Analysis of Interference and Performance
in Heterogeneously Deployed LTE systems
MATTIAS BERGSTR ¨ OM
Master of Science Thesis
Stockholm, Sweden 2010
Analysis of Interference and Performance
in Heterogeneously Deployed LTE systems
MATTIAS BERGSTR ¨ OM
Master of Science Thesis performed at
Wireless Access Networks, Ericsson Research
September 2010
Supervisor: Konstantinos Dimou
Examiner: Ben Slimane
KTH School of Information and Communications Technology (ICT)
Department of Communication Systems (CoS)
CoS/RCS 2006-TRITA-ICT-EX-2011:6
Mattias Bergstr¨ c om, September 2010
Tryck: Universitetsservice AB
Abstract
Heterogeneous network deployment has been advocated as a mean to enhance the performance of cellular networks, but at the same time heterogeneous de- ployments give rise to new interference scenarios which are not seen in homoge- neous deployments. This report includes five studies pertaining heterogeneous network deployments which is based on simulations of LTE in high detail on the lower layer protocol stack. In the first study it is investigated if results from simulated systems with ideal deployments can be generalized to realistic low power node deployments, which is seen to be the case.
Three heterogeneous network configurations, specified by 3GPP, were com- pared to a macro-only system. It is observed that the gain from low power nodes is strongly connected to the distribution of UEs. If the UE distribution is uniform the UE throughput gain is below 100 % while if the UEs are highly clustered a UE throughput gain of 400 % is achieved.
The configuration with uniform UE distribution was further analyzed and it was seen that in a low load system the average UE throughput gain from low power nodes is below 20 %. In a low loaded system with uniform UE distribution adding low power nodes is not a good way of enhancing the system performance.
A study investigating the gain of low power node range extension showed that SINR problems arise if the range of the low power nodes is extended, however the system as a whole gets increased throughput. The same applies for UE throughput. The main reasons are macro layer offloading & reduced interference created by the macro layer.
It is showed that if more low power nodes are added the UE throughput gain per low power node increases. It is also showed that a system with two range extended low power nodes outperforms a system with four low power nodes without range extension. Inter-low power node interference is seen not to be a problem in the simulated system configurations.
iii
Acknowledgements
I would like to express my gratitude to my supervisor, Konstantinos Dimou, for his valuable input, guidance and commitment through this project. Kon- stantinos has always been supportive and found time for discussions around the project.
I am thankful to Johan Lundsj¨o, manager at RAN Architecture & Protocols, for giving me the opportunity to do this project here in Ericsson.
I would also like to thank my examiner, Ben Slimane, my colleagues; Peter Moberg, Gunnar Mildh, Michael Eriksson and Robert Baldemair for their input and discussions around the topic of this project and Jessica ¨Ostergaard for reminding me to go home after too long days in the office.
v
Contents
1 Introduction 1
1.1 The wireless system . . . 1
1.1.1 First generation . . . 1
1.1.2 Second generation . . . 1
1.1.3 Third generation . . . 2
1.1.4 Fourth generation . . . 2
1.2 Problem statement . . . 2
1.3 Thesis outline . . . 2
2 What is interference? 5 2.1 Frequency hopping . . . 6
2.2 Spatial multiplexing . . . 6
2.3 Beam forming . . . 7
2.4 Interference cancellation . . . 8
3 Fourth Generation cellular networks 11 3.1 Long Term Evolution . . . 12
3.1.1 OFDM . . . 12
3.1.2 Spectrum flexibility . . . 13
3.1.3 Multiple antenna technology . . . 13
3.1.4 Hybrid ARQ with soft combining . . . 13
3.2 LTE-Advanced . . . 13
3.2.1 Carrier aggregation . . . 14
3.2.2 Higher order MIMO . . . 14
3.2.3 Coordinated Multi-Point transmission and reception . . . 14
3.2.4 Heterogeneous network deployment . . . 15
4 New interference scenarios in Heterogeneous Networks 17 4.1 Downlink . . . 18
4.1.1 Low power eNB interference to macro UE . . . 18
4.1.2 Macro eNB interference to low power node UE . . . 18
4.2 Uplink . . . 18
4.2.1 Macro UE interference to low power eNB . . . 18
4.2.2 Low power node UE interference to macro eNB . . . 18
4.3 Crucial factors . . . 19
4.3.1 Cell association . . . 19
4.3.2 P0 offset . . . 21 vii
viii Contents
5 Impact of misplacement of low power nodes 23
5.1 Background . . . 23
5.2 Simulation details . . . 23
5.2.1 Performance Measurements . . . 23
5.2.2 Configurations . . . 25
5.2.3 System parameters . . . 25
5.2.4 Traffic model . . . 26
5.3 Results . . . 27
5.3.1 Performance overview . . . 27
5.3.2 User distribution . . . 27
5.3.3 Interference . . . 28
5.3.4 SINR . . . 30
5.3.5 Cell Throughput . . . 32
5.3.6 UE Throughput . . . 35
5.4 Conclusions . . . 37
6 Analysis of 3GPP system configurations 41 6.1 Simulation details . . . 41
6.1.1 Configurations . . . 41
6.1.2 System parameters . . . 42
6.1.3 Traffic model . . . 42
6.1.4 User distribution . . . 42
6.2 Uplink results . . . 43
6.2.1 Performance overview . . . 44
6.2.2 Cell throughput . . . 44
6.2.3 Interference . . . 48
6.2.4 SINR . . . 49
6.2.5 UE Throughput . . . 52
6.3 Downlink results . . . 53
6.3.1 Performance overview . . . 54
6.3.2 Cell throughput . . . 54
6.3.3 SINR . . . 55
6.3.4 UE Throughput . . . 59
6.4 Summary . . . 60
6.5 Conclusions . . . 61
7 Analysis of 3GPP system configurations - Low load 63 7.1 Simulation details . . . 63
7.1.1 Configurations . . . 63
7.1.2 System parameters . . . 63
7.1.3 Traffic model . . . 63
7.1.4 User distribution . . . 64
7.2 Results . . . 64
7.2.1 SINR . . . 64
7.2.2 Cell throughput . . . 65
7.2.3 UE Throughput . . . 67
7.3 Conclusions . . . 69
Contents ix
8 Analysis of 3GPP system configurations - Range extension 71
8.1 Simulation details . . . 71
8.1.1 Configurations . . . 71
8.1.2 System parameters . . . 71
8.1.3 Traffic model . . . 71
8.1.4 User distribution . . . 72
8.2 Results . . . 72
8.2.1 Cell Throughput . . . 72
8.2.2 Interference . . . 72
8.2.3 SINR . . . 76
8.2.4 UE Throughput . . . 81
8.2.5 Summary . . . 89
8.3 Conclusions . . . 89
9 Analysis of 3GPP system configurations - Multiple low power nodes 91 9.1 Simulation details . . . 91
9.1.1 Configurations . . . 91
9.1.2 System parameters . . . 92
9.1.3 Traffic model . . . 92
9.1.4 User distribution . . . 92
9.2 Results . . . 92
9.2.1 Cell Throughput . . . 92
9.2.2 Interference . . . 93
9.2.3 SINR . . . 95
9.2.4 UE Throughput . . . 97
9.3 Summary . . . 99
9.4 Conclusions . . . 99
10 Conclusions, proposal and future work 103 10.1 Conclusions . . . 103
10.2 Proposal . . . 105
10.2.1 Existing ICIC schemes . . . 105
10.2.2 Fractional Frequency Reuse . . . 105
10.2.3 Proposed scheme . . . 108
10.3 Proposed further studies . . . 110
10.4 Alternative technology . . . 111
Bibliography 113
List of Tables
3.1 Cell spectral efficiency requirements in IMT-Advanced. . . 11
3.2 Cell edge user spectral efficiency requirements in IMT-Advanced. 12 5.1 System parameters. . . 26
5.2 Uplink throughput. The numbers in the parentheses are the gains compared to the reference case. . . 27
5.3 Percentage of UEs connected to the low power nodes. . . 29
