Vous êtes sur la page 1sur 74

Technische Universität München

Lehrschul für Hochfrequenztechnik

Prof. Dr. Techn. Peter Russer

Master Thesis
Complexity-Reduced Wideband Beamforming

Ahmet Coskun

25.11.2004

Supervised by:

Prof. Dr. Techn. Peter Russer


Dr. Tuan Do-Hong

This is a technical report of master thesis carried out at the Technische Universität München in

Munich, Germany. This is in partial fulfillment of the degree leading to the Masters of Science in

Microwave Engineering. I would like to express my thanks to my supervisor Dr. Tuan Do-Hong

extensively for the wonderful help, guidance and support he provided throughout my thesis. I am also

grateful to Mr. Biscontini for his support.

2
In memory of my grandfather

3
4
Contents

List of Figures.......................................................................................................................6
List of Tables.........................................................................................................................7
Abstract.................................................................................................................................8
Chapter 1.............................................................................................................................10
1.1. Smart Antennas for Wireless Communication Systems.............................................10
1.1.1. Smart Antenna Classifications................................................................................11
1.1.2. Benefits of Smart Antennas....................................................................................14
1.1.3 Applications of Smart Antennas..............................................................................20
1.2. Antenna Arrays .........................................................................................................25
1.2.1. Basic Antenna Array Parameters............................................................................26
1.2.2. Linear Arrays..........................................................................................................30
1.2.3. Pattern Multiplication.............................................................................................32
1.2.4. Planar Arrays...........................................................................................................33
Chapter 2.............................................................................................................................35
2.1. Introduction................................................................................................................35
2.2. Continuous Apertures.................................................................................................35
2.3. Discrete Apertures......................................................................................................38
2.4. Sparse Linear Arrays..................................................................................................40
2.4.1. Minimum Redundancy and Minimum Hole Arrays...............................................41
2.4.2. Simulation Results..................................................................................................43
Chapter 3.............................................................................................................................57
3.1.Introduction.................................................................................................................57
3.2. Beamforming..............................................................................................................57
3.2.1. Digital Beamforming vs. Analog Beamforming.....................................................57
3.2.2. Narrowband Beamforming vs. Wideband beamforming........................................59
3.3. Frequency-Invariant Wideband Beamforming............................................................63
3.4.Frequency-Domain Frequency-Invariant Beamformer (FDFIB)...................................65
3.4.1.Simulation Numerical Results.................................................................................66
Conclusion..........................................................................................................................69
References...........................................................................................................................71

5
List of Figures

6
List of Tables

7
Abstract

With the exponential growth of the demand for wireless communication systems, people are faced with

two main problems: Capacity and coverage area. Smart antennas offer many ways to improve

wireless system performance because of having the capability to overcome the major impairments

(multipath fading, delay spread, co-channel interference) of the wireless communication systems.

Smart antennas can provide potential improvements including capacity increase, range extension,

data rate increase, interference suppression, multipath diversity, and new services. Therefore, the

technology of smart antenna (SA) has received enormous interest worldwide in recent years.

There are many levels of smart antennas systems, ranging from those that require less signal

processing to extremely advanced signal processing systems requiring antenna arrays at both the

transmitter and receiver. Smart antennas offer a new form of multiple access, which is known as

space-division multiple access (SDMA). The latest form of SDMA uses adaptive antenna arrays, which

is mainly dependent on digital beamforming techniques. In this approach, the main beam of the

radiation pattern is directed towards the desired user directions, while the nulls or the side lobes with

very low levels are adjusted towards undesired users or interferers. Furthermore, the radiation pattern

can be adjusted to receive multipath signals that can be combined. This approach maximizes the

signal-to-interference and noise ratio (SINR) of a desired signal. However, digital beam forming

requires a large number of elements (including antenna elements, receiver modules, A/D converters

etc.) as well as more computational effort, resulting in high cost and system complexity.

Therefore, in this thesis, the first objective will be to reduce the number of elements, while retaining

same array size and same main beamwidth/side-lobe level (SLL). To reach to our aim, we will be

mainly concerned with the design of the array geometry. We will use non-uniform inter-element

spacing in the array instead of traditional uniform element spacing. These kinds of arrays are also

called sparse sampled (or thinned) arrays, which provide a large aperture with few antenna elements.

Sparse arrays are antenna arrays that originally were adequately sampled, but where several

elements have been removed. Sparsely sampled arrays have been used or proposed in several fields

such as radar, sonar, ultrasound imaging and seismics. We will show that this approach is also very

suitable for the smart antenna area.

On the other hand, in future wireless communication systems, wideband signals will be used to fulfil

the requirements of higher data services. Therefore, smart antennas for wideband signals (wideband

8
smart antennas) will be the key solution for reliable high-data-rate wireless channels. However, when it

is desired to receive signals over a broad band of frequencies, the problem of wide-banding an

antenna array arises. Because the beamwidth of a linear array decreases as frequency increases.

Thus, as a second objective, we will investigate a frequency-invariant beam-forming scheme within a

truly wide bandwidth. We first summarize traditional frequency-invariant beam-forming methods and

then propose a frequency-domain frequency-invariant beam former. Moreover, this method is suitable

to operate with arbitrary antenna arrays like uniformly/non-uniformly spaced arrays or one/multi

dimensional arrays. Therefore, we are able to apply the method to the array geometries, which have

been proposed in the first part.

9
Chapter 1

1.1. Smart Antennas for Wireless Communication Systems

It is foreseen that in the future an enormous increase in traffic will be experienced for mobile and

personal communications systems. As the number of mobile subscribers increases rapidly, combined

with a demand for more sophisticated mobile services requiring higher data rates, the operators are

forced to investigate different methods to add more capacity into their networks [1]. An application of

antenna arrays has been suggested in recent years for mobile communication systems to overcome

the problem of limited channel bandwidth, thereby satisfying an ever growing demand for a large

number of mobiles on communication channels [2].

In the literature, adaptive antennas or intelligent antennas are sometimes preferred instead of the

expression ‘smart antennas’. However, we will always prefer to use smart antenna expression. Smart

antenna technology currently provides a viable solution to capacity-strained networks and lends itself

to the migration to third generation (3G) and fourth generation (4G) mobile communication systems.

This new method is to separate the users by their position, exploiting the fact that users normally are

positioned randomly in a cell. A smart antenna is an antenna system that is able to direct the beam at

each individual user, allowing the users to be separated in the spatial domain.

The technology of smart antennas for mobile communications has received huge interest worldwide in

recent years. The main motivation for the smart antennas is the capacity increase, however we should

take into account the other benefits which smart antennas provide like range increase, better signal

quality and new services.

Many base station antennas have up till now been omnidirectional or sectored. This leads to "waste"

of power owing to radiation in other directions than toward the user. Furthermore, other users will

experience the power radiated in other directions as interference. The idea of smart antennas is to use

antenna patterns that are not fixed, but adapt to the current conditions. This can be visualized as the

antenna directing a beam toward the communication partner only. Smart antennas will result in a

much more efficient use of the power and spectrum, increasing the useful received power as well as

reducing interference.

10
conventional
The difference between a smart antenna and a dumb antenna is the property of having and adaptive

and fixed lobe pattern, respectively [3]. The main philosophy is that interferes seldom have the same

geographical location as the user. By maximizing the antenna gain in the desired direction and

simultaneously placing minimal radiation pattern in the directions of the interferers, the quality of the

communication link can be significantly improved.

Not the antenna itself, but rather the complete antenna system including the signal processing is

adaptive or smart. As in Figure 1.1, a smart antenna system combines multiple antenna elements with

a digital signal processing capability to optimise its radiation and/or reception pattern automatically in

response to the signal environment. A smart antenna consists of antenna elements, whose signals are

processed adaptively in order to exploit the spatial dimension of the mobile radio channel. In the

simplest case, the signals received at the different antenna elements are multiplied with complex

weights, and then summed up; the weights are chosen adaptively. Such a system can automatically

change the directionality of its radiation patterns in response to its signal environment (current channel

and user characteristics). This can dramatically increase the performance characteristics (such as

capacity) of a wireless system.

Figure 1.1: Basic principle of smart antennas

1.1.1. Smart Antenna Classifications

Smart antenna systems are classified on the basis of their transmit strategy, into the following three

types. This kind of definition is also called “levels of intelligence” [3].

11
• Switched Beam Antennas

This is also called switch beam. Switched-beam systems consist of multiple narrow beams, the best of

which is used to serve the subscriber as it moves through the coverage of the cell. They have only a

basic switching function between separate directive antennas or predefined beams of an array. The

setting that gives the best performance, usually in terms of received power, is chosen as in Figure 1.2.

Because of the higher directivity compared to a conventional antenna, some gain is achieved. Such an

antenna is easier to implement in existing cell structures than the more sophisticated adaptive arrays,

but it gives a limited improvement.

