DRA T
Energy Efficiency Performance Improvements Through Two-Tier Cellular Networks
ZHIHAO ZHENG
Master of Science Thesis
Stockholm, Sweden 2012
i
Energy Efficiency Performance Improvements Through Two-Tier Cellular Networks
ZHIHAO ZHENG
Master of Science Thesis performed at the Radio Communication Systems Group, KTH.
October 2012
Examiner: Professor Jens Zander
Supervisor: Sibel Tombaz
ii
KTH School of Information and Communications Technology (ICT) Radio Communication Systems (RCS)
C
Zhihao Zheng, October 2012
iii
Abstract
Mobile communication network is consuming 0.5 percent of the global energy supply alone these days. While the unrelenting increase in user capacity requirement, which is expected to grow 1000 times in 10 years, will lead to even more energy consumption in this filed. On the one hand, the energy cost which has covered half of the operators’ operation fee will continually increase; On the other hand, global warming problem may be still more severe due to the rising amount of carbon emissions caused by the increasing energy consumption.
Although the system spectrum efficiency is improved significantly by the introduction of
latest cellular network standards 3G and 4G, and the incremental enhancements in electronics
and signal processing are bringing the energy consumption down in base stations, these
improvements are still not enough to match the huge increase in energy consumption which is
realted to the surging increase demands for more capacity. It is clear that solutions have to be
found at the architectural layer. Besides, more and more capacity requirement is generating
from indoor environment, it has been a vital concern for operators to consider how to provide
enough service coverage with the existing macro-only network due to the complex indoor
environment and high wall penetration loss. All these cause urgent demand for more energy
efficient cellular networks to deal with the capacity challenges. One of the promising
technologies to this situation is macro-femto heterogeneous networks by offloading indoor
traffic to femtocells. Considering the shorten distance between indoor users and base stations
and spacial frequency reuse advantages, as well as the drawbacks on introduced interference
and extra power consumption by femto base stations, this report proposes a simulation
framework to study whether this network topology could enhance the system energy
efficiency or not. The results suggest that the introduction of femtocells may be a promising
method to save energy consumption in the future and meet the increasing user data rate access
requirements, especially in high user capacity demand networks, macro-femto deployment
could save more energy consumption.
iv
v
Acknowledgements
I will finish my master study in the program of wireless systems in Royal Institute of Technology (KTH) by presenting this thesis. During these six-month work, my supervisor Sibel Tombaz always helped me out from my problems and made sure that I was going to the right direction until the final report, regardless she was extremely busy with her own PHD studies. I want to express my sincere thanks and deep appreciations to her and hope her all the best for the left studies. I would like also to thank my examiner Prof. Jens Zander for his valuable comments and advices.
Last but not the least, I want to thank my family for their repetitive encouragement and
enthusiastic regarding during my thesis work and my whole study life in KTH. Especial
thanks go to my parents for their emotional supporting and continuous trust in me during my
studies oversea.
vi
vii
Contents
Abstract ... iii
Acknowledgements ... v
Contents ... vii
List of figures ... ix
Abbreviations ... xi
Chapter 1 Introduction ... 1
1.1 Background ... 1
1.2 Related Work ... 3
1.3 Problem Definition ... 4
Chapter 2 Definitions and Problem Formulation ... 5
2.1 Definitions ... 5
2.1.1 Energy Efficiency ... 5
2.1.2 Cell Coverage ... 5
2.1.3 Area Spectral Efficiency ... 6
2.1.4 Area Capacity ... 6
2.1.5 Area Power Consumption ... 6
2.2 Problem Formulation ... 7
Chapter 3 System Model and Methodology ... 9
3.1 System Model ... 9
3.1.1 Network Deployment Model ... 9
3.1.2 Spectrum Utilization Scheme ... 10
3.1.3 Propagation Model ... 10
3.1.4 Power Consumption Model ... 11
3.2 Methodology ... 12
3.2.1 Simulation Procedure ... 13
3.2.2 Simulation Parameter ... 13
Chapter 4 Simulation Results ... 15
viii
4.1 Energy Efficiency of Macro-Femto Deployment in Over-Provisioned Network... 15
4.2 Impact of Network Capacity Requirement on Total Power Consumption ... 20
4.3 Impact of Macro Base Station Transmit Power... 25
4.4 Impact of Indoor User Ratio on Total Power Consumption ... 31
Chapter 5 Conclusion ... 35
Reference ... 37
ix
List of figures
Figure 1.1 The growing gap between operator’s revenues and traffic growth ... 2
Figure 2.1 Interference generation among MBSs and FBSs ... 8
Figure 3.1 Network deployment model ... 10
Figure 3.2 Flow Chart of simulation scenarios... 12
Figure 3.3 Simulation Flow Chart ... 13
Figure 4.1 SINR map of Macro-Only Network ... 16
Figure 4.2 SINR map of Macro-Femto network ... 16
Figure 4.3 Spectral efficiency CDF performance comparison ... 17
Figure 4.4 User throughput cdf plot for different Rp deployment schemes ... 18
Figure 4.5 System integrated throughput versus femtocell penetration rate ... 19
Figure 4.6 Energy efficiency versus femtocell penetration rate ... 19
Figure 4.7 Simplified network development model for scenario 2 ... 20
Figure 4.8 User spectral efficiency CDF in ISD of 2100 meter case ... 21
Figure 4.9 User spectral efficiency CDF for ISD of 500 meter case ... 22
Figure 4.10 Area spectral efficiency versus inter-site distance ... 22
Figure 4.11 Area power consumption versus inter-site distance ... 23
Figure 4.12 Area power consumption for different area spectral efficiency targets ... 24
Figure 4.13 Area power consumption for different area capacity targets ... 24
Figure 4.14 Area power consumption over different ISD ... 25
Figure 4.15 Area spectral efficiency versus inter-site distance in scenario 3 ... 27
Figure 4.16 Area power consumption versus inter-site distance in scenario 3 ... 27
x
Figure 4.17 Area spectral efficiency, capacity and power consumption performance ... 27
Figure 4.18 Area power consumption for different area capacity targets in scenario 3 ... 28
Figure 4.19 Area power consumption comparison between scenario 2 and 3 ... 29
Figure 4.20 Area power consumption comparison between scenario 2 and 3 ... 29
Figure 4.21 Area power consumption over different ISD ... 30
Figure 4.22 Area spectral efficiency versus inter-site distance in scenario 4 ... 31
Figure 4.23 Area power consumption versus inter-site distance in scenario 4 ... 32
Figure 4.24 Area power consumption in different indoor user ratio cases ... 33
Figure 4.25 Area power consumption for different spectral efficiency targets ... 33
xi
Abbreviations
ICT: Information and Communication Technology
MBS: Macro Base Station
FBS: Femto Base Station
RF: Radio Frequency
ISD: Inter-Site Distance
Rp: Femtocell Penetration Rate
Indoor user: Users Inside of the House Outdoor user: Users Outside of the House Macro user: Users Served by Macrocells Femto user: Users Served by Femtocells
Ir: Indoor User Ratio
Ptx: Transmit Power
QoS: Quality of Service
OPEX: Operational Expenditures
HetNet: Heterogeneous Network
SINR: Signal to Interference and Noise Ratio
AWGN: Addictive White Gaussian Noise
xii
1
Chapter 1 Introduction
The introduction of this thesis work will be given in this chapter. With the background of huge increase in traffic demand in the near future, the world energy consumption causes problems to the telecommunication operators and even the whole society. Quite much work has been done in this filed, but there is still something missing. By the end of this chapter, we discuss the objective of this work and the problem will be formulated with some research questions.