5.4 Macro PRB utilization. . . 29
6.1 3GPP heterogeneous network deployment configurations. . . 41
6.2 User distribution and macro PRB utilization. . . 43
6.3 FTP upload time. . . 44
6.4 Uplink throughput. The numbers in the parentheses are the gains compared to the reference case. . . 45
6.5 FTP download time. . . 54
6.6 Downlink throughput. The numbers in the parentheses are the gains compared to the reference case. . . 55
6.7 Gains from adding low power nodes in the different configurations compared to the reference case. . . 61
7.1 User distribution between macro eNB and low power nodes and macro PRB utilization. . . 64
7.2 Average uplink SINR per UE. . . 65
7.3 Average downlink SINR per UE. . . 65
7.4 Average uplink cell throughput per cell. . . 67
7.5 Average downlink cell throughput per cell. . . 68
7.6 Average uplink UE throughput per UE. . . 68
7.7 Average downlink UE throughput per UE. . . 69
8.1 User distributions and macro PRB utilization. . . 74
8.2 Gains from 8 dB range extension for the different configurations. 89 9.1 User distributions and macro PRB utilization. . . 92
9.2 Spectral efficiency vs. number of low power nodes per macro cell area. . . 93
9.3 UE throughput gain and UE throughput gain per low power node. Measured on the fiftieth percentile. . . 100
9.4 Gains from different number of low power nodes without and with 8 dB range extension compared to the reference case. . . 100
xi
List of Figures
2.1 Interference between two terminals TA and TB. . . 5
2.2 Example of beam forming. . . 8
3.1 Representation of bandwidth resources in LTE. . . 12
3.2 Examples of carrier aggregation. . . 14
3.3 Example of beam forming. . . 15
3.4 Joint processing of signals. . . 16
4.1 Heterogeneous deployment example. . . 17
4.2 Interference from low power eNB to macro UE. . . 18
4.3 Interference from macro eNB to low power node UE. . . 19
4.4 Interference from macro UE to low power eNB. . . 19
4.5 Interference from low power node UE to macro eNB. . . 20
4.6 Illustration of RSRP and path loss based cell association. . . 21
5.1 Distribution of UEs between macro and low power nodes. . . 28
5.2 Interference received by base stations. . . 30
5.3 CDF - average low power node uplink UE SINR. . . 31
5.4 CDF - average macro uplink UE SINR. . . 32
5.5 CDF - average uplink UE SINR including all UEs. . . 33
5.6 Average uplink cell throughput per cell. . . 33
5.7 Different low power node cell sizes depending on distance to macro node. . . 34
5.8 CDF - average uplink low power node cell throughput. . . 35
5.9 CDF - average uplink macro cell throughput. . . 35
5.10 CDF - average uplink macro cell area throughput. . . 36
5.11 CDF - average uplink low power node UE throughput. . . 37
5.12 CDF - average uplink low power node UE throughput. . . 37
5.13 CDF - average uplink UE throughput including all UEs. . . 38
6.1 User distribution between macro eNB and low power nodes in configuration 1, 4a and 4b. . . 42
6.2 Average uplink cell throughput per cell. . . 45
6.3 Path loss from one macro eNB and two low power nodes. The cell borders are marked with vertical lines. . . 46
6.4 CDF - average uplink low power node cell throughput. . . 47
6.5 CDF - average uplink macro cell throughput. . . 47
6.6 CDF - average uplink macro cell area throughput. . . 48
6.7 Time average uplink interference per PRB per cell. . . 48 xiii
xiv List of Figures
6.8 CDF - average low power node uplink UE SINR. . . 50
6.9 CDF - average macro uplink UE SINR. . . 50
6.10 CDF - distance from macro eNBs to their macro UEs. . . 51
6.11 CDF - average uplink UE SINR including all UEs. . . 51
6.12 CDF - average uplink low power node UE throughput. . . 52
6.13 CDF - average uplink macro UE throughput. . . 53
6.14 CDF - average uplink UE throughput including all UEs. . . 53
6.15 Average cell throughput. . . 55
6.16 CDF - average downlink low power node cell throughput. . . 56
6.17 CDF - average downlink macro cell throughput. . . 56
6.18 CDF - average downlink macro cell area throughput. . . 57
6.19 CDF - average low power node downlink UE SINR. . . 57
6.20 CDF - average macro downlink UE SINR. . . 58
6.21 CDF - average downlink UE SINR including all UEs. . . 59
6.22 CDF - average downlink low power node UE throughput. . . 60
6.23 CDF - average downlink macro UE throughput. . . 60
6.24 CDF - average downlink UE throughput including all UEs. . . . 61
7.1 Average uplink SINR per UE. . . 65
7.2 Average downlink SINR per UE. . . 66
7.3 Average uplink cell throughput per cell. . . 66
7.4 Average downlink cell throughput per cell. . . 67
7.5 Average uplink UE throughput per UE. . . 68
7.6 Average downlink UE throughput per UE. . . 69
8.1 User distribution between macro eNB and low power nodes in configuration 1, 4a and 4b with and without 8 dB range extension. 72 8.2 Average cell throughput per cell. . . 73
8.3 Time average uplink interference per PRB per cell. . . 75
8.4 Average SINR per UE. . . 77
8.5 CDF - average low power node UE SINR. . . 78
8.6 Path loss from macro eNB and macro UE. . . 79
8.7 Uplink interference from macro layer to low power node layer. . . 79
8.8 Downlink interference from macro layer to low power node layer. 79 8.9 CDF - average macro UE SINR. . . 80
8.10 CDF - average UE SINR including all UEs. . . 82
8.11 5 percentile SINR. . . 83
8.12 CDF - average low power node UE throughput. . . 84
8.13 CDF - average macro UE throughput. . . 85
8.14 CDF - average UE throughput including all UEs. . . 86
8.15 Example of a system map for configuration 4a. . . 87
8.16 Legend to figure 8.15. . . 88
8.17 5 percentile UE throughput. . . 88
9.1 Average cell throughput per cell. . . 93
9.2 Spectral efficiency vs. number of low power nodes per macro cell area. . . 94
9.3 Time average uplink interference per PRB per cell. . . 94
9.4 Average SINR per UE. . . 96
9.5 CDF - average low power node UE SINR. . . 96
List of Figures xv
9.6 CDF - average macro UE SINR. . . 97
9.7 CDF - average low power node UE throughput. . . 98
9.8 CDF - average macro UE throughput. . . 98
9.9 CDF - average UE throughput including all UEs. . . 99
10.1 Performance evaluation of ICIC schemes. . . 106
10.2 Static reuse ICIC scheme. . . 107
10.3 Fractional Frequency Reuse ICIC scheme. . . 107
10.4 Allocation order based ICIC scheme. . . 108
10.5 FFR scheme protecting UEs in range extended region of OA low power node cells. . . 109
10.6 FFR scheme protecting UEs in range extended region of CSG low power node cells. . . 110
10.7 Reception of transmission grant and downlink data transmission simultaneously. . . 111
List of Abbreviations
AMPS Advanced Mobile Phone System
CA Carrier Aggregation
CB Coordinated Beam Forming
CoMP Coordinated Multipoint transmission and reception
CS Coordinated Scheduling
CSG Closed Subscriber Group CSG Closed Subscriber Group
eNB E-UTRAN Node B
FDD Frequency-Division Duplexing FDMA Frequency-Division Multiple Access FFR Fractional Frequency Reuse
HARQ Hybrid Automatic Repeat Request
HeNB Home E-UTRAN Node B
HII High Interference Indication ICIC Inter-cell Interference Coordination ITU International Telecommunication Union
JP Joint Processing
JT Joint Transmission
LTE Long Term Evolution
LTE-Advanced Long Term Evolution-Advanced MIMO Multiple-Input-Multiple-Output NAT Network Address Translation NMT Nordic Mobile Telephone
OA Open Access
xvii
