ssp-mt-22 doc.docx

8/10/2019 SSP-MT-22 DOC.docx

1/56

CHAPTER 1INTRODUCTION

1.1Introduction

In communication systems, aduplex communication system is a point-to-point

system composed of two connected parties or devices that can communicate

with one another in both directions, simultaneously. An example of a duplex

device is a telephone. The people at both ends of a telephone call can speak at

the same time; the earphone can reproduce the speech of the other person as

the microphone transmits the speech of the local person, because there is a

two-way communication channel between them.

Duplex systems are employed in many communications networks, either to

allow for a communication "two-way street" between two connected parties or

to provide a "reverse path" for the monitoring and remote adjustment of

equipment in the field.

Systems that do not need the duplex capability use instead simplex

communication in which one device transmits and the others just "listen."

Examples are broadcast radio and television, garage door openers, baby

monitors, wireless microphones, radio controlled models, surveillance cameras,

and missile telemetry. There are two types of duplex communications. They

are:


2/56

1. Half Duplex

2. Full Duplex

A half-duplex (HDX) system provides communication in both directions, but

only one direction at a time (not simultaneously). Typically, once a party begins

receiving a signal, it must wait for the transmitter to stop transmitting, before

replying (antennas are of trans-receiver type in these devices, so as to transmit

and receive the signal as well).

An example of a half-duplex system is a two-party system such as a walkie-

talkie, wherein one must use "Over" or another previously designated command

to indicate the end of transmission, and ensure that only one party transmits

at a time, because both parties transmit and receive on the same frequency.

A full-duplex (FDX), or sometimes double-duplex system, allows

communication in both directions, and, unlike half-duplex, allows this to

happen simultaneously. Land-line telephone networks are full-duplex, since

they allow both callers to speak and be heard at the same time, the transition

from four to two wires being achieved by a Hybrid coil. A good analogy for a

full-duplex system would be a two-lane road with one lane for each direction.

Two-way radios can be designed as full-duplex systems, transmitting on one

frequency and receiving on another. This is also called frequency-division

duplex. Frequency-division duplex systems can be extended to farther


3/56


4/56

duplex mode. The latter operation requires a spatial separation between

transmit and receive antennas to reduce loop-back interference from the

transmit antennas to the receive antennas.

From signal processing point of view AF relays offer interesting challenges,

especially when the AF relay operates infull-duplex mode: Adaptive algorithms

are required for loop-back interference cancellation. Furthermore, the effect of

interference must be incorporated into analytical performance studies. Spectral

shaping of the transmitted signal requires advanced techniques for digital filter

design. The research benchmarks AF relays with DF relays taking into account

the aforementioned issues. We cooperate with High-frequency and microwave

engineering group to gain understanding of the actual propagation

environment and loop-back interference with full-duplex relays

Full-duplex infrastructure relays


5/56


6/56

resources into orthogonalresources and results in an orthogonalization of the

transmissions and receptionsperformed by a wireless device. Consequently, all

currently deployed wireless devicesoperate in half-duplex fashion, where same

frequency simultaneous transmission andreception of signals is not possible.

The key challenge in achieving full-duplex wireless communications, where a

devicecan transmit and receive signals over-the-air at the same time and in the

samefrequency band, is the large power differential between the self-

interference createdby a devices own wireless transmissions and the received

signal of interest comingfrom a distant transmitting antenna. This large power

differential is due to the factthat the self-interference signal has to travel much

shorter distances than the signalof interest. The large self-interference spans

most of the dynamic range of the Analogto Digital Converter (ADC) in the

received signal processing path, which in turn dramaticallyincreases the

quantization noise for the signal-of-interest. Thus to achievefull-duplex it is

essential to mitigate the self-interference of thereceivedsignal is the same

channel. Hence,spectrum resources are utilized efficiently but as a downside

therelay is subject to loop interference(LI) due to signal leakagefrom the relays

transmission to its own reception. The earlierliterature often pessimistically

sees this self-interference as aninsurmountable problem, and resorts to the

half-duplex modeby allocating separate time slots or frequency bands for

relayreception and transmission. This is a simple way to avoid interferenceby

splurging spectrum.

1.3 Problem Outline


7/56

There are so many methods proposed to mitigate the loopback self interference

of Relays in full duplex communication. The previous approach used the

concept of time domain cancellation to mitigate the self interference; the main

drawback of this Time domain Cancellation is its blindness to the spatial

domain, e.g., low rank of channel matrix is not expected to result in better

isolation. Additionally, the scheme is sensitive to both channel estimation noise

and transmit signal noise. In fact, TDC adds a new signal in the relay input

whichmay actually lead to degraded isolation compared to pure

naturalisolation with high channel estimation noise. The one and only

advantageof time-domain cancellation is that it does not distort thedesired

signal or reduce the input and output dimensions of therelay.

1.4 Objective:

This work focuses on technical problemin full-duplex relaying: How to mitigate

the loop interferenceefficiently?For investigating and comparing several

solutions, our main motivation is to improve the spectral efficiency ofrelay

systems by avoiding the need of two channel uses for one end-to-end

transmission that is inherent for half-duplex relays. Throughout the work, the

mitigation schemes are categorized into three subtypes: A) natural isolation, B)

time-domain cancellation, and C) spatial suppression. Consequently, we aim at

showing that the loop interference can be mitigated sufficiently and, thus, the

full-duplex mode becomes a feasible and viable alternative for the half-duplex

mode.


8/56

1.5 Methodology:

In this work we set up a generic system model that explicitlyaccounts for the

loop interference, the relay processingdelay, and the imperfections of the side

informationexploited in the mitigation of loop interference. Thesespecifics are

important we summarize Natural isolation that is needed in order toavoid relay

receiver saturation, and digital MIMO time-domaincancellation, which

generalizes theschemes used in SISO repeaters. We propose and analyze

novelspatialsuppression schemes based on antenna selection, beam

selection,null-space projection, and minimummean square error (MMSE)

filters. For every scheme, we explicitly the minimize the self-interference as the

optimization target and providegeneral solutions for the optimal filters.