1.1 Background
A hundred-fold to thousand-fold of data traffic is forecasted to be demanded by 2020, which is due to the unrelenting demand of subscriptions from smart-phones, laptops and tablets [1]. To make it further, data from Cisco Visual Networking Index shows global mobile data traffic will increase 18- fold between 2011 and 2016 growing at a compound annual growth rate of 78 percent, reaching 10.8 Exabyte per month by 2016. The global mobile data traffic per month will surpass 10 Exabyte in 2016 [2]. This creates a big challenge for the existing cellular networks and operators.
It has been shown that the explosive development of Information and Communication Technology
(ICT) has become a major contributor to world’s energy consumption, which is 10% of the worldwide
energy consumption already in 2010. To meet this surging capacity increase challenge in wireless
broadband access will further increase the world energy consumption, which causes the increasing
network operation fee for the operators; while on the other hand, the global warming problem will
become even serious. The current cellular systems based on third generation (3G) and 4G (LTE)
improve system spectrum efficiency extremely (more than 10 times over 2G and 4G is 3-4 times over
3G), but the drawbacks beyond these schemes are the requirements of very linear power amplifiers
which is much less energy efficient. Data shows, to offer the same level of coverage, a LTE network
requires 60 times more energy consumption from a 2G network [1]. What’s more, the increasing
energy prices make the situation even worse. In some countries, the Energy bills that the operators
have to pay for already constitutes around 50% of their operational expenditures (OPEX) [3], and the
2
expanding energy bills lead to the increasing gap between traffic demand and operators’ revenues before new technologies come up as shown in Fig. 1.1 [4].
The other problem is global warming caused by huge carbon emissions. Among these, ICT is estimated to account for 2-4 percent of worldwide carbon emissions. The power consumption during
the operating phase of the equipment takes around 40-60 percent of carbon emissions. A significant part of these emissions about one sixth comes from telecommunication networks [5]. By 2020, these emissions are expected to double if no initiatives are taken to reduce this footprint. The increasing concern of global warming issue pushes the government to commit to reduce their carbon emissions, for example, the United Kingdom has set goal to reduce carbon emission to 80 percent of the 1990 level in the coming 60 years from then [6]. Based on the previous discussion, it is quite necessary for the future cellular networks to support improved system capacity, at the same time, save more power consumptions.
Ongoing incremental improvements in electronics and signal processing are bringing down the total energy consumption in BSs, but these improvements are not enough to match the orders-of-magnitude increase in energy consumption cause demands for more capacity. It is clear that solutions to this problem have to be found at the architectural level [1], where heterogeneous network (HetNet) is an attractive means. A typical HetNet is composed of multiple radio access technologies, architectures, transmission solutions, and base stations of varying transmission power [7]. A wide area network can be coved by macro base stations (MBSs), while small base stations like Microcells, Picocells or Femtocells are deployed to provide capacity holes and fulfill users’ expectations for high-speed mobile broadband services. Moreover, a low power small base station has significantly lower transmission power than its surrounding MBSs. It may be a good solution to provide huge capacity at the expense of lowest extra power consumption compared with the existing homogeneous network.
Furthermore, 74% of future traffic will be generated from indoor environment by 2015 as shown in [8]. It is known that network capacity and data throughput depend on signal quality, but indoor signal
Figure 1.1 The growing gap between operator’s revenues and traffic growth
3
is by nature the worst due to the complex indoor environment and high wall penetration loss. The indoor coverage is the number one customer complaint by now.
Based on these facts, macro-femto heterogeneous network is a promising method, where a femtocell is a small, low-power cellular base station operating in licensed spectrum; it is typically designed for use in a home or small business, with a range on the order of 10 meters, while the macrocell is providing coverage service, especially for users outside of the buildings. For the operators, the attractions of a femtocell are improvements to both coverage and capacity, especially indoor. Also very little upfront cost to the operation fee, and the power consumption of BSs is taken by customers. For the customers, they may benefit from the improved coverage and potentially better signal quality and battery life.
Compared with the conventional approaches like Wifi, which requires dual-mode handsets, femtocells promise fixed mobile convergence with existing handsets without re-configuration [9].
1.2 Related Work
There are numerous studying papers related to energy efficiency of macro-femto heterogeneous cellular networks. Among these, most papers support femtocells to be energy efficient, but there is also a few disagree.