xviii List of Figures
OFDM Orthogonal Frequency-Division Multiplexing
OI Overload Indication
PRB Physical Resource Block
QAM Quadrature Amplitude Modulation QPSK Quadrature Phase-Shift Keying
RE Range extension
RNTP Relative Narrowband Downlink TX Power RSRP Reference Signal Received Power
SIC Successive Interference Cancellation SINR Signal-to-Interference-plus-Noise Ratio TDD Time-Division Duplexing
TDMA Time-Division Multiple Access TTI Transmission Time Interval
UE User Equipment
UMTS Universal Mobile Telecommunications System WCDMA Wideband Code-Division Multiple Access
Chapter 1
Introduction
1.1 The wireless system
The usage of cellular systems has been growing since the systems got deployed in the 1980s. The International Telecommunication Union (ITU) estimated 4.6 billion mobile subscriptions globally in 2009. In recent years the cellular systems have also started to be used for data traffic and in 2008 the number of mobile broadband subscriptions overtook the number of fixed broadband subscriptions.
What we want to achieve with a cellular system is to offer connections to the users anywhere at any time. The user demands of the cellular systems have also increased as the years have passed and new network architecture and technolo- gies are needed. After the first generation of cellular system was introduced in the 1980s a new generation has come about around once a decade. The fourth generation cellular systems is planned to be deployed in 2011.
1.1.1 First generation
The first generation of cellular systems, 1G, was introduced in the 1980s and was targeting voice communication. 1G systems are analogue where the users are separated in the frequency domain, so called Frequency Division Multiple Access (FDMA). NMT and AMPS are examples of 1G systems.
1.1.2 Second generation
The second generation of cellular systems, 2G, was digital. The digitalization of the system made it possible to send data traffic, enabling low rate data services such as SMS. The 2G systems also had higher capacities than the preceding analogue system because of the digitalization. The traffic could be compressed and multiplexed also in time, so called Time division Multiple Access (TDMA).
This gave more degrees of freedom which increased the capacity because of higher utilization of the bandwidth. Compared to 1G systems, where a channel was assigned a terminal even during times when it did not transmit, the second generation technologies could let several users transmit in parallel through time.
GSM is the most widespread 2G system.
1
2 Chapter 1. Introduction
1.1.3 Third generation
In the third generation cellular systems, 3G, the throughput was further in- creased which made services such as video calls possible. One of the most used 3G technologies is Universal Mobile Telecommunications System (UMTS) which uses Wideband Code-Division Multiple Access (WCDMA) to separate the users. WCDMA uses near-orthogonal codes to spread the terminals signals over a wider bandwidth making their signals look like Gaussian noise to each other. Since the terminals all use the same bandwidth, in which their signals appears as noise to each other, adding a terminal effectively adds noise. A new terminal can be added to the system as long as the noise is not exceeding a critical level. WCDMA is therefore said to have a soft terminal limit compared to a hard terminal limit as in the case with TDMA or FDMA where there is a fixed number of channels.
1.1.4 Fourth generation
For the fourth generation of cellular systems, 4G, the requirements are fur- ther increased and will have peak data rates of 100 Mbps for downlink and 50 Mbps in uplink. One promising technology to meet the 4G-standard is Long Term Evolution-Advanced (LTE-Advanced). LTE-Advanced is an evolution of a technology named LTE which has not fully met the requirements to be called a fourth generation technology. The requirements are found in [1].
Key technologies in LTE-Advanced that are making it possible to meet the requirements are Carrier Aggregation, multiple antennas, heterogeneous deploy- ment and coordinated transmissions between different base stations. LTE and LTE-Advanced are described in more detail in section 3.1 and 3.2 respectively.
1.2 Problem statement
The bandwidth used in radio communication is a scarce commodity and as the demands on the networks increase there is a need to make more efficient use of the bandwidth. To enhance the performance of cellular networks the following deployment approaches have been suggested; denser macro base station de- ployment, more advanced macro base stations and heterogeneous deployments.
Macro base stations are expensive and might take long time to deploy. Hetero- geneous deployments is an alternative in which lower power base stations are deployed where there are clusters of users with high traffic demands or in areas where the macro base stations has bad coverage. The low power base stations are cheaper and can be deployed without making a big impact on the rest of the network. On the downside, new interference scenarios follows heterogeneous deployments. This report will discuss interference scenarios and performance problems associated with heterogeneous deployment. Possible countermeasures will be presented and assessed.
1.3 Thesis outline
This report has the following structure.
1.3. Thesis outline 3
In chapter 2 a background to what interference is and how it arises will be presented. Common ways to mitigate interference in cellular systems will also be explained.
Chapter 3 gives an introduction to LTE and LTE-Advanced and their main technologies.
Chapter 4 explains the new interference scenarios associated with heteroge- neous network deployment.
Five studies have been performed for this report. First a study investigating the impact of misplacement of low power nodes is found in chapter 5. In simu- lations, unlike the reality, the placement of low power nodes is often ideal. The purpose of the study is to see how ideal versus non-ideal deployment affects the system performance.
3GPP has presented a set of system configurations which should be consid- ered when simulating heterogeneous networks. In chapter 6 these configurations have been simulated and the performance has been analyzed to find possible problems related to heterogeneous deployments.
In one of the configurations it was seen that adding low power nodes will not give much gain. In chapter 7 this configuration has been further analyzed, this time with lower load to see the benefits from adding low power nodes in that configuration.
To increase the gain from the low power nodes their cell sizes can be in- creased, so called range extension. A study pertaining range extension is found in chapter 8.
In chapter 9 a study is presented where the number of low power nodes is varied to see how the spectral efficiency and other performance measurements are affected. Another question this study answers is how serious interference between low power nodes is for the performance.
Conclusions, proposal and future work is found in chapter 10.1. The ma- jor interference problem seen arised when the range of low power nodes was extended. An ICIC-scheme is proposed to mitigate this interference. Joint Scheduling between Home eNBs and macro eNBs is proposed as future work.
