Investigation and mitigation of Spatially Correlated Fading in

Massive MIMO

Habtamu Lema Adera

A Thesis Submitted to

Department of Electronics and Communication Engineering

School of Electrical Engineering and Computing

Presented in Partial Fulfillment of the Requirement for the Degree of Master’s in

Electronics and Communication Engineering

Office of Graduate Studies

Adama Science and Technology University

June, 2024
Adama, Ethiopia
Investigation and mitigation of Spatially correlated fading in mas-
sive MIMO

Habtamu Lema Adera

Main Advisor Dr. Demissie Jobir Gelmecha

Co-Advisor Dr. Rajeev Kumar Shakya

A Thesis Submitted to
Department of Electronics and Communication Engineering

School of Electrical Engineering and Computing

Office of Graduate Studies

Adama Science and Technology University

June, 2024
Adama, Ethiopia
 DECLARATION
I hereby declare that this Master Thesis entitled “Investigation and mitigation of Spatially
correlated fading in massive MIMO” is my original work. That is, it has not been submitted
for the award of any academic degree, diploma or certificate in any other university. All sources
of materials that are used for this thesis have been duly acknowledged through citation.

Habtamu Lema Adera ______________ __________

Name of student Signature Date

i|Page
 RECOMMENDATION OF ADVISORS
We, the advisors of this thesis, hereby certify that we have read the revised version of the thesis
entitled “Investigation and mitigation of Spatially correlated fading in massive MIMO”
prepared under our guidance by Habtamu Lema Adera submitted in partial fulfillment of the
requirements for the degree of Masters of Science in Electronics and Communication Engineer-
ing. Therefore, we recommend the submission of revised version of the thesis to the department
following the applicable procedures.

Dr. Demissie Jobir Gelmecha ______________ _______________

Major Advisor Signature Date

Dr. Rajeev Kumar Shakya ______________ _______________

Co-advisor Signature Date

ii | P a g e
 APPROVAL PAGE OF M.SC. THESIS
We, the advisors of the thesis entitled “Investigation and mitigation of Spatially correlated
fading in massive MIMO” and developed by Habtamu Lema Adera, hereby certify that the
recommendation and suggestions made by the board of examiners are appropriately incorpo-
rated into the final version of the thesis.

Dr. Demissie Jobir Gelmecha ______________ __________

Advisor Signature Date

Dr. Rajeev Kumar Shakya ______________ __________

Co-Advisor Signature Date

iii | P a g e
 APPROVAL OF BOARD OF REVIEWERS
We, the undersigned, members of the Board of Examiners of the thesis by Habtamu Lema
Adera have read and evaluated the thesis entitled “Investigation and mitigation of Spatially
correlated fading in massive MIMO” and examined the candidate during open defense. This
is, therefore, to certify that the thesis is accepted for partial fulfillment of the requirement of the
degree of Master of Science in Electronics and Communication Engineering.

________________ ______________ __________

Chairperson Signature Date

________________ ______________ __________

Internal Examiner Signature Date

________________ ______________ __________

External Examiner Signature Date

Finally, approval and acceptance of the thesis is contingent upon submission of its final copy
to the Office of Postgraduate Studies (OPGS) through the Department Graduate Council
(DGC) and School Graduate Committee (SGC).

________________ ______________ __________

Department Chair Signature Date

________________ ______________ __________

School Dean Signature Date

________________ ______________ __________

Postgraduate Dean Signature Date

iv | P a g e
 ACKNOWLEDGMENT
First of all, I would want to thank the Almighty God who provided me the strength and ability
to complete the thesis successfully. After the almighty God, I would like to convey my ardent
gratitude and compliments to my advisor Dr. Demissie Jobir Gelmecha for being the corner-
stone of this endeavor. He provided constant motivation and direction during the period of
questions and uncertainties, which helped me to continue with my thesis. I'm also grateful to
my co-advisor, Dr. Rajeev Kumar Shakya, for his helpful recommendations and ongoing sup-
port. Also I would like to offer my special appreciation to Dr. Ram Sewak Singh for their
advice while I faced challenges on my thesis. I would also like to thank the Electronics and
Communication Engineering (ECE) program for providing me with the opportunity to complete
this wonderful thesis on the topic " Investigation and mitigation of Spatially correlated fading
in massive MIMO System," which will also help me in the future in doing a lot of research and
learning about so many things. I am very thankful to them.

Finally, I would want to thank my parents for their endless counsel, as well as others who helped
me directly or indirectly while working on my thesis.

v|Page
 Table of Contents
DECLARATION ........................................................................................................................ i
RECOMMENDATION OF ADVISORS .................................................................................. ii
APPROVAL PAGE OF M.SC. THESIS .................................................................................. iii
APPROVAL OF BOARD OF REVIEWERS .......................................................................... iv
ACKNOWLEDGMENT .............................................................................................................v
LIST OF TABLES .................................................................................................................... ix
LIST OF FIGURE .......................................................................................................................x
LIST OF ACRONMYS AND ABBREVIATIONS................................................................. xii
ABSTRACT ............................................................................................................................ xiii
CHAPTER ONE .........................................................................................................................1
INTRODUCTION ................................................................................................................. 1
1.1 Background of the Study...............................................................................................1
1.1.1 Massive MIMO Techniques .......................................................................................3
1.1.2 Channel Estimation in Massive MIMO .....................................................................3
1.2 Statement of the problem ..............................................................................................7
1.3 Objective .......................................................................................................................8
1.3.1 General Objective.......................................................................................................8
1.3.2 Specific Objectives.....................................................................................................8
1.4 Scope of the Study ........................................................................................................8
1.5 Significance of the Study ..............................................................................................9
1.6 Thesis Contribution .......................................................................................................9
1.7 Limitation of the Study .................................................................................................9
1.8 Organization of the Thesis. .........................................................................................10
CHAPTER TWO ......................................................................................................................11
LITERATURE REVIEW..................................................................................................... 11
2.1 Introduction. ................................................................................................................11
2.2 Related works. .............................................................................................................13
CHAPTER THREE...................................................................................................................21
MATERIAL AND METHODS ........................................................................................... 21
3.1 Overview .....................................................................................................................21
3.2 Material .......................................................................................................................21
3.3 Methods of the study ...................................................................................................21
vi | P a g e
 3.4 System Model for spatially correlated fading in massive MIMO. ..............................24
3.5 Channel Estimation for spatially correlated fading in massive MIMO ......................26
3.5.1 Least Square Channel Estimator(LS) .......................................................................28
3.5.2 Minimum mean square error channel estimator (MMSE). ......................................29
3.5.3 Element-Wise Minimum Mean Square Error Channel Estimator(EW-MMSE). ....30
3.5.4 Uplink Spectral Efficiency using the MMSE Estimator. .........................................30
3.5.5 Uplink Spectral Efficiency using the EW-MMSE Estimator. .................................31
3.5.6 Uplink Spectral Efficiency using the LS Estimator .................................................32
3.5.7 Downlink Spectral Efficiency using the MMSE Estimator .....................................34
3.5.8 Downlink Spectral Efficiency using the EW-MMSE Estimator..............................36
3.5.9 Downlink Spectral Efficiency using the LS Estimator ............................................37
CHAPTER FOUR .....................................................................................................................39
4. RESULT AND DISCUSSION ........................................................................................ 39
4.1 Performance of estimation quality of channel estimator in terms of NMSE. .............40
4.1.1 Comparison of estimation quality of channel estimator in terms of NMSE. ...........41
4.2 Result of Uplink Spectrum Efficiency Under Different Channel estimator. ..............42
4.2.1 Average Uplink spectrum efficiency for different channel estimator. .....................42
4.2.2 Cumulative distribution Uplink spectrum efficiency for different channel
estimator ............................................................................................................................43
4.2.3 Average Uplink spectrum efficiency under different channel estimator where all have
LOS path. ..........................................................................................................................44
4.2.4 Average Uplink spectrum efficiency under different pilot reuse factor and different
channel estimator. .............................................................................................................46
4.2.5 Average Uplink spectrum efficiency with spatially uncorrelated Racian and Rayleigh
fading under different channel estimator. .........................................................................47
4.2.6 Average Uplink spectrum efficiency for different channel estimator and different
Precoding schemes. ...........................................................................................................48
4.3 Result of Downlink spectrum efficiency under different channel estimator. .............49
4.3.1 Average Downlink spectrum efficiency for different channel estimator. ................50
4.3.2 Average Downlink spectrum efficiency for different channel estimator and different
Precoding schemes. ...........................................................................................................51
5. CONCLUSION AND RECOMMENDATION ....................................................................53
5.1 Conclusion ..................................................................................................................53
5.2 Recommendation.........................................................................................................54

vii | P a g e
REFERENCES..........................................................................................................................55

viii | P a g e
 LIST OF TABLES
Table 2. 1: A Summary of some related work. .........................................................................19
Table 4. 1 Parameters used for Simulations. .............................................................................39

ix | P a g e
 LIST OF FIGURE
Figure 1. 1 (Ngo, 2015) Demand for mobile data traffic and number of connected devices. ...1

Figure 1. 2 (Sofi et al., 2018) Channel Estimation in Massive MIMO. ......................................4

Figure 1. 3 (Nivethitha et al., 2015) Slot structure in systems with pilot –based channel
estimation. ...................................................................................................................................5

Figure 1. 4 (Siddiqui et al., 2023) FFD and TDD modes ...........................................................7

Figure 2. 1(a) (Kansal et al., 2019) Downlink (b) Uplink............................11

Figure 2. 2 (Ali et al., 2018) Uplink and Downlink transmission in a Massive MIMO
system. .......................................................................................................................................12

Figure3. 1 Methodology. ...........................................................................................................22

Figure 3. 2 Flow chart of Spectrum Efficiency analysis ...........................................................23

Figure 4. 1 NMSE in the estimation of a spatially correlated channel, based on the local
scattering model with Gaussian angular distribution, for different estimators. ........................41

Figure 4. 2 Average Uplink Sum Spectrum efficiency for 10 user as function of number of
antennas of BS under different Channel estimator and Fading channels..................................43

Figure 4. 3 CDF curve of the Uplink Spectrum Efficiency under different Channel
estimator. ...................................................................................................................................44

Figure 4. 4 Average Uplink sum of Spectrum efficiency for 10 user is shown as a function of
BS antennas for various channel estimator where all UE-BS have a LOS Path. ......................45

Figure 4. 5 Uplink Spectrum Efficiency for different Pilot reuse factor with Various channel
estimator. ...................................................................................................................................47

Figure 4. 6 Uplink Spectrum Efficiency for spatially uncorrelated Rician and Rayleigh fading
under different channel estimation. ...........................................................................................48

x|Page
Figure 4. 7 Average UL sum SE when using MMSE, EW-MMSE, or LS channel estimators,
for a setup with M = 100 BS antennas and K = 10 UEs per cell. Three different combining
schemes are considered. ............................................................................................................49

Figure 4. 8 Downlink Spectrum Efficiency for 10 user is shown as function of BS antenna for
various channel estimation and fading channel.. ......................................................................51

Figure 4. 9 Average DL sum SE when using the MMSE, EW-MMSE, or LS channel estimators,
for a setup with M = 100 BS antennas and K = 10 UEs per cell. Three different precoding
schemes. ....................................................................................................................................52

xi | P a g e
 LIST OF ACRONMYS AND ABBREVIATIONS

AP Access Point
BS Base Station
CDF Cumulative distribution function
CE Channel estimation
CSI Channel state information
DL Downlink
EW-MMSE Element-Wise minimum mean square error
FDD Frequency division duplex
i.i.d independent and identical distributed random variable
L-MMSE Local minimum mean square error
LOS Line-of-sight
LS Least square
MIMO Multi-Input-Multi-output
MMSE Minimum mean square error
M_MMSE Multicell Minimum Mean-Squared Error
MRC Maximum ratio combining
NLOS Non-line-of-sight
UE User Equipment
ULA Uniform linear array
UL Uplink
UPA Uniform planar array
RZF Regular –zero-forcing
RIS Reconfigurable intelligent surface
SE Spectral efficiency
SINR Signal –to-interference-plus-noise ratio
TDD Time-division Duplex

xii | P a g e
 ABSTRACT
Massive MIMO (Multiple Input Multiple Output) systems, characterized by the use of a large
number of antennas at the base station, offer significant improvements in capacity and spectral
efficiency. However, the benefits of massive MIMO can be hindered by spatial correlation fad-
ing. Spatial correlation fading refers to the phenomenon where the fading characteristics of
multiple antenna channels in a MIMO system are correlated. This correlation arises due to
insufficient spatial separation between antennas, limited scattering environments, or specific
propagation conditions. In order to mitigate these issues, this thesis explores channel estimator
for accurate estimation of the fading channel to improve spectrum efficiency of massive MIMO.
The study begin with provides statistical properties for the minimum mean squared error
(MMSE), element-wise (EW-MMSE), and least-squares (LS) channel estimations in this model.
Additionally, analyze rigorous closed-form uplink (UL) and downlink (DL) spectral efficiency
(SE) expressions. Comprehensive simulations on MATLAB and an analysis comparing the LS,
MMSE and EW-MMSE under different precoder in order to estimate accurate channel state
information, the MMSE demonstrate its better performance. The simulation results show that
the SE is higher with different values of SNR, NMSE, coherence block length, number of BS
antennas and number of Ues when using MMSE channel estimator and lower when using LS
channel estimator.

Key words: Massive-MIMO, Rician Fading, Spatial Correlation, Spectrum Efficiency, Chan-
nel Estimation

xiii | P a g e
 CHAPTER ONE
INTRODUCTION
1.1 Background of the Study
In the recent past, there has been a tremendous increase in data traffics within both mobile
and fixed networks. This increase can be largely attributed to the rising proliferation of
smartphones, tablets, laptops, amongst many other wireless devices that heavily rely on the
consumption of data. Figure 1.1 provides a visual representation of the increasing demand
for mobile data traffic and the growing number of connected devices. Estimations indicate
that global mobile data traffic is to grow six fold from the 2014 levels, reaching 15.9 Exa-
byte’s per month in 2018. Secondly, by 2018, it is projected that mobile devices and con-
nections will be 10.2 billion. In light of this shifting demand, there is need for the establish-
ment of other technologies. In the case of wireless data traffic, one needs to note that wire-
less throughput purely refers to the rate at which data is transmitted wireless basically meas-
ured in bits per second)(Ngo, 2015):

Throughput = Bandwidth (Hz) × Spectral efficiency (bits/s/Hz)(Ngo, 2015).

Figure 1. 1 (Ngo, 2015) Demand for mobile data traffic and number of connected devices.

The future generation of wireless networks is expected to support a significant number of

devices and connections. The objective is to foster a connected society where various enti-
ties such as sensors, automobiles, drones, medical devices, and wearable technology can
utilize mobile networks to establish connections with each other and engage in interactions
with humans(Engineering, 2021). Massive multiple-input multiple-output (MIMO) is an

1|Page
emerging technology that significantly improves MIMO capabilities, outperforming the ex-
isting state-of-the-art by several orders of magnitude. Massive MIMO's performance is de-
pendent on spatial multiplexing, which requires the base station to have correct channel
knowledge for both uplink and downlink communications. Obtaining exact channel infor-
mation for the up-link is quite simple because the terminals transmit pilot signals(Björnson
et al., 2020). The base station uses these pilot signals to calculate the channel replies for
each terminal. Downlinking is more difficult. Massive MIMO can boost capacity by tenfold
or more while also improving radiated energy efficiency by 100 times. Massive MIMO ca-
pacity gains are due to the effective use of spatial multiplexing. While it is widely assumed
that spatial channel correlation is detrimental to MIMO communications, this is not the case
when multiuser communications are carried out using single-antenna user equipment (UE).
The overall network performance is defined by the combination of spatial correlation ma-
trices of the UE(Zhi et al., 2023). Typically, the UEs are physically separated by a large
distance, producing statistically uncorrelated channels. Furthermore, while each UE's chan-
nel may exhibit strong spatial correlation at the base station (BS), the spatial correlation
matrices can differ substantially amongst UEs. These features differentiate multiuser mas-
sive MIMO channels from point-to-point MIMO channels, where spatial channel correlation
is observed from both the transmitter and the receiver, and the channel between transmit
and receive antennas displays comparable spatial correlation patterns(Li et al., 2022).

The time-division duplex (TDD) mode is a typical configuration for massive MIMO sys-
tems, in which uplink and downlink transmissions occur at various time intervals yet within
the same frequency resource. Several benefits encourage the use of TDD method. To begin,
coherent antenna processing requires only the base station (BS) to have channel information.
Second, the overhead for uplink estimation is proportional to the number of terminals, not
the number of antennae (M), allowing for scalability with regard to the number of service
antennas. Furthermore, fundamental estimation theory shows that adding more antennas to
the BS does not reduce the quality of estimate per antenna. In fact, the estimation quality
improves with M if a known correlation structure exists between the channel responses
across the antenna array(Björnson et al., 2016).

