Multi-Cell Uplink Radio Resource Management: A LTE Case Study

Multi-Cell Uplink Radio Resource
Management
A LTE Case Study
Naizheng Zheng
PhD Thesis
2011,
Alborg, Denmark
Supervisors:
Professor Preben E. Mogensen, PhD
Aalborg University, Denmark
Co-supervisors:
Jeroen Wigard, PhD
Nokia Siemens Networks, Aalborg, Denmark
Klaus I. Pedersen, PhD
Istvan Z. Kovacs, PhD
Aalborg University
Department of Electronic System
Radio Access Technology Section
Niels Jernes Vej 12, Aalborg 9220, Denmark
Phone: +45 99408645
Email: nz@es.aau.dk
www.aau.dk
ISSN 6666-8888, ISBN 111-666-888
c
Copyright 2010
by Naizheng Zheng
All rights reserved. The work may not be reposted without the explicit permission
of the copyright holder.
Abstract
Long Term Evolution (LTE) is the next generation mobile broad-band network
and its standardization has been finalized by 3rd Generation Partnership Project
(3GPP) in Release 8 (Rel8). In order to ensure the long-term competitiveness
for the next decade and beyond, the study item on LTE-Advanced (LTE-A) has
been started as the next evolution step to investigate how LTE can become a
real Fourth Generation (4G) network.
This study explores the enhancement of LTE network in the Uplink (UL) direction at system level. In the earlier literature, most of the studies were focused
on the single-cell Radio Resource Management (RRM) optimization, where the
performance of multi-cell RRM is being less investigated. The inter-cell interference is the major concern in the LTE network. By exploiting the multi-cell
solution, the impact of inter-cell interference can be limited, and the overall
network performances can be further enhanced.
Antenna downtilting is an efficient way to reduce the inter-cell interference in
both UL and Downlink (DL) direction. With a certain optimal antenna downtilting angle, the received signal power is improved within its own serving cell
and the inter-cell interference to the other neighboring cells is also reduced.
However, if the antenna is downtilted too aggressively, it may result in insufficient coverage and mobility support. In this study, the mechanical antenna
downtilting is firstly investigated in the UL LTE and the interaction of antenna
downtilting together with UL Fractional Power Control (FPC) is also analyzed.
Based on the antenna downtilting study, it can be foreseen that the User
Equipment (UE)s who are close to the cell-border still suffer from the degradation of high level of inter-cell interference and the low signal quality due to the
propagation loss. To solve this problem, the Coordinated Multi-Point (CoMP)
solution is investigated. CoMP is an advanced technique for interference mitigation which is proposed in the LTE-A as one of the features to further reduce
ii
the impact of inter-cell interference. Theoretically by applying CoMP, the intercell interference could be converted into the useful signal and being completely
eliminated. For the UL LTE application, the UL CoMP in the form of both
macro diversity reception and joint reception are investigated in this study. The
joint effort of UL CoMP reception together with Interference Cancellation (IC)
technique and UL FPC are thoroughly analyzed. Besides, the multi-cell Coordinated Packet Scheduling (CPS) is also investigated in this study based on the
UL CoMP joint reception, where a simple CPS algorithm is studied for a cluster
of neighboring cells to jointly allocate the UEs served in their cells.
Handover (HO) is another effective technique to mitigate the inter-cell interference. A simple HO decision algorithm is being proposed in this study by utilizing Reference Signal Received Power (RSRP) measurement in the DL LTE.
The performance is compared with the traditional Power Budget (PBGT) algorithm, where the proposed integrator algorithm has the advantages of requiring
less parameter setup for the realistic application.
Dansk Resum
e
Long Term Evolution (LTE) er seneste version af mobilt bredb

and, hvor standardisering af frste version af 3rd Generation Partnership Project (3GPP)
netop er afsluttet- kaldet 3GPP Version 8. For at sikre en fortsat konkurrence inden for bredb
ands standard, har 3GPP startet arbejdet med LTEAdvanced (LTE-A), som i nogle tilflde ogs
a kaldes Fourth Generation (4G).
Dette studium har fokus p
a systemforbedringer og ydeevne af LTE og LTEA af linket fra mobilen til antennemasten. I den eksisterende litteratur har
tidligere studier primrt fokuseret p
a undersgelser med uafhngig radio resource management (RRM) per mobil celle, hvorfor multi-celle RRM aspekter
stadig krver nye studier. Specielt er interferensen mellem celler vigtig for LTE.
Det er f.eks muligt at begrnse den totale interferens i et netvrk ved at udnytte
brugen af multi-celle RRM.
Antenne tilting er en effektiv metode til at begrnse effekten af interferens
mellem celler. Ved at optimere antenne tilt vinklen er det muligt at optimere den
modtagne effekt fra de nskede mobiler, mens interferensen fra andre mobiler
minimeres. Men hvis antenne tilt vinklen bliver for stor, opn
as den modsatte
effekt, hvor der opn
as et tab i stedet for en gevinst. I dette studium er mekanisk
antenne tilt blevet undersgt for LTE i kombination med effektiv kontrol af
mobilers transmissionseffekt.
Baseret p
a frnvnte antenna tilt studier blev det observeret, at mobiler p
a
grnsen mellem to celler typisk oplever meget interferens. For at lse dette
problem er lsninger baseret p
a multi-celle koordineret modtagelse af signaler
fra mobiler undersgt - ogs
a kaldet CoMP med engelsk forkortelse. CoMP er
en forholdsvis avanceret teknik, som ogs
a undersges for mulig standardisering
til LTE-A. Brugen af CoMP kombineret med avancerede modtagere med aktiv
1 Translated
by Klaus I. Pedersen, Nokia Siemens Network (NSN) - Aalborg, Denmark
iv
interferens-undertrykkelse og koordinerede transmissioner i naboceller er blevet
undersgt. Algoritmer af forskellig kompleksitet er blevet udviklet og undersgt,
og resultater er genereret, som viser fordele ved brug af s
adanne teknikker.
Brug af optimeret Handover (HO) er en anden effektiv teknik, som kan bruges
til at kontrollere interferensen mellem naboceller. En simpel HO beslutningsalgoritme er foresl
aet i dette studium, baseret p
a mobil-m
alinger af modtaget
effekt fra forskellige celler. De opn
aede resultater viser, at den foresl
aede algoritme er attraktiv, da den har et mindre antal parametre, som skal konfigureres
i forhold til mange andre algoritmer i litteraturen.
Preface and
Acknowledgments
This dissertation is the result of a three years research project carried out at the
Radio Access Technology (RATE) section, Institute of Electronic Systems (ES),
Aalborg University, Denmark. The study is under the supervision and guidance
of Professor Preben E. Mogensen (Aalborg University, Denmark), Dr. Jeroen
Wigard (Nokia Siemens Networks, Aalborg, Denmark), Dr. Klaus I. Pedersen
(Nokia Siemens Networks, Aalborg, Denmark) and Dr. Istvan Z. Kovacs (Nokia
Siemens Networks, Aalborg, Denmark).
First, I would like to thank my supervisors for their advice, guidance and patience. It has been an honor for me to work with a group of supervisors who are
not only technically knowledgeable, but also very understanding when it comes
to personal issues. Every one of them has contributed significantly to this work.
Further, I would like to thank the colleagues and secretaries from both Aalborg
University and Nokia Siemens Networks Aalborg. Thanks for their inspiring
discussions, friendly assistance and collaboration. Our friendship will be marked
in my memory forever.
Of course, the current work cannot be accomplished without the strong support
and understanding from my parents, my parents in-law, my wife Qi Zhao, my
lovely daughters DanYu Zheng and XiYu Zheng. Thanks for their constant love
and affection.
Naizheng Zheng
Aalborg, Denmark, December 2010
vi
Abbreviations
Abbreviations used in the thesis are listed below for quick reference. The abbreviations are additionally defined at their first occurrence.
Acronyms
2-D
2-Dimensional
2G
2nd Generation
3-D
3-Dimensional
3G
3rd Generation
3GPP
3rd Generation Partnership Project
4G
Fourth Generation
aGW
Access Gateway
AC
Admission Control
ACK
Acknowledgement
AMC
Adaptive Modulation and Coding
AMI
Average Mutual Information
ARPU
Average Revenue per User
ARQ
Automatic Repeat ReQuest
ATB
Adaptive Transmission Bandwidth
AVI
Actual Value Interface
AWGN
Additive White Gaussian Noise
BLER
BLock Error Rate
viii
Abbreviations
BS
Base Station
CA
Carrier Aggregation
CAPEX
Capital Expenditures
CAZAC
Constant Amplitude Zero Auto-Correlation
CB
Coordinated Beamforming
CC
Chase Combining
CDMA
Code Division Multiple Access
CDF
Cumulative Density Function
CLPC
Close-Loop fractional Power Control
CN
Core Network
CoMP
Coordinated Multi-Point
CP
Cyclic Prefix
CPS
Coordinated Packet Scheduling
CRC
Cyclic Redundancy Check
CSI
Channel State Information
CWS
Combining Window Size
dB
Decibel
dBm
Decibel relative to 1 mW
DL
Downlink
DMRS
DeModulation Reference Signal
eNB
Evolved NodeB
EESM
Exponential Effective SINR Metric
EGC
Equal Gain Combining
EPS
Evolved Packet System
FDPS
Frequency-Domain Packet Scheduling
FFT
Fast Fourier Transform
FTB
Fixed Transmission Bandwidth
FPC
Fractional Power Control
GA
Gaussian Approximation
GSM
Global System for Mobile Communication
HARQ
Hybrid Automatic Repeat reQuest
HO
Handover
HOM
Handover Margin
HPBW
Half Power Beam Width
HSDPA
High-Speed Downlink Packet Access
HSPA
High-Speed Packet Access
0.0 Acronyms
HSUPA
High-Speed Uplink Packet Access
IC
Interference Cancellation
ICIC
Inter-Cell Interference Coordination
IIR
Infinite Impulse Response
IMT-A
International Mobile Telecommunication-Advanced
IoT
Interference over Thermal noise
IP
Internet Protocol
IPS
Independent Packet Scheduling
IR
Incremental Redundancy
ISD
Inter-Site Distance
ITU
International Telecommunication Union
JP
Joint Processing
KPI
Key Performance Indicator
LA
Link Adaptation
LLR
Log Likelihood Ratio
LOS
Line of Sight
LTE
Long Term Evolution
LTE-A
LTE-Advanced
ms
Millisecond
MAC
Medium Access Control
MAI
Multiple Access Interference
MCS
Modulation and Coding Scheme
MIMO
Multiple Input Multiple Output
MISO
Multiple Input Single Output
ML
Maximum-Likelihood
MME
Mobility Management Entity
MMSE
Minimum Mean Square Error
MRC
Maximal Ratio Combining
MU-MIMO
Multi-User MIMO
MUD
Multi-User Detection
NACK
Negative Acknowledgement
OFDM
Orthogonal Frequency Division Multiplexing
OFDMA
Orthogonal Frequency Division Multiple Access
OI
Overload Indicator
OLLA
Outer Loop Link Adaptation
OLPC
Open-Loop fractional Power Control
ix
Abbreviations
OPEX
Operating Expenses
PAPR
Peak-to-Average Power Ratio
PBGT
Power Budget
PC
Power Control
PDCP
Packet Data Convergence Protocol
PDU
Protocol Data Unit
PF
Proportional Fair
PHY
Physical Layer
PIC
Parallel Interference Cancellation
PL
Path Loss
PRB
Physical Resource Block
PS
Packet Scheduling
PUSCH
Physical Uplink Shared Channel
RAT
Radio Access Technology
RoF
Radio over Fiber
RS
Reference Symbol
QoS
Quality of Service
QPSK
Quadrature Phase Shift Keying
RAN
Radio Access Network
Rel8
Release 8
RLC
Radio Link Control
RN
Relaying Nodes
RR
Round Robin
RRC
Radio Resource Control
RRM
Radio Resource Management
RS
Reference Symbols
RSRP
Reference Signal Received Power
RTT
Round-Trip Time
SAE
System Architecture Evolution
SC
Selection Combining
SC-FDMA
Single-Carrier Frequency Division Multiple Access
SIC
Successive Interference Cancellation
SIMO
Single Input Multiple Output
SNR
Signal-to-Noise Ratio
SU-MIMO
Single-User MIMO
SINR
Signal-to-Interference-plus-Noise Ratio
0.0 Symbol Notations
SNR
Signal-to-Noise Ratio
SRS
Sounding Reference Signal
TDPS
Time-Domain Packet Scheduling
TTI
Transmission Time Interval
TTT
Time-to-Trigger
TU
Typical Urban
UE
User Equipment
UL
Uplink
UMTS
Universal Mobile Telecommunications System
UTRAN
Universal Terrestrial Radio Access Network
VoIP
Voice over Internet Protocol
WCDMA
Wideband Code Division Multiple Access
Symbol Notations
A
Gain of Antenna Radiation Pattern
Receive Antenna Index
Index of Serving eNB
Distance to Boresight Cell Border
Distance between BS and UE
fc
Carrier Frequency
Channel Matrix
Complex Channel Gain
hBS
Height of Base Station
hUE
Height of User Equipment
Received Interference Power
IPSD
I
Interference Spectral Density
Index of User Equipment
UE Measured PL from DL RS
Nt
Number of Transmit Antenna
Nr
Number of Receive Antenna
NHARQ
Number of HARQ Transmissions
Ni&q
Number of In-phase and Quadrature Representation
NPRB
Number of Assigned PRB to one User
Average of Received Interference Power
xi
xii
Abbreviations
Nprb
Thermal Noise of one PRB
NPSD
Noise Spectral Density
Nresolution
Number of Quantized Resolution
Nsubcarrier
Number of OFDM Sub-carrier per PRB
Nsymbol
Number of OFDM Symbol per PRB
Nu
Number of Users Transmitting on the same PRB
Ptx
UE Transmission Power
Prx
eNB Received Power
Pmax
Maximum UE Transmission Power
P0
Cell or User-specific Parameter for Fractional Power Control
Index of Sounded PRB
q-th HARQ Transmission
Set of Simultaneously Sounded PRBs or CSI Resolution
Received Signal Vector
Element of CSI Resolution
Received Power of Sounding Reference Signal
Transmit Signal
T
Tb
Acknowledged Throughput
Estimated Achievable Throughput
Averaged Acknowledge Throughput
Index of Instant Time
Index of User Equipment
Detector Weight
x-axis
y-axis
z-axis
Cell or User-specific PL Compensation Factor
CSI Forgetting Factor
PF Filter Coefficient
CSI Measurement Error
CSI
Standard Deviation of CSI Measurement Error
CLPC Commands Signaled from eNB to UE
mcs
MCS-dependent Power Offset set by eNB
Angle between Direction of Interest and Antenna Boresight
Angle Deviation from the Horizontal Plane
Interference plus Noise Vector
SINR Output
Contents
Abstract
Dansk Resum
e
iii
Preface and Acknowledgments
Abbreviations
vii
Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
Symbol Notations
xi
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 Thesis Introduction
1.1
Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
Long Term Evolution - LTE . . . . . . . . . . . . . . . . . . . . .
1.3
Further Evolution of LTE - LTE-Advanced . . . . . . . . . . . .
1.4
Interference Management Issues . . . . . . . . . . . . . . . . . . .
1.5
Study Objectives and Scope . . . . . . . . . . . . . . . . . . . . .
1.6
Scientific Methods Employed . . . . . . . . . . . . . . . . . . . .
xiv
CONTENTS
1.7
Novelty and Contributions . . . . . . . . . . . . . . . . . . . . . .
10
1.8
Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2 Radio Resource Management in Uplink LTE
15
2.1
System Architecture of LTE . . . . . . . . . . . . . . . . . . . . .
15
2.2
Signaling and Support for Uplink RRM . . . . . . . . . . . . . .
17
2.3
Hybrid Automatic Repeat Request - HARQ . . . . . . . . . . . .
19
2.4
Link Adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.5
Fractional Power Control . . . . . . . . . . . . . . . . . . . . . .
22
2.6
Dynamic Packet Scheduling . . . . . . . . . . . . . . . . . . . . .
24
2.7
Interaction of Related Uplink RRM Entities . . . . . . . . . . . .
25
2.8
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
3 Antenna Tilting in Homogeneous LTE
27
3.1
Antenna Tilting . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
3.2
Modeling of Mechanical Downtilting . . . . . . . . . . . . . . . .
30
3.3
Influence of Antenna Downtilting . . . . . . . . . . . . . . . . . .
34
3.4
Simulation Assumptions . . . . . . . . . . . . . . . . . . . . . . .
39
3.5
System-Level Evaluations . . . . . . . . . . . . . . . . . . . . . .
40
3.6
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
4 Uplink CoMP in the Form of Macro-Scopic Combining
55
4.1
Coordinated Multi-Point . . . . . . . . . . . . . . . . . . . . . . .
56
4.2
UL CoMP Study Scenario . . . . . . . . . . . . . . . . . . . . . .
58
CONTENTS
xv
4.3
Macro Diversity Reception . . . . . . . . . . . . . . . . . . . . . .
59
4.4
Ideal Interference Cancellation . . . . . . . . . . . . . . . . . . .
62
4.5
IC-based Macro Diversity with Power Control . . . . . . . . . . .
64
4.6
65
4.7
Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . .
68
4.8
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
5 Uplink CoMP Joint Reception
83
5.1
CoMP Joint Reception . . . . . . . . . . . . . . . . . . . . . . . .
84
5.2
LTE X2-Interface for CoMP Applications . . . . . . . . . . . . .
87
5.3
System Modeling of MMSE/MMSE-SIC Receiver for CoMP Joint

Reception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
5.4
LTE RS Modeling for CoMP Applications . . . . . . . . . . . . .
97
5.5
98
5.6
Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . 103
5.7
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6 Coordinated Packet Scheduling for CoMP Joint Reception
115
6.1
Multi-cell Coordinated Packet Scheduling . . . . . . . . . . . . . 116
6.2
CPS Allocation Algorithm Design
6.3
Simulation Assumptions . . . . . . . . . . . . . . . . . . . . . . . 121
6.4
Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . 122
6.5
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
. . . . . . . . . . . . . . . . . 116
xvi
CONTENTS
7 Main Conclusion and Future Work Recommendations
127
7.1
Antenna Tilting in Homogeneous UL LTE . . . . . . . . . . . . . 128
7.2
UL CoMP in the Form of Macro Diversity Reception . . . . . . . 129
7.3
UL CoMP Joint Reception and Coordinated Packet Scheduling . 130
7.4
Overall Conclusion and Future Work Recommendations . . . . . 131
A Performance of Integrator Handover Algorithm
133
A.1 DL LTE Handover . . . . . . . . . . . . . . . . . . . . . . . . . . 134

A.2 DL LTE Hard Handover Procedure . . . . . . . . . . . . . . . . . 135
A.3 Modeling of HO Measurement in DL LTE . . . . . . . . . . . . . 135
A.4 DL LTE Hard Handover Decision Algorithm . . . . . . . . . . . . 138
A.5 Simulation Assumptions . . . . . . . . . . . . . . . . . . . . . . . 139
A.6 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . 142
A.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
B System-Level Simulator Description
147
B.1 System-Level Simulator . . . . . . . . . . . . . . . . . . . . . . . 147

B.2 Key Performance Indicator - KPI . . . . . . . . . . . . . . . . . . 150
B.3 Simulation Assumptions . . . . . . . . . . . . . . . . . . . . . . . 151
Chapter
1
Thesis Introduction
The purpose of this initial chapter is to give an overview of the whole PhD
study and the thesis. In Section 1.1, an overview of 3G cellular communication
system is presented. As the main focus in this PhD study, the Long Term
Evolution (LTE) network is described in Section 1.2. The requirements for LTEAdvanced network are defined and proposed key technologies to achieve them
are discussed in Section 1.3. Interference management is one of the important
issues for further optimizing the LTE network and it is discussed in Section 1.4.
In Section 1.5, the objectives and scope of this PhD study are specified while
the employed scientific method is described in Section 1.6. The novelty and
main contributions of the PhD study are described in Section 1.7, and finally
the organization of the thesis is presented.
1.1
Preliminaries
Nowadays, more and more people become mobile subscribers. The global economy recession did not stop people from using the mobile communication services.
Until the year 2009, the number of worldwide mobile subscribers has reached
4.3 billion and it is estimated that there will be 5.8 billion mobile subscribers by
2013 globally [1]. Mobile phones have become an important part of everybodys
daily live. Voice service through the mobile phone is not the only function anymore. In recent years, more and more mobile subscribers start checking their
email, surfing the web, downloading music and even playing real-time games on
their wireless devices [2]. So there is a rapid growth in demand of broadband
wireless data service. The operator assessment in [3] has shown that data traffic
has increased to a level more than 10 times over the voice traffic in the year 2009
as shown in Figure 1.1 and the analyst forecast report shows that, because of
Thesis Introduction
Figure 1.1: Mobile Traffic Growth [3]

increasing competition and price reduction, the declines in mobile voice Average
Revenue per User (ARPU) will continue in the year 2010. However, the progress
in broadband data service will ensure that the total ARPU grows [4]. In order
to meet the requirements of serving mobile subscribers, the mobile operators
must develop their short and long term technology strategies based on new and
innovative mobile data services.
Right now in the wireless industry, there are several paths or solutions which
can lead the mobile operator to the future mobile broadband. Each mobile
operator will take one path over the other depending on their own business
strategies and timetables. But one ultimate goal has been agreed that the
new technology should be an efficient Internet Protocol (IP) wireless network
capable of supporting voice, video, messaging and data services [5]. LTE is such
a promising technology which can meet the needs of future IP-based services [6],
and more and more mobile operators have converged on the LTE technology,
as they believe that LTE will offer them and their customers the most benefits
and the best interests.
LTE evolved from the first 3rd Generation (3G) network, Universal Mobile
Telecommunications System (UMTS), which comes after the 2nd Generation
(2G) Global System for Mobile Communication (GSM) specifications and is
standardized by the 3GPP since 1998 [7]. In the first release (Release 99), a new
Radio Access Network (RAN), which is called Universal Terrestrial Radio Access
Network (UTRAN), was introduced together with a new air interface called
Wideband Code Division Multiple Access (WCDMA). WCDMA is a wideband
spread spectrum air interface that utilizes code division multiple access, and
sometimes it is used as a synonym for UMTS [7]. As shown in Figure 1.2,
1.1 Preliminaries
Figure 1.2: Evolution of 3GPP Family
WCDMA enables DL peak data rate of 2 Mbit/s and UL peak data rate of
384 kbit/s with latency of 150 ms on a common 5 MHz bandwidth. In order
to preserve the future competitiveness compared to the other technologies, the
3GPP standardization body started the evolution of WCDMA technology by
introducing the 3.5G network High-Speed Packet Access (HSPA), which includes
the DL evolution High-Speed Downlink Packet Access (HSDPA) in Release 5
and the UL evolution High-Speed Uplink Packet Access (HSUPA) in Release 6
[8]. HSDPA improves the DL peak data rate to 14.4 Mbit/s and reduces the
network latency to around 100 ms. HSUPA further enhances the UL peak data
rate to 5.7 Mbit/s and reduce another 30 ms network latency. The continuing
evolution of HSPA in Release 7, Release 8, Release 9 and beyond1 named HSPA+
or HSPA evolved, provides even higher data rate, lower latency and higher
spectral efficiency. As shown in Figure 1.2, HSPA+ in Release 9 can achieve the
DL peak data rate up to 84 Mbit/s and UL peak data rate of 23 Mbit/s on 10
MHz bandwidth. Meanwhile, the round trip time latency is reduced below 50
ms [10].
HSPA+ and LTE will probably coexist in parallel for many years. Many UMTS
or HSPA operators have decided to use the HSPA+ as an upgrade path to
the future LTE, because the HSPA+ can deliver remarkable data rate to meet
1 HSPA+ in Rel-8 reaches 42 Mbit/s by combining 2x2 Multiple Input Multiple Output
(MIMO) and high order modulation (64QAM) in 5 MHz bandwidth or by utilizing high order
modulation and multi-carrier in 10 MHz bandwidth. HSPA+ in Rel-9 combines multi-carrier
and MIMO in 10 MHz to reach 84 Mbit/s peak rates. Uplink multi-carrier double the uplink
peak data rate to 23 Mbit/s. For HSPA+ in Releases beyond Rel-9, it may expand multicarrier to 20 MHz and utilize combinations of multi-carrier and MIMO to reach peak data
rates exceeding 100 Mbit/s in the DL and 23 Mbit/s in the UL [9]
Thesis Introduction
the current needs of advanced mobile subscribers with simple, affordable, and
incremental cost to the existing HSPA network. However, when the mobile
operators reach their network capacity limits with all the available technologies,
the deployment of the LTE network is definitely required in order to provide
much higher data capacity in the future. [4][11].
1.2
Long Term Evolution - LTE
LTE was firstly introduced and specified by the 3GPP in Release 8. It enables
the mobile operators to operate network in scalable bandwidth up to 20 MHz,
i.e. 1.4, 3, 5, 10, 15 and 20 MHz [12][13]. With 20 MHz bandwidth, LTE
enables the peak data rates exceeding 300 Mbit/s (4x4 MIMO) in the downlink
and 75 Mbit/s (64 QAM) in the uplink with significantly reduced round trip
delay around 10 ms [14][15]. In order to achieve such a challenging improvement,
LTE introduces a new radio access technology together with MIMO technology
in the physical layer and a simple radio network architecture for the higher layer
[16].
Orthogonal Frequency Division Multiple Access (OFDMA) has been selected
as the radio interface in the DL LTE [17]. OFDMA can be regarded as an
extension of the OFDM to the multiuser scenarios, in which, instead of assigning
all the available sub-carriers to one user, a subset of sub-carriers is allocated by
the base station exclusively to each user in order to accommodate multiple
user transmissions simultaneously. The frequency selectivity enabled multiuser
diversity is an intrinsic advantage of OFDMA over other multiple access methods
[18]. Therefore, by applying the radio resource management in the OFDMA
systems, such as a variety of sub-carrier assignment, modulation coding scheme
selections and power allocation, it can provide the Quality of Service (QoS)
guarantees [19]. Concerning the UE transmit power efficiency, Single-Carrier
Frequency Division Multiple Access (SC-FDMA) has been selected as the radio
interface in the UL LTE. Because the overall SC-FDMA transmit signal is
a single carrier signal, its Peak-to-Average Power Ratio (PAPR) is relatively
low compared to the case of OFDMA which produces a multi-carrier signal
[19]. Besides, the SC-FDMA maintains most benefits of OFDMA and improved
coverage. However, as discussed in [20] and also specified in UL LTE, the
sub-carrier of SC-FDMA need to be allocated continuously to a single user in
order to minimize the effect of frequency offset. This constraint will be a very
challenging criteria when designing radio resource allocation schemes [19].
MIMO is one of the technologies which can provide better radio link reliability
and/or higher data rate without using extra bandwidth or transmission power
1.3 Further Evolution of LTE - LTE-Advanced
[21]. In general, MIMO schemes include diversity, multiplexing and beamforming. The diversity improves the reliability of the unpredictable wireless channel.
By transmitting or receiving the same signal multiple times in the frequency
selective channel, the signals are getting faded at the same time are very rare.
So the average received signal quality can be improved. At rich scattering environment and high Signal-to-Interference-plus-Noise Ratio (SINR) with good
channel estimation, the multiplexing can transmit independent signals in different antennas to boost the data transmission rate. If the antennas are quite
correlated, the channels in different antennas are behaving almost the same. The
beamforming can tune the antenna beam to the expected user. So the signal
quality of the expected user can be increased and interference to the other users
is reduced. Compared with the conventional Single-user MIMO for improving
of per user data rate, the Multi-user MIMO leverage multiple users as spatially
distributed resources to increase the average cell throughput. The nature of
LTE adopted OFDMA and SC-FDMA is also very well suited for the MIMO
operation because it simplifies the MIMO channel equalization in the frequency
selective environment [22]. In order to limit the feedback overhead in real application, a codebook-based MIMO is specified in the LTE Release 8 [12][13]. In
the DL direction, OFDMA-MIMO schemes, such as transmit diversity, spatial
multiplexing (SU-MIMO and MU-MIMO with 2x2 or 4x4 configuration) and
dedicated reference signal-based beamforming, are supported. Considering the
power consumptions in the mobile side, with the configuration of 1 transmit
antenna and 2 or 4 receive antennas, so far only the MU-MIMO schemes are
supported in the UL direction.
LTE comes hand in hand with System Architecture Evolution (SAE), an evolution of the Core Network (CN) towards a flat, packet only and all-IP based
architecture. As presented in Chapter 2 and shown in Figure 2.1, the flat network is composed of only two node types, the Evolved NodeB (eNB) and the
Access Gateway (aGW). The RRM functionalities are performed independently
in each eNB in a distributed manner. The eNBs are interconnected with each
other by means of the X2 interface. It is assumed that there always exist an X2
interface between the eNBs that need to communicate with each other, e.g. for
support of handover of UEs.
1.3
Further Evolution of LTE - LTE-Advanced
LTE is commonly considered as 3.9G network since it does not fully comply with
the International Mobile Telecommunication-Advanced (IMT-A) next generation mobile network requirement [23] specified by the International Telecommunication Union (ITU) [24]. In order to maintain the long-term competitiveness
Thesis Introduction
of LTE, the 3GPP is now working on further evolving the LTE towards 4G
LTE-A network. 3GPP has set its own requirements for LTE-A [25]. It aims at
reaching or even exceeding the IMT-A requirements as well as its own defined
requirements.
Peak Data Rate: 1 Gbit/s in DL and 500 Mbit/s in UL
Latency: Control Plane from Idle to Connect less than 50 ms and User
Plane less than 10 ms
Peak Spectrum Efficiency: 30 bit/s/Hz in DL and 15 bit/s/Hz in UL with
MIMO configuration 8x8 and 4x4 respectively
Average Spectrum Efficiency: 2.6 bit/s/Hz in DL and 2.0 bit/s/Hz in UL
with MIMO configuration 4x2 and 2x4 respectively
Cell-edge Spectrum Efficiency: 0.09 bit/s/Hz in DL and 0.07 bit/s/Hz in
UL with MIMO configuration 4x2 and 2x4 respectively
Mobility: Support mobility across the cellular network for various mobile
speeds up to 350km/h or perhaps even up to 500km/h depending on the
frequency band
Compatibility: Backward compatible with the Release 8 LTE, so that
both Release 8 terminals can work in an LTE-A network and an LTE-A
terminal can operate in a Release 8 LTE network [26][27].
Spectrum Allocation: Extended bandwidth support up to 100 MHz
In order to meet the above challenging targets, several potential technologies,
such as Carrier Aggregation (CA), advanced MIMO , Coordinated Multi-Point
(CoMP) and Relaying Nodes (RN), are being investigated in 3GPP as part of
the study item [26]. To reach the high peak data rate targets as shown in above,
the transmission bandwidth is being extended from the maximum 20 MHz up
to 100 MHz. Considering the backwards compatibility requirements with Rel-8
LTE, CA is being considered as the method to extend the bandwidth, where
multiple component carriers are aggregated to provide the necessary bandwidth.
LTE terminals receive/transmit on one component carrier, whereas LTE-A terminals may receive/transmit on multiple component carriers simultaneously to
reach the higher bandwidths [28]. In addition to wider bandwidth, advanced
MIMO with 8x8 antenna configuration in the DL and 4x4 in the UL allow
the peak spectral efficiency exceeding the requirement. Co-channel interference
limits the system capacity, especially the cell-edge data rate. To mitigate the
interference, CoMP is being extensively discussed within the context of LTEA. The basic idea behind CoMP is to apply tight coordination at different
1.4 Interference Management Issues
cell sites to reduce the co-channel interference floor, thereby improving the cell
edge user performance [26]. The targeted high data rates by LTE-A requires a
tighter infrastructure. The deployment of RN can improve the signal strength
and extend the cell coverage. As no wired backhaul is required, RNs provide an
attractive, simple to install and cost-efficient solution for dense cell deployments
[26]. In this PhD study, UL CoMP issues are explored. They are described and
presented in the later chapters.
1.4
Interference Management Issues
Unlike the WCDMA network, where the intra-cell interference and the near-far
effect issue are the main interest, in the LTE network the frequency domain
orthogonality ideally removes the intra-cell interference. The inter-cell interference becomes the major concern, which is typically due to the small frequency
reuse factor for obtaining higher spectrum efficiency [29], namely the reuse factor of 1 when all frequencies are utilized in every cell. The inter-cell interference
hinders the LTE network performances, especially for the users at the cell-edges
or at bad coverage locations. Also with the tendency of decreasing macro-cell
Inter-Site Distance (ISD) for the same number of UEs per cell, limiting the
inter-cell interference from each cell becomes more and more important.
In general, there are mainly three approaches which can be used for UL interference mitigation. These are interference randomization, interference cancellation,
and interference coordination [30]. Interference randomization does not really
reduce the interference, but rather randomizes the interferences for example in
the time or frequency domain and achieves the diversity gain. The interference
cancellation is used to cancel the strongest interference. The conventional interference cancellation technique is based on the advanced signal processing in the
transceiver with/without multiple antennas. The interference coordination minimizes the interference level by taking advantage of the efficient RRM techniques
to coordinate the frequency band allocation, transmission power assignment or
antenna parameter settings in the nearby cells. Besides, in the UL LTE, the
UL power control is the most important interference management technique.
The power control is changed to provid the required SINR while at the same
time controlling the inter-cell interference. The optimal operation of UL power
control is very important for achieving good UL LTE performance.
CoMP is an advanced interference mitigation technique proposed in LTE-A for
further optimizing the overall performance of LTE network. CoMP coordinate
multiple network nodes with distributed and/or centralized structures [26]. The
coordination requires control or even user data information to be exchanged
Thesis Introduction
among the cooperating nodes through dedicated communication links. The

