Addis Ababa University School of Graduate Studies Addis Ababa Institute of Technology

ADDIS ABABA UNIVERSITY
SCHOOL OF GRADUATE STUDIES

ADDIS ABABA INSTITUTE OF TECHNOLOGY
SCHOOL OF ELECTRICAL AND COMPUTER
ENGINEERING
Antenna Spacing Effects on Indoor

MU-MIMO Channel Capacity
By
Feyissa Endebu
Advisor:
Dr. Murad Ridwan
A Thesis submitted to Addis Ababa University in partial

fulfillment of the requirements for the degree of Masters of
science in Communication Engineering
April, 2016
ADDIS ABABA UNIVERSITY
SCHOOL OF GRADUATE STUDIES
SCHOOL OF ELECTRICAL AND COMPUTER

ENGINEERING
Antenna Spacing Effects on Indoor

MU-MIMO Channel Capacity
BY:
FEYISSA ENDEBU NURGI
ADDIS ABABA INSTITUTE OF TECHNOLOGY
APPROVAL BY BOARD OF EXAMINER
Chairman,Dept. of Graduate Committee Signature
Dr. Murad Ridwan
Advisor Signature
Internal Examiner Signature
External Examiner Signature

Dedication
Dedicated to my son Egnuma and his mom...
i
Declaration
I, the undersigned, declare that this thesis work is my original work, has not been
presented for a degree in this or any other universities, and all sources of materials
used for the thesis work have been fully acknowledged.
Feyissa Endebu
Name Signature
Place: Addis Ababa
Date of submission
This thesis has been submitted for examination with my approval as a university
advisor.
Dr. Murad Ridwan

Advisor Signature
ii
Abstract
In theory, multi-user MIMO is more immune to most of propagation limitations

plaguing single-user MIMO (SU-MIMO) systems, such as channel rank loss or an-
tenna correlation. However, in this thesis work, it has been considered the effect of
antenna spacing on indoor multi-user MIMO sum-rate channel capacity at BS (base
station) and user separation at UE (user equipment) side.
This can be analyzed by using correlation based-analytical approximation method,

and its simulation results. These simulation results were obtained using MATLAB
simulation tools. The simulation result shows the comparison of correlated MU-
MIMO sum-rate channel capacity for all scenarios considered with non-correlated
stochastic channel capacity.
Based on the result obtained it is better to use antenna spacing above 12 λ (half-
carrier wave length) to get optimum correlated sum-rate channel capacity in in-
door LOS compared to uncorrelated iid gaussian non-fading channel. In non-line of
site(NLOS) MU-MIMO BC, antenna spacing from 15 λ to 12 λ is enough to get better
sum-rate channel capacity than other correlated channel compared to uncorrelated
Rayliegh fading channel capacity. Similarly, using Ricean fading channels, it is bet-
ter to use antenna spacing above 12 λ to get optimum sum-rate channel capacity.
KEY WORDS: MU-MIMO, Indoor Scenario, Antenna Spacing, User separa-

tion, Sum-Rate Channel Capacity
iii
Acknowledgment
Foremost, I would like to thank my advisor, Dr.Murad Ridwan, for his guidance,
patience and full support during the past one year. I am grateful to him for his
encouragement and patience, especially during the time I needed these most. I am
constantly surprised by his technical intuition, and I hope I have learned a bit from
him in the way of thinking and carrying out thesis work.
My life here has been an enjoyable journey due to all my friends, brother and
all the kind people I have met, at AAiT or outside AAiT. Their characters have
enriched my own. In particular, I want to thank Tesfaye Mola and Gadissa Endebu
for their endless support in material and financial.
Last, but not the least, I want to thank all my family members in Debra Berhan
and Ambo for their support.
iv
Contents
Declaration ii
Abstract iii
Acknowledgment iv
List of Figures vii
List of Tables ix
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Multiuser Multiple-Input Multiple-Output (MU-MIMO) . . . 2
1.1.2 Multiuser Multiple-Input Multiple-Output (MU-MIMO) basics 3
1.1.3 Multi-User MIMO (MU-MIMO) Advantages . . . . . . . . . . 4
1.1.4 Challenges of Multi-User MIMO (MU-MIMO) . . . . . . . . . 5
1.2 The Research Statements of Problem . . . . . . . . . . . . . . . . . . 6
1.3 Literature Review and Related Works . . . . . . . . . . . . . . . . . . 7
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 General Objectives . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.2 Specific Objectives . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Research Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Thesis Outlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 MIMO Wireless Communication 11

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 MIMO Transmission Schemes . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Single Input Single Output (SISO) . . . . . . . . . . . . . . . 12
2.2.2 Single Input Multiple Output (SIMO) . . . . . . . . . . . . . 13
2.2.3 Multiple Input Single Output (MISO) . . . . . . . . . . . . . 13
2.2.4 Multiple Input Multiple Output (MIMO) . . . . . . . . . . . . 14
2.3 MIMO Channel Capacity . . . . . . . . . . . . . . . . . . . . . . . . . 17
v
2.3.1 Capacity of Deterministic MIMO Channel . . . . . . . . . . . 17
2.3.2 Capacity of Non-Deterministic or Random MIMO Channels . 18
2.4 MIMO Channel Models . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.1 MIMO Channel Model Classification . . . . . . . . . . . . . . 19
2.5 Statistical Properties of the Channel Matrix . . . . . . . . . . . . . . 22
2.5.1 Degrees of Freedom and Diversity . . . . . . . . . . . . . . . . 22
2.5.2 Dependency on Antenna Spacing . . . . . . . . . . . . . . . . 23
2.5.3 iid Rayleigh Fading Model . . . . . . . . . . . . . . . . . . . . 23
2.5.4 Spatial Correlation . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5.5 Singular Values and Eigenvalues . . . . . . . . . . . . . . . . . 26
2.6 Analytical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6.1 Correlation-Based Analytical Models . . . . . . . . . . . . . . 27
3 Multi-User MIMO Communication 31

3.1 Introductions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Channel State Information . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.1 Instantaneous CSI . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.2 Statistical CSI . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Antenna Spacing and Correlations . . . . . . . . . . . . . . . . . . . . 35
3.3.1 Antenna Radiation Pattern . . . . . . . . . . . . . . . . . . . 35
3.3.2 Antenna Lobe . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Grating Lobes in Antenna Arrays . . . . . . . . . . . . . . . . 37
3.4 Uplink-Multiple Access Multiuser-MIMO . . . . . . . . . . . . . . . . 38
3.5 Downlink-Broadcasting Multiuser-MIMO . . . . . . . . . . . . . . . . 40
3.6 Indoor Scenario Consideration . . . . . . . . . . . . . . . . . . . . . . 41
4 Simulation Result and Discussion 43

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 Flow Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Antenna Spatial Correlation . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Channel Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.5 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . . 46
4.5.1 Antenna Radiation Pattern . . . . . . . . . . . . . . . . . . . 46
4.6 Spatial correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5 Conclusion and Recommendation 59

5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2 Drawbacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Recommendation for Future Works . . . . . . . . . . . . . . . . . . . 60
vi
Bibliograpy 62
Appendix 65
vii
List of Figures
1.1 MU-MIMO block diagram [3] . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Block diagram of multi-user MIMO up link system [5] . . . . . . . . . 3
1.3 Block diagram of multi-user MIMO down link system [5] . . . . . . . 4
2.1 SISO model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 SIMO model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 MISO model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 MIMO model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 MIMO channel capacity with increasing number of antennas [15] . . . 16
2.6 Classification of MIMO channel and propagation models[13] . . . . . 20
2.7 Geometry of a MIMO channel with transmits and receives linear an-
tenna arrays [13] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1 Antenna Pattern Parameters [27] . . . . . . . . . . . . . . . . . . . . 37

3.2 MIMO-MAC with single antenna users . . . . . . . . . . . . . . . . . 38
3.3 MIMO-BC with M antenna base stations . . . . . . . . . . . . . . . . 40
4.1 Simulation flow chart . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.2 Antenna radiation pattern for different spacing . . . . . . . . . . . . . 47
4.3 Effects of antenna spacing on spatial correlation at BS . . . . . . . . 48
4.4 Effects of user separation in LOS on spatial correlation . . . . . . . . 49
4.5 Effects of user separation in NLOS on spatial correlation . . . . . . . 49
4.6 (4xTX, 4xRX) MIMO-BC Sum-rate channel capacity which shows
the effect of antenna spacing below 0.5λ as compared to iid Rayleigh
fading channel capacity in indoor NLOS. . . . . . . . . . . . . . . . . 50
the effect of antenna spacing above 0.5λ as compared to iid Rayleigh
fading channel capacity in indoor NLOS. . . . . . . . . . . . . . . . . 51
the effect of antenna spacing above 0.5λ as compared to Gaussian iid
channel capacity in indoor LOS. . . . . . . . . . . . . . . . . . . . . . 52
viii
the effect of antenna spacing below 0.5λ as compared to Gaussian iid
channel capacity in indoor LOS. . . . . . . . . . . . . . . . . . . . . . 53
4.10 (4xTX, 4xRX) MIMO-MAC Sum-rate channel capacity which shows
the effect average separation distance of sparsely randomly distributed
users in 25m coverage areas in and indoor NLOS. . . . . . . . . . . . 54
the effect average separation distance of randomly distributed users
in 10m coverage areas in an indoor LOS. . . . . . . . . . . . . . . . . 55
the effects antenna spacing below and above 0.5λ in indoor scenarios. 56
the effects antenna spacing below and above 0.5λ in indoor scenarios. 57
the effects random user separation in indoor scenarios . . . . . . . . . 58
ix
List of Tables
3.1 Indoor correlation matrix . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 44
x
List of Abbreviation and
Acronyms
AP Access Point
AoA Angle of Arrival
AoD Angle of Departure
AWGN Additive White Gaussian Noise
BC Broadcasting Channel
BS Base Station
CDMA Code Division Multiple Access
CSI Channel State Information
DAS Distributed Antenna System
DPC Dirty Paper Coding
DSP Digital Signal Processing
EVD Eigen Value Decomposition
FDMA Frequency Division Multiple Access
FNBW First-Null Beam Width
GSCM Geometry-based Stochastic Channel Models
HPBW Half-Power Beam Width
IEEE Institute of Electrical and Electronics Engineers
iid independent and identically distributed
LOS Line-Of-Sight
xi
MAC Multiple Access Channel
MEA Multiple Element Antenna
MIMO Multiple-Input Multiple-Output
MISO Multiple Input Single Output
MMSE Minimum Mean Square Error
MU Mobile Unit
MU-MIMO Multi-User Multiple-Input Multiple-Output
NLOS Non-Line-Of-Sight
PAS Power Azimuth Spectrum
QRD Queer Resource Decomposition
RRU Radio Resource Unit
SDMA Space Division Multiple Access
SIMO Single Input Multiple Output
SISO Single Input Single Output
SNR Signal to Noise Ratio
SU-MIMO Single-User Multiple-Input Multiple-Output
SVD Singular Value Decomposition
TDMA Time Division Multiple Access
UE User Equipment
WLAN Wireless Local Area Network
3GGP Third Generation Group Partnership
⊗ Kronecker-Product
Λ Diagonal of eigenvalue
λ Carrier wavelength
ρ Correlation coefficient
xii
φ Angle of incidence
Schur-Hadamard product
Ω Coupling matrix
A Angular power spectrum
∆ Angular spread
Σ Diagonal singular value
xiii
Chapter 1
Introduction
1.1 Motivation
Wireless communication systems have become more and more important as they
provide a flexibility and user friendly of application. Because of broadband wireless
systems benefit, from accurate channel characterization there is growing interest
in broadband wireless multiple-input multiple-output channel models and channel
characterizations [1]. Most of the broadband traffic in today’s world can be ac-
counted from indoor environments. For supporting such high rate applications in
indoor environments, and to provide a future proof infrastructure installation, opti-
cal fibers are becoming very popular.
Wireless coverage, on the other hand, offers the users with the much needed
freedom of mobility. It is expected that by 2016, [29] more than 70 percent of the
data traffic will be video traffic, but more important, it is estimated that 85 percent
of the total traffic will come from indoors locations, with others raising this number
up to 95 percent [29]. The indoor propagation scenario contains Line-Of-Sight (LOS)
and more of Non-Line Of-Sight.
Multiple-Input-Multiple-Output (MIMO) systems offer significant increase data

throughput without additional bandwidth or transmit power. It uses multiple an-
tennas at transmitter and receiver front end. Since the use of antenna arrays in
wireless communication systems provides many advantages. For example channel
capacity can be greatly increased with increasing antenna array at both links [1, 2].
1
1.1.1 Multiuser Multiple-Input Multiple-Output (MU-MIMO)
Multi-user multiple-input multiple-output (MU-MIMO) systems are the wireless

