Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
1
CHAPTER 1
INTRODUCTION
As the field of wireless communication grows, profound interest
has been created in improving the wireless network performance in terms of
spectral efficiency, link reliability and data rate. Multiple Input Multiple
Output (MIMO) technology provides all these benefits as it combats fading
and interference (Mehrzad Biguesh and Alex B.Gershman 2006, Yijia Fan
and John Thomson 2007). However, the main drawback of MIMO technology
is that placement of multiple antennas into a small mobile terminal which
faces the practical difficulty of size limit (Vahid Tarokh et al 1998, Feifei Gao
et al 2008). In order to overcome this drawback, communications using
devices referred to as relays are employed in wireless communication
networks (Van der Meulen 1968, Thomas Cover and Abbas A.El Gamal
1979). Relays are actively studied and considered in the standardization
process of next generation mobile broadband wireless communication
standards such as Third Generation Partnership Project (3GPP), Long Term
Evolution-Advanced (LTE-A), Institute of Electrical and Electronics
Engineers (IEEE) 802.16j multi-hop relay and IEEE 802.16m advanced air
interface for Fourth Generation (4G) working groups (Yang Yang et al 2009).
1.1 EVOLUTION OF WIRELESS RELAY NETWORK
Wireless relay network is a type of collaborative communications
(Yang Yang et al 2009) in which a Relay Station (RS) helps to forward user
information from neighboring Mobile Station (MS) to a local Base Station
2
(BS) (3GPP TR 36.814 V1.2.1). Primarily, in a cellular wireless network,
relays act as repeaters (Jun Ma et al 2011) in forwarding the information from
source node to destination node when there is no direct line of sight path
between them. Basically, a wireless network with relays or virtual antennas is
an arrangement of spatially dispersed nodes placed between source node and
destination node intended for expanding communication range or increasing
communication rate. The concept of using relays in communication systems
was proposed by Van der Meulen in 1971 and was referred as 3 terminal
communication systems (Van der Meulen 1971). Later, it was coined as
Wireless Relay Network (WRN) by Michael Gastpar and Martin Vetterli in
2002 (Michael Gastpar and Martin Vetterli 2002). However, the first
computer based WRN was constructed in the early 1970s by Norman
Abramson (Norman Abramson 1970, Scott Guthery 1997). Figure 1.1 shows
a Wireless Relay Network.
Figure 1.1 Wireless Relay Network
The principle of operation of a wireless relay network is that it
transmits the data in two phases namely phase I and phase II. In phase I, the
3
source node transmits an information signal to relay node through a broadcast
channel. The relay node receives the signal and performs signal processing
tasks referred as relaying strategies (Alireza Shahan Behbahani et al 2008). In
phase II, the processed signal from the relay node is forwarded to the
destination node. In this form of WRN scenario, spatial diversity or antenna
diversity is obtained, as the information signal from the source node reaches
the destination node by passing through two independent fading paths.
Usually spatial diversity techniques provide an additional diversity gain
without incurring an expenditure of transmission time or bandwidth (Rohit U
Nabar et al 2004). By virtue of this form of operation, a WRN gains attention
and it is considered as an effective means to compensate signal fading due to
multipath propagation and shadowing. Signal fading is compensated through
exploitation of spatial diversity provided by the relay nodes in WRN. Also,
WRN attains prominence due to efficient use of power resources yielding
reduced power levels in increasing the battery life of a Wireless Sensor
Network (WSN) (Bo Wang et al 2006).
Currently, WRN is used for realizing long-range communication,
by deployment of relay nodes at intermediate locations along the longer
ranges when direct line of sight path is in deep fade (Himal A.Suraweera and
Jean Armstrong 2007). It also provides flexibility to meet temporary
communication demands under certain scenarios, due to the non requirement
of a fixed infrastructure (Jun Ma et al 2009, Jun Ma et al 2011). Further,
WRN boosts signal strength in coverage holes, in thick buildings, in
underground tunnels or on the cell edges (Chan-Byoung Chae et al 2008, Jun
Ma et al 2011), as they are easier, faster, cheaper to deploy, posses reduced
terminal radiation and longer battery life (John Boyer et al 2004).
As the relay nodes in WRN play an instrumental role in providing
these advantages, three major factors govern its successful operation. They
4
are i) consideration of the relay nodes, ii) relay network topology and iii)
relaying strategy or relaying protocols (Feifei Gao et al 2008, Alireza Shahan
Behbahani et al 2008).
1.2 RELAY NODE CONSIDERATION
The first major factor in improving the performance of WRN is the
selection of suitable scenario to place the relay nodes. In general, there are
two possible scenarios of considering relay nodes. In the first scenario, the
relay nodes can be obtained from a telecommunication agency and in the
second, it can be obtained by cooperating terminals of other users. The second
scenario is referred to as Cooperative Communication (Andrew Sendonaris
et al 2003, Feifei Gao et al 2008).
Cooperative Communication is a form of communication by which
each user acts as a relay for a certain period, has its own information to
transmit and provides coordination between the source node and one or more
relay nodes thereby resulting in spatial diversity (Andrew Sendonaris et al
2003). This is an innovative manner of realizing spatial diversity gain in a
distributed fashion, which is also referred as cooperative diversity or
cooperation diversity (Nicholas Laneman et al 2004). Ultimately, this
enhances reliability, power efficiency, spectral efficiency and data rate in
comparison to MIMO communications.