Figure 1.2: Switched beam antennas

• Dynamically Phased Arrays

The beams are predetermined and fixed in the case of a switched beam system. A user may be in the

range of one beam at a particular time but as he moves away from the centre of the beam, the

received signal becomes weaker and an intracell handover occurs. Once a signal becomes too weak,

the switching centre reassigns a new traffic channel closer to the phone and asks the phone to tune to

this new channel. This is known as handover, a process that is generally transparent to the mobile

user. But in dynamically phased arrays, by including a direction of arrival (DOA) algorithm for the

user’s signal as in Figure 1.3, continuous tracking can be achieved. So even when the intra-cell

handover occurs, the user’s signal is received with an optimal gain. DOA of user is first estimated, and

then the beam-forming weights are calculated in accordance with the DOA of user. In this case also,

the received power is maximized.

12
Figure 1.3: Dynamically-phased arrays

• Adaptive Antenna Arrays


DOA
Adaptive antenna arrays can be considered the most intelligent of all. In this case, DoA algorithm for

determining the direction toward interference sources is added as well. The radiation pattern can be

adjusted to null out the interferers. In addition, by using special algorithms and space diversity

techniques,

The radiation pattern can be adapted to receive multipath signals, which are combined.

These techniques will maximize the SINR (signal to interference and noise ratio) of a desired signal.

This procedure is also known as “optimum combining”, “adaptive beam-forming” or “digital beam-

forming”. In summary, adaptive antenna arrays can make three type of optimisation to the received

signal as in Figure 1.4.

DOA
1) DoA estimation for the desired user

2) DoA estimation for the interference (to null out)

3) Multipath user signal combination

13
Figure 1.4: Adaptive antenna arrays

1.1.2. Benefits of Smart Antennas

We will explain the advantages of smart antennas briefly. More information can be found in [2], [3], [4],

[5], [6], [7].

• Capacity and spectrum efficiency improvement

Smart antennas can increase the capacity of a wireless communication system significantly. Spectrum

efficiency implies the amount of traffic a system with certain spectrum allocation can handle. Channel

capacity refers to the maximum data rate a channel of given bandwidth could sustain. An increase in

the number of users of the communication system without a loss of performance causes the spectrum

efficiency and capacity increase.

The spectrum efficiency E , measured in channels/km2/MHz, is expressed as

Bt Bch 1
E= = (1.1)
BtNcAc B chN cA c

where Bt is the total bandwidth of the system available for voice channels (transmit or receive), in

MHz, Bch is the bandwidth per voice channel in MHz, Nc is the number of cells per cluster, and Ac is

the area per cell in square kilometres. The capacity of a system is measured in channels/km 2 and is

given by

14
Bt N ch
C = EB t = = (1.2)
BchNcAc N cA c

Bt
where Nch = is the total number of available transmit or receive voice channels in the system.
Bch

The actual number of users that can be supported can be calculated based on the traffic offered by

each user and the number of channels per cell. From (1.2), it is evident that capacity can be increased

in several ways. These include increasing the total bandwidth allocated to the system, reducing the

bandwidth of a channel through efficient modulation, decreasing the number of cells in a cluster, and

reducing the area of a cell through cell splitting. If somehow more than one user can be supported per

RF channel, this will also increase capacity.

Increased capacity can be achieved when the SNR level is improved through digital signal processing

techniques that place desired signals in or near the narrower main beam of the multibeam antenna

and place interferers into the pattern side lobes and/or nulls [8]. This can mainly accomplished in two

ways in smart antenna systems. First, the increased quality of service resulting from the reduced co-

channel interferences and reduced multipath fading may be traded to increase the number of users,

leading to increased spectrum efficiency and capacity improvement. Second, an array may be used in

order to create additional channels by forming multiple beams without any extra spectrum allocation,

which results in potentially extra users and thus increases the spectrum efficiency.

• Range extension

The coverage, or coverage area, is simply the area in which communication between a mobile and the

base station is possible. In sparsely populated areas, extending coverage is often more important than

increasing capacity. In such areas, the gain provided by adaptive antennas can extend the range of a

cell to cover a larger area and more users than would be possible with omnidirectional or sector

antennas. In Fig. 1.5, it is easy to grasp the different approaches between a conventional

omnidirectional antenna and smart antenna. Hexagons in both cases show the coverage area of the

wireless system.

A coverage area varies with antenna gain as

15
Ac ∝ G (2 γ ) (1.3)

where γ is the path loss exponent, which is typically between 3 and 4, G is either transmit or receive

antenna gain, and the gain of the other antenna is held constant.

(a) Omnidirectional antenna (b) Smart antenna

Figure 1.5: Range extension using a smart antenna

From (1.3), it can be seen that, the range increase can be achieved by improving the gain. Smart

antennas can improve the gain through more antenna directivity and interference reduction. Range

increase of the wireless network system means that base stations can be placed further apart,

potentially leading to a more cost-efficient deployment. The antenna gain compared to a single

element antenna can be increased by an amount equal to the number of array elements, e.g., an

eight-element array can provide a gain of eight (9 dB) [3].

• Reduction in multipath fading and delay spread

Wireless communication systems are limited in performance and capacity by three major impairments

[9], as shown in Figure 1.6.

16
Multipath fading is caused by the multipaths that the transmitted signal can take to the receive

antenna. The signals from these paths add with different phases, resulting in a received signal

amplitude and phase that vary with antenna location, direction and polarization, as well as with time

(with movement in the environment).

Figure 1.6: Wireless system impairments

The second impairment is delay spread, which is difference in propagation delays among the multiple

paths. A desired signal arriving from different directions is delayed due to the different travel distances
involved. Here, the main concern is that multiple reflections of the same signal may arrive at the

receiver at different times. This can result in intersymbol interference (or bits crashing into one

another) that the receiver cannot sort out. When this occurs, the bit error rate (BER) rises and

eventually causes noticeable degrading in signal quality.

A smart antenna with the capability to form beams in certain directions and nulls in others is able to

cancel some of these delayed arrivals in two ways. First, in the transmit mode, it focuses energy in the

required direction, which helps to reduce multipath reflections causing a reduction in the delay spread.

Second, in the receive mode, multipath fading is compensated by diversity combining technique, by

adding the signals belonging to different clusters after compensating for delays and by cancelling

delayed signals arriving from directions other than that of the main signal. A frequency-hopping system

might be used for correcting fading effects as well. More detailed information about multipath fading

and delay spread reduction techniques can be found in [2].

17
• Reduction in co-channel interference

Co-channel interference occurs when the same carrier frequency reaches the same receiver from two

separate transmitters. The signals that miss an intended user can become interference for users on

the same frequency in the same or adjoining cells. An antenna array allows the implementation of

spatial filtering, which may be exploited in transmitting as well as in receiving mode in order to reduce

co-channel interference. In the transmitting mode, the antenna is used to focus the radiated energy by

forming a directive beam in the area, where a receiver is likely to be. This, in turn means that there is

less interference in the other directions where the beam is not pointing. The co-channel interference

generated in transmit mode could be further reduced by forming specialized beams with nulls in the

directions of other receivers [2]. This scheme deliberately reduces the transmitted energy in the

direction of co-channel receivers and hence requires knowledge of their positions. In the receive

mode, it is not necessary to have a priori knowledge of the positions of the co-channel interferences,

however requires some information concerning the desired signal, such as the direction of its source,

a reference signal, such as a channel sounding sequence, or a signal that is correlated with the

desired signal.

• Signal quality improvement or higher data rates

Reduced co-channel interference, multipath fading and delay spread also leads to an improvement

(reduction) in bit error rate (BER) and symbol error rate (SER) for a given signal-to-noise ratio (SNR).

This means that better signal quality and higher data rates can be achieved for a communication

system. In noise and interference limited environments, the gain that can be obtained with a smart

antenna can be exchanged for lower BER. Experimental results showed that, in a direct sequence

code division multiple access (DS-CDMA) system, if the smart antenna that is employed at the base

station of the central cell can achieve a radiation pattern with a beamwidth of 20˚ (ideal or effective),

then an improvement of 1-7 orders of magnitude for the BER can be accomplished with average side

lobe levels (ideal or effective) between –10 dB and –20 dB, respectively [7].

• Reduced transmit power

Sometimes, the array gain cannot be used for range extension due to limitations on the maximum

EIRP (effective isotropic radiated power). Furthermore, recent public worries over health issues

18
stemming from electromagnetic radiation will almost certainly force the governing/standardisation

bodies to change the current radiation standards in the future and adopt a lower emission policy, in

particular for cellular communication systems. In such cases, one can exploit the base station array

gain to reduce the power transmitted by the mobile. This reduction is also crucial, since it relaxes the

battery requirements and therefore the talk times are increased and the size/weight of the handsets is

reduced.

Moreover, if the received power requirement at the mobile remains same with an M element array at

the base station, then the output power from the base station power amplifiers can be reduced by M -2,

which will reduce the total transmitted power from the array by M-1. It is obvious that, it will reduce the

cost of a system, because high-power amplifiers are expensive hardware components of the system.

• Reduction in handover rate

If the mobile phone’s movement causes the signal to become too weak, the switching centre, which

monitors the signal strength arriving from the phone at the base station, reassigns a new traffic

channel via a base station closer to the phone and asks the phone to tune this new channel. This

procedure is known as handover or handoff, a process that is generally transparent to the mobile user.