In [10], two scenarios for HetNets, namely a joint macro-relay network and a joint macro-femto network with different relay and femtocell deployment densities are evaluated. The results indicate that compared with macro-centric networks, macro-relay joint networks are both energy and cost efficient, while macro-femtocell networks reduce the networks total-cost-of-ownership at the expense of increased energy consumption, and the energy and cost gains are highly sensitive to the OPEX (Operational expenditure) model adopted. Also in [11], it shows the macrocell offloading benefits in interference limited system and noise limited system, and the latter one can gain 30% to 100% in system capacity, but almost no gain is achieved to the former case.
However, simulation in [12] is taken in a cellular UMTS network shows that co-channel deployment of femtocells can be achieved with only minor impact on the macrocell throughput, and very high theoretical femtocell throughputs for both uplink and downlink can be achieved, it is an energy efficient method. The paper [13] is structured from operator and environment views to analyze area power consumption for a specific area spectral efficiency target, the results show that 100% and 60%
of femto penetration ratios are required to obtain the most energy efficient system from these two aspects separately. In [14], a large-scale femtocell deployment is addressed, where the femto base stations (FBSs) is completely switched off when it is not involved in an active call. The results show that an average of 37.5% power consumption can be reduced, which can be better for even higher femtocell traffic scenario. Also an energy consumption modeling framework in the paper of [15] is presented, where various network environments and femtocell access policies are simulated, and femtocell is proved to be a greener technology that reduces the total energy consumption in a cellular network. In [14], various heterogeneous deployment strategies on area power consumption are discussed. Different cell radiuses and station types are simulated to show huge decrease in the area power consumption with the introduction of small, low power base stations at the edge of the macrocells.
So far, most related work is done to analyze the femtocell penetration rate on the system energy
efficiency performance, but how the base station transmit power and indoor user ratio influence the
4
system energy efficiency outcome is still missing, and how the system performs at different network capacity demands are continued to be discussed in this thesis.
1.3 Problem Definition
It is well known that as the radio wave propagates, there are different kinds of attenuations that may influence user received signal quality. Particularly, in macro-only network, the users inside of the houses experience big signal strength loss due to the wall penetration loss, leading to unsatisfied indoor received signal power, let alone high speed data rate access. Obviously, these connections which are subject to strong attenuations are the most expensive in terms of macrocell resource. But it can be more efficient for those users if they are served by small indoor base stations (i.e. FBSs) which are much closer to them, in this way, the user received signal quality could be better. Meanwhile, FBSs can reuse the frequency spectrum which is allocated to macrocells to improve the system capacity. Also, macrocells could benefit from the offloaded services to small cells and enjoy larger system coverage.
But femtocells also bring some drawbacks: huge interference among in-tier femtocells and co-tier communications with MBSs operating on the same spectrum band may be caused. In the meantime, extra FBS deployments could increase the total network power consumption compared with macro- only networks. So it is the objective of this report to study whether the improvement in system capacity could compensate for the extra power consumption caused by femtocells.
Along the way to the objective, there are several research questions we are going to tackle:
• How will the system capacity and power consumption perform with femtocell deployment?
• In high user traffic load network, how will the system power consumption depend on the femtocell deployment? How about in low traffic load case?
• How will the indoor user ratio affect the system energy consumption performance?
The rest of the report is organized as follows: Chapter 2 gives definitions and problem formulation in
mathematical method; Chapter 3 formulates the system model and assumptions, also we explains the
methodologies employed targeting at the thesis objective and research questions; Chapter 4 shows the
analysis procedure and results in different scenarios; Finally, the conclusion is drawn in chapter 5 and
some future work is also included.
5
Chapter 2
Definitions and Problem Formulation
In this chapter, several concepts used in this thesis will be explained, including energy efficiency, cell coverage, area spectral efficiency, area capacity and area power consumption. They will be given by means of physical meaning as well as mathematical formulas. By the end, the thesis problem is modeled and further formulated through mathematical methods.
2.1 Definitions
2.1.1 Energy Efficiency
In order to provide a certain amount of products or services, energy efficiency sets the goal of efforts to reduce the amount of energy required [16]. In wireless networks, it defines the ratio between the service achieved and the effort expensed correspondingly, i.e. energy, where the higher values represent more efficient network. To be more specific, here we define energy efficiency as the ratio of the total network throughput over the energy consumption within a given period, T, where the unit is bits/Joule [17] or it models how many data rate (bit/s) is achieved when consuming 1 watt power, i.e.
bps/w in the entire system, it is formulated as follows:
EE =
(2.1) Here
and
represent system throughput (bit/s) and power consumption (watt) respectively. Energy efficiency will be explained more in section 2.2.
2.1.2 Cell Coverage
Cell coverage represents the area that a base station can provide its service to, to be more specific, it is defined to be the fraction of cell area where the received power is above a certain level. It is formulated to be [18]:
1
m in(
r x( ) )
A
C r P r P d r d
A φ
= ∫ ⋅ Ρ ≥ (2.2)
6
Where A denotes the cell area and P
is the minimum received power threshold. The coverage is calculated through the integration of users positions whose received power is no less than P
. In this thesis, the cell coverage is provided through MBSs, no contribution from FBSs is considered.
2.1.3 Area Spectral Efficiency
The area spectral efficiency is defined to express the achievable rates in a network per unit bandwidth per unit area, commonly measured in bit per second per Hertz per square kilometer (bps/Hz/km2) [3].
In an ideal Addictive White Gaussian Noise (AWGN) channel, spectral efficiency is expressed as
Sx = F SINRx # = log
'1 + SINRx (2.3)
It is a function of user signal to interference and noise ratio (SINR). Where x is a variable which
represents a specific user unit location. Taking the expectation of each user spectral efficiency, area spectral efficiency is expressed in formula 2.4 [3]:
1 1
[ ( )] ( )
A
S E S X S X x dx
A A
= = ∫ = (2.4)
Where A denotes the reference area.
2.1.4 Area Capacity
In telecommunications, capacity is used to express the amount of information that can be reliably transmitted through a frequency channel. Here, we use area capacity as a performance metric, which is defined to be the average of total user throughput per unit area and it is expressed as [13]:
1
n 2 nTP B log (1 SINR ) A
n= ∑ + (2.5) Where B
is the user bandwidth for the +
,user and SINR
is the corresponding signal to noise and interference ratio. The unit is expressed with Mbps/km2.