WRN also allows mobile terminals to act as relays and participate
in information transmission when they are neither the initial source node nor
the final destination node. This aspect of a relay node or mobile terminal in a
WRN makes it suitable for extending the signal and service coverage of a BS
in a cellular network, Wireless Local Area Networks (WLAN), ad-hoc and
hybrid Networks (John Boyer et al 2004). However, cooperative
communication is a promising technique for next generation cellular wireless
5
system and is considered by several task groups in the family of IEEE 802.16,
mainly in IEEE 802.16j and IEEE 802.16m standards (Sebastien Simoens et
al 2010, Kamran Ettimad and Max Riegel 2010).
International MobileTelecommunications -2000 IMT-2000
IEEE 802.16e, IEEE 802.16j IEEE 802.16m
Wideband Code Division Multiple Access (W-CDMA) Worldwide interoperabilityHigh Speed Downlink Packet Access (HSDPA) for Microwave accessHigh Speed uplink Access (HSPUA) (WiMAX)High Speed Packet Access Plus (HSPUA+)Long Term Evolution (LTE)Long Term Evolution Advanced (LTE-A)- 3GPPCode Division Multiple Access (CDMA) 2000 1xUltra Mobile Broadband (UMB) -3GPP2
High Speed Wireless Access Services
Figure 1.2 High Speed Wireless Access Services
Wireless systems to achieve high speed wireless access services are
classified into two categories, as shown in Figure 1.2. They are International
Mobile Telecommunications-Advanced (IMT-A) which is the name defined
by the International Telecommunication Union (ITU) for 4G mobile wireless
broadband communication system and Worldwide interoperability for
Microwave Access (WiMAX) which was approved to become a 3G standard
in the ITU IMT-2000 (Yang Yang et al 2009, Ian F Akyilidiz et al 2010,
Yongming et al 2010) standards family for transmission of data through
wireless communication.
IEEE launched 802.16j working group to develop relay based
multi-hop techniques for WiMAX standards. IEEE 802.16j is a Mobile Multi-
hop Relay (MMR) standard for WiMAX networks created by IEEE in March
6
2006 (IEEE standard/Part16/2009). It mainly intended to enhance the
performance of IEEE 802.16e with the use of relay station (Yang Yang et al
2009). The main objectives of introducing IEEE 802.16j (Vasken Genc et al
2008) are to extend the coverage area, enhance throughput and system
capacity, saving battery life of source node and minimizing relay node
complexity. Primarily, there are two working groups which contributed to the
development of IEEE 802.16j standard (IEEE 802.16j/D9 2009). The first
group IEEE 802.16 working group on broadband wireless (David Soldani and
S.Dixit 2008) access standard supports the development of broadband
Wireless Metropolitan Area Network (WMAN) and the second group is the
WiMAX (Chackchai So-In et al 2009), forum which certifies and promotes
broadband wireless products.
Generally two types of relays are defined in IEEE 802.16j task
group namely Type-I (Non-Transparency) and Type-II (Transparency) (Yang
Yang et al 2009).
1.2.1 Type-I (Non-Transparency)
Type-I relays are used to establish communication between local
BS and MS located far away from the BS. A relay link is established between
a MS and a RS and an access link between BS and RS. The access from the
MS to the BS takes place through the relay station(s). The Type-I relay station
transmits common reference signals and control information of the BS to the
MS in order to communicate with the MS and connect the same to the BS.
This process allows to increase capacity and improve the service coverage, as
the remote mobile station units are connected to the base station through the
Type-I relay station (Yang Yang et al 2009, 3GPP TR 36.814, V1.2.1).
7
1.2.2 Type-II (Transparency)
If the MS is within the vicinity of the BS, then Type-II relays are
used. In this case, the RS does not transmit common reference signal or
control information to the MS. The local MS service quality and the link
capacity is improved by connecting the MS to the BS via the relay station in
presence of a direct communication link between the BS and MS. The overall
capacity of the system is improved by achieving multipath diversity and
transmission gains of the local mobile station (Yang Yang et al 2009).
1.3 RELAY NETWORK TOPOLOGY
In a WRN, one of the significant factors to be considered for
improving the network performance is network topology (Younsun Kim et al
2008). There are two types of relay transmission topologies namely
i) Serial Relay Transmission Topology
ii) Parallel Relay Transmission Topology
1.3.1 Serial Relay Transmission Topology
Serial relay transmission is used for long distance communication
and range-extension in regions having shadows. In serial relay topology,
signal propagates from one relay to another relay and the channels of
neighboring hop are orthogonal to avoid any interference. The advantage of
this topology is that it provides power gain. However, the drawback of serial
relay transmission is that it suffers from multi-path fading. In outdoors and
non-line of sight communication, signal wavelength is large and installation
of multiple antennas is not possible.