industry’s current frontier towards satisfying the increasing demand of wireless high
speed services. The optimal signal processing technique in the down-link of MU-
MIMO system is dirty-paper coding (DPC).Unfortunately, DPC is far too complex
for practical implementation, and merely serves as the theoretical benchmark [2].
Most practical signal processing techniques that have been proposed are linear,
except for popular vector-perturbation based schemes. In general, a linear processing
result in capacity losses (in Shannon sense) compared to DPC, but is still preferred
due to complexity reasons [1, 2, 3].In Figure 1:1, the base station (BS) transmit
different signal through H1 , H2 and H3 channels to M S1 , M S2 and M S3 mobile
users.However, this thesis work it was considered the effects antenna spacing on
indoor MU-MIMO channel capacity typically encountered in building at BS (base
station) antennas and user separation at the UE (user equipment) side.
Figure 1.1: MU-MIMO block diagram [3]
In a multiuser MIMO (MU-MIMO) system, a base station communicates with

multiple users. On the down-link known as the MIMO broadcast channel, the base
station sends different information streams to the users. On the up-link, the base
station receives different information from the users. Other variations of MU-MIMO
involve full or partial multi-cast of data. Note that while MU-MIMO is often dis-
cussed in the context of cellular communication, it could conceivably be used in
wireless local area networks or in wireless ad-hoc networks[3].
2
1.1.2 Multiuser Multiple-Input Multiple-Output (MU-MIMO)
basics
The Multi-User-Multiple-Input-Multiple-Output (MU-MIMO) provides a method-

ology whereby spatial sharing of channels can be achieved. This can be achieved at
the cost of additional hardware filters and antennas. But the incorporation does not
come at the expense of additional bandwidth as is the case when technologies such as
FDMA, TDMA or CDMA are used. When using spatial multiplexing in MU-MIMO
the interference between the different users on the same channel is accommodated
by the use of additional antennas, and additional processing when enables the spa-
tial separation of the different users [1, 3]. There are two scenarios associated with
Multi-user MIMO (MU-MIMO) [3, 4]:
1. Uplink - Multiple Access Channel (MAC) MU-MIMO: The development of

the MU- MIMO- MAC is based on the known single user MIMO concepts
broadened out to account for multiple users [2, 3]. The BS will coherently
detect the signals transmitted as x1 , x2 , ..., xk from K users. Then the base
station receive data as D1 , D2 , D3 , ..., DK and the received signal vector y1 ,
y2 , ..., yk together with knowledge of the CSI. When Q1 , Q2 , ..., Qk are the
transmitted signal covariance. Since MT 1 , MT 2 , ...,MT K and 1, 2, 3, ..., MR
are the transmitter and receiver antenna elements. The channel estimate can
be obtained from up link training [4].
Figure 1.2: Block diagram of multi-user MIMO up link system [5]
2. The MIMO-BC is the more challenging scenario: The optimum strategy in-
volves pre-interference cancellation techniques known as Dirty Paper Coding
(DPC) [2, 3]. The MU-MIMO on the down link is especially interesting be-
cause the MIMO sum capacity can scale with the minimum of the number of
3
base station antennas and the sum of the number of users times the number
of antennas per user. Where x1 , x2 , x3 , ..., xk and F1 , F2 , F3 , ...., Fk are the
transmitter signal vector and symbols. And where, G1 , G2 , G3 , ..., Gk and y1 ,
y2 , y3 , ..., yk are receiver devices and received signal vectors.
Figure 1.3: Block diagram of multi-user MIMO down link system [5]
This means that MU-MIMO can achieve MIMO capacity gains with a multiple
antenna base station and a bunch of single antenna mobile users. This is of particular
interest since real estate for multiple antennas is limited on small hand held devices
[4].
1.1.3 Multi-User MIMO (MU-MIMO) Advantages

Multi-user MIMO (MU-MIMO) offers some significant advantages over other
techniques [4, 5]. These are:
• MU-MIMO systems enable a level of direct gain to be obtained in a multi-

ple access capacity arising from the multi-user multiplexing schemes. This is
proportional to the number of base station antennas employed.
• MU-MIMO appears to be affected less by some propagation issues that affect

single user MIMO systems. These include channel rank loss and antenna
correlation although channel correlation still affects diversity on a per user
basis; it is not a major issue for multi-user diversity.
• MU-MIMO allows spatial multiplexing gain to be achieved at the base station

without the need for multiple antennas at the UE. This allows for the produc-
tion of cheap remote terminals the intelligence and cost is included within the
base station.
4
1.1.4 Challenges of Multi-User MIMO (MU-MIMO)
• Channel state information: in order to achieve high spatial multiplexing gain,
the BS needs to process the received signals coherently. This requires accurate
and timely acquisition of CSI. This can be challenging, especially in high
mobility scenarios [4].
• There exists multiuser interference, hence complicated interference reduction

or cancellation techniques should be used. For example, maximum likelihood
multiuser detection for up-link dirty paper coding (DPC) techniques for down
link, and interference alignment [4].
• Since several users are served on the same time-frequency resource, scheduling
schemes which optimally select the group of users depending on the pre-coding
detection schemes, CSI knowledge etc., should be considered. This increases
the cost of the system implementation [4].
• Pilot contamination: in practical cellular networks, due to the limitation of the

channel coherence interval, non-orthogonal pilot sequences have to be utilized
in different cells. Therefore, the channel estimate obtained in a given cell is
contaminated by pilots transmitted by users in other cells. This effect, called
pilot contamination, reduces the system performance [4].
5
1.2 The Research Statements of Problem
Wireless channel has an application over a wide area but its performance differs
from environment to environment (i.e. indoor and outdoor) because wireless chan-
nels feature fading, shadowing, interference, and other impairments that make the
channel unpredictable. In today’s wireless networks, there is an increasing demand
for service quality, high data rates, network coverage, and lesser processing time.
The scarcity of two fundamental resources for communications, namely, energy and
bandwidth, is a serious challenge to fulfill these demands.
Nowadays, [29] indoor wireless network users are the predominant and it is esti-
mated that by 2016, 85 percent of the total traffic will come from indoors locations in
any kind of data traffic. MU-MIMO has the potential to combine the high through-
put achievable with MIMO processing with the benefits of space division multiple
accesses (SDMA). And, in an indoor scenario there are possibilities of transmitting
base station and multiple users to be in the same building, on the same floor and
may be in the same offices; which are the headache for mobile network operator to
keep the user orthogonality in space.
Here in this thesis, a strong attention is given on the problem arrived because of
spatial correlation. Spatial correlation affects the performance of wireless network
by having an effect on the channel capacities. As a result here, to overcome the
problem, the optimization of antenna spacing at BS and user separation on UE
side is studied as an alternative means of alleviating the problem arrived because of
spatial correlation.
6
1.3 Literature Review and Related Works
Multi User MIMO is advanced MIMO system used for spectral efficiencies and
throughput optimizations. Since, multiple antennas exists at the base station and at
the UE side, it has been expected that multiple users are using single base station
or multiple base stations with multiple antenna arrays spatial multiplexing. In
indoor scenarios, it is known that the receiver and the transmitter can be in the
same building, on the same floor may be in the same offices. For this reason, in
indoor scenarios the most interesting thing is to mitigate interference due to spatial
correlation by keeping sufficient antenna spacing in indoor MU-MIMO to enhance
the sum-rate channel capacity.
The following related literatures revealed some interesting technologies and meth-
ods of mitigating interference at both transmitter and receiver sides of MU-MIMO
and indoor MU-MIMO system.
Fredrik et al. investigated indoor MU-MIMO systems in narrow corridor environ-

ments, typically encountered in office buildings, universities, hospitals, etc. Started
by performing a theoretical investigation of downlink signal processing techniques
in order to establish what channel parameters influence the performance. Then,
analyzed extensively channel measurements in order to obtain an understanding of
the behavior of the predominant parameter [6].
Robert W. Heath Jr. et al. investigated that multiuser MIMO is an effective

strategy for DAS even in the presence of out-of cell interference. It considered
a narrow band DAS channel model with independent Gaussian distributed small-
scale fading and distance-dependent path-loss. It compare centralized multiuser
MIMO zero-forcing beam-forming with distributed multiuser MIMO zero-forcing
beam-forming across all RRUs in a cell, in terms of area spectral efficiency [7].
Kyung et al. studied the transmit antenna selection in multi-user MIMO systems
with pre-coding. The optimum and reduced complexity sub-optimum antenna selec-
tion algorithms are introduced. QR-decomposition (QRD) based antenna selection
is investigated and the reason behind its sub-optimality is analytically derived. It in-
troduce the conventional QRD-based algorithm and propose an efficient QRD-based
transmit antenna scheme that is both implementation and performance efficient[8].
Pohl et al. examined the relation between antenna spacing and capacity of
MIMO channels for indoor environments using physical channel model which is a
ray tracing program. In [9], it has been confirmed also by measurements that in rich
7
scattering environments an antenna spacing below 0.5λ is to reach nearly the full
capacity predicted for multiple-antenna arrays in ideal and uncorrelated Rayleigh
fading channels.
Furthermore, this research gap has been identified when talking about the an-
tenna spacing effect at BS and user separation on UE sides in an indoor MU-MIMO
channel capacity by using correlation based analytical channel modeling which was
not done as far as dealt with all these literature and related works.
8
1.4 Objectives
1.4.1 General Objectives

The main objective of this thesis work is to find the effects of antenna spacing
at BS and user separation on UE sides on an indoor MU-MIMO channel capacities
and its contribution for further capacity enhancement.
1.4.2 Specific Objectives

The specific objectives of this thesis can be summarized as follows:
• To evaluate the effect antenna spacing at BS and user separation on MU sides

in an indoor LOS and NLOS on spatial correlations.
• To evaluate an indoor MU-MIMO channel capacity with varying antenna spac-

ing at BS and users separation on MU sides in an indoor LOS and NLOS in
both MIMO-BC and MIMO-MAC.
• To get the effects of antenna spacing at BS and user separation on MU sides

on an indoor MU-MIMO channel capacity and finding sufficient spacing of
antennas at BS.
• To reduce interference due to indoor propagation scenarios and indoor multi-

users, this means to reduce the effect of spatial correlation in indoor environ-
ments.
• To compare an indoor MU-MIMO channel capacity with varying antenna spac-

ing at BS and users separation on UE sides in an indoor LOS and NLOS in
both MIMO-BC and MIMO-MAC with i.i.d Gaussian channel capacity and
i.i.d Rayleigh fading channel capacity respectively.
9
1.5 Research Methodologies
In this thesis work, it has been evaluated the effects of antenna spacing at BS
and user separation at mobile unit sides by correlation based-analytical channel
modeling. And also obtained better antenna spacing for optimum sum-rate channel
capacity in MU-MIMO BC and effects user separation on sum-rate channel capacity
in MU-MIMO MAC. The methods used to achieve the desired objectives of this
thesis were as follows. First, related literature about MU-MIMO channel capacity
and correlation based-analytical channel modeling reviewed. Second, the spatial
correlation and indoor MU-MIMO channel capacity was combined by analytical
approach. Then, has been MATLAB simulation tool is used for the analysis of
correlation based-analytical channel modeling.
1.6 Thesis Outlines
This thesis work contains different parts. These are, Introduction part which in-
cludes the technological overview of the topic, the literature review, the statement of
problems, the objective and methodologies. Part two is about MIMO wireless com-
munication which contains MIMO schemes, MIMO channel capacity and MIMO
channel modeling. Part three is about Multiuser MIMO communication which in-
cludes channel state information, MU-MIMO broad casting, MU-MIMO multiple
access control and indoor scenario considerations. Part four contains simulation
results and discussion. And part five includes conclusion and recommendation for
future works.
10
Chapter 2
MIMO Wireless Communication
2.1 Introduction
The increase in spectral efficiency offered by MIMO systems is based on the

utilization of space (or antenna) diversity at both the transmitter and the receiver.
Due to the utilization of space diversity, MIMO systems are also referred to as
multiple-element antenna systems (MEA). With a MIMO system, the data stream
from a single user is de-multiplexed into nT separate sub-streams. The number
nT equals the number of transmit antennas. Each sub-stream is then encoded into
channel symbols. It is common to impose the same data rate on all transmitters,
but adaptive modulation rate can also be utilized on each of the sub-streams [12].
The signals are received by nR receive antennas. With this transmission scheme,
there is a linear increase in spectral efficiency compared to a logarithmic increase in
more traditional systems utilizing receive diversity or no diversity[25].
The high spectral efficiencies attained by a MIMO system are enabled by the fact
that in a rich scattering environment, the signals from each individual transmitter
appear highly uncorrelated at each of the receive antennas. When the signals are
conveyed through uncorrelated channels between the transmitter and receiver, the
signals corresponding to each of the individual transmit antennas have attained
different spatial signatures. The receiver can use these differences in spatial signature
to simultaneously and at the same frequency separate the signals that originated
from different transmit antennas[12, 25].
It has been well recognized that MIMO is a key technology to fully exploiting
multi-path propagation in broadband wireless communication systems. Spatial mul-
tiplexing gain can be obtained via the potential de-correlation between the channel
coefficients of MIMO radio channel[25].
11
However, the spatial multiplexing gain depends on the multi-path richness, since
non-fading channel doesn’t include de-correlation and neither MIMO capacity gain.
Also, channel model with zero correlation would give too optimistic results, since in
field the correlation is rarely zero. Hence, it is important to have realistic channel
model in evaluating the real MIMO performance[25].
2.2 MIMO Transmission Schemes

Depending upon number of antennas used in transmission, MIMO transmission
scheme is divided into SISO, SIMO, MISO, and MIMO for wireless communication
system as shown bellow [14].
2.2.1 Single Input Single Output (SISO)