availability of these links, together with their capacity and latency, determines
the feasible type of coordination. CoMP is a multi-cell multi-user solution which
involves the techniques such as Joint Processing (JP) and Coordinated Packet
Scheduling (CPS)/Coordinated Beamforming (CB). The JP coherent joint DL
transmission/UL reception to/from geographically separated antennas. With
CPS/CB, the decisions for packet scheduling/selection of beams in one cell also
considers interference situation in neighboring cells. So an even more dynamic
and adaptive inter-cell interference coordination can be achieved. Based on the
Release 8 LTE, specifications are required in the DL CoMP for both eNB and
UE, while less specification effort is foreseen to have support for UL CoMP
[26][28].
1.5
Study Objectives and Scope
The object of the PhD study is to investigate some potential multi-cell RRM
techniques for limiting the impact of inter-cell interference in the UL LTE and
further enhance the overall LTE network performances. As discussed in the
previous section, the interference management is an important issue in the LTE
network. In this study, the developed multi-cell RRM techniques focus on the
system level solution and their performance evaluation.
The eNB antenna downtilting is one of the conventional multi-cell solutions to
relieve the effect of inter-cell interference. By downtilting the eNB antenna, both
cell-edge and system throughput are expected to be improved with increasing
signal strength in the serving cell and decreasing received inter-cell interference
from the neighboring cells. The potential of antenna downtilting in the UL
LTE is investigated first and the optimal antenna downtilting angle need to be
identified, which is used for the later studies. Meanwhile, as stated earlier, UL
power control is an important technique to achieve the good UL LTE performance. The interaction of antenna downtilting with UL power control should
be studied in order to find the optimal power control parameter setup.
Based on the antenna downtilting study, it can be foreseen that the UEs close to
the cell-border still suffer from the degradation of high level of inter-cell interference and the low signal quality due to the propagation loss. In order to solve
this problem, the potential benefits of using CoMP techniques are investigated.
The performance of UL CoMP receptions in the form of both macro diversity
combining and joint schemes are studied respectively. With CoMP macro diversity reception, the serving and coordination cell received signals of CoMP
UEs are processed separately and macro-combined in the serving cell. Whereas
1.6 Scientific Methods Employed
for the CoMP joint reception, the serving and coordination cell received signals
are jointly processed in the serving cell. By applying the multi-cell coordinated
Packet Scheduling (PS) together with the CoMP joint reception scheme, the
performance gain of combined effort need to be evaluated compared with the
standing alone solution. As presented earlier, the CoMP technique is a newly
proposed promising candidate for efficient interference management in LTE-A
network. Several interesting issues need to be studied, such as how to utilize the
CoMP technique in the LTE network, how much gain the CoMP scheme can
achieve, and how feasible the CoMP solutions to be implemented in the future
LTE-A product are, e.g. the impact on the existing LTE backhaul requirements.
All these questions will be investigated in this PhD study.
Of course, this PhD study cannot cover all the aspects of interesting research.
A certain study delimitation has been defined. LTE/LTE-A based network in
the UL direction is the main focus in this Ph.D study. In order to make the
study realistic, the LTE framework and design guidelines are employed in the
analysis. The algorithm design and evaluation of the multi-cell RRM techniques
at system-level is the main concern.
1.6
Scientific Methods Employed
The system-level performance of LTE network depends on a large number of parameters and the complicated interaction among the system entities makes it too
complex or sometimes impossible to formulate a theoretical framework. Therefore, the computer-aided simulation approach is adopted in this PhD study.
The basic idea of computer-aided design is to use a computer model to design
networks and new features for networks. Features as well as parameter settings
can be simulated before actually implementing them in the network [31]. The
accuracy of using simulation approach depends on the network function modeling and radio environment modeling. A good modeling can be build to express
more realistic and valid LTE network with less simplifying assumptions. An
important aspect of this study is to modeling and verification of modeling assumptions. It involves work on mathematical modeling and deriving abstraction
models applied in the simulator.
A semi-static UL system-level simulator was employed in this study, where the
system models applied in the simulator take the 3GPP recommended modeling
assumptions and guidelines for LTE into account, as described in [32]. The
system models contain the detailed implementation of Link Adaptation (LA)
based on Adaptive Modulation and Coding (AMC), explicit PS together with
the Hybrid Automatic Repeat reQuest (HARQ) process, concrete fractional
10
Thesis Introduction
power control and link-to-system mapping technique suitable for SC-FDMA.

The traffic type, such as 3GPP-recommended infinite/full buffer and realistic
finite buffer, is also implemented. The simulation results presented in this PhD
thesis have been generated and analyzed through massive computer simulations
by using the developed system models and implemented system-level simulator
features during the period of the PhD study. Besides, the final results and
conclusions have also been examined with other similar or 3GPP studies for
accuracy comparison.
1.7
Novelty and Contributions
The main contribution of this PhD study is the analysis, understanding and
further improvement of UL multi-cell RRM techniques in terms of interference
management issues in the LTE network. Especially, the investigating and designing work related to the UL LTE CoMP study, which provides the contributions not only from the academics point of view but also from the industrys
interests. The corresponding evaluation work involves the conceptual design,
system modeling, software development and performance analysis. One important contribution of the study is the system-level simulator development. A
lot of time have actually spend on the modeling, implementing and testing of
features in the simulator. Several topics are addressed in this PhD study period
as presented in the following.
The first topic of the study is the investigation of mechanical antenna downtilting scheme in the UL LTE network. In the open literature, antenna tilting has
been studied a lot on the Code Division Multiple Access (CDMA)-based systems [33][34]. However, the utilization of antenna tilting depends on the applied
radio access technology. WCDMA uses soft handover and cell breathing, therefore require a different antenna downtilting strategy than the LTE network. In
this study, the network-based antenna tilting was evaluated together with the
3GPP agreed UL open-loop power control scheme. The optimal antenna downtilting angle has been investigated for different inter-site-distances ranging from
500 meters to 1732 meters. The parameters of open-loop power control were
evaluated and used as reference for the later investigations. This contribution
has been published in:
Naizheng Zheng, Per-Henrik Michaelsen, Jens Steiner, Claudio Rosa and
Jeroen Wigard, Antenna Tilt and Interaction with Open Loop Power
Control in Homogeneous Uplink LTE Networks, in Proceedings of the
IEEE International Symposium on Wireless Communication Systems, pp.
693-697, Reykjavik, Iceland, October, 2008
1.7 Novelty and Contributions
11
The second topic of the study is the UL CoMP related, where the UL CoMP in
the form of macro diversity reception was first investigated. In the existing literature, [35] had the similar study which was conducted with the 2-Dimensional (2D) antenna pattern. Stronger inter-cell interference is expected with 2-D antenna pattern which results in higher CoMP performance gain. In this study,
the performance of UL CoMP macro diversity reception was investigated with
the 3-Dimensional (3-D) antenna pattern. The mechanical antenna downtilting
is applied with the optimal angle based on the previous study. The study also
presents, by combination of interference cancellation and UL close-loop power
control schemes, the overall UL LTE network can be further optimized in both
CoMP Intra-Site and Inter-Site scenarios. This contribution has been published
in:
Naizheng Zheng, Malek Boussif, Claudio Rosa, Istvan Z. Kovacs, Klaus I.

Pedersen, Jeroen Wigard and Preben E. Mogensen, Uplink Coordinated
Multi-Point for LTE-A in the Form of Macro-Scopic Combining, in Proceedings of the IEEE Vehicular Technology Conference (VTC), pp.1-5,
Taipei, China, May, 2010
The third topic of the study evaluates the UL CoMP joint reception. Most of
the published articles of CoMP joint reception were concentrate on the theoretical research [36][37][38] and DL CoMP investigations [39][40][41]. Based on the
assumption of full network cooperation, tremendous CoMP gain has been reported in the theoretical CoMP studies [36]. However, it is practically infeasible
to cooperate network over a large scale due to the implementation challenges,
such as constrained network backhaul and imperfect channel estimations. In
this study, the performance of UL CoMP joint reception was investigated in
the CoMP scenario with limited cooperation area and compared with the UL
CoMP macro diversity reception. The requirements of LTE X2-interface for
both applications were also analyzed. The application of coordinated packet
scheduling has also been studied in the Intra-Site scenario with UL CoMP joint
reception to further optimize the overall network performance. This study is
based on the realistic Minimum Mean Square Error (MMSE)/Successive Interference Cancellation (SIC) receiver and investigated with different cooperation
scenarios. Besides, the recommendations have also given for the Inter-Site scenario from the future industrial implementation interests. The work is planned
to be submitted to:
Naizheng Zheng, Gilberto Berardinelli, Claudio Rosa, Istvan Z. Kovacs,

Klaus I. Pedersen, Jeroen Wigard and Preben E. Mogensen, The performance of UL CoMP with Joint Reception in the LTE-Advanced networks
12
Thesis Introduction
Naizheng Zheng, Claudio Rosa, Istvan Z. Kovacs, Klaus I. Pedersen, Jeroen
Wigard and Preben E. Mogensen, Joint Reception of Uplink Coordinated
Multi-Point with Coordinated Packet Scheduling
The last topic of the study is about HO issue in the DL LTE. This topic is
independent from the other UL studies because of changing research fundings.
But the handover technology itself is another multi-cell solution to combat the
inter-cell interference. Hard HO has been standardized in the DL LTE Rel8.
The traditional HO decision method has been studied a lot in the GSM-based
network [42]. LTE network requires seamless mobility services, therefore it needs
a faster decision algorithm and can be easily utilized in the future deployment.
In this study, a hard HO algorithm is proposed based on the LTE RSRP measurements. Compared with the traditional algorithm, it requires less HO setup
parameters, but provides identical overall performance in the DL LTE. The
results of this study have been published in:
Naizheng Zheng and Jeroen Wigard, On the Performance of Integrator
Handover Algorithm in LTE Networks, in Proceedings of the IEEE Vehicular Technology Conference (VTC), pp.1-5, Calgary, Canada, September,
2008
1.8
Thesis Outline
The PhD thesis is organized as follows:

Chapter 2: Radio Resource Management in Uplink LTE - This chapter
presents an overview of the LTE system architecture and general descriptions of the RRM functionalities in the UL LTE.
Chapter 3: Antenna Tilting in Homogeneous LTE - This chapter presents
the multi-cell interference mitigation technique by applying the mechanical
antenna tilting in the homogeneous UL LTE networks. The interactions of
antenna downtilting with the open-loop fractional power control scheme
have also been investigated.
Chapter 4: Uplink CoMP in the form Macro-Scopic Combining - This
chapter presents the basic UL CoMP structure and the UL CoMP reception by use of macro diversity combining. The study is based on the
Maximal Ratio Combining (MRC) receiver. With the combination of ideal
IC and close-loop FPC, the upper bound of UL CoMP in the form of macro
diversity reception can be studied.
1.8 Thesis Outline
13
Chapter 5: Joint Uplink CoMP Reception - This chapter presents the UL

CoMP with multi-cell multi-user joint detection. The evaluation results
are compared with the corresponding CoMP macro diversity scheme based
on the realistic MMSE/SIC receiver. And the requirements of LTE X2interface have also been analyzed for both applications from the future
practical application interests.
Chapter 6: Coordinated Packet Scheduling for Joint Uplink CoMP - This
chapter presents a multi-cell coordinated packet scheduling algorithm which
can further improve the performance of the UL CoMP joint reception in
the Intra-Site scenario compared with the studies conducted in Chapter
5.
Chapter 7: Main Conclusion and Future Work - The chapter provides a
summary of the overall study and discusses future research issues.
Appendix A: Performance of Integrator Handover Algorithm - This chapter evaluates the performance of a proposed hard handover algorithm in
the DL LTE based on the RSRP measurement for different handover parameters.
Appendix B: Quasi-dynamic System Level Simulator Description - This
appendix provides the detailed description of the implemented UL system
level simulator.
14
Thesis Introduction
Chapter
Radio Resource Management

in Uplink LTE
This chapter outlines the study related Radio Resource Management (RRM)
entities in the UL LTE network, where the functionality and modeling issue of
each entity is briefly described and the interactions among the entities are also
discussed.
In Section 2.1, the flat LTE network architecture together with the corresponding protocols in each stack is introduced. Following the protocol layers with
bottom up approach in eNB, in Section 2.2, the signaling used for the higher
layer UL RRM is first presented. In Section 2.3, Hybrid Automatic Repeat
reQuest (HARQ) process is described. Link Adaptation (LA) functionality
which includes Fractional Power Control (FPC), Adaptive Modulation and
Coding (AMC) and Outer Loop Link Adaptation (OLLA) is discussed in Section 2.4. In Section 2.6, the dynamic Packet Scheduling (PS) issue is presented.
Finally, in Section 2.7, the interaction works among different RRM entities are
also illustrated.
2.1
System Architecture of LTE
In order to meet the low latency constraints in LTE, 3GPP has specified a
simple, flat and IP-based network architecture as part of the SAE effort. The
new flat architecture only contains two node types [43], which are the eNB and
the Mobility Management Entity (MME)/aGW, as shown in Figure 2.1. It
reduces the number of network elements in the access path which saves the time
it takes to access the radio and core network resources. From the cost savings
16
Radio Resource Management in Uplink LTE
Figure 2.1: LTE Flat Architecture and Protocol Stacks in eNB

aspect, the LTE-SAE architecture reduces both Operating Expenses (OPEX)
and Capital Expenditures (CAPEX), which means that only the two node types
must scale in capacity in order to accommodate large increases in data volumes
[44].
As shown in Figure 2.1, the eNBs are connected to the core network over the
so-called S1-interface and the X2-interface is specified for connections between
the eNBs. The X2-interface is the key interest in this study for the CoMP application. The basic functionalities of X2-interface include error message handling,
load management and mobility support[19]. The error message handling allows
reporting of general error situations. The load management used by eNBs to
indicate resource status or counteract traffic load imbalance between neighboring cells with the aim of improving the overall system capacity. The mobility
support allows the eNB to handover a certain UE to another eNB, where forwarding of user plane data, status transfer, and UE context release function are
parts of the mobility support [45][46][47][48].
The UTRAN eNB is interfacing with the UE. As shown in Figure 2.1, it hosts
protocol layers, such as Physical Layer (PHY), Medium Access Control (MAC),
Radio Link Control (RLC), and Packet Data Convergence Protocol (PDCP) to
the user-plane. Correspondingly it also offers Radio Resource Control (RRC)
functionality to the control-plane [16]. The eNB contains many RRM entities
which include HARQ, LA, dynamic PS, Admission Control (AC), persistent PS
and HO [19]. In the following sections, the functionality and modeling issue of
the study related UL RRM entities are briefly described.
2.2 Signaling and Support for Uplink RRM
2.2
17
Signaling and Support for Uplink RRM
Buffer status reports and power headroom reports are the main signalings used
for the UL RRM, where the UE buffer status reports provide the knowledge of
UE buffer status to the eNB scheduler to make sure enough resource allocation
to the UE and the power headroom reports transfer the UE transmit power
information to the eNB for performing correct RRM decisions at eNB, e.g.
allocating correct transmission format including bandwidth and modulation and
coding scheme to the UE. The Channel State Information (CSI) is extracted
from the UL Reference Symbols (RS) and utilized by the PS and LA entities
to support the channel-aware scheduling and the AMC respectively. In this
section, CSI is presented in detail and the readers are referred to [19] for further
readings about the buffer status report and power headroom report issues.
The CSI can be seen as the SINR measurement of Sounding Reference Signal
(SRS) [49], where SRS is introduced in the UL LTE as a wider band RS typically transmitted in the last SC-FDMA symbol of a 1ms subframe. User data
transmission is not allowed in this block, which results in about 7 percent reduction in UL capacity [50]. Practically, the SRS is an optional feature which
can be turned off in a cell. Users with different transmission bandwidth can
then share this sounding channel in the frequency domain. The received SRS is
estimated in the eNB and provide information on UL channel quality. SRS can
be transmitted over a fractional or full scheduling bandwidth. By applying the
Constant Amplitude Zero Auto-Correlation (CAZAC) sequences and the UL
synchronous transmission, the orthogonality guarantee the simultaneous transmission of SRS among the users using the same transmission bandwidth without
interfering with each other. In the UL LTE, the orthogonal CAZAC sequences
only apply to the users in the same cell or intra-cell users [19]. For the real application, the SRS parameter setup, such as SRS bandwidth, period, duration
and sub-band hopping sequence will impact the accuracy of the corresponding
SINR measurements [19].
In this study, it is assumed that, in every Transmission Time Interval (TTI),
the CSI of each active user in the corresponding cell is available at the eNB over
the entire scheduling bandwidth. Assume the MRC combining method, the CSI
estimation of user i on Physical Resource Block (PRB)1 p at instant time t is
1 PRB in LTE is defined as the minimum time and frequency domain scheduling granularity which consists of 12 consecutive OFDM sub-carriers and 14 OFDM symbols. Since no
Exponential Effective SINR Metric (EESM) model is used in the simulator, the fast fading
resolution in the frequency domain is on a PRB basis and not on a sub-carrier basis.
18
modeled as [49]:
CSIi,p,t =
Nr
X
Si,a,r0 ,t
p,t
X r0 R
10 10
(Ib(i),a,r0 ,t + Nprb )
a=1
(2.1)
r 0 R
where:
Nr is the number of receive antennas at the eNB b where user i is served

R is the set of simultaneously sounded PRBs within the CSI resolution of
PRB i. The size of R is the so-called CSI resolution.
S i,a,r0 ,t is the SRS power received from user i at time t on PRB r and
antenna a.
Ib(i),a,r0 ,t is the averaged interference signal power received at eNB b, at
instant time t, on PRB r and antenna a. The CSI interference component
is calculated via the exponential averaging over a certain time window
as shown in Equation 2.2, which is due to the dynamic scheduling and
variability of the instantaneous interference conditions in the UL LTE. It
has been shown in [51] that it is beneficial for the UL channel estimation
and overall performance.
Ib(i),a,r0 ,t = I b(i),a,r0 ,t + (1 ) Ib(i),a,r0 ,t1
(2.2)
In Equation 2.2, is a system parameter that can be used to control the

averaging period of the interference used in CSI measurements.
Nprb represents the thermal noise of one PRB
p,t is the introduced CSI measurement error which is a zero mean Gaussian distribution random variable with standard deviation of CSI . The
random variables p,t and p,t+a are uncorrelated for a 6= 0. The CSI
resolution is directly related to the standard deviation of the expected
measurement errors. Based on the previous study [52], with CSI resolution of 2 to 3 PRBs, it is reasonable to have a standard deviation of the
measurement error of approximately 1 dB.
2.3 Hybrid Automatic Repeat Request - HARQ
2.3
19
Hybrid Automatic Repeat Request - HARQ
In order to combat the data transfer errors, the LTE supports two levels of
re-transmissions for providing reliability, the MAC layer HARQ and RLC layer
Automatic Repeat ReQuest (ARQ). In general, the HARQ gives the additional
information to the receiver that enables it to prevent a certain amount of errors
if the initial transmission cannot be avoided, and the ARQ is required to handle
the residual errors that are not corrected by the HARQ [19].
Figure 2.2: LTE HARQ Timing for a single Uplink Packet

The HARQ in LTE is based on the use of a stop-and-wait procedure. An
example of the UL HARQ process is shown in Figure 2.2. As it can be seen,
once the UL packet is transmitted from the UE, the eNB will decode it and
provide the feedback with either an Acknowledgement (ACK) or a Negative
Acknowledgement (NACK). If a NACK is received in the UE, a re-transmission
will be triggered, either in the form of Incremental Redundancy (IR) or Chase
Combining (CC), otherwise a new UL packet will be sent out. A so-called
synchronous HARQ re-transmission is applied in the UL LTE study [49], which
means that the re-transmission of HARQ block occurs at a pre-defined periodic
interval. By doing so, no explicit signaling is required to inform the receiver
about the re-transmission schedule. The HARQ can also be utilized in LTE
either adaptively or non-adaptively, where the adaptive means the possibility of
changing transmission parameters, e.g. resource allocation and Modulation and
Coding Scheme (MCS), in the subsequent re-transmissions [49].
As also shown in Figure 2.2, the whole HARQ process takes roughly about
8 ms, where around 3 ms is estimated for the eNB/UE processing time and
1 ms for the transmission delay [19]. In Chapter 5 UL CoMP study, it can be
foreseen that the 3 ms eNB processing time brings a big challenge for the CoMP
operation with limited X2-interface latency.
In this study, the HARQ process with ideal Chase Combining is considered, and
20
the SINR after HARQ combining is modeled as [53]:

NHARQ
SIN RNHARQ =
SIN Rq
(2.3)
q=1
where SIN RNHARQ represents the combined SINR after NHARQ transmissions
and SIN Rq denotes the SINR of the q-th transmission. In this study, the HARQ
allows a maximum of three retransmissions before discarding a transmission
block, i.e. NHARQ = 4.
2.4
Link Adaptation
To respond to the fast variation of wireless channel and maximize the spectrum
efficiency, LTE includes a collection of techniques which are referred to as Link
Adaptation (LA). It contains the adaptation mechanisms such as AMC, OLLA
and FPC [19].
2.4.1
Adaptive Modulation and Coding
The basic function of AMC is to select the most suitable MCS for transmissions
according to the changing channel environments. There are a lot of studies
related to the AMC, as shown in [54]. With better channel quality or at high
SINR region, higher order of MCS can be utilized to provide higher spectral
efficiency. As shown in Figure 2.3, with a certain pre-defined or expected BLock
Error Rate (BLER) target at the first transmission (i.e. 20% BLER) and the UE
experienced SINR value, the MCS schemes can be selected in order to maximize
the expected throughput. The AMC can be applied either on a fast or a slow
basis. In this study, the TTI-based fast AMC is used, where it has been shown
in [51] that, because the fast AMC can better explore the fast variation channel
by allocating a higher order MCS, the fast AMC has much better performance
than slow AMC in terms of average cell throughput.
2.4.2
Outer-Loop Link Adaptation
OLLA is required to compensate for the fast AMC errors, where the errors are
typically due to the CSI measurement, link adaptation delay and interference
2.4 Link Adaptation
21
Figure 2.3: SINR vs BLER
variability. These errors cause the experienced BLER at first transmission to

deviate from the predefined target [55]. In LTE, the application of OLLA is not
standardized but rather vendor specific. The advantages of introducing OLLA
have already been shown for both HSDPA [56] and LTE DL [57]. In this study,
a simple OLLA algorithm is utilized in the UL LTE [58], where by monitoring
the received ACK/NACK of each user, an offset parameter is utilized and used
as an input to the AMC to stabilize the overall LA performance as shown in
Equation 2.4.
CSIi,f inal = CSIi OLLAof f
(2.4)
It should be noted that the same offset OLLAof f is applied to all the CSI
reports of a given UE across the frequency domain. In order to avoid saturation
as a result of unexpected errors, the dynamic range of the offset parameter is
defined within a certain interval. In this study, the OLLA offset is equal to 0.5
Decibel (dB) for all the UEs and the the OLLA offset range is within [4.0, 4.0]
dB.
22
2.5
Fractional Power Control
The main role of power control in UL LTE is to limit inter-cell interference while
respecting minimum SINR requirements [19]. It has been agreed in the 3GPP
meeting to utilize the FPC scheme in the UL LTE [59] as expressed in Equation
2.5 in dBm:
Ptx = min{Pmax , P0 + 10 log10 NP RB + L + mcs + f (i )}

{z
} |
|
{z
}
Open-Loop
Component
(2.5)
Close-Loop
Component
As it can be seen from the Equation 2.5, the FPC consists of two parts, which
are the Open-Loop fractional Power Control (OLPC) component and the CloseLoop fractional Power Control (CLPC) component, where:
Ptx is the UE transmit power,
Pmax is the maximum UE transmit power,
P0 is a broadcasted cell/user-specific parameter,
NPRB is the number of assigned PRB to a certain UE,
is the cell/user-specific Path Loss (PL) compensation factor,
L is the PL of the DL RS measured in the UE,
mcs is a UE-specific parameter signaled from the upper layer RRC,
i is a user-specific correction value with a relative or absolute value
depending on the f()-function.
2.5.1
Open-Loop Fractional Power Control
If the CLPC component is not applied, the Equation 2.5 can be simplified as:
Ptx = min{Pmax , P0 + 10 log10 NP RB + L}
(2.6)
From Equation 2.6, it can be seen the UE transmission power is strongly dependent on the selected open-loop power control parameters (P0 and ) and the
propagation scenario, i.e. path loss distribution. The transmission power has an
2.5 Fractional Power Control
23
impact on the received interference level and consequently on the distribution

of the scheduled SINR, which on the other hand directly impacts the system
and user spectral efficiency. As a result the choice of parameters P0 and are
very important when trying to find a optimum operation point considering both
coverage and cell throughput performance.
Basically, the OLPC can be operated at either full or fractional compensation
of the PL. With full compensation of the PL, by applying the value equal
to 1.0 in Equation 2.6, the same SINR will be received at the eNB for all the
UEs unless the Pmax constraint takes effect. In order to fight for the inter-cell
interference in the UL LTE, fractional compensation of the PL can be utilized
by applying the value for example equal to 0.6 or 0.8 in Equation 2.6. The
scheme allows compensation part of the PL so that the UEs with higher PL will
operate with a low SINR requirement and will likely generate less interference
to the neighboring cells. Based on the previous study and presented in [29], in
an interference-limited scenario, the cell coverage improves as the interference
increases until the first cell-edge user start reaching the maximum transmit
power. Then, to an increase of P0 (and hence of interference) corresponds a
decrease in coverage. From the coverage perspective, the optimal interference
operating point is higher for lower value. Generally a lower value of improves
the cell throughput performance up to approximately =0.6 [29].
2.5.2
Close-Loop Fractional Power Control
The CLPC command can also be applied to combat the inter-cell interference
or to correct the PL measurement errors. The general understanding is that
CLPC is slow and a-periodic in the UL LTE. The 3GPP specifications allow 2
types of CLPC commands, which are:
Absolute CLPC command: the UE applies the offset based on the latest
OLPC command as reference.
Cumulative CLPC command: the UE applies the offset based on the latest
transmission power value as reference.
The application of the CLPC algorithm can be adopted with either the Intra-cell
approach or the Inter-cell strategy. For the Intra-cell approach, the close-loop
component simply adjusts the UE transmit power based on the measurements.
By utilizing the Overload Indicator (OI) through the LTE X2-interface connections, the interference based inter-cell CLPC can be conducted [60]. In this
24
study, the inter-cell CLPC has also been utilized as presented in the later CoMP
topic.
2.6
Dynamic Packet Scheduling
The goal of dynamic PS is to efficiently utilize the spectrum resources and maximize the cell capacity, while making sure that the minimum QoS requirements
for the Evolved Packet System (EPS) bears are fulfilled [19]. In reality, the
wireless channel is varying in both time and frequency scale. The dynamic PS
exploits the multi-user diversity by multiplexing the UEs in both time and frequency domain and allocating the UE with the favorable conditions on a certain
transmission resource.
Figure 2.4: LTE Dynamic Packet Scheduling

The allocation algorithm of dynamic PS can be designed in many different ways.
In this study, the allocation algorithm is divided into Time-Domain Packet
Scheduling (TDPS) and Frequency-Domain Packet Scheduling (FDPS) as shown
in Figure 2.4, where the TDPS indicates a phase of user selection, and the FDPS
indicates a phase of PRB allocation. This two step structure is beneficial because
of the low computational complexity, where the FDPS only has to consider a
subset of UEs who are selected by the TDPS according to the selection criteria
or metrics.
The selection criteria or metrics of TDPS and FDPS can be defined according
to different provisions, such as the Round Robin (RR), Proportional Fair (PF),
maximum CSI measurement or QoS requirement. With different scenarios, e.g.
traffic type, number of UEs or even cell size, the performance of the selected
metrics performs differently.
In the FDPS, the UE can be allocated with either Fixed Transmission Bandwidth
(FTB) or Adaptive Transmission Bandwidth (ATB) based on the selection criteria metrics. With FTB, the number of PRBs assigned to every UE are identical. However, for the OFDM-based LTE network, bandwidth scalability is one
2.7 Interaction of Related Uplink RRM Entities
25
of the key features for improving the spectrum efficiency and providing QoS.
The ATB exploits the bandwidth flexibility and takes, such as UE buffer status,
traffic type, cell load and UE power limitation, into consideration by assigning
different portions of bandwidths to different UEs. As presented in [49] with unbalanced cell load, in a Macro Case-1 scenario, the ATB-based PS guarantees a
high bandwidth utilization which results in a higher cell throughput than FTB
but lower outage user throughput because of increase noise level. In a Macro
Case-3 scenario, the ATB is able to provide a gain in both average cell throughput and outage user throughput depends on the power settings. It is clear that
ATB is the obvious choice for having the best scheduler. But the disadvantage
of ATB is that there are many options/combinations to evaluate for deciding
the best PRB allocation for each TTI. Therefore, for the sake of simple PS, the
FTB is considered in this study.
2.7
Interaction of Related Uplink RRM Entities
As shown in Figure 2.5, the UL RRM entities do not work alone, but interact
with each other. Especially there are a lot of interactions between the dynamic
PS and other functionalities, e.g. signaling manager entities and LA entities.
Figure 2.5: Interactions of Upink LTE RRM
Interaction of Dynamic PS with Signaling Manager Entities: The

signaling manager entities include the buffer status manager, power headroom manager, CSI manager, HARQ manager and QoS manager as shown
in Figure 2.5. Some of the interactions between the signaling manager entities (e.g. the buffer status manager and the power headroom manager)
26

and the dynamic PS have already been briefly discussed in Section 2.2.
If the dynamic PS utilizes the maximum CSI measurements as the selection criteria metrics in either TDPS or FDPS, it also has to directly
interact with the CSI manager to acquire the CSI information for each
UE per TTI on a certain PRB. For the synchronous HARQ in UL LTE,
the dynamic PS needs to apply the information from HARQ manager in
order to prioritize the pending HARQ re-transmissions over the other first
transmission UEs. If the designed dynamic PS algorithm guarantees the
QoS, the knowledge of QoS requirements should be provided from the QoS
manager.
Interaction of Dynamic PS with LA Entities: The fundamental part
of the UL RRM is the interaction between dynamic PS and LA entity. In
the UL LTE, the AMC is the link between the dynamic PS and the CSI
manager as shown in Figure 2.5. It is responsible for providing the channel
quality information to the dynamic PS for a given UE on a particular
transmission bandwidth. Once the dynamic PS has assigned a specific
UL resource to a UE, the AMC is also responsible for selecting the most
appropriate MCS to the corresponding UE.
2.8
Summary
In this chapter the state of art background knowledge of RRM entities in the
UL LTE network is introduced. The main functionalities of study related RRM
entities, such as HARQ, LA, AMC, OLLA, FPC and PS, are briefly described.
Besides, the signaling used for the UL LTE RRM and the interaction among
different RRM entities are illustrated. The issues of system-level modeling in
each entity are also presented and will be utilized in the later performance
evaluation in this study.
Chapter
3
Antenna Tilting in
Homogeneous LTE
For LTE networks, the frequency domain orthogonality ideally removes the
intra-cell interference. The inter-cell interference becomes the major concern
from an overall interference point of view, typically due to the application of
frequency reuse factor1 1 for obtaining higher spectrum efficiency. Antenna
tilting is one of the inter-cell interference reduction techniques. A well defined
antenna tilting scheme minimizes the interference in the cellular network without losing coverage. Dependent on the target, by tilting the BS antenna in an
optimal manner, the mobile radio network can be optimized for coverage enhancement, capacity improvement, interference reduction, power consumption
saving, and traffic/cell load balancing.
In this study, the mechanical antenna downtilting is being studied in the UL
LTE. The optimal angle of mechanical antenna downtilting is identified, and
the parameter settings for the Open-Loop fractional Power Control are selected.
They are used as a baseline for the later simulation studies as well as a good
guideline for the future practical LTE network applications.
In Section 3.1, the state of the art of antenna tilting is presented. The modeling
issues of mechanical antenna downtilting are presented in Section 3.2. In Section
3.3, the influence of antenna tilting to the mobile network is illustrated, and the
issues are briefly discussed together with the LTE FPC. In Section 3.4, the
simulation assumptions for the mechanical downtilting study are described, and
the simulation results are presented in Section 3.5. Finally, in Section 3.6, the
conclusions for the mechanical downtilting study are made.
1 Frequency reuse factor indicates how often the same frequency channel can be used in a
given system. Smaller frequency reuse factor can be used to limit the co-channel interference.
28
Antenna Tilting in Homogeneous LTE
Figure 3.1: Mechanical Antenna Downtilting
Figure 3.2: Electrical Antenna Downtilting
3.1
Antenna Tilting
Antenna tilting or beam tilting is defined as the process of adjusting the angle
of antenna main beam below or above the horizontal (Azimuth) plane. Positive
and negative tuning angles are also referred to as antenna down-tilting and uptilting respectively. Fundamentally, the antenna tilting can be implemented in
two ways, mechanical tilting or electrical tilting.
In general, the mechanical tilting is to adjust the physical angle of the mounted
antenna brackets. As shown in Figure 3.1, with mechanical downtilting, it
lowers the angle of the antenna main lobe below the horizontal plane on one
side, and it also raises the antenna back lobe above the horizontal plane on
the other side. The effective mechanical downtilting angle corresponds to the
physical one exactly in the main lobe direction and decreases as a function
of horizontal direction, where the antenna radiation pattern is not changed
from the antenna side lobe direction [61]. With even higher downtilting angels,
the radiation pattern of mechanical downtilted antenna is shrinked from the
boresight direction and getting wider from the sides, as shown in Figure 3.3. In
comparison with the mechanical downtilting as shown in Figure 3.2 and Figure
3.3, the electrical downtilting does not change the physical angle of an antenna.
Instead it adjusts the antenna radiation pattern by utilizing the phase shifter to
lower the antenna lobes in all the geometrical horizontal directions. Practically,
3.1 Antenna Tilting
29
Figure 3.3: Practical Antenna Radiation Pattern [62]
the purely mechanical tilting and fixed electrical tilting can be implemented
either independently or jointly.
In the open literature, antenna tilting has been studied a lot on the CDMAbased systems. Based on the simulation study in [63] the average tilting applied
in WCDMA is about 10o 12o for dense urban area and 6o 8o for suburban
area. The above studied tilting angle is generally larger than what is experienced in actual deployment because of simplified channel model applied in the
analytical studies [64]. Concerning the gains from antenna tilting there is wide
range of quoted values, ranging from ten to several hundred percent depending
on the the reference used [65][66]. Typically the practical gain from antenna tilting is smaller than the simulation studies. From [33][34], the WCDMA capacity
can be improved up to 20-30% for small macro cells with antenna downtilting.
If using network-wide tilt, the gains is in the order of 10-20%, whereas cell
based tilt optimization may provide 20-30% enhancement [67]. However, the
use of antenna tilting depends on the applied radio technology. The Multiple
Access Interference (MAI) and near-far problem are the main concern [68] for
the asynchronous transmission in the UL CDMA. Whereas, in the UL LTE,
the synchronous transmission preserves the intra-cell orthogonality and the cochannel inter-cell interference becomes the key aspect. WCDMA also uses soft
handover and therefore require a different tilting strategy than the LTE that
adopts hard handover. Besides, the cell breathing1 utilized in the CDMA systems somewhat compensates for the changes in antenna titling and consequently
requires quite large changes to observe an effect on the network [69].
1 Cell breathing is defined as the constant change in the geographical area covered by the cell
tower. When the cell becomes heavily loaded, it shrinks, and the lightly loaded neighboring
cells expand. In this way, the user traffic from the overloaded cell is redirected to neighboring
cells and the overall system is load balanced.
30
For the planning and deployment of real networks, the selection of a reasonable
mechanical downtilting angle should be emphasized in order to maximize the
overall network performances and guarantee the QoS. In this study, the mechanical antenna downtilting is being studied in UL LTE and used as a baseline
for the later simulation studies.
3.2
Modeling of Mechanical Downtilting
When evaluating the benefits of different features in system-level simulations, it

is important that the simulated cellular network represents a realistic network.
For many system-level simulators, only the 2-D antenna pattern, the horizontal
antenna pattern in the Azimuth plane, was considered [70]. The vertical antenna pattern in the Elevation plane was implicitly modeled via the maximum
antenna gain pattern. However, numerous publications [71][72] have reported
on the impact of modeling the vertical antenna pattern on the network capacity,
coverage and interference. For the investigation of mechanical downtilting in this
study, the vertical antenna pattern has a major influence on the received signal
strength. So it is very important and necessary to introduce a 3-D antenna
pattern to any advanced system-level simulator, which considers the influences
of both horizontal and vertical patterns.
3.2.1
3-D Antenna Pattern Modeling
The adopted antenna pattern models in this study are characterized by the Half
Power Beam Width (HPBW) and expressed in Equation 3.1 and 3.2 in dB [43]:
"
A () = min 12
"
A () = min 12
, A,m
3dB