2|Page
1.1.1 Massive MIMO Techniques
Spatial diversity, spatial multiplexing, and beamforming are the three key ideas that form
the foundation of massive MIMO. The principle behind multiple-input multiple-output
(MIMO) is that a radio signal traveling between a transmitter and a receiver is filtered by its
environment, creating several signal paths due to reflections from obstacles like buildings.
The time delays, degrees of attenuation, and travel directions of the reflected signals will
change when they reach the receiving antenna. The placement of many receiving antennas
results in the somewhat different versions of the signal that each antenna receives, these
versions can be added numerically to improve the transmission signals quality. Because the
reception antennas are spatially separated, this strategy is known as spatial diversity. Spatial
diversity is also achieved by distributing the radio signal across multiple antenna, with each
antenna sending changed versions of the signal in some cases. While spatial diversity im-
proves radio link reliability, spatial multiplexing boosts radio link capacity by utilizing nu-
merous transition paths as additional data transmission channels. Spatial Multiplexing sends
numerous data streams between the transmitter and receiver, considerably improving per-
formance and allowing a single transmitter to accommodate multiple network users. Instead
of broadcasting to a wide area, beamforming uses contemporary antenna technology to fo-
cus a wireless signal in a specific direction. Beamforming is an important wireless method
that supports Massive MIMO in increasing network capacity and throughput. It directs the
wireless signal in a certain direction, decreasing interference and allowing for larger antenna
arrays(Wang et al., 2020)

1.1.2 Channel Estimation in Massive MIMO

Massive multiple-input multiple-output (MIMO) is a scaled version of multiuser MIMO
(MU-MIMO) that uses a large number of base station (BS) antennas to improve spectral
efficiency through spatial multiplexing and basic linear processing at the BS. Pilot contam-
ination can occur when pilots are reused in various cells and the BS obtains channel state
information (CSI). Pilot contamination can have a significant impact on the asymptotically
achievable rate in large MIMO systems in independent and identically distributed (i.i.d)
Rayleigh fading channels, and various pilot decontamination approaches have recently been
proposed in the literature. Wireless channels are complex, and fading can cause frequency
and time selective behavior. Signature phenomena include reflection, diffraction, scattering,
and Doppler Shift. Doppler Shift refers to the frequency difference between emitted and
received waves when the distance between the receiver and source changes.

3|Page
As a result, channel estimation is a critical approach in wireless communication systems,
especially mobile networks. In order to conduct signal equalization and retrieve the broad-
cast data, the receiver must have a correct understanding of the channel state infor-
mation(Thomas et al., n.d.).

Determining the precise CSI can be a complex task in wireless channels. The wireless chan-
nel is prone to various impairments such as multipath fading, shadowing, and interference,
which can significantly distort the transmitted signal. Without accurate CSI, the receiver
will be unable to properly equalize the received signal and mitigate the effects of the wire-
less channel. This can lead to degraded signal quality, reduced data rates, and unreliable
communication. So, effective channel estimation is vital to enable coherent detection and
decoding at the receiver in mobile wireless networks. The complexity of this task under-
scores the importance of developing robust and efficient channel estimation algorithms to
support high-performance wireless communication. Therefore, In MIMO, several antennas
are used on both the sending and receiving sides, as illustrated in Figure 1.2(Sofi et al.,
2018).

Figure 1. 2 (Sofi et al., 2018) Channel Estimation in Massive MIMO.

Most channel estimation techniques in wireless systems rely on the use of training sequences
on transmission. These training sequences consist of pilot symbols that are known to the
receiver. The pilot symbols are transmitted along with the actual data sequence. The receiver
uses these known pilot symbols to estimate the current channel state information (CSI). This
estimated CSI is then used to equalize and detect the data symbols that immediately follow
the training sequence. The underlying assumption is that the channel remains unchanged

4|Page
during the coherence time interval, which is the duration over which the channel character-
istics can be considered static.

Figure 1.3 illustrates a typical time slot structure in systems that perform channel estimation
based on these training sequences. The time slot is divided into a training sequence segment
followed by the actual data transmission. This approach allows the receiver to continuously
update the CSI estimation by periodically inserting known training sequences within the
data transmission. This is crucial for maintaining reliable communication in time-varying
wireless channels(Nivethitha et al., 2015).

Figure 1. 3 (Nivethitha et al., 2015) Slot structure in systems with pilot –based channel
estimation.

Massive MIMO systems rely largely on spatial multiplexing to obtain performance im-
provements. For this to work well, the base station must have an accurate understanding of
the channel conditions. In the uplink, users send training sequences to the base station,
which estimates the channel replies for each particular user. This channel state information
(CSI) allows the base station to recognize signals from different users in a coherent manner.
However, obtaining precise CSI for the downlink can be more difficult. Unlike the uplink,
the base station must know the forward channel answers to each terminal ahead of time in
order to execute precoding and beamforming to properly transmit to the users(Björnson &
Demir, 2024).

Obtaining correct downlink CSI can be more difficult due to time-varying channels and the
huge number of antennas at the base station. In Massive MIMO systems, efficient channel
estimation techniques are critical to ensuring that the base station has the required CSI for
both the uplink and the downlink. Massive MIMO operation relies largely on good channel
estimates and data capture to enable continuous and reliable transmission. To estimate CSI,
pilot data is transferred between the base station and the smart node to ensure optimal radio
link connections. The technique is divided into two categories based on resource allocation
(time and frequency). FDD is a time-continuous phenomenon, whereas TDD comprises dis-
continuous transmission slots as illustrate on figure 1.4 below(Siddiqui et al., 2023).

5|Page
The system in TDD Massive MIMO topologies is designed to take advantage of channel
reciprocity. This means that for a given coherence interval, the forward (downlink) and re-
verse (uplink) channel replies are expected to be the same. Using channel reciprocity, the
base station can use the predicted channel responses from the uplink training sequences to
describe the downlink channel as well. This enables the base station to obtain the necessary
channel state information (CSI) for downlink transmission without requiring additional
downlink pilots. Because the uplink and downlink channels are reciprocal, the base station
can simply apply the uplink CSI estimates for subsequent downlink transmissions within
the same coherence interval. This considerably minimizes the overhead requirements for
downlink CSI acquisition. This method is a fundamental feature of TDD Massive MIMO
systems since it simplifies channel estimation and allows for more efficient use of available
resources. The assumption of channel reciprocity is crucial in realizing the full potential of
massive MIMO in TDD designs(Rajani et al., 2018).

In contrast to TDD systems, FDD Massive MIMO topologies have non-reciprocal uplink
and downlink channels. This means that the base station cannot simply use the uplink chan-
nel estimations for downlink broadcast. Instead, FDD systems require a closed-loop setup
for channel estimation. In uplink channel estimation, the base station calculates the reverse
(uplink) channel using training sequences given by users. This helps to lower the computa-
tional load and power consumption on mobile devices. To estimate the forward (downlink)
channel, the base station must first send its own training sequences. The users then offer the
base station with limited input on the predicted downlink channel state information (CSI).
This feedback-based method is required in FDD systems since channel reciprocity cannot
be assumed. The base station needs the downlink CSI to precoding and beamforming the
downlink transmission(ScholarWorks et al., 2016).

Massive MIMO systems for TDD designs necessitate the same number of training se-
quences as the number of users since the base station can exploit channel reciprocity to reuse
uplink channel estimates for downlink transmission. However, for FDD systems, the num-
ber of training sequences required is equal to the number of antennas deployed at the base.
Because, the base station must transmit its own training sequences to all antennas in order
to acquire the downlink channel state information (CSI) via feedback from the users(Sid-
diqui et al., 2023).

6|Page
 Figure 1. 4 (Siddiqui et al., 2023) FFD and TDD modes

1.2 Statement of the problem

The massive MIMO system is the technology lead to meet answer of the problem in high
dense urban environment or areas with a high number of connected device as a promising
solution to meet the growing need for faster data rates and enhance spectral efficiency in
wireless communication systems. But, from an information-theoretical perspective, the
spectral efficiency (SE) performance of a wireless communication system depends heavily
not only on the signal-to-noise ratio (SNR), but also on the degree of spatial correlated fad-
ing. Since, in real-world circumstances, the SE of massive MIMO is influenced by spatial
correlated fading, which is caused by LOS path along with scattered multipath components
and dependencies among different antenna elements. So, in order to overcome the problem
occurred due to spatial correlated fading, channel estimation techniques are introduced in
this study. Furthermore, to improve the SE of Massive MIMO under spatial correlated fad-
ing MMSE, EW-MMSE and LS channel estimation techniques are developed.

7|Page
1.3 Objective
1.3.1 General Objective
The main objective of this thesis is to investigate and Mitigation the impact of spatially
correlated fading channels on massive MIMO system.

1.3.2 Specific Objectives.

 To investigate the impact of channel estimation errors on the spectral efficiency of
uplink and downlink transmission in massive MIMO system.
 To evaluate the performance of various cumulative distribution function (CDF)
curves for the UL SE per UE for different channels in Massive MIMO systems.
 To study and analyzes the performance of minimum mean squared error (MMSE)
and Element-wise (EW-MMSE) for spatially correlated fading in Massive MIMO
systems.
 To compare the average UL SE with uncorrelated fading as a function of the number
of antennas.

1.4 Scope of the Study

This thesis is focus on optimizing spectral efficiency of massive MIMO performance in
spatial correlated fading channels. Massive MIMO is an advanced wireless communication
technology that leverages a large number of antennas at the base station to efficiently cater
to multiple users concurrently. In wireless communication scenarios, Rician fading channels
are frequently encountered, characterized by a combination of line-of-sight (LOS) and non-
line-of-sight (NLOS) components that contribute to the overall signal propagation. The
work offers a comprehensive analysis of channel estimation techniques and spectral effi-
ciency in multi-cell Massive MIMO systems under spatially correlated Rician fading chan-
nels. Furthermore, the scope of the study may extend to investigate the impact of spatial
correlation on the performance metrics of massive MIMO system such as spectral efficiency
and system capacity. Overall, the scope of the study encompasses analysis an accurate math-
ematical model that captures the spatial correlation and Racian fading characteristics in mas-
sive MIMO system, with the aim of achieving better performance, optimizing the design
and deployment of massive MIMO systems.

8|Page
1.5 Significance of the Study
In telecommunications, the high data rate is essential for the smooth operation of wireless
communications. Massive MIMO is a wireless communication technique that utilizes a sub-
stantial number of antennas at the base station to simultaneously serve multiple users. In a
spatially correlated Rician fading channel, the fading characteristics are influenced by the
correlation between different antennas and the presence of a dominant line-of-sight compo-
nent. This study can assess the spectrum efficiency achieved by using massive MIMO in
such channels, leading to improved wireless communication. Since in practical scenarios,
the channels between different antennas are often correlated and understanding this corre-
lation is crucial for designing efficient communication system. The significant of this thesis
is that to enhance the understanding of the interaction between massive MIMO systems and
realistic channel conditions, leading to improve design and performance of spectrum effi-
ciency wireless communication networks. By incorporating the impact of spatially Rician
fading the study seeks to enhance the quality of the received signals, energy efficiency,
reliability, robustness of the communication system and overall system performance. Over-
all, a study on the investigation and mitigation correlated Rician fading channel in massive
MIMO system can contribute to the advancement of based on performance by channel ca-
pacity, signal-to-noise ratio, and overall reliability of communication system.

1.6 Thesis Contribution

The major contributions of this thesis are as follows:
1. Applied different types of channel estimation techniques to alleviate the degradation of
the spectrum efficiency in spatial correlation fading channel, which occurs due to phys-
ical proximity of antenna element and both line of sight and scattered components.
Moreover, the performance comparisons of various type of channel estimation tech-
niques performed in order to choose the best channel estimator to improve spectrum
efficiency.
2. Depending upon the results found from the comparison, MMSE channel estimator
demonstrated better performance to improve spectrum efficiency.

1.7 Limitation of the Study

This thesis is limited only to simulation based results. That means the performance of im-
proving spectrum efficiency in spatial correlated fading channels are evaluated based on
simulation results only.

9|Page
1.8 Organization of the Thesis.
The thesis has been organized into five chapters.

Chapter 1 briefly discusses the background of the study. The problem statement, scope of
the study, significance of the thesis and thesis objectives are also laid out here with its lim-
itation.

Chapter 2 starts by discussing a brief introduction about Overview of massive MIMO and
Massive MIMO techniques. Then functions of Massive MIMO are discussed in short. This
chapter covers a related work of thesis.

Chapter 3 presents the research method that has been used in the thesis and the system
models for spatial correlated fading in massive MIMO. Also, this chapter introduced mate-
rials that have been used for simulating the results.

Chapter 4 includes the results and discussion for the Uplink and Downlink Spectrum effi-
ciency of multi-cell massive MIMO system are examined under spatial correlated Racian
and Rayliegh fading channel.

Lastly, the Conclusion and Recommendation of the thesis are presented.

10 | P a g e
 CHAPTER TWO
LITERATURE REVIEW
2.1 Introduction.
Massive MIMO, also known as large-scale MIMO is a term used to describe MIMO systems
with a significant number of antennas in which a base station equipped with M antennas
communicates to K users concurrently. Concurrently refers to the transmission of signals to
and from each user at the same time and frequency. Figure 2.1 represents the configuration,
the transmission from the base station to the users is considered the downlink, while the
transmission from the users to the base station is termed the uplink and it shows the base
station with M antennas talking with multiple single-antenna users. MIMO (multiple-input
multiple-output) is a radio antenna system that uses multiple antennas at both the transmitter
and receiver to improve signal quality, throughput and capacity. It accomplishes this through
two major strategies, spatial diversity and spatial multiplexing techniques. MIMO spatial
diversity methods convey separate and independently encoded data signals called streams
over the same time and frequency resource, resulting in increased data capacity. Spatial
multiplexing techniques employ the same time and frequency resources to broadcast several
data streams. In massive MIMO, the transmitter sends several streams to different users at
the same time and frequency increasing network capacity. Adding more antennas to handle
more streams can boost spectral efficiency and capacity, but power sharing and interference
among users result in falling benefits and eventually, losses(Kansal et al., 2019).

Figure 2. 1(a) (Kansal et al., 2019) Downlink (b) Uplink

Traditional MIMO systems use a small number (usually two) of antennas to broadcast and
receive signals in several directions at the same time. Figure 2.2 depicts how the base station
employs antennas to direct signals to the specified receiver in the downlink and divide them
for uplink transmission. To reduce interference and enhance signal-to-noise ratio (SNR),
consider using multiple antennas rather than delivering numerous data streams. As a result,
11 | P a g e
in a Massive MIMO system the base station employs its huge number of antennas to down-
link transmission and route the signal to the preferred receiver via techniques such as beam-
forming. This reduces interference and increases the signal-to-noise ratio (SNR) at the re-
ceiver. Additionally, uplink reception divides the numerous signals from different users,
allowing for effective uplink transmission. This spatial segmentation of users aids in the
management of interference and improves overall system performance.

The MIMO system provides enhanced wireless connectivity without the need for more spec-
trum, giving it a substantial advantage over traditional networks. Finally, massive MIMO
can expand the capacity of a communication system by supporting more users with the same
amount of resources(Ali et al., 2018).

Figure 2. 2 (Ali et al., 2018) Uplink and Downlink transmission in a Massive MIMO
system.

12 | P a g e
2.2 Related works.
In (Abebaw et al., 2023), analyze the spectral efficiency of an uplink multi-cell massive
MIMO cellular communication system under fading channels. The paper proposes a model
for evaluating area throughput using the linear zero forcing uplink combining scheme and
considers the impact of different factors such as pilot reuse factor, number of users, number
of base station antennas and propagation conditions on spectral efficiency. The paper also
compares the performance of different channel estimators including LS, MMSE, and EW-
MMSE and considers the impact of line-of-sight and non-line-of-sight communication over
Rician fading channels. The paper may not extensively address the practical implementation
challenges and considerations associated with deploying the proposed model and findings
in real-world cellular communication systems.

In (Satya et al., 2023), a novel NOMA-Massive MIMO scheme are introduced to achieve
higher spectral and power efficiency in 5G networks. In this paper where NOMA allows
multiple users to share the same time-frequency resources by allocating different power
levels and decoding orders, while massive MIMO utilizes a large number of antennas at the
base station to exploit spatial diversity and multiplexing gain. By integrating NOMA and
massive MIMO with successive interference cancellation(SIC) and zero forcing (ZF) tech-
niques, the proposed scheme aims to mitigate multiuser interference and complexity chal-
lenges. While the paper mentions the use of successive interference cancellation(SIC) and
zero forcing(ZF) techniques to mitigate multiuser interference. The effectiveness of these
methods in dynamic and interference-laden environments not extensively discussed.

In (Abraham, 2022), the paper presents a comparative analysis of the performance of a mas-
sive MIMO base station using a correlated Rician fading model and an uncorrelated model.
A statistical framework is employed, demonstrating that low-directivity antenna elements
outperform high-directivity elements in angularly constrained Rician fading with off-axis
incidence. Additionally, the paper introduces a method for handling correlated Rician fading
in large-scale antenna systems, by providing accurate approximations. The paper facilitates
effective analysis of the channel characteristics of a massive MIMO base station. Overall,
the proposed model allows for a straightforward assessment of the performance of corre-
lated Rician radio channels, considering factors such as antenna element pattern, array lay-
out and Rician fading channels.