Examples 14show why [25][28] present only simplified or suboptimalspecial

cases for some of our general schemes and we introduce the combination of

timedomaincancellation and spatial suppression for reducingthe effect of

imperfect side information in mitigation. The mitigation schemes are compared

extensivelywith simulations on bit-error rate and isolationimprovement. The

results verify that the loop interferencecan be mitigated significantly or even

eliminatedcompletely in the ideal case, but, in practice, there willbe some weak

residual interference due to imperfect sideinformation used for mitigation.

Cellular, Wi-Fi and Bluetooth networks are arguably the four mostcommonly

used wireless networks. Out of these four networks, the first one to be

deployedwas the cellular network, which operates at distances in the order of

kilometersand uses mobile devices which transmit at powers close to 30 dBm.


9/56

For these values oftransmit powers and distances between communicating

devices, it seemed unfeasibleto cancel the self-interference enough to enable

full-duplex wireless communications.

1.6 Thesis outline

The complete thesis of the above work is outlined in six chapters:

Chapter1 gives the basic introduction about the communication system and

the uses of relays in Duplex communication. It also gives the basic problem in

the last approach and as solution for that problem.

Chapter2 gives the complete literature survey of the project

Chapter3 gives the basic system model used in full Duplex communications

using relays; it also gives the mathematical representation of the signals

(original and interference).

Chapter4 provides the information about the previously used design

methodologies for the mitigation of self interference and also gives the Newly

proposed mitigation algorithms.

Chapter5 gives the performance evaluation of the proposed approach and also

gives the comparison results between the proposed approach and previously

proposed approaches.

Finally chapter6 gives the conclusions of the work


10/56

CHAPTER 2

LITERATURE OUTLINE

The literature on MIMO relaying can be classified as follows based on how the

self-interference problem is treated: 1) earlier papers, e.g., [1][11], consider

half-duplex relaying in which the loop interference is inherently avoided. Some

papers, e.g., [1][6], develop half-duplex protocols for the case in which the

direct source-destination link is blocked. Our results are directly applicable for

the full-duplex counterparts of these systems and enable more spectrally-

efficient implementation once the loop interference is appropriately mitigated.

The other papers, e.g., [7][11], exploit the direct link as an extra diversity

branch. The direct link is orthogonal by design in the half-duplex mode

whereas the destination receives superposition of the direct and relayed

transmissions in the full-duplex mode. Also for these systems, the full-duplex

counterparts are feasible with proper signal separation in the destination.

2) Some information theory-oriented papers, e.g., [10][19], study various full-

duplex relaying schemes without considering the deleterious effect of the loop

interference albeit otherwise presenting many seminal contributions. In

particular, these papers tend to provide minimal (if any at all) explanations and

references for the mitigation of the loop interference. Our results will support

this body of literature by providing validation and a retroactive reference for

a central baseline assumption not verified in detail before. 3) The smallest

group of earlier papers accounts explicitly for the effect of the loop interference


11/56

in full-duplex relaying. The early results consider exclusively single-

inputsingle-output (SISO) repeaters, see, e.g., [20][23]. For full-duplex MIMO

relays with loop interference, our literature search elicited preliminary ideas

[24], [25] and recent studies [26][31] conducted in parallel with our work

reported first in [32]. These papers tackle the problem of loop interference

mitigation in a limited scope, e.g.,restricting the system to support only one

spatial streamor providing suboptimal solutions. Moreover, the relay

processing delay is neglected in [28][30], which, asdiscussed in [32], renders

the relay practically impossibleto implement or makes the loop interference not

harmful.Reference [31] studies loop channel estimation in full-duplexMIMO

relays and, thus, supports our analysis whichstarts presuming that such side

information is alreadymade available.

Analog SISO repeaters have been employed for a long time incellular networks.

Instead, we consider modern, sophisticated,digital relays that are capable of

baseband signal processing,and, in particular, employ multiantenna

techniques. Some prototypesof full-duplex MIMO relays have already been

developed,see, e.g., [33] and [34]. After mitigating the loop interference

asshown herein, they become a viable solution to transparentlyboost the

coverage of future cellular systems.

2.2 Relay Communication

Relay communication refers to the technology that the communication

between the source and the destination is established or enhanced by one or

more than one relays. The relays can be dedicated relay stations that are built


12/56

to support the wireless link, or other mobile users that are selected to facilitate

the data transmission from the source to the destination. Relay communication

is also termed cooperative communication, which we use in this dissertation

without distinction.

From the network aspects, traditional wireless communication systems,

e.g., cellular mobile communication networks, are centralized. There, the

transmission scenarios are point-to-point (single user), one-to-many

(broadcast) or many-to-one (multi-access), which can all be categorized as

single-hop transmission. However, in relay communication, the transmission of

information from the source to the destination consists of at least two-hops.

Thismulti-hop transmissionmodel makes the relay communication scenarios

more versatile and the research on it more difficult. Although relay

communication is still a young research topic, many important results have

been achieved, which makes it a fertile field of research. We summarize the

important relaying strategies and the state-of-the-art research results in this

section

2.2.2 Advantages and Challenges of Relay Communication

Relay communication can provide the benefits that traditional single-hop

communication cannot achieve in many practical scenarios. Relay

communication has drawn wide interests from both academia and industry

[35]. For practical systems, we summarize the advantages of relay

communication over single-hop communication as follows.


13/56

Combating signal attenuation The adverse effects of wireless channels

include pathloss, shadowing and fading effects. The signal strength decays

exponentially with the distance between the source and the destination. When

the distance between the source and destination is too large, the signal

attenuation becomes too high due to pathloss, which makes it impossible for

the source and destination to communicate. By placing relays between the

source and the destination, the distance between the source and the relay and

the distance between the relay and destination is shortened. As a result, the

signal strength can be boosted a lot. Moreover, due to the signal strength

improvement, the source can use higher modulation symbol alphabets to

transmit more data in each channel use. In this way, relaying technology not

only increases the coverage of the system, but also improves the data rate

transmitted to the users.

Combating shadowing effects In large cities and hilly areas, tall buildings and

mountains typically block signals transmitted from the source to the

destination. Such effect is called shadowing. Relays provide another path to

circumvent the obstruction. In those scenarios, relaying is maybe the only way

to provide services in shadowing environments.

