SC Ofdma

viii
An Abstract
Single Carrier Orthogonal Multiple Access Technique
for Broadband Wireless Communications
by
Hyung G. Myung
Advisor: David J. Goodman
Submitted in Partial Fulfillment of the Requirements

For the Degree of Doctor of Philosophy
January 2007
Broadband wireless mobile communications suffer from multipath frequency-selective fading.

Orthogonal frequency division multiplexing (OFDM) and orthogonal frequency division
multiple access (OFDMA), which are multicarrier communication techniques, have become
widely accepted primarily because of its robustness against frequency selective fading channels.
Despite the many advantages, OFDM and OFDMA suffer a number of drawbacks; high peakto-average power ratio (PAPR), a need for an adaptive or coded scheme to overcome spectral
nulls in the channel, and high sensitivity to frequency offset.
ix
Single carrier frequency division multiple access (SC-FDMA) which utilizes single carrier
modulation at the transmitter and frequency domain equalization at the receiver is a technique
that has similar performance and essentially the same overall structure as those of an OFDMA
system. One prominent advantage over OFDMA is that the SC-FDMA signal has lower PAPR.
SC-FDMA has drawn great attention as an attractive alternative to OFDMA, especially in the
uplink communications where lower PAPR greatly benefits the mobile terminal in terms of
transmit power efficiency and manufacturing cost. SC-FDMA is currently a working assumption
for the uplink multiple access scheme in 3rd Generation Partnership Project Long Term
Evolution (3GPP LTE).
In this thesis, we first give a detailed overview of an SC-FDMA system. We then analyze
analytically and numerically the peak power characteristics and propose a peak power reduction
method that uses symbol amplitude clipping technique. We show that subcarrier mapping
scheme and pulse shaping are significant factors that affect the peak power characteristics and
that symbol amplitude clipping method is an effective way to reduce the peak power without
compromising the link performance. We investigate multiple input multiple output (MIMO)
spatial multiplexing technique in an SC-FDMA system using unitary precoded transmit eigenbeamforming with practical limitations. We also investigate channel-dependent scheduling for an
uplink SC-FDMA system taking into account the imperfect channel information in the form of
feedback delay. To accommodate both low and high mobility users simultaneously, we propose a
hybrid subcarrier mapping method using orthogonal code spreading on top of SC-FDMA and
show that it can have higher capacity gain than that of conventional subcarrier mapping scheme.
List of Contents
Curriculum Vitae
Acknowledgements
vii
An Abstract
viii
List of Figures
xii
List of Tables
xvi
Chapter 1 Introduction
1.1.
Evolution of Cellular Wireless Communications.................................................................................1
1.2.
3GPP Long Term Evolution ...................................................................................................................2
1.3.
Single Carrier FDMA................................................................................................................................5
1.4.
Objectives and Contributions..................................................................................................................6
1.5.
Organization...............................................................................................................................................8
1.6.
Nomenclature...........................................................................................................................................10
Chapter 2 Channel Characteristics and Frequency Multiplexing
13
2.1.
Characteristics of Wireless Mobile Communications Channel........................................................13
2.2.
Orthogonal Frequency Division Multiplexing (OFDM) ..................................................................17
2.3.
Single Carrier with Frequency Domain Equalization (SC/FDE)....................................................19
2.4.
Summary and Conclusions.....................................................................................................................22
Chapter 3 Single Carrier FDMA
23
3.1.
Overview of SC-FDMA System...........................................................................................................25
3.2.
Subcarrier Mapping.................................................................................................................................28
3.3.
Time Domain Representation of SC-FDMA Signals........................................................................31
3.4.
SC-FDMA and OFDMA .......................................................................................................................36
xi
3.5.
SC-FDMA and DS-CDMA/FDE........................................................................................................38
3.6.
SC-FDMA Implementation in 3GPP LTE Uplink............................................................................40
3.7.
Chapter 4 MIMO SC-FDMA
47
4.1.
Spatial Diversity and Spatial Multiplexing in MIMO Systems .........................................................48
4.2.
MIMO Channel .......................................................................................................................................49
4.3.
SC-FDMA Transmit Eigen-Beamforming with Unitary Precoding ...............................................53
4.4.
Chapter 5 Peak Power Characteristics of an SC-FDMA Signal: Analytical Analysis
63
5.1.
Upper Bound for IFDMA with Pulse Shaping ..................................................................................64
5.2.
Modified Upper Bound for LFDMA and DFDMA..........................................................................70
5.3.
Comparison with OFDM.......................................................................................................................71
5.4.
Summary and Conclusion ......................................................................................................................72
Chapter 6 Peak Power Characteristics of an SC-FDMA Signal: Numerical Analysis
74
6.1.
PAPR of Single Antenna Transmission Signals..................................................................................76
6.2.
PAPR of Multiple Antenna Transmission Signals .............................................................................80
6.3.
Peak Power Reduction by Symbol Amplitude Clipping ....................................................................83
6.4.
Chapter 7 Channel-Dependent Scheduling of Uplink SC-FDMA Systems
90
7.1.
Channel-Dependent Scheduling in an Uplink SC-FDMA System..................................................91
7.2.
Impact of Imperfect Channel State Information on CDS...............................................................96
7.3.
Hybrid Subcarrier Mapping ................................................................................................................ 106
7.4.
Summary and Conclusions.................................................................................................................. 110
Chapter 8 Conclusions
111
Appendix A Derivations of the Upper Bounds in Chapter 5
115
Bibliography
126
xii
List of Figures
Figure 2.1: Delay profile and frequency response of 3GPP 6-tap typical urban (TU6) Rayleigh
fading channel in 5 MHz band. ..................................................................................................... 14
Figure 2.2: Time variation of 3GPP TU6 Rayleigh fading channel in 5 MHz band with 2GHz
carrier frequency............................................................................................................................... 16
Figure 2.3: Transmitter and receiver structures of SC/FDE and OFDM. ..................................... 19
Figure 2.4: Dissimilarities between OFDM and SC/FDE................................................................. 21
Figure 3.1: Transmitter and receiver structure of SC-FDMA and OFDMA systems. .................. 24
Figure 3.2: Raised-cosine filter. .............................................................................................................. 27
Figure 3.3: Generation of SC-FDMA transmit symbols.................................................................... 28
Figure 3.4: Subcarrier mapping modes; distributed and localized. ................................................... 29
Figure 3.5: An example of different subcarrier mapping schemes for N = 4, Q = 3 and M = 12.
............................................................................................................................................................ 30
Figure 3.6: Subcarrier allocation methods for multiple users (3 users, 12 subcarriers, and 4
subcarriers allocated per user)........................................................................................................ 30
Figure 3.7: Time symbols of different subcarrier mapping schemes. .............................................. 35
Figure 3.8: Amplitude of SC-FDMA signals........................................................................................ 35
Figure 3.9: Dissimilarities between OFDMA and SC-FDMA........................................................... 37
xiii
Figure 3.10: DS-CDMA with FDE. ...................................................................................................... 38

Figure 3.11: Spreading with the roles of data sequence and signature sequence exchanged for
spreading signature of {1, 1, 1, 1} with a data block size of 4................................................. 39
Figure 3.12: Basic sub-frame structure in the time domain. .............................................................. 40
Figure 3.13: Physical mapping of a block in RF frequency domain (fc: carrier center frequency).
............................................................................................................................................................ 41
Figure 3.14: Generation of a block. ...................................................................................................... 41
Figure 3.15: FDM and CDM pilots for three simultaneous users with 12 total subcarriers. ........ 45
Figure 4.1: Description of a MIMO channel with Nt transmit antennas and Nr receive antennas.
............................................................................................................................................................ 50
Figure 4.2: Block diagram of a spatial multiplexing MIMO SC-FDMA system. ........................... 53
Figure 4.3: Simpified block diagram of a unitary precoded TxBF SC-FDMA MIMO system.... 54
Figure 4.4: Input-output characteristics of the quantizers. ................................................................ 57
Figure 4.5: FER performance of a 2x2 SC-FDMA unitary precoded TxBF system with feedback
averaging and quantization. ............................................................................................................ 58
Figure 4.6: FER performance................................................................................................................. 59
delays of 2, 4, and 6 TTIs. ............................................................................................................. 61
Figure 5.1: CCDF of instantaneous power for IFDMA with BPSK modulation and different
values of roll-off factor . ............................................................................................................. 68
Figure 5.2: CCDF of instantaneous power for IFDMA with QPSK modulation and different
xiv
values of roll-off factor . ............................................................................................................. 69

Figure 5.3: CCDF of instantaneous power for LFDMA with BPSK modulation and different
values of input block size N........................................................................................................... 70
Figure 5.4: CCDF of instantaneous power for IFDMA, LFDMA, and OFDM. For IFDMA, we
consider roll-off factor of 0.2. ...................................................................................................... 72
Figure 6.1: A theoretical relationship between PAPR and transmit power efficiency for ideal class
A and B amplifiers. .......................................................................................................................... 75
Figure 6.2: Comparison of CCDF of PAPR for IFDMA, DFDMA, LFDMA, and OFDMA
with total number of subcarriers M = 512, number of input symbols N = 128, IFDMA
spreading factor Q = 4, DFDMA spreading factor Q = 2, and (roll-off factor) = 0.22.. 78
Figure 6.3: Comparison of CCDF of PAPR for IFDMA and LFDMA with M = 256, N = 64, Q
= 4, Q = 2, and (roll-off factor) of 0, 0.2, 0.4, 0.6, 0.8, and 1. ............................................ 79
Figure 6.4: Precoding in the frequency domain is convolution and summation in the time
domain.k refers to the subcarrier number.................................................................................... 80
Figure 6.5: CCDF of PAPR for 2x2 unitary precoded TxBF............................................................ 81
Figure 6.6: Impact of quantization and averaging of the precoding matrix on PAPR.................. 82
Figure 6.7: PAPR comparison with other MIMO schemes. .............................................................. 82
Figure 6.8: Three types of amplitude limiter........................................................................................ 85
Figure 6.9: Block diagram of a symbol amplitude clipping method for SC-FDMA MIMO
transmission with Mt transmit antenna......................................................................................... 86
xv
Figure 6.10: CCDF of symbol power after clipping. .......................................................................... 86

Figure 6.11: Link level performance for clipping. ............................................................................... 87
Figure 6.12: PSD of the clipped signals................................................................................................ 87
Figure 7.1: Comparison of aggregate throughput with M = 256 system subcarriers, N = 8
subcarriers per user, bandwidth = 5 MHz, and noise power per Hz = -160 dBm................ 94
Figure 7.2: Average user data rate as a function of user distance with M = 256 system subcarriers,
N = 8 subcarriers per user, bandwidth = 5 MHz, and noise power per Hz = -160 dBm..... 95
Figure 7.3: Block diagram of an uplink SC-FDMA system with adaptive modulation and CDS
for K users. ....................................................................................................................................... 97
Figure 7.4: System throughput vs. SNR for the 8 classes of QAM................................................ 103
Figure 7.5: Aggregate throughput with CDS and adaptive modulation......................................... 104
Figure 7.6: Aggregate throughput with CDS and constant modulation (16-QAM) with mobile
speed of 60 km/h.......................................................................................................................... 104
Figure 7.7: Aggregate throughput with CDS and adaptive modulation with feedback delay of 3
ms and different mobile speeds. .................................................................................................. 105
Figure 7.8: Conventional subcarrier mapping and hybrid subcarrier mapping. ............................ 107
Figure 7.9: Block diagram of an SC-CFDMA system. ..................................................................... 108
Figure 7.10: Comparison between SC-FDMA and SC-CFDMA in terms of occupied subcarriers
for the same number of users...................................................................................................... 108
Figure 7.11: Aggregate throughputs for hybrid subcarrier mapping method and other
conventional subcarrier mapping methods with CDS and adaptive modulation................. 109
xvi
List of Tables
Table 2.1: Transmission bandwidths of current / future cellular wireless standards..................... 15
Table 3.1: Parameters for uplink SC-FDMA transmission scheme in 3GPP LTE. ....................... 42
Table 3.2: Number of RUs and number of subcarrriers per RU for LB........................................ 43
Table 4.1. Summary of feedback overhead vs. performance loss..................................................... 60
Table 6.1: 99.9-percentile PAPR for IFDMA, DFDMA, LFDMA, and OFDMA ........................ 78
Table 7.1: SNR boundaries for adaptive modulation........................................................................ 103
Chapter 1
Introduction
1.1.
Evolution of Cellular Wireless Communications
During the 1950s and 1960s, researchers at AT&T Bell Laboratories and companies around
the world developed the idea of cellular radiotelephony. The concept of cellular wireless
communications is to break the coverage zone into small cells and reuse the portions of the
available radio spectrum. In 1979, the worlds first cellular system was deployed by Nippon
Telephone and Telegraph (NTT) in Japan and thus began the evolution of cellular wireless
communications [1], [2].
The first generation of cellular wireless communication systems utilized analog
communication techniques and its focus was on accommodating voice traffic. Frequency
modulation (FM) and frequency division multiple access (FDMA) were the basis of the first
generation systems. AMPS (Advanced Mobile Phone System) in US and ETACS (European
Total Access Cellular System) in Europe were among the first generation systems.
The second generation systems saw the advent of digital communication techniques
which greatly improved spectrum efficiency. Also they vastly enhanced the voice quality and
made possible the packet data transmission. The main multiple access schemes are time
division multiple access (TDMA) and code division multiple access (CDMA). GSM (Global
System for Mobile) which is based on TDMA and IS-95 which is based on CDMA are two
most widely accepted second generation systems.
In the mid-1980s, the concept for IMT-2000 (International Mobile Telecommunications2000) was born at the ITU (International Telecommunication Union) as the third generation
(3G) system for mobile communications [3]. Key objectives of IMT-2000 are to provide
seamless global roaming and to provide seamless delivery of services over a number of media
via higher data rate link. In 2000, a unanimous approval of the technical specifications for 3G
system under the brand IMT-2000 was made and UMTS/WCDMA (Universal Mobile
Telecommunications System/Wideband CDMA) and cdma2000 are two prominent standards
under IMT-2000 both of which are based on CDMA. IMT-2000 provides higher transmission
rates; a minimum speed of 2 Mbps for stationary or walking users and 348 kbps in a moving
vehicle whereas second generation systems only provide speeds ranging from 9.6 kbps to 28.8
kbps. Since the initial standardization, both WCDMA and cdma2000 have evolved into socalled 3.5G; UMTS through HSD/UPA (High Speed Downlink/Uplink Packet Access) and
cdma2000 through 1xEV-DO Rev A (1x Evolution Data-Optimized Revision A).
Currently, 3rd Generation Partnership Project Long Term Evolution (3GPP LTE) is
considered as the prominent path to the next generation of cellular system beyond 3G.
1.2.
3GPP Long Term Evolution
3GPPs work on the evolution of the 3G mobile system started with the Radio Access
Network (RAN) Evolution workshop in November 2004 [4]. Operators, manufacturers, and
research institutes presented more than 40 contributions with views and proposals on the
evolution of the Universal Terrestrial Radio Access Network (UTRAN) which is the
foundation for UMTS/WCDMA systems. They identified a set of high level requirements at
the workshop; reduced cost per bit, increased service provisioning, flexibility of the use of
existing and new frequency bands, simplified architecture and open interfaces, and allow for
reasonable terminal power consumption. With the conclusions of this workshop and with
broad support from 3GPP members, a feasibility study on the Universal Terrestrial Radio
Access (UTRA) and UTRAN Long Term Evolution started in December 2004. The objective
was to develop a framework for the evolution of the 3GPP radio access technology towards a
high-data-rate, low-latency, and packet-optimized radio access technology. The study focused
on means to support flexible transmission bandwidth of up to 20 MHz, introduction of new
transmission schemes, advanced multi-antenna technologies, signaling optimization,
identification of the optimum UTRAN network architecture, and functional split between
RAN network nodes.
The first part of the study resulted in an agreement on the requirements for the Evolved
UTRAN (E-UTRAN). Key aspects of the requirements are as follows [5].
Peak data rate: Instantaneous downlink peak data rate of 100 Mbps within a 20
MHz downlink spectrum allocation (5 bps/Hz) and instantaneous uplink peak
data rate of 50 Mbps (2.5 bps/Hz) within a 20 MHz uplink spectrum allocation.
Control-plane capacity: At least 200 users per cell should be supported in the
active state for spectrum allocations up to 5 MHz.
User-plane latency: Less than 5 ms in an unloaded condition (i.e. single user with
single data stream) for small IP packet.
Mobility: E-UTRAN should be optimized for low mobile speed from 0 to 15

km/h. Higher mobile speeds between 15 and 120 km/h should be supported with
high performance. Mobility across the cellular network shall be maintained at
speeds from 120 to 350 km/h (or even up to 500 km/h depending on the
frequency band).
Coverage: Throughput, spectrum efficiency, and mobility targets should be met

for 5 km cells and with a slight degradation for 30 km cells. Cells ranging up to
100 km should not be precluded.
Enhanced multimedia broadcast multicast service (E-MBMS).
Spectrum flexibility: E-UTRA shall operate in spectrum allocations of different

sizes including 1.25 MHz, 1.6 MHz, 2.5 MHz, 5 MHz, 10 MHz, 15 MHz, and 20
MHz in both uplink and downlink.
Architecture and migration: Packet-based single E-UTRAN architecture with

provision to support systems supporting real-time and conversational class traffic
and support for an end-to-end quality-of-service (QoS).
Radio resource management: Enhanced support for end-to-end QoS, efficient

support for transmission of higher layers, and support of load sharing and policy
management across different radio access technologies.
The wide set of options initially identified by the early LTE work was narrowed down in
December 2005 to a working assumption that the downlink would use Orthogonal Frequency
Division Multiple Access (OFDMA) and the uplink would use Single Carrier Frequency
Division Multiple Access (SC-FDMA). Supported downlink data modulation schemes are
QPSK, 16QAM, and 64QAM, and possible uplink data modulation schemes are /2-shifted
BPSK, QPSK, 8PSK and 16QAM. They agreed the use of Multiple Input Multiple Output
(MIMO) scheme with possibly up to four antennas at the mobile side and four antennas at the
base station. Re-using the expertise from the UTRAN, they agreed to the same channel coding
type as UTRAN (turbo codes). They agreed to a transmission time interval (TTI) of 1 ms to
reduce signaling overhead and to improve efficiency.
The study item phase ended in September 2006 and the LTE works are scheduled to
conclude in early 2008 and produce a technical standard. More technical details on 3GPP LTE
are at [6] and [7].
1.3.
Single Carrier FDMA
Ever increasing demand for higher data rate is leading to utilization of wider transmission
bandwidth. Broadband wireless mobile communications suffer from multipath frequencyselective fading. For broadband multipath channels, conventional time domain equalizers are
impractical for complexity reason.
Orthogonal frequency division multiplexing (OFDM), which is a multicarrier
communication technique, has become widely accepted primarily because of its robustness
against frequency-selective fading channels which are common in broadband mobile wireless
communications [8]. Orthogonal frequency division multiple access (OFDMA) is a multiple
access scheme which is an extension of OFDM to accommodate multiple simultaneous users.

OFDM/OFDMA technique is currently adopted in wireless LAN (IEEE 802.11a & 11g),
WiMAX (IEEE 802.16), and 3GPP LTE downlink systems. Despite the benefits of OFDM
and OFDMA, they suffer a number of drawbacks including: high peak-to-average power ratio
(PAPR), a need for an adaptive or coded scheme to overcome spectral nulls in the channel,
and high sensitivity to frequency offset.
Single carrier frequency division multiple access (SC-FDMA) which utilizes single carrier
modulation and frequency domain equalization is a technique that has similar performance
and essentially the same overall complexity as those of OFDMA system [9]. One prominent
advantage over OFDMA is that the SC-FDMA signal has lower PAPR because of its inherent
single carrier structure. SC-FDMA has drawn great attention as an attractive alternative to
OFDMA, especially in the uplink communications where lower PAPR greatly benefits the
mobile terminal in terms of transmit power efficiency and manufacturing cost. SC-FDMA has
two different subcarrier mapping schemes; distributed and localized. In distributed subcarrier
mapping scheme, users data occupy a set of distributed subcarriers and we achieve frequency
diversity. In localized subcarrier mapping scheme, users data inhabit a set of consecutive
localized subcarriers and we achieve frequency-selective gain through channel-dependent
scheduling (CDS). SC-FDMA is currently a working assumption for the uplink multiple access
scheme in 3GPP LTE [10].
1.4.
Objectives and Contributions

Introduction to SC-FDMA: SC-FDMA is a new radio interface technique that is currently
being adopted in 3GPP LTE uplink. In this thesis, we give a detailed overview of an
SC-FDMA system and explain its transmit and receive process. We also illustrate the
similarities to and differences with an OFDMA system.
Time domain representation of the transmit symbols of an SC-FDMA signal: In this thesis, we
derive the time domain representation of the transmit symbols for each subcarrier
mapping scheme.
Analysis of unitary precoded transmit eigen-beamforming (TxBF) for SC-FDMA with limited
feedback: In this thesis, we numerically analyze the performance of a unitary precoded
TxBF SC-FDMA system with limited feedback. We show the impacts of feedback
quantization/averaging and feedback delay on the link level performance and also on
the transmit PAPR characteristics.
Analytical analysis of the peak power of an SC-FDMA signal: In this thesis, we derive an
analytical upper-bound on the distribution of the instantaneous power of an SCFDMA signal using Chernoff bound. We also characterize the peak power distribution
for each of the subcarrier mapping scheme. We show analytically that an SC-FDMA
signal has indeed lower peak power than an OFDM signal.
PAPR characteristics of an SC-FDMA signal: PAPR is an important measure that affects

the transmit power efficiency. In this thesis, we numerically characterize the PAPR for
different subcarrier mapping considering pulse shaping. We consider both single
antenna and multiple antenna transmissions.
Peak power reduction by symbol amplitude clipping method: In this thesis, we propose a symbol
amplitude clipping method to reduce the peak power of the SC-FDMA transmit signal.
The proposed method effectively reduces the peak power while hardly affecting the
link level performance.
Channel-dependent resource scheduling with hybrid subcarrier mapping: Channel-dependent

scheduling (CDS) results in high capacity gain when channel state information (CSI) is
accurate but the gain decreases when the quality of CSI becomes poor. In this thesis,
we propose a hybrid subcarrier mapping scheme utilizing orthogonal code spreading
for cases where there are both high and low mobility users at the same time. Our
proposed scheme yields higher capacity gain when high mobility users are dominant.
1.5.
Organization
The following is the outline of the remainder of the thesis.