13 | P a g e
In (Shi et al., n.d.), the focus of this study is on the utilization of the minimum mean square
error (MMSE) channel estimator for estimating the channels of aggregated reconfigurable
intelligent surfaces (RIS). The paper presents an analysis of the uplink spectral efficiency in
a cell-free massive MIMO system that incorporates spatially correlated RIS and multi-an-
tenna access points (AP). Numerical results are provided to validate the analysis and assess
the system's performance, considering the spatial correlation of RIS elements and AP an-
tennas. The proposed techniques are compared with local minimum mean squared error (L-
MMSE) and maximum ratio (MR) combining methods. The comparative results indicate
that L-MMSE combining outperforms MR combining when there is a large number of RIS
elements. Furthermore, the number of AP antennas has a more pronounced impact on sys-
tem performance when there are fewer RIS elements, while the spatial correlation of RIS
elements has a more significant negative effect on system performance compared to that of
the AP antennas.

In (Zaheera, 2022), the paper examines the spectral efficiency of a cell-free massive multi-
ple-input multiple-output (MIMO) system that operates in the presence of Rician fading
channels. The analysis considers the impact of the random phase in the line-of-sight (LoS)
path, which accounts for phase-shifts caused by mobility and phase noise. The study also
takes into consideration the availability of prior information at the access points (APs) and
compares the performance of different techniques including regularized zero-forcing (RZF),
phase-aware minimum mean square error (MMSE), non-aware linear MMSE (LMMSE)
and least-square (LS) estimators. The comparison reveals that the performance degradation
resulting from the absence of phase information is influenced by the length of the pilot signal
and remains minimal when pilot contamination is low.

In (Chaves et al., 2022), the purpose of this work is to investigate the computational com-
plexity of various user selection algorithms and evaluate their performance in various sys-
tem configurations while taking into account perfect and partial channel state information
(CSI) under the LoS propagation channel. In addition, the research introduces and examines
the computational complexity of ICIBS, a new user selection technique. Finally, the paper
provides practical guidelines for massive MIMO system design, as well as a summary of
the main findings of massive MIMO systems under LOS propagation. But, the research only
addresses the LoS propagation model, which may not be representative of all practical cases
and also does not give a comprehensive comparison of the proposed ICIBS algorithm with
other state-of-the-art user selection algorithms.

14 | P a g e
In (Ueoka et al., 2022), analysis the performance of cell-free massive MIMO networks un-
der different channel fading, specifically Rayleigh and Rician fading. In this paper it is used
a simulation-based approach to investigate the performance of cell-free massive MIMO net-
works under different channel fading. Overall, the challenges of achieving reliable channel
state information (CSI) in outdoor scenarios, where the low degrees of channel hardening
can lead to poor spectral efficiency performance and the results indicate that Cell-Free sys-
tems achieve higher spectral efficiency when the channel behaves as a Rician fading and
that statistical channel state information(CSI) is unreliable in outdoor scenarios. The study
recommends alternative downlink (DL) channel estimation methods, such as blind estima-
tion or DL pilot-based, to improve spectral efficiency in outdoor scenarios. But, the study
only considers the downlink (DL) spectral efficiency (SE) and does not investigate other
performance metrics, such as uplink (UL) SE or energy efficiency.

In (Amadid et al., 2022), the Channel Estimation (CE) process in a Massive MIMO system,
particularly during the up-link (UL) phase is explained. This study compares the perfor-
mance of various channel estimation strategies, such as the Minimum Mean Square Error
(MMSE) estimator, the Least Squares (LS) estimator and the Modified MMSE (M-MMSE)
estimator, for both uncorrelated and correlated channels. Also proposed was a Uniform Pla-
nar Array (UPA) configuration for correlated channels, which delivers greater performance
than the typical Uniform Linear Array (ULA) arrangement. Overall, the research finds that
the proposed methodologies can greatly improve channel estimation in Massive MIMO sys-
tems, emphasizing the relevance of spatial correlation in an urban network. However, the
research does not address the effect of other forms of fading channels, such as Rayleigh or
Rician fading on the performance of the suggested channel estimating algorithms. Further-
more, the research does not address the impact of the various types of pilot sequences on
the performance of the channel estimation algorithms.

In (Naraiah et al., 2021), focuses on the issues of improving spectral efficiency in wireless
communications to meet the needs of 5G technology. The research examined certain beam
management approaches in millimeter-wave communications for 5G networks. This evalu-
ation could entail determining the efficiency of beamforming technologies in enhancing
spectral efficiency, capacity and coverage in wireless networks. The comparison of different
methodologies for enhancing spectral efficiency in Massive MIMO systems has a limited
scope, perhaps omitting alternative approaches or strategies that could be advantageous for
5G communications.

15 | P a g e
In (Fading et al., 2020), the analysis focuses on analyzing the performance of huge MIMO
systems with users equipped with numerous antennas taking into account a generalized
channel. The study considers a channel model that contains both line-of-sight propagation
and spatially correlated multipath fading. To examine spectrum efficiency, the study em-
ploys the Weichselberger model, which allows for the inclusion of multiple representations
such as Kronecker and virtual channel models as special cases. According to the study, in
the case of jointly correlated fading, uplink spectrum efficiency does not increase forever as
the number of base stations increases. This constraint can be solved by providing precoding
at the user side, although this technique introduces additional overhead. In the scenario of
omnidirectional uplink transmissions, the negative impact of the joint correlation structure
of the channel can be mitigated by serving users with a single antenna.

In (Sadraei et al., 2020), it examines the instantaneous spectrum efficiency of multi-cell

large MIMO systems using correlated Rician channels. The work suggests closed-form ap-
proximations for the ergodic spectrum efficiency of the Maximum Ratio Combining (MRC)
detector in the presence of Least Squares (LS) and Minimum Mean Square Error (MMSE)
estimators. The analysis focuses on the scenario of poor channel state information (CSI). A
novel approach is presented for analyzing the spectrum efficiency of correlated Rician chan-
nels with poor CSI. In this method, each user's Signal-to-Interference-Plus-Noise Ratio
(SINR) is determined using the equivalent channel, which is created by cascading the esti-
mated channel and the detector. This technique uses all available CSI information. Overall,
the paper presents a comprehensive analysis of the spectrum efficiency for correlated Rician
channels in the presence of imperfect CSI, employing a novel methodology that incorporates
the estimated channel and the detector to calculate the SINR of each user.

In (Wikiman & Idowu-bismark, 2019), it analyzes the performance of large MIMO systems
in a Rician channel using a zero-forcing precoder to improve spectral efficiency and capac-
ity. The study's goal is to validate large MIMO technology's ability to thrive over several
fading channels while also exploiting spatial diversity to increase spectral efficiency. The
research uses simulations and analysis to demonstrate that huge MIMO systems can achieve
high throughput in a Rician channel. The adoption of a zero forcing precoder reduces inter-
ference and increases the system's spectral efficiency. The results confirm the hypothesis
that massive MIMO technology may effectively leverage spatial diversity to increase per-
formance in difficult and fading conditions. Overall, the study provides evidence that mas-

16 | P a g e
sive MIMO systems can deliver significant enhancements in throughput and spectral effi-
ciency particularly in Rician fading channels, thus validating the potential and expectations
associated with this technology.

In (Rajmane & Sudha, 2019), it looks into improving spectral efficiency in 5G wireless
communication using huge Multiple Input Multiple Output (MIMO) devices. The work fo-
cuses on downlink spectral efficiency analysis in a single circular cell scenario, assuming
that the base station has previous knowledge of channel state information (CSI) and applies
linear precoding techniques. A comparison is provided between the spectral efficiency per-
formance of Uniform Linear Array (ULA) and sub-ULA configurations, emphasizing the
advantages of various array arrangements. The study focused on suppressing inter-user in-
terference through the optimization of antenna array topologies and linear precoding tech-
niques. By reducing interference, the system's spectral efficiency was improved. The paper
has made simplifying assumptions in the modeling and analysis such as perfect Channel
State Information (CSI) at the base station, which may not fully represent real-world sce-
narios where CSI estimation errors exist and while the paper compares ULA and sub-ULA
configurations, it may not have explored a wide range of antenna array configurations, po-
tentially overlooking other configurations that could offer improved spectral efficiency.

In (Shlezinger et al., 2019), focuses on the spectral efficiency (SE) of uplink massive MIMO
systems that allow Base Stations (BSs) to collaboratively decode received signals. This re-
search presented three decoding schemes, collaborative decoding of signals from related
user terminals (UTs) at each BS, decoding signals from all (UTs) in the network at each BS
and dividing the data transmission phase between cells to eliminate inter-cell interference.
The research also discussed an optimized network that combines these approaches to in-
crease SE by allowing BSs to decode some inter-cell interference while considering the rest
as noise. Simulation tests show that allowing BSs to do joint decoding results in higher SE
gains than separate linear decoding and that the usual strategy of separate linear decoding is
inadequate especially when interference is prevalent. The paper primarily considers non-
cooperative uplink massive MIMO systems without delving into the complexities of practi-
cal deployment challenges such as hardware constraints, synchronization issues, pilot con-
tamination and feedback overhead. These real-world factors can significantly impact the
performance of massive MIMO systems but are not extensively addressed in the paper. Ad-
ditionally, its limitations lie in the lack of extensive empirical validation consideration of

17 | P a g e
practical deployment challenges and the idealized assumptions made in the analysis. Ad-
dressing these limitations in future research could enhance the applicability and relevance
of the findings in real-world wireless communication systems.

In (Abid et al., 2019), a low-complexity detector based on a two-layer linear receiver struc-
ture was proposed to prevent interference and improve uplink performance analysis for mas-
sive MIMO. The suggested two-layer processing techniques have much reduced computa-
tion complexity than the M-MMSE receiver, as demonstrated in this study. Analytical ex-
pressions for residual interference power at the output of the first-layer M-MMSE pro-
cessing are studied, considering intra-cell and inter-cell interference, and the effects of
small-scale fading, large-scale losses (e.g., path loss, shadow fading), and the number of
antenna elements in each subset on residual interference power. But, the paper primarily
considers the performance of the proposed schemes in the context of specific system param-
eters and scenarios, such as a limited number of interferers and a fixed antenna array size.
The generalizability of the results to diverse network configurations and operating condi-
tions may be limited, and further investigation under varying interference levels, antenna
array sizes, and network topologies could provide a more comprehensive understanding of
the proposed techniques effectiveness. Additionally, the paper focuses on uplink perfor-
mance analysis, and the implications of the proposed linear processing schemes on downlink
transmission and overall system efficiency are not extensively explored. Investigating the
impact of the proposed schemes on downlink communication, resource allocation, and sys-
tem-level performance optimization could offer a more holistic view of their benefits and
limitations in practical massive MIMO deployments.

In (Wu et al., 2017), to increase channel estimation accuracy in the presence of pilot con-
tamination, two algorithms are proposed, one based on LOS components and other based
on data-aided iteration. This paper presents a novel channel estimation scheme that effec-
tively addresses pilot contamination in massive MIMO systems operating in Rician fading
channels. By leveraging the statistical information of channels and exploiting the property
that the LOS component is not affected by pilot contamination with a large number of base
station antennas, the proposed scheme accurately estimates the LOS components of users
served by a base station. The suggested approach assumes that the Line-of-Sight (LOS)
components are present between base stations and serviced users, but no LOS components
exist between base stations and users from other cells. This assumption may not always hold
in practice and also compares the proposed scheme to existing schemes in terms of spectral

18 | P a g e
efficiency, a broader range of performance metrics such as latency, energy efficiency, or
robustness to varying channel conditions could provide a more comprehensive evaluation.

Table 2. 1: A Summary of some related work.

Authors Year Title Techniques and Concept Gaps or Concept

Covered not Covered
Abebaw 2023 Spectral Effi- The paper uses a simula- may not extensively
et al ciency Analysis tion-based approach to address the practi-
for Uplink evaluate the spectral effi- cal implementation
Multi-cell Mas- ciency of an uplink multi- challenges and con-
sive MIMO Cel- cell massive MIMO cellular siderations associ-
lular Communi- communication system un- ated with deploying
cation System der fading channels the proposed model
under Fading
Channels
Jens 2022 Statistics of the The statistical framework This paper focuses
Abraham Effective Mas- are proposed for perform only on angular
sive MIMO high-directivity elements in constricted Rician
Channel in Cor- angular constricted Rician fading. It doesn’t
related Rician fading with off-axis inci- include spherical
Fading dence wave model Rician
fading
Shi et al., 2022 Uplink Perfor- Minimum mean square er- Even though it im-
n.d mance of RIS- ror (MMSE) channel esti- proves
aided Cell-Free mator is Proposed to esti- the performance
Massive MIMO mate the aggregated RIS with a
System Over channels. large number of
Spatially Corre- RIS elements, the
lated Channels proposed technique
is not robust to the
beamforming de-
sign to enable the
implementation of

19 | P a g e
 RIS-aided CF M-
MIMO networks.

Fathima 2022 Large-scale In this paper A combined This paper focuses

Zaheera Fading Decod- scheme of RZF only on Rayleigh
ing based Mas- equalizer-based inter cell fading channel. It
sive MIMO interference mitigation ap- doesn’t include
System with proach is employed for en- other types of Fad-
Regularized hancing the performance of ing like Racian fad-
Zero Forcing Massive ing channel
Precoder MIMO system
Fading et 2020 Massive MIMO It proposed to analysis of It is only analyzed
a with Multi-An- spectrum efficiency utilizes the performance
tenna Users un- the Weichselberger model of multi-cell mas-
der Jointly Cor- and considers a channel sive MIMO with
related Rician model that incorporates single-antenna us-
Fading line-of-sight propagation ers under
and spatially correlated correlated Rician
multipath fading fading and pilot
contamination

20 | P a g e
 CHAPTER THREE
MATERIAL AND METHODS
3.1 Overview
This thesis reviews several related papers and articles from publications such as IEEE,
Springer, and Elsevier, as well as books on Massive MIMO systems. Based on the studied
publications and with the problem statement in mind, the methodology and flow chart given
in section 3.3 are followed to build and simulate the proposed methods/techniques.

3.2 Material
This study focused on the examination and mitigation of spatial correlated fading in massive
MIMO systems. The MATLAB/R2023 software platform is used to investigate and mitigate
spatial correlated fading in massive MIMO systems. MATLAB is a high-performance pro-
gramming language for technical computing. It is also computing software, which includes
technical boxes and Simulink. Developed algorithms using MATLAB, analyzed data and
created models.

3.3 Methods of the study

To accomplish the above-mentioned general and specific objectives of the thesis, proceed
through the following methodology and flow chart as shown in figure 3.1 and figure 3.2,
respectively.

Literature Review: the thesis work is based on prior research on the subject by experts or
scholars. So, by examining the works of these academics and acquiring relevant and im-
portant knowledge about the proposed thesis work. It is reviewed using a range of sources,
such as books, journals, papers, the internet, and so on. As a result, the literature review
includes reading articles, conference papers, books, journals, and exploring the internet for
relevant topics (i.e. the impact of spatial correlated fading channels on a Massive MIMO
system).

Problem Identification: based on examined papers from different publications, journals,

and conferences, the research gaps identified for research inquiry.

System Designing: proceed to system designed after identification of research gap for thesis
work.

21 | P a g e
Simulation of designed System using MATLAB: after modeled of proposed system de-
signed, simulated using MATLAB tools.

Result and Discussion: based on simulation results, the performed discussions are done on
the result and also performance analyzed are done on proposed techniques.

Documentation: After analyzed the results of the designed system, the whole work of the
research study is concluded by suggesting an implied recommendation and future work to
be done on the areas.

Literature Reviewer

Identifying Research Problem

System designing

Simulation of designed Sys-

tem using MATLAB

Result and Discussion

Documentation

Figure3. 1 Methodology.
Figure 3. 1 Methodology.

22 | P a g e
 start

Determine the SE of each chan-

nel estimator through precoder
Trigger off the parameter (M, i,
𝜏p, 𝜏c, 𝜏u, BW, f, fc, ASD…...)

Compare the SE of MMSE,

User transmits uplink training EW-MMSE and LS channel es-
signal to BS antennas timators varying pilot reuse
factor

By using MMSE, EW-MMSE

and LS channel estimator esti-
mate CSI
Simulate the SE of each chan-
nel estimator using different
BS antennas, Ues, coherence
BS determines the uplink vec- block length & SNR
tor for precoding uplink com-
bining.

Examine the SNIR for each

channel estimator using precod-
ing receiver combining

Determines NMSE for

each estimator

Is it NMSE minimized
No Is it SNIR maximized Yes

Figure 3. 2 Flow chart of Spectrum Efficiency analysis

23 | P a g e
3.4 System Model for spatially correlated fading in massive MIMO.
This section introduces a system model created specifically for a multicellular huge MIMO
system running in Time Division Duplex (TDD). During the creation of the multicellular
massive MIMO system model, relevant channel models are carefully selected and designed.
The system model shown in Figure 3.2, accounts for the true characteristics of a multicellu-
lar massive MIMO system operating under Rician fading. This Time Division Duplex
(TDD) technique allows several cells to share the same time-frequency resources. All user
equipment concurrently uses the available frequency-time resources for both uplink pilot
and data transmission functions. Taken into consideration a Massive MIMO system with 𝑀
cells, where each cell is made up of one base station (BS) that supports 𝑖 single-antenna
user equipment (UEs) and has 𝑁𝑗 antennas. When the system is in TDD mode, the channel
responses throughout a coherence block of 𝜏𝑐 samples are constant. Furthermore, consider
for granted that there is no correlation between the channel realizations of any two coherence
blocks. The carrier frequency and extrinsic parameters like UE mobility and propagation
environment dictate the size of 𝜏𝑐 . The samples have three distinct uses, 𝜏𝑝 samples for pilot
signals in the uplink (UL) domain, 𝜏𝑢 samples for data transmission in the UL domain, and
𝜏𝑑 samples for data transmission in the downlink (DL) domain, where 𝜏𝑐 = 𝜏𝑝 + 𝜏𝑢 + 𝜏𝑑 .
The channels UL and DL are estimated by the received pilot signals and sends back the
channel state information (CSI) to the BS.

