Naval Postgraduate School: Monterey, California

NAVAL
POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA
THESIS
ANALYSIS AND SIMULATION OF DISADVANTAGED

RECEIVERS FOR MULTIPLE-INPUT MULTIPLE-
OUTPUT COMMUNICATIONS SYSTEMS
by
Tracy A. Martin
September 2010
Thesis Advisor: Frank Kragh

Second Reader: Tri Ha
Approved for public release; distribution is unlimited

THIS PAGE INTENTIONALLY LEFT BLANK
REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instruction,
searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send
comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to
Washington headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA
22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188) Washington DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE
3. REPORT TYPE AND DATES COVERED
September 2010 Masters Thesis
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
Analysis and Simulation of Disadvantaged Receivers for Multiple-Input Multiple-
Output Communications Systems
6. AUTHOR(S) Tracy A. Martin
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Naval Postgraduate School REPORT NUMBER
Monterey, CA 93943-5000
9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
N/A AGENCY REPORT NUMBER
11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official policy
or position of the Department of Defense or the U.S. Government. IRB Protocol number ______N/A__________.
12a. DISTRIBUTION / AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
13. ABSTRACT (maximum 200 words)
This thesis investigates the simulation performance of Multiple-Input Multiple-Output (MIMO) communication
systems. Uncoded narrowband and wideband MIMO systems are presented and simulated. The system is further
refined with the addition of Space-Time Codes (STC) and Channel State Information at the Transmitter (CSIT). A
disadvantaged receiver is subsequently introduced to the system lacking the optimization enjoyed by the native
receiver. Simulation and analysis was conducted with multiple modulation schemes and antenna configurations.
Rayleigh and Rician fading models are developed and simulated as the wireless channel. Receiver performance
results based on MATLAB simulation are compared and presented.
14. SUBJECT TERMS Multiple-Input Multiple-Output (MIMO), Orthogonal Frequency Division 15. NUMBER OF
Multiplexing (OFDM), Binary Phase Shift Keying, Quadrature Phase Shift Keying, Quadrature PAGES
Amplitude Modulation, Rayleigh Fading Channel, Rician Fading Channel, Space-Time Code (STC), 106
Disadvantaged Receiver, Linear Diversity Precoding, Channel State Information (CSI) 16. PRICE CODE
17. SECURITY 18. SECURITY 19. SECURITY 20. LIMITATION OF
CLASSIFICATION OF CLASSIFICATION OF THIS CLASSIFICATION OF ABSTRACT
REPORT PAGE ABSTRACT
Unclassified Unclassified Unclassified UU
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18
i
ii
ANALYSIS AND SIMULATION OF DISADVANTAGED RECEIVERS FOR

MULTIPLE-INPUT MULTIPLE-OUTPUT COMMUNICATIONS SYSTEMS
Tracy A. Martin
Captain, United States Marine Corps
B.S., United States Naval Academy, 2004
Submitted in partial fulfillment of the

requirements for the degree of
MASTER OF SCIENCE IN ELECTRICAL ENGINEERING
from the
NAVAL POSTGRADUATE SCHOOL

September 2010
Author: Tracy A. Martin
Approved by: Frank Kragh

Thesis Advisor
Tri Ha
Second Reader
R. Clark Robertson
Chairman, Department of Electrical and Computer Engineering
iii
iv
ABSTRACT
This thesis investigates the simulation performance of Multiple-Input Multiple-Output

(MIMO) communication systems. Uncoded narrowband and wideband MIMO systems
are presented and simulated. The system is further refined with the addition of Space-
Time Codes (STC) and Channel State Information at the Transmitter (CSIT). A
disadvantaged receiver is subsequently introduced to the system lacking the optimization
enjoyed by the native receiver. Simulation and analysis was conducted with multiple
modulation schemes and antenna configurations. Rayleigh and Rician fading models are
developed and simulated as the wireless channel. Receiver performance results based on
MATLAB simulation are compared and presented.
v
vi
TABLE OF CONTENTS
I. INTRODUCTION........................................................................................................1
A. BACKGROUND ..............................................................................................1
B. OBJECTIVES AND METHODOLOGY ......................................................1
C. BENEFITS OF STUDY...................................................................................2
D. ORGANIZATION ...........................................................................................2
II. OVERVIEW OF MIMO COMMUNICATIONS .....................................................3
A. LOWPASS EQUIVALENCY .........................................................................3
B. THE MIMO SYSTEM MODEL ....................................................................7
C. MULTIPATH CHANNEL MODELS..........................................................11
1. Rayleigh Fading Model .....................................................................13
2. Rician Fading Model..........................................................................15
D. ORTHOGONAL FREQUENCY DIVISION MULTIPLEXING .............16
1. OFDM-MIMO Transmission............................................................17
2. OFDM Reception ...............................................................................19
E. MAXIMUM LIKELIHOOD DETECTION................................................20
F. SPACE TIME CODING ...............................................................................22
G. CHANNEL STATE INFORMATION AT THE TRANSMITTER..........24
H. SUMMARY ....................................................................................................27
III. ANALYSIS AND SIMULATION ............................................................................29
A. SIMULATION METHODOLOGY AND ASSUMPTIONS......................29
B. UNCODED NARROWBAND ......................................................................31
1. BPSK ...................................................................................................34
2. QPSK...................................................................................................35
3. 16QAM................................................................................................37
4. 64QAM................................................................................................39
C. SPACE-TIME CODED NARROWBAND ..................................................40
D. SPACE-TIME CODED ORTHOGONAL FREQUENCY DIVISION
MULTIPLEXING..........................................................................................45
E. NARROWBAND WITH CHANNEL STATE INFORMATION AT
THE TRANSMITTER ..................................................................................49
F. UNKNOWN RECEIVER..............................................................................53
G. SUMMARY ....................................................................................................58
IV. COMPARISON AND RESULTS.............................................................................61
A. RAYLEIGH AND RICIAN CHANNEL MODEL COMPARISON.........61
B. UNCODED AND SPACE-TIME CODED COMPARISON .....................63
C. CODED NARROWBAND AND WIDEBAND COMPARISON ..............66
D. TRANSMITTER WITH AND WITHOUT CSIT COMPARISON..........67
E. DISADVANTAGED RECEIVER COMPARISON ...................................70
1. Disadvantaged Receiver in a Rayleigh Channel .............................72
2. Disadvantaged Receiver in a Rician Channel (K = 1) ....................73
3. Disadvantaged Receiver in a Rician Channel (K = 4) ....................75
vii
F. SUMMARY ....................................................................................................77
V. CONCLUSION ..........................................................................................................79
A. SUMMARY ....................................................................................................79
B. SIGNIFICANT RESULTS............................................................................79
C. RECOMMENDATIONS FOR FUTURE WORK......................................81
LIST OF REFERENCES ......................................................................................................83
INITIAL DISTRIBUTION LIST .........................................................................................85
viii
LIST OF FIGURES
Figure 1. A bandpass signal (a) and the lowpass equivalent (b) (After [2]). ....................5
Figure 2. A modulator (a) and demodulator (b) (After [2]). .............................................6
Figure 3. MIMO System Block Diagram (After [2],[3]). .................................................8
Figure 4. Updated MIMO System Block Diagram (After [2],[3])..................................10
Figure 5. OFDM-MIMO Transmission Block Diagram (After [2]). ..............................18
Figure 6. CP of an OFDM Symbol (After [10])..............................................................19
Figure 7. OFDM-MIMO Reception Block Diagram (After [2]).....................................20
Figure 8. ISI Representation at the Receiver (After [10])...............................................20
Figure 9. Alamouti Space-Time Coding Scheme (After [14])........................................22
Figure 10. Equivalent MIMO model (After [4]). ..............................................................26
Figure 11. Signal Space for (a) BPSK, (b) QPSK, (c) 16QAM, (d) 64QAM. ..................30
Figure 12. MIMO System Block Diagram (After [2],[3]). ...............................................32
Figure 13. Simulated BER performance of a 2x2 MIMO system with BPSK
modulation. ......................................................................................................35
Figure 14. Simulated BER performance of a 2x2 MIMO system with QPSK
modulation. ......................................................................................................36
Figure 15. Simulated BER performance of a 2x2 MIMO system with 16QAM
modulation. ......................................................................................................38
modulation. ......................................................................................................40
Figure 17. Alamouti space-time coding scheme (After [14]). ..........................................41
Figure 18. Simulated BER performance of a 2x2 MIMO system with Alamouti
space-time coding and QPSK modulation. ......................................................43
space-time coding and 16QAM modulation. ...................................................44
Figure 21. Basic implementation of STC with MIMO-OFDM (After [5]).......................46
Figure 22. Simulated BER performance of a 2x2 MIMO-OFDM system with
Alamouti space-time coding and QPSK modulation. ......................................47
Figure 23. Simulated BER performance of a 2x2 MIMO-OFDM system with
Alamouti space-time coding and 16 QAM modulation...................................48
Figure 24. Simulated BER performance of a MIMO-OFDM system with Alamouti
Figure 25. Closed loop model. ..........................................................................................50
Figure 26. Dominant eigenmode transmission in a 2x2 MIMO system. ..........................52
modulation using dominant eigenmode transmission when the transmitter
has channel state information...........................................................................53
Figure 28. A fixed MIMO system optimized for channel AB. .........................................54
Figure 29. Disadvantaged receiver in a fixed MIMO system optimized for channel
AB. ...................................................................................................................55
ix
Figure 30. Simulated BER performance of the disadvantaged receiver in a Rayleigh
channel, for 2x2, 2x4, and 2x8 MIMO systems...............................................56
Figure 31. Simulated BER performance of the disadvantaged receiver in a Rician
channel (K=1), for 2x2, 2x4 and 2x8 MIMO systems.....................................57
Figure 32. Simulated BER performance of the disadvantaged receiver in a Rician
channel (K=4), for 2x2, 2x4, and 2x8 MIMO systems....................................58
Figure 33. Simulated BER performance of various modulation schemes with different
channel characteristics, 2x2 MIMO system.....................................................63
Figure 34. Simulated BER performance of uncoded and coded 2x2 MIMO system
with QPSK modulation in a Rayleigh channel. ...............................................64
Figure 35. Simulated BER performance of a 2x2 MIMO system, various modulation
schemes, uncoded and using Alamouti space-time coding, with different
channel characteristics. ..................................................................................66
Figure 36. Simulated BER performance of a 2x2 MIMO system using Alamouti
space-time coding, wideband and narrowband in various channels with
64QAM modulation. ........................................................................................67
Figure 37. Simulated BER performance of a transmitter with and without CSIT for a
2x2 MIMO system in various channels using QPSK modulation. ..................68
Figure 38. Simulated BER performance of a 2x2 MIMO system with CSIT to that of
a system with Alamouti space-time coding both in a Rayleigh channel
using QPSK modulation. .................................................................................70
Figure 39. Disadvantaged receiver in a fixed MIMO system optimized for channel
AB. ...................................................................................................................71
Figure 40. Simulated BER performance of a disadvantaged receiver in a fixed MIMO
system, Rayleigh channel.................................................................................72
system, Rician channel (K = 1)........................................................................74
Figure 42. BER performance of a disadvantaged receiver in a fixed MIMO system,
Rician channel (K = 4).....................................................................................76
x
LIST OF TABLES
Table 1. Characteristics of channel fading (After [6],[7]). ............................................13