Chapter 2 gives a general overview of OFDM and frequency domain equalization. We first
characterize the wireless mobile communications channel. Then, we give an overview of
OFDM which is a popular multicarrier modulation technique, and we explain single carrier
modulation with frequency domain equalization (SC/FDE) and compare it with OFDM.
Chapter 3 introduces SC-FDMA. We first give an overview of SC-FDMA and explain the
transmission and reception operations in detail. We describe the two flavors of subcarrier
mapping schemes in SC-FDMA, distributed and localized, and briefly compare the two, and
we derive the time domain representation of SC-FDMA transmit signal for each subcarrier
mapping mode. Then, we give an in-depth comparison between SC-FDMA and OFDMA and
we also compare SC-FDMA with direct sequence spread spectrum code division multiple
access (DS-CDMA) with frequency domain equalization. Lastly, we illustrate in detail the SCFDMA implementation in the physical layer according to 3GPP LTE uplink and we describe
the reference signal structure of SC-FDMA.
Chapter 4 investigates MIMO techniques for an SC-FDMA system. We first give a general
overview of MIMO concepts and describe the parallel decomposition of a MIMO channel for
narrowband and wideband transmission. Then, we illustrate the realization of MIMO spatial
multiplexing in SC-FDMA. We introduce the SC-FDMA TxBF with unitary precoding
technique and numerically analyze the link level performance with practical considerations.
Chapter 5 analyzes the distribution of the instantaneous peak power of an SC-FDMA
signal and shows analytically that an SC-FDMA signal has statistically lower peak power than
an OFDM signal. Using the Chernoff bound, we first derive an upper-bound of the
complementary cumulative distribution function (CCDF) of the instantaneous power for
IFDMA with pulse shaping and show the bounds for BPSK and QPSK with raised-cosine
pulse shaping filter. We also derive a modified upper-bound of the CCDF of the
instantaneous power for LFDMA without pulse shaping. Then, we compare the results with
the analytical CCDF of PAPR for OFDM.
Chapter 6 investigates the PAPR characteristics of an SC-FDMA signal numerically. We
first characterize the PAPR for single antenna transmission of SC-FDMA. We investigate the
PAPR properties for different subcarrier mapping schemes. Then, we analyze the PAPR
characteristics for multiple antenna transmission. Specifically, we numerically analyze the
CCDF of PAPR for a 2x2 unitary precoded TxBF SC-FDMA system described in chapter 4.
10
Lastly, we propose a symbol amplitude clipping method to reduce peak power and show the
link level performance and frequency domain aspects of the proposed clipping method.
Chapter 7 presents a channel-dependent scheduling (CDS) method for an SC-FDMA
system in the uplink communications. We first give a general overview of CDS in an uplink
SC-FDMA system. We also analyze the capacity for the two subcarrier mapping schemes.
Then, we investigate the impact of imperfect channel state information (CSI) on CDS. We
analyze the data throughput of an SC-FDMA system with uncoded adaptive modulation and
CDS when there is a feedback delay. Lastly, we propose a hybrid subcarrier mapping scheme
using direct sequence spreading technique on top of SC-FDMA modulation. We show the
throughput improvement of the hybrid subcarrier mapping over the conventional subcarrier
mapping schemes.
Chapter 8 presents a summary of the work and concluding remarks.
1.6.
Nomenclature
3GPP
3rd Generation Partnership Project
BER
Bit Error Rate
BPSK
Binary Phase Shift Keying
CAZAC
Constant Amplitude Zero Auto-Correlation
CCDF
Complementary Cumulative Distribution Function
CDM
Code Division Multiplexing
CDMA
Code Division Multiple Access
CDS
Channel-Dependent Scheduling
11
CP
Cyclic Prefix
CSI
Channel State Information
DFDMA
Distributed Frequency Division Multiple Access
DFT
Discrete Fourier Transform
FDE
Frequency Domain Equalization
FDM
Frequency Division Multiplexing
FDMA
Frequency Division Multiple Access
FER
Frame Error Rate
FFT
Fast Fourier Transform
GSM
Global System for Mobile
IBI
Inter-Block Interference
ICI
Inter-Carrier Interference
IDFT
Inverse Discrete Fourier Transform
ISI
Inter-Symbol Interference
IFDMA
Interleaved Frequency Division Multiple Access
LB
Long Block
LFDMA
Localized Frequency Division Multiple Access
LMMSE
Linear Minimum Mean Square Error
LTE
Long Term Evolution
MIMO
Multiple Input Multiple Output
MMSE
Minimum Mean Square Error
12
OFDM
Orthogonal Frequency Division Multiplexing
OFDMA
Orthogonal Frequency Division Multiple Access
PAPR
Peak-to-Average Power Ratio
PSD
Power Spectral Density
QAM
Quadrature Amplitude Modulation
QPSK
Quadrature Phase Shift Keying
RU
Resource Unit
SB
Short Block
SC
Single Carrier
SC/FDE
Single Carrier with Frequency Domain Equalization
SC-FDMA
Single Carrier Frequency Division Multiple Access
SFBC
Space-Frequency Block Coding
SM
Spatial Multiplexing
SNR
Signal-to-Noise Ratio
SVD
Singular Value Decomposition
TDM
Time Division Multiplexing
TDMA
Time Division Multiple Access
TTI
Transmission Time Interval
TxBF
Transmit Eigen-Beamforming
WCDMA
Wideband Code Division Multiple Access
Chapter 2
Channel Characteristics and Frequency Multiplexing

Multiplexing
In this chapter, we first characterize the wireless mobile communications channel. In section 2.2,
we give an overview of orthogonal frequency division multiplexing (OFDM) which is a popular
multicarrier modulation technique and in section 2.3, we explain single carrier modulation with
frequency domain equalization (SC/FDE) and compare it with OFDM.
2.1.
Characteristics of Wireless Mobile Communications Channel
In a wireless mobile communication system, a transmitted signal propagating through the

wireless channel often encounters multiple reflective paths until it reaches the receiver [11]. We
refer to this phenomenon as multipath propagation and it causes fluctuation of the amplitude
and phase of the received signal. We call this fluctuation multipath fading and it can occur either
in large scale or in small scale. Large-scale fading represents the average signal power attenuation
or path loss due to motion over large areas. Small-scale fading occurs due to small changes in
position and we also call it as Rayleigh fading since the fading is often statistically characterized
with Rayleigh probability density function (pdf). Rayleigh fading in the propagation channel,
which generates inter-symbol interference (ISI) in the time domain, is a major impairment in
wireless communications and it significantly degrades the link performance.
13
14
3GPP 6-Tap Typical Urban (TU6) Channel Delay Profile
Frequency Response of 3GPP TU6 Channel in 5MHz Band

2.5
1
2
Channel Gain [linear]
Amplitude [linear]
0.8
0.6
0.4
0.5
0.2
1.5
3
Time [ sec]
2
3
Frequency [MHz]
Figure 2.1: Delay profile and frequency response of 3GPP 6-tap typical urban (TU6) [12] Rayleigh
fading channel in 5 MHz band.
When characterizing the Rayleigh fading channel, we can categorize it into either flat fading
channel or frequency-selective fading channel. Flat fading occurs when the coherence bandwidth,
which is inversely proportional to channel delay spread, is much larger than the transmission
bandwidth whereas frequency-selective fading happens when the coherence bandwidth is much
smaller than the transmission bandwidth. We show an example of the impulse response and
frequency response of a frequency-selective fading channel in Figure 2.1.
We can also characterize the multipath fading channel in terms of the degree of time
variation of the channel; slow fading and fast fading. The time-varying nature of the channel is
directly related to the movement of the user and the users surrounding, and the degree of the
time variation is associated with the Doppler frequency. Doppler frequency fd is given by
fd =
(2.1)
15
where v is the relative speed of the user and is the wavelength of the carrier. For a given
Doppler frequency fd, the spaced-time correlation function R(t) specifies the extent to which
there is correlation between the channels response in t time interval and it is given by
R ( t ) = J 0 ( 2 f d t )
(2.2)
where J0() is the zero-order Bessel function of the first kind. As the Doppler frequency
increases, the correlation decreases at a given time interval. An example of the time variation of
a fading channel is illustrated in Figure 2.2.
As wireless multimedia applications become more wide-spread, demand for higher data rate
is leading to utilization of a wider transmission bandwidth. For example, Global System for
Mobile Communications (GSM) system, which is a popular second generation cellular system,
uses transmission bandwidth of only 200 kHz but the next generation cellular standard 3GPP
Long Term Evolution (LTE) envisions bandwidth of up to 20 MHz which is 100 times the
bandwidth of GSM. Table 2.1 illustrates the transmission bandwidth in current and future
cellular wireless standards.
Table 2.1: Transmission bandwidths of current / future cellular wireless standards.
Generation
2G
3G
3.5 ~ 4G
Standard
Transmission Bandwidth
GSM
200 kHz
IS-95 (CDMA)
1.25 MHz
WCDMA
5 MHz
cdma2000
5 MHz
3GPP LTE
Up to 20 MHz
WiMAX (IEEE 802.16)
Up to 20 MHz
16
Mobile speed = 3 km/h (5.6 Hz doppler)
0
5
5
4
4
3
3
2
2
1
Frequency [MHz]
1
0
Time [msec]
(a)
Mobile speed = 60 km/h (111 Hz doppler)
0
5
5
4
4
3
3
2
2
1
Frequency [MHz]
1
0
Time [msec]
(b)
Figure 2.2: Time variation of 3GPP TU6 Rayleigh fading channel in 5 MHz band with 2GHz
carrier frequency; (a) user speed = 3 km/h (Doppler frequency = 5.6 Hz); (b) user speed = 60
km/h (Doppler frequency = 111 Hz).
17
With a wider transmission bandwidth, frequency selectivity of the channel becomes more
severe and thus the problem of ISI becomes more serious. In a conventional single carrier
communication system, time domain equalization in the form of tap delay line filtering is
performed to eliminate ISI. However, in case of a wide band channel, the length of the time
domain filter to perform equalization becomes prohibitively large since it linearly increases with
the channel response length.
2.2.
Orthogonal Frequency Division Multiplexing (OFDM)
One way to mitigate the frequency-selective fading seen in a wide band channel is to use a
multicarrier technique which subdivides the entire channel into smaller sub-bands, or subcarriers.
Orthogonal frequency division multiplexing (OFDM) is a multicarrier modulation technique
which uses orthogonal subcarriers to convey information. In the frequency domain, since the
bandwidth of a subcarrier is designed to be smaller than the coherence bandwidth, each
subchannel is seen as a flat fading channel which simplifies the channel equalization process. In
the time domain, by splitting a high-rate data stream into a number of lower-rate data stream
that are transmitted in parallel, OFDM resolves the problem of ISI in wide band
communications. More technical details on OFDM are at [8], [13], [14], [15], [16], and [17].
In summary, OFDM has the following advantages:
For a given channel delay spread, the implementation complexity is much lower than
that of a conventional single carrier system with time domain equalizer.
Spectral efficiency is high since it uses overlapping orthogonal subcarriers in the

frequency domain.
18
Modulation and demodulation are implemented using inverse discrete Fourier

transform (IDFT) and discrete Fourier transform (DFT), respectively, and fast
Fourier transform (FFT) algorithms can be applied to make the overall system
efficient.
Capacity can be significantly increased by adapting the data rate per subcarrier
according to the signal-to-noise ratio (SNR) of the individual subcarrier.
Because of these advantages, OFDM has been adopted as a modulation of choice by many
wireless communication systems such as wireless LAN (IEEE 802.11a and 11g) and DVB-T
(Digital Video Broadcasting-Terrestrial).
However, it suffers from the following drawbacks [18], [19]:
High peak-to-average power ratio (PAPR): The transmitted signal is a superposition

of all the subcarriers with different carrier frequencies and high amplitude peaks
occur because of the superposition.
High sensitivity to frequency offset: When there are frequency offsets in the
subcarriers, the orthogonality among the subcarriers breaks and it causes intercarrier interference (ICI).
A need for an adaptive or coded scheme to overcome spectral nulls in the channel:
In the presence of a null in the channel, there is no way to recover the data of the
subcarriers that are affected by the null unless we use rate adaptation or a coding
scheme.
19
SC/FDE
{ xn }
Add
CP/
PS
Channel
Remove
CP
Npoint
DFT
Equalization
Add
CP/
PS
Channel
Remove
CP
Npoint
DFT
Equalization
Npoint
IDFT
Detect
OFDM
Npoint
IDFT
{ xn }
Detect
* CP: Cyclic Prefix, PS: Pulse Shaping
Figure 2.3: Transmitter and receiver structures of SC/FDE and OFDM.
2.3.
Single Carrier with Frequency Domain Equalization (SC/FDE)
For broadband multipath channels, conventional time domain equalizers are impractical because
of the complexity (very long channel impulse response in the time domain). Frequency domain
equalization (FDE) is more practical for such channels. Single carrier with frequency domain
equalization (SC/FDE) technique is another way to fight the frequency-selective fading channel.
It delivers performance similar to OFDM with essentially the same overall complexity, even for
long channel delay [18], [19]. Figure 2.3 shows the block diagram of SC/FDE and compares it
with that of OFDM.
In the transmitter of SC/FDE, we add a cyclic prefix (CP), which is a copy of the last part
of the block, to the input data at the beginning of each block in order to prevent inter-block
20
interference (IBI) and also to make linear convolution of the channel impulse response look like
a circular convolution. It should be noted that circular convolution problem exists for any FDE
since multiplication in the DFT-domain is equivalent to circular convolution in the time domain
[20]. When the data signal propagates through the channel, it linearly convolves with the channel
impulse response. An equalizer basically attempts to invert the channel impulse response and
thus channel filtering and equalization should have the same type of convolution, either linear or
circular convolution. One way to resolve this problem is to add a CP in the transmitter that will
make the channel filtering look like a circular convolution and match the DFT-based FDE.
Another way is not to use CP but perform an overlap and save method in the frequency
domain equalizer to emulate the linear convolution [20].
SC/FDE receiver transforms the received signal to the frequency domain by applying DFT
and does the equalization process in the frequency domain. Most of the well-known time
domain equalization techniques, such as minimum mean-square error (MMSE) equalization,
decision feedback equalization, and turbo equalization, can be applied to the FDE and the
details of the frequency domain implementation of these techniques are found in [21], [22], [23],
[24], [25], and [26]. After the equalization, the signal is brought back to the time domain via
IDFT and detection is performed.
Comparing the two systems in Figure 2.3, it is interesting to find the similarity between the
two. Overall, they both use the same communication component blocks and the only difference
between the two diagrams is the location of the IDFT block. Thus, one can expect the two
systems to have similar link level performance and spectral efficiency.
21
OFDM
SC/FDE
DFT
DFT
Equalizer
Detect
Equalizer
Detect
Equalizer
Detect
Equalizer
IDFT
Detect
(a)
OFDM symbol
SC/FDE symbols
time
(b)
Figure 2.4: Dissimilarities between OFDM and SC/FDE; (a) different detection processes in the
receiver; (b) different modulated symbol durations.
However, there are distinct differences that make the two systems perform differently as
illustrated in Figure 2.4. In the receiver, OFDM performs data detection on a per-subcarrier
basis in the frequency domain whereas SC/FDE does it in the time domain after the additional
IDFT operation. Because of this difference, OFDM is more sensitive to a null in the channel
spectrum and it requires channel coding or power/rate control to overcome this deficiency. Also,
the duration of the modulated time symbols are expanded in the case of OFDM with parallel
transmission of the data block during the elongated time period.
22
In summary, SC/FDE has advantages over OFDM as follows:
Low PAPR due to single carrier modulation at the transmitter.
Robustness to spectral null.
Lower sensitivity to carrier frequency offset.
Lower complexity at the transmitter which will benefit the mobile terminal in
cellular uplink communications.
Single carrier FDMA (SC-FDMA) is an extension of SC/FDE to accommodate multi-user

access, which will be the subject of the next chapter.
2.4.
Summary and Conclusions
Broadband mobile wireless channel suffers from severe frequency-selective fading which causes
the variation of received signal strength. OFDM is a multicarrier technique that overcomes the
frequency-selective fading impairment by transmitting data over narrower subbands in parallel.
Despite the many benefits, OFDM has limits including: high peak-to-average power ratio
(PAPR), high sensitivity to frequency offset, and a need for an adaptive or coded scheme to
overcome spectral nulls in the channel.
Single carrier modulation with frequency domain equalization (SC/FDE) technique is
another way to mitigate the frequency-selective fading. SC/FDE delivers performance similar to
OFDM with essentially the same overall complexity and has advantages including; low PAPR,
robustness to spectral null, lower sensitivity to carrier frequency offset, lower complexity at the
transmitter which will benefit the mobile terminal in cellular uplink communications.
Chapter 3
Single Carrier FDMA
Single carrier frequency division multiple access (SC-FDMA), which utilizes single carrier
modulation and frequency domain equalization, is a technique that has similar performance
and essentially the same overall complexity as those of orthogonal frequency division multiple
access (OFDMA) system. SC-FDMA is an extension of single carrier modulation with
frequency domain equalization (SC/FDE) to accommodate multiple-user access. One
prominent advantage over OFDMA is that the SC-FDMA signal has lower peak-to-average
power ratio (PAPR) because of its inherent single carrier structure [9]. SC-FDMA has drawn
great attention as an attractive alternative to OFDMA, especially in the uplink communications
where lower PAPR greatly benefits the mobile terminal in terms of power efficiency. It is
currently a working assumption for uplink multiple access scheme in 3GPP Long Term
Evolution (LTE) or Evolved UTRA [6], [7], [10].
In this chapter, we first give an overview of SC-FDMA and explain the transmission and
reception operations in detail. In section 3.2, we describe the two flavors of subcarrier
mapping schemes in SC-FDMA and briefly compare the two. In section 3.3, we derive the
time domain representations of SC-FDMA transmit signal for each subcarrier mapping mode.
23
24
SC-FDMA
Npoint
DFT
{ xn }
{X k }
Subcarrier
Mapping
M-
point
{ X } IDFT
l
{ xm }
Add CP
/ PS
DAC
/ RF
Channel
Detect
Npoint
IDFT
Subcarrier
Demapping/
Equalization
Mpoint
DFT
Remove
CP
RF
/ ADC
Subcarrier
Mapping
Mpoint
IDFT
Add CP
/ PS
DAC
/ RF
OFDMA
{ xn }
Channel
Detect
Subcarrier
Demapping/
Equalization
Mpoint
DFT
Remove
CP
RF
/ ADC
* CP: Cyclic Prefix, PS: Pulse Shaping
Figure 3.1: Transmitter and receiver structure of SC-FDMA and OFDMA systems.
In section 3.4, we give an in-depth comparison between SC-FDMA and OFDMA. In
section 3.5, we compare SC-FDMA with direct sequence spread spectrum code division
multiple access (DS-CDMA) with frequency domain equalization and show the similarities
between the two. In section 3.6, we illustrate in detail the SC-FDMA implementation in the
physical layer according to 3GPP LTE uplink. In section 3.7, we describe the reference (pilot)
signal structure of SC-FDMA.
25
3.1.
Overview of SC-FDMA System
Figure 3.1 shows a block diagram of an SC-FDMA system. SC-FDMA can be regarded as
DFT-spread OFDMA, where time domain data symbols are transformed to frequency domain
by DFT before going through OFDMA modulation. The orthogonality of the users stems
from the fact that each user occupies different subcarriers in the frequency domain, similar to
the case of OFDMA. Because the overall transmit signal is a single carrier signal, PAPR is
inherently low compared to the case of OFDMA which produces a multicarrier signal.
The transmitter of an SC-FDMA system converts a binary input signal to a sequence of
modulated subcarriers. At the input to the transmitter, a baseband modulator transforms the
binary input to a multilevel sequence of complex numbers xn in one of several possible
modulation formats. The transmitter next groups the modulation symbols {xn} into blocks
each containing N symbols. The first step in modulating the SC-FDMA subcarriers is to
perform an N-point DFT to produce a frequency domain representation Xk of the input
symbols. It then maps each of the N DFT outputs to one of the M (> N) orthogonal
subcarriers that can be transmitted. If N = M/Q and all terminals transmit N symbols per
block, the system can handle Q simultaneous transmissions without co-channel interference. Q
is the bandwidth expansion factor of the symbol sequence. The result of the subcarrier
mapping is the set X l (l = 0, 1, 2, M-1) of complex subcarrier amplitudes, where N of the
amplitudes are non-zero. As in OFDMA, an M-point IDFT transforms the subcarrier
amplitudes to a complex time domain signal xm . Each xm then are transmitted sequentially.
The transmitter performs two other signal processing operations prior to transmission. It
26
inserts a set of symbols referred to as a cyclic prefix (CP) in order to provide a guard time to
prevent inter-block interference (IBI) due to multipath propagation. The transmitter also
performs a linear filtering operation referred to as pulse shaping in order to reduce out-ofband signal energy.
In general, CP is a copy of the last part of the block, which is added at the start of each
block for a couple of reasons. First, CP acts as a guard time between successive blocks. If the
length of the CP is longer than the maximum delay spread of the channel, or roughly, the
length of the channel impulse response, then, there is no IBI. Second, since CP is a copy of
the last part of the block, it converts a discrete time linear convolution into a discrete time
circular convolution. Thus transmitted data propagating through the channel can be modeled
as a circular convolution between the channel impulse response and the transmitted data block,
which in the frequency domain is a point-wise multiplication of the DFT frequency samples.
Then, to remove the channel distortion, the DFT of the received signal can simply be divided
by the DFT of the channel impulse response point-wise or a more sophisticated frequency
domain equalization technique can be implemented.
One of the commonly used pulse shaping filter is the raised-cosine filter. The frequency
domain and time domain representations of the filter are as follows.
,0 f
T
2T
T
1 1
1+
T
P( f ) = 1 + cos
f
f
,
2T
2T
2T

2
1+
0
, f
2T
(3.1)
27
P(f)
p(t)
0.8
=0
= 0.5
0.8
=1
0.6
0.6
0.4
0.4
=1
0.2
0
0.2
-0.2
0
Frequency
=0
Time
= 0.5
Figure 3.2: Raised-cosine filter.
p(t ) =
sin ( t / T ) cos ( t / T )
t /T
1 4 2t 2 / T 2
(3.2)
where T is the symbol period and is the roll-off factor.