When the number of mobiles in a cell exceeds its capacity, cell splitting is used to create new cells,

each with its own base station and new frequency assignment. This result in an increased handover

due to reduced cell size. Smart antennas increase the capacity by creating independent beams using

more antenna elements instead of cell splitting. Each beam is adapted as the mobiles change their

positions. The beam follows a cluster of mobiles or a single mobile, and no handover is necessary as

long as the mobiles served by different beams using the same frequency do not cross each other.

• Support of new services


An important use of adaptive antennas in future wireless systems will be direction finding. Smart

antennas can provide user location information which opens a road to the value added services like

enhanced emergency services, traffic congestion monitoring (by tracking the vehicles equipped with

cellular phones), location-sensitive billing, on-demand location specific services (roadside assistance,

tourist information, electronic yellow pages), vehicle and fleet management.

19
1.1.3 Applications of Smart Antennas

In this section, we will describe two main applications of smart antenna systems. The first one space-

division multiple access (SDMA) is also known as frequency reuse in angle (or simply angle reuse).

SDMA uses beam-forming/directional antennas to support more than one user in the same frequency

channel. The second application, which is called multiple-input multiple-output (MIMO) system, is also

very popular nowadays. In MIMO systems, multiple antenna elements at both (reception and

transmission) end of the transmission link are used. This technique can dramatically increase the

quality of the communication system. The other applications of smart antenna systems are array gain,

diversity gain, and channel estimation. More information about these applications can be found in [10,
11
].

• Space-division multiple access (SDMA)

In wireless systems, there are several methods used for sharing the communication channel among

multiple users. The most popular methods are to separate the users in time – Time Division Multiple

Access (TDMA), in frequency – Frequency Division Multiple Access (FDMA) and by code – Code

Division Multiple Access (CDMA).

Space is truly one of the final frontiers when it comes to new generation wireless communication

systems. Filtering in the space domain can separate spectrally and temporally overlapping signals

from multiple mobile units. Thus, the last stage in the development of multiple access forms is the full

space-division multiple access (SDMA). The spatial dimension can be exploited as a hybrid multiple

access technique complementing FDMA, TDMA and CDMA. This SDMA approach enables multiple

users within the same radio cell to be accommodated on the same frequency and time. The system

can allocate multiple users on the same cell, on the same frequency and on the same time slot, only

separated by angle (spatial domain).

There are various forms of SDMA approach, which provides improvement in the capacity and quality

over omnicells. These are sectorial cells, sectorial beams and adaptive beams as the latest form.

These forms are illustrated in Figure 1.7.

20
(a) 7-cell system with 120˚ sectors (b) 4-cell system with 60˚ sectors

(c) 60˚ sectorial beams within a cell (d) Adaptive beam forming for SDMA
in a 7-cell system

Figure 1.7: Various SDMA approaches

In a frequency-reuse (channel-reuse) system [2], the term radio capacity is usually used to measure

the traffic capacity. The radio capacity Cr is defined as [6]

M
Cr = (1.4)
K .S

where M denotes the total number of frequency channels, K denotes the cell reuse factor, and S

denotes the number of sectors in a cell. In the case of omnicells ( S = 1 and K = 7 ) the radio

capacity is Cr = M 7 channels per cell.

21
Sectorial cells can be exploited to reduce interference. Figure 1.7(a) and (b) show two kinds of

sectorial cell systems: the 7-cell with three 120˚ sectors ( S = 3 and K = 7 ) and 4-cell with six 60˚

sectors ( S = 6 and K = 4 ). In these systems, each sector has a set of unique designated channels.

The mobile user moving from one sector or one cell to another sector or cell requires an intracell

handover.

If directional antennas are used, the capacity can be further improved. In the case of K = 7 , each cell

has a set of M K frequency channels. One can use six directional antennas to cover 360˚ in a cell

and divide the whole set of frequency channels that are assigned to the cell into two subsets, which

are alternating from sector to sector. In this arrangement, there are three co-channel sectors using

each subset in a cell, as shown in Figure 1.7(c).

The ultimate form of SDMA is to use independently steered high-gain beams at the same carrier

frequency to provide service to individual users within a cell, as shown in Figure 1.7(d). That is, a high

level of capacity can be achieved via frequency reuse within a cell. To carry out frequency reuse within

a cell, a certain level of spatial isolation of co-channel signals is required to maintain an acceptable

carrier-to-interference ratio. Adaptive beam forming can provide such a spatial isolation by pointing a

beam at the mobile user and at the same time nulling out the interference from co-channel users.

Therefore, spectrum efficiency can be improved.

A comparison of the capacity and SIR for various systems is presented in [6], as shown in Table 1.1.

Table 1.1: Radio capacity and carrier-to-interference ratio in SDMA

K S Capacity C/I
M
Omnicells 7 1 chs./ cell 18 dB
7
M
120˚ sectorial cells 7 3 chs./ sec tor 24.5 dB
21
M
60˚ sectorial cells 4 6 chs./ sec tor 26 dB
24
3M
60˚ sectorial beams 7 6 chs./ cell 20 dB (worst case)
7
MN
N adaptive beams 7 1 chs./ cell 18 dB (worst case)
7

22
• Multiple-input multiple-output (MIMO) systems

The MIMO technology figures prominently on the list of recent technical advances with a chance of

resolving the bottleneck of traffic capacity in future Internet-intensitive wireless networks. MIMO

communication systems can be defined simply [12], by considering a link for which the transmitting end

as well as at the receiving end is equipped with multiple elements. Such a setup is illustrated in Figure

1.8. Coding, modulation, and mapping of the signals onto the antenna may be realized jointly or

separately.

Figure 1.8: Diagram of a MIMO wireless transmission system

The core idea behind MIMO is that signals at both ends are "combined" in such a way that they either

create effective multiple parallel spatial data pipes (increasing therefore the data rate), and/or add

diversity to improve the quality (bit-error rate or BER) of the communication.

MIMO systems can be viewed as an extension of the smart antenna systems. A key feature of MIMO

systems is the ability to turn multipath propagation into a benefit a user. MIMO effectively takes

advantage of random fading and when available, multipath delay spread, for multiplying transfer rates.

The prospect of many orders of magnitude improvement in wireless communication performance at no

cost of extra spectrum (only hardware and complexity are added) is largely responsible for the

success of MIMO as a topic for new research.

Consider the multi-antenna system diagram in Fig. 1.8. A compressed digital source in the form of a

binary data stream is fed to a simplified transmitting block encompassing the functions of error control

coding and (possibly joined with) mapping to complex modulation symbols (quaternary phase-shift

keying (QPSK), M-QAM, etc.). The latter produces several separate symbol streams, which range

from independent to partially redundant to fully redundant. Each is then mapped onto one of the

multiple TX antennas. Mapping may include linear spatial weighting of the antenna elements or linear

23
antenna space–time precoding. After upward frequency conversion, filtering and amplification, the

signals are launched into the wireless channel. At the receiver, the signals are captured by possibly

multiple antennas and demodulation and demapping operations are performed to recover the

message. The level of intelligence, complexity, and a priori channel knowledge used in selecting the

coding and antenna mapping algorithms can vary a great deal depending on the application.

In the conventional smart antenna terminology, only the transmitter or the receiver is actually equipped

with more than one element, being typically the base station (BTS), where the extra cost and space

have so far been perceived as more easily affordable than on a small phone handset. Traditionally, the

intelligence of the multiantenna system is located in the weight selection algorithm rather than in the

coding side. In a MIMO link, the benefits of conventional smart antennas are retained since the

optimisation of the multiantenna signals is carried out in a larger space, thus providing additional

degrees of freedom. In particular, MIMO systems can provide a joint transmit-receive diversity gain, as

well as an array gain upon coherent combining of the antenna elements (assuming prior channel

estimation). Instead of demonstrating these gains rigorously, we will give an example of the

transmission algorithm over MIMO that is known as spatial multiplexing.

In Fig. 1.9, a high-rate bit stream (left) is decomposed into three independent -rate bit sequences

which are then transmitted simultaneously using multiple antennas, therefore consuming one third of

the nominal spectrum. The signals are launched and naturally mix together in the wireless channel as

they use the same frequency spectrum. At the receiver, after having identified the mixing channel

matrix through training symbols, the individual bit streams are separated and estimated. This occurs in

the same way as three unknowns are resolved from a linear system of three equations. This assumes

that each pair of transmit receive antennas yields a single scalar channel coefficient, hence flat fading

conditions. However, extensions to frequency selective cases are indeed possible using either a

straightforward multiple-carrier approach (e.g., in orthogonal frequency division multiplexing (OFDM),

the detection is performed over each flat subcarrier) or in the time domain by combining the MIMO

space–time detector with an equalizer. The separation is possible only if the equations are

independent which can be interpreted by each antenna “seeing” a sufficiently different channel in

which case the bit streams can be detected and merged together to yield the original high rate signal.

Iterative versions of this detection algorithm can be used to enhance performance.