Combating fading effects The fading effects arise due to the multipath

propagations that lead to the fluctuations in received signals. Diversity is an

effective way to combat the signal fluctuation due to the fading effects. The

cooperative diversity introduced by the cooperative communication brings


14/56

higher link reliability to the users [36,37], where multiple independently faded

signals from the source and the relay are combined at the destination.

Low costFuture cellular communication systems will move to higher frequency.

As a result, the coverage of each cell will shrink a lot compared to present

cellular communication systems. Building more base stations can be the

solution, but the cost of building those base stations will be very high. A low-

cost alternative will be building relays to extend the coverage of each cell. Thus

relay communication provides low-cost solutions for future generation wireless

communication systems.

Infrastructure-less network: In traditional cellular networks, the whole

system operation depends on the centralized control, e.g., from the base

station. However, in military services or due to the disasters like earthquakes,

infrastructure-less networks such as ad hoc networks are preferable. Such

networks do not rely on a preexisting infrastructure such as dedicated routers

or base stations. Instead, each node participates in the routing by forwarding

data for other nodes. That is each node can act as a relay, and the choice of

relay nodes are determined dynamically based on the network connectivity.

Despite all those benefits that may be available by incorporating cooperative

communication into future wireless communication systems, there are also

challenges for implementing cooperative communications. Those challenges

include:


15/56


16/56

interference to other users. Furthermore, sending data to the destination via

relays leads to increased traffic for the whole system.

Spectral efficiency loss The major problem of current relays is that they

cannot transmit and receive data using the same time-frequency channel. This

half-duplex constraint leads to the spectral efficiency loss compare to direct

transmissions.

2.3 Relaying Strategies

There are three relaying strategies discussed by Peters et. al [38] which are

one-way relaying, two-way relaying and shared relaying as illustrated in the

following 2 [6]. As shown in the figure, the eNodeB is equipped with one

antenna per sector and one RN serving a single UE in its vicinity. On the other

hand, the relay station nodes are shared between eNodeBs of three adjacent

cells which use the same frequency.

The concept of one-way relaying is illustrated in the following Fig.2.3. The

datatransmission is divided into four frames as denoted by the number: In the

downlink, 1) the eNodeB transmits to RN, followed by 2) RN forwards the signal

to UE. Then, during uplink, 3) UE transmits to RN and finally 4) RN forwards

UEs signal to eNodeB.

As an enhancement to one-way relaying, two-way relaying is more efficient

where the data transmission is done in two phases as shown in Fig.2.4. During

the first phase, both eNodeB and UE transmit their signals to the RN and then

in second phase, after proper signal processing, the RN forwards the signals to


17/56

both eNodeB and UE. Therefore, the transmission duration would be half of the

time taken for one-way relaying.

Shared relaying is cost-saving as number of RNs to be deployed is reduced by

allowing the RN to be shared by three cells. Also, as mentioned in [38], shared

relay has advantage over one-way relaying compared to two-way relaying. This

is due to the interference that might occur during the simultaneous

transmissions of two-way relay, combining with the fact that the shared relay

itself has to handle the multiple signals from eNodeBs of the three adjacent

cells.

a) One-way and two-way relaying


18/56

b) Shared relaying

Fig.2.2. Relaying strategies with frequency reuse of factor 6 where each cell is

divided into 6sectors. a) Frequency reuse pattern for one-way and two-

wayrelays deployed in one cell. b)Frequency reuse pattern for shared relay

deployed in 3 adjacent cells.


19/56

Fig.2.3. One-way relaying

Fig.2.4. Two-way relaying

2.4 Relay Transmission Schemes

Over the past decade, numerous relay transmission schemes have been

developed tobe implemented in our cellular network technology. In [38], the

transmissiontechniques include:


20/56


21/56

real signalswith noise and interference. Thus, those undesired signals are also

amplified andretransmitted along with the original signals.

Another relaying strategy is decode-and-forward where the signals are decoded

bythe relay node, re-encoded and lastly forwarded to desired destination. In

this relayingstrategy, noise and interference are discarded from being

transmitted together with thereal signals but with the price of longer delay due

to decoding and re-encodingprocess. The relay structures can be categorized

into Layer 2 (L2) relay and Layer 3(L3) relay, depending on its function. The

transmissions involved can be both inbandand outband as well, as in L1 relay.

In the later chapters we will see the in brief about the relaying strategies.


22/56

CHAPTER 3SYSTEM MODEL

Relaying, i.e., multihop communication, is a promisingtechnique to provide

lower transmit powers, higherthroughput and more extensive coverage in

future wirelesssystems. Likewise, single-hop multiple-input multiple-output

(MIMO) transmission has attracted wide research interest, andemerging

wireless systems utilize extensively MIMO techniquessuch as spatial division

multiplexing. Hence, if relaysare used, they need to be equipped with antenna

arrays aswell to avoid a key-hole effect, i.e., squashing multiple spatialstreams

through a rank-one device. This paper focuses on thecombination of MIMO and

relaying techniques and developsnew baseband signal processing techniques to

improve spectralefficiency.

An essential classification of relaying techniques is betweenfull-duplex and

half-duplex operation modes. In fact, the choiceof the operation mode is a

fundamental tradeoff between spectralefficiency and self-interference. A full-

duplex relay receivesand transmits at the same time on the same channel.

Hence,spectrum resources are utilized efficiently but as a downside therelay is

subject to loop interference (LI) due to signal leakagefrom the relays

transmission to its own reception. The earlierliterature often pessimistically

sees this self-interference as aninsurmountable problem, and resorts to the

half-duplex modeby allocating separate time slots or frequency bands for

relayreception and transmission. This is a simple way to avoid interferenceby

splurging spectrum.