Figure 3.2 shows the raised-cosine filter graphically in the frequency domain and time
domain. Roll-off factor changes from 0 to 1 and it controls the amount of out-of-band
radiation; = 0 generates no out-of-band radiation and as increases, the out-of-band
radiation increases. In the time domain, the pulse has higher side lobes when is close to 0
and this increases the peak power for the transmitted signal after pulse shaping. We further
investigate the effect of pulse shaping on the peak power characteristics in chapters 5 and 6.
Figure 3.3 details the generation of SC-FDMA transmit symbols. There are M subcarriers,
among which N (< M) subcarriers are occupied by the input data. In the time domain, the
input data symbol has symbol duration of T seconds and the symbol duration is compressed
28
{ xn }
DFT
(N-point)
N
T
{X k }
Subcarrier
Mapping
N
T
{ X }
l
IDFT
(M-point)
M >N
N
T = T
M
{ xm }
M
T
N , M : number of data symbols

T , T : symbol durations
Figure 3.3: Generation of SC-FDMA transmit symbols. There are M total number of subcarriers,
among which N (< M) subcarriers are occupied by the input data.
to T = (N/M)T seconds after going through SC-FDMA modulation.
The receiver transforms the received signal into the frequency domain via DFT, de-maps
the subcarriers, and then performs frequency domain equalization. Because SC-FDMA uses
single carrier modulation, it suffers from inter-symbol interference (ISI) and thus equalization
is necessary to combat the ISI. The equalized symbols are transformed back to the time
domain via IDFT, and detection and decoding take place in the time domain.
3.2.
Subcarrier Mapping
There are two methods to choose the subcarriers for transmission as shown in Figure 3.4;
distributed subcarrier mapping and localized subcarrier mapping. In the distributed subcarrier
mapping mode, DFT outputs of the input data are allocated over the entire bandwidth with
zeros occupying the unused subcarriers, whereas consecutive subcarriers are occupied by the
DFT outputs of the input data in the localized subcarrier mapping mode. We will refer to the
localized subcarrier mapping mode of SC-FDMA as localized FDMA (LFDMA) and
29
X 0
X0
Zeros
X 0
Zeros
X0
X1
X1
Zeros
X2
X N 1
X N 1
Zeros
Zeros
X M 1
X M 1
Distributed Mode
Localized Mode
Figure 3.4: Subcarrier mapping modes; distributed and localized.

distributed subcarrier mapping mode of SC-FDMA as distributed FDMA (DFDMA). The
case of M = QN for the distributed mode with equidistance between occupied subcarriers is
called Interleaved FDMA (IFDMA) [27], [28], [24]. IFDMA is a special case of SC-FMDA
and it is very efficient in that the transmitter can modulate the signal strictly in the time
domain without the use of DFT and IDFT.
An example of SC-FDMA transmit symbols in the frequency domain for N = 4, Q = 3
and M = 12 is illustrated in Figure 3.5 and a multi-user perspective is shown in Figure 3.6.
From a resource allocation point of view, subcarrier mapping methods are further divided
into static and channel-dependent scheduling (CDS) methods. CDS assigns subcarriers to
users according to the channel frequency response of each user. For both scheduling methods,
distributed subcarrier mapping provides frequency diversity because the transmitted signal is
30
M = QN
12= 34
{X~
M > QN
12> 24
M = QN
12= 34
{ xn } :
x0
x1
x2
{X k } :
X0
X1
X2
X3
X0
X1
X2
X3
X0
X1
X2
X3
X0
X1
X2
X3
l , IFDMA
{X~
{X~
l , DFDMA
l , LFDMA
}
}
x3
2
N 1
j
nk
DFT X k = xn e N
n=0
, N = 4
frequency
*M : Total number of subcarriers, N : Data block size, Q : Bandwidth spreading factor
Figure 3.5: An example of different subcarrier mapping schemes for N = 4, Q = 3 and M = 12.
Terminal 1
Terminal 2
Terminal 3
subcarriers
Distributed Mode
subcarriers
Localized Mode
Figure 3.6: Subcarrier allocation methods for multiple users (3 users, 12 subcarriers, and 4 subcarriers
allocated per user).
31
spread over the entire bandwidth. With distributed mapping, CDS incrementally improves
performance. By contrast, CDS is of great benefit with localized subcarrier mapping because
it provides significant multi-user diversity. We will discuss this aspect in more detail in chapter
7.
3.3.
Time Domain Representation of SC-FDMA Signals
We derive the time domain symbols without pulse shaping for each subcarrier mapping
scheme. In the subsequent derivations, we will follow the notations in Figure 3.3.
3.3.1. Time Domain Symbols of IFDMA
For IFDMA, the frequency samples after subcarrier mapping
X l / Q , l = Q k
X l =
0 , otherwise
(0 k N 1)
{ X l } can be described as follows.

(3.3)
where 0 l M 1 and M = Q N .
Let m = N q + n ( 0 q Q 1 , 0 n N 1 ). Then,
1
xm ( = x Nq + n ) =
M
=
M 1
m
j 2 l
1 1
M
Xle
=
Q N
l =0
1 1
Q N
N 1
Xke
N 1
X e
k =0
j 2
X e
k =0
j 2
m
k
N
Nq + n
k
N
k =0
1 1
=
Q N
=
j 2
N 1
n
k
N
(3.4)
1
1
xn = x( m )
mod N
Q
Q
The resulting time symbols { xm } are simply a repetition of the original input symbols
32
{ xn } with a scaling factor of 1/Q in the time domain.

When the subcarrier allocation starts from rth subcarrier ( 0 < r Q 1 ), then,
X l =
X l / Qr
, l = Q k + r (0 k N 1)
(3.5)
0 , otherwise
xm ( = x Nq + n ) =
=
1
M
M 1
m
j 2 l
1 1
M
X
e
=
l
Q N
l =0
1 1
Q N
N 1
X ke
X ke
j 2
m mr
j 2 k +
M
N
k =0
mr
M
k =0
1 1
=
Q N
=
Nq + n
j 2
k
N
N 1
N 1
X
k =0
j 2
n
k
N
(3.6)
j 2 mr
e M
1 j 2 mr
1 j 2 mr
e M xn = e M x( m )
mod N
Q
Q
Thus, there is an additional phase rotation of e
j 2
mr
M
when the subcarrier allocation
starts from rth subcarrier instead of subcarrier zero. This phase rotation will also apply to the
other subcarrier mapping schemes as well in the same case.
3.3.2. Time Domain Symbols of LFDMA
For LFDMA, the frequency samples after subcarrier mapping { X l } can be described as
follows.
X , 0 l N 1
X l = l
0 , N l M 1
(3.7)
Let m = Q n + q , where 0 n N 1 and 0 q Q 1 . Then,
xm = xQn + q
1
=
M
M 1
m
j 2 l
1 1
M
X
e
=
l
Q N
l =0
N 1
X e
l =0
j 2
Qn + q
l
QN
(3.8)
33
If q = 0, then,
Qn
j 2
l
1 1 N 1
1 1
X l e QN =
Q N l =0
Q N
1
1
= xn = x( m )
mod N
Q
Q
xm = xQn =
N 1
If q 0, since X l = x p e
j 2
p
l
N
N 1
X le
j 2
n
l
N
l =0
(3.9)
, then (3.8) can be expressed as follows.
p=0
xm = xQn+ q
1 1
=
Q N
=
1 1
Q N
1 1
=
Q N
N 1
X e
x e
x
p =0
1
1 e
Q
Q N
1 1
=
Q N
( n p ) q
+
j 2
l
Q N
N
l =0 p = 0
1 e
N 1
Qn +q
l =0
N 1 N 1
j 2
j 2 ( n p )
q
Q
(n p) q
+
j 2
Q N
N
1 e
1
q
j 2
Q
j 2
N 1
1 1
=
Q N
=
1 1
Q N
N 1 j 2 ( n p ) + q l
x p e N Q N
l =0
p =0
N 1
1 e
N 1
x
p =0
1 e
j 2
q
Q
(3.10)
(n p) q
+
j 2
Q N
N
xp
p =0
N 1 j 2 p l j 2 QQn+Nq l
N
x p e
e
l =0 p =0
N 1
1 e
(n p) q
j 2
+
Q N
N
As can be seen from (3.9) and (3.10), LFDMA signal in the time domain has exact copies
of input time symbols with a scaling factor of 1/Q in the N-multiple sample positions and inbetween values are sum of all the time input symbols in the input block with different
complex-weighting.
3.3.3. Time Domain Symbols of DFDMA
For DFDMA, the frequency samples after subcarrier mapping { X l } can be described as
follows.
34
X , l = Q k
X l = l / Q
0 , otherwise
(0 k N 1)
(3.11)
where 0 l M 1 , M = Q N , and 1 Q < Q . Let m = Q n + q ( 0 n N 1 , 0 q Q 1 ).

Then,
xm (= xQn + q ) =
1 M 1 j 2 Mm l
Xle
M l =0
j 2
1 1 N 1
=
X ke
Q N k =0
(3.12)
Qn + q
Qk
QN
If q = 0, then,
Qn
j 2
Qk
j 2 Qk
1 1 N 1
1 1 N 1
xm = xQn =
X k e QN =
X ke N
Q N k =0
Q N k =0
(Q n )mod N
Q n
j 2
k
j 2
k
1 1 N 1
1 1 N 1
N
N
=
X ke
=
Xke
Q N k =0
Q N k =0
1
1
= x(Q n )
= x( Q ( m ) )
mod N mod N
mod N
Q
Q
N 1
If q 0, since X k = x p e
j 2
p
k
N
(3.13)
, (3.12) can be expressed as follows after derivation.
p=0
xm = xQn + q
Q
j 2 q 1
1
Q
= 1 e
Q
N
N 1
x
(
p=0
1 e
p
p
Qn

Qq
j 2
+
QN
N
(3.14)
It is interesting to see that the time domain symbols of DFDMA have the same structure
as those of LFDMA.
Figure 3.7 shows an example of the time symbols for each subcarrier mapping mode
35
{ xn }
x0
x1
x2
x3
{Q x
x 0 x1
x2 x 3 x 0 x1 x2 x 3 x 0 x1 x2 x 3
{Q x
m , LFDMA
x0 *
* x1 * * x2 * * x 3 * *
{Q x
m , DFDMA
x0 *
* x2 * * x0 * * x 2 * *
m , IFDMA
time
3
* = ck ,m xk
k =0
, ck ,m : complex weight
Figure 3.7: Time symbols of different subcarrier mapping schemes.

0.5
IFDMA
LFDMA
DFDMA
Amplitude [linear]
0.4
0.3
0.2
0.1
10
20
30
Symbol
Figure 3.8: Amplitude of SC-FDMA signals.
40
50
60
36
based on Figure 3.5 for M = 12, N = 4, Q = 3, and Q = 2. Figure 3.8 shows the amplitude of
the signal for each subcarrier mapping for M = 64, N = 16, Q = 4, and Q = 3 and we can see
more fluctuation and higher peak for LFDMA and DFDMA.
3.4.
SC-FDMA and OFDMA
Figure 3.1 includes a block diagram of an OFDMA transmitter. It has much in common with
SC-FDMA. The only difference is the presence of the DFT in SC-FDMA. For this reason SCFDMA is sometimes referred to as DFT-spread or DFT-precoded OFDMA. Other similarities
between the two include: block-based data modulation and processing, division of the
transmission bandwidth into narrower sub-bands, frequency domain channel equalization
process, and the use of CP.
However, there are distinct differences that make the two systems perform differently as
illustrated in Figure 3.9. In terms of data detection at the receiver, OFDMA performs it on a
per-subcarrier basis whereas SC-FDMA does it after additional IDFT operation. Because of
this difference, OFDMA is more sensitive to a null in the channel spectrum and it requires
channel coding or power/rate control to overcome this deficiency. Also, the duration of the
modulated time symbols are expanded in the case of OFDMA with parallel transmission of
the data block during the elongated time period whereas SC-FDMA modulated symbols are
compressed into smaller chips with serial transmission of the data block, much like a direct
sequence code division multiple access (DS-CDMA) system.
37
OFDMA
SC-FDMA
DFT
DFT
Subcarrier
Demapping
Subcarrier
Demapping
Equalizer
Detect
Equalizer
Detect
Equalizer
Detect
Equalizer
IDFT
Detect
(a)
Input data symbols

OFDMA symbol
SC-FDMA symbols*
* Bandwidth spreading factor : 4
time
(b)
Figure 3.9: Dissimilarities between OFDMA and SC-FDMA; (a) different detection processes in
the receiver; (b) different modulated symbol durations.
38
3.5.
SC-FDMA and DS-CDMA/FDE
Direct sequence code division multiple access (DS-CDMA) with FDE is a technique that
replaces the rake combiner, commonly used in the conventional DS-CDMA, with the
frequency domain equalizer [29]. A rake receiver consists of a bank of correlators, each of
which correlate to a particular multipath component of the desired signal. As the number of
multipaths increase, the frequency selectivity in the channel also increases and the complexity
of the rake combiner increases since more correlators are needed. The use of FDE instead of
rake combing can alleviate the complexity problem in DS-CDMA. Block diagram of DSCDMA/FDE is shown in Figure 3.10.
{ xn }
Spreading
Add
CP/
PS
Channel
Remove
CP
Mpoint
DFT
Equalization
Mpoint
IDFT
Despreading
Detect
Figure 3.10: DS-CDMA with FDE.

The transmitter of DS-CDMA/FDE is the same as the conventional DS-CDMA except
for the addition of CP. The FDE in the receiver removes the channel distortion from the
received chip symbols to recover ISI-free chip symbols. For small spreading factors, the link
performance of the rake receiver significantly degrades because of the inter-path interference
and FDE has a much better performance. For large spreading factors, both have similar
performances. More technical details of the DS-CDMA/FDE system can be found in [29].
SC-FDMA is similar to DS-CDMA/FDE in terms of the following aspects:
39
x0
x1
x2
Data Sequence
Conventional
Spreading
x3
Signature Sequence
x0 x0
x0 x0 x1 x1
x 1 x 1 x 2 x 2 x2 x2 x 3 x 3 x 3 x 3
time
Signature Sequence
Exchanged
Spreading
x0 x1
x 2 x 3 x 0 x 1 x 2 x 3 x0 x1 x 2 x 3 x 0 x 1 x 2 x 3
Data Sequence
x0 x1
x 2 x 3 x 0 x 1 x 2 x 3 x0 x1 x 2 x 3 x 0 x 1 x 2 x 3
time
Figure 3.11: Spreading with the roles of data sequence and signature sequence exchanged for spreading
signature of {1, 1, 1, 1} with a data block size of 4.
Both spread narrow-band data into broader band.
They achieve processing gain or spreading gain from spreading.
They both maintain low PAPR because of the single carrier transmission.
An interesting relationship between orthogonal DS-CDMA and IFDMA is that by

exchanging the roles of spreading sequence and data sequence, DS-CDMA modulation
becomes IFDMA modulation [30], [31]. An example of this observation is illustrated in Figure
40
3.11. We could see that the result of the spreading with exchanged roles is in the form of
IFDMA modulated symbols in Figure 3.7.
One advantage of SC-FDMA over DS-CDMA/FDE is that channel dependent resource
scheduling is possible to exploit frequency selectivity of the channel.
3.6.
SC-FDMA Implementation in 3GPP LTE Uplink
In this section, we describe the physical layer implementation of SC-FDMA in 3GPP LTE
frequency division duplex (FDD) uplink according to [10].
3.6.1. Modulation and Channel Coding
Available data modulation schemes are /2-offset BPSK, QPSK, 8-PSK, and 16-QAM. Turbo
code based on 3GPP UTRA Release 6 is used for forward error correcting (FEC) code.
3.6.2. Sub-frame Structure
In 3GPP LTE, the basic unit of a transmission is a sub-frame. Figure 3.12 shows the basic
sub-frame structure in the time domain.
SB
#2
CP
LB #5
CP
LB #4
CP
LB #3
CP
LB #2
CP
SB
#1
CP
LB #1
CP
CP
1 sub-frame = 0.5 ms
LB #6
Figure 3.12: Basic sub-frame structure in the time domain.

A sub-frame which has duration of 0.5 ms consists of six long blocks (LB) and two short
blocks (SB). Cyclic prefix (CP) is added in front of each block. Long blocks are used for
control and/or data transmission and short blocks are used for reference (pilot) signals for
coherent demodulation and/or control/data transmission. Both localized and distributed
41
subcarrier mapped data use the same sub-frame structure. The minimum transmission time
interval (TTI) for uplink is equal to the sub-frame duration, 0.5 ms. It is possible to
concatenate multiple sub-frames into longer uplink TTIs and the TTI configuration can be
either semi-static or dynamic in such case. In 3GPP LTE uplink, subcarriers in the guard band
area are intentionally set to zero amplitude to maintain similar subcarrier structure as that of
downlink OFDMA, which is illustrated in Figure 3.13.
Total Bandwidth (M subcarriers)
Occupied Subcarriers
fc
Frequency (Subcarrier)
Figure 3.13: Physical mapping of a block in RF frequency domain (fc: carrier center frequency).
Figure 3.14 illustrates the generation of a block and Table 3.1 shows the numerology for
different spectrum allocations.
{ x0 , x1 , xN 1}
Serial-toParallel
subc arrier
Zeros
M-1
Subcarrier
Mapping
Npoint
DFT
Zeros
Figure 3.14: Generation of a block.
Mpoint
IDFT
Parallel-toSerial
{ x0 , x1 , xM 1}
42
Table 3.1: Parameters for uplink SC-FDMA transmission scheme in 3GPP LTE.
Bandwidth
(MHz)
Sub-frame
duration
(ms)
LB size (s/# of
occupied subcarriers
/FFT size)
SB size (s/# of
occupied subcarriers
/FFT size)
20
0.5
66.67/1200/2048
33.33/600/1024
(4.13/127)
or (4.39/135)
15
0.5
66.67/900/1536
33.33/450/768
(4.12/95)
or (4.47/103)
10
0.5
66.67/600/1024
33.33/300/512
0.5
66.67/300/512
33.33/150/256
2.5
0.5
66.67/150/256
33.33/75/128
1.25
0.5
66.67/75/128
33.33/38/64
CP duration
(s/# of
subcarriers)
(4.1/63)
or (4.62/71)
(4.04/31)
or (5.08/39)
(3.91/15)
or (5.99/23)
(3.65/7)
or (7.81/15)
Note that the first CP in the sub-frame has longer duration to include ramp up and ramp
down time.
For data and control transmission, we can organize the subcarrier resources in the
frequency domain into a number of resource units (RU). In other words, we can assign the
subcarriers in groups or chunks of certain number rather than individually. Each RU consists
of a number of localized or distributed subcarriers within a long block. Table 3.2 gives the
number of RUs and the number of subcarriers per RU for different bandwidth allocations.
The number of subcarriers per RU may change in the future based on the outcome of the
current interference coordination study. One or more RUs can be assigned to a user. When
43
more than one localized RUs are assigned to a user, they should be contiguous in frequency
domain and when more than one distributed RUs are assigned, the subcarriers assigned
should be equally spaced.
Table 3.2: Number of RUs and number of subcarrriers per RU for LB.
Bandwidth (MHz)
1.25
2.5
5.0
10.0
15.0
20.0
Number of available subcarriers
75
150
300
600
900
1200
Number of available RUs
12
24
36
48
Number of subcarriers per RU
25
25
25
25
25
25
3.6.3. Reference (Pilot) Signal Structure

Uplink reference signals are transmitted within the two short blocks as mentioned in 3.6.2.
They are received and used at the base station for uplink channel estimation in coherent
demodulation/detection and for possible uplink channel quality estimation for uplink channel
dependent scheduling. We should note that the bandwidth of each subcarrier for the reference
signal is twice that for the data signal since the short block has half the length of a long block.
We can construct the reference signal using CAZAC (Constant Amplitude Zero AutoCorrelation) sequences such as Zadoff-Chu polyphase sequences. Zadoff-Chu CAZAC
sequences are defined as
j 2 r k 2 + qk ,
L 2
e
ak =
r k ( k +1)
j 2 L 2 + qk
e
k = 0,1,2,, L 1; for L even
(3.15)
, k = 0,1,2,, L 1; for L odd
44
where r is any integer relatively prime to L and q is any integer [32].