24
Figure 1.9: Basic spatial multiplexing (SM) scheme with three TX and three RX antennas yielding
three-fold improvement in spectral efficiency

A strong analogy can be made with code-division multiple-access (CDMA) transmission in which

multiple users sharing the same time/frequency channel are mixed upon transmission and recovered

through their unique codes. Here, however, the advantage of MIMO is that the unique signatures of

input streams (“virtual users”) are provided by nature in a close-to- orthogonal manner (depending

however on the fading correlation) without frequency spreading, hence at no cost of spectrum

efficiency. Another advantage of MIMO is the ability to jointly code and decode the multiple streams

since those are intended to the same user. However, the isomorphism between MIMO and CDMA can

extend quite far into the domain of receiver algorithm design.

1.2. Antenna Arrays

In many applications, it is necessary to design antennas with directive characteristics (very high gains)

to meet the demands of long-distance wireless communications. This can be accomplished by forming

an assembly of radiating elements in an electrical and geometrical configuration. This antenna, formed

by multielements, is referred to as an array. There are five-control mechanisms in an antenna array in

order to shape the overall pattern of the antenna. These are:

25
1. The geometrical configuration of the overall array (linear, circular, rectangular, etc)

2. The relative displacement between the elements

3. The excitation amplitude of the individual elements

4. The excitation phase of the individual elements

5. The relative pattern of the individual elements

Before we start to introduce antenna arrays, some antenna array parameters, which will be used in the

second and third chapters to describe the performance of an antenna array, should be provided. More

information about terms and definitions that are commonly used in the study of antennas and arrays

can be found in [13,14]. Afterwards we will explain linear arrays (one dimensional) and planar arrays

(two dimensional). Meanwhile we will introduce pattern multiplication, which let us pass to two-

dimensional arrays from one-dimensional arrays easily.

1.2.1. Basic Antenna Array Parameters

• Radiation pattern

An antenna radiation pattern or beam pattern is defined as a mathematical function or a graphical

representation of the radiation properties of the antenna as a function of space coordinates.

• Main lobe

The main lobe (also called major lobe or main beam) of an antenna radiation pattern is the lobe

containing the direction of maximum radiation power.

• Side lobes

Side lobes are lobes in any direction other than that of the main lobe (intended lobe). For an equally

weighted linear array, the first side lobe (i.e., the one nearest the main lobe) in the radiation pattern is

about 13 dB below the peak of the main lobe.

26
• Beamwidth

The beamwidth of an antenna is the angular width of the main lobe in its far-field radiation pattern.

Half-power beamwidth (HPBW), or 3-dB beamwidth, is the angular with measured between the points

on the main lobe that are 3 dB below the peak of the main lobe.

An example of radiation (beam) pattern is depicted in Figure 1.10. HPBW is usually expressed in

angle (azimuth or elevation angle).

Figure 1.10: A radiation pattern and its associated side lobes and beamwidths

• Grating Lobe

Only the lobe centered at the center angle θ = 0 (or at the beam-steering angle θ 0 if it is not equal to
zero) is the desired lobe. All additional lobes that fall into the real or visible region are called grating

lobes [15]. In an antenna array, if the element spacing is too large, extra main lobes (grating lobes) will

be formed on each side of the array plane. To prevent grating lobes the spacing between antenna

elements should be properly designed. The spacing between the antenna elements should be

maximum half wavelength. That is,

λ
d≤ (1.6)
2

27
Since lowering the array spacing below this upper limit only provides redundant information and

directly conflicts with the desire to have as much as aperture as possible for a fixed number of antenna

elements, for uniform linear arrays (ULA) d = λ 2 is used.

In Figure 1.11, we compare the beam patterns with different element spacing while keeping the

aperture size constant. The element spacings are λ 4 , λ 2 , λ , 2λ for an equal-sized apertures of
10λ with 40, 20, 10, and 5 elements, respectively.

(a) d = λ 4, M = 40 (b) d = λ 2, M = 20

(c) d = λ , M = 10 (d) d = 2λ , M = 5

Figure 1.11: Beampatterns of uniform linear arrays for different element spacing with an equal-sized
aperture

The beam patterns for d = λ 4 and d = λ 2 spacings are identical with equal 3-dB beamwidths

around 7° and equal first side lobes having a height of –13 dB. The oversampling for the array with an

element spacing of λ 4 does not provide any extra information and therefore does not improve the

28
beamformer response in terms of resolution. Actually, if the sensors are too close together

(oversampling case), spatial discrimination suffers (worse angular resolution), because of the smaller

than necessary aperture. In this example, in order to keep aperture size constant, we made the

number of elements twice. Thus, resolution remained same. In the case of the undersampled arrays (
d = λ and 2λ ), we see similar main beamwidth response, however the additional peaks in the beam

pattern appear at ±90 for d = λ and in even closer for d = 2λ . As we described before, these extra
o

lobes are grating lobes. Grating lobes create spatial ambiguities; that is, signals incident on the array

from the angle associated with a grating lobe will look just like signals from the direction of interest.

The beamformer has o means of distinguishing signals from these various directions. In some

applications, grating lobes may be acceptable if it is determined that it is either impossible or very

improbable to receive returns from these angles.

• Array aperture and beamforming resolution

The aperture is the finite area over which an antenna element collects spatial energy. In general, the

designer of an array yearns for as much aperture as possible. The greater the aperture, the finer the

resolution of the array. The resolution can be defined as the ability to distinguish between closely

spaced sources. Improved resolution results in better angle estimation [16]. The angular resolution of

an antenna array is measured in beamwidth. Usually 3-dB beamwidth is used for the comparison of

resolution capabilities of different array structures. The narrower the 3-dB beamwidth, the better the

resolution of the array.

In order to illustrate the effect of aperture on resolution, we compare the beam patterns for M =

4,8,16, and 32 with interelement spacing fixed at d = λ 2 (nonaliasing condition). Therefore, the

corresponding apertures in wavelengths are D = 2λ , 4λ , 8λ , 16λ . It can be observed in Figure 1.12

that, increasing the aperture yields narrower main lobe width and thus provides a better resolution.

29
(a) M = 4 (b) M = 8

(c) M = 16 (d) M = 32

Figure 1.12: Beampatterns of uniform linear arrays for different aperture sizes with equal element
spacing

1.2.2. Linear Arrays

In Figure 1.13, a uniformly spaced linear array is depicted with M identical isotropic elements. This is
the most common structure due to its low complexity. It can perform beam forming in azimuth angle

within an angular sector. Each element is weighted with a complex weight Wm with m = 0,1, ⋅⋅⋅, M -1 ,

and the interelement spacing is denoted by d . If a plane wave impinges upon the array at angle φ

where φ is the angle between broadside of the array and the direction from the wavefield (usually
called azimuth angle), the wavefront arrives at element m , since travel-distance between two

neighbor elements is d sin φ . By setting the phase of the signal at the origin arbitrarily to zero, the

30
phase lead of the signal at element m relative to that at element 0 is β md sin φ , where β = − 2π λ
and λ = wavelength.

Adding all the element outputs together gives array factor AF . The array factor represents the far-

field radiation pattern of an array of isotropically radiating elements.

M −1
AF = W0 + W1e j β d sin φ + W2e j 2 β d sin φ + ⋅⋅⋅ ⋅⋅ = ∑ Wm e jk β d sin φ (1.7)
m =0

Figure 1.13: A uniformly spaced linear array

The array factor in (1.7) can also be expressed in terms of vector inner product as

AF (φ ) = WT w (1.8)

where

W = [W0 W1 ⋅⋅ ⋅ WM −1 ]T (1.9)

31
is the weighting vector and

w = [1, e j β d sin φ , ⋅⋅⋅, e j ( M −1) β d sin φ ]T (1.10)

is the array propagation vector that contains the information on the angle of arrival of the signal. If the

complex weight is

Wm = Am e jmα (1.11)

where the phase of the m element leads that of the (m − 1) element by α , the array factor
th th

becomes

M −1
AF (φ) = ∑
A em j ( βmd sin
φ α
m+)
(1.12)
m =0

If α = − β d sin φ0 , a maximum response of AF (θ ) will result at the angle φ0 . That is, the antenna

beam has been steered towards the wave source in φ0 direction.

1.2.3. Pattern Multiplication

We have only considered arrays of isotropic elements until now. An isotropic element can transmit or

receive energy uniformly in all directions. However, the isotropic antenna is just a mathematical fiction.

In practice, all antenna elements have nonuniform radiation patterns. Let us consider an array

consisting of identical antenna elements that have radiation patterns decided by e(φ ,θ ) . The principle

of pattern multiplication states that the beampattern of an array is the product of the element pattern

and the array factor [6]. That is, the array beampattern G (φ ,θ ) is given by

G (φ ,θ ) = e(φ ,θ ) AF (φ ,θ ) (1.13)

32
where AF (φ ,θ ) is the array factor. It contains the geometric information of the array, that is, the

position coordinates of the elements. The first term e(φ ,θ ) is usually called element pattern. Its effect

on the array is determined by the current excitation across the element.