𝑗
The channel response denoted as ℎ𝑀𝑖 ∈ ℂ𝑁𝑗 , which represents the propagation from users
𝑖 in cell 𝑀 to the base station (BS) within the 𝑗 𝑡ℎ cell, is modeled using Rician fading chan-
𝑗
nels. Where the BS index is indicated by the superscript ℎ𝑀𝑖 , while the cell and UE indexes
are identified by the subscript. In this context, examined spatial correlated fading channels,
𝑗
where ℎ𝑀𝑖 , ∀𝑗, 𝑚 ∈ 1 … . 𝑀 and ∀𝑖 ∈ 1 … . 𝐼 represents a specific instance of the circularly
𝑗
symmetric complex Gaussian distribution. So The channel coefficient ℎ𝑀𝐼 between the 𝑗 𝑡ℎ
antenna at the base station and the 𝑁 𝑡ℎ antenna of the 𝑖 𝑡ℎ user can be modeled as a complex-
valued random variable.

𝑗 𝑗 𝑗
ℎ𝑀𝑖 ~ 𝑁𝐶 (ȟ𝑀𝑖 , 𝑅𝑀𝑖 ) (3.1)

𝑗
Where ȟ𝑀𝑖 ∈ ℂ𝑁𝑗 is corresponding to the LOS component with zero mean and unity vari-
𝑗
ance assumed to be 𝑁𝐶 (0, 𝐼𝑚 ) and 𝑅𝑀𝑖 ∈ ℂ𝑁𝑗 × 𝑁𝑗 is corresponding to the NLOS compo-
nent of positive semi-define covariance matrix of spatial correlation. The small scale fading
24 | P a g e
 𝑗 𝑗
is characterized by a Gaussian distribution, while ȟ𝑀𝑖 and 𝑅𝑀𝑖 represent the macroscopic
propagation effects, encompassing path loss, shadow fading, as well as the antenna gains
and radiation patterns at the transmitter and receiver. This statistical model provides a com-
pact way to describe the characteristics of the Massive MIMO channel, which is essential
for analyzing and designing efficient communication systems that leverage the spatial de-
grees of freedom offered by Massive MIMO architectures. Since there is a lot of channel
loss so the average channel gain from one of the antennas at BS 𝑗 to UE 𝑖 in cell 𝑀 is try to
determine by the normalized trace. Therefore, it is expressed as:

𝑗 1 𝑗
𝛽𝑀𝑖 = 𝑡𝑟( 𝑅𝑀𝑖 ) (3.2)
𝑁𝑗

𝑗
Where 𝛽𝑀𝑖 represents the combined effect of path loss and shadowing on the wireless chan-
nel between the base station and the user which called large scale fading coefficient

𝑗
ℎ𝑚𝑘

BS desired Cell j
√𝑝𝑚1 𝑥𝑚𝑖
MMSE,
𝑗 EW-MMSE,
ℎ𝑚𝑘
mth LS

√𝑝𝑚𝑖 𝑥𝑚𝑖 nj 𝑥𝑚1

̃
M-
MMSE
𝑗
ℎ𝑚𝑘 , RZF, 𝑥𝑚𝑖
̃
zz ∑ y
MR
√𝑝𝑗1 𝑥𝑗1 𝑥
̃𝑗𝑘

jth ℎ𝑚𝑘
𝑗

\
√𝑝𝑗𝑖 𝑥𝑗𝑖

Figure 3. 3 System Block Diagram Model for Uplink Massive MIMO.

25 | P a g e
3.5 Channel Estimation for spatially correlated fading in massive MIMO
Channel estimation is critical in Massive MIMO (Multiple-Input Multiple-Output) systems,
particularly where spatial correlation fading is a concern. Massive MIMO uses a base station
with a high number of antennas to serve several users at the same time, resulting in better
spectral efficiency and system capacity. However, due to the presence of spatial correlation,
channel estimation becomes more complicated. Spatial correlation is a statistical link be-
tween the channel coefficients of different antennas. In Massive MIMO systems, the chan-
nels between the base station antennas and user equipment’s (UEs) are often correlated due
to the proximity of antennas and similar propagation environments. This correlation can
affect the accuracy of channel estimation and subsequently degrade system performance if
not properly considered. Therefore, it need to estimating the channel state information (CSI)
between the base station and multiple users in a massive multiple-input multiple-output
(MIMO) system, taking into account the spatial correlation of the fading channels. To ad-
dress this issue, channel estimation algorithms for spatial correlation fading in Massive
MIMO systems must take into account the spatial correlation structure of the channels. This
can be accomplished by utilizing the known statistical properties of the spatial correlation,
such as Toeplitz or circulant structures, to design efficient channel estimation. These tech-
niques aim to accurately estimate the channel coefficients while taking into account the spa-
tial correlation effects to improve system performance. Therefore, the channel estimation
on this thesis considered under the uplink pilot based channel estimation 𝜏𝑝 sample in each
coherence block are reserved, and allowing space for 𝜏𝑝 pilot sequences that are mutually
orthogonal. A proper pilot assignment is a demanding task for minimizing interference in
the transmission channel. While it is critical to dedicate more pilots for accurate channel
estimation and to reduce pilot contamination, spectral efficiency suffers as more resources
during the coherence interval are devoted to pilots for estimation rather than the actual data
payload for transmission. These pilot sequences are assigned to different User Equipment
(UEs), and the same sequences are used by UEs across many cells. The predetermined pilot
sequence of UE 𝑖 in cell 𝑗 Consider 𝛷𝑗𝑖 as the pilot sequence for the 𝑖 𝑡ℎ UE in the 𝑗 𝑡ℎ cell.
Each element of 𝛷𝑗𝑖 is multiplied by the uplink transmit power √𝑝𝑗𝑖 and transmitted as the
signal 𝑥𝑗𝑖 over a duration of 𝜏𝑝 uplink samples. This results in the received uplink signal,

denoted as 𝑦𝑗𝑝𝑖𝑙𝑜𝑡 ∈ ℂ𝑁𝑗 × 𝜏𝑝 at BS 𝑗. Therefore, the expression for the received pilot signal
is as follows:

26 | P a g e
 𝑖𝑗 𝑀 𝑖𝑀
𝑗 𝑗 𝑗
𝑦𝑗𝑝𝑖𝑙𝑜𝑡 = ∑ √𝑝𝑗𝑖 ℎ𝑗𝑖 + ∑ ∑ √𝑝𝑀𝑘 ℎ𝑀𝑖 𝛷𝑀𝑘 + 𝑛𝑗𝑝𝑖𝑙𝑜𝑡 (3.3)
𝑖=1 𝑚=1 𝑘=1
𝑚≠1

Where 𝑛𝑗𝑝𝑖𝑙𝑜𝑡 ∈ ℂ𝑁𝑗 × 𝜏𝑝 consists of independent and identically distributed complex Gauss-
2 𝑗
ian elements with zero mean and variance 𝜎𝑢𝑙 . These element are used to estimate ℎ𝑀𝑖 ,

where BS 𝑗 multiplies 𝑦𝑗𝑝𝑖𝑙𝑜𝑡 by UE’s pilot sequence 𝛷𝑀𝑘

∗
can be express as:

𝑘𝑗
𝑗 𝑗 𝑗
𝑦𝑗𝑝𝑖𝑙𝑜𝑡 𝛷𝑚𝑖
∗
= √𝑝𝑗𝑖 ℎ𝑗𝑖 𝛷𝑗𝑖𝑇 𝛷𝑚𝑖
∗ ∗
+ ∑ √𝑝𝑗𝑖 ℎ𝑗𝑖 𝛷𝑗𝑖 𝛷𝑚𝑖
𝑖=1
𝑖≠𝑘
𝑀 𝑘
𝑗
𝑖
+ ∑ ∑ √𝑝𝑀𝑖 ℎ𝑀𝑖 ∗
𝛷𝑀𝑖 𝛷𝑀𝑖 + 𝑛𝑗𝑝𝑖𝑙𝑜𝑡 𝛷𝑚𝑖
∗
(3.4)
𝑚=1 𝑖=1
𝑚≠𝑗

In order to estimate the channel for the 𝑖 𝑡ℎ user within the cell using the signal mentioned
above, a projection of 𝑦𝑗𝑝𝑖𝑙𝑜𝑡 onto 𝛷𝑗𝑖∗ is performed to obtain 𝑦𝑗𝑚𝑖
𝑝𝑖𝑙𝑜𝑡
. The procedure can be
described as follows:

𝑖𝑗 𝑀 𝑖𝑚
𝑗 𝑗
𝑦𝑗𝑝𝑖𝑙𝑜𝑡 = √𝑝𝑗𝑖 ℎ 𝑗𝑖 𝛷𝑗𝑖𝑇 𝛷𝑗𝑖∗ + ∑ √𝑝𝑗𝑘 ℎ𝑗𝑘 𝑇 ∗
𝛷𝑗𝑘 𝛷𝑗𝑖 + ∑ ∑ √𝑝𝑘𝑚 ℎ𝑚𝑘 𝛷𝑗𝑖∗
𝑘=1 𝑚=1 𝑘=1
𝑘≠𝑖 𝑚≠𝑗
+ 𝑛𝑗𝑝𝑖𝑙𝑜𝑡 𝛷𝑗𝑖∗ (3.5)
𝑝𝑖𝑙𝑜𝑡 𝑗
𝑦𝑗𝑗𝑖 = √𝑝𝑖𝑘 ℎ𝑚𝑖 𝜏𝑝 + 𝑛𝑗 𝛷𝑗𝑖∗ (3.6)
𝑝𝑖𝑙𝑜𝑡 𝑝𝑖𝑙𝑜𝑡
Note that for all, 𝑦𝑗𝑗𝑘 = 𝑦𝑗𝑚𝑖 where (𝑀, 𝑖) ∈ 𝜗𝑗𝑘 , due to the utilization of the same pilot
for all these UEs. When operating within its own cell, there is a potential for significant
interference. Therefore, it is crucial to carefully assign pilots for the desired cellular network
in order to effectively utilize the desired pilot sequence. The value of 𝜗𝑗𝑘 is determined as
follows:

𝜗𝑗𝑘 = { (𝑚, 𝑖): 𝛷𝑚𝑖 = 𝛷𝑗𝑘 ; 𝑚 = 1, 2 … … 𝑀 𝑎𝑛𝑑 𝑖 = 1, 2 … . 𝐾𝑚 }

The UL-DL duality drives a straightforward precoding design principle select DL precoding
vectors depending on UL receive combining vectors as:
𝜗𝑗𝑖
𝚆𝑗𝑖 = (3.7)
‖𝜗𝑗𝑖 ‖
Where

27 | P a g e
 𝜗𝑗𝑖𝑀−𝑀𝑀𝑆𝐸 𝑤𝑖𝑡ℎ 𝑀 − 𝑀𝑀𝑆𝐸 𝑝𝑟𝑒𝑐𝑜𝑑𝑖𝑛𝑔.
[𝜗𝑗𝑖 … … … 𝜗𝑗𝑖𝑗 ] = { 𝜗𝑗𝑅𝑍𝐹 𝑤𝑖𝑡ℎ 𝑅𝑍𝐹 𝑝𝑟𝑒𝑐𝑜𝑑𝑖𝑛𝑔.
𝜗𝑗𝑀𝑅 𝑤𝑖𝑡ℎ 𝑀𝑅 𝑝𝑟𝑒𝑐𝑜𝑑𝑖𝑛𝑔.

Therefore, these precoding techniques depend on the UL transmit power utilized during
pilot signaling, but none of them depends on the DL transmit power (which however appears
in the DL effective SINR expression). One notable benefit of employing the receive com-
bining vectors for transmit precoding is that the computational complexity of computing the
precoding vectors reduced.

𝑝𝑖𝑙𝑜𝑡
The prepared gotten pilot flag 𝑦𝑗𝑚𝑖 ∈ ℂ𝑁𝑗 could be an adequate measurements for as-
𝑗
sessing ℎ𝑀𝑖 . presently consider channel estimation at the user end, each user receives the
pilot signals and estimates the channel response by using techniques such as minimum mean
square error estimation, element-wise minimum mean square error estimation, least squares
estimation.

3.5.1 Least Square Channel Estimator(LS)

The Least Square channel estimator works on the premise of minimizing the sum of squared
errors between the estimated channel response and the received signal. The Least Square
channel estimator estimates the channel response by solving a series of linear equations that
minimize the sum of squared errors between the received and estimated signals. This esti-
mate procedure is commonly conducted using matrix operations, which describe the re-
ceived signal and the channel response as matrices. In this channel estimator the transferred
data rate and pilots are signified by 𝑥 between user equipment and the base station.

𝑦 𝑝𝑖𝑙𝑜𝑡 = ℎ ∗ 𝑥 (3.8)

Where 𝑦 𝑝𝑖𝑙𝑜𝑡 is the received pilot signal, ℎ is the unknown channel from UEs to BS antenna,
and 𝑥 represents both the data rate and pilot length.

The unknown channels between UEs and BS antennas can be estimated using the received
pilot signal as:

ℎ = 𝑦 𝑝𝑖𝑙𝑜𝑡 ∗ 𝑥̅ (3.9)

28 | P a g e
 𝑗 𝑗
If the BS has no prior knowledge of 𝑅𝑚𝑖 and ℎ̅𝑚𝑖 , the non-Bayesian LS estimator is utilized
𝑗 𝑗
to estimate the propagation channel ℎ𝑚𝑖 . The LS estimate is the vector of ȟ𝑚𝑖 that minimizes
𝑗
‖𝑦 𝑝𝑖𝑙𝑜𝑡 − √𝑝𝑚𝑖 𝜏𝑝 ȟ𝑚𝑖 ||2. Which, in this case is:

𝑗 1 𝑝𝑖𝑙𝑜𝑡
ȟ𝑚𝑖 = 𝑦𝑗𝑚𝑖 (3.10)
√𝑝𝑚𝑖

The least square channel estimator and estimator error correlated random variable and dis-
tributed as:

𝑗 1 𝑝𝑖𝑙𝑜𝑡 1
ȟ𝑚𝑖 ~ 𝑁ℂ ( 𝑦̅𝑗𝑚𝑖 , ( 𝛷𝑚𝑖 )−1 (3.11)
√𝑝𝑚𝑖 𝜏𝑝 𝑝𝑚𝑖 𝜏𝑝

𝑗 𝑗 1 1 𝑗 𝑗
ℎ̃𝑚𝑖 ~ 𝑁ℂ ( ℎ̅𝑚𝑖 − 𝑦 𝑝𝑖𝑙𝑜𝑡 , ( 𝛷𝑚𝑖 )−1 − 𝑅𝑚𝑖 ) (3.12)
√𝑝𝑚𝑖 𝜏𝑝 𝑗𝑚𝑖 𝑝𝑚𝑖 𝜏𝑝

If reusing the chosen cellular system pilot increases capacity and reduces contamination,
simplifying the expression as below.

𝑗 1 𝑗 𝑗
ȟ𝑚𝑖 = (√𝑝𝑚𝑖 𝜏𝑝 ℎ𝑚𝑖 + ∑(𝑚,𝑖)\(𝑗,𝑚) √𝑝𝑚𝑖 𝜏𝑝 ℎ𝑚𝑖 + 𝑛𝑗 𝛷𝑗 ) (3.13)
√𝑝𝑚𝑖 𝜏𝑝

3.5.2 Minimum mean square error channel estimator (MMSE).

This channel estimator, estimate the channel response between the transmitter and receiver.
The MMSE channel estimator in massive MIMO systems with spatial correlation fading
seeks to reduce the mean square error between the estimated channel response and the true
channel response. As a result, channel estimators employ observed pilot signals to estimate
𝑗
an unknown channel parameter (h). The estimated channel value is defined ȟ𝑀𝑖 . The
MMSE channel estimator is designed to generate channel state information(CSI) at the re-
ceiver while minimizing interference errors. Using a pilot book that consists of mutually
orthogonal sequences, the Minimum Mean Squared Error (MMSE) estimation of the chan-
𝑗
nel ℎ𝑀𝑖 is performed based on the observation 𝑦𝑗𝑝𝑖𝑙𝑜𝑡 in (3.3).

