• 沒有找到結果。

前瞻長程演進異質性網路的頻譜與能源效率之分析

N/A
N/A
Protected

Academic year: 2021

Share "前瞻長程演進異質性網路的頻譜與能源效率之分析"

Copied!
73
0
0

加載中.... (立即查看全文)

全文

(1)

國 立 交 通 大 學

電信工程研究所

碩 士 論 文

前瞻長程演進異質性網路的

頻譜與能源效率之分析

Investigation of Spectral and Energy Efficiency

in LTE-A Heterogeneous Networks

研究生:謝宗展

指導教授:王蒞君

(2)

前瞻長程演進異質性網路的

頻譜與能源效率之分析

Investigation of Spectral and Energy Efficiency

in LTE-A Heterogeneous Networks

研 究 生:謝宗展

Student:Tsung-Chan Hsieh

指導教授:王蒞君 Advisor:Li-Chun Wang

國 立 交 通 大 學

電信工程研究所

碩 士 論 文

A Thesis

Submitted to Institute of Communications Engineering College of Electrical and Computer Engineering

National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Master of Science

In

Communication Engineering June 2012

Hsinchu, Taiwan, Republic of China

(3)

前瞻長程演進異質性網路的

頻譜與能源效率之分析

學生:謝宗展

指導教授:王蒞君 教授

國立交通大學

電機學院電信工程研究所

摘要

在此篇論文中,我們探討應用階層式基地台合作技術之第三代合

作夥伴前瞻長程演進異質性網路系統的頻譜與能源效率。我們發現同

時考慮系統架構與合作傳輸方法的設計能夠達到綠能傳輸之目的。因

為基地台合作之技術可以有效降低共同通道干擾與改善接收訊號品

質,且系統架構對於系統之表現有相當的影響。特別是考慮了同時結

合細胞間與細胞內基地台合作之方法。這篇論文呈現了利用上述之聯

合設計方法於應用階層式基地台合作技術之異質性網路系統表現的

改善。模擬結果顯示,採用我們提出的聯合設計方法之異質性網路系

統比起傳統單用戶多輸入多輸出系統是一個有效提高頻譜與能源效

率的方法。

(4)

Investigation of Spectral and Energy Efficiency

in LTE-A Heterogeneous Networks

A THESIS Presented to

The Academic Faculty By

Tsung-Chan Hsieh

In Partial Fulfillment

of the Requirements for the Degree of Master in Communication Engineering

Institute of Communications Engineering College of Electrical and Computer Engineering

National Chiao-Tung University

June, 2012

(5)

Abstract

In this thesis, we investigate both spectral and energy efficiency of hierarchical base station cooperation techniques in heterogeneous networks (HetNet) of the 3rd Gener-ation Partnership Project (3GPP) Long Term Evolution-Advanced (LTE-A) system. We find that joint consideration of system architectures and cooperation schemes will achieve energy-efficient transmission. Because the co-channel interference can be mitigated significantly and the signal quality is able to be improved through base station coordination techniques, and system architectures have great impacts on the system performance. Especially, coordinated multi-point (CoMP) techniques of dif-ferent levels, e.g., intra and inter-site CoMP, are considered jointly. We address the performance improvements of HetNet system with hierarchical base station coordi-nation techniques by the joint design of both system architecture and coordinated transmission scheme. The proposed joint design methodology is a promising solution for improving spectral and energy efficiency compared to the conventional single user multi-input-multi-output (SU-MIMO) system.

(6)

Acknowledgments

Foremost, I would like to appreciate Professor Li-Chun Wang who has guided me for more than three years. As the lyric goes, I once was lost, but now I am found. I was so unfamiliar with what the true research is, but I have increased some more understanding now.

Secondly, I want to thank all my laboratory members in Mobile Communications and Cloud Computing Laboratory at the Institute of Communications Engineering in National Chiao-Tung University. It is they that provide me so much assistance to finish my research. Especially, Tsung-Ting Chiang spent a lot of his free time helping me solve many problems.

Finally, I owe an enormously debt of gratitude to my parents for their great support these years.

(7)

III

Contents

Abstract I

Acknowledgements II

List of Tables VI

List of Figures VII

1 Introduction 1

1.1 Motivation . . . 2

1.2 Issues . . . 2

1.3 Thesis Outline . . . 4

2 Background 5 2.1 Why Green Communication? . . . 5

2.2 Coordinated Multi-point Techniques . . . 6

2.3 Literature Survey . . . 8

3 System Models 10 3.1 Cell Layout . . . 10

3.1.1 Homogeneous Network SU-MIMO System . . . 10

3.1.2 Heterogeneous Network CoMP System . . . 12

3.2 Channel Model . . . 14

(8)

3.2.2 Sptial Channel Model . . . 15

3.3 Power Consumption Model Description . . . 17

3.3.1 Network Power Consumption . . . 17

3.3.2 Base Station Power Consumption . . . 18

4 Heterogeneous CoMP Networks Simulator 22 4.1 Codebook-based Precoding . . . 22

4.2 Transmission Equations . . . 25

4.2.1 Single UE Case . . . 26

4.2.2 Multiple UE Case . . . 28

4.3 CoMP Schemes . . . 29

4.4 Proportional Fair Scheduling . . . 30

4.5 Exponential Effective SINR Mapping (EESM) . . . 32

4.6 Hybrid Automatic Repeat Request (HARQ) . . . 32

4.7 Energy Efficiency . . . 33

5 Tradeoff Design of Spectral and Energy Efficiency in HetNet Sys-tems 34 5.1 System Architectures . . . 34

5.2 CoMP Transmission Techniques . . . 35

5.3 Design Procedures . . . 38 6 Numerical Results 40 6.1 Simulation Assumptions . . . 40 6.2 Simulation Baseline . . . 40 6.3 Intra-site CoMP . . . 42 6.3.1 Effect of RRH Deployment . . . 42

6.3.2 Effect of Cell Architecture . . . 45

6.3.3 Tradeoff between Spectral and Energy Efficiency . . . 45

6.4 Inter plus Intra-site CoMP . . . 49 IV

(9)

6.5 Reference Signal Received Power-Based RRH Selection . . . 53

7 Conclusions 55

7.1 Thesis Summary . . . 55 7.2 Suggestions for Future Research . . . 56

Bibliography 57

Vita 60

(10)

VI

List of Tables

2.1 Comparison of Our Work and Related Works . . . . 9

3.1 Deployment regulations . . . . 12

3.2 Components of the site power . . . . 17

3.3 Power consumption for various BSs . . . . 21

4.1 Codebook for Two Antenna Ports . . . . 23

4.2 Codebook for Four Antenna Ports . . . . 24

6.1 Simulation Parameters . . . . 41

6.2 2x2 SU-MIMO Simulation Results Comparison . . . . 42

(11)

VII

List of Figures

1.1 The relation between EE and SE. . . 4

2.1 Various CoMP scenarios. . . 7

3.1 Cell architecture of SU-MIMO systems. . . 11

3.2 Cell architecture of CoMP systems. . . 13

3.3 Parameters of 3GPP SCM. . . 15

3.4 BBU+RRU based system. . . 19

3.5 Power model of a BS. . . 20

4.1 Two cell scenario of CS/CB. . . 31

5.1 Sector antenna architectures. . . 36

5.2 Two kinds of CoMP schemes. . . 37

5.3 Joint design procedure for energy-efficient transmission. . . 39

6.1 Comparison of spectral efficiency for different RRH locations. . . 44

6.2 Comparison of energy efficiency for different RRH locations. . . 44

6.3 Comparison of spectral efficiency for different system architectures. . 46

6.4 Comparison of energy efficiency for different system architectures. . . 46

6.5 Tradeoff between spectral efficiency and energy efficiency of the intra-site CoMP scheme. . . 47

(12)