Table 2. Alamouti scheme transmission sequence (After [14]).....................................23
Table 3. Eb N o required to achieve various BER for different channel models, 2x2
MIMO system with QPSK modulation............................................................62
Table 4. Eb N o required to achieve a certain BER in a Rayleigh channel, 2x2
MIMO system using QPSK modulation, with and without Alamouti space-
time coding.......................................................................................................65
Table 5. Eb N o required to achieve a BER in Rayleigh channel and Rician
channels when the system has CSIT, 2x2 MIMO system. ..............................69
Table 6. Eb N o comparison for select values of BER, disadvantaged receiver with
various antenna configurations over a Rayleigh fading channel. ....................73
various antenna configurations over a Rician channel (K = 1)........................75
various antenna configurations over a Rician channel (K = 1)........................77
xi
xii
LIST OF ACRONYMS AND ABBREVIATIONS
A/D Analog to Digital
AWGN Additive White Gaussian Noise
BER Bit Error Rate
BPSK Binary Phase-Shift Keying
CDF Cumulative Distribution Function
CP Cyclic Prefix
CSI Channel State Information
CSIR Channel State Information at the Receiver
CSIT Channel State Information at the Transmitter
D/A Digital to Analog
DFT Discrete Fourier Transform
FFT Fast Fourier Transform
IDFT Inverse Discrete Fourier Transform
IFFT Inverse Fast Fourier Transform
IID Independent Identically Distributed
ISI Intersymbol Interference
LOS Line of Sight
NLOS Non-Line of Sight
ML Maximum Likelihood
MIMO Multiple-Input Multiple-Output
MISO Multiple-Input Single-Output
OFDM Orthogonal Frequency Division Multiplexing
PSD Power Spectral Density

xiii
PDF Probability Density Function
QAM Quadrature Amplitude Modulation
QPSK Quadrature Phase-Shift Keying
RF Radio Frequency
SISO Single-Input Single-Output
SNR Signal to Noise Ratio
STC Space-Time Code
SVD Singular Value Decomposition
WSS Wide-Sense Stationary
xiv
EXECUTIVE SUMMARY
Multiple-Input Multiple-Output (MIMO) communications systems are fundamentally

different from earlier wireless systems. In MIMO systems, multiple antennas are used
for transmission and reception, thus multiple communication paths or channels are
developed. These channels can then be utilized to increase the data rate or reliability of
the system. Data rate is increased by sending different symbols over multiple channels at
the same time. On the other hand, reliability is increased by sending multiple copies of
the same symbol, or similar redundancy, over the multiple channels.
This ability of MIMO systems to provide increased reliability and data rate
without requiring an increase in bandwidth or power has resulted in MIMO schemes
being adopted in standards such as IEEE 802.11n, IEEE 802.16, and Long Term
Evolution (LTE) systems, representing most of the WLAN, WMAN, and 4G mobile
telephone industry [1]. Our adversaries use conventional wireless devices, and we can
expect them to use these MIMO devices also, as they grow in market dominance. We
benefit from a thorough understanding of our adversaries tools. Furthermore, MIMO
systems have the potential to greatly improve signal reception and reliability for friendly
forces. As the military continues to develop and further rely on high data rate, highly
mobile communications, signal reception and reliability is often found to be the weakest
link. Mobile networking, in particular, relies heavily on reliable delivery, otherwise
routing protocols degenerate and retransmissions swamp the network. By developing and
utilizing a robust MIMO system, decision makers can be assured that timely and critical
communications will get through. This work seeks to expand that understanding.
Specifically, this thesis evaluates the bit error rate performance of various MIMO
system configurations into which a second, disadvantaged receiver is introduced. The
analysis begins with a basic two transmit and two receive antenna MIMO system
utilizing maximum likelihood decoding. The system is simulated in MATLAB
employing BPSK, QPSK, 16QAM and 64QAM over both Rician and Rayleigh channel
models in order to demonstrate and analyze the bit error rate performance. The system is
improved with the application of the Alamouti space-time coding scheme and simulated.
xv
A comparison of the uncoded and coded systems is conducted. Specifically, when
comparing a 2x2 MIMO system utilizing QPSK to that of Alamouti space time coded 2x2
MIMO system utilizing QPSK, we see that the coded system provides the same
performance while requiring 10 dB less power. The performance improvement seen is
due to the diversity gain provided by the Alamouti scheme. Additionally, the
orthogonality of the coding sequence transforms the decision at the receiver from a vector
operation into a scalar operation, reducing the receiver complexity.
Next, the system is further refined with the addition of channel state information
at the transmitter. This information allows the transmitter to optimize its transmission
according to the channel conditions, further improving bit error performance. Results
show that a system with channel state information needs 12 dB less power than the
equivalent uncoded system for equal performance. In addition, knowledge of the channel
provides the receiver with an array gain equal to 2 dB, resulting in increased performance
over that of the space-time coded system without channel state information.
Into this robust system, a second receiver is introduced that does not enjoy the
same optimization as the intended receiver. This would be important whenever there are
multiple receivers attempting to receive the same transmission at distinct locations, as
might be the desired configuration in a wireless network employing broadcast
transmissions. The performance was evaluated over both Rayleigh and Rician fading
channels with various antenna configurations. In the case of a Rayleigh channel, when
the number of receive antennas is increased to eight, the performance is 3 dB worse than
the intended receiver. In a Rician channel, where the line of sight path has equal power
to that of the scattered paths, results show that when the receiver employs four antennas
the performance increases to 2 dB better than the intended receiver. Finally, when the
disadvantaged receiver is simulated at a location receiving more signal power via line of
sight from the transmitter with two receive antennas, the second receiver performs as well
as the intended receiver. Thus, depending on the configuration of the disadvantaged
receiver it can achieve an equal or superior bit error rate performance than the intended
receiver.
xvi
ACKNOWLEDGMENTS
I would like to thank my advisor, Professor Frank Kragh, for his guidance and
support during the development of this thesis.
xvii
xviii
I. INTRODUCTION
A. BACKGROUND
Multiple-input Multiple-output (MIMO) communications systems are

fundamentally different from earlier wireless systems. As their name implies, MIMO
systems are designed to have additional antennas at the transmitter and receiver,
providing spatial diversity not available to single antenna systems. Thus, information can
be sent and received over multiple channels.
This gives the communications engineer greater opportunity for robust system
designs. One such example is the capability of a MIMO system to transmit at a higher
rate and more reliably by using the stronger (less faded) channels more and the weaker
(more faded) channels less. As it is discussed in this thesis, that implies the transmission
is optimized for the location of one receiver, which is unfortunate in the case of broadcast
transmissions. This research seeks to determine how a second receiver would perform
given that the transmission is optimized for the location of the first receiver.
B. OBJECTIVES AND METHODOLOGY
The goal of this work was to investigate the performance of a second receiver
when presented into an existing MIMO system optimized for the first receiver. The
research approach takes a basic system, integrates techniques to improve performance
and concludes with the introduction of a second receiver into the system.
Development begins with the simulation of a simple uncoded narrowband MIMO

system over Rician and Rayleigh fading channels. To which the Alamouti space-time
coding scheme is applied and simulated. Then, the application of Orthogonal Frequency
Division Multiplexing (OFDM) is investigated. The system is completed by introducing
Channel State Information at the Transmitter (CSIT), where the result is a more robust
communications link. Finally, a second receiver is introduced to the optimized system
and the performance is simulated. Various scenarios are then investigated in an effort to
improve the disadvantaged receiver performance.
1
C. BENEFITS OF STUDY
MIMO systems are becoming increasingly commonplace, including IEEE

802.11n, IEEE 802.16, and Long Term Evolution (LTE) systems, which represent most
of the WLAN, WMAN, and 4G mobile telephone industry [1]. Our adversaries use
conventional wireless devices, and we can expect them to use these MIMO devices also,
as they grow in market dominance. We benefit from a thorough understanding of our
adversaries tools. This work seeks to expand that understanding. Furthermore, MIMO
systems have the potential to greatly improve signal reception and reliability for friendly
forces. As the military continues to develop and further rely on high data rate, highly
mobile communications, signal reception and reliability is often found to be the weakest
link. Mobile networking, in particular, relies heavily on reliable delivery, otherwise
routing protocols degenerate and retransmissions swamp the network. By developing and
utilizing a robust MIMO system, it is highly likely that timely and critical
communications will get through.
D. ORGANIZATION
The thesis is organized into five chapters. Chapter II introduces the MIMO
system. It provides a brief background into methods used throughout the thesis. These
include channel models, OFDM-MIMO, maximum likelihood (ML) detection, space time
coding and channel state information. Chapter III discusses the analysis and simulation
of MIMO receivers that involve various configurations of the methods introduced in the
previous chapter. Chapter IV examines each of these configurations and provides a
comparison of the results. Chapter V provides a summary and gives recommendations
for future study.
2
II. OVERVIEW OF MIMO COMMUNICATIONS
This chapter provides background on the techniques used in designing and

analyzing MIMO systems. Specifically, this chapter introduces a lowpass equivalent
MIMO system model upon which the remaining topics will build. Rayleigh and Rician
fading channels are then discussed, providing a framework for future analysis. Next,
OFDM implementation utilizing the Discrete Fourier Transform (DFT) is presented.
Maximum Likelihood (ML) detection at the receiver is introduced. Then, space-time
codes are developed using the Alamouti scheme. Finally, channel state information is
defined for application at the transmitter.
A. LOWPASS EQUIVALENCY
Wireless channels often have different spectral characteristics than that of the
information signal. Therefore, before entering the channel it is often necessary to match
the spectral characteristics of the signal to that of the channel. In other words, the
information signal is at baseband, a lower frequency, and needs to be converted to
passband, a higher frequency for transmission. At the receiver, the reverse takes place,
converting from passband to baseband. Thus, when analyzing or designing
communication systems it is common practice to represent the system in lowpass
equivalent form. Representing the system in this manner greatly simplifies analysis since
the dependency on the carrier frequency is eliminated. Additionally, signal processing
limitations require that we manipulate the information signal at lower frequencies in
order to maintain reasonable sampling rates. This section will develop an equivalent
lowpass form for bandpass signals [2], [3].
Given a bandpass signal g ( t ) , whose Fourier transform is G ( f ) , then the
analytic signal is defined as [2]

g + (t ) = g (t ) + j g (t ) (2.1)
3
where g + ( t ) denotes only positive frequency components are present and g ( t )
1
represents the Hilbert Transform of g ( t ) , i.e., g ( t ) = * g ( t ) , where the asterisk
t
denotes convolution.
Taking the Fourier transform of Equation (2.1) yields [1]

G+ ( f ) = G ( f ) + j ( j sgn ( f ) G ( f ) ) (2.2)
.
Equation (2.2) can be simplified by using the property, sgn ( f ) = 2u ( f ) 1, where
u ( f ) is the unit step function, to get
G+ ( f ) = 2G ( f ) u ( f ) (2.3)
From Equation (2.3) it is clear that the analytic signal represents the bandpass signal at
twice the amplitude and consisting of positive frequencies only. Next, let the lowpass
equivalent of g ( t ) , denoted as gl ( t ) , be defined as [2]
gl ( t ) = g+ ( t ) e j 2 fct (2.4)
Taking the Fourier transform and substituting Equation (2.3) results in

Gl ( f ) = 2G ( f + f c ) u ( f + f c ) (2.5)
Equation (2.5) shows in the frequency domain, the lowpass equivalent signal or
complex envelope is just a scaled and shifted version of the components of the bandpass
signal that have positive frequencies. The positive frequency components contain all the
information necessary to reconstruct the original signal. In this case, it is scaled by two
and shifted from the carrier frequency to the zero frequency or baseband. This result is
illustrated in Figure 1.
4
Figure 1. A bandpass signal (a) and the lowpass equivalent (b) (After [2]).
In the time domain
( )
g l ( t ) = g ( t ) + j g ( t ) e j 2 f c t
.
(2.6)
Rearranging and isolating the real part of Equation (2.6) produces a representation of a
bandpass signal in terms of its lowpass equivalent [2]
g ( t ) = Re g l ( t ) e j 2 f c t (2.7)
.
Additionally, the lowpass equivalent signal can be represented by the in-phase

and quadrature components of the baseband signal as [2]
gl ( t ) = gi ( t ) + jgq ( t ) (2.8)
.
Substituting Equation (2.8) into Equation (2.6) yields [2]
gi ( t ) = g ( t ) cos ( 2 fct ) + g ( t ) sin ( 2 fct ) and