Single input single output (SISO) is less complex and easier to make for wireless
communication system to transmit and receive signal. Assume input data stream
is S, channel is h11 and output data stream be the Y . Antenna configuration and
input output relation of SISO system is shown in Figure 2.1. The channel capacity
is poor as compare to other technique but system design is not complex.
Figure 2.1: SISO model
CSISO = B log2 det(1 + SN R) (2.1)
Where C is channel capacity, B is bandwidth of the signal, SN R is signal to noise

ratio. SISO are advantageous in terms of the simplicity. It does not require process-
ing in terms of diversity schemes. The throughput of the system depends upon the
channel bandwidth and signal to noise ratio. In some conditions, these systems are
exposed to the issues like multi-path effects[14].
12
2.2.2 Single Input Multiple Output (SIMO)
SIMO refers to the familiar wireless configuration with a single antenna at the
transmitter and multiple antennas at receiver site. Now we assume we have two
receiving signals Y1 and Y2 with different fading channel coefficient h1 and h2 with
input data stream S. Antenna configuration and input output relation of SIMO
(Receive Diversity) system is shown by Figure 2.2.
Figure 2.2: SIMO model
The channel capacity has not increased. The multiple receive antennas can help
us get a stronger signal through diversity. The SIMO channel capacity is shown,
CSIM O = Mr B log2 det(1 + SN R) (2.2)
where B is bandwidth, SNR is signal to noise ratio and Mr is the number of

antennas used at the receiver side[14].
2.2.3 Multiple Input Single Output (MISO)

MISO system has multiple antennas at the transmitter and single antennas at
receiver site. Now we assume we have two transmitting signals S1 and S2 with
different fading channel coefficient h1 and h2 with output data stream Y . Antenna
configuration and input output relation of MISO (transmit diversity) is given by
Figure 2.3.
13
Figure 2.3: MISO model
The channel capacity has not really increased because we still have to transmit
two signals at a time .The MISO capacity is,
CM ISO = Mt B log2 det(1 + SN R) (2.3)
where B is bandwidth, SNR is known as signal to noise ratio and MT is the number
of antennas used at the transmitter side[14].
2.2.4 Multiple Input Multiple Output (MIMO)
MIMO is a method of transmitting multiple data streams at the transmitter

side and also receiving multiple data streams at the receiver side. MIMO antenna
configuration describes that use of multiple transmit and multiple receive antennas
for a single user produces higher capacity, spectral efficiency and more data rates
for wireless communication. When the data rate is to be increased for a single user,
this is called single user MIMO (SU-MIMO) and when the individual streams are
assigned to various users; this is called multiuser MIMO (MU-MIMO) [14]. Antenna
configuration and input output relation of MIMO (transmit diversity) is shown by
Figure 2.4.
14
Figure 2.4: MIMO model
From the above Figure 2.4, output user data stream y = Hs + n (input output
relation of MIMO channel), where S = [S1 , S2 ,..., SM ]T is the transmitted signal
vector of order MT ×1, y = [y1 , y2 ,..., yM ]T is the received signal vector of order MR ×1
, and n = [n1 , n2 ,..., nM ]T is the Additive White Gaussian noise (AWGN) of order
MR ×1. Let us consider a MIMO system with MT transmit antennas and MR receive
antennas, denote the impulse response between the j th (j = 1, 2,..., MT ) transmit
antenna and the ith (i = 1, 2,..., MR ) receiving antenna. Then MIMO channel can be
represented using an MR by MT matrix format H is,
H = MR × MT (2.4)
where hij is a complex Gaussian random variable that models fading gain between
the ith transmit and j th receive antenna. If a signal Sj (t) is transmitted from the j th
transmitted antenna, the signal receive at the ith receive antenna. The input output
relation is [14, 15],
Mt
X
yi (t) = hi,j Si (t), i = 1, 2, ..., t, Mr (2.5)
j=1
Here we take Mt transmit and Mr receive antennas with input data stream is S and
output data stream is Y . MIMO has higher capacity as compare to other system.
The MIMO capacity is given by,
CM IM O = Mt Mr B log2 (1 + SN R) (2.6)
Where C is capacity, B is bandwidth, SNR is signal to noise ratio. Since Mt is the

number of antennas used at the transmitter side and Mr is the number of antennas
used at receiver side. Since the MIMO channel capacity increases with increasing
number of transmit and receive antennas, it can be as shown in Figure 2.5 follow.
15
Figure 2.5: MIMO channel capacity with increasing number of antennas [15]
16
2.3 MIMO Channel Capacity
MIMO channel capacity can be evaluated for the following MIMO channel matrix
cases:
1. Channel Matrix H is deterministic
2. Channel Matrix H is a random matrix
2.3.1 Capacity of Deterministic MIMO Channel

For a MIMO system of Mt transmit and Mr receive antennas received vector y
can be written as r
Ex
y= Hx + n (2.7)
Mt
Where Ex is the energy of the transmitted signals and H is time-invariant channel
matrix of Mr by Mt and n = [n1 , n2 ,..., nR ]T is a noise vector. Moreover, the
autocorrelation of transmitted vector is defined as
Rxx = E[xxH ] (2.8)
The autocorrelation of received matrix is defined as
Ryy = E[yy H ]
r ! r !
Ex Ex H H
=E Hx + n H x + nH
Mt Mt

Ex H H H
=E Hxx H + nn
Mt
(2.9)
Ex
= E(HxxH H H ) + E[nnH ]
Mt
Ex
= HE[xxH ]H H + E[nnH ]
Mt
Ex
= HQx H H + Rnn
Mt
Where H H is the complex conjugate transpose (Hermitian) of the matrix H in [14].

Now, the capacity of SISO channel can be easily be extended to MIMO capacity
in bps/Hz when Channel State information (CSI) is not known at the transmitter
which can be expressed as follows:

Ex
C = log2 det IMR + HH H (2.10)
MT N0
17
Ex
where M T
=γ is average Signal-to-Noise ratio of the receiving antenna and IMR is
identity matrix. For no CSI at the transmitter the covariance matrix Rxx is identity
matrix i.e.Qx = I. Now, when the MIMO channel capacity when Channel State
Information (CSI) is known at the transmitter in which case channel matrix H is a
full rank matrix can be written as following

max
γ H
C= tr(Qx )=MT log2 det IMR + HQx H (2.11)
MT
2.3.2 Capacity of Non-Deterministic or Random MIMO Chan-

nels
The practical MIMO channels are essentially non-deterministic and time varying.
Therefore, the channel matrix H is considered random and we assume random chan-
nel is ergodic process[13, 14]. MIMO channel capacity can be found by time average
of the deterministic channel capacity in bps/Hz which can be given as follows:

max γ
HQx H H

C=E tr(Qx )=MT log2 det IMR + (2.12)
MT
This is considered as ergodic channel capacity of the MIMO channels. In general,

MIMO channels are not independent and identically distributed (i.i.d). The channel
correlation is closely related to the capacity of MIMO channels. Here we consider the
capacity of MIMO channel when the channel gain between transmitter and receiver
are correlated[13]. Now, we can consider the following correlated Kronecker channel
model as
p p
H = Rr Hw Rt (2.13)
Where Rt is the correlation matrix and reflects the correlation between transmit
antennas.Rr is the correlation matrix reflecting the correlation between receive an-
tennas and Hw denoted the iid Rayleigh fading channel gain matrix. The diagonal
entries of Rr and Rt are constrained to be unity. Then, substituting equation (2.14)
into (2.10) we can write MIMO channel capacity as

γ p H
C = log2 det IMR + Rr Hw Rt HwH Rr2 (2.14)
MT
Condition: If MT = M = MR , Rt and Rr are full rank and SNR is high we can

approximate above relation as

γ
C ≈ log2 det Hw HwH + log2 det(Rr ) + log2 det(Rt ) (2.15)
MT
18
Equation (2.16) above shows that MIMO channel capacities have been reduced
due to correlation between transmit and receive antennas by an amount equal
to log2 det(Rr ) + log2 det(Rt ). Increase in correlation usually causes a decrease in
SNR[13, 14].
2.4 MIMO Channel Models
A set of spatial channel model parameters are specified that have been devel-
oped to characterize the particular features of MIMO radio channels. SISO channel
models provide information on the distributions of signal power level and Doppler
shifts of received signals. MIMO channel models, which are based on the classical
understanding of multi-path fading and Doppler spread, incorporate additional con-
cepts such as Angular Spread, Angle of Arrival, Power-Azimuth-Spectrum (PAS),
and the antenna array correlation matrices for the transmitter (Tx ) and receiver
(Rx ) combinations[13].
2.4.1 MIMO Channel Model Classification
A variety of MIMO channel models, many of them based on measurements. The

proposed model can be classified in various ways. A potential way of distinguishing
the individual model is with regard to type of channel that is being considered,
i.e., narrow (flat fading) vs. broadband (frequency-selective) models, time-varying
vs. time structure. In contrast, broadband (frequency-selectivity) channels require
additional modeling of the multi-path channel characteristics. With time-varying
channels, one additionally requires a model for the temporal channel evolution ac-
cording to certain Doppler characteristics[13].
19
Figure 2.6: Classification of MIMO channel and propagation models[13]
The fundamental distinction is between physical channel models and correlation

based-analytical channel model models. Physical channel models characterize an
environment on the basis of electromagnetic wave propagation by describing the
double directional multi-path propagation between the location of the transmit (Tx )
array and the location of the receive (Rx ) array. They explicitly model wave propa-
gation parameters like the complex amplitude, AoD, AoA, and delay of MPC. More
sophisticated models also incorporate polarization and time variation. Depending
on the chosen complexity, physical model allow for an accurate reproduction of radio
propagation[13].
Physical models are independent of antenna configurations (antenna pattern,

number of antennas, array geometry, polarization, mutual coupling) and system
bandwidth. Physical MIMO channel models can further be split into determinis-
tic models, geometry-based stochastic models, and non-geometric stochastic mod-
els. Deterministic models characterize the physical propagation parameters in com-
pletely deterministic manner. With geometry-based stochastic channel models (GSCM),
the impulse response is characterized by the laws of wave propagation applied to
specific Tx , Rx and scatterer geometries, which are chosen in a stochastic (random)
manner. In contrast, non-geometric stochastic models describe and determine phys-
ical parameters (AoD, AoA, delay, etc.) in a completely stochastic way by prescrib-
ing underlying probability distribution functions without assuming an underlying
geometry[13].
20
In contrast to physical models, analytical channel models characterize the im-
pulse response (equivalently, the transfer function) of the channel between the in-
dividual transmit and receive antennas in a mathematical/analytical way without
explicitly accounting for wave propagation. The individual impulse responses are
subsumed in a (MIMO) channel matrix. Analytical models are very popular for syn-
thesizing MIMO matrices in the context of system and algorithm development and
verification.Analytical models can be further subdivided into propagation-motivated
models and correlation-based models. The first subclass models the channel ma-
trix via propagation parameters. Correlation-based models characterize the MIMO
channel matrix statistically in terms of the correlations between the matrix en-
tries. Popular correlation-based analytical channel models are the Kronecker model
[7, 8, 9, 10] and the Weichselberger model [11].
For the purpose of comparing different MIMO systems and algorithms, various
organizations defined reference MIMO channel models which establish reproducible
channel conditions. With physical models this means to specify a channel model, ref-
erence environments, and parameter values for these environments. With analytical
models, parameter sets representative for the target scenarios need to be prescribed.
Examples for such reference models are the ones proposed within 3GGP, COST 259,
COST 273, IEEE 802.16a, e, and IEEE 802.11n [7, 8, 9].
21
2.5 Statistical Properties of the Channel Matrix
2.5.1 Degrees of Freedom and Diversity

