transmission. LG Electronics, August 28th – September 1st, 2006.

[16] 3GPP TSG RAN WG1 #46, R1-062497. Link Performance of Frequency Hopping in LTE Uplink

Localized Transmission. Huawei, October 9th – 13th, 2006.

[17] N. Himayat, S. Talwar, A. Rao, and R. Soni. Interference Management for 4G Cellular Standards.

IEEE Communications Magazine , 48(8): 86-92, August 2010.

[18] M. Sternad, T. Ottosson, A. Ahlén, and A. Svensson. Attaining both Coverage and High Spectral Efficiency with Adaptive OFDM Downlinks. In Proc IEEE 58th Vehicular Technology Conference

2003 (VTC 2003- Fall), volume 4, pages 2486-2490, October 2003.

[19] E. Oh, M. Cho, S. Han, C. Woo, and D. Hong. Performance Analysis of Reuse-Partitioning-Based

Subchannelized OFDMA Uplink Systems in Multicell Environments. IEEE Transactions on Vehicular

Technology, 57(4):2617–2621, July 2008.

218 Bibliography

[20] D. Liang, S. Zhu, W. Liu, and W. Wang. A Frequency Reuse Partitioning Scheme with Successive

Interference Cancellation for OFDMA uplink Transmission. In Proc. of the 20th IEEE International

Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2009), pages 1362–1366, Tokyo, Japan, September 2009.

[21] C. He, F. Liu, H. Yang, C. Chen, H. Sun, W. May, and J. Zhang. Co-channel Interference Mitigation in

MIMO-OFDM System. In Proc. of the International Conference on Wireless Communication, Networking and Mobile Computing 2007 (WiCom 2007), pages 204–208, Shanghai, China, September

2007.

[22] M. Assaad. Optimal Fractional Frequency Reuse (FFR) in Multicellular OFDMA System. In Proc IEEE 68th Vehicular Technology Conference 2008 (VTC 2008- Fall), September 2008.

[23] I. Katzela, and M. Naghshineh. Channel Assignment Schemes for Cellular Mobile Telecommunication

Systems: A Comprehensive Survey. IEEE Personal Communications, 3(3):10–31, June 1996.

[24] 3GPP TS 36.211 V1.2.0. Technical Specification Group Radio Access Network, Physical channels and

modulation (release 8). June 2007.

[25] 3GPP TSG RAN WG1#42 R1-050507. Soft frequency reuse scheme for UTRAN LTE. Technical

document, Huawei, Athens, Greece, May 2005.

[26] M. Bohge, J. Gross, and A. Wolisz. Optimal Power Masking in Soft Frequency Reuse based OFDMA Networks. In Proc. Of the 15th European Wireless Conference 2009 (EW 2009), pages 162-166,

Aalborg, Denmark, May 2009.

[27] Y. Xiang, J. Luo, and C. Hartmann. Inter-cell Interference Mitigation through Flexible Resource Reuse in OFDMA based Communication Networks. In Proc. of the 13th European Wireless Conference 2007

(EW 2007), pages 1-7, Paris, France, April 2007.

[28] IEEE 802.16m-08/702r3. Adaptive Frequency Reuse in IEEE 802.16m. IEEE 802.16 Broadband Wireless Access Working Group, Technical Report, July 2008.

[29] L. Jorgušeski, and R. Prasad. Downlink Resource Allocation in Beyond 3G OFDMA Cellular Systems.

In Proc. of the 18th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2007), pages 1–5, Athens, Greece, September 2007.

[30] K. T. Kim, and S. K. Oh. An Incremental Frequency Reuse Scheme for an OFDMA Cellular System

and Its Performance. In Proc. of the 67th IEEE Vehicular Technology Conference (VTC-Spring 08), pages 1504–1508, Singapore, May 2008.

[31] S. K. Oh, and K. T. Kim. Coordinated Inter-Cell Resource Allocation Approach for Better Frequency

Reuse in OFDMA Cellular Systems. In Proc. of the 2nd international conference on Ubiquitous information management and communication (ICUIMC 2008), pages 169–174, Suwon, Korea, January

2008.

[32] K. T. Kim, and S. K. Oh. A Universal Frequency Reuse System in a Mobile Cellular Environment. In Proc. 65th IEEE Vehicular Technology Conference (VTC-Spring 07), pages 2855–2859, Dublin,

Ireland, April 2007.

[33] 3GPP TSG RAN WG1#49bis R1-072974. Downlink Interference Coordination. Technical document,

Nokia & Nokia Siemens Networks, Orlando, FL, USA, June 2007.