23/56

3.1 System model:

Let us consider the generic wireless multihop network illustratedin the upper

left corner of Fig. 1. The network comprisesnodes operating in both half-duplex

and full-duplex modes andit is not restricted to any specific multihop routing

protocol ormultiple access strategy for the simultaneous transmissions.Wethen

focus on two-hop communication through any full-duplexrelay (R) node from a

set of source (S) nodes to a set of destination(D) nodes as illustrated in the

lower right corner of Fig. 1.The full-duplex relay receives and transmits

simultaneously onthe same frequency which necessitates to model explicitly

theresulting loop interference (LI) signal.The sources and the destinations have

in total transmitand receive antennas, respectively, and the relay is

equippedwith receive and transmit antennas. Before applyingmitigation

techniques the relay is likely implemented with spatiallyseparated receive and

transmit antenna arrays which constitutes natural isolation. However, the

followingresults are also applicable in full-duplex relaying with asingle antenna

array which is optimistically considered in [26].We set in this special

case.

3.1.1. Signal Model

The signal model is built upon frequency-flat block-fadingchannels as in the

majority of related papers, see, e.g., [1][19],[25], [26], [28][30]. This implies

that the system exploits


24/56

Fig.3.1. A wireless multihop network containing a full-duplex relay subject to

loop interference.

orthogonal frequency division multiplexing (OFDM) for broadbandtransmission

over multipath channels, and the signalmodel represents a single narrowband

subcarrier.

For time instant , let matrices , , and

represent the respectiveMIMO channels from all sources to

the relay, from the relayoutput to the relay input, and from the relay to all

destinations.

The sources transmit the combined signal vector ,and the relay

transmits signal vector while it simultaneouslyreceives signal

vector . This createsa feedback loop from the relay output to the

relay input throughchannel .


25/56

The relaying protocol is denoted by the generic function

which generates an output sample based on the sequence ofinput samples and

causes integer processing delay . Wefocus on the mitigation of the loop

interference and, thereby,keep the proposed schemes transparent and

applicable with most of the readily available relaying protocols. Remark 1: The

processing delay is strictly positive because we consider wideband

transmission over multipath channels in contrast to [28][30]. In particular,

omitting the delay causes severe causality problems in the practical

implementation of relaying protocols: It is impossible to process a subcarrier

and retransmit the OFDM symbol before the respective OFDM symbol is first

completely received and demodulated. Furthermore, the loop signal may not be

harmful at all with zero processing delay because the relay transmission only

amplifies the same input signal. See [32] for more discussion on the

consequences of neglecting the processing delay. Finally, the respective

received signals in the relay and in thedestinations can be expressed as

where and are additive noisevectors in the relay

and in the destinations, respectively. Allsignal and noise vectors have zero

mean. Signal and noise covariancematrices are denoted by ,


26/56

, and . For clarity, we willomit the time

indices in the rest of the work.

3.1.2. Side Information for Mitigation Techniques

We consider mitigation techniques that can be implementedtransparently, i.e.,

using only information that the relay is expectedto know by design or is able to

measure by itself. In otherwords, mitigation may exploit knowledge of only

, and . However, we assume that the available side informationis

degraded due to the following non-idealities which manifestthemselves in the

form of noise. In this paper, the noise is assumedto be completely unknown for

the mitigation schemeswhile some additional information such as the

covariance ornorm bounds of the errors could facilitate a robust approach.

1) Channel Estimation Noise:The relay may exploit anyoff-the-shelf technique

or one of the schemes developed specificallyfor full-duplex relays [23], [31] to

obtain the respectiveestimates and of and .We model the

practicallynon-ideal estimation process by defining estimation noises and

such that the estimates differ from the truechannel values:

All elements of and are assumed to be independent(both mutually

and from the corresponding channels)circularly symmetric complex Gaussian

random variables. Thevariance of the estimation noise is defined by relative

estimation error such that


27/56

for all i,j. Analogous relation holds between and .

2) Transmit Signal Noise:The relay knows perfectly the digitalbaseband signal

it generates, but the actual transmittedsignal cannot be exactly known. This

is because any practicalimplementation of conversion between baseband and

radiofrequency is prone to various distortion effects such as carrierfrequency

offset, oscillator phase noise, AD/DA conversion imperfections,I/Q imbalance,

and power amplifier nonlinearityamong others.We model the joint effect of all

imperfections byintroducing additive transmit distortion noise such that

Furthermore, we model all elements of with independentzero-mean random

variables, and define their variance with relativedistortion . The covariance

matrix of the transmit noisebecomes

We assume that and are uncorrelated which implies that

in which


28/56

CHAPTER 4

MITIGATION OF LOOP INTERFERENCE

In migitation of loop interference in full duplex MIMO relays we first decouple

the mitigation of loop interferencefrom the design of the relaying protocol and


29/56

develop solutionsthat transform the relay to an

equivalentinterference-free relay. Here and

represent the input and output dimensions (or the number of spatialstreams)

reserved for the relaying protocol.

The target is to make residual loop interference so infinitesimalthat it can be

regarded simply as additional relay inputnoise. Thus, we transform the signal

model from (2) to

where and are the respective receiveand transmit signal

vectors of the equivalent interference-freerelay , and

represent therespective equivalent MIMO channels from all

sources to theinterference-free relay and from the interference-free relay to

alldestinations, and is the equivalent receiver noisevector including

all residual loop interference after mitigation.

The covariance matrix of is .

Remark 2:The equivalent interference-free relay appliesrelaying protocol

to obtain from according to (1). By decouplingthe mitigation from the

protocol, the relay may adopt,directly or after minor modifications, any of the

protocols designedfor cases without loop interference in [1][19]. However,the

system setup or the relaying protocol may still affect thechoice of and .

4.1 Reference Mitigation Schemes


30/56

4.1.1) Natural Isolation:The relay installation should guaranteesome natural

isolation (represented by ) to facilitate theusage of signal processing

techniques which provide additionalman-made isolation. This is because, in

practice, the dynamicrange of the relay receiver circuitry is limited, and, thus,

largedifference in power levels may saturate the receiver renderingany attempt

to recover the desired signal futile.With separated receive and transmit antenna

arrays, naturalisolation arises from the sheer physical distance between the

arrays,and rational installation guarantees obstacles in betweenthe arrays to

block the line-of-sight. For this purpose, the installationmay exploit

surrounding buildings or add a shieldingplate [22]. Furthermore, antenna

elements can be directionaland pointed at opposite directions [20], [22], and

their polarizationsmay be orthogonal. If the same antenna array is used forboth

receiving and transmitting as in [26], all natural isolationcomes solely from the

duplexer connecting the input and outputfeeds to the same physical antenna

element. However, isolationoffered even by the most high-end duplexers may

not be sufficientfor communication.