6.7 Comparison of spectral efficiency for different transmission schemes in the single UE case. . . 51 6.8 Comparison of energy efficiency for different transmission schemes in

the single UE case. . . 51 6.9 Comparison of spectral efficiency for different transmission schemes in

the multiple UE case. . . 52 6.10 Comparison of energy efficiency for different transmission schemes in

the multiple UE case. . . 52 6.11 Comparison of spectral efficiency for different RRH density with selection. 54 6.12 Comparison of energy efficiency for different RRH density with selection. 54

(13)
(14)

1

CHAPTER 1

Introduction

Due to the explosive growth in information and communication traffic as well as demands of better quality of service (QoS) from subscribers, it is estimated that the information and communication technology (ICT) energy consumption is rising at 15-20 percentages per year, in other words, doubling every five years, and such striking increase does not seem to slow down soon. It is reckoned that the ICT industry is responsible for 3 percent of the worldwide annual electrical energy consumption, causing 2-4 percent of world’s carbon dioxide emissions [1] [2]. With the increasing awareness of gradual depletion of non-renewable resources and harmful environmental impacts caused by carbon dioxide, it is the social responsibility that cellular network operators should be devoted to develop energy-efficient telecommunication systems. Aside from environmental aspects, the energy cost accounts a great portion of network operators’ overall expenditure, and the electric bill is more than $10 billion dollars per year [3]. At present, almost 80 percentages of electrical power for system operation is attributed to radio access network (RAN) [4]. Therefore, efficient transmission schemes and network architectures benefit not only ecological but also economical aspects. Such improvements can be fulfilled in two ways: optimizing base stations (BSs) via more efficient and traffic load adaptive modules, and innovating radio access point deployment strategies as well as transmission technologies to lower the energy consumption and achieve the required system performance.

(15)

1.1

Motivation

The traditional network system design mainly focused on spectral efficiency (SE). Energy efficiency (EE) only received little attention, and seldom deemed as a vital performance indicator. As energy-saving issue becomes more crucial, green communi-cation gets more and more important. Therefore, not only SE but also EE should be considered in the system design. There are several techniques to increasing the system throughput. First of all, deploying small cells with macro-cells to build a heteroge-neous network (HetNet), which is able to enhance received signal power to guarantee acceptable signal to interference plus noise ratio (SINR). Secondly, higher data rate can be rendered through coordinated multi-point (CoMP) techniques which allow BSs to process signals jointly for mitigating interference [5] [6]. It is true that CoMP and HetNet are promising techniques to improve the system capacity [7]. However, it comes with a price: additional energy consumption. Because increasing BSs needs more power consumption accordingly, and additional backhaul connections among cooperating nodes as well as signal processing power to perform CoMP schemes also need more energy. As mentioned above, both spectral and energy efficiency are vi-tal performance indicators. Hence, the objective of this thesis is to assess spectral and energy efficiency of CoMP transmission in HetNet systems, and to find the ap-propriate transmission scheme and deployment strategy to achieve energy-efficient transmission.

1.2

Issues

To address the characteristic of the relation between those two performance met-rics [8], we consider a point-to-point transmission in additive white Gaussian noise (AWGN) channel. In accordance with Shannon’s capacity equation, given the system

(16)

bandwidth W and the transmit power P , the maximal reliable transmission rate is

R = W log2(1 + P

W N0

), (1.1)

where N0 denotes the noise power spectral density. According to SE and EE

defini-tions, SE and EE are

ηSE = R W = log2(1 + P W N0 ), (1.2) and ηEE = R P = W P log2(1 + P W N0 ) (1.3)

respectively. From (1.2) and (1.3), we can get 2ηSE = 1 + P

W N0

. (1.4)

Then, the SE-EE relation can be expressed as

ηEE =

ηSE

(2ηSE − 1)N

0

, (1.5)

which is plotted in Fig. 1.1. From (1.5), as ηEE approaches to zero, ηSE tends to

infinity. In contrast, when ηEE approaches to zero, ηSE converges to 1/(N0ln 2). It

is impossible to satisfy both metrics in the same time.

However, (1.5) is specific to point to point instead of networks transmission. If more practical constrains and transmission strategies are taken into consideration, such as transmission techniques, modulation and coding schemes, transmission dis-tances, and resource allocation algorithms, the SE-EE curve will not be as monotonic as shown in Fig 1.1 [9]. Therefore, it is worthy of assessing the SE-EE relation for the future energy-efficient communication system design.

(17)

ņņ

Ŕņ

Figure 1.1: The relation between EE and SE.

1.3

Thesis Outline

The remainder of the thesis is organized as follows. In Chapter 2 we introduce the background of our work. System models are illustrated in Chapter 3. In chapter 4, we address our simulation methodology of heterogeneous network CoMP systems. The procedure of our joint design is described in Chapter 5. Then, the numerical and simulation results are shown in Chapter 6. Finally, Chapter 7 concludes the thesis.

(18)

5

CHAPTER 2

Background

2.1

Why Green Communication?

The way people access information has been revolutionized by the consecutively re-newing technology. Dramatic growth on data traffic and the requirements for ubiq-uitous access have triggered tremendous expansion of network infrastructure and correspondingly ascending escalation of energy needs. The estimated growing rate of network subscribers is 20 percent per year, i.e., doubled every five years. It is reckoned that the ICT industry is accountable for three percent of world’s annual electrical power consumption and two to four percent of worldwide carbon dioxide emission which put great threats on global environment. There have been shown that about three billion mobile handsets and around three million base sites worldwide. The electricity bill is more than 10 billion dollars each year [10]. From the oper-ators’ perspective, reducing energy consumption not only diminishes carbon print, but also saves operating expenditure costs, so the social responsibility and operation profit are both catered. In fact, over 80 percent of the total ICT energy consump-tion is attributed to the radio access network. Studies have clearly indicated that the power drain of mobile handset equipments is far lower than that of BSs. Hence, energy reduction schemes are supposed to mainly focus on BSs. Such a goal can be achieved by renovating existing network structures with more energy efficient

(19)

deploy-ment strategies or applying novel transmission techniques of BSs. Briefly speaking, energy-efficient technologies will be indispensable for helping ICT industries face chal-lenges in a more and more energy-constrained future.

2.2

Coordinated Multi-point Techniques

Cooperation among BSs for data transmissions to one or more user equipment (UE) is known as a key technique to reach the requirement of the IMT-Advance in terms of both overall and cell-edge system throughput of cellular communication networks. The network multiple-input-multiple-output (MIMO) technique is also referred as CoMP by 3GPP. Moreover, it is thought to be one of potential contributors in en-hancing energy efficiency of upcoming LTE-A systems. There are four downlink CoMP deployment scenarios [11]:

(1) Scenario 1: Homogeneous network with intra-site CoMP, as shown in Fig. 2.1(a). (2) Scenario 2: Homogeneous network with high transmit power remote radio heads (RRHs). RRHs are linked to the macro-cell with high speed backhaul, as shown in Fig. 2.1(b).

(3) Scenario 3: Heterogeneous network with low power RRHs within the macro-cell coverage. RRHs share different cell IDs with the macro-cell, as shown in Fig. 2.1(c).

(4) Scenario 4: Heterogeneous network with low power RRHs within the macro-cell coverage. RRHs share the same cell ID with the macro-cell, as shown in Fig. 2.1(c).

Although Scenarios 1 and 2 are developed the earliest, scenario 4 has the most potential and advantages.