8
1.3.2 Parallel Relay Transmission Topology
Parallel relay transmission is employed in WRN to increase
robustness against multi-path fading (Helmut Bolcskei et al 2006). In this
topology, the signal propagates through multiple relays in the same hop(Wael
Jafar et al 2010) and the destination node combines the signals received with
the help of various combining schemes. The advantage of this scheme is that
it provides power gain and diversity gain. A typical WRN topology is shown
in Figure 1.3. The topologies are described as follows
i) A single antenna source node, a single antenna relay node and
a single antenna destination node.
ii) A single antenna source node, multiple relay nodes with
single antenna and a single antenna destination node.
iii) A source node with multiple antennas, a relay node with
multiple antennas and a destination node with multiple
antennas.
iv) Multiple source nodes with single antenna relay node with
multiple antennas and multiple destination nodes with single
antennas.
v) Multiple source nodes with multiple antennas, multiple relay
nodes with multiple antennas and multiple destination nodes
with multiple antennas.
9
Figure 1.3 WRN Network Topology
1.4 RELAYING STRATEGY
The third factor which facilitates the WRN performance is the
relaying strategy or relaying protocols. There are two major classifications of
relaying strategy namely Non-Regenerative relaying strategy and
Regenerative relaying strategy as (Xiaojun Tang and Yingbo Hua 2007, Olga-
Munoz Mediana et al 2007) shown in Figure 1.4. Non-regenerative relaying
strategy (Ronghong Mo et al 2010) mainly comprises of Amplify and
Forward (AF) and Compress and Forward (CF) relaying strategies.
Regenerative relaying strategy comprises of Decode and Forward (DF), and
Demodulate and Forward (DMF) relaying strategies (Alireza Shahan
Behbahani et al 2008).
10
Figure 1.4 WRN Relaying strategy
1.4.1 Amplify and Forward Relaying Strategy
In AF relaying strategy as shown in Figure.1.5, the signal
broadcasted from the source node is received by the relay node and the relay
node sends a scaled version of it to the destination node. Mainly, the relay
node performs a linear signal processing task (or) an amplification operation
on the transmitted signal (Feifei Gao et al 2008). AF relaying strategy uses
varying gain or constant gain to limit the transmit power at the relay node.
Varying-gain relaying scheme (Mazen Omar Hassna and Mohammed Slim-
Alouini 2004) is based on the knowledge of the instantaneous source to relay
channel coefficients at the corresponding relays and maintains constant
transmit power at the relays. Whereas, constant gain relaying scheme reduces
system complexity and maintains long-term average transmit power at each
relay. Fixed gain relaying scheme is less complex as it does not require the
knowledge of the fading channel realization (and hence channel estimation) at
the relay node (Foroogh.S.Tabataba et al 2011). Functionally, AF relays
Non-Regenerative relaying strategy Regenerative relayingstrategy
Amplify and Forward Compress and Forward Decode and Forward Demodulate and Forward
Relaying strategy
11
resemble traditional analog relays. The main advantages of AF relaying
strategy are
Its inherent simplicity (Woraniti Limpakoom et al 2009),
Lower computational complexity in terms of less processing
burden (load) on the relay (Mazen Omar Hassna and
Mohammed Slim-Alouini 2004, Chirag S. Patel and Gordon
L. Stuber 2007),
Energy saving in power limited systems (Woraniti
Limpakoom et al 2009).
Very short delay as it only amplifies the signal (Rui Zhang
et al 2009).
The drawback of AF relaying strategy is
Noise accumulation along the transmission path when the
channel is in deep fade (Karim G.Seddik et al 2007).
In general, AF relaying strategy is suitable in applications where
high complexity is not acceptable. When AF relaying strategy is applied to a
wireless relay network, it is referred to as Amplify and Forward Wireless
Relay Network (AFWRN) (Shashibhushan Borade et al 2007). AFWRN
relays operate in half-duplex or full duplex mode. In half duplex mode,
communication is supported in both directions, but only one direction at a
time. Typically, when a destination node receives a signal from the source
node, the destination node must wait till the source node to stop transmitting
the signal before it replies. A full duplex relay is practically realizable because
it transmits and receives at the same time but it is hardly implementable in
12
reality (Rui Zhang et al 2009). An example of the half-duplex system is a
walkie-talkie. There are several benefits of using half-duplex over full-duplex.
The most important one is its lower implementation complexity, whereas for a
full-duplex system, simultaneous transmission and reception of signals
requires precise design for the component.
Figure 1.5 Amplify and Forward Wireless Relaying Strategy
1.4.2 Compress and Forward Relaying Strategy
A CF or estimate and forward or observe and forward or quantize
and forward relaying strategy quantizes the received signal. More precisely,
the relay employs source coding with side information at the destination node.
This scheme is also known as Wyner Ziv coding (Sebastien Simoens et al
2010). The CF relaying strategy is efficient in cases where source node to
relay node and source node to destination node channels are of comparable
quality, and relay node to destination node channel is also good.