3dB
2
(3.1)
2
, A,m
(3.2)
where , 180 180 is defined as the angle between the direction of interest and the antenna boresight, where the antenna boresight is specified as the
direction in which the maximum antenna gain presents, 3dB represents the 3
dB horizontal HPBW in degrees and A,m represents the maximum attenuation
of horizontal pattern. For the vertical pattern, the same structure as the horizontal pattern is used, where , 90 90 is defined as the angle deviation
3.2 Modeling of Mechanical Downtilting
31
Figure 3.4: Antenna Gain Patterns: Measured vs. Model

from the horizontal plane, 3dB represents the vertical HPBW in degrees, and
A,m represents the maximum attenuation of the vertical pattern.
In Figure 3.4, the utilized antenna pattern models are compared with the practical antenna patterns, Kathrein 742215 , which are measured at 2140 MHz
[62]. The adopted antenna pattern model better matches the practical antenna
pattern in the main lobe direction and captures the main characteristics of the
practical antenna. The horizontal HPBW 3dB is 70o (i.e. +/- 35o at -3 dB)
with back lobe gain cutoff A,m at -20 dB. The gain suppression at the theoretical cell border between sectors at 60o is -9 dB. The vertical gain pattern is a
parabolic approximation of the practical pattern. It has a vertical HPBW 3dB
around 10o (i.e. +/- 5o at -3 dB). The vertical back lobe gain is chosen to cutoff
at the first side lobe null with A,m of -11 dB.
Practically, the antenna can be described by the proposed model or even measured data along the horizontal and vertical pattern cuts1 respectively. In order to describe the antenna gain with respect to the angle and in a given
direction, a process of interpolation is needed. In the literature, different interpolation approaches have been investigated by trying to express the full-sphere
1 Practically, the measurement of 3-D antenna pattern is done by measuring a number of
2-D patterns. The 2-D pattern is referred to as a pattern cut. Pattern cuts can be obtained
by fixing azimuth pattern and varying elevation pattern or vice versa.
32
measurements as precise as possible or depicting the full-sphere gain with only a

small set of parameters. As presented in [71][72] and [73], a certain weighting of
the horizontal and vertical pattern models can improve the overall modeling accuracy. In order to make it simple and feasible for the system-level simulations,
the two patterns can simply be added with equal unity weights as expressed in
Equation 3.3 in dB:
A (, ) = A () + A ()
(3.3)
In order to restrict the extra attenuation introduced by the vertical pattern as

shown in Figure 3.4, the maximum front to back ratio of A (, ) is limited by
max [A (, ) , A,,m ], where A,,m is equal to 25 dB according to the 3GPP
agreement [43].
3.2.2
3-D Antenna Mechanical Tilting Modeling
The modeling of mechanical antenna tilting in the system-level simulator is

done by rotating the patterns in the 3-D coordinate system. Theoretically, the
rotation of the 3-D coordinate can be divided into a composition of the rotation
of angle in the horizontal plane around the Z-axis and the rotation of angle
in the vertical plane around the Y-axis. It can be expressed as:

cos () cos ()
xnew
ynew = sin ()
cos () sin ()
znew
sin () cos ()
cos ()
sin () sin ()
xorig
sin ()
0 yorig
cos ()
zorig
(3.4)
where xorig , yorig , and zorig are the coordinate axis before the rotation, and
xnew , ynew and znew are the coordinate axis after the rotation.
If assumed that there is no rotation in the horizontal plane with the angle =0,
the above Equation 3.4 can be simplified to:

xnew
cos ()
ynew =
0
znew
sin ()
xorig
0 sin ()
1
0 yorig
0 cos ()
zorig
(3.5)
According to the geometry rule, angle is measured anti-clockwise. Therefore, for
3.2 Modeling of Mechanical Downtilting
33
Figure 3.5: Model Of Mechanical Downtilting Rotation

the mechanical downtilting, which is measured as a clockwise angle, the rotation
angle should be defined as as shown in Equation 3.6:
xorig
xnew
cos () 0 sin ()
yorig
ynew = 0
1
0
sin () 0 cos ()
znew
zorig
(3.6)
Since the unit vector yorig does not change, the xorig and zorig can then be
expressed as cos (orig ) and sin (orig ), where orig is defined as positive for
downwards tilt and sometimes it is also called antenna vertical beamwidth factor. Because a certain antenna vertical beamwidth will influence the antenna
downtilting angle. The above Equation 3.6 can be simplified to:
xnew
znew

=

cos () sin ()
cos (orig )
cos ( + orig )
=
sin () cos ()
sin (orig )
sin ( + orig )
(3.7)
In a scenario, as shown in Figure 3.5, with Base Station (BS) antenna height hBS ,
UE height hUE and UE to BS distance d, the antenna mechanical downtilting
angle can be expressed as shown in Equation 3.8:

= arctan
hBS hU E
d

orig
(3.8)
In Figure 3.6, the vertical antenna gain versus distance with different antenna
downtilting angle has been shown. With downtilting angle = 0o , the max-
34
Figure 3.6: Vertical Antenna Gain versus Distance along x-axis ( = 0)
imum vertical antenna gain appears at infinite far away distance, and it is expected that the maximum vertical antenna gain value is shifted to the shorter
distances with the increase of downtilting angles.
3.3
Influence of Antenna Downtilting
As shown in Figure 3.6, by ignoring the azimuth effects, the antenna gain is
only dependent on the distance, which can be thought as an adjustment to the
distance dependency of pathloss.
In the open environment, the effects of antenna downtilting can be fairly estimated by calculating the vertical angle between the base station antenna and
the mobile antenna. However, in the urban area, the vertical angle is calculated
by using the rooftop height instead of the mobile height in order to include the
obstruction from the surrounding buildings. In [74], the prediction model by
using the average building height proves better fit to measurements with different antenna downtilting angle. As a consequence, the impact of antenna tilting
reduces and almost disappears if nearby buildings have almost equal height of
3.3 Influence of Antenna Downtilting
35
the BS antenna. In addition to this effect, the vertical antenna pattern is influenced by reflection and scattering from mounting structures, e.g. as presented in
[75], the rooftop, which is a dielectric structure with finite conductivity, is much
closer to the antenna and thus will have a significant influence on the radiation
pattern. In this study a simplified model is applied in order to demonstrate
the effect of antenna downtilting. Generally, analytical studies based on the
simplified models lead to larger antenna tilting angles than experienced as best
practice in actual deployment [76].
Considering the link budget of received signal power Prx in dBm as:
Prx = Ptx P LdB SFdB + A (, ) [dBm]
(3.9)
where Ptx is the transmit power, PL is the propagation path loss, SF is the
channel slow/shadowing fading, and A(, ) is the gain of antenna pattern.
For the vertical antenna downtilting, if the user is moving along the antenna
boresight direction, there is no variation in the horizontal antenna gain A ().
The total antenna gain variation only depends on the vertical antenna gain
A ().
The modified Okumura-Hata model is adopted by 3GPP for the macro cellular
simulation and it is also considered in this study. The model is described in
detail in [77][78] and can be expressed as:
P LdB
=
=
80 18 log10 (hBS ) + 21 log10 (fc )

+ 40 160 103 hBS log10 (d)
(3.10)
128.1 + 37.6 log10 (d) [dB]
(3.11)
where fc is the carrier frequency. And by setting fc = 2.0 GHz and hBS = 15
m, the final distance dependent PL model in dB is shown in Equation 3.11.
In Figure 3.7, the vertical antenna gain A () is added to the path propagation
model for showing the signal strength in relation to the distance away from the
BS for different antenna tilting angles. As it can be seen, with increasing downtilting angle, the A () is increased at a distance close to the BS and decreased
at far-away distances from the BS. This means that, with a certain optimal
antenna downtilting angle, the received signal power becomes stronger within
its own serving cell, and the inter-cell interference to the other neighboring cells
is also reduced.
36
Figure 3.7: Path Gain plus Antenna Gain vs. Distance along x-axis ( = 0)
In this study, a simple antenna downtilting method is proposed for the practical
LTE network application based on the simple 2-eNB scenario. As shown in
Figure 3.8, UE-1 and UE-2 are served by eNB-1 and eNB-2 respectively. UE-2
acts as an interference to the UL UE-1 reception in the eNB-1 and it is located at
for the worst interference scenario.
the boresight border with distance D = ISD
3
According to Equation 3.8, the vertical antenna gain at UE-1 and UE-2 can
then be expressed as:

GUE1 = A(arctan

GUE2 = A(arctan
hBS hUE
d
hBS hUE
D
tilt )
(3.12)
tilt )
(3.13)
And based on Equation 3.11, the pathloss from the eNB-1 to UE-1 and UE-2
can be calculated as:
PLUE1 = 128.1 + 37.6 log10
p
(hBS hUE )2 + d2
[dB]
(3.14)
3.3 Influence of Antenna Downtilting
37
Figure 3.8: Antenna Downtilting of 2-eNB Scenario
PLUE2 = 128.1 + 37.6 log10
p
(hBS hUE )2 + D2
[dB]
(3.15)
At different downtilting angle, the eNB-1 received SINR value of UE-1 with
respect to distance, d, are shown in Figure 3.9 for Macro Case-1 scenario and
with fixed BS and UE height it can be expressed as:
SINRUE1 (tilt , d)
PLUE1 (d) + GUE1 (tilt , d)

PLUE2 (D) GUE2 (tilt , D)
(3.16)
For the Macro Case-1 scenario, the boresight cell border distance D = 288 m.
As it can been seen from Figure 3.9, the SINR value around the cell border area
is optimized at 14o downtilting. According to Figure 3.6, with 6o downtilting
the main lobe of vertical antenna pattern points to the boresight cell border.
To reach the optimal 14o downtilting, the eNB antenna needs to be further
downtilted 8o , which is exactly the angle of the first notch in the vertical pattern
as shown in Figure 3.4. So by pointing the first notch of vertical pattern to the
boresight cell border, the performance of cell-edge UEs can be optimized. This
method combines the geometry of cell sector with the knowledge of antenna
pattern, which is also supported in the later system level simulation results as
shown in Figure 3.17.
The system-level investigation of the mechanical antenna downtilting in the UL
LTE will be presented in the following sections. One of the key interests of the
study is the interaction of antenna downtilting with the UL FPC. As shown in
the Equation 3.9, if no closed-loop FPC is applied for simplicity, the Ptx can
38
Figure 3.9: SINR vs. Distance along x-axis ( = 0)

be derived from the open-loop FPC algorithm as described in the Equation 2.6,
where the Path Loss of the DL RS measured in the UE is a key component to
determine the UE transmit power. By downtilting the BS antenna, the overall
PL curve will be changed as shown in the Figure 3.7, which directly influences
the open-loop FPC algorithm. Combining Equation 2.6 and Equation 3.9, the
UE transmission power Ptx can be described as:
Ptx = min{Pmax , P0 + 10 log10 NP RB + L (A ())}
(3.17)
As shown in Equation 3.17, with constant P0 , and the number of assigned PRB
value NP RB , the UE transmit power will be varied, and the network performance
will be effected. This interaction between antenna downtilting and OLPC is
investigated in Section 3.5.2.
3.4 Simulation Assumptions
3.4
39
Simulation Assumptions
The performance evaluation is done by using a multi-cell system-level simulator,

which is developed by following the guidelines presented in [32] and described in
Appendix B. The simulation assumptions and parameters related to the results
discussed in this chapter are listed in Table 3.1, where the full list of default
simulation parameters is presented in Table B.1.
Table 3.1: Simulation Assumptions of Mechanical Antenna Downtilting
Parameter
Number of UEs
Deployment Scenario
HPBW
BS Antenna Gain
Antenna Downtilt Angle
P0 and value
Max UE Transmit Power

Receiver Type
Minimum UE to BS Distance
UE Height
BS Height
Traffic Model
Assumptions
10 UEs/cell
Macro Case-1: ISD=500 m
Horizontal: 70o
Vertical: 10o
14 dBi
0o to 20o with interval of 2o
Macro case-1: P0 =-58 dBm and =0.6
250 mW
MRC
35 m
1.5 m
32 m
Full Buffer
As it can be seen, the simulations are carried out in the homogeneous scenario
where all the UEs are uniformly distributed and balance loaded with 10 UEs/cell
in the whole network area. Both the Macro case-1 and Macro case-3 scenarios
are investigated in this study, where the Macro Case-1 scenario is characterized
by a small cell radius with ISD of 500 m, and the Macro Case-3 scenario is
indicated by a large cell radius with ISD of 1732 m. The antenna pattern of
horizontal HPBW 70o and vertical HPBW 10o is utilized. All the BSs have the
same height and are downtilted with the same angle from 0o to 20o with an
interval of 2o in each simulation. The BS antenna is located at the roof top
with the height of 32 m above ground, and the height of UEs is 1.5 m. The
settings of open-loop FPC parameters are based on the previous work in [29],
where the study is only considered the 2-D antenna pattern. A full or infinite
buffer traffic model are applied in the study simulations, which means that the
UEs always have data to transmit from their buffer.
40
3.5
System-Level Evaluations
In the following, the system-level evaluation of antenna downtilting in the UL

LTE is presented. The Macro Case-1 scenario is first analyzed in detail, where
both the open-loop Fractional Power Control parameters and the effect of vertical antenna beamwidth are thoroughly investigated in the Macro Case-1 scenario. The performance of antenna downtilting in the Macro Case-3 scenarios is
also presented afterwards. In the following simulation results presented, in order
to make it easy to be read the figures, only the curves with selected downtilting
angles are presented.
3.5.1
Macro Case-1 Scenario
As shown in Figure 3.10, the Cumulative Density Function (CDF) of UE PL

distribution is presented at different downtilting angles, where the PL for each
UE contains the components of PL attenuation, slow channel fading and 3-D
antenna pattern effects. By downtilting the antenna from 0o to 12o , because
the downtilted antenna mainbeam is pointing toward its own cell, the PL attenuations are reduced for all the UEs. Since the cell-center UEs have better
bearing of the downtilted beam, as seen from 4o to 12o , they show much better
improvement than the cell-edge UEs. Further downtilting the antenna angle, as
seen from 12o to 20o , the cell starts losing its coverage, which in terms of the
UEs close to the cell-edge have increasing PL attenuations. However, there are
still a small amount of the cell-center UEs who can benefit from the downtilted
vertical beam.
The CDF of UE transmit power distribution is shown in Figure 3.11. Since only
the open-loop component of FPC is utilized at different downtilting angle, the
curve shape of UE transmit power distribution follows the UE PL distribution
as presented in Figure 3.10. Because the maximum transmit power Pmax is
applied as shown in Equation 2.6, the UE transmit power is limited at 24 dBm
(10 log10 250 mw). At 0o , about 10% of the UEs are transmitted with maximum
power. With downtilting angle to 12o , the UEs operated at saturated power are
reduced to about 5%. However, by increasing the downtilting angle to 20o , more
than 20% of the UEs are transmitted at maximum power because of the impact
of shrinking cell coverage.
The CDF of eNB received signal power from each UE is shown in Figure 3.12. It
presents the joint effect of UE PL together with UE transmit power. To compare
with the previous two figures, it has to be realized that the representation of
cell-edge UEs is now located at the lower left corner of the plot or 5% outage of
3.5 System-Level Evaluations
41
Figure 3.10: Case-1 Path Loss Distribution for fixed OLPC parameters P0 =
58 and = 0.6
Figure 3.11: Case-1 UE Transmit Power Distribution for fixed OLPC parameters
P0 = 58 and = 0.6
42
the CDF curve, where in the previous two plots the cell-edge UEs are located
at the upper right corner of the plot or 95% outage of the CDF curve.
Figure 3.12: Case-1 Received Power Distribution for fixed OLPC parameters
P0 = 58 and = 0.6
Similar to the previous UE PL distribution, by increasing the antenna downtilting angle, the average received signal power is increased until downtilting
angle at 12o . For the cell-edge UEs who typically have poor signal quality, the
average received signal power are working equally good for both 4o to 12o . But
for the cell-center UEs, their performance can still be improved without losing
the cell-edge UE performance because of better bearings of the vertical antenna
gain and the average received signal power is maximized at 12o . When the
downtilting angle is larger than 12o , the downtilting antenna beam is pointing
to the area too close to the eNB, and a huge percentage of UEs cannot benefit
from the vertical antenna gain any more, which leads to the decreased average
received signal power.
In Figure 3.13, the CDF of Interference over Thermal noise (IoT) is shown, and
the IoT is defined as [7]:
IoT =
(IPSD + NPSD )
NPSD
(3.18)
43
where IPSD is the interference spectral density, and NPSD represents the noise
spectral density. The IoT indicates the level of interference in the system assuming the noise as reference. In case there is no interference in the system, the
IoT is equal to 1.0 in linear or 0.0 dB.
As expected in Figure 3.13, by increasing the downtilting angle, the IoT strength
is decreased. To compare the non-downtilt case with downtilt angle at 12o , at
50% of the CDF, there is around 4 dB reduction of the IoT, and at 95% of the
CDF there is around 2 dB reduction of the IoT.
Figure 3.13: Case-1 Interference over Thermal Noise for fixed OLPC parameters
P0 = 58 and = 0.6
In Figure 3.14, the CDF of average UE SINR is plotted. At 50% of the CDF, for
the downtilting angle at 12o , with the joint effect of increasing the received signal
power and decreasing the received inter-cell interference strength as presented
in above, there are about 5 dB increases of the SINR compared with the non-tilt
case. For the downtilting angle at 16o , even though the average received signal
power is about 0.6 dBm less than the downtilting angle at 12o , the average IoT
strength is about 1.9 dB higher, and the SINR is still about 1.0 dB higher than
the 12o case. For the cell-edge user at 5% outage, with the 16o case, the IoT
reduction can not compensate for the decreased received signal power anymore,
which makes the overall SINR performance worse than the 12o downtilting case.
44
Figure 3.14: Case-1 Received SINR Distribution for fixed OLPC parameters
P0 = 58 and = 0.6
The CDF of average UE throughput is presented in Figure 3.15. At different

downtilting angles, the UE throughput curves have the same trend as the average
SINR curves presented in Figure 3.14. In order to improve the overall network
performances for both cell-center and cell-edge UEs, the amount of antenna
downtilting angle is always a trade-off between the cell coverage and capacity,
where the cell coverage is defined as the CDF of UE throughput at 5% level.
In Figure 3.16, the cell coverage and cell capacity versus different antenna downtilting angles are shown respectively. As it can be seen, the cell coverage is
maximized at 14o . With a further increase of the downtilting angle, the cell
coverage starts shrinking, which results in a poor performance or lower SINR
for the cell-edge UEs. For the cell capacity, it is still further improved and maximized at downtilting angle of 16o , because of the cell-center UEs better bearing
of the downtilted antenna pattern. With a further increase of the downtilting
angle above 16o , even the cell-center UEs cannot benefit much from the vertical
antenna gain any more, which makes the cell capacity decrease.
Based on the above analysis, for the Macro Case-1 scenario, the optimal enhancement by downtilting angle at 14o is the best choice to achieve increased
overall network performance.
45
Figure 3.15: Case-1 UE Throughput Distribution for fixed OLPC parameters

P0 = 58 and = 0.6
Figure 3.16: Case-1 Average Cell Throughput vs. Cell-edge Throughput with
= 0.6
46
3.5.2
Interaction of Antenna Tilting with Open-Loop fractional Power Control - Case 1
In the previous section, the settings of open-loop FPC parameters are based on
the work in [29], which only considered the 2-D antenna pattern. In this section,
the chosen of open-loop FPC parameters are evaluated under the 3-D antenna
pattern configuration.
In Figure 3.17 and 3.18, the interaction between antenna tilting and OLPC is
shown with = 0.6 and = 1.0 in terms of average cell throughput and 5%
outage cell throughput.
Figure 3.17: Case1 Average Cell Throughput vs. 5% Outage with = 0.6
As it can be seen from the = 0.6 case in Figure 3.17, different colored curve
represents the antenna downtilting angles, and variation of P0 value is shown for
each individual downtilting angle. In general, both the 5% outage and average
cell throughput is improving with the increasing P0 value until a certain optimum point, i.e. -58 dBm at 12o case. This is due to the fact that both cell-edge
and cell-center UEs can improve the received SINR by increasing of transmit
power, and the optimum point can simply be estimated from the Equation 2.6,
where by taking the 12o case as an example, with the simulation assumption of
6 PRBs and the total PL of 124 dB as shown in 95% outage of Figure 3.10, the
47
Figure 3.18: Case-1 Average Cell Throughput vs. 5% Outage with = 1.0
optimal P0 value is equal to -58 dBm (10 log10 250 (10 log10 6 + 0.6 124)).
By further increasing the P0 value over the optimal point, many cell-edge UEs
start transmitting at the maximum power which generates higher interference
to the other UEs. For the cell-center UEs in Figure 3.17, they can still benefit
slightly from the increasing received power and result in average cell throughput
gains. But for the 5% cell-edge UEs, their performance is decreased considerably
by the increased interference, i.e. at 12o case the varying of P0 value from -58
dBm to -50 dBm reduces the 5% outage throughput dramatically.
The plots with = 1.0 case are shown in Figure 3.18. As expected, both 5%
outage and average cell throughput are enhanced with increase of downtilting
angle until the optimal angle of 14o and the optimal P0 value of -106 dBm
(10 log10 250 (10 log10 6 + 1.0 124)). To compare with = 0.6, utilizing
= 1.0 maximizes the cell-edge UE throughput by compromising the cellcenter UE performances. At optimal downtilting angle of 14o with = 1.0
and P 0 = 106 dBm, the 5% outage and average cell throughput is about 5.5
Mbps and 7.8 Mbps respectively as shown in Figure 3.17, whereas for = 0.6
with P 0 = 62 dBm as shown in Figure 3.18, the 5% outage and average cell
throughput is about 3.4 Mbps and 9.0 Mbps respectively.
48
3.5.3
Macro Case-3 Scenario
The antenna downtilting scheme has also been evaluated under the Macro Case-3
scenario. Compared with the interference-limited Macro Case-1 scenario, Macro
Case-3 is more noise-limited with larger ISD. As discussed in Section 3.3, the
vertical antenna downtilting scheme is an efficient inter-cell interference reduction technique, and as presented in the previous Section 3.5.1, both the 5%
outage and average cell throughput can be enhanced with the optimal antenna
downtilting in the Macro Case-1 scenario. So how much gain can the antenna
downtilting help under the noise-limited scenario?
Figure 3.19: Case-3 UE Transmit Power Distribution for fixed OLPC parameters
P0 = 64 and = 0.6
In Figure 3.19, the CDF of UE transmit power is shown at different downtilting
angle from 0o to 8o with interval of 2o . As it can be seen, with downtilting
angle of 0o , about 20% of UEs are transmitted with maximum power. The
further downtilting of antenna does not reduce the UE PL, but rather loses the
cell coverage, where more UEs start operating at sacturated power in order to
maintain their own performances. At downtilting angle of 8o there are nearly
40% maximum power UEs.
The CDF of IoT is shown in Figure 3.20, as expected under the noise-limited
scenario, the average interference is already very small, i.e. at 50% the IoT is
49
Figure 3.20: Case-3 Interference over Thermal Noise for fixed OLPC parameters
P0 = 64 and = 0.6
about 2 dB with dowtilting angle of 0o . By further downtilting the antenna, the

reduced interference is very marginal, i.e. at downtilting of 8o there is about 1
dB reduction at 50%.
As discussed in Section 2.5, the FPC scheme partially compensates the PL,
which means that not all the UEs have the same SINR requirements. In the
Macro Case-1 scenario as presented in Section 3.5.2, the cell-edge UEs are able
to meet the SINR requirements imposed by the FPC algorithm at different
downtilting angles, since only a few cell-edge UEs (about 5%-10% of total users
shown in Figure 3.11) are transmitted with maximum power as they are in an
interference-limited scenario. However, when noise becomes the main constraint,
with different downtilting angles as shown in Figure 3.19, the cell-edge UEs are
in any case transmitting at maximum power so they tend to experience a similar
SINR regardless of the PL compensation factor. This behavior predicts a less
noticeable effect of the FPC algorithm with antenna downtilting on the 5%outage performance in Macro Case-3 scenario.
Based on the above transmit power and interference analysis, the CDF of received SINR distributions is shown in Figure 3.21. It is no surprise that there
is nearly no SINR improvement by using antenna downtilting under the Macro
50
Figure 3.21: Case-3 Received SINR Distribution for fixed OLPC parameters
P0 = 64 and = 0.6
Figure 3.22: Case-3 Average Cell Throughput vs. Cell-edge Throughput with
= 0.6
51
Case-3 scenario. For each downtilting angle, the probability peaks in the curve
between 20% and 40% range are caused by the cell-edge UEs operated at maximum transmit power, where the corresponding number of percentage for each
angle can also be seen in the Figure 3.19.
The 5% outage and average cell throughput versus the antenna downtilting
angle are shown in Figure 3.22. As it can be seen, the 2o is the optimal antenna
downtilting angle for the Macro Case-3 scenario in terms of both cell-edge and
average UE throughput.
3.5.4
Effect of Antenna Beamwidth
The effect of different vertical antenna HPBW has also been evaluated in this
study under the Macro Case-1 scenario.
Figure 3.23: Case-1 5%-outage vs. Average Cell Throughput with Vertical Antenna Pattern 7.8o , 10.0o and 12.0o for fixed OLPC parameters P0 = 58 and
= 0.6
In Figure 3.23, three vertical antenna patterns with HPBW equal to 7.8o , 10.0o
and 12.0o are investigated. As it can be seen, the narrower vertical antenna
pattern, i.e. HPBW=7.8o , provides higher performance at optimal downtilting
angle in both 5% outage and average cell throughput. However, it is also very
52
sensitive to the antenna downtilting angle. If the antenna is downtilted over

the optimal angle, the 5% cell-edge UE performance is decreased dramatically.
Whereas with wider vertical antenna pattern, the overall performance can be
maintained over a certain downtilting angle range. Practically, it can tolerate
some extent of downtilting angle bias by installation or tuning defects. In 3GPP,
the vertical antenna pattern with HPBW equal to 10o is used for system-level
evaluation.
3.6
Conclusions
In this chapter, the mechanical antenna downtilting was studied in the UL

LTE. The modeling and influence of mechanical antenna downtilting has been
discussed first. A simple antenna downtilting model has been proposed for
the practical LTE application. By pointing the first notch of vertical antenna
pattern to the boresight cell border, the mechanical downtilted antenna can
optimize the cell border user throughput.
The study followed by the system-level evaluations, where the simulations have
been performed with network-based mechanical antenna downtilting in the homogeneous UL LTE network, where both Macro Case-1 and Macro Case-3 scenario have been studied. As a sum-up, the performance gain is presented in
Table 3.2 compared with the downtilting case of 0o .
Table 3.2: Performance Gain of Mechanical Antenna Downtilting
Parameter
Macro Case-1:
= 0.6, P0 = 58
= 1.0, P0 = 106
Macro Case-3:
= 0.6, P0 = 64
Throughput Gains at Downtilting Angle

14o
16o
5% outage Aver. Thrpt. 5% outage Aver. Thrpt.
296%
147%
242%
153%
239%
111%
196%
121%
2o
5% outage
Aver. Thrpt.
13%
4%
As it can be seen, being an inter-cell interference mitigation technique, the

mechanical downtilting works efficiently under the interference-limited Macro
Case-1 scenario. Independent of OLPC parameters applied, the optimal antenna
downtilting angle for Macro Case-1 is 14o in terms of cell-edge user performance
and 16o in terms of average user throughput. Under the noise-limited Macro
Case-3 scenario, the application of antenna downtilting is less effective and the
optimal antenna downtilting angle is about 2o . It should be mentioned here
3.6 Conclusions
53
that the presented results and conclusions are based on a simplified channel
model and valid for the homogeneous network only. For a real deployment the
optimal tilting angles are actually smaller depending on the ISD distribution.
The results are also based on the 3GPP antenna patterns, which in reality can
look much different, hence lead again other optimal tilt angles.
The interaction of antenna downtilting together with the LTE open-loop FPC
has also been investigated under the Macro Case-1 scenario. The impact of
mechanical antenna tilting can be neglected. The optimization of UL open-loop
FPC and antenna tilting can be conducted independently.
The effect of antenna vertical HPBW has also been analyzed under the Macro
Case-1 scenario. The results show that the narrower the vertical antenna pattern, the higher performance can be achieved at the optimal downtilting angle.
However, it is also very sensitive to the over downtilting of antenna, where both
cell-edge user performance and average user throughput are decreased dramatically after the optimal angle. By utilizing wider vertical antenna pattern, the
overall network performance can be maintained over a certain downtilting angle
range. Practically, it can also tolerate some extent of downtilting bias by installation or tuning defects. In 3GPP application, the vertical antenna pattern
with HPBW equal to 100 is is used for system-level evaluation purpose.
54
Chapter
Uplink CoMP in the Form of