2.1.5 Area Power Consumption
The area power consumption is introduced as the average power consumed in a network divided by the corresponding average covered cell area measured in Watts per square kilometer ( -//0
'). For a given cell power consumption
and network area of 1, the area power consumption is defined to be [19]:
P 23 =
456789:; (2.6) For the heterogeneous network we are considering, there are only two types of base stations: MBS and
FBS. Let P
<and P
=denote the power consumption for each MBS and FBS, and N
<>?, N
=>?represent
the number of MBS and FBSs, the accumulated network power consumption is
7
P
@ABCDE= N
<>?∙ P
<+ N
=>?∙ P
=(2.7)
2.2 Problem Formulation
It is our objective to find whether the introduction of macro-femto two-tier cellular deployment is energy efficient or not, compared with the macro-only deployment. However, it is clear that there will be many parameters that may affect the results, which include:
1. The number of installed femtocells which decides how many users can be served by femtocells;
2. The percentage of users that are inside of the house;
3. Power consumption model which indicates how to calculate the power consumption with load consideration or not;
4. Macro base station transmit power determines both the system capacity and system power consumption;
5. Spectrum utilization scheme which defines weather the spectrum should be shared or divided between two tier networks;
6. Environment area is related to different network densification and inter-site distance;
7. System bandwidth determines the system capacity and so on.
To see more clearly, we re-write the energy efficiency formula in formula 2.8:
2 1
* (
* )
*
1
u ser
N
n n
n etw o rk n
n etw o rk M B S M B S F B S F B S
B lo g S IN R E E R
P N P N P
=
+
= =
+
∑
(2.8)
Where Shannon–Hartley theorem is utilized in the numerator part to calculate system throughput, G
is the bandwidth of the +
,user, HIJ represents its corresponding SINR. The denominator part shows the system power consumption. To be more specific, Fig. 2.1 (in the next page) shows how the interference among MBSs and FBSs are generated, where the red phones are users served by MBS, called macro users; while the white phones are served by FBSs, which are called femto users. The red lines represent interference and the black lines denote the wanted signal. The SINR formulas for macro user and femto user are shown separately in formula 2.9 and 2.10.
_
_ _
,
n M B S M a c r o U s e r
i M B S j F B S
i M B S i n j F B S
S IN R P
P P N o is e
−
∈ ≠ ∈
= ∑ + ∑ + (2.9)
_
_ _
,
n F B S F e m to U s e r
i F B S j M B S
i F B S i n j M B S
S IN R P
P P N o is e
−
∈ ≠ ∈
= ∑ + ∑ + (2.10)
In formula 2.9, for macro users,
_LMNrepresents the received power from the +
,MBS which is a
function of MBS transmit power (
O_LMN); the denominator part signifies interference from
surrounding FBSs which is a function of femtocell penetration rate (Rp, which is used to describe the
percentage of houses that are installed with a FBS) and FBS Transmit Power (
O_PMN), noise is also
included at the end. Another factor that may affect macro user SINR is inter-site distance (ISD, the
8
minimum distance between any two base station sites) parameter as they are randomly distributed in the network area when we take an average consideration of the system performance. For the femto users (formula 2.10),
O_LMN, Rp and ISD also have an effect on their SINR performance.
Therefore, combined with formula 2.8, all these parameters affecting user SINRs shape system energy efficiency performance together. Besides, the system spectrum bandwidth and indoor User Ratio (Ir, which defines the percentage of users that are generated inside of the houses) also contribute to the final energy efficiency performance. Formula 2.11 combines all these together .
QQ ∝ P S
T_UVW, S
T_ZVW, M, [
\, ]N^, ]
#
PS
T_UVW, S
T_ZVW, [
\(2.11)
Figure 2.1 Interference generation among MBSs and FBSs
So here, our goal is to investigate how the conclusion is regarding the energy “effectiveness” of macro-femto deployment with these parameters.
Interference Wanted Signal
Indoor user
without FBS
9
Chapter 3
System Model and Methodology
In this thesis, we are going to study how the introduction of user-deployed femtocells impacts the system power consumption and capacity in a cellular network. We are going to explain the models we used in our simulations from the aspects of network deployment, spectrum utilization scheme, path loss and power consumption models in this chapter. After that, a brief description of the scenarios we used to find the answer will be given, then, the simulation procedure and parameters are discussed.
3.1 System Model
3.1.1 Network Deployment Model
For this study, A LTE cellular network with OFDMA technique is deployed. The requency reuse
factor is 1. The downlink (the transmission path from a cell site to a cell phone) of a hexagonal grid of
19 macrocells with 3 sectors per cell is considered; the MBSs are put in the cell center as shown in
Fig. 3.1. The central reference cell is surrounded by two tiers of interferers. The MBSs are installed
and transmitting radio waves with 3 directional antennas pointing to each sector with a separation of
120 degree (shown in the reference cell). Besides the MBSs, one-floor houses are randomly distributed
over the network area. Among all the deployed houses, some of them are installed with a FBS. For a
given femtocell penetration rate (Rp which is used to describe the percentage of houses that are
installed with a FBS), FBSs is installed by the users who are willing to pay for the extra cost related to
FBSs power consumption and installation fee in their houses with random positions. Omni-directional
antennas are used for each FBS. The users are divided into two parts: users inside of the house (indoor
users) and users outside of the house (outdoor users). Also for a given indoor user ratio (Ir, which
means the percentage of users that are generated inside of the house), we use randomly distributed user
number and positions in each house, so there might be some houses with 1 user, while others have
none or even 3 users as shown in Fig. 3.1. Each house is assumed to have up to 8 users, and FBSs are
assumed to be turned off if no user exists in their house. All the outdoor users are randomly distributed
over the cell area. Besides, based on the serving base station, indoor users are further divided to be
femto user or macro user as mentioned before, the same is for outdoor users, they will connect to the
base station that gives them the highest received power.