13
1.4.3 Decode and Forward Relaying Strategy
A DF is an example of regenerative relaying strategy in which the
relay node first verifies the correctness of the information sent from the
source node by decoding all the information. Then it re-encodes the
information and forwards it to the destination node as shown in Figure 1.6. If
DF is employed in WRN, it is referred to as Decode and Forward Wireless
Relay Network (DFWRN). In DFWRN, channel estimation is similar to that
in a traditional point to point system. Since relays are geographically
distributed and different relays come from different mobile terminals, the
individual power constraint for each relay needs to be considered (Feifei Gao
et al 2008a). The advantage of this DF relay strategy is that it involves
decoding and encoding of information bits at relay nodes. However, the
drawbacks of DF relaying strategy are
Longer propagation delay, higher delay tolerance causing
security problems (Yongming Huang et al 2010) and
increased computational complexity due to decoding and
encoding.
Suffers from error propagation (Karim G. Seddik et al 2007)
Highly non trivial and the complexity increases significantly
as the number of relay node grows (Bo Wang et al 2006).
14
Figure 1.6 Decode and Forward Wireless Relaying Strategy
1.4.4 Demodulate and Forward Relaying strategy
In Demodulate and Forward (DMF) relaying strategy, the relay
demodulates each received symbol individually, remodulates and retransmits
them to the destination node (Alireza Shahan Behbahani et al 2008).
Demodulate and Forward relaying strategy is an alternative to decode and
forward relaying strategy to reduce receiver power consumption due to
channel decoding at the relay as well as to minimize the overall delay at the
destination node.
Among the various relaying strategies, AF is found highly suitable
for parallel relay networks due to its ability to pass on soft information
(Krishna Srikanth Gomadam and Syed Ali Jaffar 2009). It is also used for
adhoc wireless systems in which high implementation complexity for
encoding and decoding is rarely acceptable (Yanwu Ding et al 2009). Due to
the advantages offered by AF relaying strategy, this thesis mainly pertains to
amplify and forward wireless relay networks and its performance analysis. In
addition a small section on DFWRN is also dealt herein this thesis.
15
1.5 SYSTEM MODEL FOR THREE TERMINAL AFWRN
Consider an AFWRN with three terminals namely a source node, a
relay node and a destination node as shown in Figure 1.7 with single antenna
in each of the nodes (Yu Bi and Yanwu Ding 2012).
Figure 1.7 Three Terminal AFWRN
As shown in Figure 1.7, a source node communicates with a
destination node directly and also with the help of a relay node. The AFWRN
assumes perfect synchronization among the nodes, employs half-duplex
transmissions with an orthogonal transmit scheme in non overlapping time
slots. In AFWRN, information data transfer between the source node to
destination node is accomplished in two phases namely phase I and phase II.
In phase I, signal is transmitted from the source node to the destination node
and also to the relay node as the channel is a broadcast channel. The received
signal at the destination node (Woraniti Limpakom et al 2009) is given as
dssdsd nxfy (1.1)
where sdf is the Channel Impulse Response (CIR) or Channel State
Information (CSI) between the source node and the destination node, sx is the
transmitted signal and dn is the noise at the destination node. Similarly, the
received signal at the relay node (Woraniti Limpakom et al 2009) is
represented as
16
rssrsr nxgy (1.2)
where srg is the CSI between the source node and the relay node sx is the
transmitted signal and rn is the noise at the relay node. During phase II, the
received signal at the relay node sry is amplified by and forwarded to the
destination node. The received signal at the destination node is expressed as
rd rd sr dy h y n (1.3)
where rdh is the CSI between the relay node and the destination node and is
the amplifying gain (Yu Bi and Yanwu Ding 2012). The gain is chosen
such that it satisfies the power constraint at the relay and it is given as
22Ssr
r
PgP (1.4)
where SP is the source node power, rP is the relay node power and 2 is the
noise variance. The CSI defined by srsd gf , and rdh represent the effects of
path-loss, shadowing and frequency non-selective fading. The terms rn and
dn capture the effects of receiver noise and interference at relay node and
destination node respectively. These terms are modeled as mutually
independent, circularly symmetric, complex Additive White Gaussian Noise
(AWGN) with zero mean and variance 2 ( Woraniti Limpakom et al 2009).
1.6 SYSTEM MODEL FOR THREE TERMINAL MIMO AFWRN
Consider a three terminal MIMO AFWRN equipped with SN
antennas at the source node ,S relN antennas at the relay node R and DN
17
antennas at the destination node D as shown in Figure 1.8
Figure 1.8 Three Terminal MIMO AFWRN
The MIMO channel matrix between the source node and the relay
node is represented as G and from the relay node to the destination node as
H respectively (Kyoung-Jae Lee et al 2010). It is assumed that the elements
of channel matrices are Independent Identically Distributed (IID) and
experience Rayleigh flat fading as assumed in 3 terminal Single Input Single
Output (SISO) AFWRN system. MIMO channel matrices are represented
mathematically by Srel NNCG and relD NNCH and its entries are assumed to
be circularly symmetric complex Gaussian random variables with zero mean
and unit variance.
In phase I, the source node transmits a 1SN data signal vector s
through the MIMO channel ,G to the relay node and the received signal
vector 1relN at the relay node (Kyoung-Jae Lee et al 2010) is given as
RR nGsy (1.5)
where Rn is the 1relN AWGN vector at the relay node. In phase II, at the
relay node R , the 1relN received signal vector Ry is multiplied by an
relrel NN amplifying matrix F and the signal is forwarded to the destination
18
node. The amplifying matrix is defined as PF , where is the fixed
amplifying gain at the relay node and P is an relrel NN unitary Precoding
matrix and it is usually a diagonal matrix or permutation of a diagonal matrix.