The set of Zadoff-Chu CAZAC sequences has the following properties:
Constant amplitude.
Zero circular autocorrelation.
Flat frequency domain response.
Circular cross-correlation between two sequences is low and it has constant

magnitude provided that L is a prime number.
Orthogonality among uplink reference signals can be achieved using the following
methods:
Frequency division multiplexing (FDM): Each uplink reference signal is

transmitted across a distinct set of subcarriers. This solution achieves signal
orthogonality in the frequency domain.
Code division multiplexing (CDM): Reference signals are constructed that are
orthogonal in the code domain with the signals transmitted across a common set
of sub-carriers. As an example, individual reference signals may be distinguished
by a specific cyclic shift of a single CAZAC sequence.
Time division multiplexing (TDM): Each user will transmit reference signal in
different short blocks. Orthogonality is achieved in the time domain.
A combination of the methods above.
Figure 3.15 illustrates an example of FDM and CDM pilots.
45
User 1
User 2
User 3
subcarriers
FDM Pilots
subcarriers
CDM Pilots
Figure 3.15: FDM and CDM pilots for three simultaneous users with 12 total subcarriers.
3.7.
Single carrier FDMA (SC-FDMA) is a new multiple access scheme that is currently being
adopted in 3GPP LTE uplink. SC-FDMA utilizes single carrier modulation and frequency
domain equalization, and has similar performance and essentially the same overall complexity
as those of OFDMA system. A salient advantage over OFDMA is that the SC-FDMA signal
has lower PAPR because of its inherent single carrier transmission structure. SC-FDMA has
drawn great attention as an attractive alternative to OFDMA, especially in the uplink
communications where lower PAPR greatly benefits the mobile terminal in terms of transmit
power efficiency and manufacturing cost.
SC-FDMA has two different subcarrier mapping schemes; distributed and localized. In
distributed subcarrier mapping scheme, users data occupy a set of distributed subcarriers and
we achieve frequency diversity. In localized subcarrier mapping scheme, users data inhabit a
set of consecutive localized subcarriers and we achieve frequency-selective gain through
channel-dependent scheduling (CDS). The two flavors of subcarrier mapping schemes give
the system designer a flexibility to adapt to the specific needs and requirements.
46
SC-FDMA also bears similarities to direct sequence CDMA (DS-CDMA) with FDE in
terms of; bandwidth spreading, processing gain from spreading, and low PAPR because of the
single carrier transmission. An advantage of SC-FDMA over DS-CDMA/FDE is the ability to
utilize channel dependent resource scheduling to exploit frequency selectivity of the channel.
Chapter 4
MIMO SCSC-FDMA
Multiple input multiple output (MIMO) techniques gathered much attention in recent years as a
way to drastically improve the performance of wireless mobile communications. MIMO
technique utilizes multiple antenna elements on the transmitter and the receiver to improve
communication link quality and/or communication capacity [33]. A MIMO system can provide
two types of gain; spatial diversity gain and spatial multiplexing gain [34]. Spatial diversity improves
the reliability of communication in fading channels and spatial multiplexing increases the
capacity by sending multiple streams of data in parallel through multiple spatial channels.
Transmit eigen-beamforming (TxBF) with unitary precoding is a spatial multiplexing technique
that utilizes the eigen-structure of the channel to generate independent spatial channels.
Practical considerations of TxBF that affect the performance and the overhead of the system
are precoder quantization/averaging and feedback delay.
In this chapter, we first give a general overview of MIMO concepts. In section 4.2, we
describe the parallel decomposition of a MIMO channel for narrowband and wideband
transmission. We also illustrate the realization of MIMO spatial multiplexing in SC-FDMA. In
section 4.3, we introduce the SC-FDMA TxBF with unitary precoding technique and
47
48
numerically analyze the link level performance with practical considerations.
4.1.
Spatial Diversity and Spatial Multiplexing in MIMO Systems
The basic idea behind spatial diversity techniques is to combat channel fading by having multiple
copies of the transmitted signal go through independently fading propagation paths. At the
receiver, multiple independently faded replicas of the transmitted data signal are coherently
combined to achieve a more reliable reception. Spatial diversity gain is the SNR exponent of the
error probability and intuitively, it corresponds to the number of independently faded paths. It is
well known that in a system with Nt transmit antennas and Nr receive antennas, the maximal
diversity gain is NtNr assuming the path gains between the individual transmit and receive
antenna pairs are independently and identically distributed (i.i.d.) Rayleigh-faded [35]. Smart
antenna techniques [36], Alamouti transmit diversity scheme [37], and space-time coding [38] are
some of well known spatial diversity techniques.
If spatial diversity is a means to combat fading, spatial multiplexing is a way to exploit
fading to increase the data throughput. In essence, if the path gains among individual transmitreceive antenna pairs fade independently such that the channel matrix is well-conditioned,
multiple parallel spatial channels can be created and data rate can be increased by transmitting
multiple streams of data through the spatial channels. Spatial multiplexing is especially important
in the high-SNR regime where the system is degree-of-freedom limited as opposed to power
limited in the low-SNR regime. A scheme achieves a spatial multiplexing gain of r if the
supported data rate approaches rlog(SNR) (bits/s/Hz) and we can think of the multiplexing
gaing as the total number of degrees of freedom to communicate. In a system with Nt transmit
49
antennas and Nr receive antennas for high SNR, Foschini [39] and Telatar [40] showed that the
capacity of the channel increases linearly with m = min(Nt, Nr) if the channel gains among the
antenna pairs are i.i.d. Rayleigh-faded. Thus there is m number of parallel channels or m number
of degrees of freedom. BLAST (Bell Labs Space-Time Architecture) [39] is a typical spatial
multiplexing technique.
For a given MIMO channel, both diversity and multiplexing gains can be achieved
simultaneously but there is a fundamental trade-off between the two gains. For example, Zheng
and Tse [41] showed that the optimal diversity gain achievable by any coding scheme of block
length larger than Nt + Nr 1 with multiplexing gain m (integer) is precisely (Nt m)( Nr m)
for i.i.d. Rayleigh slowly-fading channel with Nt transmit and Nr receive antennas. This implies
that out of the total resource of Nt transmit and Nr receive antennas, it is as though m transmit
and m receive antennas were used for multiplexing and the remaining Nt m transmit and Nr
m receive antennas provided the diversity. In summary, higher spatial diversity gain comes at the
price of lowering spatial multiplexing gain and vice versa.
4.2.
MIMO Channel
First, we consider the narrowband MIMO channel. A point-to-point narrowband MIMO

channel with Nt transmit antennas and Nr receive antennas is illustrated in Figure 4.1. We can
represent such a MIMO system by the following discrete-time model.
50
h11
hNr 1
x1
y1
h21
y2
x2
x Nt
hNr Nt
yNr
h11 h1Nt
H =
h
N r 1 hNr Nt
Figure 4.1: Description of a MIMO channel with Nt transmit antennas and Nr receive antennas.
y1 h11 h1Nt x1 n1

+
=
y h
hN r Nt xNt nNr
Nr
Nr 1

=y
=H
=x
(4.1)
=n
y = H x+n
where x is the Nt 1 transmitted signal vector, y is the Nr 1 received signal vector, n is

the Nr 1 zero-mean complex Gaussian noise, and H is the NrNt complex matrix of channel
gains hij representing the gain from jth transmit antenna to ith receive antenna.
By singular value decomposition (SVD) theorem, channel matrix H can be decomposed as
follows.
51
H = UDV H
(4.2)
where U is an NrNr unitary matrix, V is an NtNt unitary matrix, D is an NrNt non-negative

diagonal matrix, and (i) H is a Hermitian (conjugate transpose) operation. The diagonal entries
of D are the non-negative square roots of the eigenvalues of HH H , the columns of U are the
eigenvectors of HH H , and the columns of V are the eigenvectors of H H H .
We can rewrite (4.1) using (4.2) as
y = UDV H x + n
(4.3)
Multiplying U H on both sides of (4.3), it becomes

H
UH y =U
U DV H x + U H n

=I
U y = DV x + U n
H
(4.4)
Let y = U H y , x = V H x , and n = U H n , then,

y = Dx + n
(4.5)
where n has the same statistical properties as n since U is a unitary matrix. Thus n is an Nr
1 zero-mean complex Gaussian noise.
We can see from (4.5) that the MIMO transmission can be decomposed into up to m =
min(Nt, Nr) parallel independent transmissions, which is the basis for the spatial multiplexing
gain.
For a wideband MIMO channel, the entire band can be subdivided into subbands, or
subcarriers. Then, the MIMO channel for each subcarrier becomes a narrowband MIMO
channel. Let the entire band be subdivided in to M subcarriers, then for kth subcarrier
52
( 0 k M 1 ),
Y1, k H11,k H1Nt ,k X 1,k N1, k

+
=
YN , k H N 1,k H N N ,k X N ,k N N , k
r
r
r t
t
r

=Yk
=Hk
= Xk
(4.6)
= Nk
Yk = H k X k + N k
where X k is the Nt1 transmitted signal vector, Yk is the Nr1 received signal vector, N k is
the Nr1 zero-mean complex Gaussian noise, and Hk is the NrNt complex matrix of channel
gains Hij,k representing the gain from jth transmit antenna to ith receive antenna. Similarly with a
narrowband MIMO channel, Hk can be decomposed into U k DkVkH and (4.6) can be expressed
as
Yk = Dk X k + N k
(4.7)
where Yk = U k H Yk , X k = Vk H X k , and N k = U k H N k .
We can use spatial multiplexing MIMO techniques in SC-FDMA by applying the spatial
processing on a subcarrier-by-subcarrier basis, somewhat similar to MIMO-OFDM [42], in the
frequency domain after DFT. A block diagram of a spatial multiplexing MIMO SC-FDMA
system is shown in Figure 4.2.
N-point
DFT
N-point
DFT
Spatial Mapping
53
Subcarrier
Mapping
M-point
IDFT
Add CP /
PS
DA C
/ RF
Subcarrier
Mapping
M-point
IDFT
Add CP /
PS
DA C
/ RF
Detect
N-point
IDFT
Detect
N-point
IDFT
Spatial Combining
/ Equalization
MIMO
Channel
Subcarrier
De-mapping
M-point
DFT
Remove
CP
RF
/ A DC
Subcarrier
De-mapping
M-point
DFT
Remove
CP
RF
/ A DC
Figure 4.2: Block diagram of a spatial multiplexing MIMO SC-FDMA system.
4.3.
SC-FDMA Transmit Eigen-Beamforming with Unitary Precoding
Transmit eigen-beamforming (TxBF) with unitary precoding utilizes the eigen-structure of the
channel to achieve spatial multiplexing [43], [44]. Figure 4.3 illustrates a simpified block diagram
of a unitary precoded TxBF SC-FDMA MIMO system. We describe below the basic processes
of unitary precoded TxBF system. We assume the channel is time-invariant and the channel
estimation is perfect, and we follow the notations in (4.6).
First, the MIMO channel matrix Hk for subcarrier k is decomposed as follows using SVD
for each subcarrier at the receiver.
H k = U k DkVkH
(4.8)
54
Xk
Unitary
Precoding
X k = Vk X k
MIMO Channel
Hk
Hk X k
Nk
Vk = F (Vk )
Yk
Receiver
(LMMSE)
Zk
Feedback
Processing: F ()
(Av eraging
& quantization of V ks)
Figure 4.3: Simpified block diagram of a unitary precoded TxBF SC-FDMA MIMO system.
We feedback Vk to the transmitter and use it as the precoding matrix for transmission. We
may quantize and average the Vks at the receiver to reduce feedback overhead. We denote the
quantization and averaging processes as F() and Vk = F (Vk ) . The quantized and averaged Vk s
are then sent to the transmitter via the feedback channel. The transmitter precodes the data
signal X k with Vk as follows.
X k = Vk X k
(4.9)
After the transmitted signal propagates through the MIMO channel Hk, the received signal
Yk is represented as follows.
Yk = H k X k + N k
(4.10)
At the receiver, we perform linear MMSE (LMMSE) estimation for joint equalization and
MIMO detection. Let Z k = Ak Yk where Ak is an Nt Nr complex matrix. Then, our goal is to
55
find Ak such that mean square error between X k and Z k ( E X k Z k
} ) is minimum. By
applying orthogonality principal, we obtain the following solution for Ak .
Ak = RXX ,k H kH H k RXX ,k H kH + RNN ,k
(4.11)
where H k = H kVk , RXX ,k = X k X kH , and RNN ,k = N k N kH .

There are two practical factors that impact the performance of precoded TxBF; precoder
quanization/averaging and feedback delay. Since the receiver feedbacks the precoding matrix to
the transmitter, the receiver often quantizes and also averages across all subcarriers the
precoding matrix to reduce feedback overhead. We could expect some degradation in
performance because we lose some of precoder information during quantization and averaging
processes. Feedback delay becomes a major factor in the performance of precoded TxBF when
the channel is changing very fast since the precoder is extracted directly from the current
channel matrix.
In the following sections, we give numerical analysis for a 2x2 uplink MIMO system to
quantify the impact of imperfect feedback on the link level performance. We base the simulation
setup on 3GPP LTE uplink and we use the following simulation parameters and assumptions.
Carrier frequency: 2.0 GHz.
Transmission bandwidth: 5 MHz.
Symbol rate: 7.68 million symbols/sec.
TTI length: 0.5 ms.
Number of blocks per TTI: 6 long blocks (LB).
56
Number of occupied subcarriers per LB: 128.
FFT block size: 512.
Cyclic Prefix (CP) length: 32 symbols.
Subcarrier mapping: Distributed.
Pulse shaping: Time domain squared-root raised-cosine filter (rolloff factor = 0.22)
with 2x oversampling.
Channel model: 3GPP SCME C quasi-static Rayleigh fading with mobile speed of 3
km/h [45]. We assume the channel is constant during the duration of a block.
Antenna configurations: 2 transmit and 2 receive antennas
Data modulation: 16-QAM for 1st stream and QPSK for 2nd stream.
Channel coding: 1/3-rate turbo code.
We use linear MMSE equalizer described in (4.11).
We assume no feedback error and perfect channel estimation at the receiver.
4.3.1. Impact of Imperfect Feedback: Precoder Quantization/Averaging

We directly quantize the 2x2 precoding matrix Vk in the form of Jacobi rotation matrix which is
a unitary matrix [46]. We use the following matrix to quantize Vk.
cos k
Vk =
sin k e jk
sin k e jk
cos k
(4.12)
where for Bamp bits to quantize the amplitude component,
k ( 2l 1)
42
Bamp
; l = 1, , 2
Bamp
(4.13)
57
4
0.8
Output phase [rad]
Output amplitude [linear]
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
Input amplitude [linear]
(a)
-2
-4
-4
-2
0
2
Input phase [rad]
(b)
Figure 4.4: Input-output characteristics of the quantizers; (a) Amplitude quantizer (1 bit); (b) Phase
quantizer (2 bits).
and for Bphase bits to quantize the phase component,
2
l
B
2 phase
; l = 1, , 2
B phase
(4.14)
Figure 4.4 shows the input-output quantization characteristics. For phase quantization, we
use the first antenna as a reference and normalize the phase of the second antenna element with
regards to that of first antenna element. To reduce the overhead of the feedback, {Vk}s are
averaged over multiple continuous subcarriers.
Figure 4.5 shows the Monte Carlo simulation results for frame error rate (FER)
performance of the 2x2 SC-FDMA unitary precoded TxBF system. We can see that the
performance loss at 1% FER with regards to no averaging and no quantization case is 1.3 dB for
the worst case (25-subcarrier averaging and 2-bit phase-only quantization).
58
FER, 2x2 TxBF, SCME C 3km/h, 16QAM-QPSK 1/3, Ideal ChEst, No Feedback Delay, Jacobi Direct Quant.
10
-1
FER
10
-2
10
-3
10
-4
10
No avr., no quant.
No avr., phase-only(2 bits) quant.
No avr., amp.(1 bit)-phase(2 bits) quant.
25 SCs avr., phase-only(2 bits) quant.
25 SCs avr., amp.(1 bit)-phase(2 bits) quant.
12 SCs avr., phase-only(2 bits) quant.
12 SCs avr., amp.(1 bit)-phase(2 bits) quant.
1
6
7
SNR [dB]
10
11
12
averaging and quantization.
Figure 4.6 (a) shows the effect of different quantization bit resolutions (no averaging) and
Figure 4.6 (b) shows the effect of different averaging sizes (1-bit amplitude and 2-bit phase
quantization). As we use more bits and smaller averaging size, we see less performance
degradation.
Table 4.1 summarizes the tradeoff between feedback overhead and performance loss based
on the results in Figure 4.5. Performance loss is the SNR level increase at 1% FER point with
regards to that of no averaging and no quantization result. As we increase the feedback overhead,
59
FER, 2x2 TxBF, SCME C 3km/h, 16QAM-QPSK 1/3, Ideal ChEst, No Feedback Delay, Jacobi Direct Quant., No avr.
0
10
-1
FER
10
-2
10
-3
10
No quant.
Phase-only (2 bits) quant.
Amp. (1 bit)-phase (2 bits) quant.
-4
10
5
6
7
8
SNR per Antenna [dB]
10
11
12
(a)
FER, 2x2 TxBF, SCME C 3km/h, 16QA M-QPSK 1/3, Ideal ChEst, No Feedback Delay, Jacobi Direct Quant.
10
-1
FER
10
-2
10
No avr., no quant.
12 SCs avr., quant.
25 SCs avr., quant.
50 SCs avr., quant.
100 SCs avr., quant.
-3
10
-4
10
* 1 bit amp. / 2 bit phase quantization

1
5
6
7
8
10
11
12
(b)
Figure 4.6: FER performance; (a) with different quantization bit resolutions (no averaging); (b) with
different averaging sizes (1-bit amplitude and 2-bit phase quantization).
60
that is, increase the amount of precoding information, there is less performance loss.
Table 4.1. Summary of feedback overhead vs. performance loss.
Averaging size
# of
(subcarriers per RU) RUs
Quantization
resolution (bits)
Feedback
overhead
Performanc
e loss
25
12
Phase: 2, Amp.: none
2 bits 12 = 24 bits
1.3 dB
25
12
Phase: 2, Amp.: 1
3 bits 12 = 36 bits
0.6 dB
12
25
Phase: 2, Amp.: none
2 bits 25 = 50 bits
1 dB
12
25
Phase: 2, Amp.: 1
3 bits 25 = 75 bits
0.5 dB
4.3.2. Impact of Imperfect Feedback: Feedback Delay

Delay is inevitable in any realistic feedback system. When the channel is changing fast due to the
users motion or change of the environment, the precoder may be outdated and it will affect the
receiver performance. Figure 4.7 shows the impact of feedback delay for different user speeds.
We consider feedback delays of 2, 4, and 6 TTIs with no feedback quantization or averaging.
We can see from Figure 4.7 (a) that at low speed of 3 km/h, there is only a trivial loss of
performance even for the longest delay of 6 TTIs. But as the user speed becomes faster, we can
see the performance degradation become more severe as illustrated in Figure 4.7 (b). Clearly, in
high mobility and long feedback delay situations, TxBF with unitary precoding does not perform
well at all.
4.4.
MIMO technique utilizes multiple antenna elements on the transmitter and the receiver to
improve communication link quality and/or communication capacity and it has two aspects in
61
FER, 2x2 TxBF, SCME C 3km/h, 16QAM-QPSK 1/3, Ideal ChEst, No Quant.
FER
10
10
10
10
FER, 2x2 TxBF, SCME C 150km/h, 16QAM-QPSK 1/3, Ideal ChEst, No Quant.
10
-1
10
FER
10
-2
-3
No delay
2 TTI delay
4 TTI delay
6 TTI delay
10
-4
10
4
6
(a)
10
10
-1
-2
-3
No delay
2 TTI delay
4 TTI delay
6 TTI delay
-4
10
15
20
(b)
delays of 2, 4, and 6 TTIs; (a) user speed of 3 km/h; (b) user speed of 150 km/h.
terms of performance improvement; spatial diversity and spatial multiplexing. Spatial diversity
improves the reliability of communication in fading channels and spatial multiplexing increases
the capacity by sending multiple streams of data in parallel through multiple spatial channels.
Transmit eigen-beamforming (TxBF) with unitary precoding is a spatial multiplexing
technique that utilizes the eigen-structure of the channel to generate independent spatial
channels. Practical considerations of TxBF that affect the performance and the overhead of the
system are precoder quantization/averaging and feedback delay. Quantizing and averaging the
precoding matrix reduces the feedback overhead but the link performance suffers.
In this chapter, we gave an overview of the realization of MIMO spatial multiplexing in SCFDMA. We specifically analyzed the unitary precoded TxBF technique on an SC-FDMA system.
62
We showed with numerical simulations that there is a trade-off between the feedback overhead
and the link performance loss. We also showed that feedback delay is another impairment that
degrades the performance especially for the high speed mobile users.
Chapter 5
Peak Power Characteristics

Characteristics of an SCSC-FDMA Signal:
Analytical Analysis
In this chapter, we analytically characterize the peak power of SC-FDMA modulated signals
using the Chernoff bound. We will analyze the peak power characteristics numerically in
chapter 6. In [47], Wulich and Goldfeld showed that the amplitude of a single carrier (SC)
modulated signal does not have a Gaussian distribution and that it is difficult to analytically
derive the exact form of the distribution. As an alternative, they explained a way to derive an
upper bound for the complementary distribution of the instantaneous power using the
Chernoff bound. We follow their derivation and apply it to SC-FDMA modulated signals to
characterize the peak power analytically.
2
The peak power of any signal x(t) is the maximum of its squared envelope x (t ) .
However, for a continuous random process, max x (t )
random process with discrete values where max x (t )
could be unbounded. Even for
is bounded, it may occur at very low
probability which is not very useful in practice. Rather, the distribution of x (t )
is more
useful and we describe it with the complementary cut-off probability. For a given w, cut-off
63
64
probability is defined as Pr x (t ) w = F x ( t )
distribution function (CDF) of
( w ) , where
F x ( t ) 2 ( w ) is the cumulative
x (t ) , and the complementary cut-off probability is
Pr x (t ) w = 1 F x ( t ) 2 ( w ) . Pr x (t ) w
2
is also referred to as complementary
cumulative distribution function (CCDF).