10
3.1.2 Spectrum Utilization Scheme
There are some paper work using resource partitioning scheme between macrocell and femtocell, where the whole frequency bandwidth is splitted between macrocell and femtocell [19]. But there are also some papers working with co-channel spectrum scheme where the whole bandwidth is allocated to every MBS and reused by each small base station. Considering the system capacity might be improved significantly because of the special frequency reuse, as well as the scarcity of frequency bandwidth resource, the latter scheme is chosen in our work.
In this thesis, different users may be allocated with different bandwidths, which depend on the number of users who are served by the same base station with him. We used a simple Round Robin scheduling algorithm which allocates an equal part of frequency resource to each user without taking user channel conditions into account. As a result, the average bandwidth per user allocated is the whole bandwidth divided by the number of served users who have the common serving base station. As a result, the average user bandwidth decreases if more users are served by the same base station.
3.1.3 Propagation Model
As the radio wave propagates, path loss may be due to many effects, including free-space loss, refraction, diffraction, reflection, aperture-medium coupling loss, and absorption [20]. In our report, the downlink channel between base station and users is modeled to comprise distance dependent path loss and a slowly varying lognormal shadow fading, neglecting fast fading. We choose the path-loss model for macro users as is used in [21]:
10
10
15.3 37.6 log 15.3 37.6 log
macro
ow
R OutdoorUser
PL R L IndoorUser
+
=
+ +
(3.1)
Where R is the distance between MBS and users, and L
owrepresents the outdoor wall
Figure 3.1 Network deployment model
11
penetration loss. While for a femto user, the path-loss model is also according to [17]:
10 2
10 10 2
38.46 20 log 0.7
max(2.7 42.8 log ,38.46 20 log ) 0.7
D iw
femto
ow D iw
R d qL IndoorUser
PL R R L d qL OutdoorUser
+ + +
=
+ + + + +
(3.2)
Where 0.7 d
2Dtakes account of penetration loss due to walls inside of the house, and qL
iwrepresents the loss caused by the penetrated floors, it will be zero because one-floor houses are deployed in our simulation. For the femto users who are in different house with their served FBS, the path loss should be added by another wall penetration loss based on the outdoor user in formula 3.2.
Different shadow fading standard lognormal variances for open wide area and indoor environments are used: 8 dB for open environment and 4 dB for indoor area [21].
3.1.4 Power Consumption Model
The Power consumption model we used is based on EARTH project: Energy Aware Radio and Network Technology, where a linear approximation of the power consumption model is justified:
3_
= `J
a[b∙
c+ ∆
e# 0 <
≤
i3OJ
a[b∙
j_e,
= 0 (3.3) Where
is the antenna radiated power, P
maxdenotes the maximum antenna transmit power. P
orepresents power consumption with no traffic load (it is actually estimated using the power consumption calculated at a reasonably low output power, assumed to be 1% of P
max), which is independent of transmit power due to signal processing, battery backup and site cooling. ∆
prepresents the relationship between load dependent power consumption and the radiated power caused by PA amplifier and feeder losses as well as site cooling. N
TRXis the number of RF transmitter and
sleep
P is the power consumption for the site that is in sleep mode, but no sleep mode is considered in this thesis. The parameter for MBS and FBSs are shown in Table. 3.1 [22].
Based on this consumption model, the maximum consumed power for MBS is 672 watt; while for a FBS, the maximal consumed power is 5.6 watt at full load case and 4.8 watt is needed for the idle state.
Nearly 120 FBSs together consume the same amount of power with a single MBS. Also only 0.8 watt is saved if one FBS is turned off for idle state from full load state, which is slight for the whole network power consumption, as a result, the network performance may not differ too much due to this FBS scheme.
Base Station
Type N
TRXP
max(W) P
o(W) ∆
pP
sleepMacro (MBS) 3 20 130 4.7 75.0
Femto (FBS) 1 0.1 4.8 8.0 2.9
Table 3.1 Power consumption model for MBS and FBS
12
3.2 Methodology
The previous chapter gives us a basic platform and model to study the energy efficiency performance improvement through the introduction of femtocells. Considering all the factors we are going to study as mentioned by the end of chapter 2, we will first look at the energy efficiency performance on a capacity over-provisioned cellular network, where femtocell penetration rate effect will be studied, then, in the following scenarios, inter-site distance, MBS transmit power and indoor user ratio will be analyzed. To be more specific, we are going to explore the following scenarios:
Scenario 1: The femtocell penetration rate effect on the system energy efficiency is studied in this scenario. The energy efficiency is indicated by bit/s/w. By the system throughput and power consumption, the maximum achieved energy efficiency will be compared among different Rp deployment cases.
Scenario 2: The goal of this scenario is to compare system power consumption at the same capacity fulfillment premise. Different Rp deployment schemes are compared by means of area power consumption [ -//0
']. Inter-Site Distance is introduced as a control parameter. Two relationships are studied: area capacity versus ISD and area power consumption versus ISD.
Scenario 3: The effects of different MBS transmit power on area power consumption are studied. The MBS transmit power is adjusted in different ISD cases to guarantee a minimum received power at the cell border. Minimum area power consumption is obtained in different network deployment schemes;
the one who consumes the least power is the most energy efficient.
Scenario 4: This scenario studies how the percentage of users inside of the house (Ir) shapes the system energy efficiency performance. The same method is used as scenario 2 to tell which Ir scheme is the most energy efficient.
The simulation scenarios can be seen clearly from Fig. 3.2.
Figure 3.2 Flow Chart of simulation scenarios
13
3.2.1 Simulation Procedure
Fig. 3.3 illustrates the flow chart of our simulation. In the simulation, “Monte Carlo method” is adopted to obtain a reasonable result. Some assumptions have been mentioned in the previous chapter.
The simulation starts by creating a network deployment map including the locations of MBSs and houses. Then it randomly generates the positions of indoor users and outdoor users before creating a shadow fading map. Later, within every iteration, the following process is carried out: a certain number of houses with femtocells are chosen randomly according to different femto penetration rates;
users are moved around; Calculate the distance matrix between the users with their served and interfered base stations; Create shadow fading matrix for all the connected paths and add it to the path loss matrix obtained through propagation models; Calculate SINR for each users and then throughput, power consumption and energy efficiency. After certain iterations (generally we use 300 times), all the comparison metrics are averaged and plotted.