The 1DN received signal vector at the destination node in a 3 terminal
MIMO AFWRN system (Kyoung-Jae Lee et al 2010) is represented as
DeRD nHFnHFGsy (1.6)
where Den is the 1DN noise vector at the destination node. Let HFGW
be the SD NN overall channel matrix between the source node and the
destination node and DeRD nHFnn be the 1DN overall noise vector at
the destination node. Then, Equation (1.6) is represented as
DD nWsy (1.7)
1.7 CHANNEL ESTIMATION FOR AF WIRELESS RELAY
NETWORK
Channel estimation in an AFWRN (Tao Cui et al 2007, Feifei Gao
et al 2008) is a technique of estimating the wireless channel coefficients or
CSI between a transmitter or source node and a receiver or a destination node.
The wireless channel in an AFWRN normally contributes to undesirable
effects of time dispersion, attenuation in magnitude and reduction in phase in
the signal. These undesirable effects have to be eliminated, before data
detection in an AFWRN in case if linear modulation schemes are to be
applied at the source node for data transmission.
Channel estimation in an AFWRN is carried out using estimators
which are based on estimation theory. Estimation theory forms the basis for
many signal processing applications intended for extracting information.
19
Extracting information ultimately implies estimation of the values of a group
of parameters. The estimates are usually obtained from a mathematical
operator named as Estimator. Mathematically, an estimator may be thought of
as a rule that assigns a value to the estimation parameter of interest for each
realization of the observed signal (Steven M. Kay 1993).
The parameters of interest in a group can be Deterministic or
Random but are unknown. Estimation of parameters in a group can be
classified into two types namely Classical estimation and Bayesian
estimation. Estimation of Deterministic parameters in a group is referred as
Classical estimation and estimation of Random parameters in a group is
referred to as Bayesian estimation. The classical estimation approaches are
Cramer-Rao Lower bound (CRLB), Rao-Blackwell-Lehman-Scheffe, Best
Linear Unbiased Estimator (BLUE), Maximum Likelihood (ML) estimator,
Least Squares (LS) estimator and Method of Moments (MOM). The
Bayesian estimation approaches are Minimum Mean Square Error (MMSE)
estimator, Maximum APosteriori Estimator (MAP), Linear Minimum Mean
Square Error (LMMSE) estimator (Steven M. Kay 1993).
In addition to the above classification based on the parameter of
interest, basically, an estimator can be classified into two types based on the
property it satisfies. An unbiased estimator is defined as the one that on an
average the estimator yields true value of the unknown parameter. Unbiased
estimators tend to have a symmetric Probability Density Function (PDF)
centered about the true value of the parameter. Within this class of unbiased
estimators, the estimator with minimum variance exists. In general, a
Minimum Variance Unbiased (MVU) estimator does not exist although
several methods are available to find them and those methods rely on Cramer-
Rao Lower bound and the concept of a sufficient statistic. If a minimum
variance unbiased estimator does not exist, or if the previous approaches fail,
20
a constraint is placed on the estimator to be linear in data. This results in an
easily implementable estimator. On the other hand, a biased estimator is the
one that is characterized by a systematic error, which presumably should not
be present (Steven M. Kay 1993) and its performance is always very poor.
Performance of any estimator obtained will be critically dependent
on the PDF assumptions and estimators in general need to be optimal. In
searching for optimal estimators an optimality criterion needs to be adopted.
A natural one is the Mean Square Error (MSE) (Steven M. Kay 1993) which
measures the average mean squared deviation of the estimator from the true
value. Unfortunately, adoption of this natural criterion leads to unrealizable
estimators, which implies that it cannot be written as a sole function of data.
Figure 1.9 Block Diagram of Wireless Communication System
As shown in Figure 1.9, the technique of channel estimation
actually forms an estimate of the amplitude and phase shift of the wireless
channel with the aid of the estimation algorithms and training data signal or
the pilot signal. This is essential to eliminate the effect of the wireless channel
and thereby makes data detection more efficient in an AFWRN. Hence,
21
channel estimation attains significance as an important tool for determining
receiver performance in an AFWRN. Channel estimation technique in an
AFWRN (Aris S. Lalos et al 2008) depends mainly on the pilot signal energy,
channel estimation algorithms, and the environment conditions. CSI for an
AFWRN can be obtained through three techniques:
Training based channel estimation technique
Blind channel estimation technique
Semi-blind channel estimation technique
In training based channel estimation technique for AFWRN,
training data signals or pilot signals that are known a priori to the destination
node are transmitted into the wireless channel from the source node. The
commonly employed training data signals (Feifei Gao et al 2008, Oomke
Weikert and Udo Zolzer 2007) are presented in the Table 1.1.
Table 1.1 Training Data signals
Training Data Sequence Mathematical DescriptionAll one vector sequence 11..111N1 is an 1N vector
with onesZadoff Chu sequence – It is ageneralized chirp like polyphasesequence. The elements of thissequence have same magnitude inboth time and frequency domain toreduce Peak to Average Power Ratio(PAPR) problem
oddisNNnN
qnQnj
evenisNNnN
qnQnj
e
enx
;1....2,01;)21(
;1....2,01;)2(
)(
where q and Q are integers in whichQ is relatively prime to N
Perfect Root of Unity (PRUS)sequence - It is constructed for anylength N using Frank-Zadoff Chusequences.