Chapter 2
What is interference?
The capacity C of a communication channel with bandwidth B, such as the channel between a mobile phone and a base station, follows equation 2.1 ac- cording to Shannon’s Theorem.[2] SINR is the Signal-to-Interference-plus-Noise Ratio and is discussed below.
C = B × log2(1 + SIN R) (2.1)
If a transmitter TAtransmits a signal to its desired receiver RA, at the same time as a transmitter TB transmits a signal, not only will RA receive TAs signal but also the signal from TB. See figure 2.1. At the receiver the signals will superposition and from RAs point of view TBs signal will be interference. Sig- nal quality is in general quantified with Signal-to-Interference-plus-Noise Ratio (SINR). High interference leads to low SINR meaning low quality of the wanted signal.
TA
TB
RA RB
Figure 2.1: Interference between two terminals TA and TB.
In digital communication the receiver is trying to detect the transmitted data. The lower the SINR is the harder it is for the receiver to correctly detect the transmitted signal. When the SINR is below a threshold correct detection is not possible. This means that if the number of simultaneously transmitting users within a bandwidth is too high no detections will be correct.
In cellular networks the served area is divided in to smaller zones called cells.
A cell will have a base station and the terminals in the cell will be connected to 5
6 Chapter 2. What is interference?
its base station. To make efficient use of the bandwidth different cells can use the same bandwidth. This reuse of bandwidth introduces some interference as there is a possibility of terminals in different cells using the same bandwidth at the same time. To counter interference different methods can be used, some of which are discussed in this chapter.
2.1 Frequency hopping
When a terminal gets assigned a channel it can either be assigned a free channel or a channel used by other terminals in other cells. If assigned a free channel the terminal will not experience any interference. If assigned a channel used by another terminal they will interfere each other until one stops transmitting.
To counter this problem the terminals can at regular time intervals change channel. There will be a possibility of another collision but since the terminals will only stay in their channel for a limited time they will only be affected by the interference until the next frequency hop.1 The effect of frequency hopping can be seen as spreading the interference through time.
What is needed? The transmitter and the receiver need to agree on the hopping pattern.
• Pros
– Interference gets averaged though time which gives a more reliable transmission.
• Cons
– Transmitters and receivers need some complexity to make them able to change frequency during transmission.
– The transmitter and receiver needs to communicate in advance to agree on the hopping pattern.
2.2 Spatial multiplexing
The principle of spatial multiplexing is to increase the number of available trans- mission channels between transmitter and receiver. This can be achieved by hav- ing multiple transmitting and receiving antennas, a so called MIMO antenna setup. According to Shannon’s theorem the capacity is given by:
C = B × log2(1 + SIN R)
In a MIMO system with Ntantennas at the transmitter and Nrantennas at the receiver, theoretically, NL = min(Nt, Nr) different, uncorrelated paths can exist between them. The capacity of each channel is:
C = B × log2
1 + Nr
NL
SIN R
1As long as the terminals are not unlucky and jumps to the same channel again.
2.3. Beam forming 7
This gives a total capacity of:
C = B × NL× log2
1 + Nr
NL
SIN R
In theory, the capacity is increasing linearly with the number of channels, which can be created by adding antennas.[3]
In order for the receiver to demultiplex the data from the links it needs to know the properties of the created channels. This is achieved by having the transmitter transmit a known reference signal. The receiver estimates the channel properties from the received version of the reference signal and then tells the transmitter how it should code the data onto the antennas in order to get the best transmission.[4]
What is needed? Multiple antennas at the transmitter and receiver are needed. The receiver also needs to do channel estimation and feed it back to the transmitter.
• Pros
– Theoretical linear increase of the capacity within a given bandwidth.
• Cons
– Complex antenna structures.
– Channel estimation is required.
– Communication between the transmitter and received is needed.
2.3 Beam forming
Beam forming is to change the antenna beam pattern by use of array antennas.
The phase and amplitude of the signal is adjusted at each antenna element to form the beam pattern. The antenna beam pattern can be changed so that the main lobe is pointed towards a desired transmitter/receiver to achieve high antenna gain or to point the nulls in direction of undesired transmitters/receivers to avoid interference, see figure 2.2.[3]
To form the antenna beam the antenna array needs several elements spaced sufficiently far apart. Due to size limitations of mobile terminals beam forming is not suitable for terminals.
What is needed? Array antennas and feedback of measurements to the trans- mitter which are used to adjust the beam pattern is needed.
• Pros
– Transmitted power can be reduced due to higher antenna gains in main lobe.
– Interference can be reduced.
• Cons
8 Chapter 2. What is interference?
Figure 2.2: Example of beam forming.
– Advanced antenna structure with multiple antenna elements is needed.
– Pre-coding of the signal needs to be done before transmission.
– Not suitable for mobile terminals.
– Needs to sense the direction of the mobile terminals.
– Signaling between the terminal and the base station is needed.
2.4 Interference cancellation
In cellular networks several users can use the same bandwidth at the same time and therefore interfere each other. If a receiver can estimate the interfering signals they can cancel the interference by subtracting it. There are several ways of doing this, one of which is called Successive Interference Cancellation (SIC).
In SIC the transmitters are given different code words with which they en- code the signals before transmission. The receiver will try to demodulate and decode one of the signals from the received compound signal to extract its mes- sage. If successfully extracted the message is re-encoded, re-modulated and subtracted from the original signal. The procedure is repeated until all signals have been extracted.
As signals get subtracted the SINR is getting higher in each recursion. The most effective way of extracting the signals is therefore by starting with the highest SINR signal.[5] If a decoding error is made the wrong signal will be subtracted which will destroy the compound signal and the error will in that sense propagate to the next step.
What is needed? The receiver needs to know how each signal is modulated and encoded in order to decode and demodulate them. The structure differs de- pending on which cancellation method is used and can be more or less complex.
2.4. Interference cancellation 9
• Pros
– Ability to extract multiple signals which are interfering each other.
• Cons
– Complex receiver structure.
– Delay due to signal processing.
– Not always possible to decode.
Chapter 3
Fourth Generation cellular
networks
The International Telecommunication Union (ITU) has set the requirements for the fourth generation telecommunication systems, also called IMT-Advanced.[1]
The requirements are as follows:
• Peak spectral efficiency of 15 bit/s/Hz and 6.75 bit/s/Hz in downlink and uplink respectively1 .
• Data latencies of maximum 10 ms in both uplink and downlink.
• Latencies of maximum 50 and 150 ms for intra- and inter-frequency han- dovers respectively.
• Scalable bandwidth up to 40 M Hz.
• Increased cell spectral efficiency according to table 3.1. The test environ- ments are described in [6].
Test environment Downlink (bit/s/Hz/cell) Uplink(bit/s/Hz/cell)
Indoor 3 2.55
Microcellular 2.6 1.8
Base coverage urban 2.2 1.4
High speed 1.1 0.7
Table 3.1: Cell spectral efficiency requirements in IMT-Advanced.
• Increased cell edge user spectral efficiency according to table 3.2. The test environments are described in [6].
• Interworking with other radio access systems.
• Unicast and multicast broadcast services.