Given the statistical model, one can quantify the spatial multiplexing capability
of a MIMO channel [13]. With probability 1, the rank of the random matrixH a is
given by
rank(H a ) = min(Nr , Nc ) (2.16)
where Nr is number of non-zero rows and Nc is number of non-zero columns. This

yields the number of degrees of freedom available in the MIMO channel. The number
of non-zero rows and columns depends in turn on two separate factors:
• The amount of scattering and reflection in the multi-path environment. The

more scatterers and reflectors there are, the larger the number of non-zero
entries in the random matrix H a , and the larger the number of degrees of
freedom.
• The lengths Lt and Lr of the transmit and receive antenna arrays. With small
antenna array lengths, many distinct multi-paths may all be lumped into single
resolvable path. Increasing the array apertures allows the resolution of more
paths, resulting in more non-zero entries of H a and an increased number of
degrees of freedom.
In a slow fading environment, another important parameter is the amount of diver-

sity in the channel. This is the number of independent channel gains that have to
be in a deep fade for the entire channel to be in deep fade. In the angular domain
MIMO model, the amount of diversity is simply the number of non-zero entries in
Ha .
Note that channels that have the same degrees of freedom can have very different
amounts of diversity. The number of degrees of freedom depends primarily on the
angular spreads of the scatters/reflectors at the transmitter and at the receiver,
while the amount of diversity depends also on the degree of connectivity between
the transmit and receive angles. In a channel with multiple-bounced paths, signals
sent along one transmit angle can arrive at several receive angles. Such a channel
would have more diversity than one with single-bounced paths with signal sent along
one transmit angle received at a unique angle, even though the angular spreads may
be the same [13].
22
2.5.2 Dependency on Antenna Spacing
Focusing on the case of critically spaced antennas (i.e., antenna separations ∆t

and ∆r are half the carrier wavelength). Now the critical question is that; What is
the impact of changing the antenna separation on the channel statistics? To answer
this question, we vary the antenna separation at BS, or equivalently the number
of antenna elements. Placing antennas more sparsely reduces the resolution of the
antenna array and can reduce the number of degrees of freedom and the diversity of
the channel. Placing the antennas more densely adds spurious basis vectors which
do not correspond to any physical directions, and does not add resolvability. In
terms of the angular channel matrix H a , this has the effect of adding zero rows and
columns; in terms of the spatial channel matrix H, this has the effect of making the
entries more correlated [13].
Increasing the antenna separation within a given array length L does not increase
the number of degrees of freedom in the channel. What about increasing the antenna
separation while keeping the number of antenna elements n the same? This question
makes sense if the system is hardware-limited rather than limited by the amount
of space to put the antenna array in. Increasing the antenna separation this way
reduces the beam width of the n angular basis beam forming patterns but also
increases the number of main lobes in each.If the scattering environment is rich
enough such that the received signal arrives from all directions, the number of non-
zero rows of the channel matrix H a is already n, the largest possible, and increasing
the spacing does not increase the number of degrees of freedom in the channel [13].
On the other hand, if the scattering is clustered to within certain directions,

increasing the separation makes it possible for the scattered signal to be received in
more bins, thus increasing the number of degrees of freedom. In terms of the spatial
channel matrix H, this has the effect of making the entries look more random and
independent. At a base-station on a high tower with few local scatterers, the angular
spread of the multi-paths is small and therefore one has to put the antennas many
wavelengths apart to de-correlate the channel gains [13, 14].
2.5.3 iid Rayleigh Fading Model
A very common MIMO fading model is the i.i.d. Rayleigh fading model: the
entries of the channel gain matrix H[m] are independent, identically distributed
and circular symmetric complex Gaussian. Since the matrix H[m] and its angular
23
domain representation H a [m] are related by
H a [m] = Ur∗ H[m]Ut (2.17)
And Ur and Ut are fixed unitary matrices; this means that Ha should have the same
i.i.d. Gaussian distribution as H. Thus, using the modelling approach described
here, we can see clearly the physical basis of the i.i.d Rayleigh fading model, in
terms of both the multipath environment and the antenna arrays. There should be
a significant number of multi-paths in each of the resolvable angular bins, and the
energy should be equally spread out across these bins. This is the so called richly
scattered environment. If there are very few or no paths in some of the angular
directions, then the entries in H will be correlated. Moreover, the antennas should
be either critically or sparsely spaced. If the antennas are densely spaced, then some
entries of H a are approximately zero and the entries in H itself are highly correlated.
However, by a simple transformation, the channel can be reduced to an equivalent
channel with fewer antennas which are critically spaced [13, 14].
Compared to the critically spaced case; having sparser spacing makes it easier
for the channel matrix to satisfy the i.i.d. Rayleigh assumption. This is because
each bin now spans more distinct angular windows and thus contains more paths,
from multiple transmit and receive directions. This substantiates the intuition that
putting the antennas further apart makes the entries of H less dependent. On
the other, if the physical environment already provides scattering in all directions,
then having critical spacing of the antennas is enough to satisfy the i.i.d. Rayleigh
assumption.
Due to the analytical tractability, we will use the i.i.d. Rayleigh fading model
quite often to evaluate performance of MIMO communication schemes, but it is
important to keep in mind the assumptions on both the physical environment and
the antenna arrays for the model to be valid [13].
2.5.4 Spatial Correlation

The originality of MIMO techniques lies in the exploitation of the spatial or
double-directional structure of the channel. This structure is also largely responsible
for the performance of MIMO systems. Therefore, it does make sense to introduce
a way to characterize the spatial properties of a multi-antenna channel, or more
specifically, the space-only correlation of the channel. For a MIMO channel with a
limited number of antennas, we can switch from the double-directional notation to
24
a matrix notation, and therefore define the spatial correlation matrix as
R = ε(vec(H H )vec(H H )H ) (2.18)
This matrix is a nt nr ×nt nr positive semi-definite Hermitian matrix, which describes

the correlation between all pairs of transmit-receive channels. Depending on which
channel pairs are chosen, several correlations can be defined:
• If both channels share the same transmit and receive antennas m and n,
ε(H(n, m)H(n, m)∗ ) Simply represents the average energy of the channel be-
tween antenna m and antenna n.
(nq)
• If both channels share the same transmit antenna, rm = ε(H(n, m)H(q, m)∗ )
represents the receive correlation between channels originating from transmit
antenna m and impinging upon receive antennas n and q.
(mp)
• If both channels share the same receive antenna, tn = ε(H(n, m)H(n, p)∗ )
represents the transmit correlation between channels originating from transmit
antennas m and p and arriving at receive antenna n.
• If both channels originate from different transmit antennas, and arrive at differ-
ent receive antennas, ε(H(n, m)H(q, p)∗ ) is simply defined as the cross-channel
correlation between channels (m, n) and (q, p).
The facts that all these covariance are correlations between channels, despite the
fact that some are called transmit or receive antenna correlations for simplicity. It
is also convenient to define transmit and receive correlation matrices Rt and Rr as
Rt = ε(H H H) (2.19)
Rr = ε(HH H ) (2.20)
For homogeneous channels, the correlation coefficient between two individual chan-
nels is directly related to the joint angular power spectrum A(Ωt , Ωr ), while transmit
and receive correlation is related to At (Ωt ) respectively (Ar (Ωr )). As an example,
assuming 2-D propagation, the transmit correlation between two antennas can be
written as Z 2π
t= ejφt (θt ) At (θt )dθt (2.21)
0
where, for convenience, the transmit azimuth power spectrum with respect to the
relative azimuth.Two extreme cases are worth mentioning here. The first case corre-
sponds to a very rich scattering environment around the transmitter with a uniform
1
distribution of the energy, so that At (θt ) ≈ 2π . The transmit correlation then reads
25
as Z 2π
1
t= ejΦt (θt ) At (θt )dθt
2π 0
Z 2π
1 dt
(2.22)
= ej2π( λ )cos(θt ) dθt
2π 0
2πdt
= J0 ( )
λ
The transmit correlation only depends on the spacing between the two considered
antennas. The second extreme case occurs when scatterers around the transmit
array are concentrated along a narrow direction θ(t,0) , i.e. at (θt ) → δ(θt − θ(t,0) ).
Such channel is also known as degenerate and causes a very high transmit correlation
approaching one.
dt
t = ejΦt (θ(t,0) ) = ej2π( λ ) cos(θ(t,0) ) (2.23)
Interestingly, the scattering direction is directly related to the phase of the transmit
correlation. Using the finite scatterer representation, and assuming that all paths
are independent and have the same average power normalized to one, the antenna
covariance are directly related to the steering vectors as
∗ T
Rt = (At AH T
t ) = At At (2.24)
∗ T
Rr = (Ar AH T
r ) = Ar Ar (2.25)
Where Ar and At represent the nr × ns, r and nt × ns, t matrices whose columns
are the steering vectors related to the directions of each path observed at Rx and
Tx (these matrices are in general not unitary).It is straight forward to prove that
the first equation above is indeed equivalent to (2.19) in this case. Finally, it is
interesting to consider the case when the energy spreading is very large at both
sides and when the antenna inter-element spacing dt and dr are sufficiently large.
In such scenarios, the various elements of H become uncorrelated, and R becomes
diagonal[13, 14].
2.5.5 Singular Values and Eigenvalues

The channel matrix H is not always full rank. Denoting by r(H) the rank of H,
the singular value decomposition (SVD) of the nr × nt channel matrix reads as
H = UH ΣH VH (2.26)
Where UH and VH are nr × r(H) and nt × r(H) unitary matrices, and
ΣH = diag(σ1 , σ2 , ..., σ(r(H) ) (2.27)
26
is the diagonal matrix containing the ordered singular values of H. Denoting by
n the minimum of nt , nr , the matrix W = HH H (for nt > nr ) or W = H H H
(for nt < nr ) is an n × n positive semi-definite Hermitian matrix, whose eigenvalue
decomposition (EVD) is given by
W = Uw Λw UwH (2.28)
WhereΛw = diag(λ1 , λ2 , . . . , λn ) contains the eigenvalues of W (i.e. the squared

singular values of H, r(H) of which are non-zero and equal to λ1 , . . . , λ(r(H)) =
2
(σ12 , σ22 , . . . , σ(r(H)) ). For notational convenience, we always write λk ≡ λk (W )[13].
2.6 Analytical Models
2.6.1 Correlation-Based Analytical Models

Various narrowband analytical models are based on a multivariate complex Gaus-
sian distribution of the MIMO channel coefficients (i.e., Rayleigh or Ricean fading).
The channel matrix can be split into a zero-mean stochastic part Hs and a purely
deterministic part Hd . The Ricean MIMO channel HRicean can be modeled as the
sum of a constant HLOS and a variable Rayleigh component caused by scattering,
r r
Kr jΦ0 1
HRicean = e HLOS + HRayleigh (2.29)
1 + Kr 1 + Kr
Figure 2.7: Geometry of a MIMO channel with transmits and receives linear antenna
arrays [13]
Where Kr is the Ricean factor, Φ0 is the phase shift of the signal due to
27
propagation from a transmit antenna element to a receive antenna element, and
HRayleigh = Hw . Given a MIMO system with a uniform linear array of Nt transmit
antennas and a uniform linear array of Nr receive antennas, as shown in Figure 2.7,
when the distance R between the two arrays are very large, the HLOS matrix can
be derived as[13, 14].
 
1 ejθ . . . ej(Nt −1)θ
 e−jθ 1 . . . ej(Nt −2)θ 
 
HLOS = dr dTt = .. .. .. ..  (2.30)

 . . . .


−j(Nr −1)θ
e e−j(Nt −2)θ . . . 1
Where dr and dt are transmit and receive steering vectors, and θ is the phase shift
between two adjacent array elements. The second equality holds only when transmit
and receive antennas are almost parallel. The existence of a Ricean component
reduces the multipath richness. For a given SNR, the Ricean channel has a capacity
lower than that of the Rayleigh fading channel, but is higher than that of the AWGN
channel; it approximates the capacity of the Rayleigh fading channel as Kr → 0,
and that of the AWGN channel as Kr → 1. However, for a given transmit power,
the Ricean channel introduces a higher SNR, since there are no obstructions. For
simplicity, we thus assume Kr = 0 i.e. HRicean = HRayleigh . In its most general form,
the zero-mean multivariate complex Gaussian distribution of h = vecH is given by
1 −1
f (h) = exp(−hH RH h) (2.31)
(π nk detRH )
The mk ×mk matrix RH = E[hhH ] is known as full correlation matrix and describes
the spatial MIMO channel statistics. It contains the correlations of all channel
matrix elements. Realizations of MIMO channels with distribution (2.32) can be
obtained by
1
H = unvec(h), h = RH2 g (2.32)
1 1 H
Here, RH2 denotes an arbitrary matrix square root (i.e., any matrix satisfying RH2 RH2 =
RH ) and g is a mk × 1 vector with i.i.d. Gaussian elements with zero mean and
unit variance. Note that direct use of (2.33) in general requires full specification of
RH which involves (mk)2 real-valued parameters. To reduce this large number of
parameters, several different models were proposed that impose a particular struc-
ture on the MIMO correlation matrix. Some of these models will next be briefly
reviewed[13].
28
The iid Model
The simplest analytical MIMO model is the i.i.d. model (sometimes referred to
as canonical model). Here RH = ρI, i.e., all elements of the MIMO channel matrix
H are uncorrelated (and hence statistically independent) and have equal variance
ρ. Physically, this corresponds to a spatially white MIMO channel which occurs
only in rich scattering environments characterized by independent MPCs uniformly
distributed in all directions. The i.i.d. model consists just of a single parameter (the
channel power ρ) and is often used for theoretical considerations like the information
theoretic analysis of MIMO systems[8, 9, 10].
The Kronecker Model
The so-called Kronecker model was used in [19, 20, 21] for capacity analysis
before being proposed by [22] in the framework of the European Union SATURN
project . It assumes that spatial Tx and Rx correlations are separable, which is
equivalent to restricting to correlation matrices that can be written as Kronecker
product
RH = RT x ⊗ RRx (2.33)
with Tx and Rx correlation matrices RT x = E[H H H], RR x = E[HH H ], respectively.