(20)

Macro cell Coordination area (a) Scenario1. Macro cell High Tx power RRH Optical fiber Coordination area (b) Scenario2. Macro cell Low Tx power RRH Optical fiber Coordination area (c) Scenario3/4.

Figure 2.1: Various CoMP scenarios. 7

(21)

In general, CoMP techniques can be categorized into two classes, which are joint processing (JP) and coordinated scheduling / coordinated beamforming (CS/CB). Most of current researches on CoMP schemes have focused on JP. In the class of JP, by sharing channel state information among BSs through network backhaul, the sig-nal can be processed jointly before transmitting and sigsig-nals from other cells may assist the communication instead of being treated as a detrimental interference. The same information to the target UE is simultaneously transmitted from different co-operating BSs through the same frequency resource and coherently or non-coherently combined at the receiver side, which improves received signal quality and mitigates interferences. In terms of CS/CB, the data to single UE is only available at and trans-mitted from one point in the cooperating set for a time frequency resource, while user scheduling/beamforming decisions are made among coordinated cells.

2.3

Literature Survey

HetNet is thought to be power efficient for deploying low power nodes to extend the cell coverage and improve system capacity. It is true that SE may increase with the denser network. However, both [12] and [13] indicated that as the number of pico cells exceeds a certain amount, the gain in system throughput cannot compensate the extra power consumption of pico sites, which degrades EE. Hence, small cell deployment is suggested being limited under a certain degree. The work [14] further analyzed HetNet in terms of EE and success probability, indicating that there exists an optimal pico-macro density ratio that maximizes EE, which offers the knowledge in how to establish green heterogeneous networks. [15] pointed out that despite that overall EE is improved, but the macro-cell performance is slightly degraded due to interferences from small cells. In brief, increasing small cells is not always beneficial for improving the system throughput.

(22)

Table 2.1: Comparison of Our Work and Related Works

Heterogeneous Coordinated System SE-EE

Networks Multi-Point Architecture Analysis

[17] × [16] × × [18] × × [14] × × [15] × Our Work

CoMP techniques can improve network throughput significantly, but it also requires more power to sustain such operations. In [16], authors compared system capacity of various density cooperative networks. Their results shows the denser network is, the SE higher will be. Nevertheless, the relation of energy efficiency is not drawn. Tradeoff between cell throughput gains via inter-site CoMP scheme and increased power dissipation has been investigated under various cell dimensions and cooperation cluster sizes [17]. A similar investigation has been conducted in [18]. The energy efficiency analysis of joint transmission (JT) CoMP in homogeneous and heterogeneous network is addressed. However, both [17] and [18] did not take the concept of cooperation among macro-cell and lower power sites into consideration. We analyze the system capacity and power consumption of heterogeneous CoMP networks under various system settings to seek energy-efficient transmission schemes and system architectures. Related researches are summarized and compared with our work at Table 2.1.

(23)

10

CHAPTER 3

System Models

In this chapter, we illustrate our system models thoroughly. At the beginning, we present the system architecture of our simulator, including the homogeneous network for single user multi-input-multi-output (SU-MIMO) systems and the heterogeneous network for CoMP systems. Channel model and radio environment are introduced in the second section. The power consumption model is given in the last section.

3.1

Cell Layout

3.1.1

Homogeneous Network SU-MIMO System

MIMO transmission techniques have been studied extensively in the past. Spatial multiplexing has drawn much attention due to the capability of increasing spectral efficiency. Spatial multiplexing of multiple data streams to single target UE in the same time-frequency resource is referred as SU-MIMO. We set the homogeneous net-work SU-MIMO system composed by 19 cells with hexagonal gird as our simulation baseline, which is shown in Fig. 3.1. Each cell is divided into three sectors, and each sector is equipped with a directional antenna.

(24)

R

UE Macrocell

Signal from macrocell

Figure 3.1: Cell architecture of SU-MIMO systems.

(25)

3.1.2

Heterogeneous Network CoMP System

In accordance with 3GPP-A Rel-11, downlink CoMP techniques can be categorized into four scenarios, among of which scenario 4 is the most promising. We choose scenario 4 as the simulation environment so as to combine the concepts of HetNet and CoMP techniques. The cellular system is composed of 19 hexagonal cells, where each cell is divided into 3 sectors, and every sector is equipped with directional antennas as shown in Fig. 3.2. The benefit of using directional antenna is that the signal can be concentrated to be transmitted within some range, which makes the signal more strengthened and interferences among sectors can be mitigated. RRHs are deployed in each sector and connected with the macro-cell by 109 bytes level optical

fibers. Fibers are able to carry a large amount of data. In this way, remote units can exchange information with the macro-cell, and the signal processing can be jointly performed by RRHs and the macro-cell in a centralized manner. Minimum distances between the BS an UE are listed in Table 3.1.

Table 3.1: Deployment regulations

Minimum Distance Macro-RRH > 75m RRH-RRH > 40m Macro-UE > 35m RRH-UE > 10m 12

(26)

R

UE

RRH

Signal from macrocell Signal from RRH Backhaul

Macro cell

Figure 3.2: Cell architecture of CoMP systems.

(27)

3.2

Channel Model

3.2.1

Radio Environment

Based on [19], we consider path loss model, shadowing fading and antenna pattern in the radio environment. The path loss model is a major factor in the analysis and design of the link budget of wireless communication systems, describing how the power density of an electromagnetic wave reduces as it propagates through space. Path loss models of macro-cell and RRH in decibel are defined as:

P Lmacro(d) = 128.1 + 37.6· log10(d), (3.1)

and

P LRRH(d) = 140.7 + 36.7· log10(d) (3.2)

respectively, where P Lmacro and P LRRH denote the power attenuation from

macro-cell/RRH to UE, and d is the distance between the UE and macro-macro-cell/RRH. More-over, P L can be further converted into the form of effective gain:

GP L = 10−(P L(d)/10). (3.3)

The shadowing fading originates from the obstacles on propagation paths. The dis-tribution is modeled by a log-normal random variable with zero mean. 8 dB standard deviation for macro-cell to UE, and 10 dB standard deviation for RRH to UE, re-spectively. The horizontal antenna pattern is defined as:

AH(φ) =− min[12(

φ φ3dB

)2, Am], (3.4)

where φ denotes the horizontal angle between BS and UE, φ3dB is 70 degree, and Am

is 25 dB. It can also be written as the form of effective gain:

GAP(φ) = 10[AH(φ)/10]. (3.5)

(28)

3.2.2

Sptial Channel Model

The Spatial channel model (SCM) is a standardized model developed by 3GPP for evaluating cellular MIMO systems. In our work, we adopt SCM urban scenario [20]. Assume that there are N resolvable paths in each link from a BS to a UE, and each path consists of M irresolvable subpaths. A simplified plot of SCM is shown in Fig. 3.3. BS W UE W , , n m AoD D , n AoD d n AoA, d , , n m AoD q , , n m AoA q BS q UE q V q ŃŔġłųųŢź ŏ ŖņġłųųŢź ŖņġŕųŢŷŦŭ ŅŪųŦŤŵŪŰů ŏ őŢŵũġů ŃŔġłųųŢźġŃųŰŢťŴŪťŦ ŖŴŦųġłųųŢźġ ŃųŰŢťŴŪťŦ , , n m AoA D ŔŶţűŢŵũġŮ Figure 3.3: Parameters of 3GPP SCM.