Macro-Scopic Combining
Coordinated Multi-Point (CoMP) is an advanced technique for interference mitigation which is proposed in the LTE-A study item as one of the features to
further reduce the impact of co-channel inter-cell interference [26]. For the UL
LTE application, the CoMP schemes include CoMP reception and coordinated
packet scheduling. The CoMP reception is to process signals received from multiple cells which are separated or not (inter- or intra-site, in the description) and
the coordinated PS can be seen as an extension of the Inter-Cell Interference
Coordination (ICIC) scheme already present in LTE by coordinating different
cell sites in terms of the scheduling decision (PRB allocation) in frequency domain.
In this chapter, the potential of CoMP reception in the form of macro diversity combining is analyzed. As shown in the following investigations, with the
combination of Interference Cancellation (IC) and Close-Loop fractional Power
Control scheme together with macro diversity reception, the overall UL LTE
network performance can be further improved.
In Section 4.1, the state of the art of UL CoMP is presented, and the defined
UL CoMP scenario for this study is described in Section 4.2. In Section 4.3, the
concept of macro diversity reception is briefly discussed, and the principle of
utilizing ideal IC is explained in Section 4.4. In Section 4.5, the applied CLPC
scheme is described based on the IC-based macro diversity reception. In Section
4.6, the modeling assumptions for this study are depicted, and the performance
evaluation results are presented in Section 4.7. Finally, the conclusions for the
macro diversity reception study are made in Section 4.8.
56
Uplink CoMP in the Form of Macro-Scopic Combining
Figure 4.1: CoMP Structure [79]
4.1
Coordinated Multi-Point
Under the interference-limited LTE environment, the main goal for applying
UL/DL CoMP techniques is to enhance both the cell-edge UE performance and
the average user throughput in an LTE-A system by mitigating the inter-cell
interference [26]. Theoretically, the inter-cell interference can be transformed
into the useful signal and being completely eliminated.
The basic idea behind CoMP is to make the multiple cells or network nodes cooperate via information exchange with centralized and/or distributed structures
[26] as shown in Figure 4.1. With centralized structure, the collaboration eNBs
are connected to a controller unit by means of fast backhaul network, e.g. Radio
over Fiber (RoF) links. In this way, the central processing of exchanged information can be performed within the controller unit. Alternatively, with the
distributed or decentralized structure, the eNBs are inter-connected through
microwave or RoF links, i.e. X2-interface in LTE, and a master eNB has the
central processing functionality to joint process the exchanged information and
coordinate the so-called slave/coordination eNBs. In general, the distributed
structure can be applied in the local small area and the centralized structure
can be utilized in a wider region which might contain several localized CoMP
with distributed manner.
In the practical application, the CoMP cooperation can be applied with either
intra-site or inter-site coordination as also shown in Figure 4.1. With the intrasite coordination, the sectorized eNB naturally contains the central processing
4.1 Coordinated Multi-Point
57
unit to control its own cells. Therefore, the latency of information exchange is
ideally zero (Practically due to the hardware issues, the latency may be nonzero). Besides, all coordinated sectors belong to the same site could be driven
by the same clock. So the synchronization of a certain UE to all the coordinated
sectors can be achieved. On the contrary, for the inter-site coordination, the cooperation strategies with information exchange heavily depends on the capacity
and latency of the links between the coordinated eNBs, and the synchronization
in both time and frequency of all coordinated sectors is a technical challenge in
practice [80].
For the UL LTE applications, the CoMP involves the techniques such as CoMP
reception and coordinated packet scheduling. The CoMP reception implies reception of UE transmitted signal at multiple geographically separated eNB antennas. Independent of the CoMP cooperation being utilized, it is practically
infeasible to fully cooperate network over a large scale. Restriction of the collaboration to a reasonable and small number of cells is required. Therefore, a
so-called CoMP cooperation area needs to be defined [81].
In the open literature, there are many research work dedicated to the study
of CoMP technology. The initial design concepts of CoMP systems have been
reported in [79]. From the theoretical point of view, the common conclusion is
that CoMP systems can bring tremendous improvement in the system performance for both cell coverage and cell throughput. The corresponding theoretical results have been presented in [36][37][38] for UL and [39][40][41] for DL.
However, considering of the practical aspects, the benefits of CoMP systems
come at a high cost of complexity and additional infrastructure. Several papers
[82][83][84] and 3GPP contributions [85][86][87][88] have reported the performance evaluation results for CoMP at system level, but most of the studies
were focus on the DL CoMP investigations. Depending on the assumptions and
the complexity of studied scenarios, the reported CoMP performance gain can be
varied significantly . From [84], 52% of average cell throughput gain and 144%
of cell-edge user throughput gain have been reported with dual stream CoMP
transmission, but no detailed explanations about the CoMP UE selection and
scheduling issues. Whereas in [89], detailed CoMP UE selection are employed
under the Macro Case-1 scenario, the reported gain for cell throughput and
cell-edge throughput is 0% and 15.8% respectively. Based on the 2-D antenna
pattern assumption in [90][91], 18% and 30% of gain in terms of 5% outage and
average user throughput has been reported for the UL LTE compared to the
LTE Rel8. The field trial results have also been presented in [92] for UL CoMP
based on the cellular scenarios of 2 base stations and 2 terminals. The reported
50% practical gain of user throughput is much smaller than the 150% theoretical
prediction due to the impact of imperfect channel estimation, limited number
of usable modulation and coding schemes, and various RF impairments.
58
In this study the CoMP reception in the form of macro diversity reception is
investigated based on the defined CoMP scenarios and 3-D antenna pattern
assumption presented in the previous chapter.
4.2
UL CoMP Study Scenario
In this study, an UL CoMP scheme is assumed where the serving eNB independently performs the RRM, i.e. dynamic PS, for the serving UEs, and it can
request coordination from the neighboring eNBs for a specific UE through the
X2-interface. An ideal X2 interface is assumed in order to find the upper bound
for the performance of UL CoMP, where the link connections, e.g. optical fiber,
between the sites have unlimited bandwidth, and no delay is considered.
Figure 4.2: CoMP - Macro Diversity Layout

The maximum number of 3 neighboring eNB cooperation is considered, where
the macro diversity reception scheme was investigated with either Intra-Site
or Inter-Site scenario. For a certain UE, if the serving and coordination cells
are limited to having the same site reception, then it is called Intra-Site macro
diversity reception. As shown in Figure 4.2, for UE-A, the serving link A1,
where the signal is received in the serving cell, and the macro diversity link A2,
where the signal is received in the coordination cell, are centrally processed in
the same Site-1. In this case no X2-interface is required for information coordination between cells. When the cross site reception is also allowed through the
X2-interface coordination, then it is called Inter-Site macro diversity reception,
i.e. as shown in Figure 4.2, serving link D4 and macro diversity link D3 of the
UE-D signal are separately received in Site-1 and Site-2. By transferring the
4.3 Macro Diversity Reception
59
necessary reception information from Site 1 to Site 2 through X2-interface, the

processing of macro diversity received signals can be performed in the serving
Site-2. In this study both cases are studied, and the Intra-Site macro diversity reception is mainly used as a reference to the Inter-Site macro diversity
reception.
UL CoMP operation for all the UEs in the network is practically infeasible.
In this study the selection of CoMP UEs, i.e. UE-A and UE-D in Figure
4.2, is based on the RSRP measurement. The RSRP is defined as the linear average over the power contributions of the resource elements that carry
DL cell-specific reference signals1 within the considered measurement frequency
bandwidth [16][93]. The measurement reporting is assumed to be UE-assisted
and network-decided, where the UE measures the RSRP from the neighboring
eNBs and reports the RSRP measurements back to the serving eNB. Based on
the measurements, the serving eNB will make the decision on whether a certain
UE i should be macro diversity received or not according to the criteria:
RSRPiServeLink RSRPiOtherLinks X [dB]
(4.1)
where RSRPServeLink is the RSRP measurement taken from the serving eNB,
RSRPOtherLinks is the RSRP measurement from one of the other surrounding
eNBs except the serving eNB and X is called triggering threshold or Combining
Window Size (CWS).
If a certain UE fulfills the criteria stated in Equation 4.1, then the UE is defined
as Macro Diversity UE with the corresponding eNB.
4.3
Macro Diversity Reception
Diversity is one of the MIMO technology applications. Depending on whether

the diversity is implemented in the transmitter or the receiver, it can be determinated as transmitter diversity, i.e. Multiple Input Single Output (MISO) and
receiver diversity, i.e. Single Input Multiple Output (SIMO). In this study the
receiver diversity reception is considered. The basic idea behind receiver diver1 Downlink Reference Signal contains Cell-Specific RS used by the UE for cell search, initial
acquisition and DL channel quality measurements, UE-Specific RS used by the UE to estimate
the DL channel to decode the eNB DL transmission and MBSFN RS for broadcasting channel.
Uplink Reference Signal contains Demodulation RS used by the eNB to estimate the UL
channel to decode the UE UL transmission and Sounding RS used by the eNB to estimate
the UL channel conditions for each user to decide the best UL scheduling.
60
sity is that multiple antennas receive different versions of the same signal and
the chance of all these copies of signal being in a deep fade is small. Therefore,
the reception reliability is enhanced. The combining schemes, such as MRC,
Selection Combining (SC) or Equal Gain Combining (EGC), can be utilized to
combine the received diversity signals. In this study the MRC scheme is studied
in details and the SC scheme is utilized as a reference.
Considering the received signal model with single transmit antenna from UE i
as:
ri
Nu
X
p
p
Ptx,i hi si +
Ptx,j hj sj + n
(4.2)
j=1,j6=i
p
Ptx,i hi si +
(4.3)
where ri = [r1 r2 rNr ] is the received signal vector from UE i on the Nr

T
coordination cell receive antennas. hi = [h1 h2 hNr ] is the channel vector
of UE i on the Nr receive antennas. si is the transmitted signal of UE i with
transmit power of Ptx,i . Nu is the number of UEs transmitting on the identical
PRB. sj is the transmitted signal of interference UE j with transmit power
T
of Ptx,j . hj = [h1 h2 hNr ] is the channel vector associated with the j T
th interfering UE. n = [n1 n2 nNr ] is the noise vector on the Nr receive
antennas and represents the inter-cell interference plus noise vector over the
transmission PRBs.
For the MRC, both phase correction and amplitude weighting estimations are
required at the receiver. The MRC combiner output signal can be expressed as:
H
sbi = hH
i ri = hi
Ptx,i hi si + hH
i
(4.4)
And the SINR of MRC combiner output is:

2

2

H

H 2
H
E h i ptx,i h i si
ptx,i h H
h
E
s
s
p
h

h
i i
tx,i
i i
i i
o

n
mrc =
=
=
2

H
H

hH
E h H
i E { } h i
i h i
E h H
i
=
=

2

ptx,i h H
h

i
i
2 h H
i UNr h i
2
Nr
X
ptx,i |hi,a |
a=1
Nr
X
a=1
i,a

2

ptx,i h H
h

i
i
2 h H
i hi

ptx,i h H
h

i
i
2
(4.5)
(4.6)
(4.7)
4.3 Macro Diversity Reception
61
where UNr denotes an Nr Nr identity matrix. The inter-cell interference plus

noise power is assumed to be spatially and temporally white with variance 2 .
hi,a is the channel of UE i at a-th receive antenna.
It is intuitive to note that the term h H
i hi =
Nr
X
|hi,a | is actually the sum of
a=1
the signal channel powers across all the Nr receive antennas and the total MRC
output SINR mrc is the sum of SINR i,a at each antenna.
Unlike the MRC scheme, for the SC scheme, the receive antenna with the highest SINR is chosen for further processing. The simple SC scheme only needs
a measurement of signal power, whereas the phase correction and amplitude
weighting information is not required. The SINR of SC combiner output is:
sc =
max
a[1, ,Nr ]
{i,a }
(4.8)
As seen from Equation 4.8, the SC scheme wastes the signal energy by discarding
(Nr 1) copies of the received signal. This drawback is avoided by the MRC
scheme, as shown in Equation 4.7, which exploits all available signal copies by
multiplying a complex weight in each signal copy and then sums them up.
According to [43] and presented in Figure 2.1, the user plane protocol stack in
LTE includes the protocols MAC, RLC, and PDCP. The MAC layer scheduler
decides the transport block size of the MAC Protocol Data Unit (PDU) to be
transmitted over the air interface. Based on this decision the RLC protocol
provides on request of the MAC layer a RLC PDU of appropriate size to the
MAC protocol either by concatenation or segmentation of PDCP PDUs. From
the Equation 4.2 and 4.5, for the whole MAC transport block, the received SINR
for UE i with MRC can then be expressed as:
mac
mrc
=
Ptx,i
NPRB
Nr
X
a=1
NX
PRB
|hi,a,p |
p=1
NX
PRB
Nu
X
p=1
j=1,j6=i
Ptx,j
2
|hj,a,p | + n2
NPRB
(4.9)
where NPRB is the number of assigned PRB and n2 represents noise power.
With respect to the system-level simulations, the UL macro diversity reception
62
is also naturally applied to the CSI measurements, which has been discussed in
Section 2.2. By applying the CSI macro diversity, better channel knowledge can
be acquired for the macro diversity users to perform the LA selection and PS
allocation.
4.4
Ideal Interference Cancellation
According to Equation 4.7, for the macro diversity UE, the received SINR with
MRC reception can also be expressed as:
mrc = serve +
Nmacro
X
macro,m
(4.10)
m=1
where serve is the received SINR at the serving eNB, Nmacro is the number
of macro diversity combining links, and macro is the received SINR at the
neighboring coordination eNBs.
In general, the macro link of macro diversity UEs experiences strong interference, where the main interference comes from the UE originally served in the
coordination cell. As an example shown in Figure 4.3 on the left, UE-A and
UE-B are served by eNB-1 and eNB-2 respectively. Both the UEs are assigned on the identical PRBs and transmitted in the same TTI. The UE-A is a
macro diversity UE who fulfills the criteria shown in Equation 4.1. According
to Equation 4.10 and 4.9, the received SINR for macro diversity UE-A can then
be expressed as:
mrc,A
= serve,A1 + macro,A2
= serve,A1 +
ptx,A
NPRB
Nr
X
NX
PRB
|hA,a,p |
p=1
PRB
a=1 NX
p=1
(4.11)
Nu
X
ptx,j
ptx,B
2
2
|hj,a,p | +
|hB,a,p | + n2
N
N
PRB
PRB
j=1
j6=A&B
2
where |hB,a,p | is the channel from UE-B at p-th PRB and a-th antenna.
The UE-B, normally has a good signal quality to its own serving eNB-2, which
consequently deteriorates the macro link received SINR of UE-A by contributing
strong interference as shown in Equation 4.11. The received SINR imbalance
between the serving and the macro link very much limits the performance of
overall macro diversity scheme. However, if the advanced receiver is deployed
4.4 Ideal Interference Cancellation
63
Figure 4.3: CoMP Example Macro Diversity IC
in the eNBs, e.g. SIC receiver, then the correctly received UE-B signal in the
coordination cell can be subtracted from the total interference in the macro
link of UE-A, as shown in Figure 4.3 on the right. So the macro link received
SINR macro can be enhanced, and the total combining gain can be further
improved. For the macro diversity UE-A, the received ideal SINR under the
SIC assumption can be expressed as:
IC
mrc,A
= serve,A1 +
ptx,A
NPRB
Nr
X
NX
PRB
|hA,a,p |
p=1
PRB
a=1 NX
p=1
Nu
X
ptx,j
2
|hj,a,p | + n2
N
PRB
j=1
(4.12)
j6=A&B
In this study an ideal Interference Cancellation scheme is assumed, where the UE

served by the coordination eNB, e.g.UE-B, can always be successfully canceled
and the upper bound of macro diversity reception by applying the IC can then
be studied.
As shown in the later chapter, the MMSE receiver together with SIC can be
utilized in reality instead of ideally removing the full interference. For further
references, a more detailed system-level modeling approach for SIC receiver can
be found in [94] as discussed in 3GPP.
64
4.5
IC-based Macro Diversity with Power Control
As discussed in Section 4.4, the performance of macro diversity UEs is enhanced

by utilizing the IC-based macro diversity reception. However, the non-macro
diversity UEs do not benefit from it at all. By applying the CLPC scheme,
together with the IC-based macro diversity reception, the overall network performance can be further improved.
The basic idea behind this scheme is that, generally, the macro diversity UEs
generate high inter-cell interference to the other UEs. By lowering their transmit power, less inter-cell interferences will be generated to the network and even
the non-macro diversity UEs can also get benefit from it. The negative effect of
powering down macro diversity UEs is compensated by the macro diversity reception gain. In the UL LTE, this scheme can be achieved by using standardized
CLPC as presented in Section 2.5.
In this study the absolute CLPC command is used, where, as described in
Section 2.5.2, the UE applies the transmit power offset based on the latest
OLPC command. From the Equation 2.5, the applied power control scheme can
be expressed as:
Ptx = min {Pmax , P0 + 10 log10 NP RB + L + f (i )}
(4.13)
(
Pof f set ,
f (i ) =
0,
(4.14)
where
if U E is a macro diversity UE
if U E is not a macro diversity UE
Two strategies were studied for setting the Pof f set value. For Scheme-1, all
macro diversity UEs reduce the transmit power by the same amount, e.g. Pof f set
= 1 dB, 2 dB, 3 dB or 4 dB. This Scheme-1 is used as a baseline reference. It
is obvious that this scheme is not the optimal solution since the transmit power
for the different macro diversity UEs should not be reduced with the same
amount. Scheme-2 is a further optimization of Scheme-1, where the UEs with
higher number of macro diversity combining links can be reduced with higher
offset value. In this study the UEs with 2 macro diversity link connections are
reduced by 2 dB and the UEs with 3 macro diversity link connections are further
reduced by 3 dB.
4.6
65
The system-level evaluations are performed based on the previous antenna downtilting study described in Section 3.4. The main simulation assumptions for UL
CoMP macro diversity reception study are shown in Table 4.1:
Table 4.1: Simulation Assumptions of UL CoMP Macro Diversity Reception
Parameter
Deployment Scenario
Number of UEs
P0 and value
Packet Scheduler Metrics
BS Antenna Pattern
BS Antenna Downtilting Angle
Receiver Type
Number of Combining Cells
Combining Window Size
CoMP Scenario
Assumptions
30 UEs/cell
-58 dBm and 0.6 / -106 dBm and 1.0
TD-Round Robin, FD-Proportional Fair
Horizontal/Vertical HPBW 70o /10o
12o
MRC
Maximum 3 cells
(1 serving cell + 2 coordination cells)
3 dB
Intra/Inter-Site
As it can be seen from Table 4.1, the interference-limited Macro Case-1 scenario
is the main focus in this study, where the 3GPP standardized, 70o horizontal and
10o vertical, 3-D antenna pattern with 14 dBi antenna gain is utilized. In order
to obtain a certain amount of macro diversity UEs for the accurate statistical
analysis, 30 UEs are generated per cell with spatially uniform distribution.
The RR metric is assumed in the TDPS and PF metric is assumed in the FDPS.
For a certain UE i transmits on PRB p at TTI t, the PF metric is defined as:
M etric
P Fi,p,t
=
Tbi,p,t
T i,t
(4.15)
where Tbi,p,t is the estimated Layer-1 achievable throughput for UE i, on PRB

p and at scheduling instant time t. The T i,t is the past averaged Layer-1 acknowledge throughput for UE i at scheduling instant time t, which is calculated
by using an exponential averaging filter:
T i,t = (1 )T i,t1 + Ti,t
(4.16)
where T i,t is the Layer-1 acknowledged throughput for UE i at scheduling instant time t and is called the filter coefficient.
The locations of macro diversity UEs in the network layout are shown in Figure
4.4. As it can be seen, all UEs are randomly generated in the whole network
66
Figure 4.4: Location of Selected Macro Diversity UEs (random realization)

region with balanced load in each cell. The selected macro diversity UEs in the
Inter-Site scenario are located in the overlapping area between any 2-3 cells, but
the selected macro diversity UEs in the Intra-site scenario are only located in
the region between Intra-site cells.
With different antenna downtilting angles, the percentage of macro diversity
UEs in both Intra- and Inter- Site scenario is shown in Figure 4.5. As shown
in the plot, the amount of macro diversity UEs in the Intra-Site scenario are
first decreased from downtilting angle 0o to 6o . At 0o tilt, the wide eNB antenna main-lobe is pointing to the far-away distance which creates a large celloverlapping area among both Intra-Site and Inter-Site cells. With downtilting
of antenna angle from 2o to 6o as discussed in Section 3.3, the wide antenna
main-lobe starts touching the ground which reduces the cell-overlapping region.
By further mechanically increasing the antenna downtilting angle, as discussed
in Section 3.1 and shown in Figure 3.3, the radiation pattern of mechanically
downtilted antenna shrinks from the boresight direction and gets wider from
the sides, which increases the Intra-Site cell overlapping area and results in
Figure 4.5: Percentage of Macro Diversity UEs of varying

Downtilting Angle at CWS=3 dB
Figure 4.6: Percentage of Macro Diversity UEs of varying

CWS Values with Downtilt Angle 12o , 14o , 16o
67
68
the increasing amount of Intra-Site macro diversity UEs. On the other hand,
because of the shrinking of Inter-Site cell-overlapping region, the amount of selected macro diversity UEs in the Inter-Site scenario are reduced with increasing
downtilting angles until 16o . By further downtilting, the increasing amount of
UEs, located between Intra-Site cells, cause the total amount of macro diversity
UEs in the Inter-Site scenario to increase.
The influence of different CWS at downtilting angle of 12o , 14o and 16o is shown
in Figure 4.6. At smaller CWS value, the amount of macro diversity UEs for
both Intra- and Inter- Site scenario are very close among all three angles. With
increased CWS value at higher antenna downtilting angle, the amount of InterSite scenario UEs are decreased with increasing Intra-Site scenario UEs.
According to the above analysis, the 12o downtilting angle is applied together
with the optimal OLPC parameter P0 = 58 dBm and = 0.6 as discussed
in Section 3.5.2. This is done to create larger cell-overlapping area in favor of
macro diversity reception. From the practical application point of view, if a UE
is located too far away from its coordination eNBs, the received RS signals used
for demodulation and channel estimations are too weak. Therefore, the CWS
of 3 dB is utilized in this study, where the PL or RSRP differences between
serving link and macro links are less than or equal to 3 dB as expressed in
Equation 4.1. As shown in Figure 4.6, with CWS equal to 3 dB, there are about
7% macro diversity reception UEs in the Intra-site scenario and 23% macro
diversity reception UEs in the Inter-site scenario.
4.7
Performance Evaluation
The investigations of macro diversity reception are carried out for both fractional
and full compensation of open-loop FPC, with either = 0.6 or = 1.0 as
discussed in Section 2.5.1, where the fractional case of = 0.6 is the main
focus. As presented in the following section, the Inter-Site scenario with MRC
reception is presented and analyzed in detail. The Intra-Site scenario with MRC
and Inter-Site scenario with SC receptions are mainly utilized as a reference
guideline.
4.7.1
Evaluation of Macro Diversity with = 0.6
The received SINR distributions for Inter-site MRC UEs are shown in Figure
4.7. The macro link SINR and the total combining SINR are shown respectively
4.7 Performance Evaluation
69
for both with ideal IC and without ideal IC case.

Without the ideal IC applied in the coordination cell, the SINR difference between serve link and macro link is about 8 dB. This large SINR in-balance
leads to a marginal MRC combining gain. By exploiting the ideal IC in the
coordination cell as also shown in Figure 4.7, the received SINR in the macro
link is improved by nearly 7 dB, which results in about 2.2 dB improvement of
total MRC combining gain over the serve link SINR.
Figure 4.7: Average SINR Distribution for Inter-Site MRC

UEs with = 0.6 and P0 = 58 dBm
The reduction of SINR imbalance between serve link and macro link can also be
seen from the link switching frequency of Inter-Site SC UEs shown in Figure 4.8.
This indicator shows how often the strongest SINR is selected from the macro
links. When no ideal IC applied, the strong interference from the coordination
cell lowers the SINR of macro link, which results in nearly no selection switching
from the serving link. In Figure 4.8, 50% of Inter-Site SC UEs experience 5%
or less switching. When the ideal IC is utilized, the balanced serve and macro
links lead to more selection switching or diversity. Still with 50% of Inter-Site
SC UEs, the selection switching to the macro link is improved to about 40%.
The bar plot of 5% outage and average user throughput gains over the no macro
70
Figure 4.8: Switching Frequency for Inter-Site SC UEs with

= 0.6 and P0 = 58 dBm
diversity case1 are shown in Figure 4.9. With no ideal IC applied2 , the large
SINR in-balance results in nearly no gain improvement in the case of Inter-Site
SC reception. Because of the small amount of Intra-Site macro diversity UEs
with only 7% as shown in Figure 4.6, the overall performance improvement for
Intra-Site MRC reception case is quite marginal as well. For the case of Inter-Site
MRC reception, the 23% of Inter-Site macro diversity UEs leads to about 8%
and 3% gain in terms of 5% outage and avergage user throughput respectively.
When the ideal IC is utilized, the overall performance is enhanced for all the
three cases compared to the case with no ideal IC. For the best case, the InterSite MRC reception provides more benefits for the cell-edge UEs, which gives
about 14% gains in terms of 5% outage and 8% gain in terms of average user
throughput.
The study of CLPC schemes described in Section 4.5 is performed based on the
case of IC-based Inter-Site MRC reception. Figure 4.10 shows the distributions
of transmit power for Inter-Site reception UEs. Since no CLPC is applied, the
1 No Macro Diversity Case: which is based on the previous antenna downtilting study with
downtilting angle of 12o and depends on the application with either OLPC parameter ( = 0.6,
P0 = 58 dBm) or ( = 1.0, P0 = 106 dBm). The overall throughput gain numbers are
presented in Figure 3.17 and Figure 3.18
2 No/Without IC Case means when no IC scheme is applied
71
Figure 4.9: Throughput Gain Plot for No IC vs Ideal IC with

= 0.6 and P0 = 58 dBm
CDF curve of no macro diversity case and IC-based Inter-Site reception case
overlap and are used as baseline references for the CLPC cases. When the
CLPC is exploited, the transmit power reduction of macro diversity UE results
in the shifting curves to the left compared to the reference curves. As expected,
for CLPC Scheme-1, the mean value of the CDF curve is reduced by about 1
dB in each case. For CLPC Scheme-2, because the macro diversity UEs are
reduced either by 2 or 3 dB as presented in Section 4.5, its CDF curve is lying
between the -2 dB and -3 dB curves of Scheme-1.
The CDF curves of the average SINR for Inter-Site MRC UEs are shown in
Figure 4.11. To compare with the reference case of IC-based Inter-Site MRC
when no CLPC is applied, it is no surprise that the received SINR values are
also decreased for the CLPC cases because of transmit power reduction.
The plot of throughput gains relative to the no macro diversity case is shown
in Figure 4.12. To compare with the IC-based Inter-Site MRC without CLPC
applied, which was the the best scheme as shown in Figure 4.9, by using CLPC
Scheme-1, both 5% outage and average user throughput are enhanced by reducing the transmit power of macro diversity UEs. At the optimal case of powering
down 2dB, the 5% outage and average user throughput can be enhanced by 24%
72
Figure 4.10: Transmit Power Distribution for Ideal IC-base

Inter-Site MRC UE with CLPC ( = 0.6 and P0 = 58 dBm)
Figure 4.11: Average SINR Distribution for Ideal IC-base InterSite MRC UE with CLPC ( = 0.6 and P0 = 58 dBm)
Figure 4.12: Throughput Gain Plot with CLPC for = 0.6

and P0 = 58 dBm
Figure 4.13: Comparison of different P0 values with CLPC

Schemes for = 0.6
73
74
and 11% respectively. Further decrease of the transmit power, i.e. the transmit power reduction of 4 dB, the average user throughput performance can still
be further improved, which means the other non-macro diversity UEs can still
benefit from the CLPC. But the 5% outage user throughput is penalized by
decrease of the performance for some of the macro diversity users. By applying
the Scheme-2, the variety of the transmit power brings additional gains for both
5% outage and average user throughput over the Scheme-1. To compare with
the no macro diversity case, there are about 27% and 12% gains for the 5%
outage and average user throughput respectively.
Since the CLPC scheme applied was based on the OLPC parameters, the optimum P0 value was investigated with = 0.6 for the IC-based Inter-Site MRC
reception as shown in Figure 4.13. The blue curve in Figure 4.13 presents the
performance of IC-based Inter-Site MRC with varying of P0 values and the red
curve shows the performance of IC-based Inter-Site MRC with P0 = 58 at
different macro diversity schemes. According to the previous antenna downtilting study, the P0 = 58 dBm was the optimal value for = 0.6. As it can
be seen from the blue curve in Figure 4.13, for the macro diversity reception,
the P0 = 58 dBm still gives the optimal performance compared with other P0
values in terms of both 5% outage and average user throughput. Utilizing the
CLPC schemes on the top of P0 = 58 provides even higher gains as shown in
the trend of the red curve.
4.7.2
Evaluation of Macro Diversity with = 1.0
Compared with the OLPC parameter = 0.6, by utilizing = 1.0, UEs have
less transmit power and identical received power spectral density in the eNB.
As shown in Figure 4.14 of the two no macro diversity cases, = 1.0 is more
favorable for the cell-edge UE performance.
In this section, the average network performance of CoMP macro diversity reception with FPC parameter = 1.0 are presented with P0 = 106, which is
the optimal parameter setting as discussed in Section 3.5.2.
In Figure 4.14(a), the CDF curve of received SINR for Inter-Site MRC UEs
are shown. As expected, compared to the baseline of no macro diversity with
= 1.0 and P0 = 106 dBm, the ideal IC-based Inter-Site MRC reception
improves the performance of macro diversity UE, and there is still much higher
gain when the ideal IC scheme is applied in the coordination cell.
In Figure 4.14(b), the CDF curves of UE throughput are shown. It can be seen
that, with the joint effect of using = 1.0 and macro diversity reception, the
(a) Received SINR Distribution
75
(b) UE throughput Distribution
Figure 4.14: Average SINR & UE Throughput Distribution without CLPC with
= 1.0 and P0 = 106 dBm
macro diversity UE statistics are actually moving to the top part of the CDF
curve, i.e. around 95% region, whereas the non-macro diversity UEs are down
to the tail of the CDF curve. So the 5% outage of average user throughput is
now representing the non-macro diversity UE performance.
The plot of throughput gains relative to the baseline of no macro diversity with
= 1.0 and P0 = 106 dBm are shown in Figure 4.15 for all three combining
reception cases. Similar to = 0.6, the cancelled coordination cell interference
can always help to improve both 5% outage and average user throughput. However, the enhancement of 5% outage user throughout is very marginal. With
the best case of ideal IC-based Inter-Site MRC reception, there is only 3% gain
in terms of 5% outage, where it was 14% gain when = 0.6 is applied as shown
in Figure 4.9. This is simply because the cell-center UE does not enjoy much of
the benefits from the macro diversity receptions. However, for the cell-edge UE,
its benefit from the macro diversity reception brings about 9.5% gain in terms
of average user throughput.
In Figure 4.16, the CDF of UE throughput are shown to present the performance
gain by using CLPC Scheme-1. As it can be seen the throughput of macro
diversity UEs is reduced with the increasing of power reduction. However, the
performance of cell-center UEs around the 5% outage region is actually further
enhanced and maximized with power reduction of 3 dB.
76
Figure 4.15: Throughput Gain Plot for No IC vs IC with

= 1.0 and P0 = 106 dBm
The plot of CLPC throughput gains are shown in Figure 4.17. For the Scheme-1,
in general, the same trend of gains can be seen as described in = 0.6. Except
that the maximum gains are now optimized at power reduction of 3 dB, where
it was 2 dB for the case of = 0.6. This is due to the fact that, by applying the
cell-edge UE favorable parameter = 1.0, the performance of cell-edge macro
diversity UEs is improved, where in return, by applying the CLPC scheme on
the top, they can reduce even more power in favor of other UEs transmission.
So all in all at -3 dB of Scheme-1, there are further improvements of 18% and
5% gains in terms of 5% outage and average user throughput respectively. The
large 5% outage gain number also proves that, for the application of = 1.0,
it is very important to have the CLPC schemes applied together with IC-based
macro diversity receptions in order to have the overall enhancement of network
performance. The utilization of Scheme-2 has also been investigated where it
shows the same performance as the optimum case of Scheme-1.
Figure 4.16: Average User Throughput Distribution with