It can be shown that under the above assumption, (2.34) simplifies to the Kronecker
model
p p p
h = (RT x ⊗ RRx )g → H = RRx G RT x (2.34)
with G = unvec(g) an i.i.d. unit-variance MIMO channel matrix. The model re-
quires specification of the Tx and Rx correlation matrices, which amounts to n2 + m2
real parameters (instead of n2 m2 ).The main restriction of the Kronecker model is
that it enforces a separable AoD-AoA spectrum [20], i.e., the joint AoD-AoA spec-
trum is the product of the AoD spectrum and the AoA spectrum. Note that the
Kronecker model is not able to reproduce the coupling of a single AoD with a single
AoA, which is an elementary feature of MIMO channels with single-bounce scat-
tering. Nonetheless, the model (2.34) has been successfully used for the theoretical
analysis of MIMO systems and for MIMO channel simulation. Furthermore, it al-
lows for independent array optimization at Tx and Rx . These applications and its
simplicity have made the Kronecker model quite popular.
29
The Weichselberger Model
The Weichselberger model [20, 22] aims at obviating the restriction of the Kro-
necker model to separable AoA-AoD spectra that neglects significant parts of the
spatial structure of MIMO channels. Its definition is based on the eigenvalue de-
composition of the Tx and Rx correlation matrices, RT x = UT x ΛT x UTHx , RRx =
H
URx ΛRx URx . Here, UT x and URx are unitary matrices whose columns are the eigen-
vectors of RT x and RRx , respectively, and ΛT x and ΛRx are diagonal matrices with
the corresponding eigenvalues. The model itself is given by
H = URx (ΩG)UTTx (2.35)
where G is again an n × m i.i.d. MIMO matrix, denotes the Schur-Hadamard

product (element-wise multiplication), and Ω is an n × m coupling matrix whose
(real-valued and non-negative) elements determine the average power coupling be-
tween the Tx and Rx eigenmodes. This coupling matrix allows for joint modeling
of the Tx and Rx channel correlations. We note that the Kronecker model is a spe-
cial case of the Weichselberger model obtained with the rank-one coupling matrix
Ω = λRx λT x , where λT x and λRx are vectors containing the eigenvalues of the Tx and
Rx correlation matrix, respectively. The Weichselberger model requires specification
of the Tx and Rx eigenmodes UT x and URx and of the coupling matrix Ω. In general,
this amounts to n(n − 1) + m(m − 1) + nm real parameters [20, 22].
We emphasize, however, that capacity (mutual information) and diversity order

of a MIMO channel are independent of the Tx and Rx eigenmodes; hence, their
analysis requires only the coupling matrix Ω (nm parameters). In particular, the
structure of Ω determines which MIMO gains (diversity, capacity, or beamforming
gain) can be exploited.
30
Chapter 3
Multi-User MIMO
Communication
3.1 Introductions
In this chapter, we focus to multiuser channels and study the role of multiple
antennas in both the uplink (many-to-one) and the downlink (one-to-many). In ad-
dition to allowing spatial multiplexing and providing diversity to each user, multiple
antennas allow the base-station to simultaneously transmit or receive data from mul-
tiple users. Again, this is a consequence of the increase in degrees of freedom from
having multiple antennas [18]. Independent data streams are sent at the different
transmit antennas, and no cooperation across transmit antennas is needed.
Equating the transmit antennas with users, these receiver structures can be di-
rectly used in the uplink where the users have a single transmit antenna each but
the base-station has multiple receive antennas; [18] this is a common configuration
in cellular wireless systems. It is less apparent how to come up with good strategies
for the downlink, where the receive antennas are at the different users; thus the
receiver structure has to be separate, one for each user. However, as will see, there
is an interesting duality between the uplink and the downlink, and by exploiting
this duality, one can map each receive architecture for the uplink to a corresponding
transmit architecture for the downlink. In particular, there is an interesting precod-
ing strategy, which is the transmit dual to the receiver-based successive cancellation
strategy.
It has been shown that time division multiple access (TDMA) systems cannot
achieve a linear increase of the sum-rate capacity of multi-user (MU) MIMO systems
in the number of transmit antennas. The solution to this problem is to serve user’s
31
simultaneously using space-division multiple access (SDMA). The information the-
oretic results in [13, 14, 15, 16] and [17] have shown that it is necessary to use some
kind of Costa’s ”dirty-paper” coding (DPC) or Tomlinson-Harashima precoding to
reach the sum capacity of a multi-user MIMO downlink system. DPCs achieve the
maximum sum rate of the system and provide the maximum diversity order.
The sum rate capacity of the multi-user MIMO uplink system is achieved via an
MMSE receiver with successive interference cancellation. The sum-rate capacity of a
downlink multi-user MIMO system employing DPC and an uplink multi-user MIMO
system employing successive interference cancellation is at most min(NT , K) times
larger than the maximum achievable sum rate capacity of a system using TDMA.
Motivated by small size and low power consumption of user terminals, we focus on
the base station antenna spacing and user separations. This means that one user
will not be aware of other users sharing the same time and frequency resources and
that the base station will have the task of reducing the multi-user interference by
keeping sufficient antenna spacing[18].
In this chapter we will address the problem of generalized designs and approxi-
mations of the spatial correlation and correlation matrices in an indoor multi-user
MIMO communication system. The focus will be put on the multi-user MIMO
downlink (MIMO-BC) and uplink (MIMO-MAC) system, since in this case most
of the solutions for the downlink can be applied on the uplink in a straightforward
way. Another reason for this is that by using the MU-MIMO antenna spacing and
user separation on both the uplink and the downlink we can reduce the complexity
at the base station. Moreover, having in mind the complexity of DPCs and their
inability to combine instantaneous and long-term channel state information at the
transmitter for precoding, we give preference in our investigations to the effects of
antenna spacing at base station and user separations on the capacity of MU-MIMO
in both Line of Site (LOS) and Non-Line of Site (NLOS) indoor environment.
Channel state information at the transmitter allows us to exploit the benefits of

having multiple antennas at the base station and the user terminals to the maximum.
Channel state information can be acquired at the transmitter either if a feedback
channel is present or when the transmitter and receiver operate in time division
duplex (TDD) so that time-invariant MIMO channel transfer functions is the same
in both ways. The cost of acquiring the channel state information at the transmitter
is much lower in a TDD system, where it is possible to exploit the estimated uplink
channel for the downlink transmission due to the reciprocity principle than in a
frequency division duplex (FDD) system, where we have to rely on the feedback of
the channel state information[18].
32
On the downlink the base station will use any channel state information available
to mitigate or ideally completely eliminate multi-user interference through linear or
nonlinear (DPC or THP) precoding, which leads to significant information rate
gains. The user terminal estimates the effective channel and transmits data in the
next uplink frame. The effective channel is equal to the combined network channel
after the precoding at the base station.
However, on the uplink the base station has the possibility to use successive
interference cancellation, so the effective channel on the uplink that includes the
spatial processing does not have to be the same as on the downlink. The diversity
gains of MIMO are more desirable than spatial multiplexing gains if we take into
account the limited power available at the user terminal and therefore it is enough
that the base station transmits to the user terminal only the dominant singular
vector of the uplink effective channel[18].
3.2 Channel State Information
In wireless communication, channel state information (CSI) simply represents

the properties of a communication link between the transmitter and receiver. The
CSI describes how a signal propagates from the transmitter to the receiver and
represents the combined effect of, for example, scattering, fading, and power decay
with distance. The CSI makes it possible to adapt transmissions to current channel
conditions, which is crucial for achieving reliable communication with high data
rates in multi antenna systems [3].
The channel state information (CSI) at the transmitter is vital in MIMO systems
in order to increase the transmission rate, to enhance coverage, to improve spectral
efficiency and to reduce receiver complexity [3]. The CSI is usually estimated at the
receiving end and then quantized and fed back to the transmitting side. Basically
there are two ways that the transmitter can obtain CSI from the receiving end. The
transmitter and receiver can have different CSI.
There are basically two levels of CSI, namely instantaneous CSI and statistical
CSI. The following section describes both, the instantaneous and statistical CSI.
3.2.1 Instantaneous CSI

Instantaneous CSI is also known as short-term CSI. Instantaneous CSI means
that the current conditions of the channel are known, which can be viewed as know-
33
ing the impulse response of a digital filter [16]. This gives an opportunity to adapt
the transmitted signal to the impulse response and thereby optimize the received
signal for spatial multiplexing or to achieve low bit error rates.
3.2.2 Statistical CSI
Statistical CSI is also known as long-term CSI. Statistical CSI means that a
statistical characterization of the channel is known. This description can include the
type of fading distribution, the average channel gain, the line-of-sight component,
and the spatial correlation [16].As with instantaneous CSI, this information can be
used for transmission optimization.
The CSI acquisition is practically limited by how fast the channel conditions
are changing. In fast fading systems where channel conditions vary rapidly under
the transmission of a single information symbol, only statistical CSI is reasonable.
On the other hand, in slow fading systems instantaneous CSI can be estimated with
reasonable accuracy and used for transmission adaptation for some time before being
outdated. In practical systems, the available CSI often lies in between these two
levels; instantaneous CSI with some estimation/quantization error is combined with
statistical information [16].
The capacity of a MIMO (multi-input multi-output) channel is influenced by the

degree of CSI (channel-state information) available to both transmitter and receiver
[2]. In most instances of multi-antenna communication, the receiver can accurately
track the instantaneous state of the channel from pilot signals that are typically
embedded within the transmissions. In terms of CSI at the transmitter, on the
other hand, several scenarios are possible: In a narrowband flat-fading channel with
multiple transmit and receive antennas(MIMO), the system is modeled as
y = H.x + n (3.1)
where y and x are receive and transmit vectors, respectively. H and n are the channel
matrix and the noise vector, respectively. Let us suppose the noise is modeled as
circular symmetric complex normal with
n ∼ CN (0, S) (3.2)
where the mean value is zero and the noise covariance matrix S is known.
• Instantaneous CSI Ideally, the channel matrix H is known perfectly. Due to
34
channel estimation errors, the channel information can be represented as
V ec(Hestimate ) ∼ CN (V ec(H), Rerror ) (3.3)
where Hestimate is the channel estimate and Rerror is the estimation error covari-
ance matrix. The vectorization vec() was used to achieve the column stacking
of H, as multivariate random variables are usually defined as vectors.
• Statistical CSI In this case, the statistics of H are known. In a Rayleigh

fading channel, this corresponds to knowing that
V ec(H) ∼ CN (0, R) (3.4)
for some known channel covariance matrix R. Correlation matrix R is the

expectation of vectorized H channel matrix with its Hermitian conjugate [16].
R = E[vec(H)vec(H)† ] (3.5)
3.3 Antenna Spacing and Correlations
Antennas are a very important component of communication systems. By defini-

tion, an antenna is a device used to transform an RF signal traveling on a conductor
into an electromagnetic wave in free space. Antennas demonstrate a property known
as reciprocity, which means that an antenna will maintain the same characteristics
regardless if it is transmitting or receiving. Most antennas are resonant devices,
which operate efficiently over a relatively narrow frequency band. An antenna must
be tuned to the same frequency band of the radio system to which it is connected;
otherwise the reception and the transmission will be impaired. When a signal is fed
into an antenna, the antenna will emit radiation distributed in space in a certain
way. A graphical representation of the relative distribution of the radiated power in
space is called a radiation pattern [27].
3.3.1 Antenna Radiation Pattern
The radiation pattern or antenna pattern is the graphical representation of the

radiation properties of the antenna as a function of space [27]. That is, the antennas
pattern describes how the antenna radiates energy out into space (or how it receives
energy). It is important to state that an antenna radiates energy in all directions,
at least to some extent, so the antenna pattern is actually three-dimensional. It is
35
common, however, to describe this 3D pattern with two planar patterns, called the
principal plane patterns.
These principal plane patterns can be obtained by making two slices through the
3D pattern through the maximum value of the pattern or by direct measurement. It
is these principal plane patterns that are commonly referred to as the antenna pat-
terns. In discussions of principal plane patterns or even antenna patterns, you will
frequently encounter the terms azimuth plane pattern and elevation plane pattern.
The term azimuth is commonly found in reference to the horizon or the horizontal
whereas the term elevation commonly refers to the vertical. When used to describe
antenna patterns, these terms assume that the antenna is mounted (or measured)
in the orientation in which it will be used [27].
3.3.2 Antenna Lobe
Any given antenna pattern has portions of the pattern that are called lobes. A
lobe can be a main lobe, a side lobe or a back lobe and these descriptions refer to
that portion of the pattern in which the lobe appears. In general [27], a lobe is any
part of the pattern that is surrounded by regions of relatively weaker radiation. So
a lobe is any part of the pattern that sticks out and the names of the various types
of lobes are somewhat self-explanatory.
Radiation Lobe is a clear peak in the radiation intensity surrounded by regions

of weaker radiation intensity. Main Lobe (major lobe, main beam) is radiation lobe
in the direction of maximum radiation. Minor Lobe is any radiation lobe other
than the main lobe. Side Lobe is a radiation lobe in any direction other than the
direction(s) of intended radiation. Back Lobe is the radiation lobe opposite to the
main lobe. Half-Power Beamwidth (HPBW) is the angular width of the main beam
at the half-power points.First Null Beamwidth (FNBW) is angular width between
the first nulls on either side of the main beam [27].
36
Figure 3.1: Antenna Pattern Parameters [27]
3.3.3 Grating Lobes in Antenna Arrays
Another phenomenon associated with phased array probes is the generation of