1Assuming an antenna configuration of downlink 4 × 4, uplink 2 × 4
11
12 Chapter 3. Fourth Generation cellular networks
Test environment Downlink (bit/s/Hz/cell) Uplink (bit/s/Hz/cell)
Indoor 0.1 0.07
Microcellular 0.075 0.05
Base coverage urban 0.06 0.03
High speed 0.04 0.015
Table 3.2: Cell edge user spectral efficiency requirements in IMT-Advanced.
As discussed in section 1.1.4 LTE-Advanced is one of the most promising technologies to reach the requirements for a fourth generation wireless com- munication system. The focus in this report is on heterogeneous deployments in LTE-Advanced and we will, in section 3.2, look in to more details about LTE-Advanced.
LTE-Advanced is an evolution of LTE which will be described first.
3.1 Long Term Evolution
Long Term Evolution (LTE) is an air interface for cellular networks which is defined by 3GPP. The main components of LTE are introduced in this section.
3.1.1 OFDM
In LTE Orthogonal Frequency-Division Multiplexing-based (OFDM) transmis- sion schemes are used for both uplink and downlink transmission. OFDM can be seen as a combination of TDMA and FDMA where the time is divided in to timeslots and frequency is divided into a large set of orthogonal narrow- band channels called sub carriers. Twelve sub carriers are grouped together into a Physical Resource Block (PRB), see figure 3.1. This separation of User Equipments (UEs) means that is no interference between UEs within a cell but intercell interference exists.
f
t
Physical Resource Block
Figure 3.1: Representation of bandwidth resources in LTE.
Before transmission the transmitter parallelizes the signal to several lower rate signals which gets modulated using QPSK, 16 QAM or 64 QAM. Each
3.2. LTE-Advanced 13
low rate signal will be transmitted on a separate sub carrier. The receiver then demodulates the signals and recreates the original signal before performing detection.
3.1.2 Spectrum flexibility
LTE supports both Frequency-Division Duplexing (FDD) and Time-Division Duplexing (TDD) to separate uplink and downlink communication.
Which band and bandwidth used by LTE is not specified in the standard.
This implies that operators can deploy LTE in a variety of frequency bands. An operator which previously deployed GSM in the 900 M Hz spectrum can deploy LTE there instead and because the bandwidth is not specified the transition from GSM to LTE can be done gradually.[4]
3.1.3 Multiple antenna technology
As discussed in chapter 2, it is beneficial to have several antennas for beam forming and spatial multiplexing. In LTE the terminals (UE in 3GPP terms) and base stations (eNB in 3GPP terms) supports up to two and four antennas respectively.[7][8]
3.1.4 Hybrid ARQ with soft combining
To cope with errors created in non ideal channels Hybrid ARQ (HARQ) is utilized in LTE. The transmitted data is coupled with two sets of redundant bits. One set of which is used by the receiver to first try to correct errors and another set which later is used to detect uncorrected errors.
After that the receiver has performed the correction of possible errors and detected whether the transmission was successful or not it will send a report to the transmitter of the outcome. In case of an erroneous transmission the transmitter resends the data.
In HARQ the erroneous packets are discarded. A packet with errors can however contain some valuable information which would be lost if the packet is discarded. To avoid this waste a modification of the Hybrid ARQ scheme has been done. Hybrid ARQ with soft combining will save erroneous packets to be combined with retransmitted packets. The combination of two or more packets will be more reliable and will have higher chance of a successful detection.[4]
3.2 LTE-Advanced
The LTE standard does not fully reach the ITU requirements for a 4G system and is sometimes called 3.9G. LTE-Advanced is, however, planned to reach those requirements. 3GPP’s aim is to have peak data rates of 1 Gbps in downlink and 500 M bps in uplink in a bandwidth of 100 M Hz. The spectrum efficiency will then be 30 bit/s/Hz and 15 bit/s/Hz in downlink and uplink respectively. The key components that will make this possible are, among others, Carrier Ag- gregation, higher order MIMO, Heterogeneous network deployment and CoMP which are described below.[9]
14 Chapter 3. Fourth Generation cellular networks
LTE-A LTE-A LTE-A LTE-A LTE-A
f
(a) Multiple component carriers
LTE-A LTE-A LTE-A
Other services
f
(b) Separated component carriers
Figure 3.2: Examples of carrier aggregation.
3.2.1 Carrier aggregation
In LTE the bandwidth can, as discussed in section 3.1.2, change in size. The bandwidth can be as narrow as around 1 M Hz up to 20 M Hz. Something which is new for LTE-Advanced is that it can be deployed using several frequency bands, adjacent or not, see figure 3.2a and 3.2b. The concept is called Carrier Aggregation (CA) in 3GPP terms where the bands used are called component carriers.
Carrier Aggregation will be backward compatible with LTE UEs. LTE UEs will, however, only be able to use one component carrier at one time while LTE-Advanced UEs can use several to reach higher data rates. Carrier Aggre- gation is an important component in reaching higher data rates in the sense that operators can deploy LTE-Advanced in frequency bands they already own and gradually migrate to LTE-Advanced as described in section 3.1.2 instead of buying new bandwidth for LTE-Advanced.
3.2.2 Higher order MIMO
Multiple-Input-Multiple-Output (MIMO) antenna configurations refer to the existence of multiple antennas at the transmitter and receiver. With multiple antennas multiple channels can be created between the transmitter and receiver for so called spatial multiplexing described in section 2.2. MIMO was included in the LTE standard with support for four antennas at the base station and two antennas in the UE. In LTE-Advanced it will be possible to have eight antennas at the base station and four in the UE, or even more.
3.2.3 Coordinated Multi-Point transmission and reception
Coordinated Multi-Point transmission and reception (CoMP) is a technology aimed to improve coverage of high data rates, cell edge performance as well as overall system performance. The principle of CoMP methods is to have sev- eral eNBs coordinating their transmissions. There are two categories of CoMP;
3.2. LTE-Advanced 15
Coordinated Scheduling/Coordinated Beam forming (CS/CB) and Joint Pro- cessing/Joint Transmission (JP/JT).
The first type, Coordinated Scheduling/Coordinated Beam forming, means that the involved eNBs are coordinating the access to the resource blocks in a way so that interference will be avoided. If, for example, one eNB is communi- cating with an edge UE the neighboring eNB should then avoid schedule one of its edge UEs at the same time. Beam forming can also be used in a way so that the eNBs coordinate their beams not to interfere with each other. See figure 3.3.
Figure 3.3: Example of beam forming.
In Joint Processing/Joint Transmission several cooperating eNBs are trans- mitting to one single UE. The data which is going to be transmitted to the UE therefore needs to be available at all involved eNBs. Interference can be avoided by having the cooperating eNBs process the signals in a way so that interfering signals will destruct at the UE. To achieve this, a lot of signaling is needed to be sent over the back haul at the same time as the eNBs have access to the channel conditions.[9] See figure 3.4.
One difficulty with CoMP is that if we want the eNBs to cooperate they need to be able to exchange messages within a few milliseconds to not be obsolete when arriving. This put latency and throughput restrictions on the connections between the nodes.[10]
3.2.4 Heterogeneous network deployment
Heterogeneous network deployment refers to a network where eNBs of different transmit powers, i.e. different cell sizes, is distributed in a nonuniform manner throughout the served area. To increase the performance and offer higher data rates it is possible to add eNBs with low output power at heavy loaded areas where the signal from the macro eNB is weak. Below is a description of the base stations in LTE-Advanced is specified.