The principle of pattern multiplication (1.13) is an important result. It decomposes the array properties

into those associated with the excitation of the element and those resulting from the geometric

positioning of the elements. It shows how theorems relating to array design are independent of the

particular antenna element used to form the array. Furthermore, this principle is also very useful to

determine the array factor of a complicated array that is composed of simple subarrays.

1.2.4. Planar Arrays

In addition to placing elements along a line to form a linear array, one can position them on a plane to

form a planar array. Planar arrays can be designed two-dimensional or three-dimensional according to

the spatial requirement of the beam forming. We will only concentrate in two-dimensional planar arrays

that perform beam forming in both elevation and azimuth angles.

A two-dimensional rectangular grid array is one of the common configurations of planar arrays, as

shown in Figure 1.14.

Figure 1.14: A rectangular planar array geometry

33
A rectangular planar array can be seen as a composition of two linear arrays consisting of M-element

in one plane and N-element in another plane. The array factor for the M-element linear array is given

by

M −1
AF1 (u ) = ∑
A em j ( βmd xsinu α
m+ )

m =0
(1.14)

{ }
M −1
where sin u = sin φ cos θ and Am e
jmφ
are the complex weights. The array factor for the N-element
m =0

linear array is given by

N −1
( ) =∑
j ( βnd ysinv γn+ )
AF 2 v Ae n
n =0
(1.15)

{ }
N −1
where sin v = sin φ sin θ and An e jnφ are the complex weights. According to the principle of pattern
n =0

multiplication, the overall array factor for the rectangular array is the given by

=
AF AFuAF1(
v) ()
2
(1.16)

34
Chapter 2

2.1. Introduction

In this chapter, we will start with a brief description of finite continuous apertures, proceed with discrete

apertures. We will describe some necessary concepts like aperture smoothing function and co-array.

We will introduce two important array geometry approaches, which are known as minimum

redundancy arrays and minimum hole arrays. We will show that the minimum redundancy and

minimum hole arrays provide a narrowed main beamwidth and optimal or close to optimal peak side

lobe levels (SLL). For six active elements, we will search minimum redundancy and minimum hole

array geometries, and all the possible array geometries between minimum redundancy and minimum

hole array geometries. Finally, the chapter ends with the application of the approach to the two

dimensional case.

2.2. Continuous Apertures

Apertures are finite areas where antenna array elements gather signal energy. The aperture function
r r
w( x) embodies two kinds of information about the aperture. The spatial extent of w( x) reflects the

size and shape of the aperture. Actually, the aperture acts like a window through which we observe

the wavefield. Moreover, aperture functions can take on any real value between 0 and 1 inside the

aperture. This second aspect of aperture functions allows us to represent the relative weighting of the

field within the aperture. Aperture weighting is sometimes referred to as shading, tapering, or

apodization as well.

• Aperture Smoothing Function

r
Let us assume a space-time signal f ( x , t ) . When we observe a field through the finite aperture, the

output of the sensor

r r r
z ( x , t ) = w( x ) f ( x , t ) (2.1)

35
r
where w( x ) is the aperture or weighting function.

After calculating the space-time Fourier Transform of this relationship, we obtain

r 1

r r r r
Z (β ,ω ) =
(2π )3 ∫ W (β − l )F (l ,ω )dl
−∞
(2.2)

r
where ω represents the temporal frequency variable. Z ( β , ω ) is a convolution over wavenumber

between the Fourier Transform of the field.

The aperture smoothing function is defined as

r ∞
r r r
W ( β) =
1
3 ∫ {
β j .x dx
w (x )exp
(2π) −∞
}r (2.3)

r
This convolution means that the wavefield’s spectrum becomes smoothed by the kernel W ( β ) once

we observe it through an aperture.

• Co-array Function

r
The coarray is defined as the autocorrelation of the weighting function w( x ) [17] and for the continuous

aperture is given by

r r r r
c( χ
) ≡∫
wx χ
(+ )dx
( )wx (2.4)

r
The variable χ is called a lag, and we term its domain lag space. The coarray becomes important

when array-processing algorithms employ the wave’s spatiotemporal correlation function to


r r
characterize the wave’s energy content. The Fourier transform of c( χ ) equals ½
2
W (k )½ , the squared

magnitude of the aperture smoothing function.

36
To clarify the above discussion, let us assume a basic linear aperture. A linear aperture has an

aperture function that is nonzero only along a finite-length line segment in three-dimensional space.

For example, if we let

1, ½≤ D/2

b( x) =  (2.5)
0, otherwise

the three-dimensional aperture function w( x, y , z ) can be written as w( x, y , z ) = b( x)δ ( y )δ ( z ) .

The aperture smoothing function for the linear aperture can be found from Eq. 2.3 as

r sin k x D / 2
W (k ) = (2.6)
kx / 2

Fig. 2.1 illustrates the geometry associated with a typical linear aperture.

Figure 2.1: The linear aperture smoothing function

Because the aperture function is nonzero only along a small segment of the x axis, the aperture
r
smoothing function W depends only on the x component of the wavenumber vector k .

The co-array function for the linear aperture can be found from Eq. 2.4 as

 D −½, - D≤χ ≤D

c( χ ) =  (2.7)
0, otherwise

37
2.3. Discrete Apertures

In many practical applications, arrays composed of individual antenna elements sample the wavefield

at discrete spatial locations. We will give the aperture smoothing function and co-array function like in

the continuous case.

• Aperture Smoothing Function

Let us assume M antenna elements be placed anywhere in a three-dimensional space characterized


r r r
by the variable x . Let the wavenumber vector be β ∈ ¡ with norm β = 2π λ , and for three-
3

r r r r
dimensional space, it can be characterized by β x , β y and β z , where β x = −2π sin φ cos θ λ ,
r
β y = −2π sin φ sin θ λ and β z = −2π cos φ λ . (The minus signs are needed because a wave

propagating into the origin from the first octant (where x , y and z are all positive) would have

negative wavenumber vector characteristics). These angles are usually called azimuth angle for φ
r
and elevation angle for θ . Let the mth antenna element be located at the position xm , where the
r
array element locations are d m = ( xm , ym , z m ) ∈ ¡ and yield the element signal ym (t ) taken here to be
3

r
f ( xm , t ) , which in turn can be represented by the Fourier transform

∞ ∞ r rr r
∫ ∫ F ( β ,ω ) exp { j (ωt − β .d )} dkdω
1
y m (t ) = (2.8)
(2π )4
m
−∞ −∞

r
The wavenumber-frequency spectrum Z ( β , ω ) of the array output is given by

r 1

r r r r
Z (β ,ω ) =
(2π )3 ∫ W (β − l )F (l ,ω )dl
−∞
(2.9)

r
where W ( β ) is the aperture smoothing function, which is given by

r M −1 r r−M 1
W ( β) = ∑ =j ∑
− j (2 πλ)(sin
φ cos
θ
+ x φsin
θ φ
+sin y
βd m m cos
mz ) m
we we
m =0
m
=m 0
m (2.10)

38
For a one-dimensional array that lies in the x − axis like in Figure 1.15, this equation can be rewritten

as

M −1
W (φ) = ∑
w em − j 2π(sinφ λ )x m

m =0
(2.11)

since elevation angle is equal to zero, θ =0.

One can notice the similarity between the array factor formula for a one-dimensional uniformly spaced

array in (1.21) and aperture smoothing function formula for one-dimensional array in (2.11). Actually,

(2.11) can be thought a general beampattern formula that can be used for both equally-spaced (filled)

and non-equally spaced (sparse) linear arrays. Also, in (2.11) no steering has been taken into account,

that is φ0 = 0 . However, beampattern of sparse arrays can also be computed with (1.21), considering

the weighting of the removed element to be zero. That is, if the third, forth and sixth elements are

removed from a seven-element aperture in order to make a sparse array, with taking w3 = w4 = w6 = 0

beampattern can be computed with (1.21).

The aperture smoothing function is the output after weighting and summing all elements in the array

for a wave from infinite distance hitting the array. The aperture smoothing function determines how the

wavefield Fourier transform is smoothed by observation through a finite aperture [18].

• Co-array Function

As said before, the co-array is defined as the autocorrelation of the element weights and for the

discrete case is given by

M −l − 1
c(l ) = ∑w w
m =0
m m +l  (2.12)

where l  is the spatial lag between two elements.

39
The co-array describes the morphology of the antenna array, rather than describing the angular

response [19]. This means that, it describes the weight with which the array samples the different lags

of the incoming field’s correlation function. Usually the co-array has been used to design arrays with

as high resolution as possible. This is equivalent to having a co-array, which is as uniform as possible,

and which spans the maximum number of lags undersampled [20]. For the zero lag, l = 0 , the co-array

is always equal to the number of antenna elements.

For an M element linear array with element distance d , the co-array is related to the beampattern

with:

r 2 M −1 r
W ( β ) = ∑
l =− ( M −1)
c(l ) exp( j β ld ) (2.13)

r 2

where W ( β ) is squared magnitude of the aperture smoothing function. That is, the discrete co-array

function is equal to the inverse Fourier transform of the squared magnitude of the aperture smoothing

function.