Exploiting the signal model from (2), the mean square error(MSE) matrix of the

relay input signal is given by


31/56

Fig.4.1. Time-domain loop interference cancellation in a full-duplex MIMO

relay by subtracting an estimate of the loop signal.

which yields the loop interference power as

In the following, we presume that all means of improving naturalisolation have

been first exploited and then concentrate onsignal processing techniques to

mitigate the residual interference,i.e., the effect of . Measurements

show that naturalisolation is not often sufficient alone [20], [22], [34].

Hence,our study excludes the exceptional setups in which natural isolationis

large without any additional mitigation, e.g., a relaywith the receive array

placed outdoors and the transmit arrayproviding underground coverage in a

tunnel.


32/56

4.1.2) Time-Domain Cancellation (TDC): Cancellation is basedon the

reasonable presumption that the relay always knows itsown transmitted signal

at least approximately. If the relay canalso determine the loop channel, the

interference signal maybe replicated and removed from the received signal. In

practice,the relay may apply conventional analog precancellationto improve the

feasibility of the digital mitigation techniques forlower dynamic range. However,

the implementation of the electronicsbecomes expensive and difficult if the

respective circuitis more sophisticated than a phase shifter that removes one

(ideallythe strongest) multipath component.

The considered TDC scheme is a straightforward MIMO extensionfor earlier

SISO schemes [21][23], implemented as illustratedin Fig. 2 (similar structures

are used in [28][32]):

The relay contains a feedback loop with MIMO cancellationfilter .

Thus, (2), (5), and (7) can be related as

and

The equivalent receiver noise vector of the interference-free relay becomes

in which the residual loop interference channel is


33/56

Fig.4.2. Spatial loop interference suppression in a full-duplex MIMO relay by

using linear receive and transmit filters.

The MSE matrix of the relay input signal becomes

The first term includes the channel estimation error and thesecond term arises

due to the transmit signal noise. Cancellationcan only minimize the known

part of the first term by choosing which results in .

Thereby, (12) yieldsthe residual interference power as

If cancellation is not used, i.e., , (12) reduces to (8).


34/56

The main drawback of TDC is its blindness to the spatial domain,e.g., low rank

of is not expected to result in betterisolation. Additionally, the scheme is

sensitive to both channelestimation noise and transmit signal noise

as shownby (13). In fact, TDC adds a new signal in the relay input whichmay

actually lead to degraded isolation compared to pure naturalisolation with high

channel estimation noise. The advantageof time-domain cancellation is that it

does not distort thedesired signal or reduce the input and output dimensions of

therelay, i.e., and .

4.2. Novel Spatial Suppression Schemes

To exploit the extra degrees of freedom offered by the spatialdomain, we

propose that the relay applies MIMO receive filter and MIMO

transmit filter as illustrated in Fig. 3. Now (2), (5), and (7) can be

relatedas

and .

Throughout the work, we normalizefilter gains to and

with allschemes.

The equivalent receiver noise vector of the interference-free relay becomes

in which the residual loop interference channel is

Based on (14), the residual interference power is given by


35/56

Interference can be suppressed by designing and tominimize the first

term and/or by designing only to minimizethe second term that is due to

transmit signal noise. Since(15) is a matrix equation, it needs to be first

translated into ascalar value before formulating an optimization problem: In

thispaper, the Frobenius norm is adopted for this purpose whileother metrics,

rendering different optimization targets, are alsoavailable. On the other hand,

(16) for spatial suppression is reducedto mere natural isolation given in (9)

when and .

Spatial suppression comes at the cost of a reduction in theinput or output

dimensions comparing to TDC. However, (14)reveals readily one significant

advantage over cancellation: thereceive filter can be designed to suppress

the potential loopinterference that is due to the transmit signal noise .

The implementation differs depending on the procedure:

Independent design: One filter is designed without knowledgeof the other

filter which can be replaced byI.

Separate design: One filter is designed given the other.

Joint design: The filters are designed together.

Next we consider these procedures with antenna selection, beamselection, null-

space projection, and MMSE filtering.


36/56

4.2.1) Antenna Selection (AS):The simplified receive antennaselection scheme

studied in [25] inspires us to formulate loopinterference suppression based on

generalized antenna subsetselection. To that end, the respective receive and

transmit filtersare implemented with row and column selection matrices

(seeSection I-C) that are scaled to normalize the gains:

To reduce the gain of the residual loop interference channelgiven in (15), we

define the objective for suppression as

decreasing the known part of . By substituting (17)

The optimal joint filter design is found by calculating theFrobenius norm for all

combinations and choosingthe lowest. Although one may easily

devise suboptimal methodsof lower complexity, only global search gives the

exact optimumin the general case. However, it is feasible because the numberof

antennas is in practice reasonably small.

Let us then consider the design of to illustrate the separatefilter design (the

procedure is symmetric for designing ).Now needs to be first fixed

based on any spatial suppressionscheme and the unique solution for

issimply one of the combinations. If the transmit filter is


37/56

not known (as in independent filter design) or not yet selected,one can

substitute , which reduces the objective to .

Example 1: The scheme of [25] is limited to the special caseof and

, i.e. , :When , is

minimized by selecting

Table 1

Algorithm for optimal joint beam selection

---------------------------------------------------------------------------------------------

Design and to select rows and columns of as follows

Step1: Select in total min { + }, max { }} rows and columns such that

all combinations pick only off-diagonal elements of. For this sub solution

=0.

Step2: To satisfy objective. Select the rest of the rows and columns such that

the final selection matrices pick only the + -max{ smallest singular

values.

-------------------------------------------------------------------------------------------------

if and otherwise. In thegeneral MIMO case of any , and

, such singlecomparison is not sufficient for the optimal filter design that

issolved in the above paragraphs for different variations.

4.2.2) Beam Selection (BS):General (eigen)beam selection isbased on the

singular value decomposition (SVD) of


38/56

in which submatrices and contain the basis vectorsassociated with

zero singular values. The diagonal matrix comprises the singular

values, , sorted in descending order.