(2.9)
gq ( t ) = g ( t ) cos ( 2 fct ) g ( t ) sin ( 2 fct ) .
Solving for g ( t ) and g ( t ) yields [2]
g ( t ) = g i ( t ) cos ( 2 f c t ) g q ( t ) sin ( 2 f c t )
(2.10)
g ( t ) = g q ( t ) cos ( 2 f c t ) + g i ( t ) sin ( 2 f c t )
or, in polar form, [2]
5
g ( t ) = rg ( t ) cos ( 2 f c t + g ( t ) ) (2.11)
where rg ( t ) and g ( t ) represent the envelope and phase of g ( t ) .
Thus, we observe that Equation (2.10) defines the modulation process while
Equation (2.9) defines the demodulation process and original signal can be completely
described in lowpass equivalent form given by Equation (2.7). This conclusion is shown
in Figure 2. [2].
Figure 2. A modulator (a) and demodulator (b) (After [2]).
Next, we develop how the energy of the signal is represented when the signal is
expressed in lowpass equivalent form.
6
Given that the energy of a signal g ( t ) is defined as [2]

g ( t ) dt
2
g = (2.12)

applying Parseval's Theorem yields [2]

g (t ) G( f )
2 2
g = dt = df . (2.13)

Since G ( f ) = G+ ( f ) + G ( f ) and G+ ( f ) G ( f ) = 0 then [2]

G ( f )+G ( f )
2
g = + df

G (f) G (f)
2 2
= + df + df

= 2 G+ ( f ) df
2
(2.14)

Gl ( f )
2
=2 df

2
1
= gl .
2
This result confirms Equation (2.5): the energy of the lowpass equivalent signal is
double that in the bandpass signal. In the remaining analysis, signals and systems will be
represented by their lowpass equivalent form.
B. THE MIMO SYSTEM MODEL
A MIMO system comprised of NT spatially separated transmit antennas and N R

spatially separated receive antennas is depicted in Figure 3.
At the transmitter, the symbol transmitted on antenna n is denoted sn (t) , where
n = 1,2,..., NT and t denotes the time variable. The equivalent lowpass channel impulse
response between the receive antenna m , and the transmit antenna n , is denoted as
hmn ( ; t ) . Where denotes the delay variable.
7
Figure 3. MIMO System Block Diagram (After [2],[3]).
The composite of the complex channel gains is then given by the N R NT matrix,
H ( ; t ) which is also known as the channel matrix [2]
h11 ( ; t) h12 (; t) L h1NT (; t)

h21 (; t) h22 (; t) K h2 NT (; t)
H ( ; t) = (2.15)
M M O M

hN 2 ( ; t) hN R 2 (; t) L hN R NT (; t)
R .
The channel matrix is often described with statistical models that accurately
predict channel characteristics. This thesis will focus on Rayleigh and Rician fading
models, discussed in the next chapter.
The signal at the m th receive antenna, ym , is the sum of the transmitted signals
corrupted by the channel and noise, resulting in [2]
NT
ym (t ) = hmn (; t) sn (t )d
n =1
(2.16)
NT
= hmn (; t) * sn (), m = 1, 2..., N R
n =1
8
where the asterisk implies convolution. The system can therefore be expressed as a
system of linear equations as
y1 (t ) h11 (; t) h12 (; t) L h1NT (; t) s ()

y (t ) h (; t)
1
h22 (; t) K h2 NT (; t) s2 ()
2 = 21 (2.17)
M M M O M * M

y N R (t ) hN R 2 (; t) hN R 2 (; t) L hN R NT (; t) sNT ()
or more conveniently in matrix form as [2]

y (t ) = H ( ; t) * s ( ) (2.18)
where y represents the N R 1 received matrix and s represents the NT 1 sent matrix
[2].
The result above provides a baseline MIMO channel model. To refine this model
we make the following assumptions, taken largely from [2], that will be used throughout
this thesis:
1. Thermal noise is present at each receive antenna and is modeled as additive
white Gaussian noise (AWGN), denoted as m . Further, the AWGN can be
characterized as independent identically distributed (IID), each having zero
mean and two sided power spectral density 2 N o .
2. The receiver utilizes a matched filter, whose output is sampled at the end of
each symbol interval.
3. The intended receiver has perfect channel estimation. Though this is not
practical, it can be very accurately approximated through the use of training
symbols.
4. Antenna spacing is sufficient so that the fading on the subchannels are
uncorrelated.
Using these assumptions, the baseline MIMO model can be updated as shown in Figure
4.
9
Figure 4. Updated MIMO System Block Diagram (After [2],[3]).
Therefore, hmn ( t ) can be simplified to hmn , and the composite of the channel coefficients
is given by the N R NT matrix, H [2]
h11 h12 L h1NT

h21 h22 K h2 NT
H= (2.19)
M M M M

hN 1 hN R 2 L hN R NT
R .
The N R 1 noise matrix, is given as [2]
1

=
2
(2.20)
M

N R .
The received signal, ym , at each antenna is the sum of the transmitted signals corrupted
by the channel and noise, resulting in [2]

NT
ym = sn hmn + m , m = 1, 2..., N R . (2.21)
n =1
This can be expressed as a system of linear equations

10
y1 h11 h12 L h1NT s
y h
1 1

+ 2
K
2 = 21
h22 h2 NT s2
(2.22)
M M M M M M M

N R hN R 1
y hN R 2 L hN R NT sNT N R
or, more conveniently, in matrix form as [2]

y = Hs + . (2.23)
This result will be used throughout the thesis.
C. MULTIPATH CHANNEL MODELS
Over a wireless channel, a radio frequency (RF) signal can take numerous paths to
its destination. This is due to objects in the propagation path reflecting, diffracting, and
scattering the radio frequency signal as it travels from the transmit antenna to the receive
antenna. These paths result in multiple copies of the transmitted signal arriving at the
receiver that vary randomly in amplitude, phase and time. Collectively, this is called
multipath [5], [7].
Variations in amplitude can produce signals that are too weak to be detected.
Differences in phase produce signals that constructively and destructively interfere.
Additionally, deviation in arrival time often results in intersymbol interference. That is,
symbols which overlap each other in time. In the frequency domain, relative motion of
the transmitter, receiver, or the objects between them results in a Doppler shift
proportional to the relative velocity of the objects. The significance of these effects are a
function of the transmitted signal and either the coherence time (TC ) or bandwidth ( BC ) .
That is, how quickly the channel changes in time and frequency respectively. This allows
the effects of the fading channel to be classified by four types: slow fading, fast fading,
flat fading, and frequency selective fading. This fading classification, relative to the
signal bandwidth ( BS ) and symbol time ( TS ), is depicted in Table 1. [2], [5], [7].
The coherence time of the channel, TC depends on the Doppler spread, BD
according to the relationship [6]
11
1
TC = (2.24)
4BD
where the Doppler spread is a measure of the spectral expansion due to the relative
motion between the transmitter and receiver. If the coherence time is short relative to the
symbol duration, then the channel impulse response changes rapidly over the symbol
period, causing distortion. The channel is said to be fast fading. On the other hand, if the
coherence time is large relative to the symbol period then the channel can be considered
static over multiple symbols. The channel is said to be slow fading. Thus, classification
of a channel as either fast or slow fading is determined by how quickly the channel is
changing in time and the baseband signaling utilized [5][7].
The coherence bandwidth of the channel, BC depends on the multipath delay
spread, TD by the relationship [6]
1
BC = (2.25)
TD
where the multipath delay spread is a measure of the difference of arrival times between
the longest and shortest propagation paths. Only paths with significant energy are
considered. If propagation delay is large (i.e., large TD ) with respect to the symbol
duration, then multiple symbols overlap each other at the receiver, causing distortion.
The channel is said to be frequency selective. Conversely, if the propagation delay is
small compared to the symbol duration, then each path that the symbol takes arrives at
the receiver before the next symbol. The channel is said to be frequency non-selective or
flat faded. Classification of the channel as either frequency selective or flat faded is
determined by how quickly the channel is changing in frequency and the baseband
signaling utilized [5][7].
12
Table 1. Characteristics of channel fading (After [6],[7]).
Therefore, a channel can be characterized and is often referred to by these

metrics. This thesis will focus on two statistical models that can be used to represent a
frequency non-selective multipath channel: Rayleigh fading and Rician fading.
Additionally, the channel is considered slowly fading and the complex channel gain is
constant over consecutive symbols.
1. Rayleigh Fading Model
When the communicating antennas do not have a direct line of sight (NLOS) to
each other, the received envelope can be modeled as a Rayleigh random process and is
referred to as Rayleigh fading. In order to model Rayleigh fading, we utilize the sum of
two IID Gaussian random variables with zero mean.
Let Z NLOS represent the sum of two IID Gaussian random variables with zero
mean and equal variance
ZNLOS = X + jY (2.26)
where X ~ N ( 0, 2 ) , Y ~ N ( 0, 2 ) and the magnitude R is given by
R NLOS = X 2 +Y2 (2.27)

.
13
In order to relate Equation (2.27) to a distribution, we introduce the joint cumulative
distribution function (CDF), defined as [9]
FX1 X 2 ( x1 , x2 ) Pr [ X 1 x1 , X 2 x2 ] (2.28)
where the CDF is related to the joint probability density function (PDF) by [9]
2 FX1 X 2 ( x1 , x2 )
f X 1 X 2 ( x1 , x2 ) = (2.29)
x1x2 .
Solving for the CDF yields [9]
x1 x2
FX1 X 2 ( x1 , x2 ) = f X1 X 2 ( x1 , x2 ) dx1dx2 (2.30)

thus, expressing Equation (2.27) in terms of the CDF results in

FR ( r ) = f XY ( x, y ) dxdy (2.31)
X +Y r
2 2
For r 0 . Since X and Y are Gaussian random variables, we may further refine
Equation (2.31) as
2 x2 y2
1
FR ( r ) = e 2 e 2 dxdy
2 2
(2.32)
2
2
X +Y r
2 2
.
Simplifying Equation (2.32) and performing a change of variables results in
r 2
FR ( r ) = 1 e 2 u ( r )
2
(2.33)

where the unit step function u is given by

1 r0
u (r ) = (2.34)
0 r < 0.
Finally, given Equation (2.33), solving for the PDF yields

r2
r
fR ( r ) = e 2 2
u (r ) (2.35)
2
14
which is the PDF of a Rayleigh distribution. Thus, the sum of two IID complex Gaussian
random variables with zero mean as given by Equation (2.26), accurately model Rayleigh
fading and provide the desired channel model when NLOS components are present at the
receiver [2].
2. Rician Fading Model
For channels with a line of sight (LOS) path in addition to NLOS, the Rician
model is more appropriate. However, due to the LOS path, the Gaussian random
variables no longer can be modeled with equal mean.
The Rician model follows a Rician distribution given by [7]

(
r 2 + A2 )
r Ar
fR ( r ) = e 2 2
I0 2 u ( r ) (2.36)
2

where A 0 and I0 ( ) represents the modified Bessel function of the first kind and zero
order [7].
Alternatively, this distribution can be described in terms of the Kfactor, which is

defined as the ratio of the energy in the specular (LOS path) to the energy in the diffuse
(NLOS paths), given by [8]
A2
K . (2.37)
2 2
The second moment or average received power is the total power of the specular
and diffuse paths, given by
Pr = A2 + 2 2 . (2.38)
Solving for A2 and 2 2 and substituting Equation (2.37) results in