CLPC for = 1.0 and P0 = 106 dBm
Figure 4.17: Throughput Gain Plot for No CLPC vs CLPC

with = 1.0 and P0 = 106 dBm
77
78
4.7.3
Interaction of Antenna Tilting and UL CoMP
The interactions of antenna downtilting and macro diversity reception are analyzed in this section. In Figure 4.18, the throughput gains of Intra- and InterSite MRC receptions are shown with varying downtilting angle from 10o to 16o .
The no macro diversity case, with 12o downtilting and = 0.6, is used as a
baseline reference to compare with (0% gain).
Figure 4.18: Throughput Gain Compare for different Antenna Downtilting Angles with = 0.6 and P0 = 58 dBm
As it can be seen, for the case of Intra-Site MRC reception, the 14o antenna
downtilting can provide even better throughput gains than 12o case. It is because the increased number of selected Intra-Site macro diversity UEs as shown
in Figure 4.5 and the enhanced performance by downtilting antenna angle as
shown in Figure 3.16. Generally, there is about 5% gain for both 5% outage
and average user throughput over the 12o case. By further downtilting to 16o ,
the average user throughput can be further enhanced by the combined effect
of increased number of Intra-Site macro diversity UEs and antenna dowtilting
gain. However, the increased number of MRC UEs are mainly located around a
high SINR region. Because of losing cell coverage at 16o , the low SINR UEs can
get no benefit from the Intra-Site MRC reception, which results in the decreased
performance of 5% outage at 16o . Besides, also due to the loss of antenna downtilting gain at 10o as shown in Figure 3.16, 10o case shows a negative gain in
both 5% outage and average user throughput to compare with the 12o reference
79
case.
For the case of Inter-Site MRC reception, the 14o antenna downtilting can still
provide about 6% more gains than the 12o in both 5% outage and average user
throughput, because the selected macro diversity UEs in the Inter-Site scenario
contain the low SINR UEs, who are located at the borsight cell-border area. It
is very interesting to see that by taking the benefit of Inter-Site MRC reception,
even for downtilting angle at 16o , the 5% outage UEs can still achieve about 9%
improvement.
Figure 4.19: Throughput Gain Compared for different Antenna

Downtilting Angles with = 1.0 and P0 = 106 dBm
In Figure 4.19, the same investigation is shown for the case of = 1.0 and
P0 = 106 dBm, where the same trend can be seen as the previous case with
= 0.6 and P0 = 58 dBm. For both Intra-Site and Inter-Site MRC schemes,
the downtilting angle of 14o can provide further enhancement in both 5% outage
and average user throughput compared to the 12o downtilting.
4.7.4
Non-Ideal Cell Selection
In the previous analysis the studies were based on the ideal cell selection, where
the UEs are always connected to the highest path gain sector as its serving cell.
However, in reality, the UE selected serving cell might not be the strongest link
80
because of measurement errors. This imperfection from serving cell selection

influences the UE performance. So in this section the non-ideal cell selection
scenario is being considered, and the potential gain by using CoMP macro diversity reception is evaluated.
Figure 4.20: Throughput Gains Of Non-Ideal Cell Selection

In this study the non-ideal cell selection is modeled as the UE randomly connects
to one of sectors from the so-called virtual active set size (ASS) [95], where the
maximum number of sectors in the virtual ASS is 3.0 as default [95]. The
window margin is used to determine the sectors in the virtual ASS, where the
default setting is selected as 3.0 dB. By using this setting, as studied in [96],
about 20% of the UEs take the non-strongest link as their serving link within
the virtual ASS.
In Figure 4.20, the bar plot throughput gains are shown for all the combining
reception cases with ideal IC applied. For the Intra-Site MRC, not much gain
is observed because of limited number of Intra-Site macro diversity UEs. For
the Inter-Site scenario with MRC, the impact of non-ideal cell selection deteriorates the performance of reference case. No matter which combining scheme
is utilized, there are much higher gains in both 5% outage and average user
throughput to compare with the gain numbers shown in Figure 4.9 with ideal
cell selection.
4.8 Conclusions
4.8
81
Conclusions
In this chapter, the CoMP in the form of macro diversity reception has been
studied in the UL LTE, where both OLPC parameter with = 0.6 and = 1.0
are evaluated. The study is concentrated on the analysis of Inter-Site scenario
with MRC reception. The Intra-Site MRC and Inter-Site SC are utilized as
references. The baseline scenario is no macro diversity case with 12o downtilting
as investigated in the earlier chapter. The main conclusions from the study is
summarized in Figure 4.21 by taking Inter-Site MRC reception with = 0.6 as
an example.
Figure 4.21: Conclusion Comparison at 12o downtilting

with = 0.6 and Ideal-Cell Selection
As it can be seen, with standing alone macro diversity reception, the improvement of both 5% outage and average user throughput is very marginal. To
exploit the CLPC scheme alone, the enhancement is limited as well without the
support of ideal IC in the coordination cell. By cancelling the strongest interference in the macro diversity link, the performance of both 5% outage and average
user throughput can be largely enhanced, and by applying the CLPC scheme on
top of IC-based MRC reception, the outage gain number can nearly be doubled
together with further improvement of average user throughput. So the combination of macro diversity reception together with IC and CLPC scheme is very
important in order to maximize the overall network enhancement.
82
When evaluating the case of = 1.0, the 5% outage UEs represent the cell-center
UE performance where no macro diversity applied. The use of ideal IC-based
Inter-Site MRC reception slightly improve the average user throughput, the
outage UEs do not enjoy much of the benefit from the combining gain. However,
by applying the CLPC schemes to reduce the power of macro diversity UEs, the
outage throughput UEs are largely improved since less interference is received
from macro diversity UEs. And the effect of powering down the macro diversity
UEs is compensated by macro diversity reception gain.
The interaction of antenna downtilting angle together with macro diversity receptions has also been studied. At downtilting angle of 14o , both Intra-Site
MRC and Inter-Site MRC schemes provide better performance than at 12o .
The study with non-ideal cell selection scenario also demonstrated that, in the
practical application, macro diversity receptions can provide even higher gain
than the simulated ideal cell selection assumption.
Of course it has to be realized that the study results presented in this chapter
were based on the MRC receiver with ideal IC. By doing so, the upper bound
performance of CoMP macro diversity reception can be studied. By applying the
realistic MMSE/SIC receiver as presented in the next chapter, the performance
gain numbers will be decreased correspondingly.
Furthermore, compared with the theoretical CoMP investigations shown in the
earlier literature, the achievable CoMP gain in this study is generally lower
than the theoretical limits because of limited number of coordination eNBs and
selected number of macro diversity UEs. In the realistic network, the overall
CoMP performance may be even reduced due to the practical issues such as
channel estimation errors, synchronization and latency challenges.
Chapter
Uplink CoMP Joint Reception

In the UL LTE, CoMP reception can also be jointly applied between the paired
UEs. Comparing with CoMP macro diversity with MRC reception as presented
in Chapter 4, higher performances in terms of both 5% outage and average
user throughput are expected for CoMP joint reception to better exploration
of the channel spatial signature. On the other hand, the CoMP joint reception
requires a large amount of information exchange between eNBs. Especially for
the Inter-Site scenario, it involves the coordination of multiple network sites
through backhaul transport, and a large amount of information exchange will
result in additional burdens and requirements on the network backhaul design.
In this chapter, the performance of CoMP joint reception is investigated in
detail compared with the CoMP macro diversity with MRC reception. The
requirements of LTE network backhaul are analyzed for both cases based on the
realistic MMSE/SIC receiver. The independent PS performed in each individual
cell is utilized in this study, whereas the performance of multi-cell coordinated
PS is presented in the later chapter.
In Section 5.1, the study scenario of CoMP joint reception is described, and
the requirement on LTE X2-interface for UL CoMP applications is analyzed in
Section 5.2. In Section 5.3, the UL CoMP with MMSE/SIC receiver is presented
for joint reception. The modeling of RS for CoMP joint reception is described
in Section 5.4. In Section 5.5, the modeling assumptions for this study are
depicted, and the performance evaluation results are presented in Section 5.6.
Finally, the conclusions for the CoMP joint reception study are made in Section
5.7.
84
5.1
5.1.1
CoMP Joint Reception

Single-Cell Multi-User MIMO in UL LTE
In LTE Rel8, Single-User MIMO (SU-MIMO) cannot be supported in the UL

but the UL can support single-cell Multi-User MIMO (MU-MIMO) transparently. With only one UE transmit antenna and multiple eNB receive antennas,
3GPP has agreed to offer some rudimentary support for spatial multiplexing of
different UEs in the same cell [43]. Rel8 standard adopts 2-by-2 (2x2) single-cell
MU-MIMO in the LTE uplink, i.e. each group contains two UEs with single
transmit antenna and the eNB has two receive antennas. It may be noted that
the support of multiple transmit antennas at the UE is being considered for
LTE-A.
Figure 5.1: Single-Cell MU-MIMO (UE-B and UE-C are single-cell spatial multiplexing users at eNB-2)
The multi-user technique is known from the CDMA system. In a CDMA system,
each user assigns a unique spreading code to encode the transmitted signal and
allows multiple users to share the same frequency band and the same time slot.
At the receiver side, the multi-user data are separated on the basis of their
signature waveforms (which is the spreading code convolved with the channel
impulse response). Ideally the user signature waveforms are mutually orthogonal
so as to avoid MAI among different users. Practically the non-orthogonality due
to the near-far effect and MAI can be suppressed by using the advanced signal
processing with multiple antenna array at the eNB receiver [97].
To exploit the single-cell multi-user technique in the UL LTE, the packet scheduler allow multiple UEs to simultaneously transmit independent data in a cell on
the same assigned physical resource blocks and the users data can be separated
5.1 CoMP Joint Reception
85
in the space domain at the eNB with multiple antenna array. The UE-specific
spatial signature1 generated by the independent multipath channel over the
transmitted signal acts like the signature waveforms in the CDMA system. The
packet scheduler can allocate the spatial multiplexing UEs with the best and
most orthogonal spatial channels in order to exploit the multi-user spatial diversity2 [98]. The interference seen by the spatial multiplexing UE comes from two
parts: the intra-cell interference due to the co-scheduled UE involved in the UL
MU-MIMO and the inter-cell interference due to co-channel UEs in other cells.
By utilizing the Multi-User Detection (MUD) with multiple antenna at eNB,
the intra-cell interference between these two UEs can be significantly reduced
[16].
Many MUD algorithms have been studied from both theoretical and practical point of view with the CDMA-based system and OFDM-based SU-MIMO
transmission [99]. In general, the optimum multi-user detectors require a high
computational complexity which is not feasible for the practical implementation.
Therefore, many sub-optimum MUD algorithms have been investigated with low
complexity. Basically it can be divided into linear detectors, e.g. decorrelator
[100] or MMSE detector [101], and non-linear detector, e.g. decision feedback
detector [102][103], multi-stage detector [104] or interference cancellation detector [105]. As presented in [106][107], those MUD algorithms have been further
extended and utilized in the single-cell MU-MIMO transmission.
5.1.2
Extension of Single-Cell Multi-User MIMO
In some of the literature [82][108], CoMP is also denoted as Multi-Cell / Network

/ Cooperative MIMO. In this study, UL CoMP can be considered as a multi-cell
MU-MIMO solution which is the extension of single-cell MU-MIMO scenario.
In Chapter 4, UL CoMP in the form of macro diversity reception has been
studied. As shown in Figure 5.2(a), UE-A and UE-B are co-channel users and
served in the different cell, eNB-1 and eNB-2. In order to reduce the link
imbalance to improve the overall macro combining gain as discussed in Section
4.4, in the coordination cell eNB-2, a 2x2 multi-cell MU-MIMO can be performed
between the serve link of UE-B and the macro link of UE-A. In Chapter 4,
the upper bound of CoMP macro diversity reception has been investigated.
The mutual interference between the two cross-cell multiplexing UEs has been
ideally (full) removed. Practically the cross user interference can be suppressed
with multi-user detection receiver as exploited in the single-cell MU-MIMO.
As presented in [109], the linear MMSE detector and non-linear SIC receiver
1 Spatial
2 Spatial
signature is the response vector of an antenna array from a certain UE.

diversity is the difference of spatial signatures of UEs.
86
has fairly low complexity and near-optimum performance to compare with the
Maximum-Likelihood (ML) receiver, which requires very high complexity but
very optimal performance. In the following section, the CoMP macro diversity
reception is studied with the linear MMSE detector or non-linear MMSE-SIC
receiver, and utilized as a reference.
(a) Macro Diversity Reception
(b) CoMP Joint Reception
Figure 5.2: Structure of CoMP Joint Reception
By further extending the CoMP macro diversity scenario through the multi-eNB
coordination,, all the received signals can be exploited as desired signals rather
than treated as interference. As shown in Figure 5.2(b), with the macro link
signal of UE-B also considered in eNB-1, a 2x4 multi-cell MU-MIMO can be
formed. Based on this scenario, the CoMP technology in the form of joint reception is investigated in this chapter. As known from the earlier literature [110],
with multiple receiving antennas at eNB, the received signal strength can be
enhanced by applying the advanced baseband signal processing and the impact
of co-channel interferences can be mitigated. It is also known that the performance gain of multi-antenna signal processing is proportional to the number of
antennas. With higher number of antennas, the diversity of different branches
can be utilized to mitigate the serious channel fading and improve the reliability
of system performance. By cooperating multiply eNBs in the UL direction as
shown in Figure 5.2, a serving eNB can virtually increase its number of receiving
antennas. The received information from the coordination eNB antennas can
be exchanged through the LTE X2-interface and utilized in the serving eNB to
perform the joint signal processing of cross-cell multi-user detection. The application and performance comparison with the CoMP macro diversity reception
is the key interest in this study.
5.2 LTE X2-Interface for CoMP Applications
87
Similar to the function of PS in the single-cell MU-MIMO, the multi-cell CPS

can also be used to allocate the cross-cell spatial multiplexing UEs with the
best and most orthogonal spatial channels in order to fully exploit the multicell multi-user spatial diversity. In this chapter, the CoMP receiver with joint
reception is the main focus. No specific multi-cell CPS is applied. The two
joint detection UEs are allocated respectively by their own serving cell PS and
can be considered as paired by a random-fashion PS algorithm between the
coordination cells. In Chapter 6, the performance of CoMP joint reception will
be further investigated when the algorithm of multi-cell CPS is exploited.
5.2
LTE X2-Interface for CoMP Applications
In order to support different CoMP applications, the X2-interface in the LTE

network as presented in Figure 2.1 needs to fulfill a certain bandwidth and
latency requirements. In this section, the theoretical capacity requirement of
X2-interface is estimated.
In Figure 5.3, the eNB protocol stack for CoMP reception in the case of both
macro diversity and joint schemes are shown. For the macro diversity reception
as shown in Figure 5.3(a), in the PHY layer of coordination eNB-2, it performs
the channel estimation for both signals from UE-A and UE-B, and is followed
by the multi-user detection, i.e. MMSE or MMSE-SIC, between the two UEs.
For the macro diversity combining of UE-A, the MRC can be applied in the
Chase Combining fashion, where the packets are combined at bit level before
decoding by direct addition of the demodulation soft bits1 . By transferring the
quantized soft bits to the serving eNB-1 through X2-interface, the serving eNB1 combines the soft bits with its own demodulated soft bits received from UE-A
and proceeds with the final decoding and Cyclic Redundancy Check (CRC)
afterward.
Based on the physical resource structure in UL LTE as specified in [43], one
PRB consists of Nsubcarrier = 12 OFDM sub-carriers and Nsymbol = 14 OFDM
symbols, where 2 OFDM symbols are Reference Symbols used for demodulation. Assume the signal demodulation is performed at the coordination eNB-2,
the RS information does not have to be sent to the serving eNB-1. Assume the
Quadrature Phase Shift Keying (QPSK) is used, the demodulation soft bits are
quantized with resolution of Nresolution = 6 bits over the X2-interface transmission, for the MRC application, the capacity requirement for X2-interface per
1 Demodulation
soft bit is equivalent to values of Log Likelihood Ratio (LLR) given as:
P {b=1|r}
LLR(b) = log Pr {b=0|r} . Summation of LLR values at bit levels is mathematically equivalent
r
to perform MRC at symbol level before calculating the LLR [111].
88
(a) Macro Diversity Reception
(b) Joint Reception
Figure 5.3: Macro Diversity Reception vs. Joint Reception Backhaul Stacks
PRB per TTI can simply be estimated as:
Nsubcarrier Nsymbol Nresolution
(5.1)
= 12 subcarriers (14 2) symbols 6 bits
(5.2)
= 0.864 Mbit per PRB (180 kHz) per TTI (1 ms)
(5.3)
Assume a practical scenario, with CoMP application of 16 PRBs or 3 MHz,

the capacity requirement for X2-interface to facilitate MRC reception is about
0.86416 14 Mbit per TTI. By considering the maximum scalable bandwidth
20 MHz specified in Release 8 [43] with 100 PRBs, the maximum capacity requirement for X2-interface with MRC reception is about 0.864100 86.4 Mbit
per TTI.
Whereas for the joint CoMP reception as shown in Figure 5.3(b), in order to
5.2 LTE X2-Interface for CoMP Applications
89
perform the 2x4 multi-cell multi-user signal processing in the PHY of serving
eNB-1, the In-phase and Quadrature (I&Q) samples1 of received UE-A and UEB signals at each antenna have to be transferred from the coordination eNB-2
via the X2-interface, because the joint signal processing in the serving eNB-1
requires the channel information from the coordination eNB-2. The RS symbols
also need to be transferred through the X2-interface.
By assuming the I&Q samples quantized with 8 bits each (Ni&q = 2) for QPSK
transmission through X2-interface, with Nt = 1 transmit antenna and Nr = 2
receiving antenna, the capacity requirement for X2-interface with joint CoMP
reception per PRB per TTI can simply be estimated as:
Nsubcarrier Nsymbol Nresolution Ni&q Nt Nr
(5.4)
= 12 subcarriers 14 symbols 8 bits 2 iq 2 rx
(5.5)
= 5.38 Mbit per PRB (180 kHz) per TTI (1 ms)
(5.6)
Still by assuming the practical CoMP application with 16 PRBs or 3 MHz, the
capacity requirement for X2-interface to support joint reception is about 5.38
16 86.08 Mbit per TTI. By considering the maximum scalable bandwidth
20 MHz in Release 8 with 100 PRBs, the maximum capacity requirement for
X2-interface with joint reception is about 5.38 100 0.54 Gbit per TTI. To
compare with the X2-interface capacity requirement for CoMP MRC macro
diversity application presented in Equation 5.1, the backhaul requirement for
the CoMP joint reception is about 6 times higher.
It can be foreseen that the HARQ operations has also impact on the CoMP application. Based on the LTE Release 8 specifications, the Round-Trip Time (RTT)
for synchronous HARQ application is about 8 Millisecond (ms) as discussed in
Section 2.3. So it is expected that, in general, the X2-interface transmission
time and processing delay in the eNB should not contribute significantly to the
total delay budget for the CoMP application. Otherwise, the longer delay could
cause data stalling in the HARQ process. According to the present assumptions in the LTE Release 8, the transmission time of X2-interface, from serving
eNB-1 to coordination eNB-2, should be less than 1 ms, and the estimation
for X2-interface related eNB processing delay should also be less than 1 ms.
Otherwise, the changes for the HARQ RTT are required in the LTE-A network
for the CoMP applications.
1 An I&Q sample is the complex representation of a constellation point for a given subcarrier, which is received on a given antenna. It is the output of Fast Fourier Transform (FFT)
at the Orthogonal Frequency Division Multiplexing (OFDM) receiver chain and basically contains the amplitude and phase information where a particular sub-carrier has been modulated.
90
5.3
System Modeling of MMSE/MMSE-SIC Receiver for CoMP Joint Reception
In recent years, substantial amount of research has been conducted to the task of
MUD. With multiple antenna exploited at the CoMP receiver end, the separation of different users can be performed based on their unique spatial signature
by assuming the knowledge of the channel parameter [21][110].
MMSE detector is one of the well-known linear multi-user detectors. It can
balance the signal decoupling and noise enhancement [112]. In [113], it has
been shown that at low Signal-to-Noise Ratio (SNR) the MMSE detector converges to the conventional (single-user) receiver and at high SNR it converges
to the decorrelator detector. MMSE multi-user detector has been studied and
utilized in the CDMA-based system [114], OFDM-based SU-MIMO [115] and
MU-MIMO [116][117] system. In this study, we will exploit the MMSE-based
multi-user detector to the UL CoMP applications. The basic MMSE receiver
are first presented in Section 5.3.1 and the advanced MMSE-SIC are described
in later Section 5.3.2.
5.3.1
Linear MMSE Detector
Considering a general scenario of CoMP joint reception, Nu joint transmission

UEs have single transmit antenna each and coordinated eNBs contain Nr receive
antennas. Assuming that the channel is static over the duration of a PRB, as
shown in Figure 5.3 of the physical layer block diagram, the received signal after
Cyclic Prefix (CP) removal and FFT can then be expressed as:
rNr 1 = HNr Nu
p
Ptx,Nu Nu sNu 1 + Nr 1
(5.7)
where rNr 1 is the received signal vector with dimension Nr 1, sNu 1 is the
transmit signal vector of joint reception UEs with dimension Nu 1, PNu Nu is
the Nu Nu diagonal matrix of UE transmit power, Nr 1 is the Nr 1 inter-cell
interference plus noise vector assumed with Gaussian distribution and variance
2 . The channel matrix H represents the links between joint transmission UEs
5.3 System Modeling of MMSE/MMSE-SIC Receiver for CoMP Joint

Reception
91
and receive antennas, and it can be expressed as:
HNr Nu
h11
h21
h31
h41
..
.
hNr 1
h12
h22
h32
h42
..
.
..
.
h1Nu
h2Nu
h3Nu
h4Nu
..
.
hNr 2
hNr Nu
(5.8)
where hji denotes the complex channel gain from the transmit antenna of UE i
to the coordinated eNB receive antenna j.
After the FFT block as shown in Figure 5.3, the received signal rNr 1 is fed
into the MUD block, which performs the separation of joint reception UEs. For
the MMSE detector, the knowledge of channel transfer function is required and
practically can be estimated from the RS inserted in predefined sub-carrier positions at each UEs transmit antenna. In this study perfect channel knowledge is
assumed to be known at the CoMP receiver end. The weight of MMSE detector
W can be expressed as [118]:
WNu Nr
H
2
HH
Nr Nu HNr Nu HNr Nu + INr
1
(5.9)
where () denotes the Hermitian transpose operator and INr is the Nr Nr

identity matrix. The Nu Nr weight matrix W can be depicted as:
w11
w21
..
.
WNu Nr =
wNu 1
w12
w22
..
.
w13
w23
..
.
w14
w24
..
.
..
.
w1Nr
w2Nr
..
.
wNu 2
wN u 3
wN u 4
wNu Nr
(5.10)
where w represents the weight coefficient.

Combining the Equation 5.7 and Equation 5.10, the estimate of transmit signal
vector, b
s, can then be expressed as:
b
s =
=
WNu Nr rNr 1
WNu Nr HNr Nu
(5.11)
p
Ptx,Nu Nu sNu 1 + WNu Nr Nr 1
(5.12)
92
Take the joint reception scenario in Figure 5.2(b) as an example, with two joint
reception UEs Nu = 2 and two coordinated eNBs Nr = 4 (2 antenna per eNB),
the estimate of transmit signal of UE-A, sc
A , can then be expressed as:
sc
A
= wTA,1Nr rNr 1
= w
T
A,1Nr
hA,Nr 1
(5.13)
p
Ptx,A sA + w
T
A,1Nr
hB,Nr 1
Ptx,B sB
+wTA,1Nr Nr 1
(5.14)
where wTA,1Nr = [wA1 wA2 zA3 wA4 ] is the Ath row of WNu Nr as presented in
T
Equation 5.10 and hA,Nr 1 = [h1A h2A h3A h4A ] is the Ath column of HNr Nu
T
as presented in Equation 5.8. () denotes the Transpose operator.
According to Equation
5.14, the

2 received signal power of UE-A can be repre . The received mutual interference power from
sented by Ptx,A wT
h
A,N
1
A,1Nr
r

2
. And the
the joint pairing UE-B can be expressed by Ptx,B wT
A,1Nr hB,Nr 1
received inter-cell
h interference from the U E 6=iA&B plus noise power can be
H
H
T
and
= wTA,1Nr 2 wTA,1Nr
denoted by E wTA,1Nr Nr 1 H
Nr 1 wA,1Nr
it is assumed to be white Gaussian distributed. E [] represented the expected
value.
Considering on one PRB, the UE-A MMSE output SINR, mmse,A , can be
represented as:
mmse,A
2

Ptx,A wT
A,1Nr hA,Nr 1
=

2
H
2
T
+ wT
Ptx,B wT
A,1Nr hB,Nr 1
A,1Nr wA,1Nr
(5.15)
With NPRB assigned PRBs for the whole MAC transport block, the MMSE
mac
output SINR of UE-A, mmse,A
, can then be represented as:
Ptx,A
NPRB
mac
mmse,A
=
NX
PRB
T

wA,p,1N hA,p,Nr 1 2
r
p=1
NX
PRB
p=1
2
H
Ptx,B T
2
T
wA,p,1Nr hB,p,Nr 1 + wT
A,p,1Nr wA,p,1Nr
NPRB
(5.16)

Reception
5.3.2
93
Non-Linear MMSE-SIC Detector
For the linear MMSE detector, it assumes that linear combiner output associated to different users are corrupted only by Additive White Gaussian
Noise (AWGN). In fact the linear combiner output signals also contain residual interference, which is not Gaussian distributed [119]. Therefore, further
performance improvement of MUD can be made by using non-linear detector.
SIC is a popular non-linear MUD technique which has been widely investigated
[120]. It serially cancels the interfering user signals from the outputs of linear
detector in order of decreasing power. Compared with the Parallel Interference
Cancellation (PIC)1 , SIC performs better for systems without power control.
By removing the strong interference user signals in an iterative SIC manner,
the weak user signals can be recovered and will see a tremendous signal gain
from the interference reduction [120]. In this UL CoMP study, no power control
is applied to the UEs across the coordinated multiple cells. Therefore, SIC is the
better scheme to be used. As presented in the following section, the MMSE-SIC
detector is investigated with the CoMP joint reception.
MMSE-SIC detector is an iterative process of MMSE detector. It progressively
reduces the mutual interference of joint reception UEs by including the estimate
of transmitted data sequences from the previous users/iterations in the detection
process. As presented in [118] and shown in Figure 5.4, Turbo decoder can be
utilized jointly with the MUD detector in an iterative-loop. The output of Turbo
decoder provides an estimate of coded bit in the form of log-likelihood ratios,
which are then interleaved and soft-modulated in order to obtain an estimate of
the transmitted symbols. The soft estimates are then fed back to an interference
canceller. Starting with subtracting off the strongest joint reception user signal
from the rest of joint reception user signals in a successive fashion, the mutual
interference contributions are progressively removed and the reliability of data
estimates for each joint reception user is iteratively enhanced.
Considering the same system model as defined in Section 5.3.1, the output of
interference canceller block for the UE i on the PRB p at l -th iteration can be
written as:
rli,p,Nr 1 = rp,Nr 1 HZ{i},p,Nr 1
Ptx,Z{i}b
cm
Z{i},p,Nr 1
(5.17)
where Z = {1, 2} is the set of joint reception UE index, HZ{i} represents

1 PIC is a non-linear MUD technique. It parallely cancels the interfering user signals from
the output of linear detector and performs better for systems with power control, which ensures
equal receive power of all users and all signals can be detected simultaneously [120].
94
Figure 5.4: MMSE-SIC Structure
the column of H where the (Z {i})-th joint reception UE has been mapped.
Ptx,Z{i} is the transmit power of (Z {i})-th joint reception UE. b
cm
Z{i} is
the soft estimate symbols of the (Z {i})-th joint reception UE, which obtained
from the iteration m = l if (Z {i})-th joint reception UE has the highest SINR
or from the previous iteration m = l 1 if (Z {i})-th joint reception UE has
not the highest SINR.
After the interference cancellation, the residual error should be taken into consideration in the MUD block as shown in Figure 5.4. The output of MUD block
for the UE i on the PRB p at l -th iteration can then be represented as:
h
i
l
H
2
l
rli,p,M U D = HH
H
Q
H
+
I
p,Nr Nu Nu Nu p,Nr Nu
i,p,Nr 1
Nr ri,p,Nr 1
(5.18)
where Ql = diag [q1 , , qNu ] is the Nu Nu diagonal matrix of the residual

interference powers conditionally to the interference cancellation process, whose

Reception
95
k -th element can be denoted as:
if k = i
1,
2
bZ{i},l
if k 6= i, UE k has not the highest SINR
qk = 1
2
1
bZ{i},l1 , if k =
6 i, UE k has the highest SINR
(5.19)
2
2
where
bZ{i},l1
/b
Z{i},l
is the variance of soft modulated symbols obtained
after decoding of (Z {i})-th joint reception UE at (l 1)-th/l -th iteration.
It should be noted that, when no priori information is available at the start of

SIC iteration,
b2 = 0 and the covariance matrix Q is an identity matrix. The
Equation 5.18 acts as a linear MMSE detector. Once the priori information
is available, the covariance matrix Q is updated at each iteration with the new
symbol variance new , which will be used for the next UE or iteration prediction.
The output SINR of MUD block for UE i on the PRB p at l -th iteration can
be expressed as [94]:
l
sic,i,p
=
l
Ptx,i i,p
(5.20)
l
1 i,p
l
The factor i,p
represents the equivalent channel gain for UE i on the PRB p
at l -th iteration and can be expressed as:
l
l
H
2
1
i,p
= HH
Hi,p,Nr 1
i,p,Nr 1 (Hp,Nr Nu Q,Nu Nu Hp,Nr Nu + INr )
(5.21)
In this study, the symbol variance and instantaneous BLER are predicted at
system-level simulator without performing detailed link-level processing steps.
A fast prediction method based on Gaussian Approximation (GA) is proposed in
[94] for the iterative MMSE-SIC multiuser MIMO joint decoding. This method
exploited a compressed SINR scalar value, which is given by:
l
sic,i,compress
=
Mi1
1
NP RB
NX
P RB
!
l
Mi sic,i,p
(5.22)
p=1
where Mi is defined in [94] which is the Average Mutual Information (AMI)

associated to the constellation of user i in a Gaussian case.
96
As shown in [121], under GA, the compressed SINR is assimilated to the SNR of
a Gaussian transmission model. Therefore the compressed SINR can be used to
find an estimate of the symbol variance and BLER for the relevant MCS by using
the basic AWGN link-level-performance curves. It has shown in [122] that this
prediction model is able to provide a reliable estimate of decoder performance.
The reference AWGN curves used in this study are shown in Figure 5.5. When
c2
the SINR is high the estimate becomes more and more accurate, and therefore
approaches 1. In that case, the mutual interference can be completely removed
c2 is
from the weak UE signal. The term in the diagonal of the Q matrix, 1
then equal to 0 and this means that the residual error concerning the estimate
of the strong UE is not taken into account anymore in the MUD detector, since
the contribution of the strong UE has been completely removed.
Figure 5.5: Compressed SINR to Symbol Variance Mapping with

6 PRBs, 2 antennas, 10 MHz bandwidth at 2 GHz carrier frequency [117]
In Table 5.1, the prediction and updating procedure of symbol variance for
the MMSE-SIC detector are shown. For the number of Nl SIC iterations and
Nu joint reception UEs, the joint reception UEs are first detected when no a
priori information is available. This detection is corresponding to the linear
MMSE detector with a identity symbol covariance matrix Q. Afterward, the
joint reception users are ordered according to the SINR values. The SIC iteration
start from the highest SINR users. By computing the instantaneous SINR
5.4 LTE RS Modeling for CoMP Applications
97
Table 5.1: Link-to-System Performance Mapping of Symbol Variance for