Owing to the symmetry of the co-array, this implies

r 2 M −1 r
W (β
 =
) c+
(0) ∑
2 β j ld )
c (l )cos(
l =1
(2.14)

2.4. Sparse Linear Arrays

Sparse arrays are antenna arrays that originally were adequately sampled, but where several

elements have been removed. This is called thinning, and it results in the array being undersampled

[19]. Such undersampling, in traditional sampling theory, creates aliasing. In the context of spatial

sampling, and if the aliasing is discrete, it is usually referred to as grating lobes. In any case, this is

unwanted energy in the side lobe region. (Please state that in sparse arrays, if the positions of

antenna elements are selected appropriately, no grating lobes appear !)

40
What is the motivation for using sparse arrays rather than full arrays? The main reason for their use is

economy. Each of the elements needs to be connected to a transmitter and a preamplifier for

reception, in addition to receive and transmit beam formers. Therefore, the increase in the number of

antenna elements means the increase in cost. For example, medical ultrasound imaging is a field

where most of the sparse arrays work was done, illustrates this: Conventional 2-D scans is done with

1-D arrays with between 32 and 192 elements. 3-D ultrasound imaging is now in development and this

requires 2-D arrays in order to perform a volumetric scan without mechanical movement. Such arrays

require thousands of elements in order to cover the desired aperture. Another reason is the system

complexity. The antenna array should always have low number of elements to avoid unnecessarily

high complexity in signal processing.

2.4.1. Minimum Redundancy and Minimum Hole Arrays

In situations where one must obtain maximum spatial resolution from a limited number of antenna

elements, array configurations known as minimum redundancy arrays [21, 22


, 23
] and minimum hole

arrays [25, 24] are often employed.

To describe the minimum redundancy arrays and minimum hole arrays, we refer to co-array equation

in (2.11). If the array has more than one pair of antenna elements separated by the same distance,

these pairs produce redundant estimates of the correlation function at that lag. In this case, the co-

array of that array is said to have redundancies. Mathematically, this means, in the co-array equation,

if the co-array of a lag is greater than unity, c(l ) > 1 , then a redundancy occurs in that lag. If there is

no pair of antenna elements separated by some distance (lag) that is smaller than the aperture of the

array, the array is said to have a hole in its co-array at that location. This means, if the co-array of a

lag is zero, c(l ) = 0 , then a hole occurs in that lag. In order to have an even sampling of the incoming A figuge to
describe co-
wave field, it is required a co-array with the same weight for all lags. A perfect array is such an array. It array and

is defined as an array with a coarray with no holes or redundancies except for zero lag. Thus, each lag holes?

(excluding the zero lag) of the spatial correlation up to the lag corresponding to the array aperture is

sampled exactly once. Unfortunately, perfect arrays only exist for four or fewer elements in the array.

Therefore, we study arrays that approximate perfect arrays: the Minimum Redundancy (MR) and the

Minimum Hole (MH) arrays. They are defined by the number of redundancies R , and holes H .

Redundancy arrays are defined as an array with redundancies but no holes in its co-array. Minimum

redundancy arrays are those element configurations that have no holes and minimize the number of

redundancies. They are sometimes called redundant arrays [21] as well. Such an array has the largest

41
possible aperture for a redundancy array with a given number of antenna elements. Similarly, hole

arrays are defined as an array with holes but no redundancies in its co-array. Minimum hole arrays

minimize the number of holes in the co-array without any redundancies. These arrays are also known

as non-redundant arrays [21] or Golomb rulers [25]. Such an array has the minimum aperture possible

for a hole array with a given number of antenna elements.

The number of total elements M in the aperture as shown in [26] given by

( n − 1)n
M= +1+ H − R (2.15)
2

where n is the number of active elements, H is the number of holes and R is the number of

redundancies.

For n antenna elements, there are M ( M − 1) 2 pairwise element separations. If each pair were

separated with a different distance (no redundancies and holes were allowed), the number of total

elements M in the aperture would be

( n − 1)n
M= +1 (2.16)
2

As one can notice easily, this is the case for perfect arrays where R = H = 0 .

(2.15) implies that an array with n active elements and M total elements bounded by

M MR < M < M MH (2.17)

where M MR and M MH are the number of total elements of minimum redundancy arrays and minimum

hole arrays, respectively, must have a co-array with both holes and redundancies.

The concept of total elements and active elements might be confusing, because they don’t exist in

uniformly filled geometry. In a uniformly filled linear array, all the elements are active and therefore the

number of total elements is equal to the number of active elements as shown in Figure 2.2 (a).

42
(a) A filled array (b) A thinned array

Figure 2.2: Array geometries for filled and thinned (sparse) arrays

In Figure 2.2(a), a six-element equally spaced filled antenna array geometry is given. All the element

spacing between antenna elements is equal to d . The aperture size can be calculated easily using

D = ( M − 1)d (2.18)

which gives D = (6 − 1) d = 5d . However, in Figure 2.2(b) two elements have been removed and a

sparse array was obtained. In the array geometry, ● symbol shows the positions of active elements,

whereas × symbol shows the removed elements. In this case, we have four active elements and six

total elements (active plus removed elements). Therefore, in sparse arrays, both the number of total

elements and the number of active elements are given. With the knowledge of the number of total and

active elements, one can know the aperture size and the degree of thinning.

2.4.2. Simulation Results

• 1D Sparse Arrays (Sparse Linear Arrays)

We search minimum redundancy arrays and minimum hole arrays for six active antenna elements.

However, the set of possible apertures where minimum hole and minimum redundancy solutions are

restricted, and therefore we also study arrays with both holes and redundancies. For n = 6 , minimum

redundancy arrays are observed for M = 14 . There are 3 different minimum redundancy arrays (+3

mirrored). For n = 6 and M = 14 , when the two end elements are fixed, we obtain a total of

( 124 ) = 495 possible thinning patterns. In Figure 2.3(a), we show the maximum side lobe levels vs. 3

43
dB beamwidths of these 495 possible array geometries. As one can see, most of the arrays reside in

one part of the figure. We focus on this part in Figure 2.3(b).

(a) All array possibilities (b) Focus on some array possibilities

Figure 2.3: Maximum SLL vs. 3 dB beamwidth for a minimum redundancy array with n = 6 and
M = 14

From Figure 2.3(b), one can notice that there are only a few arrays, which satisfy low peak SLL and

narrow beamwidth at the same time. This kind of arrays lies on the lower and left boundary of this

figure.

We will first calculate co-array by using (2.11) where wm ∈ { 0,1} in this case and we obtain

beampattern (squared magnitude of the aperture smoothing function) with the information of co-array

as in Figure 2.4.
Explaining
the meaning
The one with the smallest peak sidelobe and narrowest 3 dB beamwidth from the three minimum
of 15322-
redundancy array possibilities is choosen as an example in Figure 2.4. In Figure 2.4(a), the element array. Also
state that ‘1’
positions of the aperture are given. We prefer to describe the element positions by giving distances
indicates for
between them, that is, 15322 is the geometry for this array. However, the array geometry can also be the position
described using the exact positions of the elements, that is, 11000010010101. In Figure 2.4(b), the of active
element,
resulting co-array design is shown. Only positive lags for the co-array are shown; the negative lags while ‘0’ is
mirror the positive ones due to the symmetry. Except for l = 0 , l = 2 and l = 4 l=5 (?), the co-array for removed
element.
has a uniform shape, that is c(l ) = 1 . As said before, for l = 0 , c(l ) is always equal to the number of

antenna elements, here c(l ) = 6 . Therefore, there are two redundancies (at and l = 4 5(?)) for this

44
array, R = 2 . In Figure 2.4(c), the beampattern, which is calculated using by using (2.13) is shown. It

has a 3 dB beamwidth of 6.0° and maximum side lobe of -6.3212 dB.

(a) Array geometry

(b) Co-array (positive side) (c) Beampattern

Figure 2.4: Array geometry, co-array and beampattern for a minimum redundancy array with n = 6
and M = 14

For n = 6 , minimum hole arrays are observed for M = 18 . There are 4 different minimum redundancy

arrays (+4 mirrored). For n = 6 and M = 18 , when the two end elements are fixed, we obtain a total of

( 164 ) = 1820 possible thinning patterns.


For the minimum hole array, the one with the lowest peak sidelobe but largest –3 dB beamwidth from

the three possibilities is chosen as an example in Figure 2.5. For a 6-element minimum hole array 3

dB beamwidth of 4.8° and maximum side lobe level of -5.8592 dB is observed.

(a) Array geometry

45
(b) Co-array (positive side) (c) Beampattern

Figure 2.5: Array geometry, co-array and beampattern for a minimum hole array with n = 6 and
M = 18

Regular uniform linear arrays and sparse arrays can be compared in two ways. The first is to keep the

aperture fixed while reducing the number of antenna elements to obtain a sparse array. The second

keeps the number of antenna elements fixed and extends the aperture in order to create a sparse

array. The latter is fairer for comparison, because the number of antenna elements decides the cost

and system complexity, as emphasized before. We compare uniform linear array (ULA), minimum

redundancy (MRA) and minimum hole array (MHA) for n = 6 as shown in Figure 2.6.