𝑗 𝑝𝑖𝑙𝑜𝑡 𝑗 𝑗
ȟ𝑀𝑖 = √𝑝𝑀𝑖 𝑅𝑀𝑖 𝛷𝑀𝑖 𝑦𝑗𝑚𝑖 (3.14)

𝑗 𝑗 𝑗
An estimation error ℎ̃𝑀𝑖 = ℎ𝑀𝑖 − ȟ𝑀𝑖 has the correlation matrix is represented by:
𝑗 𝑗 𝑗
𝐶𝑀𝑖 = 𝛦{ℎ̃𝑚𝑖 (ℎ̃𝑚𝑖 )𝐻 } (3.15)
𝑗 𝑗 𝑗 𝑗 𝑗
It is also represented by: 𝐶𝑀𝑖 = 𝑅𝑀𝑖 − 𝑝𝑀𝑖 𝑅𝑚𝑖 𝛷𝑚𝑖 𝑅𝑚𝑖 (3.16)

29 | P a g e
This equation calculates the MMSE from any user equipment(UE) in the network to the
base station 𝑗. The MMSE estimator minimizes the MSE of the channel estimate, expressed
as:

𝑗 𝑗 𝑗 𝑗 𝑗 𝑗
𝛦{ || ℎ𝑚𝑖 − ȟ𝑚𝑖 || 2 } = 𝛦{|| ℎ̃𝑚𝑖 ||2 } = 𝛦{ 𝑡𝑟( ℎ̃𝑚𝑖 ( ℎ̃𝑚𝑖 )𝐻 } = 𝑡𝑟(𝐶𝑚𝑖 ) (3.17)
𝑗 𝑗
The MMSE estimate ȟ𝑚𝑖 and estimated error ℎ̃𝑚𝑖 are independent random variables with
the following distribution:
𝑗 𝑗 𝑗 𝑗
ȟ𝑚𝑖 ~ 𝑁ℂ (ℎ̃𝑚𝑖 , 𝑅𝑚𝑖 − 𝐶𝑚𝑖 ) (3.18)
𝑗 𝑗
ℎ̃𝑚𝑖 ~ 𝑁ℂ (0𝑀 , 𝐶𝑚𝑖 ) (3.19)
𝑗
The estimation error covariance matrix 𝐶𝑚𝑖 is independent of mean values or the estimation
error is unaffected by LoS components, as they may be removed from the received signals.

3.5.3 Element-Wise Minimum Mean Square Error Channel Estima-

tor(EW-MMSE).
The EW-MMSE channel estimator divides the channel response into individual elements,
which are estimated separately. This enables a more accurate estimation of the channel re-
sponse, particularly in scenarios where the channel is time-varying or frequency selective.
If the BS lacks information of the complete covariance matrix, the EW-MMSE estimation
can be employed as an alternative. The EW-MMSE estimate for the channel from BS 𝑗 to
UE 𝑖 in a specific cell 𝑀.

𝑗 𝑗 𝑗 𝑗 𝑝𝑖𝑙𝑜𝑡 𝑝𝑖𝑙𝑜𝑡
ȟ𝑚𝑖 = ℎ𝑚𝑖 + √𝑝𝑚𝑖 𝐷𝑚𝑖 Ʌ𝑚𝑖 (𝑦𝑗𝑚𝑖 − 𝑦𝑗𝑚𝑖 ) (3.20)
𝑗 𝑗 𝑗
Where 𝐷𝑚𝑖 ∈ ℂ𝑁𝑗 × 𝑁𝑗 and Ʌ𝑚𝑖 ∈ ℂ𝑁𝑗 ×𝑁𝑗 are diagonal matrices with 𝐷𝑚𝑖 =
𝑗 𝑗 𝑗
𝑑𝑖𝑎𝑔( [ 𝑅𝑚𝑖 ]𝑛𝑚 ∶ 𝑛 = 1 … … … . 𝑁𝑗 ) and Ʌ𝑚𝑖 = 𝑑𝑎𝑖𝑔( [ ∑(𝑚′ , 𝑖 ′ ) 𝑝𝑚′ ,𝑖 ′ 𝜏𝑝 𝑅𝑚′ ,𝑖′ +

𝜎 2 𝐼𝑁𝑗 ]𝑛𝑚 ∶ 𝑛 = 1 … … … . 𝑁𝑗 )

3.5.4 Uplink Spectral Efficiency using the MMSE Estimator.

In a wireless system's uplink communication scenario, spectral efficiency refers to the rate
at which data may be safely sent from user equipment (UE) to base station (BS) over a given
bandwidth. When the MMSE channel estimator is used in uplink communication, the goal
is to accurately predict the channel response between the UE and BS while accounting for
channel statistical characteristics and noise. Minimizing the mean square error between the

30 | P a g e
estimated and real channel responses. Generally when data transmitted, the received signal
𝑦𝑗𝑝𝑖𝑙𝑜𝑡 ∈ ℂ𝑁𝑗 at BS 𝑗 is

𝑖𝑗 𝑀 𝑘𝑚
𝑗
𝑦𝑗𝑝𝑖𝑙𝑜𝑡 = ∑ ℎ𝑗𝑖 𝑆𝑗𝑖 + ∑ ∑ ℎ𝑚𝑖 𝑆𝑚𝑖 + 𝑛𝑗 (3.21)
𝑖=1 𝑚=1 𝑖=1
𝑚≠𝑗

2
Where 𝑛𝑗 represents additive noise expressed as 𝑛𝑗 ~ 𝑁ℂ (0𝑚 ; 𝜎𝑢𝑙 𝐼𝑀𝑖 ) also the uplink sig-
nal from user 𝑖 in cell 𝑀 is expressed by 𝑆𝑀𝑖 ∈ ℂ and has power 𝑝𝑀𝑖 = 𝛦{|𝑆𝑀𝑖 |2 }. From
the above equation the first term represents desired signal, the second term represents inter-
ference and the last term represents noise. This section analyzes the possible SE of the UL
using the MMSE estimator with various receive combining techniques. The total UL capac-
ity of UE 𝑖 in cell 𝑗 can be calculated using the following formula

𝑖𝑗 𝑀 𝑖𝑚
𝑗 𝑗 𝑗 𝑗
𝘝𝑗𝑖𝑈𝐿 = 𝘝𝑗𝑖𝑈𝐿 ȟ𝑗𝑖 𝑆𝑖 + 𝘝𝑗𝑖𝐻 ℎ̃𝑗𝑖 𝑆𝑗𝑖 + ∑ 𝘝𝑗𝑖𝐻 ℎ𝑗𝑘 𝑆𝑗𝑖 ∑ ∑ 𝘝𝑗𝑖𝑈𝐿 ℎ𝑈𝑚𝑖 𝑆𝑚𝑖 + 𝘝𝑗𝑖𝑈𝐿 𝑛𝑗 (3.22)
𝑖=1 𝑚=1 𝑘=1
𝑖≠𝑘

Therefore, the spectrum efficiency of the uplink for 𝑀 cells and user equipment 𝑖 expressed
by:
𝜏𝑈
𝑆𝐸𝑗𝑖𝑈𝐿 = 𝛦{log 2 ( 1 + 𝑆𝐼𝑁𝑅𝑗𝑖𝑀 )} 𝑏𝑖𝑡𝑝𝑠/𝐻𝑧 (3.23)
𝜏𝐶

Where the effective SNR becomes:

𝑆𝐼𝑁𝑅𝑗𝑖𝑈𝐿
𝑗
𝑝𝑗𝑖 | 𝘝𝑗𝑖𝐻 ȟ𝑗𝑖 |2
= 𝑖𝑚 𝑗 𝑖
(3.24)
∑𝑀
𝑚=1 ∑ 𝑘=1 𝑝𝑈𝑚𝑖 | 𝘝𝑗𝑖𝐻 ȟ𝑗𝑖 |2 + 𝘝𝑗𝑖𝐻 ( ∑𝑀 𝑚 2
𝑚=1 ∑𝑈𝑀𝑖=1 𝑝𝑈𝑚𝑖 𝐶𝑈𝑚𝑖 + 𝛿𝑈𝐿 𝐼𝑀𝑗 )𝑉𝑗𝑖
(𝑈𝐿,𝑘)≠(𝑗,𝑖)

3.5.5 Uplink Spectral Efficiency using the EW-MMSE Estimator.

The EW-MMSE (Element- Wise Minimum Mean Square Error) estimator is a technique
used for channel estimation and equalization in communication systems. The EW-MMSE
estimator is an enhanced version of the MMSE estimator that incorporates exponential
weighting to improve the estimation accuracy in time-varying channels. It assigns higher
weights to recent channel observations and lower weights to older observations, which helps
adapt the estimation to changing channel conditions. When it comes to uplink spectral effi-
ciency using the EW-MMSE estimator, the basic idea remains the same as with the MMSE

31 | P a g e
estimator. The goal is to estimate the channel characteristics accurately to optimize the
equalization and decoding processes and maximize the spectral efficiency of the uplink
communication. It takes into account the time-varying nature of the channel and assigns
weights to different channel observations based on their recency. This helps in tracking and
adapting to changing channel conditions, resulting in improved spectral efficiency. To cal-
culate the uplink spectral efficiency using the EW-MMSE estimator from maximum Ratio
𝑗
receive combining vector 𝘝𝑗𝑖 = ȟ𝑗𝑖 then:

𝑗 𝑗 𝑗 𝑗
𝛦{ 𝘝𝑗𝑖𝐻 ℎ𝑗𝑖 } = 𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟( 𝐷𝑗𝑖 Ʌ𝑗𝑖 ) + ‖ℎ̅𝑗𝑖 ||2 (3.25)

𝑖
𝑗
𝛦{ ||𝘝𝑗𝑖 || } = 𝑡𝑟(∑ ) + ||ℎ̅𝑗𝑖 ||2
2
(3.26)
𝑗𝑖

𝑗 𝑈𝐿
𝛦{ |𝘝𝑗𝑖𝐻 ℎ𝑚𝑘 |2 } = 𝜒𝑚𝑖
𝑗
𝑗
𝑗 𝑗 𝑗 𝑗 𝑗 𝑗 𝑗 𝑗
= 𝑡𝑟( 𝑅𝑚𝑘 ∑ ) + ( ℎ̅𝑗𝑖 )𝐻 𝑅𝑚𝑘 ℎ̅𝑗𝑖 + ( ℎ̅𝑗𝑘 )2 ∑ ℎ̅𝑚𝑘 + | ( ℎ̅𝑗𝑖 )𝐻 ℎ̅𝑚𝑘 |2
𝑗𝑖
𝑗𝑖
𝑗 𝑗 𝑗 𝑗 𝑗 𝑗
𝑝𝑗𝑖 𝑝𝑚𝑘 ( 𝑡𝑟 ( 𝐷𝑚𝑘 Ʌ𝑗𝑖 𝐷𝑗𝑖 ) )2 + 2√𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑝 𝑡𝑟 ( 𝐷𝑚𝑘 Ʌ𝑗𝑘 𝐷𝑗𝑖 )
𝑗 𝑗
+ { 𝑅𝑒{ ( ℎ̅𝑗𝑖 )𝐻 ℎ̅𝑚𝑘 (𝑚, 𝑘) ∈ 𝑝𝑗𝑖
0 (𝑚, 𝑘 ) ∉ 𝑝𝑗𝑖 (3.27)

So when plugged these expressions into effective SINR, spectrum efficiency in case of EW-
MMSE is

𝑗 𝑗 𝑗 𝑗
𝑝𝑗𝑖 ( 𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟 ( 𝐷𝑗𝑖 Ʌ𝑗𝑖 𝐷𝑗𝑖 ) + || ℎ̅𝑗𝑖 ||2 )2
𝑈𝐿,𝐸𝑊−𝑀𝑀𝑆𝐸
𝑦 = 𝑖𝑚 𝑗 𝑗 𝑗
(3.28)
∑𝑀 𝑈𝐿 ̅𝑗 2 2
𝑚=1 ∑𝑘=1 𝑝𝑚𝑖 𝜒𝑚𝑘 − 𝑝𝑗𝑖 (𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟( 𝐷𝑗𝑖 Ʌ𝑗𝑖 𝐷𝑗𝑖 + ||ℎ𝑗𝑖 || )
𝑗 𝑗
2
+ 𝜎𝑈𝑙 (𝑡𝑟(∑𝑗𝑘) + ||ℎ̅𝑗𝑖 ||2

3.5.6 Uplink Spectral Efficiency using the LS Estimator

The LS estimator is a simple and computationally efficient method for estimating the chan-
nel response between the user equipment (UE) and base station (BS). By minimizing the
sum of squared errors between the received signal and the estimated channel response. It is
a straightforward method that involves solving a set of linear equations to estimate the chan-
nel response. This simplicity makes it computationally efficient and suitable for real-time
channel estimation, enabling quick adaptation to changing channel conditions and enhanc-
ing spectral efficiency in uplink communication. The uplink spectral efficiency of a massive

32 | P a g e
MIMO system is derived using the least-squares (LS) estimator. Therefore the maximum
1 𝑝𝑖𝑙𝑜𝑡
ratio combining with 𝘝𝑗𝑖 = 𝑦𝑗𝑗𝑖 are used based on least square estimator, then
√𝑝𝑗𝑖 𝜏𝑝

𝑗 √𝑝𝑚𝑘 𝑗 𝐻 𝑗
𝛦{ ||𝘝𝑗𝑖𝐻 ||2 } = 𝑡𝑟(𝑅𝑗𝑖 ) + ∑ (ℎ̅𝑚𝑘 ) ℎ̅𝑗𝑖 (3.29)
𝑝
√ 𝑗𝑖
(𝑚,𝑘)

1 𝑗 1 𝑝𝑖𝑙𝑜𝑡 2
𝛦{||𝘝𝑗𝑖𝐻 ||2 } = 𝑡𝑟(( 𝛷𝑗𝑖 )−1 ) + 2
||𝑦̅𝑗𝑗𝑖 || (3.30)
𝑝𝑗𝑖 𝜏𝑝 𝑝𝑗𝑖 𝜏𝑝

𝑗 𝑈𝐿,𝐿𝑆
𝑝𝑗𝑖 𝜏𝑝2 {| 𝘝𝑗𝑖𝐻 ℎ𝑚𝑘 |2 } = 𝑝𝑗𝑖 𝜏𝑝2 𝜒𝑚𝑘
𝑗 𝑗 𝑝𝑖𝑙𝑜𝑡 𝑗 𝑗 𝑝𝑖𝑙𝑜𝑡 𝐻 𝑗 𝑗
= 𝜏𝑝 𝑡𝑟(𝑅𝑚𝑘 (𝛷𝑗𝑘 )−1 + 2√𝑝𝑚𝑘 𝜏𝑝 𝑅𝑒{(𝑦̅𝑗𝑗𝑖 )𝐻 ℎ̅𝑚𝑘 𝑡𝑟(𝑅𝑚𝑘 ) + (𝑦̅𝑗𝑗𝑖 ) ℎ̅𝑚𝑘 𝑡𝑟(𝑅𝑚𝑘 )
𝑝𝑖𝑙𝑜𝑡 𝐻 𝑗 𝑗
+ (𝑦̅𝑗𝑗𝑖 ) 𝑅𝑚𝑘 ℎ̅𝑚𝑘
𝑗 𝑗 𝑗 𝑗 𝑗 𝑗
𝐻
𝑝𝑚𝑘 𝜏𝑝2 |𝑡𝑟(𝑅𝑚𝑘 )|2 + 𝜒̅𝑗𝑖 𝑅𝑚𝑘 𝜒̅𝑗𝑖 + 𝜏𝑝 (ℎ̅𝑚𝑘 )𝐻 (𝛺𝑗𝑖 )−1 ℎ̅𝑚𝑘 + |𝜒̅𝑗𝑖
𝐻 ̅
ℎ𝑚𝑘 |2
𝑗 𝐻 ̅𝑗 𝑗
+ 𝑝𝑚𝑘 𝜏𝑝2 ||ℎ̅𝑚𝑘 ||4 + 2√𝑝𝑚𝑘 𝜏𝑝 𝑅𝑒{𝜒̅𝑗𝑖 ℎ𝑚𝑘 ||ℎ̅𝑚𝑘 ||2 } ( 𝑚, 𝑘) ∈ 𝑝𝑗𝑖
+ 𝑝𝑖𝑙𝑜𝑡 𝐻 𝑗 𝑝𝑖𝑙𝑜𝑡 𝑗 𝑗 𝑗 𝑝𝑖𝑙𝑜𝑡 𝐻 ̅ 𝑗
( 𝑦̅𝑗𝑗𝑖 ) 𝑅𝑚𝑘 𝑦̅𝑗𝑗𝑖 + 𝜏𝑝 ( ℎ̅𝑚𝑘 )𝐻 (𝛷𝑗𝑖 )−1 ℎ̅𝑚𝑘 + |(𝑦̅𝑗𝑗𝑖 ) ℎ𝑚𝑘 |2 (𝑚, 𝑘) ∉ 𝑝𝑗𝑘

{ (3.31)

𝑝𝑖𝑙𝑜𝑡 𝑗 𝑗 𝑗 𝑗
Where 𝜒̅𝑗𝑖 = 𝑦̅𝑗𝑗𝑖 − √𝑝𝑚𝑘 𝜏𝑝 ℎ̅𝑚𝑘 and (𝛺𝑗𝑖 )2 = (𝛷𝑗𝑖 )−1 − 𝑝𝑚𝑘 𝜏𝑝 𝑅𝑚𝑘 Therefore,
plugged these expression into effective 𝑆𝐼𝑁𝑅 got:

𝑦𝑗𝑖𝑈𝐿,𝐿𝑆
𝑗 √𝑝𝑚𝑘 ̅ 𝑗 𝐻 ̅ 𝑗 2
𝑝𝑗𝑖 |𝑡𝑟(𝑅𝑗𝑖 ) + ∑(𝑚,𝑘)∈𝑝𝑗𝑖 (ℎ𝑚𝑘 ) ℎ𝑗𝑖 |
√𝑝𝑗𝑖
= (3.32)
√𝑝𝑚𝑘 ̅ 𝑗 ̅ 𝑗 2
(ℎ )ℎ | +
𝑖𝑚 𝑈𝐿,𝐿𝑆 𝑗 √𝑝𝑗𝑖 𝑚𝑘 𝑗𝑖
∑𝑀
𝑚=1 ∑𝑘=1 𝑝𝑚𝑘 𝜒𝑚𝑘 − 𝑝𝑗𝑖 |𝑡𝑟 (𝑅𝑗𝑖 ) + ∑(𝑚,𝑘)∈𝑝𝑗𝑖 𝑗
2
𝑡𝑟((𝛷𝑗𝑖 ||𝑦̅𝑗𝑖𝑝 ||2
𝜎𝑈𝑙 ( 𝑝𝑗𝑖 𝜏𝑝 + 𝑝𝑗𝑖 𝜏𝑝2 )

To derive the Rayleigh fading equivalent of above equation, the mean vectors set to be zero.
The MMSE and LS/EW-MMSE estimators may have a minimal difference in spectral effi-
ciency (SE). However, when employing the LS estimator with Rician fading, there can be a
significant decrease in SE, which is influenced by the strength of line-of-sight (LoS) paths.
Not incorporating mean values as prior knowledge results in higher interference terms in
comparison to the MMSE estimator. The LS estimates of the UEs affected by pilot contam-

33 | P a g e
ination are equivalent up to a scaling factor. In contrast to the MMSE estimator, the magni-
𝑝𝑖𝑙𝑜𝑡
tude of the corresponding interference terms is determined by the inner product of 𝑦̅𝑗𝑗𝑖
𝑗 𝑗 𝑗
and ℎ̅𝑚𝑘 , rather than the inner product of ℎ̅𝑗𝑖 and ℎ̅𝑚𝑘 .