In this chapter, we first derive an upper-bound of the CCDF of the instantaneous power
for interleaved FDMA (IFDMA) with pulse shaping and show the bounds for BPSK and
QPSK with raised-cosine pulse shaping filter. In section 5.2, we show the modified upperbound of the CCDF of the instantaneous power for localized FDMA (LFDMA) without
pulse shaping. In section 5.3, we compare the results of sections 5.1 and 5.2 with the analytical
CCDF of PAPR for OFDM.
5.1.
Upper Bound for IFDMA with Pulse Shaping
In chapter 3 section 3.3.1, we derived the time domain representation of the IFDMA symbols
and showed that it is a repetition of the original input symbols with a scaling factor of 1/Q.
Since the IFDMA signal has the same structure as the conventional SC modulation signal, we
first derive the upper bound on the CCDF of the instantaneous power of a conventional SC
modulated signal.
We consider a baseband representation of the conventional SC modulated signal
x(t , s ) =
k =
where
{sk }k =
p (t kT )
(5.1)
are transmitted symbols, s = [ , s1 , s0 , s1 ,] , p(t) is the pulse shaping
65
filter, and T is the symbol duration. sk belongs to a modulation constellation set C of size B,
that is, sk C = {cb : 0 b B 1} and it is uniformly distributed. We assume
{sk } s
are
independent of each other.

Let us define a random variable Z for a given t0 [0, T) as follows.
Z x(t0 , s ) =
k =
where ak = sk p (t0 kT ) .
p (t0 kT ) =
{ak } s
k =
(5.2)
are also mutually independent since
{sk } s
are mutually
independent.
Since sk is uniformly distributed over C and from (5.2),
Pr [ ak = cb p (t0 kT )] = Pr [ sk = cb ] =
1
B
(5.3)
where 0 b B 1 .
Our goal is to characterize the following CCDF.
2
2
Pr x(t0 , s ) w = Pr Z w = Pr Z
(5.4)
where w 0 and = w .
Let us assume the modulation constellation set C is real and its constellation points are
symmetric. Then,
Pr Z = Pr [ Z ] + Pr [ Z ]
= 2 Pr [ Z ]
Using the Chernoff bound, the following inequality holds [48], [49].
(5.5)
66
{ }
Pr [ Z ] e E eZ
(5.6)
where is a solution of the following equation.
{ }
E Ze Z E e Z = 0
(5.7)
By expanding E {Ze Z } and E {e Z } , (5.7) becomes
k =
{ } =
E {e }
E ak eak
(5.8)
ak
for = . By solving (5.8), we obtain for a given . Since (5.8) is not in a closed form, we
evaluate the solution numerically.
2
We can upper-bound the CCDF Pr x(t0 , s ) w as follows from (5.4), (5.5), and
(5.6).
{ }
Pr x (t0 , s ) w 2e E e ak Pub , SC
k =
(5.9)
The details of the derivations for (5.8) and (5.9) are in section A.1 of appendix A.
Since it is impossible to consider the infinite span in (5.8) and (5.9), we limit the span by
only considering -Kmax k Kmax. As long as the IFDMA input block size N is larger than
Kmax, the instantaneous power distribution of the IFDMA signal will have the same upper
bound as the conventional SC signal since the input symbols are mutually independent within
the span of -Kmax k Kmax.
In the next subsections, we derive the upper bounds specifically for BPSK and QPSK
67
modulations. We consider the following raised-cosine pulse.
sin ( t / T ) cos ( t / T )
t /T
1 4 2t 2 / T 2
p(t ) =
(5.10)
where 0 1 is the roll-off factor.

5.1.1. Upper Bound for BPSK
For BPSK, we use the following constellation set.
CBPSK = {1,1}
(5.11)
Then, (5.8) becomes
K max
k = K max
1
1
p (t0 kT )e p (t0 kT ) p (t0 kT )e p ( t0 kT )

2
2
=
1 p (t0 kT ) 1 p (t0 kT )
e
+ e
2
2
(5.12)
After we determine from (5.12), the upper bound in (5.9) becomes
( BPSK )
ub , SC
1

2
2 K max
( e
K max
p ( t0 kT )
k = K max
+ e p ( t0 kT )
(5.13)
and we upper-bound the CCDF as

2
)
Pr x ( BPSK ) (t0 , s ) w Pub( BPSK
, SC
(5.14)
The details of the derivations for (5.12) and (5.13) are in section A.2 of appendix A.
)
Figure 5.1 shows Pub( BPSK
for Kmax= 8 , T = 1, and t0 = T/2 = 0.5 along with the
, SC
empirical results using Monte Carlo simulation. We considered roll-off factor of 0, 0.2, 0.4,
68
10
Pr(|Z|2 > w)
10
10
10
10
CCDF of Instantaneous Power for IFDMA (BPSK)
-1
-2
-3
= 0.6
= 0.4
-4
= 0.2
4
w [dB]
=0
6
Figure 5.1: CCDF of instantaneous power for IFDMA with BPSK modulation and different values
of roll-off factor . Dotted lines are empirical results and solid lines are the upper bounds.
and 0.6. For the Monte Carlo simulation, we generated 1000 random IFDMA-modulated
blocks with total number of symbols M = 256, number of input symbols N = 64, and
spreading factor Q = 4, and produced a histogram. We can see that the upper bound we
derived in (5.13) is valid compared to the empirical results and that the bound is rather tight in
the tail region of the distribution which we are most interested in. Another interesting
observation is the fact that the peak power at a given probability increases as the roll-off
factor of the raised-cosine filter becomes closer to zero.
5.1.2. Modified Upper Bound for QPSK
)
Figure 5.2 shows the modified upper-bound for QPSK, Pub(QPSK
, for Kmax= 8 , T = 1, and
, SC
69
10
Pr(|Z|2 > w)
10
10
10
CCDF of Instantaneous Power of IFDMA (QPSK)
-1
-2
= 0.2
-3
=0
= 0.6
10
-4
= 0.4
10
w [dB]
Figure 5.2: CCDF of instantaneous power for IFDMA with QPSK modulation and different values
of roll-off factor . Dotted lines are empirical results and solid lines are the upper bounds.
t0 = T/2 = 0.5 along with the empirical results using Monte Carlo simulation. The derivation
)
of Pub(QPSK
is in section A.3 of appendix A. We considered roll-off factor of 0, 0.2, 0.4,
, SC
and 0.6. For the Monte Carlo simulation, we generated 1000 random IFDMA-modulated
blocks with total number of symbols M = 256, number of input symbols N = 64, and
spreading factor Q = 4, and produced a histogram. To make the upper bound tighter, we used
= 2 (see section A.3 of appendix A). We can see that the modified upper bound we
derived is valid compared to the empirical results. As with the case of BPSK in section 5.1.1,
the peak power at a given probability increases as the roll-off factor of the raised-cosine pulse
70
10
Pr(|Z|2>w)
10
10
10
CCDF of Instantaneous Power for LFDMA (BPSK)
-1
-2
N=4
-3
N=8
N =32
10
-4
10
w [dB]
Figure 5.3: CCDF of instantaneous power for LFDMA with BPSK modulation and different values
of input block size N. Dotted lines are empirical results and solid lines are the upper bounds.
shaping filter becomes closer to zero.
5.2.
Modified Upper Bound for LFDMA and DFDMA
In chapter 3 sections 3.3.2 and 3.3.3, we showed that the time domain symbols of localized
FDMA (LFDMA) and distributed FDMA (DFDMA) signals have the same structure. Thus
the distributions of the instantaneous power are the same for both signals. We do not consider
pulse shaping in the following analysis because of the complexity of the derivation.
Figure 5.3 shows the modified upper bound for BPSK, Pub,LFDMA, for M = 256 and
different values of N and Q, along with the empirical results using Monte Carlo simulation.
The details of the derivation for Pub,LFDMA is in section A.4 of appendix A. For the Monte
71
Carlo simulation, we generated 1000 random LFDMA-modulated blocks and produced a

histogram. To make the upper bound tighter, we used =
(see section A.4 of appendix
A). We can see that the upper bound we derived is valid compared to the empirical results. We
can also observe that the peak power at a given probability increases as the input block size
block N increases.
5.3.
Comparison with OFDM
We can express the CCDF of the PAPR of an OFDM data block for N subcarriers with
Nyquist rate sampling as follows [50], [51].
Pr {PAPR w} = 1 1 e w
(5.15)
Let us assume the input symbols have unit power. Then we can use (5.15) and compare it
with the CCDF of the instantaneous power of IFDMA and that of LFDMA.
Figure 5.4 compares the analytical upper bounds of CCDF for IFDMA and LFDMA
with the theoretical CCDF for OFDM with the same block size. For IFDMA, we consider
roll-off factor of 0.2. For LFDMA and OFDM, the number of occupied subcarriers N is 32
and we do not apply pulse shaping filter. We consider the case of BPSK modulation. We can
see that SC-FDMA signals for both subcarrier mapping modes indeed have lower peak power
at a given probability than that of OFDM. Comparing the 99.9-percentile instantaneous power
(w such that Pr Z w = 103 ), we see that the peak power of IFDMA is 4 dB lower than
2
that of OFDM and the peak power of LFDMA is 2.2 dB lower than that of OFDM. We can
72
10
Pr(|Z|2 > w)
10
10
10
CCDF of Instantaneous Power
-1
OFDM
-2
-3
IFDMA
10
-4
LFDMA
6
w [dB]
10
12
Figure 5.4: CCDF of instantaneous power for IFDMA, LFDMA, and OFDM. For IFDMA,
we consider roll-off factor of 0.2. For LFDMA and OFDM, the number of occupied subcarriers N is
32 and we do not apply pulse shaping filter. We consider the case of BPSK modulation.
also observe that LFDMA signal has higher peak power than IFDMA signal by 1.8 dB at 99.9percentile probability.
5.4.
Summary and Conclusion
In this chapter, we investigated the peak power characteristics for IFDMA and LFDMA
signals analytically. By using Chernoff bound, we found upper bounds for both subcarrier
mapping schemes and we showed results for IFDMA with BPSK and QPSK modulations and
for LFDMA with BPSK modulation. We saw that the peak power of an SC-FDMA signal for
both subcarrier mappings is indeed lower than that of OFDM. We also observed that IFDMA
73
has the lowest peak power among the considered subcarrier mapping schemes of SC-FDMA.
With regards to the tightness of the upper bounds, the original upper bounds are rather loose
except for the case of IFDMA/BPSK and we had to modify the original upper bounds to
make them tighter. We prefer numerical analysis in order to precisely characterize the peak
power characteristics for SC-FDMA signals, which will be the subject of the next chapter.
Other interesting findings are:
For IFDMA, the roll-off factor of the raised-cosine pulse shaping filter has a
great impact on the distribution of the instantaneous power. As the roll-off factor
decreases to zero, the peak power at a given probability increases.
For LFDMA, the input block size is the major factor characterizing the CCDF of
peak power.
Chapter 6
Peak Power Characteristics of an SCSC-FDMA Signal:

Signal:
Numerical
Numerical Analysis
Peak-to-average-power ratio (PAPR) is a performance measurement that is indicative of the

power efficiency of the transmitter. In case of an ideal linear power amplifier where we achieve
linear amplification up to the saturation point, we reach the maximum power efficiency when the
amplifier is operating at the saturation point. A positive PAPR in dB means that we need a
power backoff to operate in the linear region of the power amplifier. We can express the
theoretical relationship between PAPR [dB] and transmit power efficiency as follows [52].
= max 10
PAPR
20
(6.1)
where is the power efficiency and max is the maximum power efficiency. For class A power
amplifier, max is 50% and for class B, 78.5% [53]. Figure 6.1 shows the relationship graphically
and it is evident from the figure that high PAPR degrades the transmit power efficiency
performance.
A salient advantage of SC-FDMA over OFDMA is the lower PAPR because of its inherent
single carrier structure [54]. The lower PAPR is greatly beneficial in the uplink communications
74
75
100
90
80
Efficiency [%]
70
Class B
60
50
40
30
20
Class A
10
0
0
10
PAPR [dB]
Figure 6.1: A theoretical relationship between PAPR and transmit power efficiency for ideal class A
and B amplifiers.
where the mobile terminal is the transmitter. As we showed in section 3.3, time domain samples
of the SC-FDMA modulated signals are different depending on the subcarrier mapping scheme
and we can expect different PAPR characteristics for different subcarrier mapping schemes.
In this chapter, we first characterize the PAPR for single antenna transmission of SCFDMA. We investigate the PAPR properties for different subcarrier mapping schemes. In
section 6.2, we analyze the PAPR characteristics for multiple antenna transmission. Specifically,
we numerically analyze the CCDF of PAPR for 2x2 unitary precoded TxBF SC-FDMA system
described in chapter 4. In section 6.3, we propose a symbol amplitude clipping method to reduce
peak power. We show the link level performance and frequency domain aspects of the proposed
76
clipping method.
6.1.
PAPR of Single Antenna Transmission Signals
In this section, we analyze the PAPR of the SC-FDMA signal for each subcarrier mapping mode.
We follow the notations in Figure 3.2 of chapter 3.
Let
{ xn : n = 0,1, , N 1}
be
data
symbols
to
be
modulated.
Then,
{ X k : k = 0,1, , N 1}
are frequency domain samples after DFT of
{ X
are frequency domain samples after subcarrier mapping, and
: l = 0,1, , M 1
{ xm : m = 0,1, , M 1}
are time symbols after IDFT of
{ X
{ xn : n = 0,1, , N 1} ,
: l = 0,1, , M 1 . We
represent the complex passband transmit signal of SC-FDMA x(t) for a block of data as
M 1
x(t ) = e jct xm p (t mT )
(6.2)
m =0
where c is the carrier frequency of the system, p(t) is the baseband pulse, and T is the
symbol duration of the transmitted symbol xm . We consider raised-cosine (RC) pulse and
squared-root raised-cosine (RRC) pulse which are widely used pulse shapes in wireless
communications.
We define the PAPR as follows for transmit signal x(t).
2
max x (t )
peak power of x(t )
PAPR =
= 0t MT
1 MT
2
average power of x(t )
x(t ) dt
0
MT
(6.3)
77
Without pulse shaping, that is, using rectangular pulse shaping, symbol rate sampling will
give the same PAPR as the continuous case since SC-FDMA signal is modulated over a single
carrier. Thus, we express the PAPR without pulse shaping with symbol rate sampling as follows.
PAPR =
xm
max
m = 0,1,, M 1
M 1
1
M
m=0
(6.4)
Using Monte Carlo simulation, we calculate the CCDF (Complementary Cumulative

Distribution Function) of PAPR, which is the probability that PAPR is higher than a certain
PAPR value PAPR0 (Pr{PAPR>PAPR0}). We evaluate and compare the CCDFs of PAPR for
IFDMA, DFDMA, LFDMA, and OFDMA. The following are the simulation setup and
assumptions.
We generate 104 uniformly random data blocks to acquire the CCDF of PAPR.
We consider QPSK and 16-QAM symbol constellations.
We truncate the baseband pulse p(t) from -6 T to 6 T time period and we

oversample it by 8 times and we use transmission bandwidth of 5 MHz.
We consider consecutive chunks for LFDMA.
We do not apply any pulse shaping in the case OFDMA.
Figure 6.2 contains the plots of CCDF of PAPR for IFDMA, DFDMA, LFDMA, and
OFDMA for total number of subcarriers M = 512, number of input symbols N = 128, IFDMA
spreading factor Q = 4, and DFDMA spreading factor Q = 2 . We compare the PAPR value that
is exceeded with the probability less than 0.1% (Pr{PAPR>PAPR0} = 10-3), or 99.9-percentile
78
CCDF of PAPR: QPSK, Rolloff = 0.22, N = 512, N
fft
occupied
= 128
CCDF of PAPR: 16-QAM, Rolloff = 0.22, N = 512, N

fft
10
OFDMA
= 128
OFDMA
-1
-1
10
Pr(PAPR>PAPR )
10
Pr(PAPR>PAPR )
occupied
10
-2
10
IFDMA
DFDMA
-3
10
-4
10
Dotted lines: no PS
Dashed lines: RRC PS
LFDMA
Solid lines: RC PS
6
PAPR [dB]
-2
10
DFDMA
10
12
LFDMA
-3
10
-4
IFDMA
10
Dotted lines: no PS
Dashed lines: RRC PS
Solid lines: RC PS
2
6
PAPR [dB]
10
12
(a)
(b)
Figure 6.2: Comparison of CCDF of PAPR for IFDMA, DFDMA, LFDMA, and OFDMA
with total number of subcarriers M = 512, number of input symbols N = 128, IFDMA spreading
factor Q = 4, DFDMA spreading factor Q = 2, and (roll-off factor) = 0.22; (a) QPSK; (b)
16-QAM.
PAPR. We denote 99.9-percentile PAPR as PAPR99.9%. Table 6.1 summarizes the 99.9-percentile
PAPR for each subcarrier mapping.
Table 6.1: 99.9-percentile PAPR for IFDMA, DFDMA, LFDMA, and OFDMA
Modulation
Pulse shaping
None
QPSK
16-QAM
IFDMA
DFDMA
LFDMA
OFDMA
0 dB
7.7 dB
7.7 dB
11.1 dB
RC
6.2 dB
7.7 dB
8.0 dB
N/A
RRC
5.3 dB
7.8 dB
8.7 dB
N/A
None
3.2 dB
8.7 dB
8.7 dB
11.1 dB
RC
7.8 dB
8.7 dB
9.0 dB
N/A
RRC
7.2 dB
8.7 dB
9.5 dB
N/A
* RC: raised-cosine pulse shaping, RRC: squared-root raised-cosine pulse shaping
79
CCDF of PAPR: QPSK, N = 256, N
fft
occupied
= 64
CCDF of PAPR: 16-QAM, N = 256, N
10
10
IFDMA
occupied
-1
-1
Pr(PAPR>PAPR )
10
= 64
LFDMA
IFDMA
LFDMA
10
Pr(PAPR>PAPR )
fft
-2
10
=1
=0.8
-3
10
-4
10
=0.2
=0.6
=0.4
=0
10
=0.8
10
Solid lines: without pulse shaping

Dotted lines: with pulse shaping
2
4
6
PAPR [dB]
0
(a)
10
=1
-2
10
=0.6 =0.4
=0.2
-3
-4
Solid lines: without pulse shaping

Dotted lines: with pulse shaping
2
=0
4
6
PAPR [dB]
10
(b)
Figure 6.3: Comparison of CCDF of PAPR for IFDMA and LFDMA with M = 256, N = 64,
Q = 4, and (roll-off factor) of 0, 0.2, 0.4, 0.6, 0.8, and 1; (a) QPSK; (b) 16-QAM.
We can see that all the cases for SC-FDMA have indeed lower PAPR than that of OFDMA.
Also, IFDMA has the lowest PAPR, and DFDMA and LFDMA have very similar levels of
PAPR.
Figure 6.3 shows the impact of the roll-off factor on the PAPR when using raised-cosine
pulse shaping for M = 256, N = 64, and Q = 4. We can see that this impact is more obvious in
the case of IFDMA. As the roll-off factor decreases from 1 to 0, PAPR increases significantly
for IFDMA. This observation is consistent with the results of the instantaneous peak power for
IFDMA with pulse shaping in section 5.1 of chapter 5. This implies that there is a tradeoff
between PAPR performance and out-of-band radiation since out-of-band radiation increases
with increasing roll-off factor.
80
6.2.
PAPR of Multiple Antenna Transmission Signals
In this section, we analyze the PAPR characteristics of multiple antenna transmission for SCFDMA system. Specifically, we investigate unitary precoded transmit eigen-beamforming (TxBF)
for 2 transmit and 2 receiver antennas which we described in section 4.3 of chapter 4. Since it is
difficult to derive the PAPR analytically, we resort to numerical analysis using Monte Carlo
simulation to investigate the PAPR properties of our MIMO SC-FDMA system. We use the
parameters in section 4.3 of chapter 4 for the simulations in this section.
As described in section 4.3 of chapter 4, we apply unitary precoding in the frequency
domain after DFT. Precoding in the frequency domain is convolution (filtering) and summation
in the time domain as shown in Figure 6.4. Thus we expect to see an increase in the peak power
because of the filtering and summation.
V11,k V12,k
S
, S k = 1,k
Vk =
V21,k V22,k
S 2,k
X k = Vk S k
V11,k V12,k S1,k
=
S
V
V
22, k 2, k
21,k
V11,k S1,k + V12,k S 2,k
S
+
V
S
22, k
2, k
21,k 1,k
Frequency domain
Vij = {Vij ,0 ,Vij ,1 ,,Vij , M 1}

Sl = {Sl ,0 , Sl ,1 ,, Sl , M 1}
vij = IDFT {Vij } , sl = IDFT {Sl }
v11 * s1 + v12 * s2
x=
v21 * s1 + v22 * s2
Time domain
Figure 6.4: Precoding in the frequency domain is convolution and summation in the time domain.k
refers to the subcarrier number.
81
10
-1
1st antenna
(16-QAM)
Precoded
2nd antenna
(QPSK)
Pr(PAPR>PAPR )
10
CCDF of PAPR for Unitary Precoded TxBF
10
-2
No precoding
Only summation
10
-3
2nd stream zero
10
-4
8
PAPR [dB]
10
12
Figure 6.5: CCDF of PAPR for 2x2 unitary precoded TxBF.