(a) Co-array for ULA (b) Beampatterns for ULA, MRA and MHA

Figure 2.6: Co-array for ULA and beampattern comparison for ULA, MRA and MHA with n = 6

46
In Figure 2.6(a), co-array of a uniform linear array is shown. A uniform linear array gives a discrete

triangular shaped co-array pattern. Clearly there is a very high degree of redundancy present.

Therefore, it is impossible to obtain a narrow beamwidth (good resolution) with a uniform linear array

as one can observe in Figure 2.6(b). The 3 dB bandwith beamwidth of a this six-element uniform linear

array is found of 17.2°. This is naturally far from being a good result. But However it gives a max. SLL

of -12.426 dB, which is almost twice better of the minimum redundancy and minimum hole solutions.

However, we have to remind you that there are only a few arrays, which satisfy low peak SLL and

narrow beamwidth at the same time. Minimum redundancy and minimum hole arrays are those with

optimal or close to optimal peak side lobe level. In Table 2.1, a comparison of 3 dB beamwidth and

peak side lobe levels is given for 6, 14, and 18-element ULAs, 6-element MRA, and 6-element MHA.

Table 2.1: Comparison of 6,14, and 18-element ULA and 6-element MRA and 6-element MHA

n M R H -3 dB Max. Array Geometry Array


Beamwidth SLL Type
[dB]
6 6 15 0 17.2° -12.426 11111 ULA
14 14 91 0 7.3° -13.112 1111111111111 ULA
18 18 153 0 5.7° -13.171 11111111111111111 ULA
6.1° -6.0606 13162
6 14 2 0 6.0° -6.3212 15322 MRA
6.1° -6.0606 11443
4.8° -5.8592 13625
4.4° -5.3444 13652
6 18 0 2 MHA
4.8° -5.8592 17324
4.6° -5.7285 17423

14-, and 18-element ULAs are given for the comparison while keeping the aperture fixed for minimum

redundancy and minimum hole arrays, respectively. If the aperture is fixed, MRAs and MHAs have

beampatterns with slightly narrower mainlobe beamwidth then ULAs. If one increases the number of

elements for ULAs, it naturally implies an increase in the aperture size as well. Thus, which parameter

(aperture size or number of elements) determines the main beamwidth or max. SLL? For the answer,

we can compare our minimum redundancy ( n = 6 , M = 14 ) and minimum hole ( n = 6 , M = 18 )

arrays, because only aperture size changes, number of elements is fixed. The increased aperture

leads to a narrower mainlobe. There is a slight increase in side lobe level, however this is not related

to the increased aperture, side lobe changes are due to the changes in the array’s geometry array

geometry.

47
Table 2.2 and Table 2.3 show a comparison of –3 dB beamwidth and maximum side lobe levels of

minimum redundancy (from n = 3 to n = 17 ) and minimum hole arrays (from n = 3 to n = 19 ),

respectively. These tables include all known minimum redundancy and minimum hole arrays and can

partially be found in [27, 28].

Table 2.2: Comparison of minimum redundancy arrays for n = 3, 4,...,17

n M R -3 dB Max. SLL Array Geometry


Beamwidth [dB]
[deg]
3 4 0 23.5° -4.6112 12 (perfect)
4 7 0 12.2° -5.2547 132 (perfect)
8.5° -5.4566 1332
5 10 1
8.7° -5.0110 3411
6.1° -6.0606 13162
6 14 2 6.0° -6.3212 15322
6.1° -6.0606 11443
4.6° -6.3353 136232
4.7° -5.6402 114443
7 18 4 4.9° -5.3530 111554
4.6° -5.7666 116423
4.8° -6.2271 173222
3.4° -5.9243 1366232
8 24 5
3.4° -5.6572 1194332
2.7° -6.2428 13666232
9 30 7 2.7° -5.4375 12377441
2.7° -5.7076 11(12)43332
10 37 9 2.2° -5.8592 123777441
11 44 12 1.8° -6.3603 1237777441
1.6° -6.5199 12377777441
12 51 16
1.6° -5.8598 111(20)5444433
1.4° -5.8919 111(24)54444433
13 59 20 1.4° -5.8936 11671(10)(10)(10)3423
1.4° -5.4177 143499995122
1.2° -5.9051 11671(10) (10)(10)(10)3423
14 69 23
1.2° -5.2866 11355(11)(11)(11)66611
15 80 26 1.0° -5.9791 11355(11)(11)(11)(11)66611
16 91 30 0.9° -5.8312 11355(11)(11)(11)(11)(11)66611
17 102 35 0.8° -6.1546 11355(11)(11)(11)(11)(11)(11)66611

We can observe some interesting results from Table 2.2 and Table 2.3. First, only one MRA array

geometry (15322) shows better side lobe level compared to MHA, which includes same number of

antenna elements. For all the others, MHA shows lower peak side lobe level and narrower beamwidth.

Therefore, it is a better solution to use MHA instead of MRA for a given number of antenna elements,

if there is no requirement that antenna array size should be limited. We also see that for large number

of elements, the peak side lobe level of a MHA is approaching to the peak side lobe level of a ULA.

For instance, for 18-element MHA provides peak SLL of -9.1451 dB 9.2018 dB (?), whereas same

48
number of element ULA gives -13.171 dB peak SLL. Moreover, while the increase in the number of

elements for a ULA does not result in a noteworthy decrease in peak SLL, the increase in the number

of elements for a MHA leads to a considerable improvement in peak SLL. Thus, side lobe level may

not be a problem for sparse arrays in case of the usage of large number of elements. If we increase

the number of elements both for MRA and MHA, the mainlobe becomes narrower due to bigger

aperture size, as shown in Fig 2.8(a) and (c) Fig 2.7(a) and (c) (?). However, there is not always

decrease in the peak side lobe level while increasing the number of elements, as shown in Fig 2.7(a)

and (c) Fig 2.7(b) and (d) (?).

Table 2.3: Comparison of minimum hole arrays for n = 3, 4,...,19

n M H 3 dB Max. Array Geometry


Beamwidth SLL
[deg] [dB]
3 4 0 23.5° -4.6112 12 (perfect)
4 7 0 12.2° -5.2547 132 (perfect)
6.7° -4.7564 1352
5 12 1
7.3° -5.5632 2513
4.8° -5.8592 13625
4.4° -5.3444 13652
6 18 2
4.8° -5.8592 17324
4.6° -5.7285 17423
3.0° -5.2997 136852
3.1° -5.8501 164932
7 26 4 3.2° -6.3911 217654
3.2° -6.1730 1(10)5342
3.1° -5.9691 256813
8 35 6 2.4° -6.2299 13567(10)2
9 45 8 1.8° -6.7747 147(13)2863
10 56 10 1.5° -7.0929 154(13)387(12)2
1.1° -7.3838 139(15)5(14)7(10)62
11 73 17
1.2° -7.3485 18(10)57(21)42(11)3
12 86 19 1.1° -7.6831 24(18)5(11)3(12)(13)719
13 107 28 0.8° -8.1373 23(20)(12)6(16)(11)(15)4917
14 128 36 0.8° -7.7829 5(23)(10)381(18)7(17)(15)(14)24
15 152 46 0.6° -7.0393 618(13)(12)(11)(24)(14)32(27)(10)(16)4
16 178 57 0.5° -7.5132 137(15)6(24)(12)8(39)2(17)(16)(13)59
17 200 63 0.5° -8.7866 8(23)36(21)(16)(22)(19)1(13)(11)4(35)(10)25
18 217 63 0.5° -9.2018 (11)(13)4(21)(14)5(18)(32)961(26)3(31)(12)82
19 247 75 0.4° -9.1451 492(27)(14)3(18)(16)(23)(10)(12)8(28)(40)7(19)51

We have two important results that can guide the selection of particular array geometry, aperture size

and number of elements.

• Aperture size fundamentally determines the width of the beampattern’s mainlobe, which in

turn determines the spatial resolution of the array. A greater aperture leads to a narrower

beamwidth.

49
• Both the number of antenna elements and the geometry of an array determine the level of the

beampattern’s side lobes.

(a) HPBW vs. number of elements for MRA (b) Peak SLL vs. number of elements for MRA

c) HPBW vs. number of elements for MHA (b) Peak SLL vs. number of elements for MHA

Figure 2.7: -3 dB Beamwidth/Peak side lobe level vs. number of elements for MRA (a), (b) and for
MHA (c), (d)

In Figure 2.7, the lowest sidelobe level and the lowest –3 dB beamwidth values is chosen for the

numbers where there is more then one MRA or MHA geometry. In Figure 2.7, the values of SLL and 3

dB beamwidth are chosen for smallest values. For instance, for n = 5 and M = 12 (MHA), 6.7° and

-5.5632 dB values 3 dB beamwidth of 6.7° and SLL of -5.5632 dB are used in order to create the

graphs, but they belong to different geometries.