3.2.2 Simulation Parameter
The selection of the simulation parameters are mostly based on [23]. The total bandwidth is 10 MHz.
In this report most results demonstrated are using the parameters in Table 3.2, unless otherwise noted.
Parameter Value
Carrier frequency 2.0 GHz
Total Bandwidth 10.0 MHz
Inter-site distance 500-2500m
MBS transmit power (directional) 46 dBm (varied for scenario 3) FBS transmit power (Omni) 20 dBm
MBS/FBS antenna gain 14/5 dBi
Total_House_Number (per cell) 75 Total_User_Number (per cell) 113
indoor user Ratio 50%-90%
femto-cell penetration rate (Rp) 0-1 (varied) MBS Shadow fading (standard deviation) 8 dB
Figure 3.3 Simulation Flow Chart
14
FBS Shadow fading (standard deviation) 4 dB Shadow fading correlation distance 20 m Exterior wall penetration loss 15 dB Minimum received power per user -70 dBm
Thermal noise -174 dBm/Hz
House size 10 x 10 m
2Table 3.2 Simulation system parameters
In each cell, there is a base station site installed in the center with 3 directional antennas pointing to 3 directions with 120 degree separated. The macrocell antenna pattern is shown below:
kl mG = k
i3O− min r12 t l
uv , k
jw , −x ≤ l ≤ x -yzℎ u = 70x
180 ~+ -ℎ ~y+ ~zz+ y 3mG m-+ 0 ~/
k
j= 20mG ym ~y+ y+ mG
k
i3O= 14mG 0~y00 ~y+ y+ mG
15
Chapter 4
Simulation Results
In this chapter, simulation results of the aforementioned four scenarios are discussed in separate sections, namely: i) The impact of femtocell penetration rate on energy efficiency in capacity over- provisioned system; ii) System power consumption comparison for different user requirements; iii) Adjusted MBS transmit power effect on area power consumption; iv) Indoor user ratio impact on system power consumption.
4.1 Energy Efficiency of Macro-Femto Deployment in Over-Provisioned Network
In this scenario, Rp is changing in each run of simulation to illustrate the system performance difference and a direct comparison between macro-only case and macro-femto deployment is conducted from the perspective of system throughput and power consumption. Finally the system energy efficiency is calculated to give the conclusion. In this section, we will go through the system performance by means of showing SINR map, user spectral efficiency cumulative distribution functions and then, give the system throughput and energy efficiency evaluation. Some important parameters in this scenario are shown in Table. 4.1.
Parameter Value
Inter-site distance 500 m MBS transmit power (directional) 46 dBm Total_House_Number (per cell) 75 Total_User_Number (per cell) 113
indoor user Ratio 70%
femto-cell penetration rate (Rp) 0-1 (varied)
Table 4.1 Important simulation parameters for scenario 14.1.1 SINR Map
Fig. 4.1 shows the SINR map for macro-Only network and Fig. 4.2 for macro-femto case. In Fig. 4.1,
the circles inside of the hexagon represent indoor or outdoor users and they are all served by MBSs.
16
While Fig. 4.2 shows a full femtocell deployment case, where femtocell is installed in every house.
The black squares denote the deployed houses and the circles inside of the squares are indoor users, other circles belong to outdoor users.
Figure 4.1 SINR map of Macro-Only Network
Figure 4.2 SINR map of Macro-Femto network
Theoretically, in macro-only scenario, the user who is far away from the base station will receive less signal strength due to signal propagation, which is why the user color is becoming from red until blue as they move to the cell boarder. Besides, the effect of MBS directional transmit antennas is also shown as the sectored shape. While, in Fig. 4.2, all indoor users served by FBSs will receive higher
-300 -200 -100 0 100 200 300
-250 -200 -150 -100 -50 0 50 100 150 200 250
1
SINR MAP Macro-Only
-10 -5 0 5 10 15
-300 -200 -100 0 100 200 300
-250 -200 -150 -100 -50 0 50 100 150 200 250
1
SINR MAP Macro-Femto
-10 0 10 20 30 40
Outdoor User
Houses
Indoor User
17
signal power compared with previous case when they are served by MBSs. In this case, MBSs act as the interference source. So for femto users, those who are near to the central MBS are expected to achieve worse SINR than those at the cell edge. For the outdoor users, they might experience general performance degradation because they are interfered by nearby FBSs. All these are coherent with the color changes in the map. Comparing these two figures, the introduction of femtocells improves the general user SINR, which can be seen from the SINR scalar on the right side in each figure.
Table. 4.2 and 4.3 show more analysis in detailed data (with shadow fading considered) for macro- only and macro-femto cases individually. For example, we take MUE3 at 147.5 meter from MBS in macro-only case for example. The wanted received signal power is -63.7 dBm; among all the interference and noise, -74.3 dBm is the highest dominate part, with these, a 6.2 dB SINR is obtained;
On the other hand, for a femto user FUE3 149.4 meter away from MBS, MBS signal of -65.9 dBm acts as the main dominate interferer, but the wanted signal quality is much better obtained from the indoor FBS turning to be -47.4 dBm. In this way, the SINR is improved to be around 18 dB, it is almost three times higher compared to the outdoor user who is the same far away from MBS. If we take a further look at these tables, the users who are farther to MBS achieve worse signal SINR in macro-only case. But in macro-femto scheme, the distance between user and FBS and the house positions decide user experience together.