;1,...01;)1(
;1,...01;2
)(NkoddwithNfor
NkQkj
NkwithevenNforN
Qkj
e
enx
where Q is a natural number greaterthan zero and co-prime to N
22
Blind channel estimation technique is carried out by evaluating the
statistical information of the channel and it does not rely on the knowledge of
the transmitted signal. A well known class of blind estimation algorithms is
the decision directed or decision feedback algorithms. These algorithms rely
on the demodulated and detected sequence at the receiver to reconstruct the
transmitted signal. Blind channel estimation has its advantage in that it has no
overhead loss. But, it is only applicable to slowly time-varying channels due
to its need for a long data record. The major drawback of the blind channel
estimation technique is that a decision or bit error at the receiver will cause
the construction of an incorrect transmitted signal. On the whole, the decision
error introduces a bias in the channel estimate thereby making it less accurate.
Semi-blind channel technique is a hybrid of blind and training
techniques, which utilizes pilots and other natural constraints to perform
channel estimation. The techniques commonly used for estimating the
channel coefficients in a an AFWRN are Least Squares (LS) estimator,
Minimum Mean Square Error (MMSE) estimator, Best Linear Unbiased
Estimator (BLUE) and Maximum Likelihood (ML) estimator. Also other
variants of LS estimators namely Scaled LS, Sequential LS, Ordered LS,
Weighted LS. Further, Relaxed Minimum Mean Square Error (REMMSE)
estimators and Maximum APosteriori (MAP) estimators also exist (Mehrzad
Biguesh and Alex B. Gershman 2006, Steven M. Kay 1993). This thesis
focuses mainly only on the general forms of estimators and not into its
variants.
1.7.1 Least Squares Estimator
LS is an estimator with no optimality property associated with it
and does not constrain it to be linear in data. The method of LS dates back to
1795 when Gauss used the method to study planetary motions (Steven M.
Kay 1993). LS estimator minimizes the squared difference between the
23
received signal and the assumed data or noiseless data. The advantage of LS
estimator is that it is easy to implement The drawback of LS estimator is that
it reduces signal error and does not reduce channel estimation error since it
makes probabilistic assumptions only on the data for the signal model which
is assumed (Steven M. Kay 1993,Mehrzad Biguesh and Alex B. Gershman
2006).
1.7.2 Minimum Mean Square Error Estimator
It is an optimal estimator defined to be the one which minimizes
mean square error when averaged over all the realizations of the estimation
parameter of interest and the received signal. The estimator which minimizes
the Bayesian MSE is called as Minimum Mean Square Error (MMSE)
estimator. MMSE estimator is the mean of the posterior Probability Density
Function (Steven M. Kay 1993). The advantage of MMSE estimator is that it
produces the minimum mean square error through minimization of Bayesian
MSE. The drawback of MMSE estimator is that it is based on conditional
PDF and integration of it is complex for increased number of observation
samples.
1.7.3 Best Linear Unbiased Estimator
Best Linear Unbiased Estimator (BLUE) is an estimator obtained
by restricting the estimator to be linear in data and finding the linear estimator
that is unbiased and has minimum variance characteristics. This estimator
referred to as BLUE is Best Linear Unbiased Estimation according to (Steven
M. Kay 1993) is determined only with the knowledge of only the first and
second moments of the probability density function. The first moment is the
mean and the second moment is the spread from the mean. It does not require
the complete knowledge of PDF and hence it is employed for practical
implementations (Steven M. Kay 1993, Mehrzad Biguesh and Alex B.
24
Gershman 2004). Also, BLUE posses the minimum variance among the
family of unbiased estimators making it a Minimum Variance Unbiased
(MVU) estimator. The advantage of BLUE is that it has minimum variance
characteristics. The drawback of BLUE is that it is a suboptimal estimator.
1.7.4 Maximum Likelihood Estimator
Maximum Likelihood (ML) estimator is defined as the value of the
estimation parameter which maximizes the likelihood function or PDF
(Steven M. Kay 1993). ML has asymptotic properties of being unbiased,
achieves CRLB and has a Gaussian PDF. The advantage of ML estimator is
that it is the optimal estimator as it is based on the PDF. The drawback of ML
estimator is that it is an intensive search algorithm.
In general, the channel estimation techniques use two estimation
theoretic performance metrics, namely MSE and CRLB. The first metric MSE
depends on the specified channel estimation algorithm employed at the
receiver. MSE defines the expectation of the squared value for the difference
between the estimated value and the true value of the estimation parameter of
interest. Second metric, CRLB sets a lower bound on the MSE of the channel
estimator. Also, it allows to place a lower bound on the performance of an
unbiased estimator and at worst, acts a benchmark against which the
performance of any unbiased estimator can be compared. On par with MSE of
the channel estimation, the expression for CRLB is independent of specific
channel estimation technique employed at the receiver (Steven M. Kay 1993).