3.5.7 Downlink Spectral Efficiency using the MMSE Estimator

The downlink spectral efficiency in a wireless communication system is the rate at which
data may be successfully sent from the base station (BS) to the user equipment (UE) over a
given bandwidth. The downlink spectral efficiency (SE) can be calculated using the minimal
mean square error (MMSE) estimation. The MMSE estimator is used to estimate the sent
symbols at the receiving end while taking into consideration channel circumstances. To
compute the SE, consider a wireless communication system with a fixed bandwidth and a
set of users. The MMSE estimator uses the received signal, channel state information, and
noise characteristics to estimate the broadcast symbols with the lowest mean square error.
The SE can be expressed as the achievable data rate per unit bandwidth. Mathematically, it
can be calculated from maximum ratio combining, through signal transmitting from Bas-
tion(BS) 𝑀 is :

𝑘𝑚

𝜒𝑀 = ∑ 𝚆𝑚𝑖 𝜍𝑚𝑖 (3.33)

𝑖=1

Where in this equation, 𝜍𝑚𝑖 ~ 𝑁ℂ (0, 𝑝𝑚𝑖 ) represents the downlink data signal intended for
UE 𝑖 in the cell, with 𝑝𝑚𝑖 being the signal power allocated to that UE. The transmit precod-
ing vector 𝚆𝑚𝑖 determines the spatial directivity of the transmission, and it satisfies the
condition 𝛦||𝚆𝑚𝑖 ||2 = 1, ensuring that the expected power allocated to UE 𝑖 is 𝑝𝑚𝑖 , which
can be expressed as 𝛦{||𝚆𝑚𝑖 𝜍𝑚𝑖 ||2 } = 𝑝𝑚𝑖 . This constraint helps in managing the transmit
power allocated to each UE efficiently. The received signal at UE 𝑖 in cell 𝑗 denoted as 𝑦𝑗𝑖 ∈
ℂ can be expressed as:

𝑖𝑚 𝑀 𝑖𝑚
𝑗
𝑦𝑗𝑖 = (ℎ𝑗𝑖𝑚 )𝐻 𝚆𝑗𝑖 𝜍𝑗𝑖 + ∑(ℎ𝑗𝑖 )𝐻 𝚆𝑗𝑘 𝜍𝑗𝑘 + ∑ ∑(ℎ𝑗𝑖𝑚 )𝐻 𝚆𝑚𝑘 𝜍𝑚𝑘 + 𝑛𝑗𝑖 (3.34)
𝑘=1 𝑚=1 𝑘=1
𝑘≠𝑖 𝑚≠𝑗

Where 𝑛𝑗𝑖 denotes the additive noise component which is assumed to be complex Gaussian
2 ).
with zero mean that is 𝑛𝑗𝑖 ~ 𝑁ℂ (0, 𝜎𝑑𝑙 In the equation mentioned (3.34), the terms have
the following interpretations:

34 | P a g e
The first term represents the desired signal, which is the signal intended for UE 𝑖 in the cell,
the second term corresponds to the intra-cell interference. It is the combined effect of signals
from other UEs within the same cell. This interference arises due to the simultaneous trans-
mission of multiple UEs in the cell and is represented. the third term represents the inter-
cell interference. It accounts for the interference caused by signals from neighboring cells.
This interference arises due to the limited spatial separation between cells and the overlap-
ping coverage areas. The ergodic downlink capacity of UE 𝑖 in cell 𝑗 has a lower bound,
which can be expressed as follows:

𝜏𝑑
𝑆𝐸𝑗𝑖𝐷𝐿 = log 2 ( 1 + 𝑦𝑗𝑖𝐷𝐿 ) (𝑏𝑖𝑡/ 𝑠⁄𝐻𝑧) (3.35)
𝜏𝑐

Where the effective SINR becomes:

𝑗
𝑝𝑗𝑖 |𝛦{𝚆𝑗𝑖𝐻 ℎ𝑗𝑖 }|2
𝑦𝑗𝑖𝐷𝐿 = 𝑖𝑚
(3.36)
∑𝑀 𝐻 𝑚 2 𝐻 𝑗 2 2
𝑚=1 ∑𝑘=1 𝑝𝑚𝑘 𝛦{|𝚆𝑚𝑘 ℎ𝑗𝑖 | } − 𝑝𝑗𝑖 |𝛦 {𝚆𝑗𝑖 ℎ𝑗𝑖 } | + 𝜎𝐷𝐿

By using the MMSE estimator, can obtained a mathematical expression for the downlink
SE that does not require further calculations or approximations. This closed-form expression
allows for a more straightforward evaluation of the SE in the downlink system. So a closed
form expression presented in the above equation can derive according to the following the-
𝑗
ȟ𝑗𝑖
orem if maximum ratio(MR) precoding employed with 𝚆𝑗𝑖 = is used based on
𝑗
√𝛦{|| ȟ𝑗𝑖 ||2 }

the MMSE estimator then:

𝑗 𝑗 𝑗 𝑗 𝑗
𝛦{𝚆𝑗𝑖𝐻 ℎ𝑗𝑖 } = √𝑝𝑗𝑖 𝜏𝑑 𝑡𝑟(𝑅𝑗𝑖 𝛷𝑗𝑖 𝑅𝑗𝑖 ) + ||ℎ̅𝑗𝑖 ||2 (3.37)

𝑚 2
𝛦{||ȟ𝑚 2 𝐻
𝑚𝑘 || } 𝛦{|𝚆𝑚𝑘 ℎ𝑗𝑖 | }
𝑗
𝑚 𝑚
= 𝑝𝑚𝑘 𝜏𝑑 𝑡𝑟(𝑅𝑗𝑘 𝑚
𝑅𝑚𝑘 𝛷𝑚𝑘 𝑅𝑗𝑖 + 𝑝𝑚𝑘 𝜏𝑝 (ℎ̅𝑗𝑖𝑚 )𝐻 𝑅𝑚𝑘
𝑚 𝑚
𝛷𝑚𝑘 𝑅𝑚𝑘 𝑅𝑗𝑖 + |(ℎ̅𝑗𝑖𝑚 )𝐻 ̅ℎ𝑚𝑘
𝑚 ̅𝑚 𝑚 2
|

+ (ℎ̅𝑚𝑘
𝑚 𝐻 𝑚 ̅𝑚
) 𝑅𝑗𝑖 ℎ𝑚𝑘

𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑑2 |𝑡𝑟(𝑅𝑗𝑖𝑚 𝛷𝑚𝑘

𝑚 𝑚
𝑅𝑚𝑘 )|2 + 2√𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑑 𝑅𝑒{ 𝑡𝑟( 𝑅𝑗𝑖𝑚 𝛷𝑚𝑘
𝑚 𝑚
𝑅𝑚𝑘 )
+ { ( ℎ̅𝑚𝑘 )𝐻 ℎ̅𝑗𝑖
𝑚 𝑚
(𝑚, 𝑘) ∈ 𝑝𝑗𝑖 (3.38)
0 (𝑚, 𝑘) ∉ 𝑝𝑗𝑖

Where 𝛦{ ||ȟ𝑚 2 𝑚 𝑚 𝑚 ̅𝑚 2
𝑚𝑘 || } = 𝑝𝑚𝑘 𝑡𝑟( 𝑅𝑚𝑘 𝛷𝑚𝑘 𝑅𝑚𝑘 ) + ||ℎ𝑚𝑘 || when plugged this equation

into effective SINR (3.36) give:

35 | P a g e
 𝑗 𝑗 𝑗 𝑗
𝑝𝑗𝑖 𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟( 𝑅𝑗𝑖 𝛷𝑗𝑖 𝑅𝑗𝑖 ) + 𝑝𝑗𝑖 || ℎ̅𝑗𝑖 ||2
𝐷𝐿,𝑀𝑀𝑆𝐸
𝑦 𝑖𝑚
(3.39)
∑𝑀 𝐷𝐿 𝐷𝐿 𝐷𝐿 2
𝑚=1 ∑𝑘=1 𝑝𝑚𝑘 𝜉𝑚𝑘 + ∑(𝑚,𝑘) ∈ 𝑝𝑗𝑖 \(𝑗,𝑖) 𝑝𝑚𝑘 Ґ𝑚𝑘 − 𝑝𝑗𝑖 𝘝𝑗𝑖 + 𝜎𝐷𝐿

𝑚 4
̅𝑗𝑖
|| ℎ ||
Where 𝘝𝑗𝑖𝐷𝐿 = 𝑚 𝑚 𝑚 ̅𝑚 ||2
and
( 𝑝𝑗𝑖 𝑡𝑟 (𝑅𝑗𝑖 𝛷𝑗𝑖 𝑅𝑗𝑖 ) + || ℎ𝑗𝑖

𝑝𝑚𝑘 𝜏𝑝 𝑡𝑟( 𝑅𝑗𝑖𝑚 𝑅𝑚𝑘

𝑚 𝑚
𝛷𝑚𝑘 ) + 𝑝𝑚𝑘 𝜏𝑝 (ℎ̅𝑗𝑖𝑚 )𝐻 𝑅𝑚𝑘
𝑚 𝑚
𝛷𝑚𝑘 𝑚 ̅𝑚
𝑅𝑚𝑘 ℎ𝑗𝑖
𝐷𝐿
+(ℎ̅𝑚𝑘 ) 𝑅𝑗𝑖 ℎ̅𝑚𝑘 + (ℎ̅𝑗𝑖 ) 𝑅𝑗𝑖 ℎ̅𝑚𝑘 + |(ℎ̅𝑗𝑖 ) ℎ̅𝑚𝑘 |2
𝑚 𝐻 𝑚 𝑚 𝑚 𝐻 𝑚 𝑚 𝑚 𝐻 𝑚
𝜉𝑚𝑘 = (3.40)
𝑚
𝑝𝑚𝑘 𝜏𝑝 𝑡𝑟(𝑅𝑚𝑘 𝑚
𝛷𝑚𝑘 𝑚 )
𝑅𝑚𝑘 + ||ℎ̅𝑚𝑘
𝑚
||2

Which this expression represents non-coherent interference and

𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑝2 |𝑡𝑟(𝑅𝑗𝑖𝑚 𝛷𝑚𝑘

𝑚 𝑚
𝑅𝑚𝑘 )|2 + 2√𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑝
𝑚 ̅𝑚 2 ̅𝑚
𝑅𝑒{𝑡𝑟(𝑅𝑗𝑖𝑚 𝛷𝑚𝑘𝑚
𝑅𝑚𝑘 (ℎ𝑚𝑘 ) ℎ𝑗𝑖 }
Ґ𝐷𝐿
𝑚𝑘 = (3.41)
𝑚
𝑝𝑚𝑘 𝜏𝑝 𝑡𝑟(𝑅𝑚𝑘 𝑚
𝛷𝑚𝑘 𝑚 )
𝑅𝑚𝑘 + ||ℎ̅𝑚𝑘
𝑚
||2

This equation which represents coherent interference.

3.5.8 Downlink Spectral Efficiency using the EW-MMSE Estimator.

The EW-MMSE (Element-Wise Minimum Mean Square Error) estimator is a technique
used in wireless communication systems to reduce interference and thereby enhance signal
quality. The EW-MMSE estimator uses the MMSE estimating method to reduce interfer-
ence. It takes into account the statistical aspects of the interference and noise, as well as the
received signal, in order to minimize the mean square error between the estimated and actual
signal. This decreases the influence of interference and improves the quality of the targeted
signal. Therefore, in this section, the downlink spectral efficiency of a massive MIMO sys-
tem with the EW-MMSE estimator is derived. So from downlink spectrum efficiency above
𝑗
ȟ𝑗𝑖
when applying MR precoding based on the MMSE estimator got 𝚆𝑗𝑖 = then:
𝑗
√𝛦{|| ȟ𝑗𝑖 ||2 }

𝑗 𝑗 𝑗 𝑗 𝑗
𝛦{𝚆𝑗𝑖𝐻 ℎ𝑗𝑖 } = √𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟 (𝑅𝑗𝑖 𝛷𝑗𝑖 𝑅𝑗𝑖 ) + ||ℎ̅𝑗𝑖 ||2 (3.42)

36 | P a g e
 𝐻 𝑚 2
𝛦{||ȟ𝑚 2
𝑚𝑘 || }𝛦{|𝚆𝑗𝑘 ℎ𝑗𝑖 | }
𝑗
= 𝑝𝑚𝑘 𝜏𝑝 𝑡𝑟( 𝑅𝑗𝑖𝑚 𝑅𝑚𝑘
𝑚
𝑅𝑚𝑘 ) + 𝑝𝑚𝑘 𝜏𝑝 (ℎ̅𝑗𝑖𝑚 )𝐻 𝑅𝑚𝑘
𝑚 𝑚
𝛷𝑚𝑘 𝑚 ̅𝑚
𝑅𝑚𝑘 ℎ𝑗𝑖 + |(ℎ̅𝑗𝑖𝑚 )𝐻 ℎ̅𝑚𝑘
𝑚 2
|
+ (ℎ̅𝑚𝑘
𝑚 𝐻 𝑚 ̅𝑚
) 𝑅𝑗𝑖 ℎ𝑚𝑘
𝑚 𝑚 𝑚
𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑝2 |𝑡𝑟(𝑅𝑗𝑖 𝛷𝑚𝑘 𝑅𝑚𝑘 )|2 +
+ {2√𝑝𝑗𝑖 𝑝𝑚𝑘 𝜏𝑝 𝑅𝑒{𝑡𝑟(𝑅𝑗𝑖𝑚 𝛷𝑚𝑘 𝑚 𝑚
𝑅𝑚𝑘 )(ℎ̅𝑚𝑘
𝑚 𝐻 ̅𝑚
) ℎ𝑗𝑖 } (𝑚, 𝑘) ∈ 𝑝𝑗𝑖 (3.43)
0 (𝑚, 𝑘) ∉ 𝑝𝑗𝑖

Where 𝛦{|| ȟ𝑚 2 𝑚 ̅𝑚 2
𝑚𝑘 || } = 𝑡𝑟( ∑𝑚𝑘) + || ℎ𝑚𝑘 || , by inserting these expression into the DL
effective SINR above, the resulting expression would give:

𝑦𝑗𝑖𝐷𝐿,𝐸𝑊_𝑀𝑀𝑆𝐸
𝑗 𝑗 𝑗 𝑗 ̅𝑗 2
𝑝𝑗𝑖 ( 𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟(𝐷𝑗𝑖 Ʌ𝑗𝑖 𝐷𝑗𝑖 ) + ||ℎ̅𝑗𝑖 ||2⁄( 𝑡𝑟( ∑𝑚
𝑚𝑘) + ||ℎ𝑗𝑖 || )
= 𝑗 𝑗 𝑗 𝑗
(3.44)
𝑖𝑚
𝐷𝐿
𝜒𝑚𝑘 ( 𝑝𝑗𝑖 𝜏𝑝 𝑡𝑟 ( 𝐷𝑗𝑖 Ʌ𝑗𝑖 𝐷𝑗𝑖 ) + ||ℎ̅𝑗𝑖 ||2 )2
∑𝑀
𝑚=1 ∑𝑘= 𝑝𝑚𝑘 ̅𝑚 2 − 𝑝𝑗𝑘
𝑡𝑟( ∑𝑚
𝑚𝑘) + ||ℎ𝑚𝑘 ||
𝑗 𝑗
𝑡𝑟( ∑𝑗𝑖) + ||ℎ̅𝑗𝑖 ||2
2
+ 𝜎𝐷𝐿