By choosing beam selection matrices as

the objective is transformed from (18) to

as , by definition. Filter design becomesconceptually similar

to AS, but row and column selection isbased on the effective diagonal channel

instead of .

Remark 3: Objective readily indicates that BS is superiorto AS. In (22) most row

and column combinations pickoff-diagonal elements of that are zero by

definition leadingto for many subsolutions whereas in (19) all

elementsof are practically nonzero, i.e. , for all subsolutions.In

other words, AS is optimal only when limiting thesearch space to binary

selection matrices while BS solves theoptimization target with general complex

matrices.Intuitively, the optimal joint BS could be solved by testingall

combinations as with AS. However, the diagonalizedstructure of the effective

loop channel facilitates directoffline selection based on


39/56

and such thatonly the SVD is calculated for each channel representation:

Theoptimal joint selection is obtained with the algorithm given inTable I.

If , Step 2 is omitted andBS reduces to null-space

projection discussed in the next section.On the other hand, separate and

independent filter designsapply only Step 2 for all rows and columns.

Let us assume that in thefollowing. One

straightforward illustrative solution can be obtainedwith the optimal joint BS

algorithm as

in which , and are identity matrices of ,

and dimensions, respectively. In fact,the optimal joint BS algorithm

translates (22) to

For the general case, this shows that BS may cause residual loopinterference

even if the side information is perfect. In the nextsection, this motivates to

consider the special cases of beam selectionthat ideally eliminate all

interference.

Example 2: Compared to our general BS solution, the schemeof [28] is not only

suboptimal but also limited to the specialsymmetric case of and

: The beams areselected by in which is anzero


40/56

matrix andI is an identity matrix. Thispicks the smallest

singular values transforming (22) to

This is larger than (24) obtained with (23) because the schemedoes not exploit

the possibility to suppress interference by kicking the off-diagonal elements of

in Step 1. The suboptimalitycan be also interpreted to be the consequence

ofindependent filter design instead of joint design.

4.2.3) Null-Space Projection (NSP): Next we develop spatialsuppression

schemes that can eliminate all loop interference inthe ideal case with perfect

side information similarly to TDC.This is desirable when the loop interference is

dominating butAS or general BS does not offer sufficient attenuation.

In null-space projection, and are selected suchthat the relay receives

and transmits in different subspaces,i.e., transmit beams are projected to the

null-space of the loopchannel combined with the receive filter and vice versa.

Thecondition can be stated for joint or separate filter design as

to eliminate the known part of the first term in (16). Similarly,for suppressing

the transmit signal noise, the condition becomes , partly eliminating

the second term in (16).

One solution for joint NSP can be obtained with the optimaljoint BS algorithm

given in Table I, if , and are low enough w.r.t. and .


41/56

Firstly, atotal of beams are selected in Step 1 correspondingto

different singular values. Secondly, the lastterms in (24) are zero if

. Thus, input and output

beams maycorrespond to the same singular values after Step 2 and still, i.e.

, satisfying also the condition in(26). This proves that

the BS algorithm results in null-spaceprojection whenever

This condition defines also the general existence of joint NSP,if and

are additionally constrained to be of full rank.Even if is rank deficient,

is of full rank in practicedue to the estimation noise which also causes residual

loop interference.Thereby, the condition in (27) can be alternatively

evaluatedusing the anticipated value of based on prior informationor

by defining with a threshold below whichthe singular values are

rounded to zero.

Remark 4:

For the case , the total number for antennas is minimized

with NSP selecting

orwhen (full rank), or by selectingor

when

(minimum rank).


42/56

Choosing may be preferable due to transmit noise.For separate filter

design, let us recall that the Moore-Penrosepseudo inverse is unique,

always exists, and satisfies by definition. For designing

separately given , we can, thereby, apply projection matrix

If . Separate design for is given bya similar projection

matrix which is obtained by replacing and above with and

, respectively.

Example 3:The scheme of [27] is limited to the simple specialcase of

and : When , is

guaranteed directly by . In the general MIMO case of any

and , the optimal filter design, solved in the above

paragraphs,becomes more involved.For designing one filter independently, the

above schemesmay be exploited by setting the other filter to identity.

However, simpler design is obtained by choosing

because the row space of should be in the left null space of or by

choosing because the column space of should be in

the null space of .


43/56

Joint design solutions satisfying the NSP condition in (26)are not unique in

most cases. For example, Step 1 in the optimaljoint BS algorithm allows to

choose rows and columns in differentways. Furthermore, general BS inherits

the same propertyexcept that the subsolution picking the nonzero diagonal

valuesof is unique in Step 2. Selection between the solutions withthe same

cost can be done based on any other performance criterionas illustrated by the

next example.

Example 4: The scheme of [26] is limited to the case of

and : When the SVD of theloop channel is

,is guaranteed either by

, or by , . Although not

recognized in [26],also , can be used if .

Compared to Example 3, the extra receive antenna facilitatesadditionalselection

diversity available for reducing the effectof transmit signal noise. In our general

MIMO case of any , and , the optimal filter design becomes more

involvedand the applicability of NSP is governed by (27).

4.2.4) Minimum Mean Square Error (MMSE) Filtering:The previousspatial

suppression schemes aim at minimizing the effectof loop interference at the

cost of spatially shaping the usefulsignal which does not happen with TDC. In

order to reducethe effect of this drawback with spatial suppression, a

minimumMSE scheme is developed next to both minimize the distortionand

attenuate the loop interference.


44/56

Now and as with TDC. Thus, the

MSE matrix of the relay input signal is given by

(29)

Inwhich For separate filter design

given , the minimum MSE receivefilter is derived from the condition

yielding

Which needs to be scaled to satisfy . Note thatMMSE filtering

requires knowledge of and signal covariancematrices as opposed to the other

mitigation schemes.

Condition to minimize MSE at the transmitside reduces to the

condition for null-space projection givenin (26). Therefore, the evident order for

joint filter design isto firstly minimize interference at the transmit side using

anyscheme, and then secondly design the receive filter using (30).