16 Chapter 3. Fourth Generation cellular networks
Figure 3.4: Joint processing of signals.
• Macro eNB is the top level node. The UEs should be able to reach a macro base station from anywhere within the service area. The transmit power is typically around 43 dBm. The macro eNBs are connected to each other with a dedicated back haul connection.
• Relay eNB is a low power (23 − 30 dBm) eNB with a over-the-air back haul connection to the serving macro eNB.
• Pico eNB is a low power eNB which has a dedicated back haul connec- tion. The transmit power is usually around 23 − 30 dBm. The nodes are deployed by the operator.
• Femto eNB, or Home eNB (HeNB), as they also are called, are low power nodes that the users can buy and deploy where they need. Femto eNBs are connected to the rest of the network through the Internet. Since the users, instead of the operators, deploy femto eNBs planning is not possible for the femto eNBs. The femto eNBs can operate in two modes; open access or Closed Subscriber Group (CSG). If operating in open access any UE can connect to the node while in the CSG mode only authorized UEs can connect. The owner of a femto eNB can for example give access to its family and friends.[11]
Chapter 4
New interference scenarios
in Heterogeneous Networks
Heterogeneous network deployment both has benefits and drawbacks. It is bene- ficial to add low power nodes where the macro eNBs signal has problem reaching, inside buildings for examples. It is also beneficial to add low power nodes in high user density areas to support the high traffic. On the other hand, new in- terference scenarios are created which are not seen in traditional homogeneous deployments. Section 4.1.1 to 4.2.2 describes four interference scenarios related to heterogeneous deployments. Section 4.3 discusses how cell association and the UE target output power can be adjusted to mitigate interference.
Low power node 1
Low power node 2 Low power node 3
Figure 4.1: Heterogeneous deployment example.
17
18
Chapter 4. New interference scenarios in Heterogeneous Networks
4.1 Downlink
4.1.1 Low power eNB interference to macro UE
Low power nodes and macro eNBs normally use the same spectrum. Because of this, a macro UE close to a low power node might receive a stronger signal from the low power node than from the macro eNB which results in low SINR.
This effect gets worse in cases when the distance to the macro eNB is big and when the macro UE is close to the low power node. See figure 4.2.
Macro UE Low power node UE
Figure 4.2: Interference from low power eNB to macro UE.
4.1.2 Macro eNB interference to low power node UE
In case a low power node is close to the macro eNB the UEs connected to the low power node can get interference from the macro eNB. Since the macro eNB has higher output power than low power nodes there can be cases when the low power node UEs gets a stronger signal from the macro than from the low power node. The closer the low power eNB is to the macro eNB the stronger this effect gets. See figure 4.3.
4.2 Uplink
4.2.1 Macro UE interference to low power eNB
Low power nodes will receive interference from macro UEs. The further a UE gets from the serving eNB the higher power it transmits in order to reach the eNB. This effect gets stronger when the low power node is on the macro cell edge. See figure 4.4.
4.2.2 Low power node UE interference to macro eNB
When a low power eNB is close to the macro eNB the signals from the UEs in the low power cell can reach the macro eNB and therefore create interference.
4.3. Crucial factors 19
Low power node UE Macro UE
Figure 4.3: Interference from macro eNB to low power node UE.
Macro UE
Low power node UE
Figure 4.4: Interference from macro UE to low power eNB.
This is shown in figure 4.5.
4.3 Crucial factors
Aside from the factors given in section 4.1.1 to 4.2.2 other factors can affect the interference in the system, such as cell association and P0 offsets discussed below.
4.3.1 Cell association
As discussed, high interference can arise when UEs are close to an eNB that they are not connected to. Therefore cell selection in heterogeneous networks is an important factor to the system performance. The task of assigning UEs to base stations is non-trivial and there is no universally optimal way of solving the task.
20
Chapter 4. New interference scenarios in Heterogeneous Networks
Macro UE
Low power node UE
Figure 4.5: Interference from low power node UE to macro eNB.
If, for example, the cell association is optimized for downlink transmissions the upload transmissions will suffer and vice versa.
To optimize the downlink performance the UE should be assigned to the base stations from which the strongest signal is received. In this way the higher power a base station is transmitting the bigger the cell gets. This approach in cell association is called Reference Signal Received Power (RSRP).
To optimize uplink transmissions the UEs should be assigned to the base stations to which the path loss is lowest. This way of path loss based cell association will make the UEs connect to the base station which will have the best potential to receive it.
Figure 4.6 shows these two ways of association UEs with the low power nodes. If RSRP cell association is used the low power node cell will be smaller having the blue cell border. If path loss based cell association is used the cell border will be larger and have the red cell border. In either case the UEs in the yellow region will create or receive interference.
If RSRP cell association is used the UEs in the yellow region will be con- nected to the macro eNB for optimal downlink performance. As seen in the figure the UEs in the yellow region will be closer to the low power node but connected to the macro eNB. This means that the low power node will receive a stronger version of their signal than the macro eNB and uplink performance is not optimal. They will also create interference to the low power node described in section 4.2.1.
If path loss based cell association is used the UEs in the yellow region will connect to the low power nodes. In this case they will be connected to the base station which will get the strongest version of their transmitted signal which will optimize uplink performance. In downlink there will be problems. The UEs in the yellow region gets a stronger signal from the macro eNB compared to the low power node and the signal received from the macro eNB is interference to them.
A compromise between RSRP and path loss based cell association is to use RSRP with offsets. When comparing the received power from two base stations,
4.3. Crucial factors 21
Figure 4.6: Illustration of RSRP and path loss based cell association.
say a macro eNB and a low power eNB, an offset is added to the measured received power from the low power node resulting in that the UEs will with higher probability connect to the low power node. This can be thought of as enlarging the low power cells without changing their output power and is called range extension (RE).
4.3.2 P
0offset
As we saw earlier in this chapter the difference in output power between low power nodes and macro nodes creates interference problems. A macro UE just outside the cell border of a low power node can create strong uplink interference to the low power node. See section 4.2.1.
To overcome this problem the low power node can tell its UEs to increase their output power to fend the high interference.[12]
Chapter 5
Impact of misplacement of
low power nodes
The following study will show how misplacement of low power nodes within a hot zone will affect the performance of the system.
5.1 Background
In cellular networks users tend to gather in certain areas, such as in a shopping mall or a busy square, forming so called hot zones. To support the high traffic in a hot zone a low power base station can be deployed in it.
Hot zones are often modeled in an ideal manner as a perfect circle in which a low power node is placed in the center. In reality a hot zone is defined by the location of the UEs. The shape and location of hot zones therefore change as the UEs move and the low power nodes are in general not located in the center of the hot zones. The aim of this study is to see how the performance is affected by having non perfect deployment compared to perfect deployment of low power nodes within hot zones. To investigate this there is a need to see how the distribution of UEs between the low power nodes and the macro nodes together with the SINR distributions are changing in the different deployment scenarios. The conclusions obtained for uplink are applicable to downlink as well.
5.2 Simulation details
To perform the simulations in this report a simulation tool which simulates LTE in high detail on the lower layer protocol stack has been used. System parameters such as traffic model, propagation model and deployment are input in the simulator and the output has then been processed in MatLab.