KPr
A2 =
K +1 (2.39)
P
2 2 = r
K +1 .
15
Now, the Kfactor can be used to represent a complex random variable whose
magnitude is a Rician random variable, given in Equation (2.36), by utilizing Equation
(2.39) as [8]
N ( 0, 2 )
K 1
e j + (2.40)
K +1 K +1
where the first term is from the specular path and the second term is from the diffuse
path. The diffuse path is equivalent to Rayleigh fading described in the previous section.
Thus, by Equation (2.40), Rician fading can be modeled by the sum of a LOS path
with uniform phase and NLOS paths as characterized by Equation (2.26) scaled by the
Kfactor. It is easily seen that if K = 0 , Equation (2.40) reduces to a zero mean complex
Gaussian, which can be used to model the hmn , for complex channel gains. Additionally,
as K becomes very large, Equation (2.40) becomes deterministic and would be used for
the nonfaded case [2].
D. ORTHOGONAL FREQUENCY DIVISION MULTIPLEXING
A primary reason for employing MIMO systems is to achieve higher data rates
when compared to single antenna systems. However, as the data rate increases, the
symbol time decreases relative to the multipath delay spread. As discussed in the
previous section, this produces frequency selective fading which leads to intersymbol
interference (ISI) and deep fades, severely degrading performance [10],[11].
To mitigate ISI, multicarrier modulation schemes have been developed that divide
the large bandwidth signal into L smaller subchannels, where the subchannels are
BS
characterized by BC . That is, the bandwidth of each subchannel is equal to the
L
signal bandwidth ( BS ) divided by the number of total subchannels (L). Thus, when L is
selected to be sufficiently large, the bandwidth of each subchannel is less than the
coherence bandwidth of the channel and each subchannel experiences flat fading
conditions. The subchannels are then transmitted in parallel such that the bandwidth of
the subchannels sum to that of the original signal without the effects of ISI. When the
16
subchannels are separated in frequency and manipulated to be orthogonal to each other,
the multicarrier modulation scheme is called orthogonal frequency division multiplexing
(OFDM) [3],[10].
Let sk ( t ) represent the carrier signal on the kth subcarrier [2]
sk ( t ) = cos 2 f k t (2.41)
where k = 0,1, 2,K , L 1 and f k is the center frequency of the kth subchannel. The
subchannels are orthogonal and independent of the phase between subchannels, over a
symbol interval T , when the integral of their product over that same interval is zero,
given by [2]
T
cos ( 2 f t + ) cos ( 2 f t + ) dt = 0
0
k k j j
(2.42)
where k and j represent the phase of the kth and jth subchannel respectively and are
independent of f k and f j . Selecting the OFDM subchannel frequency spacing to be
1
and maintaining the subchannel center frequencies to be integer multiples of each other
T
1
(i.e., f k +1 f k = k = 0, 2,K , L 2 ) results in the peak of each subchannel
T
corresponding to the nulls of adjacent subchannels. Thus, a set of L parallel symbols
1
spaced apart produces L orthogonal overlapping signals, providing bandwidth
T
efficiency and minimizing the effects of ISI [2].
1. OFDM-MIMO Transmission
A block diagram showing the transmission OFDM-MIMO system with two

transmit antennas ( NT = 2) is given in Figure 5. At the input, data is converted from a
binary stream into a set of L parallel symbols divided between each antenna and
modulate, where the output of the modulator is of the form X [ 0] , X [1] ,K , X [ N 1] and
17
1
the symbol rate is reduced by a factor of containing bk bits. Modulation is typically
L
chosen to be consistent across subcarriers. However, it is possible to modulate each
subcarrier with a different scheme.
When L is selected to be an integer multiple of two, then an efficient method

known as the Inverse Fast Fourier Transform (IFFT) can be used to implement the
inverse discrete Fourier transform (IDFT) given by
j 2 lk
1 L1
x [l ] = X [ k ]e L l = 0,1,K, L 1 (2.43)
L k =0
As can be seen from Equation (2.43), the IDFT takes the components of each
subcarrier and produces equivalent time domain symbols that have orthogonal properties
mentioned previously.
Figure 5. OFDM-MIMO Transmission Block Diagram (After [2]).
The next step in the transmission process is the addition of a cyclic prefix (CP).
The CP is created by replicating the data contained in the last part of the symbol and
superimposing this copy to the front of the symbol as depicted in Figure 6. The total
symbol duration is given by
Ttot = TCP + TS . (2.44)
This is done to guard against the inherent delay of a multipath channel. As

introduced in previous sections, as the symbol propagates through the wireless channel, a
delay is incurred at the receiver. The cyclic prefix is selected to be of sufficient length as
18
to prevent the delay of one symbol from corrupting the following symbol. In other
words, the cyclic prefix is used as a guard interval to keep the transmitted symbols
independent. Obviously, in order to do this, the CP length must be chosen greater than
the multipath delay spread of the channel. Another important aspect of the CP and the
reason it is appended to the front of a symbol is to protect the start point of the symbol
which aids in bit decisions at the demodulator [10].
Figure 6. CP of an OFDM Symbol (After [10]).
The symbols are then converted from parallel to a serial sequence, passed through
the digital to analog (D/A) converter, and unconverted to radio frequency (RF) for
transmission.
2. OFDM Reception
A block diagram showing the OFDM-MIMO system reception process with two
receive antennas ( N R = 2) is given in Figure 7. After reception, the signal is down
converted and sent through the analog to digital (A/D) converter. The signal is then
transformed from a serial sequence into parallel subchannels, where the CP is removed
and the discrete Fourier transform (DFT) is implemented with the Fast Fourier Transform
(FFT) which is a Fourier pair with the IDFT. Where the DFT is given by
L 1 j 2 lk
X [ k ] = x [l ]e L
(2.45)
l =0
Finally, the subchannels are demodulated and converted back into the binary bit
stream.
19
Figure 7. OFDM-MIMO Reception Block Diagram (After [2]).
An OFDM system must maintain synchronization and orthogonality in order to

accurately reproduce the original signal. If implemented correctly, the CP will maintain
orthogonality by preventing symbol overlap as illustrated by Figure 8. Synchronization
can be maintained by allocating OFDM symbols for this purpose. This thesis will
assume that the receiver has perfect synchronization and the orthogonality of the symbols
is maintained.
Figure 8. ISI Representation at the Receiver (After [10]).
E. MAXIMUM LIKELIHOOD DETECTION
Recall that the received signal, ym at each antenna in a MIMO system is the sum
of the transmitted signals corrupted by the channel and noise. This was expressed in
matrix form by Equation (2.23) and is repeated here, y = Hs + .
The corruption introduced by the wireless channel requires that the receiver must
statistically decide on which symbol was sent based on what was received. To make this
decision, the receiver utilizes knowledge of the channel and compares that to the received
20
signal. For the MIMO model presented in this thesis the optimal method to find this
decision statistic is known as Maximum Likelihood (ML) detection. [2]
Bayes' rule forms the basis of the ML detector and is given by [9]
Pr [ X 2 | X 1 ] Pr [ X 1 ]
Pr [ X 1 | X 2 ] = (2.46)
Pr [ X 2 ]
.
Let Pr [ X 1 ] represent the a priori probability that a particular symbol vector s
was sent, where s is selected from a known signal space. Let Pr [ X 2 ] denote the a priori
probability that a particular N R 1 decision statistic vector y was received.
Additionally, let Pr [ X 2 | X 1 ] indicate the a posteriori probability of y given s was
observed [12].
Observing that sent symbol s is selected from a finite set of values and the
received vector y is stochastic, Equation (2.46) can be represented more accurately as
f ( y | s ) Pr ( s )
Pr ( s | y ) = (2.47)
f (y)
Therefore, the detector needs to select the sent symbol vector that maximizes the
probability of the sent symbol vector, given the received vector. Updating Equation
(2.47) results in [12]
f ( y | s ) Pr ( s )
s$ = arg max (2.48)
s f (y)
where s$ denotes the estimate out of the detector. Assuming that the symbols are sent
with equal probability, Equation (2.48) becomes [2]
s$ = arg max f ( y | s ) (2.49)

s .
Since the noise at the receiver is assumed to be IID Gaussian random variables
with zero mean, Equation (2.49) can be represented as the minimization of the distance to
the stationary vector y , as [13]
s$ = arg min y Hs (2.50)

s .
21
Thus, the detector compares every possible symbol in the signal space to the received
symbol and selects the closest. Alternatively, Equation (2.50) can be expressed in terms
of each received symbol [2]
NR NT 2
s$ = arg min ym hmn sn . (2.51)

s
m =1 n =1
F. SPACE TIME CODING
Space-time coding implies multiple antennas either at the transmitter, receiver, or

both. They exist to exploit transmit diversity in order to increase channel capacity and
reliability. This thesis concentrates on the Alamouti space-time code, introduced in [14].
Figure 9. depicts the scheme when applied to a MIMO system consisting of two
transmitters and two receivers. At time t symbols s1 and s2 are transmitted
simultaneously. At time t + T , the negative complex conjugate of s2 and the complex
conjugate of s1 are transmitted. The result is a sequence of signals shown in Table 2.

[14].
Figure 9. Alamouti Space-Time Coding Scheme (After [14]).
Assuming that the channel remains constant over at least two consecutive
symbols, the received signals can then be expressed as [14]
22
y1 ( t ) = h11s1 + h12 s2 + 1 ( t )
y2 ( t ) = h21s1 + h22 s2 + 2 ( t )
(2.52)
y1 ( t + T ) = h11s2* + h12 s1* + 1 ( t + T )
y2 ( t + T ) = h21s2* + h22 s1* + 2 ( t + T )
where y1 ( t ) and y1 ( t + T ) correspond to the received signals at antenna one at time t
and t + T respectively. Additionally, 1 ( t ) and 1 ( t + T ) represent the noise
components at antenna one at time t and t + T respectively. As mentioned in the

previous section, hmn indicates the channel gain to antenna m from antenna n .
Table 2. Alamouti scheme transmission sequence (After [14]).
The received signals are combined and compared. Resulting in symbol estimates
given by [14]
s1 = h11* y1 ( t ) + h12 y1 ( t + T ) + h21* y2 ( t ) + h22 y2 ( t + T )

* *
(2.53)
s2 = h12* y1 ( t ) h11 y1 ( t + T ) + h22* y2 ( t ) h21 y2 ( t + T )
* *
Substituting Equation (2.52) into Equation (2.53), yields
(
s1 = h11* ( h11s1 + h12 s2 + 1 ( t ) ) + h12 h11* s2 + h12* s1 + 1 ( t + T )
*
)
(
+ h21* ( h21s1 + h22 s2 + 2 ( t ) ) + h22 h21* s2 + h22* s1 + 2 ( t + T )
*
) (2.54)
(
s2 = h12* ( h11s1 + h12 s2 + 1 ( t ) ) h11 h11* s2 + h12* s1 + 1 ( t + T )
*
)
(
+ h22* ( h21s1 + h22 s2 + 2 ( t ) ) h21 h21* s2 + h22* s1 + 2 ( t + T )
*
)
simplifying
23
(
s1 = h11 + h12 + h21 + h22
2 2 2 2
)s
1
+ h11*1 ( t ) + h12 1 ( t + T ) + h21*2 ( t ) + h22 2 ( t + T )

* *
(2.55)
(
s2 = h11 + h12 + h21 + h22
2 2 2 2
)s
2
h111 ( t + T ) + h12*1 ( t ) h212 ( t + T ) + h22*2 ( t + T )

* *
.
From Equation (2.55) it is clear that the combined symbols are heavily affected by the
channel gains. Therefore, the Alamouti scheme provides transmit diversity such that the
effects of multipath fading are minimized. The symbols s1 and s2 are sent to the
maximum likelihood detector, discussed in the previous section, and results in symbol
estimates s1 and s2 .
Another benefit of the Alamouti scheme is that it achieves full diversity while
maintaining low complexity at the receiver. This is due to the orthogonality of the
sequence of signals, shown in Table 2. The coding matrix is given by [14]
s s2*
G= 1 (2.56)
s2 s1*
where G is the Alamouti generator matrix for a 2 2 MIMO system. It can be seen that
the row vectors are orthogonal
v1 v2* = 0 (2.57)
where v1 = s1 s 2 * and v2 = s2 s1* . Due to this orthogonality, the maximum
likelihood detector no longer needs to compare each possible transmitted symbol to the
one received. Thus, the complexity of the ML detector is reduced from M 2 to 2 M
possibilities, where M represents the number of possible received symbols in a given
modulation scheme (i.e., M-QAM). It is clear that this will significantly reduce
computational time requirements [2].
G. CHANNEL STATE INFORMATION AT THE TRANSMITTER
When the transmitter has access to channel characteristics by means of a feedback

loop or another method, it is said to know the channel state. With channel state
24
information at the transmitter (CSIT), capacity can be increased by allocating more
power to antennas with favorable channels [2].
Let the channel gain matrix H be known to the transmitter and receiver. The
singular value decomposition (SVD) of H is given by
H = U V H (2.58)
where U is an NR NR matrix, V is a NT NT matrix and both are unitary, that is
U H U = I N R and V H V = I NT . is a NR NT diagonal matrix of the singular values of the
channel, denoted by i , where i is determined by the rank of H denoted by r and
given by r min ( NT , N R ) . A H denotes the Hermitian operator, A = A ( )