Figure 6.5 shows the CCDF of the PAPR for 2x2 unitary precoded TxBF. To verify that the
filtering aspect of the precoding is the dominant factor in increasing the PAPR, we also show
results for the case when there is only filtering by intentionally setting the 2nd antenna signal to
zero, that is, s2 = 0 (the graph labeled 2nd stream zero) and for the case when there is only
summation by intentionally replacing the convolution with a scalar multiplication of the first
element of vij (the graph labeled Only summation). With precoding, we can see that there is
an increase of 1.6 dB for the 1st antenna and 2.6 dB for the 2nd antenna in the 99.9-percentile
PAPR. We can also observe that the filtering aspect is the main contributor to the increase of
PAPR for precoding.
82
10
10
10
10
10
TxBF
(no avr. & no quant.)
No precoding
(QPSK)
-1
Pr(PAPR>PAPR )
10
-2
No precoding
(16-QAM)
-3
TxBF
(both avr. & quant.)
-4
TxBF
(only quant.)
TxBF
( only avr.)
-5
8
PAPR [dB]
10
12
Figure 6.6: Impact of quantization and averaging of the precoding matrix on PAPR.
10
-1
TxBF
(no avr. & no quant.)
SM
Pr(PAPR>PAPR )
10
10
10
10
-2
SFBC (QPSK)
SFBC (16-QAM)
-3
-4
TxBF
(avr. & quant.)
8
PAPR [dB]
0
Figure 6.7: PAPR comparison with other MIMO schemes.
10
12
83
Figure 6.6 illustrates the impact of quantization and averaging of the precoding matrix on
the PAPR characteristics. We use direct quantization and averaging process described in section
4.3.1 of chapter 4. We average the channel H over 25 continuous subcarriers and perform direct
quantization of the precoder matrix V using 3 bits (1 bit for amplitude and 2 bits for phase
information).
Without averaging of channel and quantization of precoding matrix, the MIMO TxBF
signal has 1.6 ~2.6 dB higher PAPR99.9% with respect to single antenna transmission. When we
apply both averaging and quantization, PAPR99.9% decreases by 0.7 dB. This is due to the fact
that the averaging and quantization of the precoding matrix smoothes the time filtering, thus
reduces the peak power.
Figure 6.7 compares the CCDF of the PAPR for unitary precoded TxBF with those of
different MIMO schemes. We consider space-frequency block coding (SFBC) and non-precoded
spatial multiplexing (SM) for comparison. Since the spatial multiplexing that we consider does
not have any precoding or spatial processing at the transmitter, it has the same PAPR as the case
of single antenna transmission. Without quantization/averaging of precoding matrix, TxBF has
higher PAPR99.9% than that of SFBC by 0.5 ~ 1 dB. However, TxBF and SFBC have similar
PAPR when we have quantization/averaging of the precoding matrix.
6.3.
Peak Power Reduction by Symbol Amplitude Clipping
One way to reduce PAPR is to limit or clip the peak power of the transmitted symbols [55].
Depending on the smoothness of the limiter, we can define three types of limiter; hard, soft, and
smooth limiter [56]. The input-out relationship of each type of amplitude limiter for a complex
84
symbol is as follows. Note that we leave the phase of the symbol as is and only modify the
amplitude part.
Hard limiter:
ym = g hard ( xm ) = Amax e jxm
(6.5)
Soft limiter:
xm , xm Amax
ym = g soft ( xm ) =
j xm
Amax e , xm > Amax
(6.6)
Smooth limiter:
ym = g smooth ( xm ) = Amax erf ( xm / Amax ) e jxm
(6.7)
where xm is the input symbol to the limiter, xm is the phase component of xm, ym is the
output symbol of the limiter, Amax > 0, and erf ( x ) =
e t dt . Figure 6.8 graphically

2
illustrates the input amplitude-output amplitude relationship of each limiter. In the subsequent
analysis, we use the soft limiter for clipping.
The problems associated with clipping are in-band signal distortion and generation of outof-band signal. Because SC-FDMA modulation spreads the information data across all the
modulated symbols, in-band signal distortion is mitigated when an SC-FDMA symbol is clipped.
To analyze the impact of symbol amplitude clipping on the link level performance, we apply
clipping to the unitary precoded TxBF MIMO system that we described in chapter 4 and show
numerical simulation results. We use the parameters and assumptions in section 4.3 of chapter 4
85
ym
Soft limiter
Hard limiter
Amax
Smooth limiter
Amax
xm
Figure 6.8: Three types of amplitude limiter.

for the simulations in this section.
Figure 6.9 shows the block diagram of a symbol amplitude clipping method for spatial
multiplexing SC-FDMA MIMO transmission for Mt transmit antenna. In our current analysis,
we apply the clipping after baseband pulse shaping. We can also consider other sophisticated
clipping methods, such as iterative clipping.
Figure 6.10 shows the CCDF of symbol power with clipping at various levels. With 7 dBmax clipping less than 1% of the symbols are clipped. Note that even with as much as 3 dB-max
PAPR clipping, only about 10% of the modulated symbols are clipped.
Figure 6.11 shows the uncoded bit error rate (BER) and coded frame error rate (FER)
performances when we apply symbol amplitude clipping. We can observe that the performance
degradation due to clipping is almost none for 7 dB-max clipping. The performance degrades
86
{x0,n ,, x0, N 1}
Sucarrier
Mapping
Add
CP/
PS
IDFT
Spatial Processing
DFT
{xM t 1,n ,, xM t 1, N 1}
Amplitude
Clipping
DAC/
RF
Channel
Sucarrier
Mapping
DFT
Add
CP/
PS
IDFT
Amplitude
Clipping
DAC/
RF
Figure 6.9: Block diagram of a symbol amplitude clipping method for SC-FDMA MIMO
transmission with Mt transmit antenna.
10
10
-1
No clipping
-2
Pr(|y |2>w)
10
CCDF of Symbol Power
10
10
-3
3 dB-max clipping
5 dB-max clipping
-4
7 dB-max clipping
0
10
w [dB]
Figure 6.10: CCDF of symbol power after clipping. ym represents the baseband symbol. Solid line
represents CCDF for antenna 1 and dashed line represents CCDF for antenna 2.
87
0
10
10
-1
10
-2
FER
Raw BER
10
-2
10
-4
10
No clipping
7 dB-max clipping
5 dB-max clipping
3 dB-max clipping
-3
10
-4
-6
10
No clipping
7 dB-max clipping
5 dB-max clipping
3 dB-max clipping
10
15
20
25
30
10
4
6
(a)
(b)
Figure 6.11: Link level performance for clipping; (a) uncoded BER; (b) coded FER.
0
3 dB-max
clipping
-10
Power [dB]
-20
-30
5 dB-max
clipping
7 dB-max
clipping
-40
-50
No clipping
-60
-4
-3
-2
-1
0
1
2
Normalized Frequency Band
Figure 6.12: PSD of the clipped signals.
10
88
slightly more when we apply 5 or 3 dB-max clipping. For 3 dB-max clipping, an error floor starts
to appear in high SNR region for uncoded BER. But when we incorporate forward error
correction coding, it mitigates this effect. Thus, we can use a modest amount of amplitude
clipping to reduce the PAPR for unitary precoded TxBF with SC-FDMA without compromising
the link level performance.
Clipping, consequently, will generate both in-band and out-of-band frequency components.
Figure 6.12 shows the power spectral density (PSD) of the clipped signals. For PSD calculation,
we use Hanning window with 1/4 of window overlapping. For 7 dB-max clipping, the spectrum
is almost the same as that of the original signal. More pronounced out-of-band components
arise when we use 5 or 3 dB-max clipping. Thus, we should control the amount of clipping
depending on the out-of-band radiation requirements.
6.4.
In this chapter, we analyzed the PAPR of SC-FDMA signals for single and multiple antenna
transmissions using numerical simulations. We showed that SC-FDMA signals indeed have lower
PAPR compared to OFDMA signals. Also, we have shown that DFDMA and LFDMA incur
higher PAPR compared to IFDMA but compared to OFDMA, it is still lower, though not
significantly. Another noticeable fact is that pulse shaping increases PAPR, thus degrades the
power efficiency and that the roll-off factor in the case of raised-cosine pulse shaping has a
significant impact on PAPR of IFDMA. A pulse shaping filter should be designed carefully in
order to reduce the PAPR without degrading the system performance.
For multiple antenna transmission, we observed that unitary precoding TxBF scheme
89
increases the PAPR because of its filtering aspect. It was interesting to find out that averaging
and quantization of the precoder matrix actually reduces the PAPR.
In order to reduce the peak power in a SC-FDMA system, we proposed a symbol amplitude
clipping method and applied it to the 2x2 unitary precoded TxBF system. Numerical simulation
results showed that moderate clipping does not affect the link level performance. But out-ofband radiation is a concern when we use clipping.
Chapter 7
ChannelChannel-Dependent Scheduling of Uplink SCSC-FDMA

Systems
As we illustrated in section 2.1 of chapter 2, wide band wireless channels experience time and
frequency-selective fading because of users mobility and multi-path propagation. In wide
band multi-user uplink communications, the channel gain of each user is different for
different time and frequency subcarrier when the channels are uncorrelated among users.
Time and frequency resources which are in deep fading for one user may be in excellent
conditions for other users. The resource scheduler in the base station can assign the timefrequency resources to a favorable user which will increase the total system throughput [57],
[58], [59]. We term the class of this adaptive resource scheduling method as channeldependent scheduling (CDS) which can greatly increase the spectral efficiency (bits/Hz).
Another way to exploit the time and frequency-selectivity in the channel to its advantage
is the adaptive modulation and coding (AMC) scheme along with subband or multicarrier
communications. AMC dynamically adapts the modulation constellation and the channel
coding rate depending on the channel condition and thus improves the energy efficiency and
the data rate [60], [61]. By applying AMC on subbands or subcarriers, we can dramatically
90
91
improve the link quality and increase the data rate for time and frequency-selective fading
channel.
In this chapter, we first give a general overview of CDS in an uplink SC-FDMA system.
We also analyze the capacity for the distributed (IFDMA) and localized subcarrier mapping
schemes. In section 7.2, we investigate the impact of imperfect channel state information
(CSI) on CDS. We analyze the data throughput of an SC-FDMA system with uncoded
adaptive modulation and CDS when there is a feedback delay. In section 7.3, we propose a
hybrid subcarrier mapping scheme using direct sequence spreading technique on top of SCFDMA modulation. We show the throughput improvement of the hybrid subcarrier mapping
over the conventional subcarrier mapping schemes.
7.1.
Channel-Dependent Scheduling in an Uplink SC-FDMA System
In this section, we investigate channel-dependent resource scheduling for an SC-FDMA

system in uplink communications. Specifically, we analyze the capacity gains from CDS for the
two different flavors of subcarrier mapping of SC-FDMA. For distributed subcarrier mapping,
we consider IFDMA. A key question of CDS is how we should allocate time and frequency
resources fairly among users while achieving multi-user diversity and frequency selective
diversity. To do so, we introduce utility-based scheduling where utility is an economic concept
representing level of satisfaction [62], [63], [64], [65], [66]. The choice of a utility measure
influences the tradeoff between overall efficiency and fairness among users. In our studies, we
consider two different utility functions: aggregate user throughput for maximizing system
capacity and aggregate logarithmic user throughput for maximizing proportional fairness. The
92
objective is to find an optimum resource assignment for all users in order to maximize the sum
of user utility at each transmission time interval (TTI). If the user throughput is regarded as
the utility function, the resource allocation maximizes rate-sum capacity ignoring fairness
among users. Therefore, only the users near the base station who have the best channel
conditions occupy most of the resources. On the other hand, setting the logarithmic user data
rate as the utility function provides proportional fairness.
In considering the optimization problem of CDS for multicarrier multiple access, it is
theoretically possible to assume that the scheduler can assign subcarriers individually.
Allocating individual subcarriers is, however, a prohibitively complex combinatorial
optimization problem in systems with a large number of subcarriers and user terminals.
Moreover, assigning subcarriers individually would introduce unacceptable control signaling
overhead. In practice, we allocate the resource in chunks, which are disjoint sets of subcarriers.
We refer to this chunk as a resource unit (RU). As a practical matter for SC-FDMA, chunkbased transmission is desirable since the input data symbols are grouped into a block for DFT
operation before subcarrier mapping. We will consider only chunk-based scheduling in the
remainder of this section. With regards to chunk structure, consecutive subcarriers comprise a
chunk for LFDMA and distributed subcarriers with equidistance for IFDMA.
Even with subcarriers assigned in chunks, optimum scheduling is extremely complex for
two reasons: 1) The objective function is complicated, consisting of nonlinear and discrete
constraints dependent on the combined channel gains of the assigned subcarriers; and 2) there
is a total transmit power constraint for each user. Furthermore, the optimum solution entails
93
combinatorial comparisons with high complexity. Instead of directly solving the optimization
problem, we can use a sub-optimal chunk allocation scheme for both IFDMA and LFDMA to
obtain most of the benefits of CDS. For LFDMA, we can apply a greedy chunk selection
method where each chunk is assigned to the user who can maximize the marginal utility when
occupying the specific chunk. For IFDMA, we can achieve the benefit of multi-user diversity
by selecting users in order of the estimated marginal utility based on the average channel
condition over the entire set of subcarriers. The users with higher channel gains may occupy a
larger number of chunks than users with lower channel gains. The details of the chunk
selection method are in [64] and [65].
Our throughput measure in this study is the sum of the upper bound on user
throughputs given by Shannons formula, C = BWlog(1+SNR) where BW is the effective
bandwidth depending on the number of occupied subcarriers and SNR is signal-to-noise ratio
of a block.
Figures 7.1 and 7.2 are the results of computer simulations of SC-FDMA with 256
subcarriers spread over a 5 MHz band. They compare the effects of channel dependent
subcarrier allocation (S) with static (round-robin) scheduling (R) for LFDMA and IFDMA. In
all of the examples, the scheduling took place with chunks containing 8 subcarriers and we
assume perfect knowledge of the channel state information.
Figure 7.1 shows two effects of applying different utility functions in the scheduling
algorithm. In the two graphs in Figure 7.1, the utility functions are the sum of user
throughputs and the sum of the logarithm of user throughputs, respectively. Each graph
94
45
45
35
40
Aggregate throughput [Mbps]
40
R-LFDMA
S-LFDMA
R-IFDMA
S-IFDMA
30
25
20
15
30
25
20
15
10
10
5
4 8 16
35
R-LFDMA
S-LFDMA
R-IFDMA
S-IFDMA
32
64
Number of users
(a)
128
5
4 8 16
32
64
Number of users
128
(b)
Figure 7.1: Comparison of aggregate throughput with M = 256 system subcarriers, N = 8 subcarriers
per user, bandwidth = 5 MHz, and noise power per Hz = -160 dBm; (a) utility function: sum of
user throughputs; (b) utility function: sum of the logarithm of user throughputs.
shows the aggregate throughput as a function of the number of simultaneous transmissions.
Figure 7.2 shows the expected user throughput at each distance from the base station for the
same utility functions. The simulation results in the figures use the following abbreviations: RLFDMA (static round robin scheduling of LFDMA), S-LFDMA (CDS of LFDMA), RIFDMA (static round robin scheduling of IFDMA), and S-IFDMA (CDS of IFDMA).
Figure 7.1 shows that for throughput maximization (utility = bit rate), the advantage of
channel dependent scheduling over round robin scheduling increases as the number of users
increases. This is because the scheduler selects the closer users who can transmit at higher data
rate. If there are more users, the possibility of locating users at closer distance to the base
station increases. As a result, the CDS achieves significant improvements for both IFDMA
95
5
R-LFDMA
S-LFDMA
R-IFDMA
S-IFDMA
Average user data rate [Mbps]
Average user data rate [Mbps]
0
0.2
0.4
0.6
0.8
User distance from target BS [km]
(a)
R-LFDMA
S-LFDMA
R-IFDMA
S-IFDMA
0
0.2
0.4
0.6
0.8
User distance from target BS [Km]
(b)
Figure 7.2: Average user data rate as a function of user distance with M = 256 system subcarriers, N
= 8 subcarriers per user, bandwidth = 5 MHz, and noise power per Hz = -160 dBm; (a) utility
function: sum of user throughputs; (b) utility function: sum of the logarithm of user throughputs.
and LFDMA. In the case of logarithmic rate utility, the CDS gain stops increasing beyond
approximately 32 users. With 32 users, maximizing logarithmic rate utility can increase system
capacity by a factor of 1.8 for LFDMA and 1.26 for IFDMA relative to static scheduling.
Figure 7.2 shows that the CDS scheme based on the logarithm of user throughput as a
utility function provides proportional fairness whose gains are shared among all users, whereas
the CDS gains are concentrated to the users near the base station when the user throughput is
considered as the utility function.
For static subcarrier assignment (round robin scheduling), a system with users each
transmitting at a moderate data rate is better off with IFDMA while LFDMA works better in
a system with a few high date rate users. Due to the advantages of lower outage probability
96
and lower PAPR, IFDMA is an attractive approach to static subcarrier scheduling. For CDS,
LFDMA has the potential to achieve considerably higher data rate. The results show that CDS
increases system throughput by up to 80% relative to static scheduling for LFDMA but the
increase is only 26% for IFDMA. We can exploit the scheduling gains in LFDMA to reduce
power consumption and PAPR by using power control to establish a power margin instead of
increasing system capacity.
7.2.
Impact of Imperfect Channel State Information on CDS
In section 7.1, we assumed perfect knowledge of the channel state information (CSI). In a
practical wireless system, CSI is often not known at all or only part of the information is
available due to limited feedback mechanism and channel estimation error. Also, the CSI may
become outdated because of the feedback delay. In a centralized resource management
scheme, the base station makes all the decision regarding the uplink transmission parameters
based on the uplink channel quality and feeds back the transmission parameter information to
the mobile terminal. A delay in the feedback mechanism is inevitable in a practical feedback
system and the feedback delay has a major impact on the system performance especially when
the channel is fading rapidly.
In this section, we investigate the impact of imperfect CSI on CDS in uplink
communications by considering outdated CSI due to feedback delay. We assume the channel
estimation is perfect and also assume there is no error during the feedback signaling. Since
there is no general closed form for the capacity equation for outdated CSI, we resort to
numerical simulations in the context of an SC-FDMA system with uncoded adaptive
97
Mobile terminals
Base station
User K
Channel K
User 2
DFT
Subcarrier
Mapping
Constellat ion
Mapping
User 1
Channel 2
IDFT
CP /
PS
Channel 1
SC-FDMA Receiver
Resource Scheduler
Data flow
Control signal flow
Figure 7.3: Block diagram of an uplink SC-FDMA system with adaptive modulation and CDS for
K users.
modulation to measure the data throughput.

Figure 7.3 describes the system under investigation. We consider K number of users
orthogonally and independently sending data to the base station and we assume that the
channel responses are independent of each other. The resource scheduler in the base station
performs the following two tasks upon acquiring the channel information for all users. First,
the scheduler searches for an optimal set of users that maximizes the total capacity. Next, for
each user, it decides the modulation constellation based on the SNR of the chunk allocated to
the user. The scheduler, then, sends the chunk allocation and modulation information back to
98
each user via the feedback control signaling channel. When there is a delay in the feedback
channel, it creates two sources of performance degradation. First, the chunk allocation
becomes outdated by the time the users transmit and it no longer is the optimal allocation.
Second, the modulation constellation for each user no longer matches the channel condition at
the time of transmission. Because of these two factors, we can expect significant decrease in
the total capacity for time-varying channels.
In our analysis, we introduce a time delay between the time instance of the channel
estimation (t1) and the actual time of data transmission from the mobile terminals (t2). We
consider a quasi-static multipath fading channel in which channel response is constant over a
TTI but varies between each TTI depending on a correlation parameter. For a Rayleigh fading
channel, we represent the correlation coefficient for time delay t = t2 - t1 as
(t ) = J 0 (2 vt / ) = J 0 (2 f D t )
(7.1)
where J 0 (i) is the zero-order Bessel function, fD = v/ is the maximum Doppler frequency, v
is the velocity of the mobile terminal, and is the wavelength of the carrier signal [11].
We model the multipath fading channel at t1 with an R-path discrete time channel impulse
response, h (t1 ) ( n) , as follows.
R 1

h (t1 ) ( n) = A Prel ( r ) wr(t1 ) n r
Ts
r =0
(7.2)
where { wr(t1 ) }s are zero mean complex Gaussian i.i.d. noise process, r is the propagation
99
delay of path r, Ts is the symbol duration, A is a normalization parameter such that the average
power of the channel equals to 1, Prel(r) is the relative power of the delay profile model for
path r, and (n) is the discrete time Dirac-delta function. Note that {r}s are integer-multiples
of Ts.
Using (7.1), we can generate the multipath fading channel at t2 , h (t2 ) (n) , as follows.
R 1
r =0
Ts
h (t2 ) ( n) = A Prel ( r ) wr(t2 ) (n
(7.3)
and
wr(t2 ) = ( t ) wr(t1 ) + 1 2 ( t ) nr
(7.4)
where nr is a zero mean complex Gaussian noise process.

We will use (7.2) and (7.3) to analyze the effect of the outdated channels on our system.
In [64], we derived the SNR of user k, k, as follows based on [67] for an MMSE equalizer
when a set of subcarriers Isub,k is assigned to the user.
k =
1
1
1
i ,k
iI sub ,k i , k + 1
(7.5)
where N is the number of subcarriers allocated to the user and i ,k is the SNR of subcarrier
i for user k. For equal power allocation scheme, we can express i ,k as
100
i ,k =
( Pk / N ) H i ,k
i2
(7.6)
where Pk is the total transmit power of user k, Hi,k is the channel gain of subcarrier i for user
k, and i2 is the noise power of subcarrier i.
In our simulations, we apply the effect of the time delay during feedback in the following
manner. We first calculate the received SNRs for the subcarrier allocations and select the
modulation constellations based on the channels at time t1 and then calculate the received
SNRs for the throughput calculations based on the channels at time t2.
For the numerical simulations, we consider the following setup and assumptions.
Carrier frequency: 2 GHz.
Transmission bandwidth: 5 MHz.
Duration of a TTI: 0.5 ms.
Blocks per TTI: 7. We assume there is no CP in a block for simplicity of the

simulation.
The size of the subcarrier chunk is the same among all users and each user
occupies only one subcarrier chunk.
The number of users is K, the total number of subcarriers M, the size of a chunk
N, the number of chunks Q, and M = QN.
Channel: 3GPP TU6 i.i.d. quasi-static Rayleigh fading channel [12]. We assume
that the channel is constant during the duration of a TTI. We also assume that all
the users have the same path loss to the base station.
101
Modulation format: Quadrature amplitude modulation (QAM). We consider SQAM where S = 2B and B is the number of bits per symbol. We have 8 classes of
QAM where B {1, 2, , 8}. B = 1 corresponds to BPSK and B = 2 to QPSK.
We assume equal transmit power and equal receive noise power for all users.
We use aggregate user throughput as the utility function of CDS.
We define the system throughput as the number of information bits per second received
without error. Let user k be assigned Sk-QAM where S k = 2 Bk and Bk is the number of bits
per symbol. Then we express the system throughput for user k as follows [68], [69].
TPk =
7 N Bk
PSR ( S k , k )
TTTI
(7.7)
where TTTI is the duration of a TTI, k is the received SNR for user k, and PSR ( i ) is the
packet success rate (PSR) which we define as the probability of receiving an information
packet correctly. The aggregate throughput TPtotal is the sum of the K individual throughput
measures in (7.7).
For an uncoded packet with a size of L bits, PSR becomes
PSR ( S , ) = {1 Pe ( S , )}
(7.8)
where Pe ( S , ) is the BER for a given constellation size S and SNR .