[34] A. Simonsson. Frequency Reuse and Intercell Interference Co-ordination in E-UTRA. In Proc. 65th

IEEE Vehicular Technology Conference (VTC-Spring 2007), pages 3091–3095, Dublin, Ireland, April 2007.

[35] IST-WINNER II, IST-4-027756, D5.4. Final Report on Link Level and System Level Channel Models.

Deliverable. September 2006.

[36] M. Schinnenburg, F. Debus, A. Otyakmaz, L. Berlemann, and R. Pabst. A Framework for

Reconfigurable Functions of a Multi-Mode Protocol Layer. In Proc. of SDR Forum 2005, pages 1-6,

Los Angeles, U.S., November 2005. URL http://www.comnets.rwth-aachen.de.

Bibliography 219

[37] Open Wireless Network Simulator (OpenWNS), Web Page, http://www.openwns.org (Retrieved

2009-10-15).

[38] D. Bültmann, M. Mühleisen, K. Klagges, and M. Schinnenburg. OpenWNS – open Wireless Network Simulator. In Proc. of the 15th European Wireless Conference 2009 (EW 2009), pages 205-210,

Aalborg, Denmark, May 2009.

[39] Z. Xie, and B. Walke. Enhanced Fractional Frequency Reuse to Increase Capacity of OFDMA Systems. In Proc. of the 3rd International on New Technologies, Mobility and Securitye 2009 (NTMS

2009), pages 1-5, Cairo, Egypt, December 2010.

[40] Z. Xie, and B. Walke. Resource Allocation and Reuse for Inter-Cell Interference Mitigation in OFDMA based Communication Networks. In Proc. of the 5th Annual International Wireless Internet

Conference 2010 (WICON 2010), pages 1-6, Singapore, March 2010.

[41] IEEE 802.16m-08/003r7. IEEE 802.16m system description document [draft]. IEEE 802.16 Broadband Wireless Access Working Group, Technical Report, July 2009.

[42] IEEE 802.16-2004. IEEE Standard for Local and metropolitan area networks, Part 16: Air Interface for Fixed Broadband Wireless Access Systems, October 2004.

[43] J. Habetha, and J. Wiegert. Network Capacity Optimization, Part I: Cellular Radio Networks. In Proc.

of the 10th Symposium on Signal Theory, pages 125 – 132, Aachen, Germany, September 2001.

[44] WiMAX system evaluation methodology. Version 1.0, WiMAX Forum, January 2007.

[45] C. U. Castellanos, D.L. Villa, C. Rosa, K. I. Pedersen, F. D. Calabrese, P.-H. Michaelsen, and J.

Michel. Performance of Uplink Fractional Power Control in UTRAN LTE. In Proc. of the 67th IEEE Vehicular Technology Conference (VTC-Spring 2008), pages 2517–2521, Singapore, May 2008.

[46] C. Hoymann. Analysis and performance evaluation of the OFDM-based metropolitan area network

IEEE 802.16. International Journal of Computer and Telecommunications Networking, 49(3): 341-363, October, 2005.

[47] Z. Xie, and B. Walke. Performance Analysis of Reuse Partitioning Techniques in OFDMA based

Cellular Radio Networks. In Proc. of the 17th International Conference on Telecommunication 2010 (ICT 2010), pages 272-279, Doha, Quatar, April 2010.

[48] Z. Xie, and B. Walke. Frequency Reuse Techniques for Attaining both Coverage and High Spectral

Efficiency in OFDMA Cellular Systems. In Proc. of IEEE Wireless Communications & Networking Conference 2010 (WCNC 2010), pages 1-6, Sydney, Australia, April 2010.

[49] 3GPP TR 36.814 V2.0.1. Evolved Universal Terrestrial Radio Access (E-UTRA); Further

advancements for E-UTRA physical layer aspects. 3rd Generation Partnership Project, March 2010.

http://www.openwns.org/

CURRICULUM VITAE

Curriculum Vitae

Name: Zheng Xie

Date of birth: January 10, 1976

Place of birth: Shanghai, V.R. China

Nationality: German

09/1991 – 07/1994 No.1 High School Affiliated to East China Normal

University, Shanghai

09/1994 – 04/1997 Studies in Environmental Engineering at Tongji University

in Shanghai

08/1997 – 09/1998 National Studienkolleg for foreign students at RWTH

Aachen University

10/1998 – 09/2005 Studies in Computer Science at RWTH Aachen University

01/2006 – 02/2011 Member of research and teaching staff, Chair of

Communication Networks (ComNets), RWTH Aachen

University

Documents

PhD Thesis on HPA OFDM and PAPR