1.8 PERFORMANCE METRICS FOR AFWRN
The metrics for analyzing the performance of AFWRN are ergodic
capacity, Bit Error Rate (BER) and outage probability.
25
1.8.1 Ergodic Capacity
Ergodic capacity is an important metric for analyzing AFWRN
performance as it specifies the ability to maintain long term constant bit rates
(Shin Jin et al 2010). Ergodic capacity refers to averaging the randomness of
the channel gain over time (Bo Wang et al 2004). It attains an important
metric as it yields an information (Thomas M. Cover and Thomas 1991)
theoretic bound (Bhuvan Modi et al 2012) on the achievable rate for reliable
communication over fading channels. But, ergodic capacity is generally
difficult to obtain (Shin Jin et al 2011). Ergodic capacity (Bo Wang et al
2005) analysis with Perfect CSI and Imperfect CSI provides an insight into
the maximum capacity (Anders Høst Madsen 2002, Gerhard Kramer et al
2003) that can be reached.
1.8.2 Bit Error Rate
Bit Error Rate is an important performance metric which analyzes
the performance of AFWRN for any specified modulation scheme. BER is
defined as the rate at which errors occur in an AFWRN when a set of
information bits are transmitted (Seungyoup Han et al 2009). The definition
of BER is defined by a simple formula
BER = number of errors / total number of bits sent
In practical situations, the wireless channel suffers from severe
fade, and ultimately degrades the overall AFWRN system performance. As a
result, proper receiver structure is to be designed to improve system
performance (Feifei Gao et al 2008, Yindi Jing and Babak Hassibi 2004).
1.8.3 Outage Probability
Outage Probability (David Tse and Pramod Viswanath 2005) is an
important metric to analyze the performance of AFWRN in slow fading
26
channels. It arises when the channel is so poor and that no scheme can
communicate reliably at a fixed data rate. The largest rate of reliable
communication at a certain outage probability is called outage capacity. An
AFWRN system is said to be in outage if the received Signal to Noise Ratio
(SNR) falls below a specific threshold SNR
1.9 RESEARCH ISSUES IN AFWRN
1.9.1 Channel Estimation
Channel estimation is one of the major issues in an AFWRN. It
attains significance as it is essential for signal detection in wireless relay
networks in order to carry out phase alignment which is required in coherent
relaying. Another important feature of channel estimation is that the
performance of a good receiver is based on impact of the channel. Generally,
channel state information is considered to be perfect or its availability is
known to destination node in an AFWRN. When perfect CSI assumption is
made at the destination node, each of the source node present in an AFWRN
completely subtracts/cancels out its own data at the broadcasting phase. Also,
channel uncertainty, due to channel estimation errors, prevents the users from
subtracting/canceling out completely their own data. This leads into self
interference and consequently system performance degradation. Moreover, in
practical environments, CSI is not available and has to be estimated.
In an AFWRN, individual channel coefficients between source
node to relay node and relay node to destination node need not be estimated
and only the overall channel coefficients needs to be estimated (Feifei Gao
et al 2008). This is because estimating the source node to the relay node and
relay node to the destination node channel coefficients has several drawbacks.
The first drawback is that the relay node must inform the destination node of
the estimate of the source node to the relay node channel, which consumes
27
bandwidth efficiency and consumes additional transmitting power. The
second drawback is that the transmission of channel estimate will suffer from
further distortions (Feifei Gao et al 2008).When channel coefficients are
estimated it results in errors and delays which deteriorate the potential
performance improvement of AFWRN. Hence, selection of the estimation
algorithm which reduces estimation error to a minimal value needs to be
considered.
1.9.2 Synchronization
Synchronization is another issue in AFWRN as the receiver has to
determine when there is a signal to demodulate and where the packets start in
order to interpret the received signal. When there is no data transmission, it is
important that the receiver is able to enter into a mode where power can be
saved. However, as soon as transmission starts the receiver has to achieve
synchronization in a very short time to prevent the loss of data.
1.9.3 Power Allocation
In AFWRN transmission power is the primary resource as it is
shared by many relay nodes and it ultimately plays a role in its lifetime and
scalability (Tony Q.S. Quek et al 2010). In AFWRN, relay terminals
cooperate with each other for data transmission, making power control of the
source terminal and power allocation issues too complicated. Hence, to
overcome complications, and to maximize AFWRN throughput, best relay
node selection with uniform power distribution between the source node and
the relay nodes needs to be considered (Jun Cai et al 2006). Under uniform
power allocation, threshold based sufficient and necessary conditions have to
be derived to facilitate the search of feasible relay node so that better network
throughput in terms of average mutual information can be achieved from user
relaying over direct transmission. With the help of the derived conditions,
28
searching time for the best relay node from a set of available relay nodes can
be reduced. From the selected relay node, an analytical expression, optimal
power allocation can be developed. Moreover, to minimize symbol error rate
for overall transmission, optimal power allocation schemes can be employed.
Generally, optimal power and power allocation schemes are obtained using
Convex Programming (Feifei Gao et al 2008) so that signal to noise ratio can
be maximized at the destination node even with partial availability of CSI.