For Nt transmit antennas at the BS and Nr receive antennas at the UE, the

channel impulse response in time domain for the nth path between the sth transmit

and uth receive antenna can be written as:

hn(t) =            h1,1,n(t) h1,2,n(t) · · · h1,Nt−1,n(t) h1,Nt,n(t) h2,1,n(t) h2,2,n(t) · · · h2,Nt−1,n(t) h2,Nt,n(t) .. . ... · · · ... ... hNr−1,1,n(t) hNr−1,2,n(t) · · · ... ... hNr,1,n(t) hNr,2,n(t) · · · hNr,Nt−1,n(t) hNr,Nt,n(t)            . (3.6) 15

(29)

Each element in hn(t) can be written as: hu,s,n(t) =Pn M Mm=1

[ejkdssin(θn,m,AoD+ψn,m)ejkdusin(θn,m,AoA)ejkV cos(θn,m,AoA−θV)t]. (3.7)

where Pnis the power of the nthpath, M is the number of subpaths per path, k is the

carrier wave number which equals to 2π/λ, λ is the carrier wavelength in meters, ds

is the distance in meters from the first antenna to the sth antenna at the BS, θn,m,AoD

is the angle between the mth subpath of the nth path and the BS array broadside,

ψn,m is the phase of the mth subpath of the nth path which uniformly distributes in

the interval [0◦, 360◦], du is the distance in meters from the first antenna to the uth

antenna at the UE, θn,m,AoA is the angle between the mth subpath of the nth path

and the UE array broadside, V is the UE velocity, and θV is the angle between the

UE travel direction and the UE array broadside. For multi-carrier OFDM systems, the channel impulse response from the sth transmit to the uth receive antenna in

frequency domain of the kth subcarrier is:

HSCM(k) =            H1,1(k) H1,2(k) · · · H1,Nt−1(k) H1,Nt(k) H2,1(k) H2,2(k) · · · H2,Nt−1(k) H2,Nt(k) .. . ... · · · ... ... HNr−1,1(k) HNr−1,2(k) · · · ... ... HNr,1(k) HNr,2(k) · · · HNr,Nt−1(k) HNr,Nt(k)            . (3.8)

Each element in HSCM(k) can be expressed as:

Hu,s(k) = F F T{[hu,s,1(t), hu,s,2(t), ..., hu,s,n(t)]}, (3.9)

where F F T{·} denotes the Fourier transform function.

(30)

3.3

Power Consumption Model Description

3.3.1

Network Power Consumption

The total network input power includes power consumption in RAN, backhaul, and signal processing for CoMP:

Ptotal = Psite+ Psp+ Pbh, (3.10)

where

Psite= PP DS + PAC+ PT rans+ PBS. (3.11)

Related details of Psite are listed in Table 3.2 [21]. Psp denotes the signal processing

power for cooperation, it can be expressed as [17]:

Psp = 58· (0.87 + 0.1 · Nc + 0.03· Nc2), (3.12)

where Nc denotes the number of cooperated sites. Pbh is the power for backhaul

connection, which can be written as:

Pbh=

R

100M · 50, (3.13)

where R denotes the data rate.

Table 3.2: Components of the site power

W att PP DS 485.23

PAC 1940.96

PT rans 650

(31)

3.3.2

Base Station Power Consumption

There are different types of BS such as macro-cell, micro-cell, pico-cell, femto-cell, and RRH. Because each kind of BS has different constituents and power figures, it is not easy to describe the precise BS architectures and corresponding power dissipation models thoroughly. In our work, we adopt the power consumption model of the baseband unit (BBU) plus remote radio unit (RRU) based BS as illustrated in Fig. 3.4 [22]. The basic idea of BBU plus RRU system is to separate the baseband part and the radio frequency part of BSs. RRUs are connected to the BBU with optical fibers, and each RRU is equipped with transceivers. The BBU is responsible for baseband signal processing and radio resource management, so the signal can be processed jointly at BBUs. It is cost-saving for radio access points deployment because smaller BBUs require less equipment room, and RRUs can be deployed more flexibly depends on traffic needs.

Each BS consists of multiple transceivers (TRXs), and each TRX is comprised of a power amplifier (PA), a radio frequency (RF) module, a baseband processor, a DC-DC supply, and an AC-DC unit (mains supply), which is shown in Fig. 3.5. As a result, we can decompose the power consumption model into four classes generally [23] [24] :

(1) Power Amplifier (PA): In this part, we consider the PA power efficiency and feeder losses which is caused by extra feedback for pre-distortion and additional signal processing. For small cells, feeder losses are elided, but necessary in macro and micro cells. The power consumption of a PA can be written as:

PP A =

Pout

ηP A(1− σf eed)

, (3.14) where ηP A is the PA efficiency, and σf eed is the feeder loss factor.

(2) Radio Frequency (RF): The RF part contains a transmitter and a receiver for 18

(32)

䣄䣄䣗

䣔䣔䣗

䣄䣄 䣒䣱䣹䣧䣴䢢䣵䣷䣲䣲䣮䣻 䣆䣅䢯䣆䣅 䣔䣈 䣒䣃 䣵䣷䣲䣲䣮䣻䣒䣱䣹䣧䣴䢢 䣆䣅䢯䣆䣅

Figure 3.4: BBU+RRU based system.

(33)

3$ 3$ 5) '&'& %% 5) &RROLQJ 0DLQ6XSSO\ 3LQ 3RXW

Figure 3.5: Power model of a BS.

downlink and uplink, respectively. Different types of RF modules are required for different types of BSs. For example, low-intermediate frequency (IF) are suitable for macro and micro sites, and zero-IF is favored by smaller cells.

(3) Baseband unit: The baseband unit deals with modulation/demodulation, channel coding/decoding, channel estimation, equalization, pre-distortion, etc.

(4) Heat emission: Any losses related to the system powering are included in this category, such as DC-DC conversion and mains supply.

According to above, given PA, RF, and baseband unit power consumption,

PP A, PRF and PBB, the power consumption for BS input is:

PBS = β[NT RX(PP A+ PRF + PBB)], (3.15)

where β is defined as:

β = 1

(1− σDC)(1− σM S)

, (3.16) and σDC, σM S,and σcool are loss factors of the DC/DC, mains supply and cooling

system. Thus, the input power for BS can be further expressed as:

PBS = NT RX · Pmax ηP A + PRF + PBB (1− σDC)(1− σM S) . (3.17) 20

(34)

We assume that the BS power consumption is proportional to the number of transceiver chains NT RX, namely, transmit/receive antenna pairs per site, and all the BSs with

the maximal load,

Pmax = 10(Pout+σf eed−30)/10. (3.18)

The related power consumption parameters of different types BSs are listed in Table 3.3.

Table 3.3: Power consumption for various BSs

Macro BBU Macro RRU Pico BBU Pico RRU

Pout [W] 0 46 0 30 σf eeder [dBm] 0 1 0 0 PA Pmax [W] 0 50.11 0 1 ηP A [%] 0 31.1 0 6.7 PP A [W] 0 161.15 0 14.92 Heat σDC [%] 8 8 9 9 emission σM S [%] 9 9 12 12 β # 1.19 1.19 1.24 1.24 RF PRF [W] 0 12.9 0 1 BB PBB [W] 29.6 0 3 0 Sectors # 3 1 1 1 Antennas # 2 2 1 1 Carriers # 1 1 1 1 Pin [W] 212.13 415.79 3.74 19.88 21

(35)

22

CHAPTER 4

Heterogeneous CoMP Networks Simulator

In this chapter, the simulation environment of the 3GPP LTE-A heterogeneous co-ordination multi-point networks will be illustrated in details, including joint signal process for base station cooperation, precoding method, scheduling method, and SINR calculation.