H * T
[4].
Substituting Equation (2.58) into the received signal vector given by Equation
(2.23), yields
y = UVH s + . (2.59)
Defining linear transformations on the received vector as
y% = UH y (2.60)
the sent vector as

s% = V H s (2.61)
and noise vector as
% = UH
(2.62)
then applying these linear transformations to Equation (2.59), yields
y% = U H ( UV H s + )
= U H UV H Vs% + U H (2.63)
%
= s% +
% is distributed identically to . When the transformation of Equation (2.61) is

where
applied to the channel input s , it is often called transmit precoding. When the
transformation of Equation (2.60) is applied to the output y , it is referred to as receiver
25
shaping. From Equation (2.63), it is clear that the SVD operation allows the channel to
be represented by rank, r parallel channels, as is illustrated in Figure 10. [2].
Figure 10. Equivalent MIMO model (After [4]).
This thesis will focus on the ability of the transmitter to reduce errors by
allocating power to the individual antennas based on the dominant eigenvalues or
singular values of the channel i . This process is repeated every symbol period to
reduce the probability of errors. If the channel state is unknown, then the power is
distributed evenly among all antennas [2],[3].
Given the following total power constraint [4]

NT
P = Pi (2.64)
i =1
where Pi is the power allocated to antenna i and P is the total power. Let the channel
capacity be given as [4]
P
N
C = B log2 1 + i 2i
T
(2.65)
i =1
26
where C is the channel capacity, B represents the bandwidth, and 2 is the noise power
of each element of . Optimization of the channel capacity in terms of the antenna
power allocation is given by [4]
2
Pi = max ,0 (2.66)
i
where denotes a constant that is adjusted until Equation (2.64) is satisfied. If the
channel gain is low, then reliable communication is unlikely and the transmitter allocates
less power to that antenna. Iterations are conducted in this manner until the transmitter
deems the remaining channels adequate, resulting in optimal power distribution [4],[5].
H. SUMMARY
This chapter introduced the MIMO system model which forms the basis for
follow-on analysis. The effects of multipath fading on a wireless channel were
introduced and two statistical models defined to aid in analysis. The fundamentals of
orthogonal frequency division multiplexing (OFDM) were presented and a MIMO-
OFDM system model was introduced utilizing the discrete Fourier transform (DFT). An
optimal detection scheme in the form of Maximum likelihood (ML) at the receiver was
discussed. Then the Alamouti Space-time coding method for a 2 2 MIMO system was
introduced. Finally, channel state information is defined for application at the transmitter
where a power allocation process was developed. MIMO systems containing various
aspects presented in this chapter are developed and simulated in the following chapter.
27
28
III. ANALYSIS AND SIMULATION
This chapter provides bit error performance analysis and simulation of the two
transmit and two receive antenna baseband MIMO model presented in the last chapter.
The analysis begins with an uncoded narrowband system and progresses by
implementing the various schemes introduced in the previous chapter, culminating with a
system utilizing CSIT. To this, a second system is added and analyzed where this second
receiver enjoys none of the benefits intended for the primary receiver. Each system was
coded and simulated in MATLAB. The source code can be requested from the manager
of the Communications Research Laboratory (CRL), at the Naval Postgraduate School.
A. SIMULATION METHODOLOGY AND ASSUMPTIONS
The following assumptions are made when conducting the simulations, which are
described in detail in the previous chapter.
Given a fixed wireless communication system comprised of two transmit and two
receive antennas, where each antenna is spatially and angularly independent from the
other. The information transmitted from each antenna uses the same modulation, either
MPSK or MQAM, selected from a signal space as illustrated in Figure 11. The
probability of bit error ( Pb ) for each scheme is inversely related to the distance between
closest symbols and is a measure of the number of bits in error per bit sent over a given
time interval. Furthermore, the modulation utilizes Gray coding, in that adjacent symbols
differ by only one bit. Both signals are transmitted with equal power, split among the
two antennas. The total received power from all receive antennas per symbol time is
given by [19]
E s =1
2
(3.1)
n
n
Due to the statistical independence of sn , Equation (3.1) can be expressed for
each individual value as
29
1
E sn =
2
(3.2)
NT
In the remaining analysis and simulations, for each modulation scheme, Eb is
determined and held constant while N o is adjusted according to the simulated Eb N o .
Figure 11. Signal Space for (a) BPSK, (b) QPSK, (c) 16QAM, (d) 64QAM.
The channel is slow frequency non-selective fading and is modeled as having an

AWGN, Rayleigh, or Rician distribution. Additionally, the channel gains hmn are
30
identically distributed and statistically independent from each other and their sum is
normalized to unity, given by [2]
E h =1
2
(3.3)
n,m
mn
Due to the statistical independence of hmn , Equation (3.3) can be expressed for
each individual value as
1
E hmn =
2
. (3.4)
NT N R
Thus, the average total energy per bit received is the transmitted energy per bit
per transmit antenna. In other words, increasing the number of transmit or receive
antennas reduces the channel gains of the system. This may seem counter intuitive, since
we would expect that a system employing a greater number of antennas would result in
more received power and thus larger energy per bit. However, this assumption is
consistent with the literature, namely [2], [3], [5], and will be utilized in this thesis.
Synchronization is maintained perfectly and the channel is known to the receiver,

but unknown to the transmitter. The receiver noise is modeled as independent and
identically distributed complex white Gaussian noise processes, each having zero mean
and power spectral density of 2 N o . Detection is accomplished by utilizing a maximum-
likelihood detector, which provides a best case for probability of error. With the
exception of equal power transmission and channel state information at the transmitter,
which will be addressed specifically, these assumptions will be used in the proceeding
analysis and simulation [2].
B. UNCODED NARROWBAND
The first task is to evaluate the performance of an uncoded MIMO system, as

depicted in Figure 12. The entire system is simulated in MATLAB utilizing equivalent
baseband form in the discrete time domain. Two million symbols are generated and
modulated with either BPSK, QPSK, 16QAM or 64QAM. This is simulated for
increasing values of Eb N o over the range of 30 dB.
31
In these simulations, the transmitter employs two antennas with which to transmit
the modulated symbols with equal power. The wireless channel is modeled as slowly
fading where the signal period is less than the coherence time. The channel gains are
simulated according to the AWGN, Rayleigh and Rician models as presented in Chapter
II, under the assumptions given in the previous section. At the receiver, a maximum
likelihood decision statistic is simulated in which the symbol with the smallest Euclidian
distance is selected. This symbol estimate is then demodulated according the appropriate
scheme and compared to the transmitted signal. Decision errors are then tallied and
calculated over each iteration of Eb N o resulting in the following bit error plots.
Figure 12. MIMO System Block Diagram (After [2],[3]).
In addition, for comparison and to validate the accuracy of the program, an

AWGN channel is simulated and the ideal bit error probability, Pb is presented for each
modulation scheme.
In order to compare various modulation schemes of an uncoded 2x2 MIMO

system, the average energy per bit is calculated for each modulation scheme with regard
to the total energy received during one symbol period.
32
Let z denote the received signal in the absence of noise, given by
NT
zm = hmn sn (3.5)
n =1 .
Therefore, the average total energy received over one symbol period is
NR
ET = E zm
2
(3.6)
m =1

where E [] is the expectation operator and z m = z m z m * . Expanding Equation (3.6) we

2
have
NR NT NT
*
ET = E hmn sn hmk sk
m =1 n=1 k =1
(3.7)
N R NT NT
= E hmn sn ( hmk sk )
*
m =1 n =1 k =1

.
Due to statistical independence, Equation (3.7) can be simplified
N R NT NT
ET = E hmn sn ( hmk sk )
*
m=1 n =1 k =1

N R NT
(3.8)
= E hmn E sn
2 2
m=1 n =1

.
Recall the previously stated assumptions,
1
E hmn =
2
N R NT
. (3.9)
1
E sn =
2
NT
Therefore the total energy received during one symbol period for a 2 2 MIMO system
is
N R NT
ET = E hmn E sn
2 2
m =1 n =1

N R NT
1 1
= (3.10)
m =1 n =1 NT N R NT
1
=
NT
33
1. BPSK
The bit error probability for BPSK in a single-input single-output (SISO) AWGN
channel is given by
2 Eb
Pb = Q (3.11)
N0
where Eb denotes the energy per bit and Gaussian Q -function is defined as

1 y2 2
Q ( x) e dy (3.12)
x 2 .
Alternatively, the complementary error function is defined as

2
erfc ( x ) e
t 2
dt (3.13)
x
and is related to the Q -function by
1 x
Q ( x ) = erfc (3.14)
2 2 .
Thus, the BER of BPSK in terms of the complementary error function is
1 Eb
Pb = erfc
2 N (3.15)
0
which is more conducive to programming in MATLAB.
Applying Equation (3.2) to the case of BPSK modulation, where one bit of
information is sent every symbol, the energy per bit is given as
1
Eb = (3.16)
NT .
Figure 13. shows the performance of BPSK in AWGN, Rayleigh, and Rician channels
with K = 1 and K = 4 . The theoretical result obtained via Equation (3.15) is also plotted,
which closely follows the simulated AWGN channel.
34
Figure 13. Simulated BER performance of a 2x2 MIMO system
with BPSK modulation.
2. QPSK
The bit error probability for QPSK in a SISO AWGN channel is equal to that of
BPSK and is given by
2 Eb
Pb = Q (3.17)
N0
where Eb denotes the energy per bit. Applying Equation (3.2) for QPSK modulation in a
2 2 MIMO system, where two bits of information are sent every symbol, the energy per
bit is given as
35
1
Eb = (3.18)
2NT
Figure 14. shows the performance of QPSK in AWGN, Rayleigh, and Rician
channels with K = 1 and K = 4 . The theoretical result obtained via Equation (3.17) is
also plotted. As expected, the results for QPSK match those for BPSK. Therefore,
BPSK will not be considered in the remaining analysis.
modulation.
36
3. 16QAM
Consider a square 16QAM constellation depicted in Figure 11. (c). The symbol
error probability for MQAM in an AWGN channel when each symbol selected with
equal probability is given by [2]
1 3log 2 M Ebavg
Ps = 4 1 Q (3.19)
M M 1 N 0

where Ebave denotes the average energy per bit. For the case of 16QAM, Equation (3.19)
becomes
4 Eb
Ps = 3Q
avg
(3.20)
5 N 0

By utilizing gray encoding, Ps can be approximated in terms of Pb by [6]
Ps
Pb (3.21)
log2 M .
Applying Equation (3.14) and Equation (3.21), the BER of 16QAM in terms of the
complementary error function,
3 2 Eb
Pb = erfc avg (3.22)
8 5 N0

The signal space of a square 16QAM constellation is depicted in Figure 11. (c) where the
average symbol energy is given by
4(2) + 8(10) + 4(18)
Esavg =
16 (3.23)
= 10
when the minimum distance between adjacent vectors is 2 [2]. The relationship between
Eb and Es is
Es
Eb = (3.24)
k
where k denotes the bits per symbol, k = log 2 M . Thus, the average energy per bit in
16QAM is found to be
37
5
Ebavg = (3.25)
2.
For the case of a 2 2 MIMO system, we have
5
Eb = (3.26)
2 NT
Figure 15. shows the performance of 16QAM in AWGN, Rayleigh, and Rician
channels with K = 1 and K = 4 . The theoretical result obtained via Equation (3.22) is
also plotted for comparison.
modulation.
38
4. 64QAM
For the square 64QAM constellation depicted in Figure 11. (d), the symbol error
probability is given by [2]
7 2 Ebavg
Ps = Q (3.27)
2 7 N 0