3.5.9 Downlink Spectral Efficiency using the LS Estimator

The least squares (LS) estimator is a statistical estimation approach that minimizes the sum
of the squares of the discrepancies between observed and predicted values. In the context of
downlink spectral efficiency, the LS estimator can be used to predict channel parameters
such as channel gains or signal-to-noise ratio (SNR), which are critical for determining the
feasible data rate. Therefore, the downlink spectral efficiency of a massive MIMO system
with the LS estimator is derived. From above general Downlink spectrum efficiency, if MR
(Maximum Ratio) precoding is used with the LS estimator, and the precoding weights are
𝑗
ȟ𝑗𝑖
defined as 𝚆𝑗𝑖 = then,
𝑗
√𝛦{ ||ȟ𝑗𝑖 ||2 }

𝑗 √𝑝𝑚𝑘 ̅ 𝑗 𝐻 ̅ 𝑗
𝑗 𝑡𝑟 (𝑅𝑗𝑖 ) + ∑(𝑚,𝑘)∈𝑝𝑗𝑖 (ℎ𝑚𝑘 ) ℎ𝑗
𝑗
𝛦{ 𝘝𝑗𝑖𝐻 ℎ𝑗𝑖 } √𝑝𝑗𝑖
𝛦{ 𝚆𝑗𝑖𝐻 ℎ𝑗𝑖 } = = (3.45)
√𝛦{ ||𝘝𝑗𝑖 ||2 } 1 𝑗 𝑝𝑖𝑙𝑜𝑡
√𝑝 𝜏 2 ( 𝜏𝑝 𝑡𝑟(( 𝛷𝑗𝑖 )−1 ) + ||𝑦̅𝑗𝑗𝑖 ||2 )
𝑗𝑖 𝑝

37 | P a g e
 𝑚 2 2 𝐷𝐿,𝐿𝑆
𝛦{ ||ȟ𝑚 2 𝐻
𝑚𝑘 || }𝛦{ |𝚆𝑚𝑘 ℎ𝑗𝑖 | } = 𝑝𝑚𝑘 𝜏𝑝 𝜒𝑚𝑘
𝑝𝑖𝑙𝑜𝑡 𝐻 ̅ 𝑚 𝑝𝑖𝑙𝑜𝑡 𝐻 𝑚 ̅ 𝑚
= 𝜏𝑝 𝑡𝑟( 𝑅𝑗𝑖𝑚 (𝛷𝑚𝑘
𝑚 −1
) ) + 2√𝑝𝑗𝑖 𝜏𝑝 𝑅𝑒{ ( 𝑦̅𝑗𝑚𝑘 ) ℎ𝑗𝑖 𝑡𝑟(𝑅𝑗𝑖𝑚 ) + ( 𝑦̅𝑗𝑚𝑘 ) 𝑅𝑗𝑖 ℎ𝑗𝑖 }
𝑝𝑖𝑙𝑜𝑡 𝐻 𝑚 𝑝𝑖𝑙𝑜𝑡
( 𝑦̅𝑗𝑚𝑘 ) 𝑅𝑗𝑖 𝑦̅𝑗𝑚𝑘 + 𝜏𝑝 (ℎ̅𝑗𝑖𝑚 )𝐻 (𝛷𝑚𝑘
𝑚 −1 ̅ 𝑚
) ℎ𝑗𝑖
𝑝𝑖𝑙𝑜𝑡 𝐻 ̅ 𝑚 2
+|(𝑦̅𝑗𝑚𝑘 ) ℎ𝑗𝑖 | (𝑚, 𝑘) ∉ 𝑝𝑗𝑖
+ (3.46)
𝑝𝑗𝑖 𝜏𝑝2 |𝑡𝑟(𝑅𝑗𝑖𝑚 )|2 + 𝜒̅𝑚𝑘
𝐻
𝑅𝑗𝑖𝑚 𝜒̅𝑚𝑘 + 𝜏𝑝 (ℎ̅𝑗𝑖𝑚 )𝐻 (𝛺𝑚𝑘
𝑚 −1 ̅ 𝑚
) ℎ𝑗𝑘
{+|𝜒̅𝑚𝑘 ℎ̅𝑗𝑖 | + 𝑝𝑗𝑖 𝜏𝑝 ||ℎ̅𝑗𝑖 || + 2√𝑝𝑗𝑖 𝜏𝑝 𝑅𝑒{𝜒̅𝑚𝑘 (ℎ̅𝑗𝑖 ) ℎ̅𝑗𝑖 } (𝑚, 𝑘)
𝐻 𝑚 2 2 𝑚 4 𝐻 𝑚 𝐻 𝑚
∈ 𝑝𝑗𝑖

To calculate the DL SINR by inserting the given expression for 𝛦{||ȟ𝑚 2

𝑚𝑘 || } =
1 𝑚 −1 𝑝𝑖𝑙𝑜𝑡 2
2 (𝜏𝑝 𝑡𝑟((𝛷𝑚𝑘 ) ) + ||𝑦̅𝑚𝑚𝑘 || into general equation of DL effective SINR (3.36)
𝑝𝑚𝑘 𝜏𝑝

yield as below.

𝑦𝑗𝑖𝐷𝐿,𝐿𝑆

𝑗 √𝑝𝑚𝑘 ̅ 𝑗 𝐻 ̅ 𝑗 2 𝑗 𝑝𝑖𝑙𝑜𝑡 2
𝑝𝑗𝑖 𝑝𝑗𝑖 𝜏𝑝2 |𝑡𝑟(𝑅𝑗𝑖 ) + ∑(𝑚,𝑘)∈𝑝𝑗𝑖 (ℎ𝑚𝑘 ) ℎ𝑗𝑖 | ⁄(𝜏𝑝 𝑡𝑟( 𝛷𝑗𝑖 )−1 + ||𝑦̅𝑗𝑗𝑖 || )
√𝑝𝑚𝑖
=
𝑗 √𝑝𝑚𝑘
𝑝𝑗𝑖 𝑝𝑗𝑖 𝜏𝑝2 |𝑡𝑟 (𝑅𝑗𝑖 ) + ∑(𝑚,𝑘)∈𝑝𝑗𝑖
√𝑝𝑗𝑖
𝐷𝐿,𝐿𝑆 𝑗 𝑗
𝑖𝑚 𝑝𝑚𝑘 𝜏𝑝2 𝜒𝑚𝑘 (ℎ̅𝑚𝑘 )𝐻 ℎ̅𝑚𝑘 |2
∑𝑀
𝑚=1 ∑𝑘=1 𝑝𝑚𝑘 𝑚 𝑝𝑖𝑙𝑜𝑡 − 𝑗 𝑝𝑖𝑙𝑜𝑡 2
2
+ 𝜎𝐷𝐿
𝜏𝑝 𝑡𝑟((𝛷𝑚𝑘 )−1 + ||𝑦̅𝑗𝑗𝑖 ||2 𝜏𝑝 𝑡𝑟((𝛷𝑗𝑖 )−1 + ||𝑦̅𝑗𝑗𝑖 ||

38 | P a g e
 CHAPTER FOUR
4. RESULT AND DISCUSSION
In this section, the simulation result of spectrum efficiency of massive MIMO in uplink and
downlink are obtained using MATLAB version 2023b. Furthermore, the simulation result
Rician fading and Rayleigh fading spectrum efficiency under different estimator are ob-
tained. Additionally, the obtained results are discussed and analyzed in these section. In this
simulation to ensure a more realistic and representative evaluation of the system perfor-
mance as the closed form spectral efficiency expressions in a typical cellular network has
wrap-around topology used in simulation setup, to ensures that all base stations receive
equal amount of interference from all direction, regardless of their position within the cel-
lular layout. Generally, each UE is allocated to the BS with the highest channel gain, taking
into account all potential combinations of UEs and BSs in the network. The position of each
UE is used to compute large-scale fading and the nominal angle between the UE and BSs.
The simulation result produced is consistent with 3GPP requirements. The simulation pa-
rameters utilized in the analysis are given in table 4.1.

The simulation results presented in this thesis are organized and discussed under three main
topics.

 Performance of estimation quality of channel estimator in terms of NMSE.

 Results of Uplink spectrum efficiency under different channel estimator.
 Result of Downlink spectrum efficiency under different channel estimator.

Table 4. 1 Parameters used for Simulations.

Parameter Description Parameter type and value Remarks

Cell type Micro cell Based on
(Zarai et al., 2011)

Cell coverage area 250m*250m Based on

(25996-800, n.d.)
Bandwidth 20MHz Based on
(25996-800, n.d.)
Number of UE per cell 𝑖 ≥ 10 It can be vary
Pilot reuse factor 𝑓=1 It can be vary

39 | P a g e
 Carrier Frequency 2GHz Based on many paper
Pilot length 𝜏𝑝 = 𝑓𝑖,since 𝑓 = 1 pilot According to
are allocated random in (3GPP TR 25.996, 2009)
each cell
Separation distance(UE to BS) 35m Based on
(25996-800, n.d.)
Shadow fading SD 𝛿𝑠𝑓 = 4 for LOS,𝛿𝑠𝑓 = 10 According to
for NLOS (3GPP TR 25.996, 2009)

Coherence Block 𝜏𝑐 = 200 Based on (25996-800,

n.d.)
Path loss exponent 3.7 for urban According to
(25996-800, n.d.)
Number of BSs antenna M = 16 It can be vary
ASD 5° According to
(3GPP TR 25.996, 2009)
Total received noise power -94dBm Based on
(25996-800, n.d.)

4.1 Performance of estimation quality of channel estimator in terms of

NMSE.
This section provides the Normalized Mean Squared Error (NMSE) in the estimation of a
spatially correlated channel is analyzed based on the local scattering model with Gaussian
angular distribution. The scenario considered involves a base station (BS) estimating the
channel of its user equipment (UE) in the presence of interference from a UE in another cell
transmitting the same pilot sequence. The simulation results underscore the significance of
accurate channel estimation in Massive MIMO systems, the impact of different estimators
on system. The simulation results present for estimation quality of channel estimator is:

 Comparison of estimation quality of channel estimator in terms of NMSE

40 | P a g e
4.1.1 Comparison of estimation quality of channel estimator in terms of
NMSE.
The performance estimation quality comparison among various channel estimator can be
analyzed in terms of the NMSE curves as shown in figure 4.1. The results are simulated at
effective SNR of the desired UE is varied from −10𝑑𝐵 𝑡𝑜 20𝑑𝐵, 𝐴𝑆𝐷 = 10° are used and
results are averaged over different nominal angle between 0° 𝑎𝑛𝑑 360° . As observed from
the graph the three channel estimators produce varying NMSEs. The MMSE estimator is
the most accurate because it takes into account spatial channel correlation when compared
to other EW-MMSE and LS. Although, the EW-MMSE estimator performs similarly to the
MMSE estimator for uncorrelated channels, it falls short of the MMSE estimate, even at
high SNR due to pilot contamination. The LS estimator works badly at low SNR (𝑁𝑀𝑆𝐸 >
1), while the trivial all-zero estimate yields an NMSE of 1. At higher SNRs, the LS estima-
tor is comparable to the EW-MMSE estimator, but their error levels differ (assuming pilot
contamination). The LS estimator can offer accurate channel direction estimates, but limited
statistical information makes scaling the channel norm challenging.

Figure 4. 1 NMSE in the estimation of a spatially correlated channel, based on the local
scattering model with Gaussian angular distribution, for different estimators.

41 | P a g e
4.2 Result of Uplink Spectrum Efficiency Under Different Channel esti-
mator.
This section provides further clarity on the simulations and analysis uplink spectrum effi-
ciency under different channel estimator. The simulation results for spatially correlated and
uncorrelated Racian, Rayleigh fading uplink spectrum efficiency under different channel
estimator are discussed here. The key simulation results presented for the uplink spectrum
under different channel estimator are:

 Average Uplink spectrum efficiency for different channel estimator.

 Cumulative distribution Uplink spectrum efficiency for different channel estimator.
 Average Uplink spectrum efficiency under different channel estimator where all
have a LOS path
 Average Uplink spectrum efficiency under different pilot reuse factor and different
channel estimator.
 Average Uplink spectrum efficiency with spatially uncorrelated Racian and Ray-
leigh fading under different channel estimator.

4.2.1 Average Uplink spectrum efficiency for different channel estimator.

Figure 4.2 depicts the average uplink (UL) sum spectral efficiency (SE) for a system with
10 users as a function of the number of base station (BS) antennas (N). The plot showcases
the sum UL SE averaged across UE sites and shadow fading realizations utilizing maximum
ratio combining with MMSE, LS, or EW-MMSE estimators. Also assume the Rayleigh fad-
ing curves with the same covariance matrices, reflecting the case when all LoS components
are blocked but small-scale fading remains. The curves are generated using closed-form
expressions from uplink spectrum efficiency. while the □ markers are generated through
Monte Carlo simulations. The markers' overlap with the curves confirms the validity of the
result analysis. The MMSE estimator turnout the highest uplink spectrum efficiency due to
its ability to account for the LoS component and spatial correlation. For Rician fading,
MMSE and EW-MMSE performance are very close to each other, but EW-MMSE outper-
forms LS due to its use the knowledge of channel mean values. For Rayleigh fading, MMSE
is the preferred estimation method. However, the total SE with LS and EW-MMSE are same
due to similar estimates up to a scaling factor. Generally, the result insights into how the UL

42 | P a g e
sum SE changes with the varying number of antennas for each channel estimator. It demon-
strates the impact of different estimation methods on the system's spectral efficiency in Mas-
sive MIMO scenarios.

Figure 4. 2 Average Uplink Sum Spectrum efficiency for 10 user as function of number of
antennas of BS under different Channel estimator and Fading channels.

4.2.2 Cumulative distribution Uplink spectrum efficiency for different

channel estimator
In this result analyzed the distribution of spectrum efficiency value across different UEs in
a multi-cell massive MIMO system. The CDF curve are show that how the spectrum effi-
ciency value is distributed across UEs in the system, capturing the variability in channel
condition and performance. So the result compares the CDF curves for different channel
estimators of the spectrum distribution among UEs under varying channel condition and
shadow fading realization. Therefore, which channel estimator performs better in terms of
SE distribution among UEs compered. From the figure 4.3 the point where the CDF curves
of different channel estimators intersect or touch each other signifies that SE value at which
the performance of those estimators is comparable or similar spectrum efficiency. The graph
indicates that for UEs with good channel conditions, the MMSE and EW-MMSE estimators

43 | P a g e
provide similar spectral efficiency performance, as the estimation errors are small and How-
ever, there is a noticeable difference between the MMSE and EW-MMSE estimators for the
UEs with the weakest channel conditions. In these cases, the MMSE estimator outperforms
the EW-MMSE estimator. Generally, that the MMSE estimator is particularly beneficial for
UEs experiencing poor channel conditions, as it is able to better utilize the knowledge of
the channel statistics to improve the estimation accuracy and, consequently, the spectral
efficiency performance.

Figure 4. 3 CDF curve of the Uplink Spectrum Efficiency under different Channel
estimator.

4.2.3 Average Uplink spectrum efficiency under different channel estima-

tor where all have LOS path.
Figure 4.4 analyzed a situation with spatially correlated Rician fading channels between
each pair of BSs and UEs, which may not be feasible in practice. In comparison to cases
when only some UEs have LoS, the presence of a LoS component leads to greater average
UL sum SEs, which improve the SE. As illustrate in graph the average UL SE as a function
of the number of BS antennas for various channel estimators, including MMSE, EW-
MMSE, and LS estimators. The plot underscores the critical role of channel estimation

44 | P a g e
methods in maximizing spectral efficiency, especially in environments where all UE-BS
pairs benefit from LoS propagation paths. So from comparison the MMSE estimator
achieves the highest UL SE among the considered estimators when all UE-BS pairs have
LoS paths, this superior performance is attributed to the MMSE estimator's capability to
exploit the known LoS components and spatial correlation in the channel. The EW-MMSE
estimator also demonstrates competitive performance in the scenario with all UE-BS pairs
having LoS paths, its performance is very close to that of the MMSE estimator, indicating
that leveraging the knowledge of channel mean values contributes to improved spectral ef-
ficiency. But he LS estimator shows comparatively lower UL SE performance in this sce-
nario, The LS estimator lacks the ability to suppress pilot interference through spatial pro-
cessing, which can limit its performance in environments with strong LoS components.

Figure 4. 4 Average Uplink sum of Spectrum efficiency for 10 user is shown as a function
of BS antennas for various channel estimator where all UE-BS have a LOS Path.

45 | P a g e
4.2.4 Average Uplink spectrum efficiency under different pilot reuse fac-
tor and different channel estimator.
In this section the average uplink spectral efficiency (UL SE) for different pilot reuse factors
and various channel estimators is analyzed. The figure 4.5 indicates that the average UL SE
as a function of the number of BS antennas for different pilot reuse factors, denoted by the
integer 𝑓, where 𝜏𝑝 = 𝑓𝑖. This setup indicates that there are 𝑓 times more pilots than user

equipment (UE) per cell, and the same subset of pilots is reused in a fraction 1⁄𝑓 of the

cells. The figure demonstrates the impact of different pilot reuse factors on the average UL
SE for various estimators. Increasing the pilot reuse factor affects the pre-log factor in the
UL SE expression, where 𝜏𝑢 = 𝜏𝑐 − 𝜏𝑝 , influencing the system's performance. Generally,
a larger reuse factor reduces pilot contamination but increases the pre-log penalty and larger
reuse factors lead to a decrease in the pre-log factor, affecting the overall system perfor-
mance in terms of spectral efficiency. Additionally, from the plot the Least Squares (LS)
estimator is shown to be more sensitive to pilot contamination due to its inability to suppress
pilot interference through spatial processing. In contrast, the Minimum Mean Squared Error
(MMSE) estimators can mitigate pilot interference, enabling them to function effectively
even with tighter reuse factors. Overall, the figure 4.5 highlights that the optimal pilot reuse
factor differs for the different estimators based on their sensitivity to pilot contamination
and pre-log penalty considerations and selecting an appropriate reuse factor is crucial for
balancing pilot interference suppression and pre-log factor optimization to maximize spec-
tral efficiency.