4.1

Codebook-based Precoding

In FDD systems, the CSI needs to be fed back from UEs to the eNB. It is well known that the system performance can be improved with full channel state infor-mation (CSI) at the transmitter. However, because the feedback channel bandwidth is limited, complete channel state feedback leads to excessive overhead. Therefore, the codebook-based precoding method is adopted to reduce the feedback overhead. The UE calculates the precoding matrix and feedback a precoding matrix indicator (PMI) as the index of a codeword. The codebook is designed in the off-line manner, and the same codebook set is available at both the transmitter and the receiver. The feedback overhead can be reduced enormously by only feeding the PMI back rather than the full precoding matrix.

Codebooks for two antenna and four transmit antenna ports are defined in [25]. Tables 4.1 and 4.2 list codewords for two and four transmit antenna ports, respectively. The Frobenius norm of each codeword is normalized to unity to remain the same

(36)

transmit power.

Table 4.1: Codebook for Two Antenna Ports

Index Rank-1 Rank-2

C0 12    1 1    1 2    1 1 1 −1    C1 12    1 −1    1 2    1 1 j −j    C2 12    1 j    C3 12    1 −j   

For four transmit antennas, the householder matrix Wc can be derived by uc,

i.e., Wc = I4 − 2ucuHc

/

uHc uc. Wac denotes the ath column vector of Wc, and Wa,bc

denotes ath and bth column vectors of Wc.

We assume that UEs can estimate the CSI perfectly based on reference signals from BSs. Singular value decomposition-based (SVD) precoding method is adopted in our work. Denote Hm,n,i as the channel matrix from the mth sector to the ith UE

in the nth sector. Then, the channel matrix is decomposed by the SVD technique:

Hm,n,i = Um,n,iDm,n,iVm,n,iH , (4.1)

where Um,n,i = [u1m,n,i· · · u Nr m,n,i]∈ C Nr×Nr, (4.2) 23

(37)

Table 4.2: Codebook for Four Antenna Ports

Index uc Rank-1 Rank-2

C0 u0 = [ 1 −1 −1 −1 ]T W10 W1,40 /√2 C1 u1 = [ 1 −j 1 j ]T W11 W1,41 /√2 C2 u2 = [ 1 1 −1 1 ]T W12 W1,42 /√2 C3 u3 = [ 1 j 1 −j ]T W13 W1,23 /√2 C4 u4 = [ 1 (−1 − j)/√2 −j (1 − j)/√2 ]T W14 W1,44 /√2 C5 u5 = [ 1 (1− j)/√2 j (1− j)/√2 ]T W15 W1,45 /√2 C6 u6 = [ 1 (1 + j)/√2 −j (−1 + j)/√2 ]T W16 W1,36 /√2 C7 u7 = [ 1 (−1 + j)/√2 j (1 + j)/√2 ]T W17 W1,37 /√2 C8 u8 = [ 1 −1 1 1 ]T W18 W1,28 /√2 C9 u9 = [ 1 −j −1 −j ]T W19 W1,49 /√2 C10 u10= [ 1 1 1 −1 ]T W110 W1,310/√2 C11 u11 = [ 1 j −1 j ]T W111 W1,311/√2 C12 u12= [ 1 −1 −1 1 ]T W112 W1,212/√2 C13 u13= [ 1 −1 1 1− ]T W113 W1,313/√2 C14 u14= [ 1 1 −1 −1 ]T W114 W1,314/√2 C15 u15= [ 1 1 1 1 ]T W115 W1,215/√2 24

(38)

Dm,n,i =            λ1m,n,i 0 · · · · 0 0 · · · 0 0 . .. 0 0 ... ... . .. ... .. . 0 . .. 0 ... ... . .. ... .. . 0 0 . .. 0 ... . .. ... 0 · · · 0 λNr m,n,i 0 · · · 0            ∈ RNr×Nt, (4.3) and Vm,n,iH = [v1m,n,i· · · vNt m,n,i] H ∈ CNt×Nt. (4.4) Note that λ1 m,n,i ≥ λ2m,n,i ≥...≥ λ Nr

m,n,i. For SVD-based precoding techniques, Vm,n,i

is exploited, and each column of Vm,n,i is called an Eigenvector of Hm,n,i(Hm,n,i)H,

which is related to an Eigenmode of the channel. The individual Eigenmode quality can be defined by each singular value of Hm,n,i. If the signal is transmitted with ND

data streams to the ith UE in the nth sector, ND column vectors from the left of Vm,n,i

will be selected as the full precoding matrix fWm,n,i. The corresponding codeword is

selected based on the minimum angle between each codeword and the full precoding matrix: a = arg max i trace ( CH i Wfm,n,i ) , (4.5)

where Ci denotes the codeword defined in Tables 4.1 and 4.2, and a is the PMI to be

fed back to BSs.

4.2

Transmission Equations

For the sake of simplicity, we discuss the following equations in the single carrier case. The OFDM multi-carrier case can be extended in the same way. Let Hm,n,i ∈ CNr×Nt

represent the channel matrix from the mth sector to the ith UE in the nth sector, and

(39)

HRRHmr,n,i ∈ CNr×Nt denotes the channel matrix from the r

th RRH in the mth sector to

the ith UE in the nth sector similarly. In accordance with (3.3), (3.5), and (3.9), the

channel matrix Hm,n,i can be written as:

Hm,n,i = HSCM

GP L· σS· GAP, (4.6)

where σS is the shadowing factor. In the same manner, HRRHmr,n,i has a similar

math-ematic form with Hm,n,i. For SU-MIMO transmission, the effective channel matrix

Hef fm,n,i ∈ CNr×Nt detected by the i

th UE is

Hef fm,n,i = Hm,n,i. (4.7)

As for CoMP transmission, RRHs share the same cell ID with the macro-cell, the whole network can be regarded as a distributed antenna system. Thus data trans-mitted by the macro-cell and RRHs are placed on the same resource blocks (RBs). Consequently, the effective channel matrix Hef fm,n,i ∈ CNr×Nt detected by the i

thUE in

the nth sector is the combination of channel matrices from the macro-cell and RRHs,

which can be written as:

Hef fm,n,i =∑ m Hm,n,i+ ∑ r HRRHmr,n,i. (4.8)

4.2.1

Single UE Case

Assume that the nth sector is the serving sector antenna of the ith UE. The received

signal matrix of the ith UE in the nth sector can be written as:

Yn,n,i=

Desired signal

z }| {

Hef fn,n,iWn,n,iXn,n,i+

Inter−cell interference

z }| {

57

m̸=n

Hef fm,n,iWm,m,iXm,m,i+ N oise

z }| {

Nn,n,i. (4.9)

The second term in (4.9) is the interference from other sectors. Thus Xn,n,i is the

desired data matrix to the ith UE in the nth sector, Wn,n,i is the precoding matrix to

(40)

the ith UE in the nth sector, and Nn,n,i∈ CNr×1 is the additive white Gaussian noise

with power spectral density σ2

n=−174dBm/Hz.

For the rank-1 transmission, the maximal ratio combining (MRC) technique is adopted. Assume that the desired data matrix is Xn,n,i= x1n,n,i and the

correspond-ing precodcorrespond-ing matrix is Wn,n,i = w1n,n,i. The desired data can be demodulated via

multiplying Yn,n,i by the MRC matrix:

Mn,n,i= [Hef fn,n,iWn,n,i]H. (4.10)

For the rank-2 transmission, the minimum mean square error (MMSE) receive tech-nique is adopted. Assume that the desired data matrix is Xn,n,i =

[

x1n,n,i x2n,n,i

] and the corresponding precoding matrix is Wn,n,i =

[

w1n,n,i w2n,n,i

]

. The desired data can be demodulated via multiplying Yn,n,i by MMSE matrices M1n,n,i ∈ C1×Nr

and M2n,n,i ∈ C1×Nr, respectively, where

M1n,n,i= [( σn2I + Hn,n,iwn,n,i2 ( Hn,n,iw2n,n,i )H)−1 Hn,n,iw1n,n,i ]H , (4.11) and M2n,n,i= [( σn2I + Hn,n,iw1n,n,i ( Hn,n,iw1n,n,i )H)−1 Hn,n,iw2n,n,i ]H . (4.12) Hence, the received SINR of the ith UE in the nth sector can be written as:

γn,n,i =

PT x Mn,n,iH ef f

n,n,iWn,n,iXn,n,i

2 PT x Mn,n,i 57 ∑ m̸=n

Hef fm,n,iWm,m,iXm,m,i

2 +∥Mn,n,iNn,n,i∥ 2 , (4.13)

where PT x is the transmit power of BS.