unwanted grating lobes or side lobes, two closely related phenomena caused by sound
energy that spreads out from the transducer at angles other than the primary path.
This phenomenon is not limited to phased array systems. This unwanted lobe also
occurs with conventional transducers as element size increases. These unwanted ray
paths can reflect off surfaces in the test piece and cause spurious indications on an
image. The amplitude of grating lobes is significantly affected by pitch size, the
number of elements, frequency, and bandwidth.
Grating lobes will occur whenever the size of individual elements in an array is
equal to or greater than the wavelength, and there will be no grating lobes when
element size is smaller than half a wavelength [27]. (For element sizes between one-
half and one wavelength, the generation of grating lobes will depend on the steering
angle.) Thus the simplest way to minimize grating lobes in a given application is
to use a transducer with a small pitch. Specialized transducer design incorporating
subdicing (cutting elements into smaller elements) and varying element spacing will
also reduce unwanted lobes.
37
3.4 Uplink-Multiple Access Multiuser-MIMO
Under this topic, we consider a K-user MIMO system over a flat fading MAC
with M antennas at BS and single antenna at each UE. Denote HK and XK the
channel matrix and the transmitted vector of UE K. The received vector from K
user y in [2, 18] is
Xk
y= HK XK + n (3.6)
k=1
Where n is a vector of complex additive white Gaussian noise (AWGN) samples

with zero mean and unit variance. HK in (3.6) are independent among different
UE’s and perfectly known at both the transmitter and the receiver.
Figure 3.2: MIMO-MAC with single antenna users
Following the common Kronecker channel correlation model for indoor NLOS,
we can write HK as
p p
HK = RRX HW RT X (3.7)
Where HW is a Rayleigh fading matrix whose entries is independent and identically

distributed complex circular symmetric Gaussian variables with zero mean and unit
variance. And RT X , RRX in (3.7) are semi-positive definite matrices that charac-
terize correlation effects of antennas at the transmitter and the receiver respectively
for the link between K-UE and the BS. Following [2, 18] the common Kronecker
channel correlation model for indoor LOS, we can write HK as
p p
HK = RRX Hz RT X (3.8)
38
Where Hz a Gaussian matrix is whose entries is independent and identically dis-
tributed with zero mean and unit variance. And similarly, RRX and RRX in (3.8)
are semi-positive definite matrices that characterize correlation effects of antennas
at the transmitter and the receiver respectively for the link between K-UE and the
BS. Where the sum-rate capacity in MACs under individual power constraints for
each channel can be formulated in [see: Appendix-A] as
K
!
X
max ∗
CM AC = tr( K
P log2 det IM + HK QK HK (3.9)
k=1 QK )≤PK
K=1
P
K
Subject to tr k=1 Q K ≤ PK , ∀K .
∗
where QK = E[XK XK ] and PK are input covariance matrix and the maximum
transmitted power of K-UE, respectively. It is well known that the spatial correlation
between two adjacent identical antennas due to the arrival of a single plane wave
can be approximated as in (2.23) a Jakes correlation function of its incident angle,
Φ, as
d
ρ(d) = e−j2π λ cos(Φ) (3.10)
where d is the antenna separation, λ is the carrier frequency wavelength and Φ

is the angle of departure (AOD) at the transmitter and angle arrival (AOA) at
the receiver spread [18]. And assumed to be 45 degree angle of departure (AOD)
at the transmitter and angle arrival (AOA) at the receiver spread in LOS and 10
degree angle of departure (AOD) at the transmitter and 22.5 degree angle of arrival
(AOA) at the receiver spread in NLOS. As we can see in (3.10) the correlation
coefficient is the function of antenna separation only. The matrix RT X ,RRX is
complex exponential correlation model characterized by
 
1 ρ∗ . . . ρ∗(N −1)
ρ 1 . . . ρ∗(N −2) 
 

RT x , RRx = .. .. .. ..  (3.11)

 . . . . 

ρ(N −1) ρ(N −2) ... 1

where N is equal to Nr or Nt , corresponding to the receive or transmit antenna array,
and ρ is the fading correlation between two adjacent receive or transmit antenna
elements. Note that for small ρ(d), the higher-order terms are negligible and the
correlation matrices take the form of diagonal matrices.
In practical cases, the degenerate channel phenomena called keyholes may arise,
where the antenna elements at both the transmitter and the receiver have very low
correlation, yet the channel matrix H has only a single degree of freedom, yielding
a single mode of communication [2]. This phenomenon is very similar to the case
39
when rich-scattering transmit and receive antennas are separated by a screen with
the wave passing through a keyhole. This model also applies for indoor propagation
through hallways, narrow tunnels or wave guides. Relay channels in the amplify-
and-forward mode can be treated as keyhole channels. Thus, low correlation is not
a guarantee for achieving high capacity.
3.5 Downlink-Broadcasting Multiuser-MIMO
We still use K and M to denote the number of UEs in the system, and the
number of antennas at the BS respectively. As we can see from Figure 3.3 below,
a single base station with multiple antennas broadcasting signals to UE with single
antenna element. At the BS we assumed uniform linear array (ULA) which is the
most common linear antenna array.
Figure 3.3: MIMO-BC with M antenna base stations
Denote HK the channel matrix between the BS and K-UE. The received vector
yK at K-UE can be written [2, 18] and [28] as
yK = HK X + nK (3.12)
where, X is the transmitted vector of the BS and nK is a vector of complex AWGN

samples with zero mean and unit variance. Similar to the MAC, HK are independent
among all UEs and perfectly known at both the transmitter and the receiver sides.
The common Kronecker channel correlation model in (3.7) and (3.8) are still used.
Note that in this case,RT X and RRX with tr(RT X ) = M and tr(RRX ) = K represent
correlation effects at the BS side and K-UE side respectively. Based on the duality
40
principle between MIMO BCs and MACs shown in [see: Appendix-A], the sum-
rate capacity of a BC can be written as
K
!
X
max ∗

CBC = tr(QK )≤P log2 det IR + HK QK HK (3.13)
K=1
Subject to tr(QK ) ≤ P .
where, P is the transmitted sum power constraint for the BS and QK the input
covariance matrix of K-UE in the dual MAC. Note that the difference between
equation (3.9) and (3.14) is the power constraint. The former is based on individual
power constraints; while the later is based on a sum power constraint.
3.6 Indoor Scenario Consideration
As we have mentioned earlier in this thesis work, our work is mainly studied for
indoor scenario. Based on indoor environment we divided into three scenarios which
is reliant on the location of UE and AP (BS) when both are LOS, NLOS and both
LOS and NLOS to each other. These are:
1. Scenario-I: when BS and MS are both in LOS
• We considered the channel is fixed and we assumed it as iid Gaussian

non-fading channel (HZ ) [8, 9, 10]. Therefore, the Kronecker channel
correlation would becomes
p p
HK1 = RRx HZ RT x (3.14)
• Then by duality principle of MIMO BC-MAC [see: Appendix A], the

uplink and downlink MU-MIMO sum-rate channel capacity in LOS can
be obtained by substituting equation (3.15) into equation (3.9) and (3.14).
2. Scenario-II: when BS and MS are both in NLOS
• In this scenario we consider the channel is variable due to multiple scat-

terer and we assumed the Rayleigh fading channel [13] which is iid com-
plex symmetric Gaussian (HW ). Therefore, the Kronecker channel corre-
lation would become,
p p
HK2 = RRx HW RT x (3.15)
• Then by using the duality principle of MIMO BC-MAC [see: Appendix
41
A], the uplink and downlink MU-MIMO sum-rate channel capacity in
NLOS can be obtained by substituting equation (3.16) into equation (3.9).
3. Scenario-III: when BS is LOS to some users and NLOS to others,
• In this scenario we consider that the channel is fading and we assumed

Ricean fading channel [13] which is the sum of fixed and variable (fading)
channels when K is the constant Ricean factor. i.e.
r r
K 1
HR = Hf ixed + Hvariable (3.16)
K +1 K +1
• Therefore, in the Kronecker channel correlation would become,

p p
HK3 = RRx HR RT x (3.17)
• Then similar to the above two scenarios, by using duality principles of

MIMO BC-MAC [see: Appendix A], we can obtain the sum-rate chan-
nel capacity in both LOS an NLOS by substituting equation (3.18) into
equation (3.9) and (3.14).
The following table contains the assumed indoor LOS and NLOS antenna correlation
in both uplink(MIMO-MAC) and downlink(MIMO-BC) and the Kronecker model.
Table 3.1: Indoor correlation matrix
Correlation matrix Uplink(MIMO-MAC) Downlink(MIMO-BC)

RT x RM S → LOS RBS → LOS
RM S → N LOS RBS → N LOS
RRx RBS → LOS RM S → LOS
RBS →N N LOS → N LOS
RM S N
RH RM S N RBS → LOS RBS N RM S → LOS
RM S RBS → N LOS RBS RM S → N LOS
42
Chapter 4
Simulation Result and Discussion
4.1 Introduction
In chapter two covers some theoretical and analytical part which helps to un-
derstand this thesis work and in chapter three, Jakes spatial correlation model and
Kronecker channel correlation model was discussed. And also, MU-MIMO channel
capacity based on duality principles of MIMO-MAC and MIMO-BC which differs
based on their power constraints was discussed. In this chapter, demonstration of
simulation results on the basis of ideas discussed in previous two chapters using
MATLAB 13.0 is going to be done.
It is obvious that, nowadays thanks to technology we have a lot of programming

languages and coding platforms available for system simulation development includ-
ing C/C++, Java, MATLAB, atoll and other modeling software. While the first
two offer much faster execution of the program, they require developing libraries or
components for each and every element, which is very time consuming. Sometimes,
open source libraries for different functions are available, but their reliability is not
guaranteed and also atoll like modeling packages can model a system but based on
pre-defined inbuilt formulas algorithms which are not modified or changed.
On the other hand, MATLAB is relatively slow in terms of execution speeds

but has numerous built-in functions and specially developed tool boxes for commu-
nications and signal processing which makes the coding both simple and straight
forward. Since the focus of this thesis is not developing new basic tools, but rather
using those tools for evaluations purposes, these built-in functions save significant
time and energy. Also, the powerful computers can run the system at a reasonably
fast pace, so for this reason, MATLAB has been chosen as the underlying platform
for all simulator and code development work.
43
For comparison one must take note that MIMO-MAC channel capacity, MIMO-
BC channel capacity is based on antenna spacing, user separation and fading model.
In order to present equitable comparison, we assume that in MIMO-BC the total
transmitted power is equally allocated among antenna elements at BS and in MIMO-
MAC case the transmitted power is based on individual power constraints. And also,
the channel state information (CSI) at both transmitter and receiver side is assumed
to be known. Angle of spread and angle of incidence is taken based on the BS and
Users location in indoor scenarios.
Table 4.1: Simulation parameters
Antenna spacing at BS From 0.1λ up to 1λ

From 0.5m up to 10m in LOS
User separation in LOS and NLOS
From 2m up to 30m in NLOS
Both 45 degree in LOS
Angle of Departure,Angle of Arrival
10 degree and 22.5 degree in NLOS respectively
Spatial Correlation Jakes spatial correlation model
Channel Correlation Kronecker channel correlation model
iid non-fading Gaussian channel for indoor LOS
Channel model Rayleigh fading channel for indoor NLOS
Ricean fading channel for indoor LOS and NLOS
Equally distributed among all antennas at BS
Transmit power
Based on individual power constraints at MS
Carrier frequency 2.4GHz
Channel State Information Known both at transmitter and receiver side
44
4.2 Flow Chart
It is essential to visualize the simulation using flow chart since it describes the
ones understanding of the whole process easily and it also gives a chance to think
where the process can be improved .Flow chart is shown in Figure 4.1 below.
Figure 4.1: Simulation flow chart
4.3 Antenna Spatial Correlation
Suppose we have M antenna at BS and K mobile User in an indoor environment,

then the antenna correlation is the interference between adjacent antenna elements
during broadcasting of the same signal to K users and multi-cast of different message
signal for K users simultaneously. Normally spatial correlation is small number
from 0 to 1. It is easy to achieve sufficient antenna spacing and very small spatial
correlation by varying the space between adjacent antenna elements without wasting
transmission power and bandwidth. However, in modern wireless technology, it is
challenging to increase the spacing between adjacent antenna elements as we like
because of modern wireless equipment size. But, it is possible to obtain adequate
45
antenna spacing to get optimum channel capacity.
4.4 Channel Capacity
Channel capacity is the upper limit of transmission rate for reliable communica-
tion. Channel capacity depends on signal to noise ratio (SNR), number of antenna
elements (in MU-MIMO) and bandwidth. In the next section, we will see the sim-
ulation result for the effects of antenna spacing and random user separation on
correlation coefficient and channel capacity. We will also see the simulation result
for MIMO-BC and MIMO-MAC channel capacity (i.e., in bits/s/Hz) of four trans-
mit and four receive antenna for Rayleigh fading channel, iid Gaussian and four
transmit and eight receive Ricean fading channel with an intention of comparing
the amount of SNR required for the same spatial correlation of the three different
channels (Rayleigh, Gaussian and Ricean fading). Since this thesis is focused to find
adequate antenna spacing for optimum MU-MIMO channel capacity, by comparing
the three graphs of Rayleigh, Gaussian and Ricean channel to be seen later, here
it has been confirmed that Ricean fading outperforms Rayleigh fading as it has a
strong signal component.
4.5 Simulation Results and Discussion
4.5.1 Antenna Radiation Pattern
For four elements uniform linear array (ULA) antennas at BS of different spacing,
[27] its radiation pattern looks like as shown in Figure 4.2 below. As we can see from
Figure 4.2, as the spacing between two adjacent antenna increases from 0.25λ to 1λ
the strength of the side-lobe also increases. This increment in side lobe strength
finally leads to unintended lobe called grating lobes. The grating lobe occurs when
the spacing between two adjacent antennas is 1λ and it spreading angle is 90 degree.
46
(a) d/λ=0.25
(b) d/λ=0.5
(c) d/λ=0.75
(d) d/λ=1
Figure 4.2: Antenna radiation pattern for different spacing
47
4.6 Spatial correlations
As mentioned earlier in chapter three, the spatial correlation was modeled by