For an additive white Gaussian noise (AWGN) channel with S-QAM modulation and
ideal coherent detection, we can upper-bound Pe ( S , ) as follows [70].
102
Pe ( S , ) 0.2e
1.5
( S 1)
(7.9)
Using this upper bound, PSR in (7.8) becomes

1.5
( S 1)
PSR ( S , ) = 1 0.2e
(7.10)
and the system throughput for user k becomes

7 N Bk
TPk =
TTTI
1.5 k
1 0.2e ( Sk 1)
(7.11)
Then, the total system throughput for K number of users is

K
TPtotal = TPk
(7.12)
k =1
Figure 7.4 shows the system throughput vs. SNR for the 8 classes of QAM. We use L =
100 bits per packet and N = 32 subcarriers per chunk. Based on Figure 7.4, we set the SNR
boundaries for the adaptive modulation. Table 7.1 describes the SNR boundaries for our
analysis.
103
4
B=8
3.5
B=7
Throughput [Mbps]
3
B=6
2.5
B=5
2
B=4
1.5
B=3
1
B=2
0.5
0
B=1
10
15
20
25
SNR [dB]
30
35
40
Figure 7.4: System throughput vs. SNR for the 8 classes of QAM. B is the number of bits per
symbol.
Table 7.1: SNR boundaries for adaptive modulation.

SNR (dB)
Bits per symbol (B)
Modulation
> 8.7
BPSK
8.7 ~ 13.0
QPSK
13.0 ~ 16.7
8-QAM
16.7 ~ 20.0
16-QAM
20.0 ~ 23.3
32-QAM
23.3 ~ 26.6
64-QAM
26.6 ~ 29.6
128-QAM
29.6 <
256-QAM
104
mobile speed = 3 km/h (f = 5.6 Hz)
mobile speed = 60 km/h (f = 111 Hz)
18
14
12
LFDMA: Static
LFDMA: CDS
IFDMA: Static
IFDMA: CDS
16
10
8
6
4
2
0
2
3
Feedback delay [ms]
LFDMA: Static
LFDMA: CDS
IFDMA: Static
IFDMA: CDS
16
14
12
10
8
6
4
2
0
2
3
Feedback delay [ms]
(a)
(b)
Figure 7.5: Aggregate throughput with CDS and adaptive modulation; (a) mobile speed = 3 km/h
(Doppler = 5.6 Hz); (b) mobile speed = 60 km/h (Doppler = 111 Hz).
mobile speed = 60 km/h (f = 111 Hz)
D
18
LFDMA: Static
LFDMA: CDS
IFDMA: Static
IFDMA: CDS
16
18
14
12
10
8
6
4
2
0
0
2
3
Feedback delay [ms]
Figure 7.6: Aggregate throughput with CDS and constant modulation (16-QAM) with mobile
speed of 60 km/h.
105
Feedback delay = 3 ms
18
LFDMA: Static
LFDMA: CDS
IFDMA: Static
IFDMA: CDS
16
14
12
10
8
6
4
2
0
20 (37)
40 (74)
60 (111)
Mobile speed [km/h] (Doppler [Hz])
80 (148)
Figure 7.7: Aggregate throughput with CDS and adaptive modulation with feedback delay of 3 ms
and different mobile speeds.
Figures 7.5 and 7.6 show the results of the Monte Carlo simulations of the investigated
system. Total number of subcarriers is M = 256, the number of chunks is Q = 16, the number
of subcarriers per chunk is N = 16, and total number of users is K = 64. We perform 2000
independent iterations of the simulation.
Figure 7.5 shows the aggregate throughputs with CDS and adaptive modulation for
different mobile speeds. For low mobile speed, we can see that the feedback delay does not
affect the system throughput. However, at high mobile speed, the throughput degrades as the
feedback delay increases and the degradation is most significant for LFDMA with CDS. This
implies that LFDMA is very sensitive to the quality of CSI when we apply CDS for fast fading
106
channel. Another interesting observation is that the static round-robin scheduling also suffers
from throughput decrease. This is due to the fact that even though there is no chunk
allocation mismatch, outdated CSI causes the adaptive modulation to be mismatched with the
current channel condition.
Figure 7.6 shows the aggregate throughput with CDS for mobile speed of 60 km/h and
constant modulation (16-QAM) instead of adaptive modulation. It further verifies the
observation in Figure 7.5 (b) for static scheduling that the cause of the throughput decrease is
the adaptive modulation process. We do not see any throughput decrease for the static
scheduling when we use constant modulation in the same channel condition.
Figure 7.7 shows the aggregate throughput with CDS and adaptive modulation with
feedback delay of 3 ms and different mobile speeds. We can see the performance impact of
the mobility of the user on each type of scheduling method. Similarly with Figure 7.5 (b),
LFDMA with CDS suffers the most throughput decrease when the mobile speed is high.
7.3.
Hybrid Subcarrier Mapping
As we can see from Figure 7.7, localized subcarrier mapping yields much higher throughput
for low speed users when we use CDS but the throughput of LFDMA with CDS becomes
lower when the users are moving at high speed. In this situation, static round-robin scheduling
using distributed subcarrier mapping is advantageous since it requires no overhead and no
computation. Also, IFDMA has an additional advantage in PAPR. Thus, localized subcarrier
mapping with CDS is more advantageous for low mobility users and IFDMA with static
scheduling is better for users with high mobility. When there are low mobility users and high
107
mobility users at the same, accommodating both types of subcarrier mapping can lead to
higher data capacity.
Conventional SC-FDMA cannot efficiently accommodate both distributed and localized
mapping schemes since subcarriers must not overlap to maintain orthogonality among users.
Using orthogonal direct sequence spread spectrum technique prior to SC-FDMA modulation,
both mapping can coexist with overlapping subcarriers as illustrated in Figure 7.8.
Terminal 1
Terminal 2
Terminal 3
Spreading Code 1
subcarriers
Spreading Code 2
Conventional
subcarriers
Hybrid
Figure 7.8: Conventional subcarrier mapping and hybrid subcarrier mapping.

We refer to this multiple access scheme as single carrier code-frequency division multiple
access (SC-CFDMA). Figure 7.9 shows the block diagram of an SC-CFDMA system and
Figure 7.10 compares SC-FDMA and SC-CFDMA in terms of occupied subcarriers for the
same number of users.
108
Npoint
DFT
Spreading
Subcarrier
Mapping
Mpoint
IDFT
SC-FDMA Modulation
Add
CP/
PS
Channel
Remove
CP
SC-FDMA Demodulation
Mpoint
DFT
Subcarrier Demapping/
Equalization
Despreading
Detect
Npoint
IDFT
Figure 7.9: Block diagram of an SC-CFDMA system.
User 1
User 1
Code 1
User 2
User 3
User 4
Code 2
subcarriers
SC-FDMA
Q (# of users) = 4
User 2
User 3
User 4
subcarriers
SC-CFDMA
Q (# of users) = QcodeQfrequency= 22 = 4
Figure 7.10: Comparison between SC-FDMA and SC-CFDMA in terms of occupied subcarriers
for the same number of users.
109
16
IFDMA: static
Hybrid
LFDMA: CDS
14
12
10
8
6
4
2
0
25/75
50/50
75/25
Low/high mobility users percentile [%/%]
Figure 7.11: Aggregate throughputs for hybrid subcarrier mapping method and other conventional
subcarrier mapping methods with CDS and adaptive modulation.
Figure 7.11 compares the aggregate throughput between hybrid subcarrier mapping
method and conventional subcarrier mapping methods when we apply CDS and adaptive
modulation. In hybrid subcarrier mapping method, we apply LFDMA with CDS to low
mobility users and IFDMA with static scheduling to high mobility users. Total number of
subcarriers is M = 256, the number of chunks is Q = 16, the number of subcarriers per chunk
is N = 16, the total number of users is K = 64, and the feedback delay is 3 ms. Low mobility
users have velocity of 3 km/h and high mobility users have velocity of 60 km/h. In the figure,
the label X/Y means that X-percentile of the users have low mobility and Y-percentile of the
users have high mobility. Clearly, when high mobility users are the majority of the entire users
(the graphs labeled 25/75 and 50/50), hybrid subcarrier mapping scheme yields higher data
throughput than the conventional subcarrier mapping schemes.
110
7.4.
In this chapter, we investigated channel-dependent scheduling (CDS) for an uplink SC-FDMA

system. Specifically, we analyzed the capacity gains from CDS for the two different flavors of
subcarrier mapping of SC-FDMA. We also looked into the impact of imperfect channel state
information (CSI) on CDS by considering the feedback delay.
We showed that localized subcarrier mapping yields highest aggregate data throughput
when we use CDS. However, we also showed that LFDMA is very sensitive to the quality of
CSI and the capacity gain quickly decreases when the channel changes very fast. For high
mobility users, IFDMA with static round-robin scheduling is more suitable.
To accommodate both low and high mobility users simultaneously, we proposed a hybrid
subcarrier mapping method using orthogonal code spreading on top of SC-FDMA. We
showed that the hybrid subcarrier mapping scheme has capacity gains over conventional
subcarrier mapping schemes when there are both low and high mobility users at the same time.
Chapter 8
Conclusions
Single carrier FDMA (SC-FDMA) is a new multiple access scheme that is currently being
adopted in uplink of 3GPP LTE. SC-FDMA utilizes single carrier modulation and frequency
domain equalization, and has similar performance and essentially the same overall complexity as
those of OFDMA system. A salient advantage over OFDMA is that the SC-FDMA signal has
lower PAPR because of its inherent single carrier transmission structure. SC-FDMA has drawn
great attention as an attractive alternative to OFDMA, especially in the uplink communications
where lower PAPR greatly benefits the mobile terminal in terms of transmit power efficiency
and manufacturing cost.
In this thesis, we gave a detailed overview of an SC-FDMA system and compared it with
the OFDMA system and also with the direct sequence CDMA (DS-CDMA) with frequency
domain equalization (FDE) system. We also explained the different subcarrier mapping schemes
in SC-FDMA; distributed and localized. The two flavors of subcarrier mapping schemes give the
system designer a flexibility to adapt to the specific needs and requirements.
Along with the introduction of SC-FDMA, we investigated the followings.
Realization of MIMO spatial multiplexing in SC-FDMA with practical limitations: We
111
112
specifically analyzed the unitary precoded transmit eigen-beamforming (TxBF)

technique on an SC-FDMA system. We showed with numerical simulations that there is
a trade-off between the feedback overhead and the link performance loss. We also
showed that feedback delay is another impairment that degrades the performance
especially for the high speed mobile users.
Analysis of the peak power characteristics of SC-FDMA signals: By using Chernoff bound and
using Monte Carlo simulations, we characterized the peak power of an SC-FDMA
signal both analytically and numerically. We saw that the peak power of an SC-FDMA
signal for any subcarrier mapping is indeed lower than that of OFDM and observed
that IFDMA has the lowest peak power among the considered subcarrier mapping
schemes of SC-FDMA. For IFDMA, the roll-off factor of the raised-cosine pulse
shaping filter has a great impact on the distribution of the instantaneous power. For
LFDMA, the input block size is the major factor characterizing the distribution of peak
power. For multiple antenna transmission, we observed that unitary precoded TxBF
scheme increases the peak power because of its filtering aspect. Interestingly, averaging
and quantization of the precoder matrix actually reduces the peak power.
Peak power reduction by symbol amplitude clipping: In order to reduce the peak power in an
SC-FDMA system, we proposed a symbol amplitude clipping method and applied it to
the 2x2 unitary precoded TxBF system. Numerical simulation results showed that
moderate clipping does not affect the link level performance. But out-of-band radiation
is a concern when we use clipping.
113
Channel-dependent resource scheduling for SC-FDMA: We investigated channel-dependent

scheduling (CDS) for an uplink SC-FDMA system. Specifically, we analyzed the capacity
gains from CDS for the two different flavors of subcarrier mapping of SC-FDMA. We
also looked into the impact of imperfect channel state information (CSI) on CDS by
considering the feedback delay. We showed that localized subcarrier mapping yields
highest aggregate data throughput when we use CDS. However, we also showed that
LFDMA is very sensitive to the quality of CSI and the capacity gain quickly decreases
when the channel changes very fast. For high mobility users, IFDMA with static roundrobin scheduling is more suitable since it does not require any computation and the
PAPR is low. To accommodate both low and high mobility users simultaneously, we
proposed a hybrid subcarrier mapping method using orthogonal code spreading on top
of SC-FDMA. We showed that the hybrid subcarrier mapping scheme has capacity
gains over conventional subcarrier mapping schemes when there are both low and high
mobility users at the same time.
For future work, we list the following items.
Impact of carrier frequency offset on the system performance: In this thesis, we did not
consider carrier frequency offset which is a common impairment at the receiver. Its
impact on the link level performance and capacity is an important issue for a practical
implementation.
Peak power characteristics of SC-FDMA pilot reference signals: We can apply the same
numerical analysis to the pilot signals proposed by 3GPP LTE and characterize the peak
114
power characteristics.
More sophisticated clipping methods, such as iterative clipping, for peak power
reduction: We showed that clipping is a very effective way to reduce peak power without
compromising the link performance. Yet, out-of-band radiation is a problem for
clipping. In order to mitigate the out-of-band radiation associated with clipping, we can
apply clipping before and after the pulse shaping and iterate the process until we reach a
satisfactory level of performance.
Impact of channel estimation error on unitary precoded TxBF MIMO SC-FDMA

system: In our analysis, we assumed that the channel estimation is without error. In real
systems, channel estimation error is unavoidable and this will have a significant impact
on our unitary precoded TxBF MIMO SC-FDMA system.
Impact of channel estimation error on CDS considering error correction coding:

Channel estimation error adds to the performance degradation in adaptive modulation
scheme and channel-dependent resource scheduling scheme. We can analyze its impact
on capacity using our adaptive modulation/CDS system.
Appendix A
Derivations of the Upper Bounds

Bounds in Chapter 5
A.1
Derivation of (5.8) and (5.9)
We can expand E {Ze Z } as follows.

E {Ze Z } = E ak exp al = E ak e al
l = k = l =
k =
ak al
= E ak e e
k =
l =
l k
k =
E ak e ak
(A.1)
} E {e }
al
l =
l k
We can also expand E {e Z } similarly as follows.
E e Z = E exp al = E e al
l =
l =
{ }
{ }
=E e
l =
al
By applying (A.1) and (A.2) to (5.7) for = , we get the following.
115
(A.2)
116
k =
E ak eak
} E {e } E {e } = 0
al
l =
lk
al
l =
E {a e } E {e } = E {e }
ak
k =
al
l =
l k
l =
By dividing each side with
(A.3)
al
E {e } , we obtain
al
l =
k =
{ } =
E {e }
E ak eak
(A.4)
ak
Using (A.2), (5.6) becomes

Pr {Z } e
A.2
E {e }
ak
(A.5)
k =
Derivations of (5.12) and (5.13)
From (5.11),
sk CBPSK
ak { p (t0 kT ), p (t0 kT )}
(A.6)
Also,
P {sk = 1} = P {sk = 1} =
1
2
1
P {ak = p (t0 kT )} = P {ak = p (t0 kT )} =
2
Using (A.6) and (A.7), (5.8) becomes
(A.7)
117
K max
k = K max
1
1
p (t0 kT )e p (t0 kT ) p (t0 kT )e p ( t0 kT )

2
2
=
1 p (t0 kT ) 1 p (t0 kT )
e
+ e
2
2
(A.8)
After we determine from (A.8), the upper bound in (5.9) becomes

)
Pub( BPSK
2e
, SC
= 2e
= 2e
=e
E {e }
K max
ak
k = K max
K max
1 p (t0 kT ) 1 p (t0 kT )
+ e
e
k = K max 2
1

2
1

2
2 K max +1
2 K max
(e
K max
p ( t0 kT )
+e
p ( t0 kT )
k = K max
( e
K max
p ( t0 kT )
k = K max
+ e p (t0 kT )
(A.9)
and we upper-bound the CCDF as

2
)
Pr x ( BPSK ) (t0 , s ) w Pub( BPSK
, SC
A.3
(A.10)
Modified Upper Bound Derivation for QPSK
For complex modulation constellations, we can express (5.4) as follows from the probability
regions in Figure A.1.
2
2
Pr x(t0 , s ) w = Pr Z w
w
w
2
2
Pr Z re + Pr Z im
2
2
= Pr Z re + Pr Z im
where
(A.11)
118
Im {Z }
Im {Z } = w
w
w
Im {Z } = w
Z =w
w
w
Re {Z }
2
w
Re {Z } = w
Re {Z } = w
Figure A.1: Probablity regions for complex modulation constellations.

Z re = Re {Z } =
K max
Re {ak } =
K max
Im {ak } =
k = K max
K max
Re {sk } p (t0 kT )
(A.12)
K max
Im {sk } p (t0 kT )
(A.13)
k = K max
,
Z im = Im {Z } =
k = K max
k = K max
, and = w . Since (A.11) is not a tightly-bounded inequality, we can instead use

2
= w to produce a tighter bound by adjusting >

Similarly with (5.5),
1
empirically.
2
119
Pr Z re = 2 Pr [ Z re ]
(A.14)
Pr Z im = 2 Pr [ Z im ]
1
1
We consider a QPSK constellation sk such that Re {sk } , Im {sk }
,
and
2
2
Re {sk } and Im {sk } are independent and uniformly distributed. Then,

Pr [ Z re ] = Pr [ Z im ]
(A.15)
Thus (A.11) becomes

2
Pr x(t0 , s ) w Pr Z re + Pr Z im
= 2 {Pr [ Z re ] + Pr [ Z im ]}
(A.16)
= 4 Pr [ Z re ]
From (A.12),
Pr {Z re } e
K max
Re{ak }
E e
k = K max
(A.17)
where is the solution of the following equation.
K max
E Re {ak } e
k = K max
E e
Re{ak }
Re{ak }
} =
(A.18)
2
Then, we can upper-bound the CCDF Pr x(t0 , s ) w as follows.
K max
2
Re a
)
Pr x (t0 , s ) w 4e E e { k } Pub(QPSK
, SC
k = K
max
(A.19)
120
where = w .
)
Similarly with the case of BPSK in section 5.1.1, we can express Pub(QPSK
as follows.
, SC
( BPSK )
ub , SC
A.4
=e
1

2
2 K max 1
2 p ( t0 kT ) 2 p (t0 kT )
+e
e
k = K max
K max
(A.20)
Modified Upper Bound for LFDMA and DFDMA
In this section, we derive the upper bound of the CCDF for LFDMA symbols which will also
apply to the case of DFDMA.
In chapter 3 section 3.3.2, we derived the time domain symbols after LFDMA modulation
without pulse shaping as follows. We follow the symbol notations in section 3.3 of chapter 3.
xm = xQn + q
1
Q x( m )mod N , q = 0
q
=
j 2 1 N 1
xp
1
1 e Q
( n p ) q
j 2
+
Q
N p=0
1 e N Q N
where m = Q n + q , 0 n N 1 , and 0 q Q 1 .
We can express Q xm as follows when q 0 .
, q0
(A.21)
121
q
j 2 1
Q
Q xm = 1 e
=e
=e
=e
q
j
Q
q
j
Q
( N 1) q Qn
QN
( N 1) q Qn
j
QN
N 1
xp
p=0
q
j
Q
1 e
(n p) q
j 2
+
Q N
N
(n p) q
j
+
Q N
N
N 1
p =0
(n p) q
j
+
Q N
N
q 1
(2 j ) sin
Q N
N 1
p =0
xp
(n p ) q
j
+
Q N
N
p
N
xp
(Qn + q )
p
(2 j ) sin

QN
N
q
sin

N 1
j Np
Q
e xp
(Qn + q )
p
p =0
N sin

QN
N

c ( n ,q , p )
=e
( N 1) q Qn
QN
N 1
c(n, q, p ) x p
p =0

(A.22)
)
Z n( LFDMA
,q
q
where c(n, q, p ) = sin
Q
(Qn + q )
p
, x p = exp ( j p N ) x p , and
N sin
QN
N
N 1
)
Z n( ,LFDMA
= p = 0 c(n, q, p ) x p .
q
Our objective now is to characterize the following CCDF of an LFDMA signal.
122
1
M
=
M 1
Pr Q x
m=0
1
M
1
=
M
=
1
M
N 1 Q 1
Pr Qx
Qn + q
n = 0 q =0
Q 1

2
+ Pr Z ( LFDMA) 2 w
Pr
Qx
w
Qn
n,q
n=0
q =1
N 1
(A.23)
Q 1
2
) 2
Pr
Pr Z n( ,LFDMA
x
w
+
w
n
q =1
n=0
N 1
2
Let us assume that xn has constant amplitude with unit power. Then, Pr xn w
becomes zero when w > 1 since xn = 1 . In such case, (A.23) becomes