(41)

4.2.2

Multiple UE Case

We assume that there are two UEs in each sector: the ith UE and the jth UE. The

received signal matrix of the ith UE in the nth sector can be written as:

Yn,n,i =

Desired signal

z }| {

Hef fn,n,iWn,n,iXn,n,i+

Inter−user interference

z }| {

Hef fn,n,jWn,n,jXn,n,j

+

Inter−cell interference

z }| {

57

m̸=n

(Hef fm,n,iWm,m,iXm,m,i+ Hef fm,n,jWm,m,jXm,m,j) +

Noise

z }| {

Nn,n,i

. (4.14)

The second term of (4.14) is the interference from the other UE in the same cell, and the third term is the interference from other cells. Likewise, we can derive the received signal matrix of the jth UE in the nth sector like (4.14) in the same manner:

Yn,n,j =

Desired signal

z }| {

Hef fn,n,jWn,n,jXn,n,j+

Inter−user interference

z }| {

Hef fn,n,iWn,n,iXn,n,i

+

Inter−cell interference

z }| {

57

m̸=n

(Hef fm,n,jWm,m,jXm,m,j+ Hef fm,n,iWm,m,iXm,m,i) +

Noise

z }| {

Nn,n,j

. (4.15)

To demodulate received signals of the ith and the jth UE, we multiply Yn,n,i

and Yn,n,j by MMSE matrix

M1n,n,i= [(σn2I + Hef fn,n,iWn,n,j(Hef fn,n,iWm,m,j)H)−1Hef fn,n,iWn,n,i]H (4.16)

and

M1n,n,j = [(σn2I + Hef fn,n,jWn,n,i(Hef fn,n,jWm,m,i)H)−1Hef fn,n,jWn,n,j]H, (4.17)

respectively. Hence, the received SINRs of the ith UE and the jthUE in the nth sector

can be written as:

γn,n,i =

PT x Mn,n,iHef fn,n,iWn,n,iXn,n,i

2 PT x Mn,n,i 57 ∑ m̸=n

Hef fm,n,iWm,m,iXm,m,i

2 +∥Mn,n,iNn,n,i∥2 , (4.18) 28

(42)

and γn,n,j = PT x Mn,n,jH ef f n,n,jWn,n,jXn,n,j 2 PT x Mn,n,j 57 ∑ m̸=n Hef fm,n,jWm,m,jXm,m,j 2 +∥Mn,n,jNn,n,j∥ 2 , (4.19)

respectively, where PT x is the transmit power of a BS.

4.3

CoMP Schemes

In the JP transmission, the precoding matrix is decided by decomposing the received effective channel matrix from multiple BSs based on the SVD method. For the intra-site JP, the effective channel matrix of the ith UE in the nth sector can be expressed

as

Hef fm,n,i = Hm,n,i+

r

HRRHmr,n,i. (4.20) For the inter-site plus intra-site JP, the effective channel matrix of the ith UE in the

nth sector can be expressed as

Hef fm,n,i = 3 ∑ m=1 Hm,n,i+ ∑ r HRRHmr,n,i. (4.21) Apart from JP, we also consider the CS/CB technique for the inter-cell coor-dination. All the three cells serve an UE in the same time for JP transmission. On the other hand, each cell serves only an UE at a time, but manages to reduce the interference to UEs in other cells for CS/CB transmission . Take a two cell system as shown in Fig. 4.1 for illustration. We can express the received signal of UE1 as

Y1 = H11W1X1+ H21W2X2+ N, (4.22)

(43)

where H11 denotes the channel matrix from BS1 to UE1, H21 denotes the channel

matrix from BS2 to UE1. W1 and W2 are precoding matrices of UE1 and UE2,

respectively. The SINR of UE1 can be written as:

SIN R1 =

P1∥M1H11W1 2

P2∥M1H21W22+ σn2

, (4.23) where M denotes the demodulation matrix, P is the transmit power, and σn2 denotes

the thermal noise power. In order to maximize SIN R1, the interference term of

(4.23) should be mitigated. Thus, once W1 is determined, we try to select a W2

which makes H21W2 approaches zero from the codebook, in other words, finding a

precoing matrix which tends to be orthogonal to the channel matrix. For the rank-1 transmission, W2 is chosen from the codebook according to

arg min

j (|H11W1(H21Wj)| H

). (4.24) For the rank-2 transmission, W2 is chosen from the codebook according to

arg min

j [tr(|H11W1(H21Wj)| H

)]. (4.25)

4.4

Proportional Fair Scheduling

Proportional fair scheduling is a compromise-based scheduling algorithm. Both fair-ness of all UEs and maximizing the network throughput are taken into account. If maximizing network throughput is the highest priority, UEs with better link quality will be always served and UEs with poor link quality will be neglected. With pro-portional fair scheduling, both transmission rate and fairness are concerned. An UE will be served if its current transmission rate is high or its past average transmission rate is low. On the contrary, if the current transmission rate of another UE is low or its past average transmission rate is high, its serving priority is lower than the former

(44)

R R UE Macro cell Interference Signal Cell 1 Cell 2 UE 1 UE 2

Figure 4.1: Two cell scenario of CS/CB.

(45)

one. By using proportional fair scheduling, the BS transmits data to the UE i∗ in the rth RB: i∗ = arg max i Ri Ti , (4.26)

where i = 1, 2, ..., N , Ri represents the current transmission rate of the ith UE in the

rth RB, and Ti denotes the past average transmission rate of the ith UE before the

rth RB.

4.5

Exponential Effective SINR Mapping (EESM)

Exponential Effective SINR Mapping (EESM) method is used to map the instan-taneous subcarrier channel state, such as instaninstan-taneous SINR, to the corresponding BLER (Block Error Rate) value. EESM maps a set of subcarriers’ SINRs into an instantaneous effective SINR. The generalized EESM is stated as:

γef f ≡ EESM(γ, β) = −β ln( 1 Nc Ncj=1 e−γjβ), (4.27)

where γj is the SINR of jthsubcarrier and Ncis the number of subcarrier per subband.

In the 3GPP Release 9, 15 different modulation and coding schemes (MCS) corre-sponding to 15 different CQI are defined in [26]. The effective SINR can be mapped to a certain MCS by SINR-BLER curve simulated by link level simulator [27]. We choose the highest MCS with BLER not exceeding 0.1.

4.6

Hybrid Automatic Repeat Request (HARQ)

Since the link level behavior is not included in our system level simulation, we model the transmission error by introducing a random variable with probability mass

(46)

tion (PMF) [28]: Pe(X) =    0.1, x = 0 0.9, x = 1 , (4.28)

where x = 0 represents the occurred error, and x = 1 represents the successful trans-mission. If there is an error, data will be retransmitted again in the next transmis-sion time interval (TTI). Suppose that we obtain a throughput T before considering retransmission. Then, the relation between retransmission time Nre which is not

exceeding three and throughput Tre is:

Tre =

T

(1 + Nre)

, (4.29)

where 0 ≤ Nre ≤ 3. The more the retransmission occurs, the lower the throughput

will be. Once the retransmission times is larger than three, the data will be discarded and the throughput of TTI will be zero.