MMSE-SIC Detector
Prediction and Updating Procedure:
Initialize the symbol covariance matrix Qli to identity matrix
(corresponding to no a priori information)
for l = 1 to l = Nl
Order the UEs in descend based on the SINR value
for i = 1 to i = Nu
l
Step-1 : Compute the instantaneous SINR sic,i,p
according to
Equation 5.20
l
l
Step-2 : Compress the SINRs of sic,i,p
into sic,i,compress
using
Equation 5.22
Step-3 : Based on the compressed SINR value, read the symbol
variance new from Figure 5.5 associated to the UEs MCS
Step-4 : Update the symbol covariance matrix Qli with new symbol
variance new = Qli+1 used for the next user prediction
end for
l
Update the symbol covariance matrix Ql+1
i=1 = Qi+1 for the next
iteration prediction
end for
of all the streams for a single user, the compressed SINR is used to find the
symbol variance from the AWGN curve shown in Figure 5.5. By updating the
corresponding symbol variance in the covariance matrix, the updated covariance
matrix is used for the detection filter computation to the next user/iteration.
5.4
LTE RS Modeling for CoMP Applications
In the UL LTE as specified in Rel8, there are two RSs, which are called DeModulation Reference Signal (DMRS) and SRS [22]. DMRS is used for channel
estimation which is needed for coherent detection and demodulation, and SRS
is used to provide information on UL channel quality. For the Physical Uplink Shared Channel (PUSCH), DMRS has the same bandwidth as the UL data
transmission and occupies the 4-th SC-FDMA symbol for each UL slot. Whereas
SRS has larger/potentially much larger transmission bandwidth and is less often
transmitted than the DMRS [15].
The DMRS transmissions are orthogonal among the Intra-cell UEs in Rel8
LTE. Interference can be mitigated because the UEs are transmitted on the
98
assigned PRBs within a cell [15]. However, for the CoMP application as shown
in Figure 5.2, CoMP UE pairs will be simultaneously transmitting with the
same PRBs across cells and estimated at separated CoMP eNBs. In order to
avoid the high interference between the DMRS among CoMP UEs, the support
of multi-cell orthogonal DMRS transmissions is required in the LTE-A. For the
SRS transmission, there is a possibility to sound from multiple intra-cell UEs
within the same frequency band by using orthogonal code in LTE Rel8 [22].
But for the CoMP application, it is also required to extend the simultaneous
orthogonal sounding for the cross-cell UEs within the same frequency band.
In this study the DMRS is not modeled explicitly by assuming the perfect channel knowledge estimated at the CoMP eNBs. So the full channel knowledge of
joint detection users is assumed for the CoMP MUD detection as described in
Section 5.3. The orthogonal SRSs are assumed for the UEs not only within a cell
but also between the cross cell joint reception UEs. Therefore, the mutual interference of joint reception UEs is ideally removed in the weight calculation shown
Equation 5.9 because of orthogonality. In practice, the multiple orthogonal SRS
can be generated by using either cyclic shift of a RS sequence or different RS
sequences [22].
5.5
The main simulation assumptions for UL CoMP joint reception study are shown
in Table 5.2:
As it can be seen from Table 5.2, the interference-limited Macro Case-1 scenario
is still the main focus in this study. Based on the previous antenna tilting
study, the 3-D antenna pattern with 70o horizontal and 10o vertical patterns
is considered. The eNB antenna is downtilted at the optimal angle of 14o as
investigated in Section 3.5.1, and it also achieves the optimum performance for
the CoMP macro diversity study as presented in Section 4.7.3. In order to
obtain a certain amount of CoMP UEs for the accurate statistical analysis, 30
UEs per cell are generated.
The maximum number of CoMP cells is limited to 2 cells (1 serving cell +
1 coordination cell) in this study. Therefore, with 2 antennas per cell, there
are in total 4 receive antenna at the CoMP eNB side. The CoMP UEs are
selected based on the RSRP measurement as described in Section 4.2 and the
selection criteria is defined in Equation 4.1. The selection window size of 3 dB
is considered, which is also a practical value used in the WCDMA soft handover
scheme [123].
99
Table 5.2: Simulation Assumptions of UL CoMP Joint Reception

Parameter
Deployment Scenario
Number of UEs
OLPC P0 and value
CLPC Pof f set
BS Antenna Pattern
Receiver Type
Number of Joint Detection Cells
PS Algorithm
Coordinated PS Algorithm
CoMP UE Selection
CoMP Scenario
Assumptions
30 UEs/cell
-58 dBm and 0.6/-106 dBm and 1.0
-1,-2,-3 or -4 dB
Horizontal HPBW 70o
Vertical HPBW 10o
14o
MMSE/MMSE-SIC
Maximum 2 cells
(1 serving cell + 1 coordination cell)
Proportional Fair
Random Pairing
RSRP-based, 3 dB window size
Intra-Site, Inter-Site, Case-1, Case-3
No specific CPS algorithm is applied in this study, which means all UEs are allocated respectively by their own serving cell packet scheduler and the selected
CoMP UEs are then paired with the corresponding co-channel UE scheduled in
the coordination cell to perform the CoMP joint detection. The two joint detection UEs can also be considered as paired by a random-fashion PS algorithm.
In each serving cell the default PF scheduling algorithm is used as described in
Section 4.6.
Both open-loop FPC parameter with = 0.6 and = 1.0 are investigated in
this study for the CoMP joint reception, where the case of = 0.6 is the main
focus and presented in detail. A simple CLPC scheme is considered in this
study by applying the Pof f set to the joint reception UEs. The Pof f set is equal
to -1,-2,-3 or -4 dB in each study case.
The case of no CoMP with 14o antenna downtilting is utilized as the baseline in
this study. Instead of the ideal IC with MRC reception investigated in the previous study, as shown in Figure 5.6, the realistic MMSE/MMSE-SIC detection
of MRC macro diversity is used as a reference to MMSE/MMSE-SIC CoMP
joint reception study.
This study is carried out in both Intra-Site and Inter-Site scenarios as described
in Section 4.6. The Intra-Site scenario limits the UL CoMP reception within
the same site cells and the Inter-Site scenario allows the UL CoMP reception
between any neighboring cells, which is also called unlimited Inter-Site scenario
in this study. So naturally, the Intra-Site scenario is included as part of the
100
(a) MMSE Macro MRC
(c) MMSE Joint
(b) MMSE-SIC Macro

MRC
(d) MMSE-SIC Joint
Figure 5.6: MMSE/MMSE-SIC MRC Macro Diversity versus MMSE/MMSESIC CoMP Joint Reception
unlimited Inter-Site scenario. The number of selected CoMP UEs with 3 dB

window size are shown in Figure 4.6 for both Intra-Site and unlimited Inter-Site
scenario. As it can be seen, there are about 8% of CoMP UEs1 for the Intra-Site
scenario and 22% of CoMP UEs for the unlimited Inter-Site scenario.
In order to further analyze the CoMP joint reception in the unlimited Inter-Site
scenario, the unlimited Inter-Site scenario is decomposed into three limited cases
as shown in Figure 5.7. The Figure 5.7 only highlights a set of coordination cells
for illustration purpose, and the full picture of the CoMP network layout can be
seen in Figure B.1. The decomposed unlimited Inter-Site scenarios are restricted
to 3 neighboring cell coordinations, where the Case-1 scenario is identical to the
1 CoMP UE: is defined as the UE located at the CoMP cooperation area, which was called
macro diversity UE in the previous macro diversity study in Chapter 4.
(a) Case-1
101
(b) Case-2
(c) Case-3
Figure 5.7: Three Limited Cases of Unlimited Inter-Site Scenario

Intra-Site scenario, the Case-2 scenario is the Inter-Site coordination with 2-Site,
and the Case-3 scenario is the 3-site Inter-Site coordination. It is worth pointing
out that these 3 cases are the general description of network coordination. Any
3 neighboring cell CoMP coordinations are covered by those 3 cases, i.e. the
coordination of cells 1, 9 and 11 as shown in Figure B.1 is actually corresponding
to the Case-2 scenario with 2 cells belonging to the same site and the third cell
from the neighboring site.
The number of selected CoMP UEs for the limited Inter-Site scenarios are shown
in Table 5.3 compared with the Intra-Site and unlimited Inter-Site scenario. As
expected, the Case-1 scenario is identical to the Intra-Site case with 8% CoMP
UEs. The Case-2 and Case-3 scenario have about 10% and 8% CoMP UEs
respectively.
Table 5.3: Number of Selected CoMP UEs
Scenario
Intra-Site
Unlimited Inter-Site
Case-1 (Limited Inter-Site)
Percentage of CoMP UEs

in total UEs
8%
22%
8%
10%
8%
The path gain distributions of CoMP UEs in the three limited Inter-Site scenarios are shown in Figure 5.8(a). As it can be seen, the CoMP UEs in the
Case-1 scenario have the highest path gain distribution, because the selected
102
(a) Path Gain Distribution
(b) Transmit Power Distribution
Figure 5.8: Comparison of Limited Inter-Site Scenarios of CoMP UEs

CoMP UEs are located at the close-by Intra-Site cell overlapping area as shown
in Figure 5.9(a) and have a good signal quality to the eNB. For the Case-3
scenario, the CoMP UEs have the lowest path gain distribution, because the
selected CoMP UEs are mainly located at the far-away cell-border region, as
shown in Figure5.9(b), with a high signal attenuation. As shown in Figure 5.6,
the CoMP UEs in the Case-2 scenario contain both Intra-Site and cross-site
coordination, which results in the path gain distribution of CoMP UEs between
Case-1 and Case-3 scenarios as shown in Figure 5.8(a).
In Figure 5.8(b), the transmit power distribution of CoMP UEs is shown. Based
on the open-loop FPC as presented in Section 2.5.1 and as expected, the Case3 scenario with low path gain CoMP UEs have the highest transmit power
distribution, and about 20% of CoMP UEs operated at maximum power. The
Case-1 scenario with high path gain CoMP UEs has the lowest transmit power
distribution which has about 10% of CoMP UEs transmitted at saturated power.
The curve of mixed Case-2 scenario is between Case-1 and Case-3 scenarios.
For the rest of the study, the Case-1 and Case-3 are going to be investigated
and presented in detail.
(a) Case-1 CoMP UE Location
103
(b) Case-3 CoMP UE Location
Figure 5.9: Location of CoMP UEs in Limited Case-1 and Case-3 Scenarios
5.6
With open-loop FPC parameter = 0.6 and P0 = 58 dBm, the performance

comparison of CoMP joint reception with CoMP macro diversity reception is
first presented under the Intra-Site and unlimited Inter-Site scenarios. The
detailed investigation of joint reception are performed afterwards for the Case-1
and Case-3 scenarios.
5.6.1
Comparison of CoMP Joint and Macro Diversity

Reception
The comparison of CoMP joint and macro diversity reception are conducted
in both Intra-Site and unlimited Inter-Site scenarios with open-loop FPC of
= 0.6 and P0 = 58 dBm, where the macro diversity with MRC scheme is
used as reference.
In Figure 5.10, the SINR distribution of CoMP UEs per TTI are shown for the
case of MMSE-SIC receiver. As expected, with better diversity of wireless channels and higher array gain by 4 CoMP eNB antennas as shown in Figure 5.6,
the CoMP joint reception has higher SINR per TTI than the macro diversity
104
MRC reception in both Intra-Site and unlimited Inter-Site scenarios. The unlimited Inter-Site scenario also contains the high path loss CoMP UEs located
at boresight cell-overlapping regions as shown in Figure 4.4. So the overall CDF
distribution of SINR per TTI is lower in the unlimited Inter-Site scenario compared with the Intra-Site scenario for both CoMP joint and macro diversity
MRC receptions.
Figure 5.10: SINR per TTI for CoMP UEs in Intra-Site and Unlimited Inter-Site
Scenarios with = 0.6 and P0 = 58 dBm
In Figure 5.11, the average SINR distribution of unlimited Inter-Site CoMP
UEs with CoMP joint reception are shown. The Non-CoMP case is utilized as
a reference, because the Intra-Site scenario is included as part of the unlimited
Inter-Site scenario, and there are only about 8% of CoMP UEs in the IntraSite scenario stated in Table 5.3. As expected, the average SINR distribution
of unlimited Inter-Site scenario shows better performance than the Intra-Site
scenario, and at high average SINR region the MMSE-SIC receiver outperforms
the MMSE receiver in both scenarios.
The throughput gain plot over the baseline, no CoMP case, is shown in Figure
5.12. The CoMP MRC macro diversity reception with MMSE and MMSE-SIC
receivers are utilized as a reference. To compare with the previous MRC receiver
with ideal IC cases as shown in Figure 4.9, the throughput gain of CoMP MRC
macro diversity reception for the realistic MMSE and MMSE-SIC receivers is
105
Figure 5.11: Average SINR of unlimited Inter-Site CoMP UEs in Intra-Site and
Unlimited Inter-Site Scenarios with = 0.6 and P0 = 58 dBm
Figure 5.12: Intra-Site and Unlimited Inter-Site Scenarios with = 0.6 and
P0 = 58 dBm
106
expected to be lower because of partial interference removal.

As it can be seen in Figure 5.12, considering the CoMP joint reception with
both MMSE and MMSE-SIC receivers, it has better performance than the corresponding CoMP macro diversity reception in both 5% outage and average
UE throughput. For the best case with MMSE-SIC receiver, there is about 6%
outage gain and 4% average UE throughput gain in the Intra-Site scenario and
about 16% outage gain and 8% average UE throughput gain in the unlimited
Inter-Site scenario.
The small improvement in the Intra-Site scenario is very limited by the number
of CoMP UEs and the weak coordination link of the joint reception pair, i.e. the
link B1 as shown in Figure 5.2(b). The improvement of the weak coordination
link of the joint reception pair can be made by using multi-cell coordinated
packet scheduling presented in Chapter 6.
In order to further investigate the performance of CoMP joint reception in the
unlimited Inter-Site scenario, the limited Inter-Site scenarios with Case-1 and
Case-3 are presented in detail in the following section.
5.6.2
CoMP Joint Reception in Limited Inter-Site Scenarios
Based on the description of Case-1 and Case-3 presented in Section 5.5, in Figure
5.13, the average SINR distribution of CoMP UEs is shown in the limited Case1 and Case-3 scenarios. The selected CoMP UEs are different in the Case-1
and Case-3, two baseline curves for Non-CoMP case are plotted. The selected
CoMP UEs in the Case-1 scenario have higher path gain distribution than the
Case-3 scenario as shown in Figure 5.8(a). As expected, the averaged SINR
distribution of selected CoMP UEs in the Case-1 scenario also have about 2 dB
higher than the Case-3 scenario.
The throughput gain plot over the baseline, no CoMP case, is shown in Figure
5.14. In the Case-1 scenario, the throughput gain plot is exactly identical to
the Intra-Site scenario as shown in Figure 5.12. To compare with the Case-3
scenario, which also contains about 8% of CoMP UEs as presented in Table
5.3, it shows 5% higher outage gain in the Case-3 scenario than in Case-1. The
reason is that, the Case-3 scenario has about 26% of 5%-outage UEs locate at the
CoMP area and selected as the CoMP UEs as shown in Table 5.4, whereas the
Case-1 scenario has only about 11% of 5% outage UEs within the CoMP area.
The CoMP reception improves the performance of CoMP UEs and consequently
results in higher 5% outage gain in the Case-3 scenario, where the performance
Figure 5.13: Average SINR Distribution in Limited InterSite Scenarios with = 0.6 and P0 = 58 dBm
Figure 5.14: Limited Inter-Site Scenarios with = 0.6 and

P0 = 58 dBm
107
108
Table 5.4: Number of CoMP UEs out of the 5%-outage UEs

at Non-CoMP Case with 14o Downtilting
Scenario
Intra-Site
Inter-Site (Unlimited)
Percentage of CoMP UEs

in the overall 5%-outage
11%
55%
11%
22%
26%
gain in the Case-1 scenario can mainly be seen in the average user throughput.
It can also be seen that from Figure 5.14, to compare with the MMSE type
of receiver, the MMSE-SIC receiver is more effective in the Case-3 scenario in
terms of 5% outage throughput, because the cell-edge UEs with low SINR can
gain more benefit from the interference cancellation than the high SINR UEs.
Comparing the CoMP joint reception with CoMP MRC macro diversity reception for both Case-1 and Case-3, the CoMP joint reception performs more
efficiently in the Case-1 scenario with high SINR as shown in Figure 5.13. For
the best case of MMSE-SIC receiver, the joint reception has about 5% gain in
the 5% outage and 2% gain in the average user throughput over the MRC macro
diversity reception. However, for the Case-3 scenario with low SINR, the joint
reception only has 0.5% gain in the 5% outage and about 1% gain in the average
user throughput over the MRC macro diversity reception.
By considering both the performance gain and capacity requirement of LTE
X2-interface, the CoMP joint reception is a good choice for the Case-1/IntraSite application. It can provide better overall performance than the CoMP
macro diversity reception and without the need of large capacity X2-interface
coordination. Whereas, the CoMP macro diversity reception with MMSE-SIC
receiver is a good candidate for the cross-site, Case-3, type of application. It can
give the sub-optimal performance gain close to the application of CoMP joint
reception and demand relatively lower capacity requirement of X2-interface as
discussed in Section 5.2.
5.6.3
CoMP Joint Reception with = 1.0
Based on the previous study in Section 3.5.2, the performance of CoMP joint
reception with open-loop FPC parameter = 1.0 and P0 = 106 dBm has also
109
been investigated in this study.

As shown in Figure 5.15, the average SINR distribution of unlimited Inter-Site
CoMP UEs is shown by using the case of no CoMP with = 0.6 as a reference.
As expected, compared with = 0.6, the utilization of = 1.0 optimizes the 5%
outage cell-edge UE throughput by compromising the performance of cell-center
UEs. The unlimited Inter-Site scenario still outperforms the Intra-Site scenario
because large number of CoMP UEs are selected, and the MMSE-SIC receiver
in both scenarios gives better gain than the MMSE receiver.
Figure 5.15: Average SINR of Inter-Site CoMP UEs in IntraSite/Inter-Site Scenarios with = 1.0 and P0 = 106 dBm
In Figure 5.16, the throughput gain plot over the no CoMP case with = 1.0
is presented. The cases of CoMP macro diversity with MRC reception are still
used as a benchmark for the CoMP joint reception. Comparing with = 0.6,
the MMSE-SIC receiver has shown marginal gain over the MMSE receiver in
both Intra-Site and unlimited Inter-Site scenarios. This is due to the fact that,
by using = 1.0, more cell-center UEs are transmitted with less power as shown
in Figure 5.17, which results in low SINR as well. For those cell-center UEs with
low SINR and being the pair of low SINR joint reception CoMP UEs, both UEs
are quite hard to be detected and demodulated, which make it ineffective for
the interference cancellation in order to improve the overall performance.
110
Figure 5.16: Intra-Site and Inter-Site Scenarios with =

1.0 and P0 = 106 dBm
Figure 5.17: UE Transmit Power Distribution of No CoMP

scenario with = 0.6 vs. = 1.0
111
To compare the CoMP joint reception with MRC macro diversity reception,
for the best case with MMSE-SIC receiver, there is about 3% and 2% gain for
Intra-Site scenario and about 5% and 6% gain for unlimited Inter-Site scenario
in terms of outage and average user throughput respectively.
5.6.4
CoMP Joint Reception with Close-loop FPC
As presented in Chapter 4, the application of close-loop FPC together with the

IC-based macro diversity reception can provide further enhancement in both
cell-edge and average user throughput compared with the standing alone ICbased macro diversity reception. In this study, the close-loop FPC scheme is
also investigated for the CoMP joint reception with realistic MMSE/MMSE-SIC
receiver.
In general, the exploited close-loop FPC scheme is to lower the transmit power
of CoMP UEs, who normally generate high inter-cell interference to the other
UEs, especially to the non-CoMP UEs. Because of less inter-cell interference
received, the performance of non-CoMP UEs can be enhanced and the negative
effect of powering down CoMP UEs is compensated by the CoMP joint reception
gain.
The combination of CoMP joint reception and CLPC scheme with MMSE-SIC
receiver has been evaluated under the unlimited Inter-Site scenario with the case
of = 0.6 and = 1.0 as shown in Figure 5.18 and Figure 5.19 respectively.
The case of No CLPC presented in the previous Figure 5.12 and Figure 5.16 is
utilized as the baseline for the CLPC cases. The CoMP MRC macro diversity
reception is utilized as a reference to the CoMP joint reception.
Considering the CoMP joint reception with CLPC, for the case of = 0.6,
the 5% outage throughput gain is optimized at -1 dB. To compare with the no
CLPC, it can provide further improvement of 4% gain in terms of 5% outage
and only 1% gain in terms of average user throughput. Futher transmit power
reduction of CoMP UE can still improve the average user throughput but with
dramatically reduced cell-edge UE performances. For the case of = 1.0, with
the optimal reduction at -2 dB, the application of CLPC together with CoMP
joint reception gives about 6% and 4% gain in terms of 5% outage and average
user throughput respectively over the no CLPC.
In Figure 5.20, the performance of CLPC for MMSE-SIC receiver under the
Intra-Site scenario is shown for both CoMP joint and MRC macro diversity reception with = 0.6 and P0 = 58 dBm. As it can be seen, by reducing the
transmit power of CoMP UEs, the Non-CoMP cell-edge UEs, who cannot get
112
Figure 5.18: Inter-Site Scenarios for Macro vs. Joint SIC Receiver
Figure 5.19: Inter-Site Scenarios for Macro vs. Joint SIC Receiver
5.7 Conclusions
113
Figure 5.20: Intra-Site Scenarios for Macro vs. Joint SIC Receiver
benefit from the CoMP reception, will get more opportunities to be scheduled
in the PF frequency domain and further improve the overall 5% outage gain.
Considering the CoMP joint reception, for the best case with optimal reduction
of -3 dB, there is about 11% gain and 5% gain over the baseline, Non-CoMP
case, in terms of 5% outage and average user throughput respectively. To compare with the no CLPC case, there is a further enhancement of cell-edge user
throughput by 5%.
5.7
Conclusions
In this chapter, the CoMP has been studied in the UL LTE in the form of joint
reception. This study is carried out with MMSE/MMSE-SIC receiver and the
investigations are performed in both Intra-Site and unlimited Inter-Site CoMP
scenarios, where the unlimited Inter-Site scenario is further divided into Case1 and Case-3 for detailed analysis. The open-loop FPC with = 0.6 and
P0 = 58 dBm is the main focus. The scheduled users are allocated by the
packet scheduler performed in each individual cell and the selected CoMP UEs
are paired with the corresponding co-channel UE scheduled in the coordination
114
cell to perform the CoMP joint detection. The two joint detection UEs can also
be considered as paired by a random-fashion/independent PS algorithm.
Figure 5.21: Macro vs. Joint SIC Receiver under the Intra- and
Inter- Site Scenarios with = 0.6 and P0 = 58 dBm
As a sum-up, the throughput gain plot over the no CoMP case is shown in Figure
5.21 with the MMSE-SIC receiver. As it can be seen in both Intra-Site and
unlimited Inter-Site scenario, the CoMP joint reception outperforms the CoMP
macro diversity with MRC reception, and the application of CLPC can further
improve the overall performance. In the unlimited Inter-Site scenario, there is
about 20% and 10% gain in terms of 5% outage and average user throughput
respectively. Under the Intra-Site scenario, the performance of CoMP joint
reception shows marginal gain in terms of both 5% outage and average user
throughput. For the best case, there is about 10% and 5% gain in terms of
outage and average user throughput respectively. The small improvement in
the Intra-Site scenario is very much limited by the number of selected CoMP
UEs.
Chapter
Coordinated Packet
Scheduling for CoMP Joint
Reception
In Chapter 5, the performance of advanced (MMSE and MMSE-SIC) receiver
for the CoMP application is investigated in the system-level without considering
the multi-cell Coordinated Packet Scheduling (CPS). Under the CoMP IntraSite scenario, the performance of CoMP joint reception shows marginal gain in
terms of both 5% outage and average user throughput.
This chapter is the extension investigation of CoMP techniques from the previous chapter. The gain potential of multi-cell CPS is investigated under the
Intra-Site scenario, where a cluster of Intra-Site cells jointly allocate the served
UEs. The exploited CPS algorithm is based on the traditional PF scheduler and
the orthogonality requirement is considered to pair the joint detection users.
This study will focus on the joint MMSE-SIC detection which gave the optimal
performance as presented in Section 5.3.2.
In Section 6.1, the multi-cell CoMP CPS is briefly introduced and the targeted
study scenario is described. In Section 6.2, the studied multi-cell CPS algorithm
is presented. In Section 6.3, the modeling assumption of the study is described
and the performance evaluation results are presented in Section 6.4. In Section
6.5, the conclusion for the CoMP joint reception with CPS is made.
116
6.1
Coordinated Packet Scheduling for CoMP Joint Reception
Multi-cell Coordinated Packet Scheduling
For the CoMP studies presented in Chapter 5, the Independent Packet Scheduling
(IPS) was assumed, where each cell schedules its own serving UEs without considering the UE allocation in the other neighboring cells. To perform the CoMP
joint detection, the allocated CoMP users were considered as paired in a random fashion with the co-channel user scheduled in the coordination cell. In
this chapter, the potential of multi-cell Coordinated Packet Scheduling (CPS) is
investigated together with the CoMP joint detection. By coordinating different
cell sites in terms of the scheduling decision (PRB allocation) in the frequency
domain, better joint detection pair could be allocated together to fully explore
the CoMP multiuser detection gain. In general, the multi-cell CPS can be seen
as an extension of the Inter-Cell Interference Coordination (ICIC) scheme already presented in the LTE network.
This study is focus on the Intra-Site CoMP, where the coordination cells are
belong to the same BS site as described in Section 4.1. Compared with the
Inter-Site CoMP, where the coordination cells are belong to different sites, it
exhibits several advantages. First, the Intra-Site CoMP is not limited by the
existing LTE network backhaul discussed in Section 5.2 since the coordination
take place within the same physical location. The Intra-Site coordination cells
can be coordinated via the equipment internal bus transfer. Therefore, there
are basically no restrictions regarding to the amount of exchanged information
and also the additional delay due to the CoMP coordination becomes almost
negligible. Besides, the coordination cells belonging to the same site could be
driven by the same clock, the synchronization of a certain UE to all coordination
cells can be readily achieved. Finally, the intra-site coordination can be realized
with the existing LTE Release 8 network since no backhaul signaling is involved
and hence basically no further standardization is required for that purpose.
6.2
CPS Allocation Algorithm Design
Three-sector BS site is assumed in this study. The exploited CPS entity takes
care of the physical resource allocation for all three intra-site cells. Therefore,
three UEs will be allocated simultaneously on the same transmission frequency
band by the CPS. An example is shown in Figure 6.1, UE-A, UE-B and UE-C
are served respectively by the Cell-1, Cell-2 and Cell-3, which are the coordination cells in one BS site/Intra-Site coordination cells. On a certain frequency
band, UE-A, UE-B and UE-C will be jointly allocated by the CPS.
6.2 CPS Allocation Algorithm Design
117
Figure 6.1: Scenario of CoMP Joint Reception with CPS
There are mainly two roles for the CPS algorithm investigated in this study,
which are pairing and scheduling. The pairing makes sure that the best joint
detection pair could be allocated together to fully explore the CoMP multiuser
detection gain. Two scheduling algorithms are used jointly. The channel aware
scheduling guarantees the QoS and the lowest path gain scheduling ensures
the generated interference of allocated UE is minimized to the joint reception
pair. Both pairing and channel aware scheduling rely on the available channel
information. The pairing relies on the full spatial complex channel matrix (shortterm) whereas the channel aware scheduling relies on the knowledge of SINR
(long-term). In the following, the utilized pairing and scheduling criteria are
defined.
6.2.1
CPS Pairing Algorithm
The theoretical work of [124] [125] shown that the MIMO capacity is influenced
by the spatial correlation. With the independent channel fading between antennas, the capacity of the MIMO system grows proportionally to the minimum
number of transmit and receive antenna, min [Nt , Nr ], for fixed transmit power.
By exploiting the multiuser diversity, the CPS can allocate the joint reception
UE pair with low channel correlation to minimize the mutual interference and
therefore maximize the UL CoMP capacity. In this study, the defined channel
118
orthogonality i,v between UE i and UE v is expressed as:

HH
i , Hv
=1
, 0 i,v 1
kHi k kHv k
i,v
(6.1)
where Hi and Hv are the complex fast fading channel of UE i and UE v, and hi
denotes the dot product. If the channel vector of the two UEs are orthogonal,
the dot product in the nominator is zero and the defined channel orthogonality
i,v between them is 1.0, else is 0.0.
The selection of best pairing UE is based on the maximum orthogonality in this
study. For a CoMP UE i, the selected UE pair v can be expressed as:
v = arg max i,v
(6.2)
i,vV
b(i)6=b(v)
where V is all the users served in one site. i and v are the UEs from the same
site but served in difference cell b(i) and b(v).
6.2.2
CPS Scheduling Algorithms
There are two scheduling algorithms considered in this study, which is the channel aware scheduling and lowest path gain scheduling.
The channel aware scheduling in this study only considered the FDPS with
FTB as presented in Section 2.6. The PF scheduling algorithm is exploited.
In a given serving cell b based on the maximum PF metric, the selected UE i
transmits on PRB p at TTI t can be expressed as:
iPF
b(i),p,t
= arg max
iVb(i)
Tbi,b(i),p,t
T i,b(i),t
!
(6.3)
where Vb(i) is all the users served in cell b(i). Tbi,b(i),p,t is the estimated Layer-1
achievable throughput for UE i, served in cell b(i), on PRB p and at scheduling
instant time t. T i,b(i),t is the past averaged Layer-1 acknowledge throughput
6.2 CPS Allocation Algorithm Design
119
for UE i served in cell b(i) at scheduling instant time t, which is calculated by

using an exponential averaging filter:
T i,b(i),t = (1 )T i,b(i),t1 + Ti,b(i),t
(6.4)
where T i,b(i),t is the Layer-1 acknowledged throughput for UE i served in cell

b(i) at scheduling instant time t and is called the filter coefficient.
The lowest path gain scheduling exploited in this study ensures that the generated interference of allocated UE to the joint reception pair is minimized. With
less received interference or higher received SINR, the CoMP joint detection
gain can be further improved. For an allocated UE pair v and g transmit on
PRB p at TTI t, the scheduled UE i based on the lowest path gain metric can
be expressed as:
iPG
b(i),p,t =
arg min
Li,b(i),p,t
(6.5)
iVb(i)
b(i)6=b(v)6=b(g)
where -L is the measured path gain of UE i in the serving cell b(i) and the UE
i, v and q are served in the Intra-Site cell b(i), b(v) and b(q) respectively.
6.2.3
General Decision Flow of CPS
The general flow of CPS scheduler assumes that scheduling and pairing decisions
are taken every TTI. For a certain frequency band, the decision flow is outlined
as follows and the corresponding flow chart are shown in Figure 6.2:
1. Three UEs (UE-A, UE-B and UE-C ) are selected from each Intra-Site cell
according
to the maximum PF

metric which is presented in Equation 6.3
PF
PF
PF
Ab(A),p,t , Bb(B),p,t , Cb(C),p,t .
2. The selected three UEs are then sorted again according to the PF metric
in descendant order and the output vector PF
cells,p,t can be expressed as:

PF
PF
PF
PF
=
Order
A
,
B
,
C
(6.6)
cells,p,t
b(A),p,t
b(B),p,t
b(C),p,t
PF,1st
where the first UE b(1st),p,t
has the highest allocation priority among the
PF,3rd
three UEs and the third UE b(3rd),p,t
has the lowest.
120
Figure 6.2: Flow Chart of Multi-cell CPS

PF,1st
3. Start from the highest priority UE b(1st),p,t
, check if the UE is a CoMP
UE:
PF,1st
PF,1st
(a) If the UE b(1st),p,t
is a CoMP UE, allocate the UE b(1st),p,t
and
PF,1stPair
find the best joint reception pair b(1stPair),p,t
based on the Equation
6.2. The allocation of last UE is selected based on the Equation 6.5,
which has the lowest generated interference to the already allocated
CoMP pair.
PF,1st
(b) If the UE b(1st),p,t
is not a CoMP UE. But the second highest priority
PF,2nd
PF,2nd
UE b(2nd),p,t
is a CoMP UE. Allocate the UE b(2nd),p,t
and find the
PF,2ndPair
best joint reception pair b(2ndPair),p,t
based on the Equation 6.2. Still
121
by considering the lowest interference to the CoMP pair, the last UE

is allocated based on Equation 6.5.
(c) If none of the UEs are CoMP UEs, then all three selected UEs are
scheduled as expressed in Equation 6.3, which is corresponding to the
independent PS performed in each cell.
Based on the above description, the algorithm generally needs to find two CoMP
PF,1st
PF,2nd
UEs and one non-CoMP UE. The pair of CoMP UE b(1st),p,t
or b(2nd),p,t
can
be selected by using Equation 6.2. If there is no pair found by using Equation
6.2, the algorithm is equivalent to the independent PS performed in each cell.
6.3
The main simulation assumptions for this UL CoMP CPS study are shown in
Table 6.1:
Table 6.1: Simulation Assumptions of Multi-Cell Coordinated PS
Parameter
Deployment Scenario
Number of UEs
OLPC P0 and value
CLPC Pof f set
BS Antenna Pattern
Receiver Type
Number of Joint Detection Cells
CPS Pairing Algorithm
CPS Scheduling Algorithm
CoMP UE Selection
CoMP Scenario
Assumptions
30 UEs/cell
-58 dBm and 0.6
-1,-2,-3 or -4 dB
Horizontal HPBW 70o
Vertical HPBW 10o
14o
MMSE-SIC
Max. 2 cells (1 serving + 1 coor. cell)
Max. Orthogonality
PF or Lowest PG
RSRP-based, 3 dB window size
Intra-Site
(All 3 cells belong to the same site)
This study is still focus on the interference-limited Macro Case-1 scenario. 30

UEs are balance-loaded in each cell with spatially uniform distribution. The
3-D antenna pattern with 70o horizontal and 10o vertical HPBW patterns is exploited. Based on the investigation presented in Section 3.5.1, the eNB antenna
is mechanically downtilted of 14o and the open-loop FPC parameter exploits
= 0.6 and P0 = 58 dBm.
122
The Intra-Site CoMP scenario is the main concentration for the investigated
CPS algorithm. The CoMP UEs are selected based on the RSRP measurement
as described in Section 4.2 and the selection criteria is defined in Equation 4.1.
The CoMP joint reception with MMSE-SIC receiver is considered which gave
the optimal performance in Chapter 5. A simple CLPC scheme is considered
in this study by applying the Pof f set to the joint reception CoMP UEs. The
Pof f set is equal to -1,-2,-3 or -4 dB in each study case.
6.4
In this section, the simulation results of investigated CPS algorithm is presented.