using Jake’s spatial correlation model which is equivalent to zero order Bessel func-
tion. From this model, as the spacing between two adjacent antenna elements and
the separation distance among randomly distributed k-users increases the spatial
correlation of antenna decreases. Here, we can conclude that by increasing the spac-
ing between antenna elements at BS and the separation distance between K-users
we can decrease the antenna spatial correlation as shown in Figure 4.3, 4.4 and 4.5
respectively. But, with the size of latest technology devices we cannot increase the
space between two adjacent antenna elements as we like. The aim of this thesis
work is not to increase antenna spacing but to find the adequate spacing between
two antenna elements. Here Figure 4.3 shows the effects of antenna spacing on
spatial correlation from 0.1λ to λ with carrier wavelength of 0.125m.
Figure 4.3: Effects of antenna spacing on spatial correlation at BS
Figure 4.4 and Figure 4.5 below shows effects of user separation on spatial cor-
relation in LOS which is from 0.5m to 10m and in NLOS from 2m to 30m. In both
cases, as separation distance between the users increases the spatial correlation co-
efficient decreases.
48
Figure 4.4: Effects of user separation in LOS on spatial correlation
Figure 4.5: Effects of user separation in NLOS on spatial correlation
49
As shown in Figure 4.6 and 4.7 below, the space between two adjacent antenna
elements has greater impact on indoor multiuser-MIMO BC in NLOS correlated
sum-rate channel capacity as compared to uncorrelated Rayleigh fading channel.
As shown below in (4xTX, 4xRX) for antenna spacing 0.1λ, 0.2λ, 0.5λ and λ, the
better antenna spacing could be 0.2λ and 0.5λ at which we can obtain better sum-
rate channel capacity compared to uncorrelated Rayleigh fading channel. Here as
shown by simulation, antenna spacing below 0.2λ or above 0.5λ minimizes the sum-
rate channel capacity. So this simulation result proves again the theoretical concept
which says in rich scattering environment 0.5λ antenna spacing is sufficient. As
shown in Figure 4.6 the channel capacity at 0.1λ and λ are minimum compared to
the uncorrelated Rayleigh fading channel capacity. The reason is at 0.1λ the spatial
correlation is high and at λ due to grating lobe where both parameter minimizes
the channel capacity.
Figure 4.6: (4xTX, 4xRX) MIMO-BC Sum-rate channel capacity which shows the
effect of antenna spacing below 0.5λ as compared to iid Rayleigh fading channel
capacity in indoor NLOS.
50
effect of antenna spacing above 0.5λ as compared to iid Rayleigh fading channel
capacity in indoor NLOS.
As shown in Figure 4.8 and 4.9 below, the multiuser MIMO-BC sum-rate channel
capacity is compared to Gaussian iid channel capacity in an indoor LOS. Here, it is
shown that the better antenna spacing in an indoor multiuser-MIMO BC LOS can
be greater than or equal to 0.5λ.
51
effect of antenna spacing above 0.5λ as compared to Gaussian iid channel capacity
in indoor LOS.
52
effect of antenna spacing below 0.5λ as compared to Gaussian iid channel capacity
in indoor LOS.
The simulation result shown in Figure 4.10 and 4.11 below indicates the effects of
average random user separation distance from 0.5m to 10m in LOS and 2m to 30m
in NLOS respectively. From this simulation result, as it is shown if the separation
distance between the randomly distributed users is too much closer or too much far
from each other it has its own impact on indoor multiuser-MIMO MAC in NLOS.
53
Figure 4.10: (4xTX, 4xRX) MIMO-MAC Sum-rate channel capacity which shows
the effect average separation distance of sparsely randomly distributed users in 25m
coverage areas in and indoor NLOS.
The Figure 4.11 below shows the effects of randomly distributed users separa-
tion distance from 0.5m to 10m in indoor multiuser-MIMO MAC sum-rate channel
capacity in LOS to BS. As shown from the simulation result in MIMO-MAC if users
are LOS to the BS they must not be closer to each other.
54
the effect average separation distance of randomly distributed users in 10m coverage
areas in an indoor LOS.
The following simulation result shown in Figure 4.12 and 4.13, shows the effects
of two adjacent antennas spacing on indoor multiuser-MIMO BC sum-rate channel
capacity when users are both in LOS and NLOS to the BS. Antenna spacing for
0.1λ, 0.2λ, 0.3λ, 0.5λ, and λ is selected. From this result we can conclude that
antenna spacing above 0.5λ is better when compared to uncorrelated Ricean fading
channels.
55
effects antenna spacing below and above 0.5λ in indoor scenarios.
56
effects antenna spacing below and above 0.5λ in indoor scenarios.
In the following simulation result in Figure 4.14 shows that the sum-rate channels
capacity of 8-users communicating with single BS of 4-antennas using Ricean fading
channel. Where dta(n) represents the random user separation distance in LOS and
dT (n) represents the random user separation in NLOS.
57
the effects random user separation in indoor scenarios
If the average random user separation increases in both LOS and NLOS and
using Ricean fading channels, the sum-rate channel capacity become the same.
58
Chapter 5
Conclusion and Recommendation
5.1 Conclusion
Based on the simulation results obtained in chapter 4, the following conclusion
has been given.
• In multi-user environment increasing the number of antenna elements at both

transmitter and receiver side is increasing the channel capacity linearly. But,
simply by increasing the number of antenna elements the channel capacity
cannot be increased. Here, as it has been seen on the simulation results, an-
tenna spacing and user separation plays the main role on the sum-rate channel
capacities of (4xTX, 4xRX) MU-MIMO system in an indoor environment.
• In an indoor NLOS the antenna spacing from 0.2λ to 0.5λ is better to get
optimum sum-rate channel capacity in indoor (4xTX, 4xRX) MU-MIMO BC
. Hence, NLOS path is rich of scatterers, generally Rayleigh fading channels
has been considered to calculate the sum-rate channel capacity.
• In an indoor LOS the antenna spacing greater or equal to 0.5λ is better to

obtain the optimum sum-rate channel capacity of (4xTX, 4xRX) indoor MU-
MIMO BC . Since in LOS the channel is fixed, iid Gaussian non-fading channel
is selected.
• Since indoor environment scenario consists of LOS and NLOS path Ricean
fading channels is selected to calculate the sum-rate capacity of (4xTX, 8xRX)
MU-MIMO BC channel capacity. Ricean fading channel is better than Rayleigh
fading channel to obtain the optimum sum-rate channel capacity with the same
antenna spacing at BS.
• In (4xTX, 4xRX) MIMO-MAC Sum-rate channel capacity, the effects of aver-

age separation distance of randomly distributed users in 25m coverage areas in
59
an indoor NLOS has been considered. In (4xTX, 4xRX) MIMO-MAC indoor
NLOS, the average separation distance of randomly distributed users in 30m
coverage areas should not be much closer to each other less than 4m to get
finest sum-rate channel capacity. Since the channel is Rayleigh fading channel.
• In (4xTX, 4xRX) MIMO-MAC Sum-rate channel capacity, the effect of average

separation distance of randomly distributed users in 10m coverage areas in an
indoor LOS has been considered. In this scenario, generally the randomly
distributed user should not be much closer to each other less than 2m since
there is no scatterer. The channel is iid Gaussian non-fading channel.
• In (8xTX, 4xRX) MIMO-MAC Sum-rate channel capacity, the effects random

user separation in indoor scenarios has been considered. In this case, since the
fading channel is Ricean, the separation distance among randomly distributed
user in both LOS and NLOS path has no much effects on the sum-rate channel
capacity of MU-MIMO system.
5.2 Drawbacks
In this thesis work, the following drawback was observed.These are,
• In uplink (MIMO-MAC) of fading and non-fading channels, the correlated

sum-rate channel capacity has no parameter of comparison. Which is in
MIMO-MAC the correlated sum-rate channel capacity is equivalent to single
user sum-rate channel capacity. Therefore, we compared the correlated sum-
rate channel capacity to each other to get better user separation for optimum
correlated sum-rate channel capacity.
• In this thesis work the correlated sum-rate channel capacity is limited to equal
number of BS antennas and K-users with single antenna element on their
device. Because, to obtain the Kronecker channel correlation model first we
need to get antenna correlations. That is, to get correlated channel, we need
to multiply both the transmitter and receiver antenna correlation with fading
or non-fading channels.
5.3 Recommendation for Future Works

The following few promising future works were proposed.
• In this thesis work , it has been considered spatial correlation effects on the
sum-rate channel capacity on different fading channels for indoor LOS and
60
NLOS where in both cases decisions are based on analytical-based channel
correlation model. The use of physical channel model which is more accurate
for indoor MU-MIMO sum-rate channel capacity could be interesting area to
investigate.
• The results presented in this thesis work are based on the simulation results
that the number of antennas at the BS and the number of k-users with single
antenna element on their devices are assumed to be equal. And the channel
state information is assumed to be known at both transmitter and receiver
sides. The above two assumptions is not usually the case in real condition
and studying the impact of real assumptions on indoor MU-MIMO sum-rate
channel capacity evaluation need to be addressed.
• This thesis work is limited to antenna spacing effects on the sum-rate channel
capacity of indoor MU-MIMO system which focuses only to obtain adequate
antenna spacing but, because of the limited size of our equipment it is impos-
sible increase the space between two adjacent antenna elements as we like. So,
we have a limited equipment size. To overcome this problem to some extent,
the idea of side-lobe canceller and beamformer should be investigated. These
two ideas can handle the effects of the two adjacent antennas on each other.
A study in this area is another work in multiuser environment.
• In this thesis work different correlation effects have been neglected. Using
different correlation techniques e.g temporal correlation, phase correlation that
can affect the indoor MU-MIMO sum-rate channel capacity. Therefore,spatial
and temporal dynamics in indoor MU-MIMO is also another future research
work.
• In this thesis work antenna correlation at base station and user separation
at UE side considered without knowing antenna state. That is, excited and
terminated antenna at both side should be considered to compute effects of
antenna correlation in all ports on channel capacity. Therefore the idea of
embedded element pattern in indoor MU-MIMO is also another future research
work.
• This thesis work is limited to the effects of antenna correlation on channel

capacity only without considering the order of diversity gain. Therefore effects
of antenna mutual coupling will be another future research work in indoor MU-
MIMO channel capacity.
61
Bibliography
[1] I. Poole Multi-User MIMO, http://www.radio-electronics.com/info/

antennas/mimo/multi-user-mu-mimo.php.[Online] (November,2014).
[2] B. Z. Maha and R. Kosai Multi User MIMO Communication: Basic As-
pects, Benefits and Challenges.,Recent Trends in Multi-user MIMO Commu-
nications,December 4, 2013.
[3] Professor R. W.Heath,MIMO-communication, http://www.profheath.org/

mimo-communication/multiple-user-mimo/.[Online](November, 2014).
[4] H. Q. Ngo, Performance Bounds for Very Large Multiuser MIMO Systems,
ISSN 0280-7971, November 12, 2012.
[5] V. Stankovic, Multi-user MIMO wireless communications, November, 2006.
[6] O. Edfors, F. Tufvesson and F. Rusek, Indoor Multi User MIMO: Measured
User Orthogonality and its impact on the choice of codings, Lund University,
Sweden, 2012.
[7] T. Wu, Y.H. Kwon, Anthony C. K. Soong, R. W. Heath Jr., Multiuser MIMO
in Distributed Antenna Systems, November, 2010.
[8] M. Kyung and H. Mohaisen, On Transmit Antenna Selection for Multiuser

MIMO Systems with Dirty Paper Coding, South Korea, September, 2009.
[9] V. Jungnickel, T. Haustein, C. von Helmolt V. Pohl, Antenna Spacing in MIMO

Indoor Channels, Berlin, Germany, 2002.
[10] B. Holter, On the capacity of the MIMO channel, Norwegian University of