4.7

Energy Efficiency

We adopt the bit per Joule as the performance metric which is widely used in the world. Denote Ttotal and Ptotal as the overall system throughput and the total system

power consumption, respectively. The energy efficiency metric is defined as:

EE(bits/J oule) = Ttotal(bits/s) Ptotal(W )

. (4.30) The EE metric shows us how many bits can be transmitted when one Joule is con-sumed. Based on (4.30), different systems are able to be evaluated the energy effi-ciency performance.

(47)

34

CHAPTER 5

Tradeoff Design of Spectral and Energy

Efficiency in HetNet Systems

The goal of this chapter is to evaluate the spectral and energy efficiency of hierarchical base station cooperation techniques in HetNet environment and propose a system design methodology for energy-efficient transmission. Because almost 80% of overall system power consumption is attributed to radio access network, we focus on two aspects of BSs, i.e., system architecture and cooperative transmission techniques.

5.1

System Architectures

For system architectures, we consider two factors: RRH deployment and sector an-tenna architectures. The RRH can provide additional signal power to UEs by cooper-ating with macro- cell. It plays a vital role in enhancing signal quality. Therefore, we want to investigate the suitable location and the number of RRHs to improve spectral and energy efficiency of the system. If RRHs are placed close to the macro-cell, cell-edge UEs will barely benefit from RRHs due to the long distance. On the contrary, if RRHs are deployed nearly at cell margin, it might cause strong interference to the cell-edge UEs of adjacent cells. In addition, we are interested in the relation of the cooperative RRHs quantity with system performance. Intuitively, increasing RRHs seems to improve signal quality more. However, it also causes more interference to

(48)

other cells and consumes power. Hence, finding a moderate numbers of RRHs or even how to decide the serving RRH is important.

Sector antenna architectures have various antenna patterns causing various transmit power distribution, and having different impact on the system performance. We consider the following three types of sector antenna architectures, as shown in Fig. 5.1.

(1) Diamond shape: Each BS is equipped with three 120 directional antennas, and the 3 dB power attenuation angle is set at 70.

(2) Pentagon shape: Basically, this architecture has the same antenna pattern as diamond shape, but central antenna beams are rotated by a certain angle, which makes a different cell sectorization figure from diamond shape to pentagonal shape. (3) Narrow beam: Each BS is equipped with three 60 directional antennas, and the

3 dB power attenuation angle is set at 30 [29].

Three adjacent central antenna beams direction of diamond-shape architecture meet at the same point. Thus, cell-edge UEs are easily affected by adjacent cells. However, because the pentagon shape are rotated by a certain angle, it causes fewer interference to other cells. Compared with two other types, the narrow beam can further reduce ICI because the virtual cellular figuration matches realistic cellular figuration better, and thus neighboring sectors cause less interfere to each other.

5.2

CoMP Transmission Techniques

In conventional MIMO systems, cell-edge UEs are more easily interfered by adjacent cells than cell-center UEs. In order to improve signal quality of cell-edge UEs, we deploy small cells within the cell to cooperate with the macro-cell processing signal

(49)

Diamond shape

(a) Diamond shape.

Pentagon shape

(b) Pentagon shape.

Narrow beam

(c) Narrow beam.

Figure 5.1: Sector antenna architectures. 36

(50)

jointly, which is called intra-site CoMP. In this way, the signal strength of cell-edge UE can be enhanced through the additional cooperative signal from small cells. In our work, joint processing (JP) technique is adopted in the intra-site level CoMP transmission. To further mitigate the interference from adjacent cells, we also consider the inter-site cooperation. Therefore, we introduce two other neighboring cells to cooperate with the observing cell. Fig. 5.2 illustrates the two types of coordination schemes mentioned above. For the inter-site level CoMP, we compare inter JP and

R

R R

Inra-site CoMP

Inter plus intra-site CoMP

Figure 5.2: Two kinds of CoMP schemes.

inter CS/CB in addition to the intra-site JP. Three cells serve an UE jointly at the same time in JP. For CS/CB, each cell serves a individual UE, but manages to mitigate interference to UEs in other cells. It is not easy to predict which one is

(51)

better. Therefore, we compare performances of two schemes by simulation to find the best joint level transmission scheme.

5.3

Design Procedures

Fig. 5.3 shows the following steps to seek the optimal design for energy-efficient transmission in HetNet environment. Firstly, we start from the intra-site transmis-sion level. We evaluate the system performance under different RRH locations and sector antenna architectures to learn the best setting. In these steps, we choose the best one with the highest spectral efficiency. Since power consumption of each setting is the same as others, spectral efficiency and energy efficiency are positively corre-lated. Followed by the previous settings, both cell throughput and energy efficiency of various number of RRHs are compared and examined to determine the better per-formance tradeoff. Hence, the number of RRHs to cooperate with the macro-cell can be determined to maximize EE. In this step, we determine the best one by energy effi-ciency. Spectral efficiency and energy efficiency can be negatively correlated because increasing RRHs consumes more power. After obtaining the best system architec-ture for the intra-site coordinated transmission, we examine different inter-site and intra-site CoMP transmission schemes to determine the most energy-efficient setting. We also select the scheme with the best capacity in this step. Then, once we know the best combination of intra and inter CoMP schemes, the proposed joint design of system architectures and CoMP transmission schemes for HetNet system is revealed.

(52)

3. Determine The Number of RRHs according to Spectral Efficiency-Energy Efficiency Chart Intra-site Inter-site HetNet

Choose The One with The Highest

Spectral Efficiency

Complete

Combine All The Best Settings for Joint Design

2. Determine

Sector Antenna Architecture according to

Spectral Efficiency-Antenna Chart

4. Determine

The Inter-site CoMP Scheme according to

Spectral Efficiency-Transmission Scheme Chart

Start

Intra-site

Choose The One with The Highest

Spectral Efficiency

Choose The One with The Highest

Energy Efficiency

1. Determine RRH Location

according to

Spectral Efficiency-Cell Radius Chart

Choose The One with The Highest

Spectral Efficiency

Figure 5.3: Joint design procedure for energy-efficient transmission.

(53)

40

CHAPTER 6

Numerical Results

6.1

Simulation Assumptions

According to [11] and [19], simulation parameter settings are listed in Table 6.1. Our simulation environment is 3GPP case 1, and the frequency division duplex (FDD) transmission mode is adopted. The cellular system consists of nineteen macro-cells with hexagonal grid, and each cell is divided into three sectors. Besides, the inter site distance (ISD) is 500 meters. The center frequency (CF) is 2 GHz, and the whole bandwidth is 10 MHz with 9 MHz available bandwidth. The pathloss and shadowing effect are considered, and the channel model is the SCM urban macro with high spread. The penetration loss is 20 dB. 10 UEs are uniformly distributed in each sector of the central cell, and the speed of UE is 3 km/hr. RRHs are equally spaced in the macro-cell coverage and share the same cell ID with the macro-cell. All 57 sectors are in the same frequency band, and the whole system is synchronized. Full buffer traffic mode is adopted. The maximum retransmission times is three.