The corresponding study by using IPS, which is presented in the Chapter 5, is
utilized as a reference.
Figure 6.3: SINR per TTI for Intra-Site CoMP UEs - IPS vs. CPS
In Figure 6.3, the SINR of CoMP UEs per TTI distribution is shown. Compared
with IPS, the applied CPS algorithm improves the received SINR of CoMP UE
for every TTI, but the improvement is quite marginal.
The throughput gain plot over the baseline, no-CoMP, case are shown in Figure
6.6. The case of IPS with MMSE-SIC receiver under the Intra-Site scenario
123
Figure 6.4: Throughput Gain Plot of Multi-Cell CPS with CLPC for Joint SIC
reception under the Intra-Site scenario with = 0.6 and P0 = 58 dBm
is shown as a reference, which is presented in Figure 5.12. Compared with

IPS, both 5% outage and average user throughput are further improved when
the multi-cell CPS are utilized. There are about 11% and 7% gain over the
baseline case in terms of 5% outage and average user throughput respectively.
By applying the close-loop FPC to power down the transmit power of CoMP
UEs as described in Section 4.5, the 5% outage user throughput can be further
optimized, and the negative effect of powering down CoMP UEs is compensated
by the CoMP joint reception gain. In Figure 6.4, the combination performance
of multi-cell CPS together with the close-loop FPC is shown. For the best case
of powering down 2 dB, there are about another 3% gain for the cell-edge user
throughput compared with the no CLPC case.
The CPS algorithm shown in Figure 6.2 only considers the pairing of the first
PF,1st
PF,2nd
two highest priority UEs b(1st),p,t
and b(2nd),p,t
. In Figure 6.5, another CPS
algorithm is presented which also considers the pairing of the lowest priority UE
PF,3rd
b(3rd),p,t
. The throughput gain comparison of the two CPS algorithms over the
baseline, no-CoMP, case are shown in Figure 6.6.
Compared with the first CPS algorithm, the second CPS algorithm shows the
reduction of cell-edge user throughput from 11% to about 9%. This is because
124
PF,3rd
Figure 6.5: Flow Chart of CPS: Pairing the lowest priority UE b(3rd),p,t
Figure 6.6: Throughput Gain Plot of Multi-Cell CPS vs IPS for Joint SIC
reception under the Intra-Site scenario with = 0.6 and P0 = 58 dBm
6.5 Conclusions
125
the second CPS algorithm gives too high priority to the CoMP UEs which results
in the non-CoMP cell-edge UEs are less often scheduled. In the CoMP Intra-Site
scenario, most of the non-CoMP cell-edge UEs locate at bore-sight cell-border
region. In order to maintain the performance of those cell-edge users, they need
to be allocated more often by the CPS scheduler. Low prioritization of them in
the CPS domain will influence the overall 5% outage user throughput.
6.5
Conclusions
In this chapter, the multi-cell CoMP CPS has been studied. The CoMP IntraSite scenario with MMSE-SIC joint reception is the main focus. The corresponding MMSE-SIC joint reception with IPS investigated in Chapter 5 is utilized as
a reference.
Figure 6.7: Multi-Cell CPS vs IPS for Joint SIC reception under the IntraSite scenario with = 0.6 and P0 = 58 dBm
As a sum-up, the comparison of multi-cell CPS with IPS is shown in Figure 6.7.
The CPS can provide better throughput gain than the optimal IPS case even
without the application of CLPC. The jointly applying the close-loop FPC with
CPS can make further optimization in the CoMP Intra-Site scenario. For the
best case of powering down 2 dB, there are about 14% and 7% gain over the
126
baseline, no CoMP, case in terms of 5% outage and average user throughput

respectively.
The exploited CPS algorithm in this study has been investigated with the coordination of 3 neighboring Intra-Site cells. There are mainly two roles for the
investigated CPS algorithm, which are pairing and scheduling. The pairing algorithm makes sure that the most orthogonal CoMP pair are allocated together
to fully explore the CoMP multiuser detection gain, and the scheduling algorithm guarantees the QoS and the lowest generated interference to the scheduled
CoMP pair. As presented in this study, in order to maintain the overall performance especially the cell-edge user performance, the user pairing and scheduling
algorithms need to be jointly considered.
Chapter
Main Conclusion and Future

Work Recommendations
The main object of this PhD thesis has been to study the potential multi-cell
Radio Resource Management (RRM) techniques for limiting the impact of intercell interference in the Uplink (UL) LTE and further optimize the overall LTE
network performance. Unlike the asynchronous UL WCDMA network, where
the intra-cell interference and the near-far effect issue are the main interest,
the frequency domain orthogonality of synchronous UL LTE network ideally
removes the intra-cell interference. The inter-cell interference becomes the major
concern with the frequency reuse factor of 1.
The conducted investigations have focused on the traditional mechanical antenna downtilting and the Coordinated Multi-Point (CoMP) feature proposed
in the LTE-A study item. The CoMP techniques in the form of both macro diversity reception and joint reception have been studied. Besides, the multi-cell
Coordinated Packet Scheduling (CPS) has also been studied with the CoMP
joint reception. The other RRM functionalities, such as Adaptive Modulation
and Coding (AMC), Outer Loop Link Adaptation (OLLA) and Hybrid Automatic Repeat reQuest (HARQ) have been considered in the analysis of results
due to the interaction with the focused techniques.
The system-level simulator described in Appendix B has been used as the main
tool for generation of study results. The simulator has been developed as part
of the PhD project together with other colleagues from both Aalborg University
and Nokia Siemens Networks (NSN). Along with the system-level designing, the
development of the simulator has required a significant time investment in terms
of modeling, software designing, implementation and testing. Part of the PhD
study was also spend on the development of dynamic system-level simulator
128
Main Conclusion and Future Work Recommendations
described in Appendix A.5 which followed a different approach than the one
presented in the main report.
The thesis is mainly divided into three parts where the first part addressed
the topic of mechanical antenna downtilting and the interaction with the UL
Fractional Power Control (FPC). The second part studied the CoMP technique
in the form of macro diversity reception. Finally, the third part presented the
UL CoMP joint reception and the coordinated packet scheduler design. In the
following sections, a summary of the whole thesis is presented, and the main
conclusions are drawn for the investigated topics. The recommendations for the
future research are given at the end.
7.1
Antenna Tilting in Homogeneous UL LTE
In Chapter 3, the mechanical antenna downtilting has been investigated in the

UL LTE. A simplified path loss model has been used to demonstrate the effect of antenna downtilting. The evaluations have been performed with the
network-based mechanical antenna downtilting, where all the eNBs are mechanically downtilted with the same angle. The scenarios with different ISD
have been investigated. Being an inter-cell interference mitigation technique,
the mechanical downtilting performs efficiently under the interference-limited
scenario. The optimal antenna downtilting angle for the interference-limited
scenario with ISD = 500 m is 14o . A simple antenna downtilting method has
been proposed by pointing the first notch of vertical antenna pattern to the
boresight cell-edge area, and the cell-edge user throughput can be optimized in
the UL LTE. Compared with the no downtilting case of 0o , there are about
296% gains in terms of cell coverage and about 147% increases in terms of cell
capacity for downtilting angle of 14o . The application of antenna downtilting
is less effective in the noise-limited scenario with ISD = 1732 m. The optimal
antenna downtilting angle is about 2o . There is about 13% gains in terms of cell
coverage and 4% increases in terms of cell capacity respctively compared with
the no downtilting case of 0o .
The UL power control is an important technique to achieve the good UL LTE
performance. The interaction of antenna downtilting together with UL openloop FPC has also been analyzed in the interference-limited scenario. The study
results have shown that the impact of mechanical antenna tilting to the chosen
of open-loop FPC parameters can be neglected.
Furthermore, the antenna tilting is very sensitive to the narrow vertical antenna
beamwidth. With narrow beam, both cell-edge user performance and average
7.2 UL CoMP in the Form of Macro Diversity Reception
129
user throughput are decreased dramatically after the optimal downtilting angle.
By utilizing wider vertical antenna pattern, the overall network performance
can be maintained over a certain downtilting angle range. Practically, it can
also tolerate some extent of downtilting bias by installation or tuning defects.
In 3GPP application, the vertical antenna pattern with HPBW equal to 10o is
standardized.
7.2
UL CoMP in the Form of Macro Diversity

Reception
CoMP is an advanced technique for interference mitigation which has been proposed in the LTE-A as one of the features to further reduce the impact of
inter-cell interference. The UL CoMP involves the techniques such as CoMP
reception and CPS. The CoMP reception implies the reception of UE transmitted signal at multiple geographically separated eNB antennas. In Chapter 4,
CoMP reception in the form of macro diversity reception has been studied.
The study has been carried out in both Intra-Site and Inter-Site CoMP scenarios.
The study results have shown that the CoMP macro diversity reception performs
better in the Inter-Site scenario. The performance in the Intra-Site scenario is
limited by the number of macro diversity UEs. In the Inter-Site scenario, both
MRC and SC schemes have been studied, and as expected the MRC scheme
performs better than the SC scheme.
It was also shown in the study that with standing alone macro diversity reception
scheme, the overall improvement is very marginal. There is only about 2%
average user throughput gain and 8% of 5% outage gain. By applying the
CLPC scheme alone without the support of ideal IC in the coordination cell,
the enhancement is very limited as well with only about 4% and 8% gain in terms
of 5% outage and average user throughput respectively. By ideally canceling the
strongest interference in the macro diversity link, the performance of both 5%
outage and average user throughput can be largely enhanced by 14% and 8%
gain, and by applying the CLPC scheme on the top of IC-based MRC reception,
the 5% outage gain number can be nearly doubled together with another 4%
gain in terms of average user throughput. So the combination of macro diversity
reception together with IC and CLPC scheme is necessary in order to maximize
the overall network enhancement.
Besides, the interaction of antenna downtilting angle together with macro diversity reception has shown the optimal performance for both outage and average
user throughput at 14o downtilting. The study with non-ideal cell selection
130
scenario also demonstrated that, in the practical application, macro diversity

reception can provide even higher gain than the simulated ideal cell selection
assumption.
7.3
UL CoMP Joint Reception and Coordinated

Packet Scheduling
In Chapter 5 and Chapter 6, the UL CoMP reception has also been investigated
with joint reception processing, where the previous studied UL CoMP macro
diversity reception is used as a reference.
For the Inter-Site CoMP scenario, the requirement of network backhaul is a
key issue for the application of CoMP reception scheme. In this study the
LTE network backhaul requirements for both CoMP reception schemes are first
analyzed. Compared with the CoMP macro diversity reception, the backhaul
requirement for the CoMP joint reception is about 10 times higher. Practically
depends on the number of quantization bits and transmission overhead utilized,
the backhaul requirement could be much higher for CoMP joint reception.
The study carried out in Chapter 5 is based on the realistic MMSE/MMSE-SIC
receiver, where the multi-cell CPS is not exploited. The investigations are also
evaluated in both Intra-Site and Inter-Site CoMP scenarios, and the Inter-Site
scenario is further divided into Case-1 and Case-3 for detailed analysis. By considering the performance gain and capacity requirement of LTE X2-interface,
the CoMP joint reception is a favorable choice for the Intra-Site application. It
can provide better overall performance than the CoMP macro diversity reception without the need of large capacity X2-interface coordination. In general,
there is about 10% cell-edge througput gain and 6% average user throughput
gain compared with the no CoMP case. The CoMP macro diversity reception
with MMSE-SIC receiver is a better candidate for the cross-site, Case-3, type
of application. It can give the sub-optimal performance gain close to the application of CoMP joint reception and demand relatively much lower capacity
requirement of X2-interface.
In Chapter 6, a multi-cell CPS algorithm is proposed and investigated in the
Intra-Site scenario with 3 coordination cells. The study is focused on the MMSESIC joint reception. The exploited CPS utilizes the multiuser diversity to allocate better joint reception pair together to fully explore the multiuser detection
gain. Compared with the corresponding Intra-Site scenario with no CoMP, the
applied CPS algorithm makes further improvement in both cell-edge and average user throughput. The optimal case shows that there are about 14% gain in
7.4 Overall Conclusion and Future Work Recommendations
131
terms of 5% outage and 7% gain in terms of average user throughput over the
baseline, no CoMP, case.
Furthermore, compared with the theoretical CoMP investigations shown in the
earlier literature, the achievable CoMP gain in this study is generally lower
than the theoretical limits because of limited number of coordination eNBs
and selected macro diversity UEs. In the realistic network, the overall CoMP
performance may even be reduced due to the practical issues such as channel
estimation errors, synchronization and latency challenges.
7.4
Overall Conclusion and Future Work Recommendations
In this study, the topic of multi-cell UL CoMP is the main focus. The overall
results have shown that the UL CoMP gain in the Macro-cell scenario with
three cell coordination is about 20% to 25% gain in terms of cell-edge user
performance. Compared with the very high gain number reported in the earlier
theoretical CoMP studies, the more realistic UL CoMP investigation presented
in this study does not show considerable gains in terms of both cell-edge and
average user throughput. One main reason for the presented small CoMP gain
is the limited number of CoMP UEs. With larger CoMP UE selection window
size, more users will be selected as the CoMP UEs and higher CoMP gain is
expected. Besides, the study is limited to the CoMP scenario with 3 cells.
In some TTIs, the strong interference might come from the cells outside the
3-cell coordination area. So larger CoMP coordination area is also expected
to give better performance. Currently the study work is ongoing with larger
CoMP coordination area. In this study the LTE X2-interface was considered
ideally with unlimited transmission bandwidth and zero delay. For the larger
area of CoMP coordination, the backhaul issues should be included for more
realistic investigation. The potential gain of UL CoMP in the heterogeneous
networks is another interesting topic which is ongoing to be investigated in the
next step work. From the future implementation point of view, Macro-cell with
remote radio head (RRH) scenario is showing a potential interests with CoMP
application.
132
Appendix
Performance of Integrator
Handover Algorithm
The LTE network aims at increasing network capacity, lowering latencies and
reducing network complexity [126]. It focuses on services in the packet-switched
domain to minimize transmission latency and increase robustness of communication. An important requirement for LTE is to provide support for IP-based
traffic with end to end QoS. Voice traffic will be supported mainly as Voice over
Internet Protocol (VoIP) enabling better integration with other multimedia services [126]. For VoIP transmission to be intelligible to the receiver, voice packets
should not be dropped, excessively delayed, or suffer varying delay/jitter. For
mobile-to-mobile communication, the maximum tolerable one way (end-to-end)
delay is 200 ms [127].
Handover/handoff (HO) is a critical procedure for QoS since it contains a socalled HO detach time, which is a gap in the data transmission. Typical values
are in the range of 20 ms [126]. For the VoIP services, it is required to have a
fast HO decision algorithm to avoid further delays and the risk of a call drop.
In this chapter, a HO decision algorithm is proposed for the LTE system. It is
evaluated in the Manhattan scenario and compared with the traditional PBGT
algorithm. As main Key Performance Indicator (KPI), the number of HOs
and SINR before and after the HO are used. In Section A.1, the state of art
for the DL LTE HO is presented. In Section A.2, the LTE intra-frequency
HO procedure is briefly discussed. In Section A.4, both PBGT and integrator
algorithm are introduced. In Section A.5, the system evaluation parameters and
setup are shown. And the simulation results and conclusions are presented in
Section A.6 and Section A.7 respectively.
134
A.1
Performance of Integrator Handover Algorithm
DL LTE Handover
Handover is part of the LTE mobility management, and it is an important functionality for QoS provisioning, especially for the delay sensitive services [126].
LTE network aims at providing seamless access to voice and multimedia services, which is achieved by supporting HO from one cell, i.e. serving cell, to
another, i.e. target cell. As presented in Section 2.1, the LTE has the decentralized network architecture, which facilitates the use of so-called hard HO1 .
In LTE Rel8, the hard HO has been standardized for the DL LTE application
[43], where the break-before-make type of connection makes the issue of serving
seamless access even more critical.
(a) Movement of a UE
(b) UE Received Signal Level
Figure A.1: Corner effect or Manhattan scenario

One of the challenging environments for the hard HO is the corner turning effect
in the Manhattan scenario as shown in Figure A.1. It happens due to the loss
of the Line of Sight (LOS) component from the serving eNB to the UE, for
example when a UE turns around a corner from one street to the other or a
moving obstacle temporarily hinders the path between an eNB and a UE. The
corner effect is very hard to predict, and it might cause a sudden large drop
(e.g. 20-30 dB) in the UE signal strength.
In UMTS, the soft HO schemes can solve the corner effect quite efficiently
compared with the hard HO schemes. Because the soft HO schemes are not
1 A hard HO is defined as the link connection to the serving cell broken before the connection
to the target cell, also known as break-before-make. On the contrary, a soft HO indicates that
the link connection to the target cell is established before the serving link connection broken,
also known as make-before-break.
A.2 DL LTE Hard Handover Procedure
135
supported in the LTE system, to deal with the corner effect, it requires fast hard
HO decision algorithm. Otherwise the call or service quality will be decreased
significantly, and a long gap may potentially lead to many lost packets or even a
lost HO command which leads to a potential call drop. For the real time services,
such as VoIP application, in order to keep the QoS criteria, it is quite important
for the fast HO decision algorithm to be implemented to avoid the call or service
drop. It is also desired that the fast hard HO decision algorithm can deal with all
the environment scenarios, such as the moving obstacles temporarily hindering
the path between an eNB and a UE which resembles the corner effect.
In general, the LTE HO can be divided into intra-LTE HO and inter-Radio Access Technology (RAT) HO, where the inter-RAT HO contains the applications
such as the HO from LTE to the 2G networks or other 3G networks. In this
study, the focus is on the DL hard HO for the intra-LTE application.
A.2
DL LTE Hard Handover Procedure
The hard intra-frequency HO in DL LTE is network-initiated, network-controlled

and UE-assisted [128], where, as shown in Figure A.2, the hard HO procedure
in DL LTE typically consists of three phases: Initialization, Preparation and
Decision [126].
The initialization phase contains three main steps which are UE measurements,
UE processing and UE reporting [19]. Firstly, the UE measures the DL RS
from both serving cell and target cells. The processing of measurements is also
performed in UE in order to filter out the effect of fast fading and measurement/estimation imperfections. At the end, the processed measurements are
reported from the UE to the serving eNB in a periodic or event based manner.
In the network side, the preparation phase contains two main steps which are
the serving eNB HO decision and the serving eNB negotiation with target eNB.
If the HO is granted by the serving eNB, the execution phase contains two main
steps as well which are the serving eNB temporary forwarding packets to the
target eNB and path switch in aGW [19].
A.3
Modeling of HO Measurement in DL LTE
The Reference Signal Received Power (RSRP) is standardized as the DL measurement in LTE. A single RSRP observation is defined as the mean measured
power per RS observed over a single sub-frame or TTI, as shown in Figure
136
Figure A.2: Intra-LTE HO Procedure
A.3. One or several such observations may be combined to form an RSRP measurement report in accordance with the specified frequency and time domain
filtering procedure [129].
As discussed in Chapter 2, the LTE utilizes the scalable bandwidth up to 20
MHz. The application of scalable bandwidth in LTE also allows to perform
the HO measurement with different bandwidth. Depending on the RSRP measurement bandwidth, the frequency selective fading will have impact on the HO
performance. With wide-band signal measurements, the impact is marginal because of the measurement averaging in the frequency domain [129]. However,
for the narrow-band signal measurements, the multi-path fading can cause the
signal power to drop rapidly, which seriously impacts the HO performance.
As shown in Figure A.3, there are a total of eight DL reference symbols per PRB
at each antenna port. This limited number of RS for the HO RSRP measurements introduces the measurement errors according to different measurement
bandwidth. In this study, the measurement error is modeled as uncorrelated
and normally distributed in dB with zero mean and dB standard deviation as
shown in Equation A.1 [55]:
4Q N (0, 2 ) [dB]
where Q is the error measured in dB.
(A.1)
A.3 Modeling of HO Measurement in DL LTE
137
Figure A.3: RS Structure in DL LTE

As shown in Table A.1, more measurement errors are expected for the smaller
RSRP measurement bandwidth which contains less number of RSs, i.e. with
the measurement bandwidth of 1.25 MHz for 6 PRBs, the corresponding measurement error standard deviation is 0.8 dB.
Table A.1: Standard Deviation of Measurement Error [55]
Measurement Bandwidth [MHz]
1.25
2.5
5
10
Number of PRBs
6
12
25
50
[dB]
0.8
0.6
0.45
0.35
138
A.4
DL LTE Hard Handover Decision Algorithm
Two HO decision algorithms are evaluated in this study, the PBGT algorithm
and the integrator algorithm.
A.4.1
Power Budget (PBGT) Algorithm
The PBGT algorithm uses both HO margin (HOM) and Time-to-Trigger (TTT)
timer to make the HO decision, as shown in Figure A.4. A HO is triggered when
the triggering condition, RSRPT > RSRPS + HOM , is fulfilled during TTT,
where RSRPS /RSRPT are the serving/target cell RSRP measurements.
Figure A.4: HO Decision Algorithms
A.4.2
Integrator Algorithm
The integrator algorithm considers both a triggering threshold and a forgetting

factor to make the HO decision. The general idea of integrator algorithm is to
integrate the RSRP differences of the serving and target cell, as shown in the
shaded area in Figure A.4, by using an Infinite Impulse Response (IIR) filter.
The HO decision is made according to the triggering condition between the
filtered RSRP differences and the triggering threshold.
A.5 Simulation Assumptions
139
In this study a special case of the first order auto regressive moving average
(ARMA) filter is used and shown below:
F DIFs j (t)
DIFs j (t)
=
=
(1 )F DIFs j (t 1) + DIFs j (t)

RSRPT (t) RSRPS (t)
(A.2)
(A.3)
where DIFs j (t) is the difference of DL RSRP measurement between the received
signal level of the serving cell s and the target cell j at the time t. F DIFs j (t)
and F DIFs j (t 1) are the filtered DIFs j (t) and DIFs j (t 1) value at the
time t between the serving cell s and the neighboring cell j. is known as the
forgetting factor or smoothing constant (0 1).
F DIFT hreshold is the HO triggering threshold. If F DIFs j (t) > F DIFT hreshold ,
then the HO is triggered immediately. The F DIFs j (t) value is influenced by
the choice of the value. If the choice of value is equal to or close to 1, it
would result in the F DIFs j (t) value more likely being reflected by the most
recent DIFs j (t) value. The value of the F DIFs j (t) will be very instantaneous
or responsive. Else, if the choice of value is equal to or close to 0, it would
result in the F DIFs j (t) value more likely being reflected by the averaged past
F DIFs j (t) value. The value of the F DIFs j (t) would be very constant or
unresponsive to the actual DIFs j (t) change.
The initial value of F DIFs j (t 1) can be defined either by averaging several
early periods of DIFs j (t) values or simply the first observed value of DIFs j (t).
In this study, the defined initial value is set at zero.
A.5
A dynamic system-level simulator for the DL LTE application is used to evaluate

the proposed integration algorithm [130]. The Manhattan scenario with micro
cells is used as the network model. In Figure A.5, the Manhattan scenario setup
is shown. The Manhattan grid setup has 28 eNBs, which are located below the
roof top, with the block size of 200 m and street width size of 30 m. The eNB
antennas are omi-directional. The statistics are only collected from the central
cells.
In Table 1, the detailed simulation parameters and their values are shown. The
active UEs are uniformly distributed over the network area, and the UEs are
allowed to move forward, backwards and turning corners (no wrap-around) with
140
Figure A.5: Micro Scenario Setup
a constant speed during the whole simulation time. When the UEs reach the
simulation border, they will simply turn around and move in the reverse direction.
As the traffic model, a standard VoIP model is used with 30 byte of packet size
and 6 byte of full header size. The RR packet scheduler is assumed during the
simulations. Path loss, shadow fading, and frequency selective fast fading have
been included in the simulation. The shadow fading is modeled as log-normal
distributed with a mean value of 0 dB and a standard deviation of 6 dB. The
spatial de-correlation distance parameter used to describe the spatial correlation
function of the shadow fading is set at 50 m. The frequency selective fast fading
is modeled by using the 3GPP standard Pedestrian (3 kmph) or Vehicular-A
model (30 kmph/120 kmph) depending on the UE moving speed.
For the RSRP measurements, the RSs are not explicitly modeled. The RS values
in one TTI are assumed to be highly correlated in both time and frequency
directions and are represented by one path loss plus fading value per PRB.
According to the 3GPP definition, the RSRP observations are only made for the
given N central PRBs, which are then averaged in frequency domain. The 3GPP
defined minimum measurement bandwidth is 1.25 MHz. During the simulation,
the measurement bandwidth is chosen to be 1.25 MHz and the corresponding
central number of PRBs to be measured is 6. The 6 PRB values are measured
independently and linear averaged afterwards. The UE sampling of the RSRP
measurement is set to be 5 ms. An RSRP value is reported to the serving eNB
every 500 ms. One report contains the exponential average in time of 100 RSRP
A.5 Simulation Assumptions
141
Table A.2: Micro-Scenario Simulation Parameters in DL LTE

Parameter
Network Layout
Number of eNBs
Number of UEs
Average number of UEs per cell
Inter-eNB Distance
eNB Height
UE Location
eNB Antenna
UE Move Speed
Duration of simulation
System Bandwidth
RSRP Measurement Bandwidth
RSRP Measurement Period
Measurement Error
Sliding Window Size
Assumptions
Micro Cells, Manhattan Grid
(Block Size-200 m,Street Size-30 m)
28
1400
50
200 m
10 m (below roof top level)
Outdoors
Omi-directional with linear gain=1
3 kmph, 30 kmph, 120 kmph)
90 sec.
5 MHz
1.25 MHz or 6 PRBs
5 ms
0.8
500 ms
samples. The exponential smoothing filter uses a forgetting factor of 0.1.

In order to evaluate the performance of the proposed HO algorithms, two KPIs:
Number of HOs and SINR have been used during the investigation of the proposed algorithms.
The Number of HOs show what the average number of HOs per UE is
by using of the proposed HO algorithms. Every HO comes with a risk
of an HO failure. In general, by lowering the number of HOs, the HO
burden to the network can be reduced and the potential degradation in
QoS due to the detach time gap introduced by the HOs can be minimized
as well. However, the number of HOs cannot be infinitely minimized.
There is always a tradeoff between the number of HOs and the signal
quality parameter.
The signal quality is evaluated by the scheduled SINR per UE. During
the evaluation the SINR is divided into SINR before HO and SINR after
HO. The SINR before HO tells the signal quality level before making the
HO and the SINR after HO tells us the signal quality level after making
the HO. In our simulation the observation time for SINR before and after
HO is 500 ms.
142
A.6
In the simulations, the influences of forgetting factor and F DIFT hreshold

to the proposed integrator algorithm are evaluated first, and followed by the
performance comparison with the PBGT algorithm.
In order to evaluate the influence of to the integrator algorithm, the value of
F DIFT hreshold is fixed to be -5. The forgetting factor varies between 0.25, 0.5,
and 1.
According to the theory, when =1, it means that all the past FDIF values
will be forgotten, and the HO only depends on the present instantaneous DIF
value. The filtered or integrated instantaneous DIF value can also easily reach
the F DIFT hreshold to trigger the HO. So it is expected that there is a larger
number of HOs for =1 than lower values of .
As shown in Figure A.6, the case of =1 has the highest number of HOs, and
the number of HOs are decreasing when is getting smaller. In Figure A.7,
all the SINR after HO are improved compared to the SINR before HO for all
the varying cases. Before making the HO, =1 has the best SINR since it
can make the HO without any delay in the serving eNB with declining RSRP
measurement, and there is about 5 dB difference at cdf probability 70% to
compare with =0.25. After the HO, it shows that =1 has the lowest SINR.
To evaluate the influence of F DIFT hreshold to the integrator algorithm, the
value is fixed to be 0.5. The F DIFT hreshold varies between -0.1 dB, -5 dB and
-10 dB.
Theoretically higher values of the F DIFT hreshold correspond to a larger Handover Margin (HOM) value in PBGT algorithm. So a lower number of HOs are
expected with a higher value of F DIFT hreshold . As it can be seen in Figure A.8,
F DIFT hreshold =-0.1 dB has the highest number of HOs and F DIFT hreshold =10 dB has the lowest number of HOs.
As shown in Figure A.9, in general, all the SINR values after HO are improved
compared to the SINR before HO for all cases. The best SINR before HO
is achieved for F DIFT hreshold equal to -0.1 dB. This is due to the fact that
this setting leads to the fastest HO decision while the slowest HO decision
(F DIFT hreshold =-10 dB) leads to the worst SINR before HO. With the same
cdf probability at 70%, there is about 5 dB difference in SINR between them.
However, after making the HO, there is a big improvement for F DIFT hreshold
=-10 dB in SINR, and the improvement for F DIFT hreshold =-0.1 dB is quite
small.
A.6 Performance Evaluation
143
Figure A.6: Number of HOs: Varying Values with F DIFT hreshold = 5dB
Figure A.7: SINR: Varying Values with F DIFT hreshold = 5dB
144
Figure A.8: Number of HOs: Varying F DIFT hreshold with =0.5
Figure A.9: SINR: Varying F DIFT hreshold with =0.5
A.7 Conclusions
145
Based on the above evaluations of the integrator algorithm, comparisons of

the integrator algorithm with the traditional PBGT algorithm are made. The
comparisons are performed in two steps.
In the first step, special parameters are used in both algorithms. For the integrator algorithm with = 1, the HO decision depends only on the F DIFT hreshold .
For the PBGT algorithm with TTT=0 ms, the HO triggering relies only on the
HOM. We set the value of F DIFT hreshold equal to the HOM. It is expected
that both algorithms are identical since HOM=F DIFT hreshold =RSRP s (t) RSRP T (t). As shown in Figure A.10 and A.11 the two algorithms are performed identically.
In the second step, more realistic parameters are chosen based on the first
step evaluation. HOM=5 dB and TTT=500 ms are used in the PBGT version
algorithm, and F DIFT hreshold =-5db and Forgetting Factor=0.5 are used in the
integrator algorithm for comparison at the speed 3 kmph, 30 kmph and 120
kmph.
As it can be seen in Figure A.10 and A.11, with this specific parameter setup,
the integrator and PBGT algorithms have also the same performances in both
Numbers of HOs and SINR before and after HO evaluations at different UE
speeds, and as expected the higher speed has a higher number of HOs.
A.7
Conclusions
In this paper, the integrator handover decision algorithm is proposed and studied. The general idea of this algorithm is to integrate the RSRP differences of
the serving and target cell.
Two parameters, FDIF threshold and Forgetting Factor , have been studied
respectively, which can be used to tune the integrator algorithm. The performances of integrator algorithm are also evaluated and compared with the
traditional PBGT algorithm in the LTE network.
The simulation results show that the integrator algorithm has the same performance as the PBGT algorithm based on the Number of HO analysis and SINR
before and SINR after HO evaluations at different UE speeds.
146
Figure A.10: Number of HOs - Comparison
Figure A.11: SINR - Comparison
Appendix
B
System-Level Simulator
Description
A semi-static system-level simulator has been used in this study for the performance assessment of the designed algorithms presented in Chapter 3, 4, 5, and
6. In Section B.1, the models used in the system-level simulator are described.
In Section B.2, the mostly used KPIs are defined. And the default simulation
assumptions and parameters are listed in Section B.3.
B.1
System-Level Simulator
The performance evaluation is made by using a multi-cell system-level simulator,

which contains 19 BSs with 3 sectors per BS (in total 57 cells) as shown in
Figure B.1. A hexagonal shape per sector has been simulated. The orientation
of antenna main lobe is indicated by the red-dashed line in Figure B.1. The
3-D antenna pattern with 70o horizontal HPBW and 10o vertical HPBW is
exploited. Wrap-around is applied to eliminate the network edge effect and
generate accurate simulation results.
The deployment scenario is defined according to [78], where Macro case-1 scenario is characterized by small cell radius with ISD of 500 m, and Macro case-3
scenario is indicated by a large cell radius with ISD of 1732 m. BS antenna is
located at the roof top with the height of 32 m above ground.
The propagation modeling consists of path-loss, shadowing and fast fading. A
20 dB outdoor-to-indoor penetration loss is included in the path-loss model for
148
System-Level Simulator Description
Figure B.1: System Level Simulator Network Layout
the Macro cell scenario [131]. Shadowing is fully correlated between cells of
the same site, while the correlation is 0.5 between different site cells. The fast
fading can be simulated with either Typical Urban (TU) power delay profile for
UE speed of 3 kmph or SCMC channel model. The TU model is a tapped delay
line implementation with uncorrelated Rayleigh fading path [132]. The SCMC
channel model is a tapped delay-line model for multiple antennas with correlation parameters derived from fixed angular parameters and assumed antenna
configurations. It can be used for simulating full MIMO channels of size up to
4x4 [133].
The UE creation or arrival process depends on the traffic model deployed. For
the full/infinite buffer traffic model, the UEs are dropped in the system at the
start of a simulation run and remain in the network until the end of the run.
One simulation consists of several simulation runs. For the finite buffer traffic
model, the UEs are generated at the beginning of the simulation. Once the UE
buffer is emptied, the UE is killed and replaced by a new UE at different network
location. The finite buffer traffic model can also be combined with AC, where
the UEs are generated in the network according to a Poisson arrival process. If
a certain AC decision criteria is fulfilled, the UE is admitted, otherwise the UE
is rejected or blocked.
For the best effort services with variable bit rate, the full buffer UEs experience
B.1 System-Level Simulator
149
different channel conditions, which result in different data rates by transmitting

a different amount of data in the same simulation time, whereas the finite buffer
UEs will transmit the same amount of data and the session time is terminated
once the buffer is emptied. In this case, the cell-edge UEs will stay longer in
the network due to low data rate compared to the cell-center UEs.
The distribution of the UEs depends on whether the load is balanced or unbalanced. The network is called balanced-load if the UEs are uniformly generated
in each cell and also each cell contains the same number of UE. Else if the UEs
are uniformly distributed in the whole network region with different number of
UEs in each cell, then it is called unbalanced-load network.
For each UE, the fast fading component is time varying at every TTI, while
the PL and shadowing components are assumed to be constant during a packet
call. The serving cell is selected according to the lowest total path-loss including
distance dependent path-loss, shadowing, and effective antenna gains.
The RRM functionalities such as LA, PS and HARQ are also accurately modeled
in the simulator for each cell. As presented in Section 2.4, the LA is modeled as
the fast AMC based on the CSI measurement. In order to maintain the BLER
target of the first transmission, the OLLA offset is used to bias the CSI before
using it for MCS selection. The dynamic PS is modeled as decoupled TD and
FD scheduler as presented in Section 2.6. With the transmission bandwidth
of 10 MHz in LTE, there are in total 50 PRBs where 48 PRBs are used for
data transmission and 2 PRBs are utilized for control signaling transmission.
Fixed transmission bandwidth with 6 PRBs is the main focus in this study.
Therefore in each TTI, 8 UEs will be scheduled in the frequency domain in
each cell according to the applied scheduling algorithm. The system model also
includes synchronous adaptive HARQ with ideal chase combining as presented
in Section 2.3. The UL Power Control (PC) is implemented according to the
standardized formula presented in Equation 2.5, where both OLPC and CLPC
components are considered.
The single simulator approach, which includes everything from single-cell linklevel processing to multi-cell network management, is too high for the required
simulation resolution [7]. Therefore, separated link-level and system-level simulators are desired. The link-to-system performance mapping function is used to
connect the separated link and system level simulations. At system-level simulation, the mapping function can be used to predict the instantaneous BLER
without performing detailed link-level processing steps. The mapping function
is estimated from the link-level simulations, which considers the facts such as the
MCS format, receiver type and channel state. The generated mapping function
should be general enough to cover different transceiver types and antenna techniques. In this study, the Actual Value Interface (AVI) method is used for the
150
link-to-system mapping [134]. The AVI tables constructed from an extensive

link-level simulations are used to map the average received SINR to the corresponding BLEP. Besides, the compressed SINR to symbol variance mapping
presented in Section 5.3.2 is used for the multiuser detection with MMSE-SIC
receiver.
B.2
Key Performance Indicator - KPI
For the study in the system-level, the KPIs used to evaluate the performance
are as follows:
1. The average cell throughput TPcell is defined as the ratio between the
total correctly decoded bits per cell and the total simulation time
TPcell =
Total correctly received bits per cell