MUE1 MUE2 MUE3 MUE4 MUE5
Distance_to_MBS (m) 40 103.3 147.5 199.1 279.3
Useful Received Power (dBm) -48.9 -57.8 -63.7 -73.3 -75.2
Interference_D1 (dBm) -68.8 -74.9 -74.3 -75.2 -75.7
Interference_D2 (dBm) -68.8 -77.8 -75.5 -75.5 -76.3
Interference_D3 (dBm) -84 -77.8 -82.1 -76.9 -76.5
Interference_D4 (dBm) -87.7 -79.4 -83.5 -77 -77.6
Interference_D5 (dBm) -88.1 -82.8 -83.5 -77.9 -79.4
Noise (dBm) -118.9 -119.3 -119.3 -118.9 -118.9
SINR (dB) 16.6 12 6.2 -4.9 -11.8
Table 4.2 Macro users received signal analysis in macro-Only network
FUE1 FUE2 FUE3 FUE4 FUE5
Distance_to_MBS (m) 45.5 89.8 149.4 199.5 256.1
Distance_to_FBS (m) 4.2 4 11.5 8.2 8.3
Useful Received Power (dBm) -25.4 -28.5 -47.4 -42.3 -37.1
Interference_D1 (dBm) -40.6 -54.6 -65.9 -69.6 -67.8
Interference_D2 (dBm) -54.1 -68.9 -77.2 -79.9 -75.3
Interference_D3 (dBm) -57.3 -71.5 -82.1 -85 -80.6
Interference_D4 (dBm) -77.4 -82.7 -83.6 -89.1 -83.8
Interference_D5 (dBm) -79.7 -84.5 -84 -89.1 -85.7
Noise (dBm) -104 -110 -110 -104 -111.8
SINR (dB) 15 25.8 17.9 26.6 29.4
Table 4.3 Femto users received signal analysis in macro-femto network
4.1.2 Area Spectral Efficiency and Area Capacity
With the previous discussions about user SINR, Fig. 4.3 gives us more intuitionistic understanding of
user spectrum efficiency based on changing Rp. The blue curve represents macro-only network, as the
curve moves to the right, the femtocell penetration rate is increasing. It is quite obvious that femtocells
18
improve users’ spectral efficiency significantly, and the more femtocells are deployed, the more gains the system achieves. As we can see from this figure, 50% of users in macro-only network (Rp=0) show a spectral efficiency below 1.5 bps/Hz, while this value improves to be around 6 bps/Hz when Rp equals 1 where all the indoor users are served by FBSs.
Figure 4.3 User spectral efficiency CDF plot for different Rp deployment schemes
Fig. 4.4 shows user average throughput cdf plot for different Rp cases. According to the co-channel spectrum scheme, the system bandwidth is reused by each FBS, the user capacity will be linear increased by the bandwidth improvement. From Fig. 4.4 we can see, for outdoor users, their throughput almost stays the same, there are two reasons: one is that they are enjoying wider bandwidth because less users are sharing with them, this pushes capacity growth; but on the other hand, the
0 2 4 6 8 10 12 14 16
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
User Spectral Efficiency [bps/Hz]
CDF
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Macro-Femto Network Rp=0
0 20 40 60 80 100 120 140
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
User Throughput [Mbps]
CDF
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Rp=0
Macro-Femto Case
Outdoor UEs
Figure 4.4 User throughput cdf plot for different Rp deployment schemes
19
interference from surrounding FBSs leads to decreased spectral efficiency. While for the FBS served indoor users, they are enjoying a surge in their throughput, these can be seen clearly from the starting point of the branches in the plot, the enhancement of both SINR and bandwidth contribute to the improved capacity.
4.1.3 System Integrated Throughput and Energy Efficiency
With the previous analysis, we arrive at Fig. 4.5 and 4.6 which give us integrated system area throughput and energy efficiency performance. As we know, this network is allowed to be over- provisioned, which means that the users may be provided much more capacity than they even need.
From the figures, both of them have a significant improvement with increasing installation of FBSs.
For integrated throughput case, it improves from around 241 Mbps in macro-only network until
Figure 4.5 System integrated throughput versus femtocell penetration rate
Figure 4.6 Energy efficiency versus femtocell penetration rate
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Femtocell Penetration Rate Area Integrated Throughput [Mbps/km2] 16657
241
69 times up
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.5 1 1.5 2 2.5 3
Femtocell Penetration Rate
Energy Efficiency [Mbps/W]
48.6 times up
0.0549
2.667
20
16657 Mbps when all the houses are installed with a FBS, it is 69 times improved! The same trend shows for energy efficiency performance which increases up to be around 49 times. But it should be noted that the gain is mostly coming from spectrum reuse and the traffic demand has been ignored.
In the study of scenario 1, we found that the introduction of femtocells enhances the system capacity much more than the extra FBSs power consumption expense. However, some users are provided as high data rate as possible, which may be needless for most of them. This cannot quantify the energy savings in wireless systems. It can be much better for the comparison between macro-only network and macro-femto network if they could hold the system performance, for instance, area capacity or area spectral efficiency. Because energy efficiency improved does not guarantee energy saving, and to make a fair comparison between different femtocell deployment and reflect the real saving, user capacity should be fixed and power consumption should be compared.
We find different MBS transmit power and ISD are two main factors that may shape the area power consumption, which will be studied in the next two scenarios, and there are two relationships we want to analyze. The first one is how the ISD influences area capacity (or spectral efficiency), the other relationship is between ISD and area power consumption. Combining these two interactions, we will get the tendency of area power consumption corresponding to increasing user demand in scenario 2.
4.2 Impact of Network Capacity Requirement on Total Power Consumption
In this scenario, we will fix the network capacity and find the required ISD to study the required power consumption. For a fixed considered area, the user and house number are fixed and they are randomly distributed in the network area as shown in Fig. 4.7 by black crosses and squares, respectively. All the MBS sites will transmit a fixed power of 46 dBm. ISD is ranged from 500 meter to 2500 meter and the femto penetration rate is also varying. Fig. 4.7 gives a simplified network model, where red stars denote FBSs, and 75 houses are generated with 30% of users are outside of the house.
-300 -200 -100 0 100 200 300
-250 -200 -150 -100 -50 0 50 100 150 200 250
1 Network Deployment
70% Indoor- Users
75 Houses
FBS (0~100%) 30%
Outdoor-
Varying ISD:
500 ~2500 m
46 dBm
Figure 4.7 Simplified network development model for scenario 2
21
Table 4.4 Simulation parameters for scenario 2
The simulation test-bed is realized with two-layer loops. In the first layer, ISD is ranging from 500 meter to 2500 meter. Corresponding to each specific ISD, the second layer loop of varying Rp is set.