Applying Equation (3.14) and Equation (3.27), the BER of 64QAM in terms of the
complementary error function is
7 1 Eb
Pb = erfc avg (3.28)
24 5 N0

and the average symbol energy given by

4(2) + 8(10) + 4(18) + 8(26) + 8(34) + 12(50) + 8(58) + 8(74) + 4(98)
Esavg =
64 (3.29)
= 42
when the minimum distance between adjacent vectors is 2 [2]. Using Equation (3.24),
the average energy per bit of 64QAM modulation is found to be
Ebavg = 7 (3.30)
.
For the case of a 2x2 MIMO system, we have
7
Eb = (3.31)
NT
Figure 16. shows the performance of 64QAM in AWGN, Rayleigh, and Rician channels.
For comparison, the theoretical result obtained via Equation (3.28) is also plotted.
39
modulation.
C. SPACE-TIME CODED NARROWBAND
Building upon the prior section, we evaluate the performance of space time
coding with a MIMO system. As before, QPSK, 16QAM and 64QAM modulations are
simulated and presented later in this section. The coding implemented follows the
Alamouti scheme, introduced in the previous chapter and the following analysis
completes that introduction by developing the maximum likelihood decision statistic for
equal and non-equal energy constellations.
Figure 17. outlines the system model and the output of the combiner, presented in
the previous chapter and reproduced here for convenience.
40
(
s1 = h11 + h12 + h21 + h22
2 2 2 2
)s 1
+ h11*1 ( t ) + h12 1 ( t + T ) + h21*2 ( t ) + h22 2 ( t + T )

* *
(3.32)
(
s2 = h11 + h12 + h21 + h22
2 2 2 2
)s 2
h111 ( t + T ) + h12*1 ( t ) h212 ( t + T ) + h22*2 ( t + T )

* *
Figure 17. Alamouti space-time coding scheme (After [14]).
The Alamouti scheme uses the following maximum likelihood decision rule to
select si iff [14]
d 2 ( y11 , h11si ) + d 2 ( y12 , h12 si ) d 2 ( y11 , h11sk ) + d 2 ( y12 , h12 sk ) (3.33)
for all i k . The squared Euclidean distance between signals x and y is given by [14]
d 2 ( x, y ) = ( x y ) ( x* y * ) . (3.34)
Therefore, applying Equations (3.34) and (3.32) to the decision rule gives
( h + h + h + h 1) s + d ( s , s )
11
2
12
2
21
2
22
2
i
2 2
1 i
( h + h + h + h 1) s + d ( s + s )
2 2 2 2 2 2
11 12 21 22 k 1 k
(3.35)
( h + h + h + h 1) s + d ( s + s )
11
2
12
2
21
2
22
2
i
2 2
2 i
( h + h + h + h 1) s + d ( s + s )
2 2 2 2 2 2
11 12 21 22 k 2 k
.
41
Expanding Equation (3.35), the maximum likelihood statistic becomes

s1
(2 2 2 2 2
) (
min h11 + h12 + h21 + h22 1 si si * s1 + si s1
*
)

(3.36)
s2

(2 2 2 2 2
) (
min h11 + h12 + h21 + h22 1 sk si* s2 + si s2
*
)
2 2
where the s1 and s2 terms of the minimum distance calculation are constant and are
neglected since they have not impact on the decision. For signals that have equal energy,
as in the case of BPSK and QPSK, the magnitudes are constant and the statistic is
s1
(
min si * s1 + si s1
*
)

(3.37)
s2
(
min si * s2 + si s2
*
)

Figure 18. shows the performance of QPSK, 16QAM and 64QAM respectively with
space time coding in Rayleigh and Rician channels.
42
Figure 18. Simulated BER performance of a 2x2 MIMO system with Alamouti space-
time coding and QPSK modulation.
43
time coding and 16QAM modulation.
44
time coding and 64QAM modulation.
D. SPACE-TIME CODED ORTHOGONAL FREQUENCY DIVISION

MULTIPLEXING
We now apply space-time coding to MIMO-OFDM. As a transmission strategy,

MIMO and OFDM are integral to 4th generation communications standards. Thus, the bit
error performance of such systems is of great interest. This discussion will focus on the
implementation of STC, relative to the OFDM system presented in the previous chapter
[1].
Consider the basic space-time coded MIMO-OFDM system as shown in Figure

21. The coding scheme is the same as detailed in the previous section; however, the
scheme is implemented in the frequency domain on consecutive tones or subcarriers.
45
Analogous to the time domain implementation, in the frequency domain it is assumed
that the channel remains constant over consecutive tones [10].
Figure 21. Basic implementation of STC with MIMO-OFDM (After [5]).
The simulation was performed with an FFT size of 512 points and a cyclic prefix
of 1 4 . That is, the last 128 points from the output of the IDFT operation were
reproduced and appended to the beginning of the symbol in the form of a cyclic prefix.
The OFDM symbols are then sent to the transmitter and transmitted through either a
Rayleigh or Rician fading channel. At the receiver, the cyclic prefix is removed and the
DFT operation is performed. The symbols are then combined and estimated in
accordance with the Alamouti scheme where the decision criterion is analogous to the
narrowband case presented in the previous section.
The simulation evaluated over two million symbols with Eb N o increasing from 0
to 30dB in increments of two dB. The results of QPSK, 16QAM and 64QAM can be
seen in Figure 22. Figure 23. and Figure 24. respectively.
46
Figure 22. Simulated BER performance of a 2x2 MIMO-OFDM system with Alamouti
space-time coding and QPSK modulation.
47
Figure 23. Simulated BER performance of a 2x2 MIMO-OFDM system with Alamouti
space-time coding and 16 QAM modulation
48
Figure 24. Simulated BER performance of a MIMO-OFDM system with Alamouti
space-time coding and 64QAM modulation.
E. NARROWBAND WITH CHANNEL STATE INFORMATION AT THE

TRANSMITTER
This section investigates the performance of a 2 2 MIMO system when the

channel state is available to the transmitter, often referred to as a closed-loop system. We
assume that the antennas are fixed and that the feedback channel provides perfect
information on the transmit channel. Figure 25. depicts the basic system. Knowledge of
the channel gains allows the transmitter to use various techniques in order to increase
capacity or reliability. The following analysis focuses on increasing reliability through
diversity techniques. One such method, which requires channel state information at the
transmitter, is dominate eigenmode transmission.
49
Dominate eigenmode transmission is also known as linear diversity precoding,
and is analogous to maximal ratio Combining (MRC) in the case of a MISO system.
Additionally, dominate eigenmode transmission lends a comparison with the Alamouti
scheme of space-time coding, which does not require CSIT. Both techniques provide full
diversity order of four, the maximum in the case of a 2 2 system. However, in the case
of dominate eigenmode transmission, the channel state information allows the transmitter
to optimally precode the signal, which provides an array gain at the receiver not found in
the Alamouti scheme.
Figure 25. Closed loop model.
Given a 2 2 MIMO system, let the same symbol be transmitted on both

antennas during each time period, weighted by a NT 1 vector w . Thus, the receiver
yields
y = Hws + (3.38)
where H represents the N R NT channel matrix, is the N R 1 noise vector, s
represents the transmitted symbol and y is the N R 1 received signal vector. At the
receiver, the signals are combined and weighted by a N R 1 vector g , given by [16]
z = gH y (3.39)
resulting in
50
z = g H ( Hws + ) (3.40)
.
In order to find the optimal weighting of the transmitted signals, the transmitter
utilizes the singular values of the channel. Recall from Chapter II that SVD operation is
given as
H = UV H (3.41)
where U and V are NT NT and N R N R unitary matrices, and is a N R NT diagonal

matrix of the singular values of the channel. It is shown in [4] that the receiver SNR is
maximized when both the postcode vector g , and the precode vector w correspond to the
right and left singular vectors of the maximum singular value of H , max . Therefore, to
maximize the receiver SNR, Equation (3.40) can be represented as
y = maxs + (3.42)
where max is a scalar that corresponds to the largest value in which is the maximum
singular value of H . Given that the max singular value of H corresponds to the left and
right singular vectors umax and vmax respectively, we can express Equation (3.42) in terms
of the SVD operation by [17]
y = umax H Hvmaxs + (3.43)
The corresponding system model is depicted in Figure 26. Thus, with channel
state information, the transmitter weighs the transmitted symbols according to the right
singular vector of the channel H . At the receiver, the symbols are summed and weighed
by the left singular vector where the signal to noise ratio is given by [3]
Es
= max
2
(3.44)
2
where Es is the symbol energy, 2 is the noise power of each element of and is
bounded by [4],[9]
2 2
Es H F Es H F
(3.45)
2 NT 2
51
and the Frobenius norm denoted as F
is given by [17]
NT NR
HF= h
i =1 j =1
2
ij (3.46)
.
Thus, dominant eigenmode transmission provides two types of gain; diversity and array.
Since the same symbol is sent from each antenna, the diversity gain is the result of having
multiple independent copies of the symbol for decision. At the receiver the optimally
weighted and independent signals are coherently combined, providing the array gain.
The array gain realized depends on the fading characteristics of the channel and is given
by E max
2
[3][5].
Figure 26. Dominant eigenmode transmission in a 2x2 MIMO system.
Figure 27. provides the performance of QPSK utilizing a dominant eigenmode

transmission in Rayleigh and Rician channels.
52
Figure 27. Simulated BER performance of a 2x2 MIMO system with QPSK modulation
using dominant eigenmode transmission when the transmitter has channel state
information.
F. UNKNOWN RECEIVER
Throughout this chapter, each section presented analysis and performance results
of various techniques used in MIMO communications. The implementation of such
schemes is often dependant on the environment in which the communication system is
operating. Additionally, the many benefits of MIMO communications rely on the
successful pairing of the transmitter and receiver, where the result of this pairing is often
53
the optimization of the transmission link. This is a preferred, indeed desired, outcome for
system designers. However, this thesis seeks to investigate how a receiver performs in
the absence of this optimization.
Consider a MIMO system containing transmitter A and receiver B as shown in

Figure 28. The system is located on fixed infrastructure and has been designed to
optimize the link between A and B, a NLOS situation. The transmitter has channel state
information, through a feedback loop with receiver B. With CSIT, the transmitter utilizes
dominant eigenmode transmission in order to improve the signal to noise ratio at receiver
B, providing a robust communications link.
Figure 28. A fixed MIMO system optimized for channel AB.
We now introduce a second receiver, C, that is disadvantaged in the sense that it

does not enjoy the same optimization with transmitter A that receiver B has. Receiver C
can be mobile or fixed. Channel AC can be either Rayleigh faded channel (NLOS) or a
Rician fading channel (LOS). If the receiver is mobile, the velocity is assumed to be such
that the assumptions given at the beginning of this chapter still hold. Additionally, the
receiver is not limited by the number of antennas available. That is, the receiver may
employ additional antennas, greater than two, in order to improve reception. Figure 29.
provides the system layout, where the dashed line represents the link that is not
optimized.
54
Figure 29. Disadvantaged receiver in a fixed MIMO system optimized for channel AB.
Recall from the previous section that a transmitter with channel state information
may precode the transmitted symbols in order to increase the signal to noise ratio at the
receiver. This result was shown in Equation (3.42) and is reproduced here for
convenience.
y = maxs + . (3.47)
Expressing Equation (3.47) in terms of the SVD operation we get an equivalent system
represented by [17]
y = ( umax
H
Hvmax s + ) (3.48)
AB
where ( ) AB denotes the characteristics of the link from transmitter A to receiver B. We
assume that receiver C has knowledge of the precoding vector or the right singular value
of the channel ( H ) AB . With this knowledge, the receiver is able to compensate for the
optimization induced by the transmitter for channel AB. This case is simulated with
55
QPSK modulation over four million symbols with Eb N o increasing by increments of
two dB. Three situations are presented: The first involves a Rayleigh channel, the second
involves a Rician channel with rice factor of one ( K = 1) , and the final situation simulates
a Rician channel with ( K = 4) . Additionally, all three scenarios are evaluated with
receiver C having 2, 4, and 8 antennas. The results are plotted in Figure 30. Figure 31.
and Figure 32. ,, respectively.
Figure 30. Simulated BER performance of the disadvantaged receiver in a Rayleigh

channel, for 2x2, 2x4, and 2x8 MIMO systems.
56
Figure 31. Simulated BER performance of the disadvantaged receiver in a Rician channel
(K=1), for 2x2, 2x4 and 2x8 MIMO systems.
57
Figure 32. Simulated BER performance of the disadvantaged receiver in a Rician channel
(K=4), for 2x2, 2x4, and 2x8 MIMO systems.
G. SUMMARY
Chapter III provided bit error rate analysis and simulation results of the concepts
introduced in Chapter II. Simulation assumptions were presented and used throughout
the chapter in an effort to provide for results that will be compared in the next chapter.
The basic uncoded narrowband MIMO system was analyzed and simulated for numerous
channel conditions as the basis for follow on analysis. Next, space time coding following
the Alamouti scheme was refined and simulated. Then orthogonal frequency division
multiplexing with space time coding was simulated. Channel state information was then
incorporated into the narrowband system and simulated. Finally, the chapter concludes
58
with a situation in which a disadvantaged receiver was introduced to an existing system
previously optimized for communication. Various results from these simulations are
presented and compared in the next chapter.
59
60
IV. COMPARISON AND RESULTS
This chapter compares the performance of the individual simulations presented

the in last chapter. The comparison begins by investigating the performance of an
uncoded system in Rayleigh and Rician channels. A space-time coded narrowband
system is compared to an uncoded system. Then, narrowband performance is compared
to that of OFDM. Next, the narrowband system is analyzed with and without channel
state information. Finally, the performance of the primary receiver is compared to that of
the disadvantaged receiver.
A. RAYLEIGH AND RICIAN CHANNEL MODEL COMPARISON
The discussion begins with an evaluation of the performance of an uncoded 2 2