4.2.5) Combining Cancellation and Spatial Suppression:

Time-domain cancellation suffers from residual interferencethat is due to the

transmit signal noise, while spatial suppressionmay need many extra antennas

for efficient mitigation. Hence,the combination could offer high isolation with

conservativenumber of antennas in the presence of transmit signal noise.


45/56

Combining yields the residual interference channel given by

, cf. (11) with mere cancellation and(15) with mere

suppression. Thus, filter design can be performedfor one scheme first if the

other scheme is consequently designedgiven the residual channel. The

implementation admitsfour variations for independent and separate filter

design [32],but we will now focus on joint filter design, which is possible inmost

scenarios.

The performance evaluation of the above presented work is illustrated in next

chapter


46/56

Chapter5Results

In this chapter, the performance evaluation of the proposed method is going to

be discussed. There is also discussion about the comparison results of the

proposed approach with the previously proposed approaches.. In the

simulations, all channels aremodeled with Rayleigh fading and the transmitted

signals areassumed to be spatially white with unit power per stream. The relay

receiver noise is white andGaussian with, and imperfect side information

usedin mitigation is generated as explained in above chapter.

0 5 10 15 20 25 300

5

10

15

20

25

30

1

2

3

4

5

6

7

8

9

10

11

12

13

1415

selected path for Communication


47/56

0 2 4 6 8 10 12 14 16 18 200

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

P?natural[db]

BER

natural isolation

TDC(.02,.02)

NSP(.02,.02)

TDC(0,.02)

NSP(0,0.02)

TDC(.02,0)

NSP(.02,0)

half duplex

0 2 4 6 8 10 12 14 160

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

antenna selection AS

F1?P(x)

3x4:0.94

2x4:1.88

3x3:2.20

1x4:3.25

2x3:3.63

2x2:5.75

1x3:6.18

1x2:10.05

1x1:22.63


48/56

0 2 4 6 8 10 12 14 160

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

x[db]

F1?P(x)

3x4:3.06

2x4:7.37

3x3:7.86

1x4:21.81

2x3:23.57

1 2 3 4 5 6 7 8 9

0

5

10

15

20

25

30

35

40

45

?H

??P1[db]

NSP,2x4

NSP,3x4

NSP,3x3

NSP,2x4

NSP,3x3

TDC,4x4

BS,3x3

BS,2x4

BS,3x3

BS,2x4

BS,3x4

BS,3x4

BS,3x4

1 2 3 4 5 6 7 8 90

5

10

15

20

25

30

35

40

45

et

??P1[db]

NSP,4x3

NSP,4x4

NSP,3x4

TDC,3x3

BS,4x4

BS,4x3

BS,4x4

BS,3x4

BS,4x3

BS,3x4


49/56

0 2 4 6 8 10 12 14 16 18 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

x[db]

F1?P(x)

BS(3):5.04

BS(2):8.15

MMSE(3):8.77

MMSE(2):14.76

TDC(1):25.20

TDC(2):25.30

TDC(3):25.34

MMSE(1):40.20

NSP(1):40.52

0 2 4 6 8 10 12 14 16 18 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

x[db]

F1?P(x)

BS(4):7.80

BS(3):11.56

TDC(1):25.25

TDC(2):25.32

TDC(3):25.35

TDC(4):25.36

NSP(2):29.67

both(4):34.49

both(3):34.49

both(2):34.49

both(1):36.81

NSP(1):40.51

0 2 4 6 8 10 12 14 16 180

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

x[db]

F1?P(x)

AS,3x4(3):1.29

AS,3x4(2):1.62

AS,4x4(4):2.20

AS,4x4(3):2.57

BS,3x4(3):5.04

BS,3x4(2):8.12

BS,4x4(4):7.86

BS,4x4(3):11.87


50/56

Chapter6CONCLUSION

Full-duplex MIMO relaying has large potential for spectrallyefficient wireless

transmission. In this work, we concentratedon solving the main associated

technical problem, i.e., the mitigationof relay self-interference. We extended the

earlier SISOcancellation schemes for the MIMO relay case and proposednew

solutions that suppress the interference in the spatial domain:antenna and

beam selection, null-space projection, andMMSE filtering. We also discussed

the issues that need to beconsidered when combining cancellation and spatial

suppression.Errors in the side information used for mitigation wereidentified as

the practical limitation to prevent complete interferenceelimination obtainable

in the ideal case. However, oursimulations illustrated that the proposed

schemes offer significantmitigation such that the residual interference may be

regardedas mere additional noise.


51/56

References:

[1] C.-B.Chae, T. Tang, R. W. Heath, Jr, and S. Cho, MIMO relaying withlinear

processing for multiuser transmission in fixed relay networks,IEEE Trans.

Signal Process., vol. 56, no. 2, pp. 727738, Feb. 2008.

[2] R. Zhang, C. C. Chai, and Y.-C. Liang, Joint beamforming andpower

control for multiantenna relay broadcast channel with QoSconstraints, IEEE

Trans. Signal Process., vol. 57, no. 2, pp. 726737,Feb. 2009.

[3] B. K. Chalise and L. Vandendorpe, MIMO relay design for multipoint-to-

multipoint communications with imperfect channel state information,IEEE

Trans. Signal Process., vol. 57, no. 7, pp. 27852796,Jul. 2009.

[4] Y. Rong, X. Tang, and Y. Hua, A unified framework for optimizinglinear

nonregenerative multicarrier MIMO relay communication systems,IEEE Trans.

Signal Process., vol. 57, no. 12, pp. 48374851,Dec. 2009.

[5] C. Li, X. Wang, L. Yang, and W.-P. Zhu, A joint source and relaypower

allocation scheme for a class of MIMO relay systems, IEEETrans. Signal

Process., vol. 57, no. 12, pp. 48524860, Dec. 2009.

[6] Y. Huang, L. Yang, M. Bengtsson, and B. Ottersten, A limited feedback

joint precoding for amplify-and-forward relaying, IEEE Trans.Signal Process.,

vol. 58, no. 3, pp. 13471357, Mar. 2010.

[7] O. Muoz-Medina, J. Vidal, and A. Augustin, Linear transceiver designin

nonregenerative relays with channel state information, IEEETrans. Signal

Process., vol. 55, no. 6, pp. 25932604, Jun. 2007.