Simulation Time
(B.1)
2. The corresponding spectral efficiency is given by the ratio of total cell

throughput divided by the bandwidth occupied:
Spectral Efficiency =
TPcell
Bandwidth
(B.2)
3. The average user throughput for the i th active user is defined as:
TPi =
Correctly received bits from user i

Session Time
(B.3)
Coverage, denoted by TPcoverage , is determined from the CDF curve of the

average user throughput taken over all the completed sessions. Coverage
is defined as the data rate corresponding to the 5% quantile in the CDF
curve, i.e., 95% of the users experience a higher average data rate than
the rate specified by the coverage parameter. This KPI indicates data
rate experienced by users around the cell edge. Further, it can be used
to differentiate packet scheduler in terms of fairness in the distribution of
throughput among users.
B.3 Simulation Assumptions
B.3
151
The default simulation assumptions and parameters are listed in Table B.1:
Table B.1: Simulation Assumptions of Mechanical Antenna Downtilting
Parameter
Simulation Time
System Carrier Frequency
System Bandwidth
Number of PRBs for Data/Control
Frequency Reuse Factor
Number of sub-carriers per PRB
TTI
PRBs per UE per TTI
Deployment Scenario
Minimum UE to BS Distance
BS / UE Height
Number of UE/BS Antennas
Receiver Type
BS Antenna HPBW
BS Antenna Gain
UE Antenna Gain
Max UE Transmit Power
FPC P0 and value
Distance dependent Path-Loss
Log-normal Shadowing
CSI Log-normal Err. Std. Dev.

CSI Resolution
OLLA Offset
OLLA Offset Range
Control channel overhead
Assumptions
5 runs, 5 s/run, 2 s/run (WarmUp)
2 GHz
10 MHz (50 PRBs)
48 PRBs / 2 PRBs
1
12
1 ms
6 PRBs
35 m
32 m / 1.5 m
1-UE Tx / 2-BS Rx Antenna
MRC/MMSE/SIC
Horizontal: 70o
Vertical: 10o
14 dBi
0 dBi
24 dBm (250 mW)
Macro case-1: -58 dBm and 0.6
Macro case-1: -106 dBm and 1.0
128.1 + 37.6log10 (d[km])
Standard deviation: 8 dB
Correlation distance: 50 m
Correlation bet. cells/sites: 1.0/0.5
1 dB
2
0.5 dB
[-4.0,4.0] dB
3/14 symbols
152
Bibliography
[1] Editors Desk, Mobile web: Latest facts and stats forecast a rosy outlook,
http://mobithinking.com/blog/latest-mobile-stats.htm, August 2009.
[2] 2008 global broadband phenomena - executive summary, Sandvine, October 2008.
[3] Hans Beijner, Ran evolution: An introduction to evo ran, Ericsson,
September 2009.
[4] 3G Americas, The mobile broadband future - hspa+ and lte: An educational feature brought to you by 3g americas and wireless week, 3G
Americas Wireless Communications, 2008.
[5] Chandra D and Harris R. J. and Shenoy N., Tcp performance for future
ip-based wireless networks, 3rd. IASTED International Conference on
Wireless and Optical Communication (WOC2003), Banff Canada, July,
2003.
[6] M. D
ttling, W. Mohr, and A. Osseiran, Radio Technologies and Concepts
for IMT-Advanced. John Wiley & Sons Ltd, 2009.
[7] H. Holma and A. Toskala, WCDMA for UMTS: Radio Access for Third
Generation Mobile Communication, 3rd edition. John Wiley & Sons Ltd,
2004.
[8] , HSDPA/HSUPA for UMTS: High Speed Radio Access for Mobile
Communication. John Wiley & Sons Ltd, 2006.
[9] QUALCOMM, Hspa+ rel-8 and rel-9, http://russia.qualcomm.com/
common/documents/articles/HSPAplus R8andR9 Benefits 081109.pdf,
September 2009.
154
BIBLIOGRAPHY
[10] H. Holma and A. Toskala, WCDMA for UMTS: HSPA Evolution and
LTE. John Wiley & Sons Ltd, 2007.
[11] Ericsson, Evo ran: 3 generation, 1 network, http://www.ericsson.com/
ericsson/corpinfo/publications/review/technology update/archive/2009/
issue 2/articles/evoran.shtml, February 2009.
[12] 3GPP Technical Specifications 36.101, Evolved universal terrestrial radio access (e-utra), user equipment (ue) radio transmission and reception
(release 8), version 8.2.0, 3GPP, 2008.
[13] 3GPP Technical Specifications 36.104, Evolved universal terrestrial radio
access (e-utra), base station (bs) radio transmission and reception (release
8), version 8.2.0, 3GPP, 2008.
[14] 3GPP Technical Report 25.913, Requirements for evolved utra (e-utra)
and evolved utran (e-utran), version 7.3.0, 3GPP, March 2006.
[15] H. Holma, A. Toskala, K. Pajukoski, and E. Tiirola, Utran long term
evolution in 3gpp, in proceeding of the IEEE international symposium
on personal, indoor and mobile radio communication (PIMRC), Helsinki,
Finland, 2006.
[16] E. Dahlman, S. Parkvall, J. Skold, and P. Beming, 3G Evolution: HSPA
and LTE for Mobile Broadband. Elsevier Science, 2008.
[17] H. Schulze and C. Lueders, Theory and Applications of OFDMA and
CDMA: Wideband Wireless Communication. John Wiley & Sons Ltd,
2005.
[18] S. Sesia, I. Toufik, and M. Baker, LTE, The UMTS Long Term Evolution:
From Theory to Practice. John Wiley & Sons Ltd, 2008.
[19] H. Holma and A. Toskala, LTE for UMTS: OFDMA and SC-FDMA Based
Radio Access. John Wiley & Sons Ltd, 2007.
[20] H. G. Myung, J. Lim, and D. J. Goodman, Single carrier fdma for uplink
wireless transmission, IEEE Vehicular Technology Magazine, vol. 1, pp.
3038, Sep.2008.
[21] D. Tse and P. Viswanath, Fundamentals of Wireless Communication,
1st ed. Cambridge University Press, 2005.
[22] E. Dahlman, H. Ekstrom, A. Furuskar, Y. Jading, J. Karlsson, M. Lundevall, and S. Parkvall, The 3g long-term evolution - radio interface concepts and performance evaluation, in proceeding of the IEEE vehicular
technology conference (VTC), Melbourne, Australia, vol. 1, May 2006.
BIBLIOGRAPHY
155
[23] ITU-R, Itu-r imt.[rest], requirements, evaluation criteria and submission

templates for the development of imtadvanced, http://www.itu.int/md/
R07-SG05-C-0068/en, 2007.
[24] S. Kumar, Techniques for efficient spectrum usage for next generation
mobile communication networks: An lte and lte-a case study, Ph.D. dissertation, Department of Electronic Systems, Aalborg University, 2009.
[25] 3GPP Technical Report 36.913, Requirements for further advancements
for e-utra (lte-advanced), 3GPP, June 2008.
[26] P. E. Mogensen, T. Koivisto, K. I. Pedersen, I. Z. Kovacs, B. Raaf, K. Pajukoski, and M. J. Rinne, Lte-advanced: The path towards gigabit/s
in wireless mobile communications, Wireless VITAE 2009, pp. livcxiv,
May 2009.
[27] E. Seidel, Progress on lte advanced - the new 4g standard, Nomor
Research GmbH, July, 2008.
[28] S. Parkvall and D. Astely, The evolution of lte towards imtadvanced,
Journal of Communications, vol. 4, p. 3, April 2009.
[29] 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 Proceedings of the IEEE Vehicular Technology
Conference (VTC) Singapore, pp. 25172521, May 2008.
[30] S. Hussain, Dynamic radio resource management in 3gpp lte, Ph.D.
dissertation, Department of Applied Signal Processing, Blekings Institute
of Technology, January 2009.
[31] Performance enhancement in a frequency hopping gsm network, howpublished=Kluwer Academic Publisher, year=2000.
[32] 3GPP Technical Report 25.814, Physical layer aspect for evolved universal terrestrial radio access (utra), 3GPP, September 2006.
[33] jarmo niemela, tero isotalo, and jukka lempiainen, Optimum antenna
downtilt angels for macrocellular wcdma networks, EURASIP Journal
on Wireless Communications and Networking, issue 5, pp. 816-827, 2005.
[34] A. Wacker, K. Sipila, and A. Kuurne, Automated and remotely optimization of antenna subsystem based on radio network performance, WPMC
pp. 752-756, October 2002.
[35] F. L., H. C., and G. R., Distributed uplink macro diversity for cooperating base stations, IEEE International Conference on ICC Workshops,
June 2009.
156
BIBLIOGRAPHY
[36] P. Marsch, S. Khattak, and G. Fettweis, A framework for determining

realistic capacity bounds for distributed antenna systems, in Proceedings
of IEEE Information Theory Workshop, Chengdu, China, pp. 22-26, October 2006.
[37] S. Venkatesan, coordinating base stations for greater uplink spectral efficiency in a cellular systems, In Proceeding of IEEE symposium, personal,
indoor and mobile radio communication, September 2007.
[38] M. N. Bacha, J. S. Evans, and S. V. Hanly, On the capacity of mimo cellular networks with macrodiversity, in Proceedings of the IEEE Australian
Theory Workshop, 7th, 2006.
[39] H. Zhang and H. Dai, Cochannel interference mitigation and cooperative processing in downlink multicell multiuser mimo networks,
EURASIP Journal Wireless Communication and Networking, vol.2004,
no.2, pp.222-235, December 2004.
[40] N.H.Dawod, I.D.Marsland, and R.H.M.Hafez, Improved transmit null
steering for mimo-ofdm downlinks with distributed base station antenna arrays, IEEE Journal Selected Areas Communication vol.24, no.3,
pp.419-426, March 2006.
[41] M. Karakayali, G.J.Foschini, and R.A.Valenzuela, Network coordination
for spectrally efficient communications in cellular systems, IEEE Wireless
Communication, vol.13, no.4, pp.56-61, August 2006.
[42] Redl, S. M., Weber, M. K., Oliphant, and M. W., An introduction to
GSM. Artech House, Boston, 1995.
access (e-utra) and evolved universal terrestrial radio access network (eutran), overall description, version 8.3.0, 3GPP, January 2008.
[44] Ericsson, Ericsson review, no. 03, http://www.ericsson.com/ericsson/
corpinfo/publications/review/2007 03/05.shtml, 2007.
access (e-utra): X2 general aspects and principals, 3GPP, 2008.
access (e-utra): X2 layer 1, 3GPP, 2008.
access (e-utra): X2 signaling transport, 3GPP, 2008.
access (e-utra): X2 application protocol (x2ap), 3GPP, 2008.
BIBLIOGRAPHY
157
[49] F. D. Calabrese, Scheduling and link adaptation for uplink sc-fdma systems: An lte case study, Ph.D. dissertation, Department of Electronic
Systems, Aalborg University, 2009.
[50] Frank Rayal, The essential aspects of lte phy - part 2, http://www.
eetindia.co.in/STATIC/PDF/201006/EEIOL 2010JUN29 RFD TA 01.
pdf?SOURCES=DOWNLOAD, 2010.
[51] C. Rosa, D. L. Villa, C. U. Castellanos, F. D. Calabrese, P. H. Michaelsen,
K. I. Pedersen, and P. Skov, Performance of fast amc in e-utran uplink,
in Proceedings of the IEEE International Conference on Communications
(ICC) Beijing China, pp. 49734977, May 2008.
[52] A. Pokhariyal, Downlink frequency-domain adaptation and scheduling,
Ph.D. dissertation, Department of Electronic Systems, Aalborg University,
2007.
[53] K. C. Beh, A. Doufexi, and S. Armour, Performance evaluation of hybrid
arq schemes of 3gpp lte ofdma system, in Proceedings of the 18th Annual
IEEE International Symposium on Personal, Indoor and Mobile Radio
Communications (PIMRC), 2007.
[54] B. Priyanto, H. Codina, S. Rene, T. Sorensen, and P. Mogensen, Initial
performance evaluation of dft-spread ofdm based sc-fdma for utra lte uplink, in Proceedings of the IEEE Vehicular Technology Conference (VTC)
Dublin Irland, pp. 15502252, April 2007.
[55] M. Anas, Uplink radio resource management for qos provisioning in long
term evolution, Ph.D. dissertation, Department of Electronic Systems,
Aalborg University, 2009.
[56] K. I. Pedersen, F. F., T. E. Kolding, L. T. F., and P. E. Mogensen,
Performance of high-speed downlink packet access in coexistence with
dedicated channels, in Proceedings of the IEEE Vehicular Technology
Conference (VTC), 2007.
[57] Klaus I. Pedersen and Akhilesh Pokhariyal, Outter loop link adaptation
for uprise lte dl version 1.4, NSN Internal Document, 2006.
[58] Claudio Rosa and Carlos Ubeda and Francesco Calabrese, Lte uplink
rrm: Concept document, NSN Internal Document, 2007.
access (e-utra): Physical layer procedure, version 8.2.0, 3GPP, 2008.
[60] M. Boussif, N. Quinteroand, F. D. Calabrese, C. Rosa, and J. Wigard, Interference based power control performance in lte uplink, in Proceedings
of the IEEE International Symposium on Wireless Communications
Systems (ISWCS) Reykjavik Iceland, pp. 698702, October 2008.
158
BIBLIOGRAPHY
[61] D. Lee and C. Xu, Mechanical antenna downtilt and its impact on
system design, in Proceedings of the IEEE 47th Vehicular Technology
Conference, 1997.
[62] Technical information and new products - kathrein antennen electronic,
Kathrein, September 2000.
[63] P. Zanier, Automated wcdma antenna tilt optimization based on network
measurements, Nokia Siemens Network - Internal Slides, Version 0.0.6,
January, 2009.
[64] 3GPP, Default enb antenna tilt values for lte-advanced evaluations,
R1-084314, November 2008.
[65] U. T
urke and M. Koonert, Advanced site configuration techniques for automatic umts radio network design, in Proceedings of the IEEE Vehicular
Technology Conference Spring, 2005.
[66] J. Niemela and J. Lempiainen, Impact of mechanical antenna downtilt
on performance of wcdma cellular network, in Proceedings of the IEEE
Vehicular Technology Conference (VTC), Spring, 2004.
[67] U. Turke and M. Koonert, Advanced site configuration techniques for automatic umts radio network design, in Proceedings of the IEEE Vehicular
Technology Conference (VTC), Spring, 2005.
[68] F. Gunnarsson, M. N. Johansson, A. Furuskar, M. Lundevall, A. Simonsson, C. Tidestav, and M. Blomgren, Downtilted base station antennas a simulation model proposal and impact on hspa and lte performance, in
Proceedings of the IEEE Vehicular Technology Conference (VTC), Spring,
2008.
[69] K. Pedersen, I. Z. Kovacs, T. Srensen, and P. Mogensen, Review of
antenna technologies for macro cells, Nokia Siemens Network - Internal
Document, 1H 2009.
[70] 3GPP Technical Report 25.996, Technical specification group radio access network - spatial channel model for multiple input multiple output
(mimo) simulations, 3GPP, September 2003.
[71] L. Thiele, T. Wirth, K. Borner, M. Olbrich, and V. Jungnickel, Modeling
of 3d field patterns of downtilted antennas and their impact on cellular
systems, International ITG workshop on Smart Antennas WSA Berlin
Germany, February 2009.
[72] F. Gil, A. Claro, J. Ferreira, C. Pardelinha, and L. Correia, A 3d interpolation method for base station antenna radiation patterns, IEEE
Antennas and Propagation Magazine 43(2), April 2001.
BIBLIOGRAPHY
159
[73] T. Vasiliadis, A. Dimitriou, and G. Sergiadis, A novel technique for the

approximation of 3-d antenna radiation patterns, IEEE Transactions on
Antennas and Propagation, 53(7), July 2005.
[74] H. Asplund, J.-H. Berg, and T. Srensen, The influence of base station antenna tilt on path loss, Nordisk Radio Symposium (NRS), Lund,
Sweden, November 1999.
[75] M. A. F, F. Josef, and B. Ernst, Pattern distortion of mobile radio base
station antennas by antenna masts and roofs, Microwave Conference,
25th European, September 1995.
[76] 3GPP, Default enb antenna tilt values for lte-advanced evaluations, r1084314, 3GPP.
[77] , Selection procedures for the choice of radio transmission technologies of the umts technical requirement, tr 101 112 (umts 30.03), 3GPP,
April,1998.
[78] 3GPP Technical Report, Universal mobile telecommunication system
(umts), selection procedure for the choice of radio transmission technologies of the umts, 3GPP, April 1998.
[79] P. Komulainen and M. Boldi, Initial report on advanced multiple antenna
systems, celtic / cp5-026, Ph.D. dissertation, CELTIC Telecommunication Solutions, 2009.
[80] V. Kotzsch and G. Fettweis, Interference analysis in time and frequency
asynchronous network mimo ofdm systems, in Proceedings of the IEEE
Wireless Communications and Networking Conference (WCNC2010),
Sydney, Australia, 18.-21., April 2010.
[81] Nokia Siemens Networks and Nokia, Setup of comp cooperation areas,
r1-091354, 3GPP, 2009.
[82] A. Tolli, Resource management in cooperative mimo-ofdm cellular systems, Ph.D. dissertation, University of Oulu, 2008.
[83] L. Thiele, T. Wirth, M. Schellmann, Y. Hadisusanto, and V. Jungnickel,
Mu-mimo with localized donwlink base station cooperation and downtilted antennas, IEEE International workshop on LTE evolution, June
2009.
[84] J.Jin, W.QiXing, G.Liu, H.Yang, B.Li, and D.Yang, A novel cooperative
nulti-cell mimo scheme for the downlink of lte-advanced system, IEEE
International workshop on LTE evolution, June 2009.
160
BIBLIOGRAPHY
[85] 3GPP, Performance evaluation of coherent su and mu comp for lte-a,

TSG-RAN1 57bis R1-092430, Jule 2009.
[86] , Preliminary comp gains for itu micro scenario, TSG-RAN1 57bis
R1-092691, Jule 2009.
[87] , Performance evaluation of downlink comp, TSG-RAN1 57bis
R1-092777, Jule 2009.
[88] , Performance estimate on comp joint transmission, TSG-RAN1
57bis R1-092479, Jule 2009.
[89] , Evaluation of dl comp gain considering rs overhead for lteadvanced, TSG-RAN1 56bis, March 2009.
[90] M. Andreas, F. Philipp, and S. Joachim, Performance of the lte uplink
with intra-site joint detection and joint link adaptation, in Proceedings
of the IEEE Vehicular Technology Conference (VTC), Spring, May 2010.
[91] V. Stencel, A. Muller, and P. Frank, Lte advanced-a further evolutionary step for next generation mobile networks, IEEE 20th International
Conference Radioelektronika, 2010.
[92] G. M., M. P., Z. Rong, and F. G., Field trial results for a coordinated
multi-point uplink in cellular systems, IEEE Smart Antennas (WSA),
2010 International ITG Workshop, February 2010.
[93] 3GPP Technical Specifications 36.214, Evolved universal terrestrial radio access (e-utra) and evolved universal terrestrial radio access network
(e-utran),physical layer measurements, version 8.0.0, 3GPP, September
2009.
[94] M. Lalam, R. Visoz, and A. Berthet, A fast prediction method to optimize link adaptation for ic receivers, in Proceedings of the IEEE ISWCS
Reykjavik Iceland, October 2008.
[95] Klaus I. Pedersen and Akhilesh Pokhariyal, Non-ideal serving cell selection for uprise lte dl, NSN Internal Document, 2006.
[96] Naizheng Zheng and Claudio Rosa and Jeroen Wigard, Intra-site sector
macro combining with interference management in uplink lte networks,
NSN Internal Document, 2008.
[97] Mohit Garg, Multi-user detection, http://www.mohrahit.in/find/
multi-user-detection.pdf, 2000.
[98] L. Z., X. P., S. T. B., and V. B., Spatial frequency scheduling for uplink sc-fdma based linearly precoded lte multiuser mimo systems, in
Proceedings of the European Transactions on Telecommunications, pp.
11, 2010.
BIBLIOGRAPHY
161
[99] A. E. Gamal and T. Cover, Multiple user information theory, In

Proceeding of IEEE, vol.68, no.12, pp. 1466-1483, December 1980.
[100] R. Lupas and S. Verdu, Linear multiuser detectors for synchronous codedivision multiple-access channels, IEEE Transaction Information Theory,
vol.35, pp.123-136, January 1989.
[101] U. Madhow and M. L. Honig, Mmse interference suppression for directsequence spread-spectrum cdma, IEEE Transaction Communications,
vol.42, pp.3178-3188, December 1994.
[102] A. D. Hallen, Decorrelating decision-feedback multiuser detector for
synchronous code-division multiple-access channels, IEEE Transaction
Communications, vol.41, pp.285-290, February 1993.
[103] M. K. Varanasi and B. Aazhang, Near-optimum detection in synchronous code-division multiple-access systems, IEEE Transaction
Communications, vol.39, pp.725-736, May 1991.
[104] , Multistage detection in asynchronous code-division multiple-access
communications, IEEE Transaction Communications, vol.38, pp.509-519,
April 1990.
[105] P. Patel and J. Holtzman, Analysis of a simple successive interference
cancellation scheme in a ds/cdma system, IEEE Journal Selection Areas
Communications, vol.12, pp.796-807, June 1994.
[106] D. Gesbert, M. Kountouris, R. W. H. Jr., C.-B. Chae, and T. Salzer,
From single user to multiuser communications: Shifting the mimo
paradigm, IEEE Signal Processing Magazine, 2007.
[107] N. Chiurtu and J. Ayadi, Multi-user multiple antenna systems in uplink ofdm communications, 16th IEEE International Conference on
Electronics, Circuits, and Systems (ICECS), December 2009.
[108] L. Thiele, M. Schellmann, T. Wirth, and V. Jungnickel, Cooperative
multi-user mimo based on limited feedback in downlink ofdm systems,
Signals, Systems and Computers, 42nd, Asilomar Conference, October
2008.
[109] J. G. Andrews, Interference cancellation for cellular systems: A contemporary overview, IEEE Wireless Communication, April 2005.
[110] R. Prasad, M. I. Rahman, S. S. Das, and N. Marchetti, Single- and
Multi-Carrier MIMO Transmission for Broadband Wireless Systems.
River Publisher, 2008.
162
BIBLIOGRAPHY
[111] K. C. Beh, A. Doufexi, and S. Armour, Performance evaluation of hybrid

arq schemes of 3gpp lte ofdma system, IEEE International Symposium
on Personal, Indoor and Mobile Radio Communications (PIMRC), 2007.
[112] B. Zhu, N. Ansari, and Z. Siveski, Convergence and stability analysis of a synchronous adaptive cdma receiver, IEEE Transaction on
Communications, December 1995.
[113] R. Sandra, D. Fernando, B. Carmen, P. Gema, G. Alberto, and A. Vicenc,
Recent advances in mimo wireless systems, Waves, pp.115-123, 2009.
[114] J. Shen and Z. Ding, Blind mmse cdma detection in multipath channels, IEEE Workshop on Signal Processing Advances in Wireless
Communications, SPAWC, 1999.
[115] G. Berardinelli, L. A. M. R. de Temino, S. Frattasi, T. B. Sorensen, P. Mogensen, and K. Pajukoski, On the feasibility of precoded single user mimo
for lte-a uplink, Journal of Communications, Vol 4, No 3 (2009), pp. 155
163, April 2009.
[116] M. Schellmann, L. Thiele, T. Haustein, and V. Jungnickel, Spatial transmission mode switching in multi-user mimo-ofdm systems with user fairness, IEEE Transactions on Vehicular Technology, 2009.
[117] H. Yuan and P. Skov, Ul multi-user mimo implementation, Nokia
Siemens Network - Internal Document, 2009.
[118] G. Berardinelli, B. E. Priyanto, T. B. Srensen, and P. Mogensen, Improving sc-fdma performance by turbo equalization in utra lte uplink,
in Proceedings of the IEEE Vehicular Technology Conference (VTC)
Singapore, pp. 25572561, May 2008.
[119] Fernando H. Gregorio, Multiuser detection for sdma ofdm, http://www.
comlab.hut.fi/opetus/333/2004 2005 slides/Multiuser detection text.pdf,
2004.
[120] V. Kuhn, Linear and nonlinear multi-user detection in coded ofdmcdma systems, International Conference on Telecommunications (ICC
01), Vol.5,Bucharest,Romania, July 2001.
[121] R. Visoz, A. Berthet, and M. Lalam, A novel fast semi-analytical performance prediction method for iterative mmse-ic multiuser mimo joint
decoding, in Proceedings of GLOBECOM 2008, 2008.
[122] Orange, R1-090395, performance prediction of turbo-sic receivers for
system-level simulations, 3GPP, 2009.
BIBLIOGRAPHY
163
[123] K. I. Pedersen, A. Toskala, and P. E. Mogensen, Mobility management

and capacity analysis for high speed downlink packet access in wcdma,
in Proceedings of the IEEE Vehicular Technology Conference, VTC-Fall,
60th, 2004.
[124] E. Telatar, Capacity of multi-antenna gaussian channels, european
Transactions on Telecommunications 10(6), pp 585-569, 1999.
[125] G. Foschini and M. Gans, On limits of wireless communications in a
fading environment when using multiple antennas, Wireless Personal
Communications 6(3), pp 311-335, 1998.
[126] Ulrich Barth, 3gpp long -term evolution/system architecture evolution
overview, Alcatel, September 2006.
[127] NGMN, Next generation mobile networks radio access performance evaluation methodology, NGMN, 2006.
[128] Leping Huang, Handover and ue measurement modeling in freac, Internal Nokia, 2006.
[129] Jean Aicard Fabien, 3gpp tsg ran wg4 r4-070440: Ls on lte measurement
supporting mobility, 3GPP, 2006.
[130] Siemens, Mobile radio simulation environment, requirement specification
version 1.0.1 ius, Internal Document, 2006.
[131] 3GPP Technical Report 25.943, Deployment aspects (release 5), version
5.1.0, 3GPP, 2002.
[132] T. B. Sorensen, P. E. Mogensen, and F. Frederiksen, Extension of the
itu channel models for wideband (ofdm) systems, in Proceedings of the
IEEE VTC Dallas Texas USA, September 2005.
[133] H. Asplund, J. Medbo, B. Goransson, J. Karlsson, and J. Skold, A simplified approach to applying the 3gpp spatial channel model, The 17th
Annual IEEE International Symposium on Personal, Indoor and Mobile
Radio Communications (PIMRC), 2006.
[134] S. Hamalainen, P. Slanina, M. Hartman, A. Lappetelainen, H. Holma,
and O. Salonaho, a novel interface between link and system level simulations, in Proceedings of ACTS Summit, Aalborg, Denmark, pp. 599-604,
October, 1997.

Multi-Cell Uplink Radio Resource Management: A LTE Case Study

Uploaded by

Copyright:

Available Formats

Multi-Cell Uplink Radio Resource Management: A LTE Case Study

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Multi-Cell Uplink Radio Resource Management: A LTE Case Study

Uploaded by

Copyright:

Available Formats

Multi-Cell Uplink Radio Resource

Long Term Evolution (LTE) er seneste version af mobilt bredb

by Klaus I. Pedersen, Nokia Siemens Network (NSN) - Aalborg, Denmark

3rd Generation Partnership Project

Adaptive Modulation and Coding

Average Mutual Information

Average Revenue per User

Automatic Repeat ReQuest

Adaptive Transmission Bandwidth

Actual Value Interface

Additive White Gaussian Noise

BLock Error Rate

Constant Amplitude Zero Auto-Correlation

Code Division Multiple Access

Cumulative Density Function

Close-Loop fractional Power Control

Coordinated Packet Scheduling

Cyclic Redundancy Check

Channel State Information

Combining Window Size

DeModulation Reference Signal

Exponential Effective SINR Metric

Equal Gain Combining

Evolved Packet System

Frequency-Domain Packet Scheduling

Fast Fourier Transform

Fixed Transmission Bandwidth

Fractional Power Control

Global System for Mobile Communication

Hybrid Automatic Repeat reQuest

Half Power Beam Width

High-Speed Downlink Packet Access

High-Speed Packet Access

High-Speed Uplink Packet Access

Inter-Cell Interference Coordination

Infinite Impulse Response

International Mobile Telecommunication-Advanced

Interference over Thermal noise

Independent Packet Scheduling

International Telecommunication Union

Key Performance Indicator

Log Likelihood Ratio

Long Term Evolution

Medium Access Control

Multiple Access Interference

Modulation and Coding Scheme

Multiple Input Multiple Output

Multiple Input Single Output

Mobility Management Entity

Minimum Mean Square Error

Maximal Ratio Combining

Orthogonal Frequency Division Multiplexing

Orthogonal Frequency Division Multiple Access

Outer Loop Link Adaptation

Open-Loop fractional Power Control

Peak-to-Average Power Ratio

Packet Data Convergence Protocol

Protocol Data Unit

Parallel Interference Cancellation