1.9.4 Information Exchange/Forward/Backward
Information exchange in an AFWRN plays an important role in
implementing any resource optimization process. This is mandatory for
amplify and forward wireless relay networks employing coherent modulation
and demodulation techniques. The information is transmitted from the source
node to relay node and relay node to the destination node in the forward
direction. In backward direction information bits are sent in as feedback from
the destination node to the relay node which is useful for applications like
Beamforming.
Beamforming is a spatial filtering or a signal processing technique
used in sensor arrays for directional signal transmission or reception. This is
achieved by combining elements in a phased array in such a way that signals
at particular angles experience constructive interference while others
experience destructive interference. Beamforming can be used at both the
transmitting and receiving ends in order to achieve spatial selectivity. The
improvement compared with omni-directional reception/transmission is
known as the receive/transmit gain. During transmission, a beamformer
controls the phase and relative amplitude of the signal at each transmitter, in
order to create a pattern of constructive and destructive interference in the
29
wavefront. While receiving, information from different sensors is combined
in a way where the expected pattern of radiation is preferentially observed.
In the receive beamformer, the signal from each antenna is
amplified by different weights. Generally different weighting patterns are
used to achieve the desired sensitivity patterns. A main lobe is produced
together with nulls and side lobes. Moreover, controlling the main lobe width
(the beam) and the side lobe levels, the position of a null can be controlled.
This is useful to ignore noise in one particular direction, while listening for
events in other directions. A similar result can be obtained on transmission.
Beamforming techniques are broadly divided into two categories,
namely conventional beamforming and adaptive beamforming. Conventional
beamformers use a fixed set of weightings and time-delays (or phasing) to
combine the signals from the sensors in the array. Primarily, it uses the
information about the location of the sensors in space and the wave directions
of interest. In contrast, adaptive beamforming techniques generally combine
this information with properties of the signals actually received by the array,
mainly to improve rejection of unwanted signals from other directions. This
process may be carried out in either time or in frequency domain.
Adaptive beamformers automatically adapt its response to different
situations. A criterion has to be set up to allow the adaptation to proceed such
as minimizing the total noise output. Because of the variation of noise with
frequency, in wide band systems it may be desirable to carry out the process
in the frequency domain. Adaptive beamforming has the characteristics of
interference signal minimization.
30
1.9.5 Interference Management
Interference arises in an AFWRN due to the broadcast nature of the
relay nodes, presence of multiple of relay nodes and multiple source nodes.
Generally, the broadcast nature in an AFWRN exposes the relays to common
interferers which results in correlated noise at the nodes (Krishna Srikanth
Gomadam and Syed Ali Jafar 2006) which is a significant drawback for
successful operation of AFWRN. Such common interference is minimized by
careful selection of relay gains so that they add out of phase. Existence of
multiple relays in a network naturally creates a multiuser scenario and
multiple sources, can produce interference among data streams. Interference
management is therefore closely related to resource optimization problems as
well as multiple access techniques in relaying systems.
Other issues of AFWRN such as level of relay mobility, time and
frequency synchronization require careful examinations but are not dealt with
this thesis. This thesis addresses only channel estimation issue and ergodic
capacity, BER metrics as they govern the successful operation of an AFWRN.
In this thesis, it is assumed that the nodes have perfect synchronization and no
interference among them.
1.10 SCOPE OF THE THESIS
Among the research issues of AFWRN, this thesis addresses the
issue of estimating the channel coefficients of WRN, since the knowledge of
channel state information plays an important role in signal detection. This
thesis analyses BLUE based channel estimation technique and the
conventional channel estimation techniques namely LS estimator and MMSE
estimator for SISO AFWRN and MIMO AFWRN. This thesis also addresses
31
the effect of channel estimation error on ergodic capacity and BER
performance of AFWRN.
1.10.1 Objectives
1. To develop a channel estimation technique for AFWRN
which provides minimal estimation error.
2. To analyze the performance of the proposed channel
estimation technique and compare its MSE performance with
the existing estimation methods for both SISO AFWRN and
MIMO AFWRN.
3. To analyze ergodic capacity and BER performances of
AFWRN with Imperfect CSI and analyze the impact of the
proposed channel estimation technique on these performance
metrics for SISO AFWRN and MIMO AFWRN.
4. To apply the proposed channel estimation technique in the
various practical system models of SISO AFWRN and MIMO
AFWRN and analyze the performance.
1.11 ORGANIZATION OF THE THESIS
In Chapter 2, a new method using the concept of BLUE is proposed
for estimating the overall channel coefficients of SISO AFWRN. Analytical
expression for MSE of the proposed technique is also derived. The MSE
performance of the proposed technique is analyzed and compared with CRLB
and other existing methods. In Chapter 3, the proposed technique is extended
for estimating the overall channel coefficients for MIMO AFWRN. The MSE
performance is also analyzed and compared with existing techniques.
32
Chapter 4, presents ergodic capacity and BER analysis of AFWRN with
Imperfect CSI in SISO and MIMO environments. In Chapter 5, the proposed
channel estimation technique is applied for a relay assisted multi user
downlink transmission system. Though, this thesis is mainly concerned with
AFWRN, using BLUE, channel estimation for DFWRN using BLUE is also
mathematically portrayed for analysis. Chapter 6, concludes the thesis, with a
note on future work.