Inside of this layer, the path loss calculation, user SINR, spectral efficiency, capacity and power consumption performance are studied. The results are shown by the form of figures as follows.
4.2.1 User Spectral Efficiency and Capacity Evaluation
To tell the influence of different ISD on the user spectral efficiency performance, we take ISD=2100 meter to make a comparison with ISD=500 meter aforementioned in scenario 1. The result shows in Fig. 4.8 and 4.9. Similar trend can be seen that user spectral efficiency increase with increasing femto penetration rate. To further, we can see 2100 meter ISD case has better spectral efficiency performance when femto-cells are introduced except the macro-only network case (the blue curve), which behaves the same with 500 meter ISD case. This implies that user SINR is independent of cell radius in a homogeneous macro network. The key reason behind this is, in the interference limited system, the path-loss between MBS to the users is increased when the user moves away from the central MBS, so both the wanted received signal from the benefit MBS and the interference from neighbor MBSs become weaker to the same degree, as a result, there is almost no difference experienced.
Figure 4.8 User spectral efficiency CDF in ISD of 2100 meter case
0 5 10 15 20 25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
User Spectral Efficiency [bps/Hz]
CDF
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Macro-Femto Network Rp=0
Parameter Value
Inter-site distance (ISD) 500-2500 m
MBS transmit power 46 dBm
Total_House_Number 75
Total_User_Number 113
indoor user Ratio 70%
femto-cell penetration rate (Rp) 0-1 (varied)
22
Figure 4.9 User spectral efficiency CDF for ISD of 500 meter case
While for the macro-femto case, all the MBSs work as interfers to femto users, the increased ISD causes that the interferenced signal is decreased; while the wanted signal strength from FBSs remains the same quality, that is why FBSs provide better spectral efficiency in larger ISD case. Consequently, users in larger cell area enjoy better throughput performance.
4.2.2 Area Spectral Efficiency and Power Consumption versus Inter-Site Distance
In this part, femtocell penetration rate is used as a control parameter, the relationships between area spectral efficiency and ISD, area power consumption and ISD are studied individually.
Figure 4.10 Area spectral efficiency versus inter-site distance
0 2 4 6 8 10 12 14 16
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
User Spectral Efficiency [bps/Hz]
CDF
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Macro-Femto Network Rp=0
500 1000 1500 2000 2500
0 5 10 15 20 25 30
Inter Site Distance [m]
Area Spectral Efficiency [bps/Hz/km2] Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Target
23
Figure 4.11 Area power consumption versus inter-site distance
From the Fig. 4.10 and 4.11 we can see, increasing ISD leads to decreasing area spectral efficiency, and area power consumption, especially for higher femto penetration case which always has larger descending gap as the network becomes sparse (ISD increases). Also, higher Rp case always has better spectral efficiency performance over lower case. For example, only 9 bps/Hz/km2 of spectral efficiency is obtained for macro-only case with an ISD of 500 meter, while this value is improved to be 27.5 bps/Hz/km2 when Rp is 1. Besides, for each fixed Rp case, the total system consumed power is fixed, so the area power consumption decreases with increasing of ISD. Because femtocells are low- power base stations, the gaps between different curves are not so significant.
4.2.3 Area Power Consumption for Different User Data Rate Targets
The figures in previous parts could give us more interesting information if a specific target of area spectral efficiency or capacity is set to meet, which is a more intuitionistic and efficient method to evaluate wireless network from a perspective of providing users the amount of data rate that they really need. For instance, some users may only need voice service, then, a data rate of 10 Mbps provided for them is a waste of radio resource. Scenario 1 is a less-efficient resource allocation scheme because user capacity could be over-provisioned. In this part, we try to analyze the system power consumption for the cases of different system spectral efficiency and capacity targets with Fig. 4.12 and 4.13.
We will take area spectral efficiency for example, from Fig. 4.10, we assume that 5 bps/Hz/km2 is demanded by each user, so we can find the corresponding ISD for different Rp cases, for instance, when 100% of FBS is installed, the network can be deployed with an ISD of 1500 meter; while the network should be with ISD of 1015 meter to fix this constraint in 40% femtocell developed network.
This indicates denser network is required when femto penetration rate is low. Then, with the corresponding ISDs for different Rp cases, the relationship between ISD and area power consumption in Fig. 4.11 tells us 800 watt need to be consumed per square kilometer in a fully developed macro- femto network, while in macro-only network, 2550 watt per square kilometer is required, which is 3 times more power consumption. With this method, for different area spectral efficiency targets, a
500 1000 1500 2000 2500
0 1000 2000 3000 4000 5000 6000 7000
Inter Site Distance [m]
Area Power Consumption [W/km2]
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
1120W 1250W
1020W 850W
750W 2300W
24
maximum ISD is matched as shown in Fig. 4.12. The same method is used for area capacity and the results are shown below:
Figure 4.12 Area power consumption for different area spectral efficiency targets
Very clear conclusion should be driven that to meet a target area spectral efficiency, macro-only case always consumes more power, additionally, the highest spectral efficiency that can be provided is 9 bps/Hz/km2 in our simulation (corresponding to ISD of 500 meter), if higher demand is required, the network should be even denser. On the other hand, the more FBSs are installed, the lower power is consumed, and higher Rp case can provide better spectral efficiency fulfillments. The Fig. 4.13 proves even better improvement when area capacity is used as targets, where macro-only network cannot meet users’ demand to achieve area capacity of over 5 Mbps/km2, femtocell is a good way to improve the system capacity performance.
0 5 10 15 20 25 30
0 1000 2000 3000 4000 5000 6000 7000
Target Area Spectral Efficiency [bps/Hz/km2]
Area Power Consumption [W/km2]
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
0 20 40 60 80 100 120 140 160
0 1000 2000 3000 4000 5000 6000 7000
Target Area Capacity [Mbps/km2]
Area Power Consumption [W/km2]
Macro-Only Rp=20%
Rp=40%
Rp=60%
Rp=80%
Rp=100%
Figure 4.13 Area power consumption for different area capacity targets