5.2.1 Performance Measurements
In this section details about the performance measurements are described. The performance measurements are calculated in the same manner in all studies in
23
24 Chapter 5. Impact of misplacement of low power nodes
this report.
PRB utilization
In each Transmission Time Interval (TTI) the PRB utilization is calculated by dividing the number of PRBs used for transmission by the total number of PRBs, according to equation 5.1. The PRB utilization is averaged over the whole simulation time.
P RB utilization = number of P RBs used f or transmission
total number of P RBs (5.1)
Interference
The base stations will sum the total received power under a time t seconds.
After t seconds the interference is calculated by subtracting the power of useful signal from the total power. The interference is calculated according to equation 5.2 and is averaged over time, PRB and cell and presented in dBm.
Interf erence = 10 × log10(T otal received power − U sef ul signal power) + 30 (5.2) The time t is 0.2 seconds in these simulations.
SINR
The SINR is the useful signal in a transmission divided by the interference plus noise, see equation 5.3. The SINR is presented in dB.
SIN R =U sef ul signal power
Interf erence (5.3)
The SINR is averaged over a time t = 0.2 s Cell throughput
The cell throughput is calculated by starting a timer and having a counter count the number of received bits. After a time t the simulator calculates the throughput according to equation 5.4 after which the number of received bits is set to zero before the counter is restarted.
Cell throughput = N umber of received bits
t (5.4)
Where t = 0.2 s.
UE throughput
The UE throughput is calculated in a similar way as the cell throughput, see equation 5.5.
U E throughput = N umber of received bits
t (5.5)
Where t = 0.2 s.
5.2. Simulation details 25
5.2.2 Configurations
Two cases have been simulated. First the low power nodes have been placed, as they often are in simulations, in the center of the hot zone, from here on refereed to as bingo deployment. Thereafter the low power nodes have been placed randomly within the hot zones, referred to as random deployment. Within the hot zones 50 % of the users are placed, while the rest of the users are distributed randomly within the system.
• No low power nodes. (Reference case)
• Bingo deployment. One hot zone per macro cell area where 50 % of the UEs are placed. A low power node is deployed in the center of each hot zone.
– 0 dB Range extension – 8 dB Range extension – 16 dB Range extension
• Random deployment. One hot zone where 50 % of the UEs are placed. A low power node is deployed at a random location within the hot zone.
– 0 dB Range extension – 8 dB Range extension – 16 dB Range extension
5.2.3 System parameters
Range extension has been achieved by changing the cell association algorithm.
The UEs measure the received signal power from the all base stations from which they receive a signal. For all low power nodes an offset is added to the received power. The UEs then connect to the base station which has the highest value.
The system parameters are found in table 5.1. The reason for not having shadow is to make the simulations run faster.
The propagation model is defined by the following two equations. The gain from a macro eNB to a UE is follows equation 5.6 and the gain from a low power node to a UE follows equation 5.7.
Gain = −35.3 − 3.76 × 10 log10(distance) + 14 − min
12
angle
70 360× 2π
, 20
(5.6) Gain = −50.6 − 3.67 × 10 log10(distance) (5.7) where distance is the distance from a UE to its base station and angle is the angle between the UE and middle of the base station antenna beam.
26 Chapter 5. Impact of misplacement of low power nodes
Parameter Value
Deployment
Number of macro base stations 7
Number of cells per macro base station 3
Hot zone radius 40 m
Cell radius 167 m
Macro to macro distance 500 m
Minimum LPN to LPN distance 75 m
Minimum LPN to macro distance 75 m
Resources
Bandwidth 10 MHz
Number of PRBs 50
Propagation
Macro propagation factor - 3.76
Macro attenuation constant - 35.3
Low power node propagation factor - 3.67 Low power node attenuation constant - 50.6
Shadow fading -
Base station specifics
Noise figure 5 dB
Macro base station output power 40 W
Macro base station antenna elements (per cell) 2 Low power base station output power 1 W Low power base station antenna elements 2
Transmit antenna ports 1
Receive antenna ports 2
UE specifics
Speed 0 m/s
Output power 0.2 W
Noise figure 9 dB
UE antenna elements 2
Transmit antenna ports 1
Receive antenna ports 2
Miscellaneous
UE scheduling algorithm Round robin
Table 5.1: System parameters.
5.2.4 Traffic model
The traffic model is chosen to comply with the Poisson based traffic model 1 specified in [9]. Users arrive in the system following a Poisson distribution with an arrival intensity of λ users per second. They upload or download one FTP packet of fixed size and then disappear from the system.
• λ: 150 UE/s system wide. (7.14 UE/s/cell)
• FTP packet size: 100 kByte
5.3. Results 27
This traffic model was chosen in order generate fixed offered traffic regardless of how the system performs in different situations. The traffic model generates the following offered traffic.
• 120 Mbps system wide.
• 5.712 Mbps per macro cell area.
Simulation time is 100 seconds during which 14947 UEs was created, i.e.
149.47 UEs / second.
5.3 Results
The following results were obtained by computer simulations. Only uplink per- formance has been analyzed in this study.
5.3.1 Performance overview
In table 5.2 the throughput performance has been summarized.
Referencecase 0dBBingo 0dBRandom 8dBBingo 8dBRandom 16dBBingo 16dBRandom Macro cell area
5.5 5.8 5.8 5.8 5.8 5.8 5.8
throughput (Mbps) (5%) (5%) (5%) (5%) (5%) (5%)
Macro cell
5.5 4.4 4.6 3.2 3.5 2.4 2.4
throughput (Mbps) Low power node
- 1.4 1.2 2.7 2.4 3.4 3.4
throughput (Mbps) Spectral efficiency
0.55 0.58 0.58 0.58 0.58 0.58 0.58
(bps/Hz/Macro cell area) (5%) (5%) (5%) (5%) (5%) (5%)
5 % UE
0.014 0.78 0.73 1.0 0.98 1.1 1.1
throughput (Mbps) (5500%) (5100%) (7000%) (6900%) (7800%) (7800%)
50 % UE
1.0 1.5 1.4 1.6 1.6 1.7 1.7
throughput (Mbps) (50%) (40%) (60%) (60%) (70%) (70%)
95 % UE
1.6 1.8 1.8 1.9 1.8 1.9 1.9
throughput (Mbps) (15.5%) (15.5%) (18.8%) (15.5%) (18.8%) (18.8%)
Table 5.2: Uplink throughput. The numbers in the parentheses are the gains compared to the reference case.
5.3.2 User distribution
The number of UEs in the hot zones is 50 % in all configurations. To cover the whole hot zone means that we should see 50 % of the UEs connection go the
28 Chapter 5. Impact of misplacement of low power nodes
low power nodes.1 In figure 5.1 and table 5.3 the percentage of UEs connection to the low power nodes is displayed.
No low power nodeBingo Random Bingo 8dB Random 8dBBingo 16dBRandom 16dB 0
500 1000 1500 2000 2500 3000 3500
Number of Users
User distributions
Macro users Low power node users
Figure 5.1: Distribution of UEs between macro and low power nodes.