6.2

Simulation Baseline

In order to evaluate the performance of heterogeneous CoMP networks, we set the SU-MIMO system as the comparison baseline. We list our simulation results and

(54)

Table 6.1: Simulation Parameters

Parameter Value

Duplex Method FDD

DL Transmission Scheme OFDMA

Number of Available Subcarriers 600

Transmit Antennas 2

Receive Antennas 2

ISD 500 meters

Macro-cell Number 19

UE Distribution Uniform in all cells

Frequency Reuse 1

Number of UE per Sector 10

UE Speed 3 km/hr

Network Synchronization Synchronized

Scheduling Method Proportional fair

HARQ Max 4 transmissions

Channel Estimation Ideal

Transmit Power Macro: 46 dBm

RRH: 30 dBm

Noise Power Density −174 dBm/Hz

Traffic Model Full buffer

Channel Model SCM urban macro

Antenna Pattern (Horizontal) AH(ϕ) =− min

[ 12 ( ϕ ϕ3dB ) , Am ] ϕ3dB = 70◦, Am= 25 dBm

Antenna Pattern (Vertical) AV(θ) =− min[12(θ−θθ3dBetilt)2, SLAv]

θetilt= 15◦, θ3dB= 10◦, SLAv= 20 dBm

Penetration Loss 20 dB

Pathloss Model Macro: 128.1 + 37.6· log10(d), d in km

RRH: 140.7 + 36.7· log10(d), d in km

Shadowing Model Lognormal with zero mean and

8 dB standard deviation

(55)

other results of the 3GPP technical report in Table 6.2 [30]. The maximal value of average spectral efficiency in 3GPP specification is 2.47 (bps/Hz/cell), and the minimal value is 2.14 (bps/Hz/cell). As for cell-edge spectral efficiency, the maximal value in 3GPP specification is 0.072 (bps/Hz/cell), and the minimal value is 0.100 (bps/Hz/cell). The average and cell-edge spectral efficiency of our work are 2.41 (bps/Hz/cell) and 0.083 (bps/Hz/cell), respectively. Our simulation results lie within the interval between the maximal and the minimal value in the 3GPP technical report.

Table 6.2: 2x2 SU-MIMO Simulation Results Comparison

Min value of Our Work Max value of

3GPP Spec. 3GPP Spec. Cell Average Spectral Efficiency 2.14 2.41 2.47 (bps/Hz/cell) 5% Cell-edge Spectral Efficiency 0.072 0.083 0.100 (bps/Hz/UE)

6.3

Intra-site CoMP

6.3.1

Effect of RRH Deployment

We notice that different RRH positions yield various system performances. In this section, we adopt the diamond shape antenna architecture first. As we can see from Fig. 6.1, when RRHs are placed 144.3 meters (0.5R) away from the macro-cell, the SE is higher than any other positions regardless of the number of RRHs. If RRHs are too close to the macro-cell, most UEs are out of RRH coverage. In contrast, if

(56)

RRHs are located near by the cell boarder, it will cause interference to cell-edge UEs in other cells. Since the power consumption is fixed despite of different distances between the RRH and macro-cell, EE is also the highest when RRHs are deployed at 0.5R. The similar trend of SE polygonal line can be seen in EE at Fig. 6.2. Thus, RRHs are always deployed 144.3 meters (0.5R) away from the macro-cell in the following simulations so as to achieve better SE and EE. The EE gain of 4 RRHs over SU-MIMO is 12 %, and the SE gain of 4 RRHs is 5.4 %.

(57)

0.4R 0.5R 0.6R 0.7R 2.2 2.25 2.3 2.35 2.4 2.45 2.5 2.55 2.6 2.65 2.7

Distance between Macrocell and RRH, Cell Radius R = 288.7 meters

Cell Average Spectral Efficiency (bps/Hz/cell)

SU−MIMO 1 RRH 2 RRHs 3 RRHs 4 RRHs

Figure 6.1: Comparison of spectral efficiency for different RRH locations.

0.4R 0.5R 0.6R 0.7R 4500 4600 4700 4800 4900 5000 5100

Distance between Macrocell and RRH, Cell Radius R = 288.7 meters

Energy Efficiency (bits/Joule) SU−MIMO

1 RRH 2 RRHs 3 RRHs 4 RRHs

Figure 6.2: Comparison of energy efficiency for different RRH locations.

(58)

6.3.2

Effect of Cell Architecture

Different cell structures have various impacts on the system performance. Spectral and energy efficiency comparison of different system architectures are shown in Figs. 6.4 and 6.3. We find that all the cell architectures have the same trend in both SE and EE curve. The best location for RRH to achieve the highest SE and EE is at 0.5-0.6R, where R is the cell radius. Moreover, the performance of narrow beam is the best, and the diamond shape is the worst. Because three adjacent directional antennas are facing toward the same point in diamond shape architecture, it will cause larger interference than the other two kinds of antenna pattern. As for narrow beam, it can avoid causing interference to other cells effectively, and thus it has the best system performance. When RRHs are deployed at 0.5R, both SE and EE are maximum. The spectral efficiency of narrow beam is 2.8660 (bps/Hz/cell), which is the highest. The second one is pentagon shape, which is 2.8363 (bps/Hz/cell). The diamond shape has the lowest energy efficiency, 2.7012 (bps/Hz/cell). The SE gain of the narrow beam over diamond shape is up to 7.9%, and the gain of the pentagon-shape is 5.9%. The energy efficiency of narrow beam is 5352 (bits/Joule), which is also the highest. The second one is pentagon shape, which is 5296 (bits/Joule). The diamond shape has the lowest energy efficiency, 5044 (bits/Joule). The SE gain of narrow beam over the diamond shape is up to 6.1%, and the gain of pentagon shape is 4.9%.

6.3.3

Tradeoff between Spectral and Energy Efficiency

Based on Fig. 6.5, we observe that the system throughput is growing with the number of RRHs. However, EE does not necessarily increase with the amount of RRHs. On the contrary, EE stops increasing when RRHs are up to a certain number. For instance, the system with 8 RRHs has the highest SE, but the EE is just 5279 (bits/Joule), which is lower than the EE of 4 RRHs, e.g., 5357 (bits/Joule).

(59)

0.4R 0.5R 0.6R 0.7R 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3

Distance between Macrocell and RRH, Cell Radius R = 288.7 meters

Cell Average Spectral Efficiency (bps/Hz/cell)

Narrow−beam Pentagonal Diamond

Figure 6.3: Comparison of spectral efficiency for different system architectures.

0.4R 0.5R 0.6R 0.7R 4400 4600 4800 5000 5200 5400 5600

Distance between Macrocell and RRH, Cell Radius R = 288.7 meters

Energy Efficiency (bits/Joule)

Narrow−beam Pentagonal Diamond

Figure 6.4: Comparison of energy efficiency for different system architectures.

數據

Figure 1.1: The relation between EE and SE.
Figure 2.1: Various CoMP scenarios.7
Table 2.1: Comparison of Our Work and Related Works
Figure 3.1: Cell architecture of SU-MIMO systems.
+7

參考文獻

相關文件

本次的作業 (netstat -na 部分 ) 即在觀看所有機 器上的 transport layer 連線. 本次的作業 (netstat -na 部分 ) 即在觀看所有機 器上的 transport layer

5/11 Network Address Translation and Virtual Private Network. 5/18 System configuration and

5/4 System configuration and log systems 5/11 Network Address Translation (NAT)
. and Virtual Private

– evolve the algorithm into an end-to-end system for ball detection and tracking of broadcast tennis video g. – analyze the tactics of players and winning-patterns, and hence

Dennis Ritchie Brian Kernighan Douglas McIlroy Michael Lesk Joe Ossanna.. Multitasking and multi-user OS

We shall show that after finite times of switching, the premise variable of the fuzzy system will remain in the universe of discourse and stability of the adaptive control system

Failure Mode and Effects Analysis (Failure Modes and Effects Analysis, FMEA) is used to analyze the product / system, the potential failure modes, identify potential trouble spots

After digest and analyze, extract valued factors of new product development to divide into quality, cost and time three sub system, then construct causal loop model of system