MIMO system in Rayleigh and Rician fading channels. Figure 14. in Chapter III, section
B, shows the performance of an uncoded MIMO system over Rayleigh and Rician fading
channels. For this comparison QPSK was used, however, the results are similar for the
other modulation schemes.
Recall from Chapter II, that the K - factor is a ratio of the power in the specular
component (LOS) to that of the diffuse component (NLOS). Clearly, as the value of K
increases, the specular component dominates and the channel approaches that of a non-
fading channel or AWGN channel. This is confirmed by the increased performance of
the system in a Rician channel to that of the Rayleigh channel. For a BER rate of 104 , a
LOS component of equal power to that of the NLOS, as seen by the receiver, results in an
improvement of 4.4 dB. Table 3. highlights the Eb N o required to achieve a particular
BER.
61
Table 3. Eb N o required to achieve various BER for different channel models, 2x2 MIMO
system with QPSK modulation.
Obviously, when the receiver has LOS to the transmitter, higher order modulation
schemes may be employed and still achieve the same (or better) BER. Figure 33.
compares QPSK, 16QAM and 64QAM with increasing power in the specular component.
It is evident that beyond a BER of 104 the higher modulation schemes outperform QPSK
due to the specular component. This implies that the specular component improves BER
more than the higher order modulation degrades the BER.
62
Figure 33. Simulated BER performance of various modulation schemes with different
channel characteristics, 2x2 MIMO system.
B. UNCODED AND SPACE-TIME CODED COMPARISON
We now investigate the performance of an uncoded system to that of a space-time

coded system. The Alamouti space-time code is designed to extract diversity when a
minimum of two transmitting antennas are present, therefore we would expect to see a
diversity gain when compared to the uncoded case. Figure 34. plots the BER
performance of coded and uncoded QPSK in a Rayleigh channel.
The performance improvement seen the in plot is due to the diversity gain
provided by the Alamouti scheme. Additionally, the orthogonality of the coding
sequence transforms the decision at the receiver from a vector operation into a scalar
63
operation, reducing the receiver complexity. Comparing the required Eb N o to achieve
a BER of 104 , we see that coding improves performance by 9.7 dB. Table 4. provides
the required Eb N o to achieve a particular BER.
Figure 34. Simulated BER performance of uncoded and coded 2x2 MIMO system with
QPSK modulation in a Rayleigh channel.
64
Table 4. Eb N o required to achieve a certain BER in a Rayleigh channel, 2x2 MIMO
system using QPSK modulation, with and without Alamouti space-time coding.
Next, we compare a coded system to an uncoded system characterized by a strong

LOS component for all three modulation schemes, shown in Figure 35. It is interesting to
see from the plot that an uncoded system over a Rician channel has a marginal
improvement in performance past 12 dB (QPSK) to that of a space-time coded system.
However, it is clear that a coded system outperforms an uncoded system even in an
environment where the antenna has a dominate line of sight.
This demonstrates the advantage of antenna placement with respect to system

performance, or if this is not possible the ability to apply the Alamouti scheme in order to
improve performance. MIMO systems allow the system designer implement multiple
techniques in order to establish robust communications. This simulation provides a
comparison of the performance of these different techniques. As the simulation shows,
Rayleigh fading degrades BER as compared to Rician fading with a strong LOS
component (K = 4) roughly the same as uncoded compared to Alamouti coded.
65
Figure 35. Simulated BER performance of a 2x2 MIMO system, various modulation
schemes, uncoded and using Alamouti space-time coding, with different channel
characteristics.
C. CODED NARROWBAND AND WIDEBAND COMPARISON
For this thesis, it was assumed that the channel provided slow frequency non-
selective fading conditions. Thus, the performance of a coded wideband system will
coincide with that of a narrowband system. Figure 36. confirms this result with a plot of
BER for 64QAM with Alamouti space-time coded wideband and narrowband 2 2
MIMO system.
66
Figure 36. Simulated BER performance of a 2x2 MIMO system using Alamouti space-
time coding, wideband and narrowband in various channels with 64QAM
modulation.
D. TRANSMITTER WITH AND WITHOUT CSIT COMPARISON
In this section, the performance of a system when the transmitter has CSIT will be
compared to a system without. With a closed loop system, the information available to
the transmitter can be applied to improve performance in a number of areas, depending
on the design of the system. In this thesis we focus on the ability of the transmitter to
increase the received SNR through the use of dominant eigenmode transmission, where
the result of this application is an array gain in addition to the diversity gain. Therefore,
we expect that the performance increase over the uncoded case will be considerable.
67
Figure 37. plots the error rate performance of a 2 2 MIMO system with and
without CSIT where both Rayleigh and Rician fading are considered. From the
illustration it is clear that there is a significant improvement when the transmitter has
knowledge of the channel. For a BER rate of 104 , the system with CSIT results in an
improvement of roughly 12 dB. Table 5. highlights the Eb N o required to achieve a
particular BER.
Figure 37. Simulated BER performance of a transmitter with and without CSIT for a 2x2
MIMO system in various channels using QPSK modulation.
68
Table 5. Eb N o required to achieve a BER in Rayleigh channel and Rician channels when
the system has CSIT, 2x2 MIMO system.
The performance trends displayed in Figure 37. are as expected; the array gain
increases the BER of the entire system for all values of Eb N o , while the diversity gain
improves BER performance for large values of Eb N o .
Figure 38. compares a system employing the Alamouti space-time code and no
CSIT to a system with CSIT and no STC in Rayleigh fading. In this figure, it is clear that
both systems offer the same diversity. However, the system employing CSIT shows a
consistent improvement of 2 dB. This is due to an array gain resulting from the coherent
combining of the precoded signals at the receiver. Thus, for the same diversity order we
see that knowledge of the channel at the transmitter improves performance by 2 dB. Of
course, this improvement in performance is at the cost of increased overhead due to the
requirements of the feedback channel.
69
Figure 38. Simulated BER performance of a 2x2 MIMO system with CSIT to that of a
system with Alamouti space-time coding both in a Rayleigh channel using QPSK
modulation.
E. DISADVANTAGED RECEIVER COMPARISON
The results in this thesis illustrate that MIMO communications allow the system
designer the ability to produce a system that specifically addresses the environment in
which it operates. That is one of the benefits of MIMO communications. When a MIMO
system is optimized using the techniques presented in this thesis, such as employing
channel state information to overcome channel characteristics specific to that radio link,
introducing a second receiver to the system could pose a serious problem, because that
optimization is specific for the first receivers channel gains.
70
In this final section, we investigate the performance of a receiver that is
disadvantaged with respect to the intended receiver. That is, the disadvantaged receiver
does not enjoy the same optimization of the transmission, as does the intended receiver.
The goal of this comparison is to integrate the topics discussed previously in order to
produce performance at the disadvantaged receiver that is equal if not better than that of
the intended receiver.
Recall, the system in question, presented in the last chapter, shown in Figure 39.
Figure 39. Disadvantaged receiver in a fixed MIMO system optimized for channel AB.
Transmitter A has channel state information of channel AB and uses dominate

eigenmode transmission in order to improve the SNR at receiver B, the intended receiver.
Consider receiver C, which is introduced to the system and does not enjoy the same
optimization as the intended receiver, B.
71
1. Disadvantaged Receiver in a Rayleigh Channel
First, we compare the performance of both receivers in a Rayleigh channel, shown

in Figure 40. The modulation for all systems in this section is QPSK.

system, Rayleigh channel.
From the simulation results, it is clear that even when the disadvantaged receiver
employs eight receive antennas, the performance is still worse than for the intended
receiver. At a BER of 106 with eight receive antennas, the difference is 1.2 dB.
However, depending on the application, this performance may prove to be adequate.
Table 6. provides performance values for specific values of Eb N o .
72
Table 6. Eb N o comparison for select values of BER, disadvantaged receiver with various
antenna configurations over a Rayleigh fading channel.
2. Disadvantaged Receiver in a Rician Channel (K = 1)
Next, we look at a situation in which receiver C has a LOS path to the transmitter.
Where there is equal power in the specular and diffuse paths, shown in Figure 41.
73
system, Rician channel (K = 1).
Clearly, performance is improved for receiver C compared to that of the Rayleigh

case of Figure 40. , as expected. From the graph, we see that with two receive antennas
performance still does not approach that of the intended receiver. However, when the
number of antennas is doubled, the performance of receiver C is comparable at low
values of Eb N o and provides superior performance at high values of Eb N o . Table 7.
highlights this improvement.
74
antenna configurations over a Rician channel (K = 1).
3. Disadvantaged Receiver in a Rician Channel (K = 4)
Finally, we look at a situation in which receiver C has a more significant LOS

path to the transmitter, illustrated in Figure 42.
75
Figure 42. BER performance of a disadvantaged receiver in a fixed MIMO system,
Rician channel (K = 4).
We observe that with two antennas, receiver C enjoys similar performance to the
intended receiver. Obviously, four antennas provide an even greater performance gain.
A transmitter with eight antennas was not evaluated since the result of four antennas is
sufficient, but it is expected that BER performance increases dramatically as the number
of antennas increases. Table 8. gives specific values for Eb N o .
76
antenna configurations over a Rician channel (K = 1).
When considering the performance of the disadvantaged receiver as the number

receive antennas increased, recall from the first section in Chapter III, that the average
total energy per bit received is equal to the transmitted energy per bit per transmit
antenna by convention. In other words, both systems have the same total average Eb . As
a result, increasing the number of receive antennas reduces the channel gains of the
system. Thus, our assumption diminishes the performance gained by increasing the
number of receive antennas at the disadvantaged receiver. Of course, if the systems had
different Eb and N o , those values would be used to make a more accurate comparison.
The gross path loss, or geometry with respect to the transmitter, may favor one receiver
over the other. In this case, we would see an even greater improvement in performance
since the greater number of receive antennas results in more received power and thus a
higher Eb .
However, with the assumptions of this thesis, the results of the simulations show
that when the disadvantaged receiver has a weak Eb N o , increasing the number of
receive antennas is one method to significantly decrease the bit error rate.
F. SUMMARY
In this chapter, several different scenarios are compared in terms of BER

performance, based on the results of the previous chapter. The first comparison
investigated the performance of an uncoded MIMO system over Rayleigh and Rician
77
channels. The resulting improvement in BER performance over a Rician channel was
presented. Next, a space-time coded system was compared to an uncoded system, where
the diversity gain provided by the Alamouti scheme was made apparent. Then,
narrowband performance was compared to that of OFDM. Subsequently, a MIMO
system was analyzed with and without channel state information. Finally, the
performance of the primary receiver in a MIMO system is quantified and compared to the
performance of a disadvantaged receiver. The disadvantaged receiver is then presented
with various configurations in order to improve it BER performance. The next chapter
concludes this thesis and provides a summary of results.
78
V. CONCLUSION
The goal of this research was to investigate disadvantaged receiver performance

compared to the intended receiver in a MIMO system with the transmission optimized for
the intended receiver. This was accomplished through a progressive development
process from the simple uncoded MIMO system to one that employs channel state
information at the transmitter. Once the final system was developed, a second,
disadvantaged receiver was added and configured such that the BER performance
matched or exceeded that of the intended receiver. Each system was simulated in
MATLAB and the BER performance was determined and compared.
A. SUMMARY
A comprehensive study of MIMO systems and technologies was presented.