46 | P a g e
 Figure 4. 5 Uplink Spectrum Efficiency for different Pilot reuse factor with Various
channel estimator.

4.2.5 Average Uplink spectrum efficiency with spatially uncorrelated Ra-

cian and Rayleigh fading under different channel estimator.
This section considered the simulation result of the average uplink spectral efficiency (UL
SE) in a scenario where spatially uncorrelated Rician and Rayleigh fading channels. The
average UL SE analyzed as a function of the number of BS antennas under different channel
estimators, including MMSE/EW-MMSE/LS with Rician fading, and MMSE/EW-
MMSE/LS with Rayleigh fading. The figure 4.6 analyzed the differences in UL SE perfor-
mance when utilizing different channel estimators under spatially uncorrelated Rician and
Rayleigh fading. The simulation result indicates that the average UL SE with uncorrelated
fading as a function of antenna count. EW-MMSE and MMSE coincide in this scenario
because the spatial covariance matrices are diagonal. Rayleigh fading results in equal per-
formance across all vectors due to a deterministic scaling factor that cancels out in SINR
formulations. But the LS estimator shows lower UL SE performance in spatially uncorre-
lated Rician fading compared to MMSE and EW-MMSE estimators.

47 | P a g e
 Figure 4. 6 Uplink Spectrum Efficiency for spatially uncorrelated Rician and Rayleigh
fading under different channel estimation.

4.2.6 Average Uplink spectrum efficiency for different channel estimator

and different Precoding schemes.
Figure 4.7 illustrates the simulation result of the average UL sum SE when M-MMSE, RZF,
and MR combining schemes are considered. From the result MMSE estimator provides the
largest SEs, as expected due to its accuracy in channel estimation and EW-MMSE instead
results in some percent reduction in SE, depending on the combining strategy. But, in case
of RZF or MR combining, the SE difference between EW-MMSE and LS estimators is
minimal. However, M-MMSE combining performs poorly with LS. The LS estimator fails
to scale channel estimates correctly in the presence of pilot contamination, instead estimat-
ing the sum of interfering channels. M-MMSE overestimates the norm of UE channels in
other cells, emphasizing the importance of minimizing inter-cell interference. Using RZF
or M-MMSE combining yields significant SE increases over low-complexity MR combin-
ing, regardless of the channel estimator utilized.

48 | P a g e
 Figure 4. 7 Average UL sum SE when using MMSE, EW-MMSE, or LS channel
estimators, for a setup with M = 100 BS antennas and K = 10 UEs per cell. Three different
combining schemes are considered.

4.3 Result of Downlink spectrum efficiency under different channel esti-

mator.
In this section the simulation result of downlink spectrum efficiency under different channel
estimator analyzed. The simulation results for spatially correlated Racian, Rayleigh fading
of downlink spectrum efficiency under different channel estimator are discussed here. The
key simulation results presented for the downlink spectrum under different channel estima-
tor are:

 Average Downlink spectrum efficiency for different channel estimator

 Average Downlink spectrum efficiency for different channel estimator and different
Precoding schemes.

49 | P a g e
4.3.1 Average Downlink spectrum efficiency for different channel estima-
tor.
In figure 4.8 the downlink spectral efficiency (DL SE) is examined as a function of the
number of base station (BS) antennas for different channel estimators. The simulation result
presents the downlink spectral efficiency (DL SE) as a function of the number of base station
(BS) antennas for different channel estimators (MMSE, EW-MMSE, LS) in scenarios with
Rician and Rayleigh fading channels. The graph analyzed that, the DL SE achieved with the
MMSE estimator in Rician fading scenarios generally exhibits the highest values among the
three estimators as the number of BS antennas increases. Because MMSE estimation opti-
mally utilizes the knowledge of Rician fading characteristics and spatial diversity, leading
to enhanced spectral efficiency in the downlink transmission scenario. The DL SE with the
EW-MMSE estimator closely follows the MMSE estimator in Rician fading scenarios,
showing competitive performance and relatively high spectral efficiency, due to EW-
MMSE focuses on element-wise mean squared error optimization, contributing to improved
spectral efficiency in Rician fading conditions. But the DL SE achieved with the LS estima-
tor in Rician fading scenarios lags behind MMSE and EW-MMSE estimators, resulting in
lower spectral efficiency values, because LS estimation may struggle to fully exploit the
benefits of Rician fading and spatial diversity, leading to reduced spectral efficiency com-
pared to MMSE and EW-MMSE. Also the simulation result analyzed DL SE in Rayleigh
fading scenarios, from this result the DL SE achieved with the MMSE estimator generally
exhibits the highest values among the three estimators as the number of BS antennas in-
creases, reason MMSE estimation optimally utilizes the available channel information and
spatial diversity, leading to enhanced spectral efficiency even in Rayleigh fading conditions.
The DL SE with the EW-MMSE/LS estimator in Rayleigh fading scenarios coincide per-
formance in improving spectrum efficiency.

50 | P a g e
Figure 4. 8 Downlink Spectrum Efficiency for 10 user is shown as function of BS antenna
for various channel estimation and fading channel..

4.3.2 Average Downlink spectrum efficiency for different channel estima-

tor and different Precoding schemes.
Figure 4.9 shows the simulation results of a comparison of the spectrum efficiency perfor-
mance of M-MMSE, RZF and MR combining schemes under difference channel estimator
techniques. From the result MMSE channel estimator produces the greatest SEs and EW-
MMSE instead results in some percent reduction in SE, depending on the combining strat-
egy. Additionally, the difference in SE between the EW-MMSE and LS estimators is quite
minor when employing RZF or MR precoding, but M-MMSE precoding performs badly
with LS. The LS estimator incorrectly scales channels to UEs in other cells, leading to an
overemphasis on mitigating inter-cell interference. To address this issue, reduce the norm
of inter-cell channel estimates.

51 | P a g e
 Figure 4. 9 Average DL sum SE when using the MMSE, EW-MMSE, or LS channel
estimators, for a setup with M = 100 BS antennas and K = 10 UEs per cell. Three different
precoding schemes.

52 | P a g e
 5. CONCLUSION AND RECOMMENDATION
5.1 Conclusion
In this thesis, the Uplink and Downlink Spectrum efficiency of multi-cell massive MIMO
system are examined under spatially correlated Racian and Rayliegh fading channel. Ac-
cording to the simulation result channel estimation methods, such as MMSE, EW-MMSE,
and LS estimators analyzed, in maximizing spectral efficiency, especially in environments
with Line-of-Sight (LoS) and NLOS propagation paths. The comparison results reveal that
the MMSE estimate outperforms other estimators for spatially coupled Rician and Rayleigh
fading, while the LS estimator has the lowest SE. Furthermore, the results show that under
Rician fading and Rayleigh circumstances, the MMSE estimator consistently obtains the
highest Downlink (DL) spectral efficiency values among the estimators tested, demonstrat-
ing its optimal use of available channel information and spatial diversity. In contrast, the LS
estimator falls behind the MMSE and EW-MMSE estimators in both Rician and Rayleigh
fading scenarios, resulting in poorer spectral efficiency values due to constraints in fully
exploiting fading benefits and spatial variety.

Overall, the simulation outcomes emphasize the significance of selecting appropriate chan-
nel estimation techniques to enhance spectral efficiency in massive MIMO systems, with
MMSE estimation proving particularly beneficial in scenarios with challenging channel
conditions. In practice, the covariance matrices and mean vectors may not be properly un-
derstood. The practical performance sits somewhere between the MMSE/EW-MMSE and
LS estimators, because the mean is likely known up to a random phase shift and the covar-
iance matrices are known with some error. These insights contribute to a deeper understand-
ing of the impact of channel estimators on system performance, and can influence the design
and optimization of massive MIMO installations.

53 | P a g e
5.2 Recommendation
In this thesis, the improvement of spectrum efficiency in Massive MIMO systems under
different channel estimator are evaluated under simulation result only using the MATLAB
tools. In the future exploring advanced channel estimation techniques beyond MMSE, EW-
MMSE, and LS estimators. Investigating machine learning-based approaches or deep learn-
ing algorithms for channel estimation in massive MIMO systems could potentially improve
spectral efficiency further. Another possibility for future research is effective interference
management strategies for massive MIMO deployments. Developing interference mitiga-
tion techniques, such as advanced precoding algorithms or interference alignment methods,
can improve spectral efficiency in interference-limited scenarios.

54 | P a g e
 REFERENCES
25996-800. (n.d.).

3GPP TR 25.996. (2009). Spacial channel model for Multiple Input Multiple Output
(MIMO) simulations (3GPP TR 25.996 version 11.0.0 Release 11). 11.0.0, 0–41.

Abebaw, Y., Shakya, R. K., Gelmecha, D. J., & Ware, E. T. (2023). Spectral Efficiency
Analysis for Uplink Multicell Massive MIMO Cellular Communication System under
Fading Channels. Journal of Electrical and Computer Engineering, 2023.
https://doi.org/10.1155/2023/6623938

Abid, W., Roy, S., & Ammari, M. L. (2019). Uplink Performance Analysis for Massive
MIMO Linear Processing. IEEE Access, 7, 180749–180760.
https://doi.org/10.1109/ACCESS.2019.2959580

Abraham, J. (2022). Statistics of the Effective Massive MIMO Channel in Correlated Rician
Fading. 3(December 2021). https://doi.org/10.1109/OJAP.2022.3147015

Akgun, B., Krunz, M., & Koyluoglu, O. O. (2017). Pilot contamination attacks in massive
MIMO systems. 2017 IEEE Conference on Communications and Network Security,
CNS 2017, 2017-Janua, 1–9. https://doi.org/10.1109/CNS.2017.8228655

Ali, A., Qureshi, I. A., Memon, A. L., Memon, S. A., & Saba, E. (2018). Spectral efficiency
of massive MIMO communication systems with zero forcing and maximum ratio
beamforming. International Journal of Advanced Computer Science and Applications,
9(12), 383–388. https://doi.org/10.14569/IJACSA.2018.091254

Amadid, J., Belhabib, A., Khabba, A., El Ouadi, Z., & Zeroual, A. (2022). Channel
Estimation Evaluation For a Massive MIMO System Considering Spatially Correlated
Channels in an Urban Network. E3S Web of Conferences, 351, 1–6.
https://doi.org/10.1051/e3sconf/202235101055

Björnson, E., & Demir, Ö. T. (2024). Introduction to Multiple Antenna Communications

and Reconfigurable Surfaces. In Introduction to Multiple Antenna Communications
and Reconfigurable Surfaces. https://doi.org/10.1561/9781638283157

Björnson, E., Hoydis, J., Labs, B., & Sanguinetti, L. (2020). * Massive MIMO Networks:
Spectral, Energy, and Hardware Efficiency. Foundations and Trends in Signal
Processing, 11(3–4), 154–655. https://doi.org/10.1561/2000000093.Simulation

55 | P a g e
Björnson, E., Larsson, E. G., & Marzetta, T. L. (2016). Massive MIMO: Ten myths and one
critical question. IEEE Communications Magazine, 54(2), 114–123.
https://doi.org/10.1109/MCOM.2016.7402270

Chaves, R. S., Lima, M. V. S., Cetin, E., & Martins, W. A. (2022). User Selection for
Massive MIMO under Line-of-Sight Propagation. IEEE Open Journal of the
Communications Society, PP(May), 1.
https://doi.org/10.1109/OJCOMS.2022.3172621

Engineering, C. (2021). Performance Analysis and Optimization in Massive MIMO Systems

Approval Page.

Fading, R., Mimo, B. M., & Users, M. (2020). Massive MIMO with Multi-Antenna Users
under Jointly Correlated Massive MIMO with Multi-Antenna Users under Jointly
Correlated Ricean Fading.

Kansal, L., Sharma, V., & Singh, J. (2019). Multiuser Massive MIMO-OFDM System
Incorporated with Diverse Transformation for 5G Applications. Wireless Personal
Communications, 109(4), 2741–2756. https://doi.org/10.1007/s11277-019-06707-1

Li, F., Sun, Q., Ji, X., & Chen, X. (2022). Scalable Cell-Free Massive MIMO with Multiple
CPUs. Mathematics, 10(11). https://doi.org/10.3390/math10111900

Ngo, H. Q. (2015). Massive MIMO: Fundamentals and System Designs. In Massive MIMO:
Fundamentals and System Designs (Issue 1642). https://doi.org/10.3384/lic.diva-
112780

Nivethitha, R., Suresh, M. N., & Ananthi, G. (2015). Channel estimation in massive mimo
systems. International Journal of Applied Engineering Research, 10(20), 18807–
18811. https://doi.org/10.22214/ijraset.2022.39775

Rajani, A., Teja, S. R., & Tech, M. (2018). Energy Efficient Data Detection for Uplink
Multiuser Massive Mimo Systems. 8(1169), 1169–1175.

Sadraei, M. H., Fazel, M. S., & Doost-hoseini, A. M. (2020). Ergodic spectral efficiency of
massive MIMO with correlated Rician channel and MRC detection based on LS and
MMSE channel estimation. https://doi.org/10.1049/iet-com.2019.0905

ScholarWorks, U., Graduate Theses, U., & Mayouche, A. (2016). Channel Estimation
Overhead Reduction for Downlink FDD Channel Estimation Overhead Reduction for

56 | P a g e
 Downlink FDD Massive MIMO Systems Massive MIMO Systems Part of the Electrical
and Electronics Commons, and the Systems and Communications Commons Citation
Cit. https://scholarworks.uark.edu/etd://scholarworks.uark.edu/etd/1688

Shi, E., Zhang, J., Wang, Z., Wing, D., Ng, K., & Ai, B. (n.d.). Uplink Performance of RIS-
aided Cell-Free Massive MIMO System Over Spatially Correlated Channels.

Shlezinger, N., Eldar, Y. C., & Mimo, A. M. (2019). On the Spectral Efficiency of
Noncooperative Uplink Massive MIMO Systems. IEEE Transactions on
Communications, 67(3), 1956–1971. https://doi.org/10.1109/TCOMM.2018.2884008

Siddiqui, M. U. A., Qamar, F., Kazmi, S. H. A., Hassan, R., Arfeen, A., & Nguyen, Q. N.
(2023). A Study on Multi-Antenna and Pertinent Technologies with AI/ML
Approaches for B5G/6G Networks. Electronics (Switzerland), 12(1).
https://doi.org/10.3390/electronics12010189

Sofi, I. B., Gupta, A., & Kausar, R. (2018). Performance evaluation of different channel
estimation techniques in MIMO system for hata channel model. Proceedings - 2nd
International Conference on Intelligent Circuits and Systems, ICICS 2018, April, 126–
132. https://doi.org/10.1109/ICICS.2018.00041

Thomas, L., Erik, G., Yang, H., & Ngo, H. Q. (n.d.). Fundamentals of Massive MIMO.

Ueoka, T. T. R., Souza, D. D., Freitas, M. M. de, Cavalcante, A. M., & Costa, J. W. (2022).
Performance of Cell-Free Massive MIMO Networks under Rayleigh and Rician fading.
1–2. https://doi.org/10.14209/sbrt.2022.1570811228

Wang, S., Wen, M., Xia, M., Wang, R., Hao, Q., & Wu, Y. C. (2020). Angle Aware User
Cooperation for Secure Massive MIMO in Rician Fading Channel. IEEE Journal on
Selected Areas in Communications, 38(9), 2182–2196.
https://doi.org/10.1109/JSAC.2020.3000837

Wikiman, O. O., & Idowu-bismark, O. (2019). Performance of Massive MIMO in a Rician

Fading Channel Using a ZF Precoder. 9(1), 1–7.
https://doi.org/10.5923/j.jwnc.20190901.01

Zaheera, F. (2022). Large-scale Fading Decoding based Massive MIMO System with
Regularized Zero Forcing Precoder. 13(03), 16–26.

Zarai, F., Boudriga, N., & Obaidat, M. S. (2011). Universal Mobile Telecommunications

57 | P a g e
 System. Handbook of Computer Networks, 2, 699–715.
https://doi.org/10.1002/9781118256114.ch46

Zhi, K., Pan, C., Ren, H., Wang, K., Elkashlan, M., Renzo, M. Di, Schober, R., Poor, H. V.,
Wang, J., & Hanzo, L. (2023). Two-Timescale Design for Reconfigurable Intelligent
Surface-Aided Massive MIMO Systems With Imperfect CSI. IEEE Transactions on
Information Theory, 69(5), 3001–3033. https://doi.org/10.1109/TIT.2022.3227538

Zuo, X., Zhang, J., Mu, X., & Yang, L. (n.d.). Channel Correlation Relied Grouped Spatial
Modulation for Massive MIMO Systems. 1–11.

58 | P a g e