2
1
M
M 1
1
2
Pr Q xm w =
M
m=0
N 1 Q 1
Pr Z
n = 0 q =1
( LFDMA ) 2
n,q
(A.24)
) 2
Similarly with (A.11), we bound Pr Z n( ,LFDMA
w as follows.
q
) 2
Pr Z n( ,LFDMA
w
q
)
Pr Re Z n( ,LFDMA
q
= Pr Re Z
( LFDMA)
n,q
)
+ Pr Im Z n( ,LFDMA
q
} + Pr Im {Z
} + Pr Im {Z
)
= 2 Pr Re Z n( ,LFDMA
q
( LFDMA)
n,q
w
2
}
} )
(A.25)
( LFDMA)
n ,q
1
where = w . Or we can use = w to produce a tighter bound by adjusting >
2
2
empirically.
Using the Chernoff bound,
123
{
E {e
)
re Re Z n( LFDMA
,q
)
Pr Re Z n( ,LFDMA
ere E e
q
Pr Im Z
( LFDMA )
n ,q
} e
im
}
}
}
)
im Im Z n( LFDMA
,q
(A.26)
where re and im are the solutions of the following equations, respectively.
{
E {Im {Z
{
( LFDMA )
n ,q
)
Re Z n( LFDMA
,q
)
E Re Z n( ,LFDMA
e
q
}e
)
Im Z n( LFDMA
,q
} {
}
} E {e
}
)
Re Z n( LFDMA
,q
E e
)
Im Z n( LFDMA
,q
}= 0
}
}= 0
}
(A.27)
From (A.22)
N 1
)
Re Z n( ,LFDMA
= c(n, q, p ) Re { x p }
q
p =0
N 1
p
p
= c(n, q, p ) cos Re { x p } sin Im { x p }
N
p =0
N

d re ( n , q , p )
N 1
= d re (n, q, p )
p =0
N 1
Im Z
( LFDMA )
n,q
} = c(n, q, p) Im {x }
(A.28)
p=0
N 1
p
p
= c(n, q, p ) sin Re { x p } + cos Im { x p }
N
p =0
N

dim ( n , q , p )
N 1
= dim (n, q, p )
p =0
where
d re (n, q, p ) = c(n, q, p ) cos ( p N ) Re { x p } sin ( p N ) Im { x p }
d im (n, q, p ) = c(n, q, p ) sin ( p N ) Re { x p } + cos ( p N ) Im { x p } .
Using (A.28),
and
124
)
Re Z n( LFDMA
,q
)
E Re Z n( ,LFDMA
e
q
N 1
N 1
= E d re (n, q, p ) exp d re (n, q, l )

l =0
p =0
N 1
N 1
= E d re (n, q, p ) e dre ( n, q ,l )
p =0
l =0
N 1
d re ( n , q , p )
d re ( n , q ,l )
= E d re (n, q, p )e
e
p =0
l =0
l p
N 1
N 1
= E d re (n, q, p )e dre ( n, q , p )
p =0
)
Im Z n( LFDMA
,q
)
E Im Z n( ,LFDMA
e
q
N 1
N 1
} E {e
= E d im (n, q, p )e dim ( n ,q , p )
p =0
d re ( n , q ,l )
l =0
l p
N 1
} E {e
d im ( n , q ,l )
l =0
l p
(A.29)
and
)
Re Z n( LFDMA
,q
E e
} = E exp N 1 d (n, q, l ) = E N 1 e d
re
N 1
= E e dre ( n , q ,l )
)
Im Z n( LFDMA
,q
E e
l =0
N 1
l =0
= E e dim ( n , q ,l )
l =0
From (A.29) and (A.30), (A.27) becomes
l =0
re ( n , q ,l )
(A.30)
125
N 1
E d re (n, q, p)ere dre ( n ,q , p )

E e
p=0
N 1
re d re ( n , q , p )
} =
E dim (n, q, p)eim dim ( n ,q , p )
E eim dim ( n ,q , p )
p=0
(A.31)
} =
and we can upper-bound the CCDF of an LFDMA signal as follows.

1
M
M 1
Pr Q x
m=0
1
M
n = 0 q =1
N 1 Q 1
N 1
n = 0 q =1
N 1 Q 1
2 e E
re
)
re Re Z n( LFDMA
,q
2 e E {e
re
l =0
re d re ( n , q ,l )
}+e
im
)
Im Z n( LFDMA
,q
+ eim E e im
N 1
E {e
im dim ( n , q ,l )
l =0
Pub, LFDMA
where re and im are the solutions of (A.31) and = w .
(A.32)
Bibliography
[1] D. J. Goodman, Wireless Personal Communications Systems, Addison-Wesley, 1997.

[2] T. S. Rappaport, Wireless Communications: Principles and Practice, Prentice Hall PTR, 2nd edition,
2002.
[3] http://www.itu.int/osg/spu/imt-2000/technology.html
[4] http://www.3gpp.org/Highlights/LTE/LTE.htm
[5] 3GPP TR 25.913 3GPP; Technical Specification Group Radio Access Network;
Requirements for Evolved UTRA and Evolved UTRAN (Release 7).

[6] H. Ekstrm, A. Furuskr, J. Karlsson, M. Meyer, S. Parkvall, J. Torsner, and M. Wahlqvist,
Technical Solutions for the 3G Long-Term Evolution, IEEE Commun. Mag., vol. 44, no. 3,
Mar. 2006, pp. 38-45.
[7] E. Dahlman, H. Ekstrm, A. Furuskr, Y. Jading, J. Karlsson, M. Lundevall, and S. Parkvall,
The 3G Long-Term Evolution - Radio Interface Concepts and Performance Evaluation,

IEEE Vehicular Technology Conference (VTC) 2006 Spring, Melbourne, Australia, May 2006.
[8] R. van Nee and R. Prasad, OFDM for Wireless Multimedia Communications, Artech House, 2000.
[9] H. G. Myung, J. Lim, and D. J. Goodman, Single Carrier FDMA for Uplink Wireless
Transmission, IEEE Vehicular Technology Mag., vol. 1, no. 3, Sep. 2006.
126
127
[10] 3GPP TR 25.814 3GPP; Technical Specification Group Radio Access Network; Physical
Layer Aspects for Evolved Universal Terrestrial Radio Access (UTRA) (Release 7).
[11] B. Sklar, Rayleigh Fading Channels in Mobile Digital Communication Systems: Part I:
Characterization, IEEE Commun. Mag., vol. 35, no. 7, Jul. 1997, pp. 90-100.
[12] 3GPP TS 45.005 3GPP; Technical Specification Group GSM/EDGE Radio Access
Network; Radio Transmission and Reception (Release 7).

[13] W. Y. Zou and Y. Wu, COFDM: An Overview, IEEE Trans. on Broadcasting, vol. 41, no. 1,
Mar. 1995, pp. 1-8.

[14] L. J. Cimini, Jr., Analysis and Simulation of a Digital Mobile Channel using Orthogonal
Frequency Division Multiplexing, IEEE Trans. on Commun., vol. COM-33, no. 7, Jul. 1985,
pp. 665-675.
[15] S. B. Weinstein and P. M. Ebert, Data Transmission by Frequency Division Multiplexing
using the Discrete Fourier Transform, IEEE Trans. Commun. Technol., vol. COM-19, Oct.
1971, pp. 628-634.
[16] C. Y. Wong, R. S. Cheng, K. B. Letaief, and R. D. Murch, Multiuser OFDM with Adaptive
Subcarrier, Bit, and Power Allocation, IEEE J. Select. Areas Commun., vol. 17, Oct. 1999, pp.
1747-1757.
[17] I. Koffman and V. Roman, Broadband Wireless Access Solutions based on OFDM Access
in IEEE 802.16, IEEE Commun. Mag., vol. 40, no. 4, Apr. 2002, pp. 96-103.
128
[18] H. Sari, G. Karam, and I. Jeanclaude, Transmission Techniques for Digital Terrestrial TV
Broadcasting, IEEE Commun. Mag., vol. 33, no. 2, Feb. 1995, pp. 100-109.
[19] D. Falconer, S. L. Ariyavisitakul, A. Benyamin-Seeyar, and B. Eidson, Frequency Domain
Equalization for Single-Carrier Broadband Wireless Systems, IEEE Commun. Mag., vol. 40,
no. 4, Apr. 2002, pp. 58-66.
[20] A. V. Oppenheim, R. W. Schafer, and J. R. Buck, Discrete-Time Signal Processing, Prentice Hall,
2nd edition, 1999.

[21] G. K. Kaleh, Channel Equalization for Block Transmission Systems, IEEE J. Select. Areas
Commun., vol. 13, no. 1, Jan. 1995, pp. 110-121.

[22] M. V. Clark, Adaptive Frequency-Domain Equalization and Diversity Combining for
Broadband Wireless Communications, IEEE J. Sel. Areas Commun., vol. 16, no. 8, Oct. 1998,
pp. 1385-1395.
[23] M. Tchler and J. Hagenauer, Linear Time and Frequency Domain Turbo Equalization,
Proc. IEEE 53rd Veh. Technol. Conf. (VTC), vol. 2, May 2001, pp. 1449-1453.
[24] R. Dinis, D. Falconer, C. T. Lam, and M. Sabbaghian, A Multiple Access Scheme for the
Uplink of Broadband Wireless Systems, Proc. IEEE GLOBECOM 04, vol. 6, Dallas, TX,
Dec. 2004, pp. 3808-3812.
[25] F. Pancaldi and G. M. Vitetta, Block Channel Equalization in the Frequency Domain,
IEEE Trans. Commun., vol. 53, no. 3, Mar. 2005, pp. 463-471.
129
[26] N. Benvenuto and S. Tomasin, Iterative Design and Detection of a DFE in the Frequency
Domain, IEEE Trans. Commun., vol. 53, no. 11, Nov. 2005, pp. 1867-1875.
[27] U. Sorger, I. De Broeck, and M. Schnell, Interleaved FDMA - A New Spread-Spectrum
Multiple-Access Scheme, Proc. IEEE ICC 98, Atlanta, GA, Jun. 1998, pp. 1013-1017.
[28] M. Schnell and I. De Broeck, Application of IFDMA to Mobile Radio Transmission, Proc.
IEEE 1998 International Conference on Universal Personal Communications (ICUPC '98), Florence,
Italy, Oct. 1998, pp. 1267-1272.
[29] F. Adachi, D. Garg, S. Takaoka, and K. Takeda, Broadband CDMA Techniques, IEEE
Wireless Comm., vol. 12, no. 2, Apr. 2005, pp. 8-18.

[30] C. Chang and K. Chen, Frequency-Domain Approach to Multiuser Detection in DS-
CDMA Communications, IEEE Comm. Letters, vol. 4, no 11, Nov. 2000, pp. 331-333.
[31] C. Chang and K. Chen, Frequency-Domain Approach to Multiuser Detection over
Frequency-Selective Slowly Fading Channels, IEEE International Symposium on Personal,

Indoors and Mobile Radio Communications (PIMRC) 2002, Lisboa, Portugal, Sep., 2002, pp. 12801284.
[32] B. M. Popovic, Generalized Chirp-like Polyphase Sequences with Optimal Correlation
Properties, IEEE Trans. Info. Theory, vol. 38, Jul. 1992, pp. 1406-1409.
[33] D. Gesbert, M. Shafi, D. Shiu, P. J. Smith, and A. Naguib, From Theory to Practice: An
Overview of MIMO Space-Time Coded Wireless Systems, IEEE J. Select. Areas Commun.,
vol. 21, no. 3, Apr. 2003, pp. 281-301.
130
[34] D. Tse and P. Viswanath, Fundamentals of Wireless Communication, Cambridge University Press,
2005.
[35] A. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless Communications,
Cambridge University Press, May 2003.

[36] M. Chryssomallis, Smart Antennas, IEEE Antennas and Propagation Mag., vol. 42, no. 3, Jun.
2000, pp. 129-136.

[37] S. Alamouti, A Simple Transmit Diversity Technique for Wireless Communications,
IEEE J. Select. Areas Commun., vol. 16, Oct. 1998, pp. 1451-1458.
[38] V. Tarokh, N. Seshadri, and A. R. Calderbank, Space-Time Codes for High Data Rate
Wireless Communication: Performance Criterion and Code Construction, IEEE Trans.

Inform. Theory, vol. 44, Mar. 1998, pp. 744-765.
[39] G. J. Foschini, Layered Space-Time Architecture for Wireless Communication in a Fading
Environment when using Multi-Element Antennas, Bell Labs Tech. J., vol. 1, no. 2, 1996, pp.
41-59.
[40] I. E. Telatar, Capacity of Multi-Antenna Gaussian Channels, Europ. Trans. Telecommu., vol.
10, Nov./Dec. 1999, pp. 585-595.

[41] L. Zheng and D. N. C. Tse, Diversity and Multiplexing: A Fundamental Tradeoff in
Multiple-Antenna Channels, IEEE Trans. Info. Theory, vol. 49, no. 5, May 2003, pp. 10731096.
131
[42] G. L. Stber, J. R. Barry, S. W. McLaughlin, Y. G. Li, M. A. Ingram, and T. G. Pratt,
Broadband MIMO-OFDM Wireless Communications, Proceedings of the IEEE, Feb. 2004,

vol. 92, no. 2, pp. 271-294.
[43] J. S. Hammerschmidt, C. Brunner, and C. Drewes, Eigenbeamforming - A Novel Concept
in Array Signal Processing, Proc. European Wireless Conference 2000, Dresden, Germany, Sep.
2000
[44] InterDigital, Extension of Uplink MIMO SC-FDMA with Preliminary Simulation Results,
3GPP TSG-RAN WG1 #44, Technical Document R1-060365, Feb. 2006.

[45] Ericsson, Elektrobit, Nokia, Motorola, and Siemens, LTE Channel Models and
Simulations, 3GPP TSG-RAN WG4 #38, Technical Document R4-060334, Feb. 2006.
[46] G. H. Golub and C. F. Van Loan, Matrix Computations, Johns Hopkins University Press, 2nd
edition, 1989.
[47] D. Wulich and L. Goldfeld, Bound of the Distribution of Instantaneous Power in Single
Carrier Modulation, IEEE Trans. Wireless Commun., vol. 4, no. 4, Jul. 2005, pp. 1773-1778.
[48] J. G. Proakis, Digital Communications, McGraw-Hill, 4th edition, 2001.
[49] R. D. Yates and D. J. Goodman, Probability and Stochastic Processes: A Friendly Introduction for
Electrical and Computer Engineers, John Wiley & Sons, 2nd edition, 2005.
132
[50] S. H. Han and J. H. Lee, An Overview of Peak-to-Average Power Ratio Reduction
Techniques for Multicarrier Transmission, IEEE Wireless Commun., vol. 12, no. 2,
Apr.
2005, pp. 56-65.

[51] H. Ochiai and H. Imai, On the Distribution of the Peak-to-Average Power Ratio in OFDM
Signals, IEEE Trans. Commun., vol. 49, no. 2, Feb. 2001, pp. 282-289.
[52] S. L. Miller and R. J. ODea, Peak Power and Bandwidth Efficient Linear Modulation,
IEEE Trans. Commun., vol. 46, no. 12, Dec. 1998, pp. 1639-1648.
[53] S. C. Cripps, RF Power Amplifiers for Wireless Communications, Artech House, 1999.
[54] H. G. Myung, J. Lim, and D. J. Goodman, Peak-to-Average Power Ratio of Single Carrier
FDMA Signals with Pulse Shaping, The 17th Annual IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications (PIMRC 06), Helsinki, Finland, Sep. 2006.
[55] InterDigital, PAPR Considerations for UL MIMO using Precoding, 3GPP TSG-RAN
WG1 #46, Technical Document R1-062161, Aug. 2006.

[56] Q. Shi, OFDM in Bandpass Nonlinearity, IEEE Trans. Consumer Electron., vol. 42, no. 3,
Aug. 1996, pp. 253-258.

[57] S. Grandhi, R. Yates, and D. J. Goodman, Resource Allocation for Cellular Radio Systems,
IEEE Trans. Veh. Technol., vol. 46, no. 3, Aug. 1997, pp. 581-587.
133
[58] L. Jorguseki, J. Farserotu, and R. Prasad, Radio Resource Allocation in Third-Generation
Mobile Communication Systems, IEEE Commun. Mag., vol. 39, no. 2, Feb. 2001, pp. 117123.
[59] S. Nanda, K. Balachandran, and S. Kumar, Adaptation Techniques in Wireless Packet Data
Services, IEEE Commun. Mag., vol. 38, no. 1, Jan. 2000, pp. 54-64.
[60] J. K. Cavers, Variable-Rate Transmission for Rayleigh Fading Channels, IEEE Trans.
Commun., vol. COM-20, no. 1, Feb. 1972, pp. 15-22.

[61] A. Goldsmith and S. G. Chua, Adaptive Coded Modulation for Fading Channels, IEEE
Trans. Commun., vol. 46, no. 5, May 1998, pp. 595-602.

[62] D. J. Goodman and N. B. Mandayam, Power Control for Wireless Data, IEEE Personal
Commun., vol. 7, no. 2, Apr. 2000, pp. 48-54.

[63] J. Lim, H. G. Myung, K. Oh, and D. J. Goodman, Proportional Fair Scheduling of Uplink
Single-Carrier FDMA Systems, The 17th Annual IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications (PIMRC 06), Helsinki, Finland, Sep. 2006.
[64] J. Lim, H. G. Myung, K. Oh, and D. J. Goodman, Channel-Dependent Scheduling of
Uplink Single Carrier FDMA Systems, IEEE Vehicular Technology Conference (VTC) 2006 Fall,
Montreal, Canada, Sep. 2006.
[65] J. Lim, Adaptive Radio Resource Management for Uplink Wireless Networks, Ph.D.
Dissertation, Polytechnic University, Jun. 2006.
134
[66] K. Oh, Impact of Channel Estimation Error in Adaptive Wireless Communication
Systems, MS Dissertation, Polytechnic University, Jun. 2006.

[67] Motorola, Simulation Methodology for EUTRA UL: IFDMA and DFT-Spread-OFDMA,
3GPP TSG-RAN WG1 #43, Technical Document R1-051335, Nov. 2005.

[68] R. J. Lavery, Throughput Optimization for Wireless Data Transmission, M.S. Dissertation,
Polytechnic University, Jun. 2001.

[69] D. J. Goodman, Z. Marantz, P. Orenstein, and V. Rodriguez, Maximizing the Throughput
of CDMA Data Communications, IEEE Vehicular Technology Conference (VTC) 2003 Fall,
Orlando, Florida, Oct. 2003.
[70] A. J. Goldsmith and S.-G. Chua, Variable-Rate Variable-Power MQAM for Fading
Channels, IEEE Trans. Commun., vol. 45, no. 10, Oct. 1997, pp. 1218-1230.

SC Ofdma

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

SC Ofdma

Hochgeladen von

Copyright:

Verfügbare Formate

viii

Submitted in Partial Fulfillment of the Requirements

Broadband wireless mobile communications suffer from multipath frequency-selective fading.

Evolution of Cellular Wireless Communications.................................................................................1

3GPP Long Term Evolution ...................................................................................................................2

Single Carrier FDMA................................................................................................................................5

Objectives and Contributions..................................................................................................................6

Chapter 2 Channel Characteristics and Frequency Multiplexing

Characteristics of Wireless Mobile Communications Channel........................................................13

Orthogonal Frequency Division Multiplexing (OFDM) ..................................................................17

Single Carrier with Frequency Domain Equalization (SC/FDE)....................................................19

Summary and Conclusions.....................................................................................................................22

Chapter 3 Single Carrier FDMA

Overview of SC-FDMA System...........................................................................................................25

Time Domain Representation of SC-FDMA Signals........................................................................31

SC-FDMA and OFDMA .......................................................................................................................36

SC-FDMA and DS-CDMA/FDE........................................................................................................38

SC-FDMA Implementation in 3GPP LTE Uplink............................................................................40

Summary and Conclusions.....................................................................................................................45

Chapter 4 MIMO SC-FDMA

Spatial Diversity and Spatial Multiplexing in MIMO Systems .........................................................48

MIMO Channel .......................................................................................................................................49

SC-FDMA Transmit Eigen-Beamforming with Unitary Precoding ...............................................53

Summary and Conclusions.....................................................................................................................60

Chapter 5 Peak Power Characteristics of an SC-FDMA Signal: Analytical Analysis

Upper Bound for IFDMA with Pulse Shaping ..................................................................................64

Modified Upper Bound for LFDMA and DFDMA..........................................................................70

Comparison with OFDM.......................................................................................................................71

Summary and Conclusion ......................................................................................................................72

Chapter 6 Peak Power Characteristics of an SC-FDMA Signal: Numerical Analysis

PAPR of Single Antenna Transmission Signals..................................................................................76

PAPR of Multiple Antenna Transmission Signals .............................................................................80

Peak Power Reduction by Symbol Amplitude Clipping ....................................................................83

Summary and Conclusions.....................................................................................................................88

Chapter 7 Channel-Dependent Scheduling of Uplink SC-FDMA Systems

Channel-Dependent Scheduling in an Uplink SC-FDMA System..................................................91

Impact of Imperfect Channel State Information on CDS...............................................................96

Hybrid Subcarrier Mapping ................................................................................................................ 106

Summary and Conclusions.................................................................................................................. 110

Appendix A Derivations of the Upper Bounds in Chapter 5

Figure 3.10: DS-CDMA with FDE. ...................................................................................................... 38

values of roll-off factor . ............................................................................................................. 69

Figure 6.10: CCDF of symbol power after clipping. .......................................................................... 86

Evolution of Cellular Wireless Communications

3GPP Long Term Evolution

Mobility: E-UTRAN should be optimized for low mobile speed from 0 to 15

Coverage: Throughput, spectrum efficiency, and mobility targets should be met

Enhanced multimedia broadcast multicast service (E-MBMS).

Spectrum flexibility: E-UTRA shall operate in spectrum allocations of different

Architecture and migration: Packet-based single E-UTRAN architecture with

Radio resource management: Enhanced support for end-to-end QoS, efficient

Single Carrier FDMA

access scheme which is an extension of OFDM to accommodate multiple simultaneous users.

Objectives and Contributions

PAPR characteristics of an SC-FDMA signal: PAPR is an important measure that affects

Channel-dependent resource scheduling with hybrid subcarrier mapping: Channel-dependent

The following is the outline of the remainder of the thesis.

3rd Generation Partnership Project

Bit Error Rate

Binary Phase Shift Keying

Constant Amplitude Zero Auto-Correlation

Complementary Cumulative Distribution Function

Code Division Multiplexing