Science and Technology,Department of Telecommunications, Norway, 2003.
[11] J.W. Wallace, Modeling the Indoor MIMO Wireless Channel, IEEE Trans. on
Antennas and Propagation, vol. 50, no. 5, pp. 591-599, May 2002.
[12] J. Wallace and M. Jensen, Statistical Characteristics of Measured MIMO Wire-

less Channel Data and Comparison to Conventional Models, in Proc. IEEE
Vehicular Technology Conference, vol. 2, no. 2, pp. 1078-1082, October 2001.
62
[13] G. Caire and S. Shamai, On the achievable throughput of a multi-antenna
Gaussian broadcast channel, IEEE Trans. on Inf. Theory., vol. 49, no. 7, pp.
1691-1706, July 2003.
[14] M.H.M. Costa, Writing on dirty paper, IEEE Trans. on Inf.Theory, vol. 29,
no. 12, pp. 439-441, May 1983.
[15] N. Jindal, and A. J. Goldsmith S. Vishwanath, On the capacity of multiple

input multiple output broadcast channels, in Proc. of the IEEE International
Conference on Communications (ICC) , New York, April 2002.
[16] B. Hochwald, and L. Swindlehurst C. Peel, A vector-perturbation technique for

near-capacity multi-antenna multi-user communication, in Proc. of the 41st
Allerton Conference on Communication, Control, and Computing October
2003.
[17] R. F. H. Fischer, and J. B. Huber C.Windpassinger, Lattice-reduction-aided

broadcast precoding, in Proc. 5th International ITG Conference on Source and
Channel Coding (SCC), vol. 6, no. 2, pp. 403-408, January 2004.
[18] D. Tse and P. Viswanath, Fundamentals of Wireless Communication, England,

UK: Cambridge University Press, 2005.
[19] Nimay Ch. Giri, Capacity Performance Comparisonof SISO and MIMO System
for Next Generation Network (NGN), International Journal of Advanced Re-
search in Computer Engineering Technology (IJARCET), vol. 3, no. 9, Septem-
ber 2014.
[20] M. Herdin, H.A Ozcelik, and E. Bonek W. Weichselberger, A Stochastic MIMO

Channel Model with Joint Correlation of Both Link Ends, IEEE Trans. on Wire-
less Comm.vol.5, no.1, January 16,2006.
[21] L. Schumacher, K. Pedersen, P. Mogensen, and F. Frederiksen J. Kermoal, A

Stochastic MIMO Radio Channel Model with Experimental Validation, IEEE J.
Sel. Areas Comm., vol. 20, no. 6, pp. 1211-1226, August 2002.
[22] G. Foschini, M. Gans, and J. Kahn D.-S. Shiu, Fading Correlation and Its Effect
on the Capacity of Multi-element Antenna Systems, IEEE Trans. Comm., vol.
48, no. 3, pp. 502-513, March 2000.
[23] F. Rashid-Farrokhi, J. Ling, and A. Lozano D. Chizhik,Effect of Antenna Sepa-

ration on the Capacity of BLAST in Correlated Channels, IEEE Comm. Letters,
vol. 4, no. 11, pp. 337-339, November 2000.
63
[24] J. Kahn, and D. Tse C.-N. Chuah, Capacity of Multi-Antenna Array Systems
in Indoor Wireless Environment, in Proc. IEEE Global Telecommunications
Conf., vol. 4, pp. 1894-1899, 1998.
[25] Tommi J.,HW Implementation of Indoor MIMO WLAN Channel Models, IEEE
802.11-03/824r0, November, 2003.
[26] Erceg V., Indoor MIMO WLAN Channel Models, IEEE 802.11 document
03/161r2, September, 2003.
[27] C. A. Balanis, Antenna Theory, Analysis and Design, John Wiley Sons, Inc.,
New Jersey, 2005.
[28] J. Lv, Y.Lu, Y. Wang, H. Zho and C.Y. Han, Antenna Spacing Effect on Indoor
MIMO Channel Capacity, IEEE Conference Publishing, vol. 3, December, 2005.
[29] S. Saunder, Small cell Forum Release program, http://www.smallcellforum.

org/about/what-we-do/20142015-round/.[Online] (November, 2014)
64
Appendix
Appendix-A: MU-MIMO sum-rate channel capac-

ity and MIMO MAC-BC duality
[2, 18]
For a MIMO system with NT transmit and NR receive antennas, we consider a

K-user MIMO system over a flat fading MAC with M antennas at BS and single
antenna at each UE. Denote HK and XK the channel matrix and the transmitted
vector of K-UE . The received vector from user is
K
X
y= HK XK + n (5.1)
k=1
here n is a vector of complex additive white Gaussian noise (AWGN) samples with
w
zero mean and unit variance. HK are independent among different UEs and perfectly
known at both the transmitter and the receiver. Therefore, the capacity of the
channel is shown as,
C = max

QK I(X; Y ) (5.2)
In which QK is the probability density function (PDF) of the transmit signal

vector x, and I(X; Y ) is the mutual information of random vectors x and y. Namely,
the channel capacity is the maximum mutual information that can be achieved by
varying the PDF of the transmit signal vector. From the fundamental principle
of the information theory, the mutual information of the two continuous random
vectors, x and y; is given as
I(X; Y ) = H(Y ) − H(Y |X) (5.3)
In which H(Y ) is the differential entropy of y and H(Y |X) is the conditional differ-
ential entropy of y when x is given. Using the statistical independence of the two
random vectors n and x in equation(5.1), we can show the following relationship:
65
H(Y |X) = H(n) Then substituting into equation (5.3),
I(X; Y ) = H(Y ) − H(n) (5.4)
Meanwhile, the auto-correlation matrix of y is given as,
Ryy = E yy H

K
! K
!H
X X
=E HK XK + n HK XK + n
k=1 k=1
K
X
H H
XK + EnnH

= E HK XK HK
k=1
XK (5.5)
H H

= E HK XK XK HK + NO IR
k=1
K
X
∗
= HK QK HK + NO IR
k=1
K
X
∗
= NO IR + HK QK HK
k=1
∗
QK = E[XK XK ] = (σP2t ) is the transmitted signal energy. Then, the mutual infor-
x
mation of y and n is respectively given as,
H(Y ) = log2 det (πeRyy ) (5.6)
H(n) = log2 det (πeNo IR ) (5.7)
Substituting into the following,
I(X; Y ) = H(Y ) − H(n)
I(X; Y ) = log2 det (πeRyy ) − log2 det (πeNo IR ) (5.8)
Then, the channel capacity of MIMO MAC channel is expressed as
K
!
X
max max ∗

CM AC = tr(QK ) I(X; Y ) = tr(QK ) log2 det IR + HK QK HK (5.9)
k=1
Under individual power constraints MIMO MAC channel capacity can be modified
66
as,
K
!
X
max max ∗

CM AC = tr( K
P I(X; Y ) = tr(QK )≤PK log2 det IR + HK QK HK
k=1 QK )≤PK
k=1
P (5.10)
K
Subjected to tr k=1 QK ≤ PK , ∀K
Then based on duality principle of Gaussian MIMO MAC-BC with multiple

transmit covariance constraints, we can obtain the MIMO BC sum-rate channel ca-
pacity. i.e. in Gaussian MIMO-BC associated optimization problem such as capacity
computation and beamforming optimization are typically non-convex and cannot be
solved directly. On feasible approach to this problem is to transform them into their
dual MAC which is easier to deal with due to their convexity property. Based on
this there are two types of BC-MAC duality. These are;
• Conventional BC-MAC duality
• Minimax BC-MAC duality
Conventional duality
The conventional BC-MAC duality is established via BC-MAC signal transfor-
mation and has been successfully applied to solve beamforming optimization, SINR
balancing, and capacity region computation. However, this conventional duality ap-
proach is applicable only to the case in which the BS of BC is subjected to single
power constraints. Therefore;
• Under single sum power constraints the capacity region of BC is identical to

that of dual MAC under the same power constraints.
• The channel matrix associated with dual MAC is the conjugate transposed
channel matrix of BC, and the noise covariance matrixes of both channels are
identity matrices.
Minimax BC-MAC duality

The sum-rate maximization problem of a BC with multiple linear constraints has
the solution as the dual MAC minimax optimization problem. Then the channel
matrix of the dual MAC is the conjugate transposed channel matrix of the BC, and
the noise covariance matrix of the dual MAC is unknown variable of the minimax
optimization problem. Generally, based on the conventional duality principle the
67
dual BC sum-rate channel capacity under single sum power constraints become as
follows:
K
!
X
∗
CBC = max max

tr(QK )≤P I(X; Y ) = tr(QK )≤P log2 det IR + HK QK HK (5.11)
k=1
Subjected to tr(QK ) ≤ P, ∀K
68
Appendix-B: Simulation Algorithms
• Step-1: Input
– Number of antennas at the BS

– Number users
– Carrier wavelength
– Angular spread
– Antenna spacing at BS and User separation
• Step-2: Using Jakes correlation model, calculate the transmitter and receiver
antenna correlation coefficients independently i.e.
ρ(d) = exp−j2πd/λcosΦ (5.12)
• Step-3: Then calculate transmitter and receiver antenna correlation matrix

separately.  
1 ρ∗ . . . ρ∗(r−1)
 ρ 1 . . . ρ∗(r−2) 
 
RT x , RRx = 
 .. .. ... ..  (5.13)
. . . 


ρ(r−1) ρ(r−2) . . . 1
Where, r is number of transmitter or receiver antenna.
• Step-4: Use different stochastic models such as, iid Gaussian non-fading chan-
nel (Hz), Rayleigh channel (Hw) and Ricean channel (HR) depending on indoor
scenarios i.e.
Hz = randn(nR , nT ) + jrandn(nR , nT ) (5.14)
1
Hw = √ (randn(nR , nT ) + jrandn(nR , nT )) (5.15)
2
s s
K 1
HR = Hz + Hw (5.16)
(K + 1) (K + 1)
• Step-5: Using Kronecker channel modeling, calculate the correlated chan-

nel matrix between the transmitter and receivers antennas for step-4 indoor
69
scenarios i.e.
1 1
H = RT2 x Hz RRx
2
1
H = RT12x Hw RRx
2
1
12
H = RT2 x HR RRx
• Step-6: Use duality principles of MIMO MAC-BC.
• Step-7: Assume equal power allocation among BS antenna elements and

based on individual power constraints at K-users.
• Step-8: Calculate correlated sum-rate channel capacity based on step 5 and

6 for both MAC and BC i.e.
K
!
X
∗
CM AC = max
tr( K QK )≤PK
P log2 det IM + HK QK HK (5.17)
k=1
k=1
P
K
Subject to tr k=1 QK ≤ PK , ∀K
where PK is individual power constraints, M is number BS antennas and
∗
QK = E[XK XK ] = (σP2t ) . And,
t
K
!
X
max ∗
CBC = tr( K
P log2 det IR + HK QK HK (5.18)
k=1 QK )≤P
k=1
P
K
Subject to tr k=1 QK ≤ P, ∀K Where P is the
total transmitted power and N is number of users.
• Step-9: Compare with non-correlated sum-rate channel capacity.
70

Addis Ababa University School of Graduate Studies Addis Ababa Institute of Technology

Uploaded by

Copyright:

Available Formats

Addis Ababa University School of Graduate Studies Addis Ababa Institute of Technology

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Addis Ababa University School of Graduate Studies Addis Ababa Institute of Technology

Uploaded by

Copyright:

Available Formats

ADDIS ABABA UNIVERSITY

SCHOOL OF GRADUATE STUDIES

Antenna Spacing Effects on Indoor

Dr. Murad Ridwan

A Thesis submitted to Addis Ababa University in partial

SCHOOL OF ELECTRICAL AND COMPUTER

Antenna Spacing Effects on Indoor

ADDIS ABABA INSTITUTE OF TECHNOLOGY

APPROVAL BY BOARD OF EXAMINER

Chairman,Dept. of Graduate Committee Signature

Dr. Murad Ridwan

Internal Examiner Signature

External Examiner Signature

Dedicated to my son Egnuma and his mom...

Place: Addis Ababa

Dr. Murad Ridwan

In theory, multi-user MIMO is more immune to most of propagation limitations

This can be analyzed by using correlation based-analytical approximation method,

KEY WORDS: MU-MIMO, Indoor Scenario, Antenna Spacing, User separa-

List of Figures vii

2 MIMO Wireless Communication 11

3 Multi-User MIMO Communication 31

4 Simulation Result and Discussion 43

5 Conclusion and Recommendation 59

1.1 MU-MIMO block diagram [3] . . . . . . . . . . . . . . . . . . . . . . 2

2.1 SISO model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Antenna Pattern Parameters [27] . . . . . . . . . . . . . . . . . . . . 37

4.1 Simulation flow chart . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.1 Indoor correlation matrix . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.1 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 44

AoA Angle of Arrival

AoD Angle of Departure

AWGN Additive White Gaussian Noise

CDMA Code Division Multiple Access

CSI Channel State Information

DAS Distributed Antenna System

DPC Dirty Paper Coding

DSP Digital Signal Processing

EVD Eigen Value Decomposition

FDMA Frequency Division Multiple Access

FNBW First-Null Beam Width

GSCM Geometry-based Stochastic Channel Models

HPBW Half-Power Beam Width

IEEE Institute of Electrical and Electronics Engineers

iid independent and identically distributed

MEA Multiple Element Antenna

MIMO Multiple-Input Multiple-Output

MISO Multiple Input Single Output

MMSE Minimum Mean Square Error

MU-MIMO Multi-User Multiple-Input Multiple-Output

PAS Power Azimuth Spectrum

QRD Queer Resource Decomposition

RRU Radio Resource Unit

SDMA Space Division Multiple Access

SIMO Single Input Multiple Output

SISO Single Input Single Output

SNR Signal to Noise Ratio

SU-MIMO Single-User Multiple-Input Multiple-Output