From table 5.3 we can see that when RSRP is used without any offset 25 % and 20 % of the UEs are connection to the low power nodes in the bingo and random case respectively. This means that 50 % and 40 % of the hot zone is covered by the low power cell. When the offset is increased a larger portion of the hot zones are covered by the low power nodes and in the case of 16 dB range extension we can see that the whole hot zone is covered. Comparing the values in the bingo and random deployment cases it is seen that the bigger the cell is, i.e. the larger offset is used, the smaller the impact of misplacement is on the number of UEs connecting to the low power nodes.
The PRB utilization in the macro layer is compiled in table 5.4.
5.3.3 Interference
Figure 5.2 is showing the average interference received by the base stations.
The difference in interference between bingo and random deployment is due to different number of UEs connecting to the low power nodes. The relation between number of UEs connecting to the low power node and the interference is discussed below.
Low power eNB
A decrease in interference to the low power nodes is observed as the offsets gets larger. This is explained by that a low power node gets the strongest interference from macro UEs surrounding the cell. The number of UEs in the hot zones is
1The UE which are not placed in the hot zones intentionally are randomly distributed throughout the system area. There is a chance that a UE not chosen to be placed in the hot zone are placed there anyway. This means that to cover the whole hot zone a low power node should actually have more than 50 % of the UEs connected to it.
5.3. Results 29
Bingo Random IncreaseBingovs.Random 0 dB 25 % 20 % 25 % 8 dB 46 % 39 % 18 % 16 dB 59 % 56 % 5 %
Table 5.3: Percentage of UEs connected to the low power nodes.
Referencecase 0dBBingo 0dBRandom 8dBBingo 8dBRandom 16dBBingo 16dBRandom Macro uplink
80 % 65 % 70 % 48 % 54 % 39 % 41 % PRB utilization
Table 5.4: Macro PRB utilization.
the same regardless of the offset abut what differs is the number of UEs which are absorbed by the low power nodes. In the case without offset, there will be a large number of surrounding UEs which are connected to the macro eNB and the interference is -93.6 dBm and -94.1 dBm in the Bingo and Random case respectively. If an offset is added those surrounding UEs are absorbed by the low power node and therefore will not interfere to it and in case of a 16 dB range extension the whole hot zone is covered and the interference is reduced to -111 dBm and -110 dBm. This effect was earlier explained in section 4.2.1.
Macro eNB
There are two factors affecting the interference to macro eNBs as the offset changes. The dominant interferers to a macro eNB are the edge UEs in neigh- boring cells and the UEs connected to low power nodes within its own cell.
• By increasing the offsets of a low power node, hence assigning more UEs to it, there will be more possible interferers to the macro eNB. UEs which earlier were intra cell UEs have become inter cell UEs when absorbed by the low power nodes and therefore will interfere with the macro cells.
• On the other hand, in neighboring cells edge UEs are absorbed by the low
30 Chapter 5. Impact of misplacement of low power nodes
Low power node Macro Macro cell area
−130
−120
−110
−100
−90
−93.6 −94.1
−104
−103
−111 −110
−102
−105 −104
−107 −106
−108 −107
−102
−96.3 −96.7
−106
−104
−109 −108
Interference (dBm)
Interference Received by Base Stations − Averages
Ref 0 dB Bingo 0 dB Random 8 dB Bingo 8 dB Random 16 dB Bingo 16 dB Random
Figure 5.2: Interference received by base stations.
power node as well. Those UEs will be closer to their serving eNB and will transmit with less output power and therefore interfere less.
From the interference reduction we can conclude that the interference added by the low power node UEs is smaller than the reduction of interference from the neighboring cells.
It can be seen that the interference from neighboring cells decreases and compensates for the interference from the low power nodes.
Overall
Summing the interference received by the low power nodes and the macro nodes shows that it is possible to get lower interference than in the homogeneous deployment case if the range is extended.
It can be concluded that the interference depends on how many UEs are handed over to the low power node. If a low power node is misplaced it will have fewer UEs connecting to it and therefore the interference will be stronger.
The difference in interference between bingo and random deployment is around 1 dB in all configurations.
5.3.4 SINR
Which modulation scheme (QPSK, 16 QAM or 64 QAM) can be used for a transmission depends on the SINR level. With high SINR higher modulation schemes can be used, hence utilizing the bandwidth more efficiently. In this section the SINR for the UEs is analyzed. The SINR is calculated from equation 5.8.
SIN R = S
I + N (5.8)
5.3. Results 31
Low power node users
In figure 5.3 a CDF over the SINR for UEs connected to low power nodes is shown. When the range is extended the following two things will happen.
1. The average distance from the low power nodes to their UEs will increase giving an average higher path loss and lower SINR.
2. The interference decreases which will give higher SINR, mainly to the edge UEs. An explanation to why edge UEs are mostly affected by the uplink interference reduction is found in section 8.2.3.
For the high percentiles the SINR seems to decrease when using range exten- sion. The reason for this is described in point 1. Worth noting is that the UEs who were connected to the low power node in the case without range extension will get higher SINR when range extension is applied due to lower interference.
In the lower percentiles the edge UEs are found. In the case without range extension the edge UEs are closer to the low power node compared to the cases with range extension. When the range is extended the edge UEs will have higher path loss which is reducing the SINR but at the same time range extension reduced the interference and since the interference reduction is larger than the higher path loss a higher SINR is achieved. The path loss from the low power node to its edge will be 8 or 16 dB when the range is extended. At the same time the interference will in those cases be 10.4 and 17.4 dB lower respectively resulting in a gain.
−200 −10 0 10 20 30 40
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SINR (dB)
CDF
Uplink UE SINR − Low power node UEs
0 dB Bingo 0 dB Random 8 dB Bingo 8 dB Random 16 dB Bingo 16 dB Random
Figure 5.3: CDF - average low power node uplink UE SINR.
Macro users
Figure 5.4 shows a CDF over the macro UE SINR. When deploying a low power node the number of UEs connecting to it will depend on its distance to the macro
32 Chapter 5. Impact of misplacement of low power nodes
eNB. A low power node on the edge of the macro cell will absorb more UEs than a low power node deployed close to the macro eNB. This means that the low power nodes will, on average, absorb more edge UEs compared to center UEs.
The observed gain in SINR in the low percentiles is not a direct gain but rather a gain coming from removing edge UEs from the macro cells which therefore will not be present in the macro SINR CDF.
The higher the offset is the more UEs will be absorbed by the low power nodes and the bigger gain is seen.
The UEs in the high percentiles are those close to the macro eNB. Those UEs are not as likely to be absorbed by the low power nodes and will only gain from lower interference.
−100 −5 0 5 10 15 20 25 30
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SINR (dB)
CDF
Uplink UE SINR − Macro UEs
Ref 0 dB Bingo 0 dB Random 8 dB Bingo 8 dB Random 16 dB Bingo 16 dB Random
Figure 5.4: CDF - average macro uplink UE SINR.
All users
A CDF for all UEs average SINR is shown in figure 5.5. The SINR is higher when the low power nodes are deployed in the center of the hot zones. We also see that the importance of bingo deployment is also reduced as the offset increases.
5.3.5 Cell Throughput
In this section the cell throughput is discussed.
Averages
Figure 5.6 shows the average cell throughput. We see the effect of the low power node offloading the macro cells. In the reference case the served traffic was 5.54 Mbps but when the low power nodes are deployed the served traffic increased to 5.81 Mbps.