52/56

[8] S. W. Peters and R. W. Heath, Jr, Nonregenerative MIMO relayingwith

optimal transmit antenna selection, IEEE Signal Process.Lett.,vol. 15, pp.

421424, 2008.

[9] S. Simoens, O. Muoz-Medina, J. Vidal, and A. del Coso, Compressand-

forward cooperative MIMO relaying with full channel state information,IEEE

Trans. Signal Process., vol. 58, no. 2, pp. 781791, Feb.2010.

[10] M. Yuksel and E. Erkip, Multiple-antenna cooperative wireless systems:

A diversity-multiplexing tradeoff perspective, IEEE Trans. Inf.Theory, vol. 53,

no. 10, pp. 33713393, Oct. 2007.

[11] S. Simoens, O. Muoz-Medina, J. Vidal, and A. del Coso, On theGaussian

MIMO relay channel with full channel state information,IEEE Trans. Signal

Process., vol. 57, no. 9, pp. 35883599, Sep. 2009.

[12] E. C. Van Der Meulen, Three-terminal communication channels,Adv.

Appl. Probab., vol. 3, pp. 120154, 1971.

[13] T. M. Cover and A. A. El Gamal, Capacity theorems for the relaychannel,

IEEE Trans. Inf. Theory, vol. 25, no. 5, pp. 572584, Sep.1979.

[14] B. Wang, J. Zhang, and A. Hst-Madsen, On the capacity of MIMOrelay

channels, IEEE Trans. Inf. Theory, vol. 51, no. 1, pp. 2943, Jan.2005.

[15] A. Hst-Madsen and J. Zhang, Capacity bounds and power allocationfor

wireless relay channels, IEEE Trans. Inf. Theory, vol. 51, no. 6,pp. 20202040,

Jun. 2005.

[16] G. Kramer, M. Gastpar, and P. Gupta, Cooperative strategies and capacity


53/56

theorems for relay networks, IEEE Trans. Inf. Theory, vol. 51,no. 9, pp. 3037

3063, Sep. 2005.

[17] Z. Zhang and T. M. Duman, Capacity-approaching turbo coding

anditerative decoding for relay channels, IEEE Trans. Commun., vol. 53,no.

11, pp. 18951905, Nov. 2005.

[18] Y. Liang, V. V. Veeravalli, and H. V. Poor, Resource allocation forwireless

fading relay channels: Max-min solution, IEEE Trans. Inf.Theory, vol. 53, no.

10, pp. 34323453, Oct. 2007.

[19] V. R. Cadambe and S. A. Jafar, Degrees of freedom of wireless networks

with relays, feedback, cooperation, and full duplex operation,IEEE Trans. Inf.

Theory, vol. 55, no. 5, pp. 23342344, May 2009.

[20] W. T. Slingsby and J. P. McGeehan, Antenna isolation measurements

for on-frequency radio repeaters, in Proc. 9th Int. Conf. AntennasPropag., Apr.

1995, vol. 1, pp. 239243.

[21] H. Hamazumi, K. Imamura, N. Iai, K. Shibuya, and M. Sasaki, Astudy of a

loop interference canceller for the relay stations in an SFNfor digital terrestrial

broadcasting, in Proc. IEEE Global Telecommun.Conf., Nov. 2000, vol. 1.

[22] C. R. Anderson, S. Krishnamoorthy, C. G. Ranson, T. J. Lemon, W.

G.Newhall, T. Kummetz, and J. H. Reed, Antenna isolation,

widebandmultipath propagation measurements, and interference mitigation

foron-frequency repeaters, in Proc. IEEE SoutheastCon, Mar. 2004, pp.110

114.


54/56

[23] K. M. Nasr, J. P. Cosmas, M. Bard, and J. Gledhill, Performance ofan echo

canceller and channel estimator for on-channel repeaters inDVB-T/H

networks, IEEE Trans. Broadcasting, vol. 53, no. 3, pp.609618, Sep. 2007.

[24] D. W. Bliss, P. A. Parker, and A. R. Margetts, Simultaneous transmission

and reception for improved wireless network performance, inProc. IEEE 14th

Workshop on Statist. Signal Process., Aug. 2007.

[25] A. Hazmi, J. Rinne, and M. Renfors, Diversity based DVB-T in-

doorrepeater in slowly mobile loop interference environment, in Proc. 10thInt.

OFDM-Workshop, Aug.Sep. 2005.

[26] H. Ju, E. Oh, and D. Hong, Improving efficiency of resource usagein two-

hop full duplex relay systems based on resource sharing andinterference

cancellation, IEEE Trans. Wireless Commun., vol. 8, no.8, pp. 39333938,

Aug. 2009.

[27] B. Chun, E.-R. Jeong, J. Joung, Y. Oh, and Y. H. Lee, Pre-nulling forself-

interference suppression in full-duplex relays, in Proc. APSIPAAnn. Summit

and Conf., Oct. 2009.

[28] P. Larsson and M. Prytz, MIMO on-frequency repeater with self-

interferencecancellation and mitigation, in Proc. IEEE 69th Veh.

Technol.Conf., Apr. 2009.

[29] J. Sangiamwong, T. Asai, J. Hagiwara, Y. Okumura, and T. Ohya,Joint

multi-filter design for full-duplexMU-MIMO relaying, in Proc.IEEE 69th Veh.

Technol. Conf., Apr. 2009.


55/56


56/56

[38]. Peters, S.W., Panah, A.Y., Truong, K.T., Heath Jr., R.W.: Relay

Architectures for 3GPPLTE-Advanced. EURASIP Journal on Wireless

Communications and Networking (2009)

[39]. Yang, Y., Hu, H., Xu, J., Mao, G.: Relay Technologies for WiMAX and LTE-

AdvancedMobile Systems. IEEE Communications Magazine 47(10), 100105

(2009)

[40]. Lo, A., Niemegeers, I.: Multi-hop Relay Architectures for 3GPP LTE-

Advanced. In:Proceedings of the 9th IEEE Malaysia International Conference

on Communications(MICC), Malaysia, pp. 123127 (2009)

ssp-mt-22 doc.docx

Documents