Specifically, the BER performance of the transmitter and receiver in several cases was
developed, to include space-time coding, OFDM, and channel state information at the
transmitter. A second receiver was then introduced into the final system and the
performance utilizing the discussed technologies was investigated.
The systems were designed and simulated in MATLAB for each case, over both
Rayleigh and Rician fading channels. The results obtained were presented in the form of
BER performance curves and compared against each other.
B. SIGNIFICANT RESULTS
The following results are provided by this thesis.
The BER performance of a MIMO system increases significantly when a LOS

component is present. This is completely consistent with intuition developed from the
study of SISO systems. When equal power is present in the specular and diffuse
components (K = 1), an improvement of 6.8 dB was observed for a bit error rate of 106 .
In addition, a system employing a higher order modulation scheme such as 64QAM over
a Rician channel is able to achieve equal, if not better, performance than a system with
79
QPSK modulation over a Rayleigh channel. This comparison was made in order to
quantify the advantage of a receiver in a Rician channel as compared to one in a Rayleigh
channel. That is important in this work as that might be the case for the disadvantaged
and intended receivers described in Chapter IV, section E.
In Chapter IV, section B, MIMO systems utilizing the Alamouti space-time code
were observed to improve the BER performance by 14 dB at a bit error rate of 106 when
compared to uncoded systems. Moreover, in high Eb N o regions, a system
communicating over a Rician channel provided increased performance to that of a coded
system over a Rayleigh channel. This comparison further demonstrates the advantage of
antenna placement with respect to system performance. This relates to the situation of a
disadvantaged receiver since the fixed system may be designed in such a way as to
exploit this result. In other words, with MIMO systems there are multiple techniques that
can be implemented in order to establish robust communications. This work primarily
focused on the situation in which the transmitter utilized channel state information in
order to establish a reliable system. However, understanding of such other techniques
benefits the designer of the disadvantaged system as well.
When the transmitter utilizes channel state information to increase receiver

diversity we observe a significant performance increase over a system that does not have
CSIT. For a BER of 106 , an improvement of 13.7 dB was revealed. Additionally, it was
observed that the system with CSIT provides an array gain of 2 dB when compared to the
system without channel state information.
When a disadvantaged receiver is added to the system it is seen that the BER
performance cannot match that of the intended receiver when both receivers experience
Rayleigh fading, even when the number of reception antennas is increased to eight.
However, when the disadvantaged receiver is located such that it has a line of sight
component (Rician fading), we observe that performance increases significantly. In the
case of equal LOS and NLOS components, a disadvantaged receiver with four reception
antennas outperforms the intended receiver. When the disadvantaged receiver has a more
80
significant LOS component, similar performance to that of the intended receiver can be
found with only two receive antennas at the disadvantaged receiver.
C. RECOMMENDATIONS FOR FUTURE WORK
Four areas have been identified for future work. First, this work assumes that the
system is located in a fixed environment. However, a more realistic environment may be
investigated by relaxing the channel assumptions to allow for frequency selective fading.
Second, the communication system can be further refined with a space-time

coding scheme and OFDM. Due to the channel conditions assumed, this thesis
developed a system that was single carrier and uncoded. If the first recommendation is
applied, then the system can be integrated with those techniques and the disadvantaged
receiver performance can then be determined under the new channel conditions. Of
course, utilizing OFDM with the first recommendation will provide similar results to the
flat fading conditions considered in this work, since each subcarrier experiences flat
fading conditions even when transmitted through a frequency selective channel.
However, a simulation can be conducted to confirm this result.
Third, in this thesis the focus was on bit error performance. Future work can
analyze the system from an information theory standpoint. That is, focus on the capacity
of the various systems and compare the capacity gains (or losses). For this, the channel
state information at the transmitter can be utilized to increase capacity rather than
improving the receiver BER as was used in this study.
Fourth, the simulations performed in this work discovered that as the number of
receive antennas employed by the disadvantaged receiver increased, so did the BER
performance. However, future work can identify the point of diminishing returns with
respect to the number of receivers. At what point do the number of antennas and the
corresponding complexity outweigh the performance improvement gained? Simulations
involving an increased number of antennas can easily determine such a threshold.
81
82
LIST OF REFERENCES
[1] J. G. Andrews, A. Ghosh, and R. Muhamed, Fundamentals of WiMAX

Understanding Broadband Wireless Networking, Upper Saddle River, New
Jersey: Prentice Hall, 2007.
[2] J. G. Proakis, and M. Salehi, Digital Communications, 5th ed. New York:
McGraw-Hill, 2008.
[3] A. Goldsmith, Wireless Communications, Cambridge, United Kingdom:

Cambridge University Press, 2005.
[4] B. Vucetic, and J. Yuan, Space-Time Coding, West Sussex, England: John Wiley
& Sons, 2003.
[5] A. J. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless

Communications, Cambridge, United Kingdom: Cambridge University Press,
2003.
[6] T. Ha, Theory and Design of Digital Communication Systems, Cambridge, United
Kingcom: Cambridge University Press, 2010.
[7] T. S. Rappaport, Wireless Communications: Fundamentals and Applications, 2d

ed., Upper Saddle River, New Jersey: Prentice Hall, 2002.
[8] D. Tse, and P. Viswanath, Fundamentals of Wireless Communication,Cambridge,

United Kingdom: Cambridge University Press, 2005.
[9] C. W. Therrien, and M. Tummala, Probability for Electrical and Computer

Engineers, Boca Raton, Florida: CRC Press LLC, 2004.
[10] R. Prasad, OFDM for Wireless Communications Systems, Boston, Massachusetts:

Artech House, 2004.
[11] S. Sesia, I. Toufik, and M. Baker, LTE - The UMTS Long Term Evolution, West
Sussex, England: John Wiley & Sons, 2009.
[12] B. Sklar, Digital Communications Fundamentals and Applications, 2d ed., Upper

Saddle River, New Jersey: Prentice Hall, 2001.
[13] X. Zhu, and R. D. Murch, Performance Analysis of Maximum Likelihood

Detection in a MIMO Antenna System," IEEE Transactions on Communications,
vol. 50, no. 2, pp. 187191, February 2002.
83
[14] S. M. Alamouti, "A Simple Transmit Diversity Technique for Wireless
Communications," IEEE Journal on Select Areas in Communications, vol. 16, no.
8, pp. 14511458, October 1998.
[15] P. J. Smith, L. M. Garth, and M. Shafi, "Performance Analysis of Multiple-Input

Multiple-Output Singular Value Decomposition Transceivers During Fading and
other Cell Interference," Microwaves, Antennas & Propagation, IET , vol.1, no.6,
pp.11111119, Dec. 2007.
[16] C. Oestges, and B. Clerckx, MIMO Wireless Communications: From Real-World

Propagation to Space-Time Code Design, Oxford, United Kingdom: Academic
Press, 2007.
[17] G. H. Golub, and C. F. Van Loan, Matrix Computations, 3d ed. Baltimore,

Maryland: The Johns Hopkins University Press, 1996.
[18] E. Teletar, Capacity of Multi-antenna Gaussian Channels," AT&T Bell Labs

Internal Technical Memo, June, 1995.
[19] F. Gray, Pulse code communication, March 17, 1953. U.S. patent no.
2,632,058.
[20] A. J. Paulraj, D. A. Gore, R. U. Nabar, and H. Bolcskei, An Overview of MIMO

CommunicationsA Key to Gigabit Wireless, Proceedings of The IEEE, vol.
92, no. 2, pp. 198217.
84
INITIAL DISTRIBUTION LIST
5. Defense Technical Information Center

Ft. Belvoir, Virginia
6. Dudley Knox Library

Naval Postgraduate School
Monterey, California
7. Chairman, Code EC
Department of Electrical and Computer Engineering
8. Associate Professor Frank Kragh, Code EC/Kh

9. Professor Tri Ha, Code EC/Ha

10. Professor Clark Robertson, Code EC/Rc

11. Professor Roberto Cristi, Code EC/Cx

12. Donna Miller

13. Marine Corps Representative

14. Director, Training and Education, MCCDC, Code C46

Quantico, Virginia
85
15. Director, Marine Corps Research Center, MCCDC, Code C40RC
Quantico, Virginia
16. Marine Corps Tactical Systems Support Activity (Attn: Operations Officer)
Camp Pendleton, California
86

Naval Postgraduate School: Monterey, California

Uploaded by

Copyright:

Available Formats

Naval Postgraduate School: Monterey, California

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Naval Postgraduate School: Monterey, California

Uploaded by

Copyright:

Available Formats

NAVAL

ANALYSIS AND SIMULATION OF DISADVANTAGED

Thesis Advisor: Frank Kragh

Approved for public release; distribution is unlimited

ANALYSIS AND SIMULATION OF DISADVANTAGED RECEIVERS FOR

Submitted in partial fulfillment of the

MASTER OF SCIENCE IN ELECTRICAL ENGINEERING

NAVAL POSTGRADUATE SCHOOL

Author: Tracy A. Martin

Approved by: Frank Kragh

This thesis investigates the simulation performance of Multiple-Input Multiple-Output

Table 1. Characteristics of channel fading (After [6],[7]). ............................................13

A/D Analog to Digital

AWGN Additive White Gaussian Noise

BER Bit Error Rate

BPSK Binary Phase-Shift Keying

CDF Cumulative Distribution Function

CSI Channel State Information

CSIR Channel State Information at the Receiver

CSIT Channel State Information at the Transmitter

D/A Digital to Analog

DFT Discrete Fourier Transform

FFT Fast Fourier Transform

IDFT Inverse Discrete Fourier Transform

IFFT Inverse Fast Fourier Transform

IID Independent Identically Distributed

ISI Intersymbol Interference

LOS Line of Sight

NLOS Non-Line of Sight

MIMO Multiple-Input Multiple-Output

MISO Multiple-Input Single-Output

OFDM Orthogonal Frequency Division Multiplexing

PSD Power Spectral Density

QAM Quadrature Amplitude Modulation

QPSK Quadrature Phase-Shift Keying

SISO Single-Input Single-Output

SNR Signal to Noise Ratio

STC Space-Time Code

SVD Singular Value Decomposition

WSS Wide-Sense Stationary

Multiple-Input Multiple-Output (MIMO) communications systems are fundamentally

Multiple-input Multiple-output (MIMO) communications systems are

B. OBJECTIVES AND METHODOLOGY

Development begins with the simulation of a simple uncoded narrowband MIMO

MIMO systems are becoming increasingly commonplace, including IEEE

This chapter provides background on the techniques used in designing and

Given a bandpass signal g ( t ) , whose Fourier transform is G ( f ) , then the

analytic signal is defined as [2]

Taking the Fourier transform of Equation (2.1) yields [1]

u ( f ) is the unit step function, to get

Taking the Fourier transform and substituting Equation (2.3) results in

In the time domain

Additionally, the lowpass equivalent signal can be represented by the in-phase

Substituting Equation (2.8) into Equation (2.6) yields [2]

gi ( t ) = g ( t ) cos ( 2 fct ) + g ( t ) sin ( 2 fct ) and

Solving for g ( t ) and g ( t ) yields [2]

or, in polar form, [2]

where rg ( t ) and g ( t ) represent the envelope and phase of g ( t ) .

+ h111 ( t ) + h12 1 ( t + T ) + h212 ( t ) + h22 2 ( t + T )

h111 ( t + T ) + h121 ( t ) h212 ( t + T ) + h222 ( t + T )