50
For a given number of antenna elements n , an array that has a larger aperture than the MRA with n

elements has holes. Similarly, an array with aperture smaller than that of the n -element MHA has

redundancies. Therefore, any array whose aperture is between that of the MRA and MHA with n

elements must have both redundancies and holes. Aperture sizes of the known minimum redundancy

and minimum hole arrays are presented in Figure 2.8. The region between the two curves represents

arrays that must have both holes and redundancies.

Figure 2.8: Aperture size vs. number of elements for known MRA and MHA

We investigate the arrays for n = 6 , which are near to perfect arrays like minimum redundancy and

minimum hole arrays, but allow redundancies and holes. The arrays with n = 6 active elements for

each of the apertures M = 14,15,...,18 are presented in Table 2.4.

Table 2.4: Properties of the arrays with minimum number of redundancies ( R ) and minimum number
of holes ( H ) for n = 14,15,...,18

M R H 3 dB Max. SLL Array


Beamwidth [dB] Geometry
6.1° -6.0606 13162
14 2 0 6.0° -6.3212 15322
6.1° -6.0606 11443
5.2° -4.1007 14621
15 2 1 5.9° -6.3635 41612
5.6° -5.8481 24341
4.8° -3.5487 14721
16 2 2 5.6° -5.7223 24153
5.1° -5.6498 15522
17 1 2 4.8° -5.6512 25531

51
5.4° -6.2830 41173
5.2° -5.8432 54421
4.8° -5.8592 13625
4.4° -5.3444 13652
18 0 2
4.8° -5.8592 17324
4.6° -5.7285 17423
We see that, near perfect arrays that allow redundancies and holes may give lower side lobe levels.

The best maximum side lobe level has been observed for ‘41612’ array geometry ( M = 15 ) which has
R = 2 and H = 1 . Another this kind of array, 41173, might also be an optimal solution with 5.4° 3 dB

beamwidth of 5.4° and -6.2830 peak side lobe level of -6.2830 dB.

• Side lobe suppression

As we have showed before, minimum redundancy and minimum hole arrays have relatively high side

lobes in comparison with uniform linear arrays. The side lobe suppression can be achieved by using

non-uniform weighting instead of uniform array weights. However, it is difficult or sometimes

impossible to achieve uniform side lobe levels for MRA and MHA over all the angles. Such

suppression can be achieved only over a limited range of angles near the main beam. The number of

side lobe peaks that can be suppressed does not excess n − 1 .

Several researchers have investigated ways to control the side lobe levels for sparse arrays. In [29], an

elegant technique is proposed to find appropriate weighting coefficients, which suppress a maximum

number of side lobes in minimum redundancy arrays to an arbitrarily specified level.

The employed technique is based on iteratively aligning the side lobe peaks to a prespecified level.

The weighting coefficients were obtained in [29], for the MRA that produce a beampattern with the side

lobe level specified as –30 dB. These weights were found to be 0.1401, -0.0966, 0.0329, 0.1303,

0.2096, 0.2308, 0.1914, 0.1135, 0.0115, 0.0337, 0.0027, for an 11-element MRA. The resulting

beampattern and 11-element MRA with uniform weighting is given for the visual comparison in Figure

2.9.

The MRA with non-uniform weightings has uniform –30 dB side lobe level of –30 dB extending to

about ± 13° from both sides of the main lobe. Compared to the uniform weighting pattern, a

substantial side lobe improvement was achieved over the region [-13°, 13°] in which 10 side lobe

peaks were suppressed to be –30 dB. Since there are only 11 elements in the array, the side peaks in

other regions cannot be controlled. However, the 3 dB beamwidth is now 2.7°, which is wider than of

52
the uniform weighting case. The 3 dB beamwidth for 11-element MRA with uniform weighting had a

value of 1.8° -3 dB beamwidth. This is a result of the trade-off between the mainlobe width and the

side lobe levels.

Figure 2.9: The beampatterns for 11-element MRA with uniform and non-uniform weightings

• 2D Sparse Arrays

There are different antenna geometries, which can be used in (smart) antenna systems. These

antenna geometries can be one-, two-, or three-dimensional, depending on the dimension of the space

one wants to access. Although increasing the dimension increases the complexity in the signal-
processing unit, it provides additional scanning (so information) in space, which might be necessary

for some applications. For example, linear arrays and circular arrays are both examples of one-

dimensional arrays and they are used for beam forming in the horizontal plane (azimuth) only. This will

normally be sufficient for outdoor environments, at least in large cells in mobile communication

systems [3]. However, for indoor or dense urban environments, two-dimensional arrays may be

necessary due to their two-dimensional beam-forming capability, in both azimuth and elevation angles.

Like one-dimensional case, we search minimum redundancy arrays and minimum hole arrays for six

active antenna elements. We will find the beampattern using (1.25), after calculating (1.23) and (1.24)

and for the sake of consistency, we will take the square of (1.25) to obtain the squared aperture

smoothing function.

53
A 6-element 1D minimum redundancy array (6-element linear minimum redundancy array) structure

corresponds to a 36-element 2D minimum redundancy array (36-element rectangular minimum

redundancy array) structure. A 36-element 2D minimum redundancy array structure and the

corresponding beampattern for this structure are given in Figure 2.10.

(a) Array structure (b) Beampattern

Figure 2.10: Array structure and beampattern for a 36-element rectangular MRA

The array stucture in Figure 2.10(a) consists of 6 × 6 = 36 active antenna elements. Active antenna

elements are showed with ● symbol, whereas × symbol shows the removed elements like in linear

arrays. The unit spacing between antenna elements (including non-active elements) is d , which is

equal to λ 2 . The aperture size is equal to 13d × 13d or 6.5λ × 6.5λ . The beampattern in Figure

2.10(b) is same for both azimuth and elevation angles because there is no steering applied. For a 36-

element 2D rectangular minimum redundancy array 3.0° 3 dB beamwidth of 3.0° and -7.8923 dB

maximum side lobe level of -7.8923 dB is obtained.

Similar to MRA, 36-element 2D rectangular minimum hole array structure and the corresponding

beampattern for this structure are given in Figure 2.11.

54
(a) Array structure (b) Beampattern

Figure 2.11: Array structure and beampattern for a 36-element rectangular MHA

This array structure is the extension of ‘13625’ geometry of the 1D linear MHA structure to 2D. For a

36-element 2D rectangular minimum hole array, 3 dB beamwidth of 2.4° and peak SLL of -8.7106 dB

are found. An interesting observation is that, extending the 1D geometry to 2D geometry, that is

doubling the number of elements, results with an improvement in the resolution twice. For 6-element

MRA and 6-element MHA, 3 dB values beamwidths changed from 6.0° and 4.8° to 3.0° and 2.4°,

respectively.

Finally, we compare two-dimensional 36-element rectangular minimum redundancy array, 36-element

rectangular minimum hole array and 36-element uniform rectangular linear array (URA) in Figure

2.12(b).

(a) Array structure for 2D ULA URA (b) Beampattern

Figure 2.12: (a) Array structure for 2D ULA URA and (b) beampattern for 36-element rectangular
MRA, 36-element rectangular MHA and 36-element URA

55
The array geometry for a two-dimensional uniform linear array is given in Figure 2.12(a). Different

types of arrays are compared in Table 2.5.

Table 2.5: Comparison of 36-element 2D ULA, MRA and MHA 36-element rectangular MRA, 36-
element rectangular MHA and 36-element URA

n M Max. SLL 3 dB Array Array Type


[dB] Beamwidth Geometry
36 36 -21.361 8.6° 11111 × 11111 ULA URA

36 196 -7.8923 3.0° 15322 × 15322 MRA

36 324 -8.7106 2.4° 13625 × 13625 MHA

56
Chapter 3

3.1. Introduction

In this chapter, we will begin with a very brief description of beamforming. Then, we will make a

comparison of classical analog beamforming and digital beamforming and continue with a discussion

of narrowband beamforming. After a quick view to narrowband beamforming, we will introduce

broadband beamforming techniques. Our aim is to investigate a broadband frequency-invariant

beamforming method. Therefore, in section 3.2, we address ourselves to frequency-invariant

beamformers. We first summarize traditional frequency-invariant beam-forming methods and then

propose a frequency-domain frequency-invariant beam former (FDFIB). We will apply the method to

the minimum redundancy and minimum hole arrays, which have been introduced in the second

chapter.

3.2. Beamforming

A beam former is a processor used in conjunction with an array antenna to provide a versatile form of

spatial filtering. The array antenna collects spatial samples of propagating waves, which are

processed by the beamformer. A beamformer performs spatial filtering to separate signals that have

overlapping frequency content but originate from different spatial locations [ 30]. The goal is to estimate

the signal coming from a desired direction in the presence of noise and interference. Beamforming

techniques are used in several different areas like radar, sonar, communication, imaging,

geophysical/astrophysical exploration and biomedical.

3.2.1. Digital Beamforming vs. Analog Beamforming

Smart antenna techniques are mainly dependent on digital beamforming (DBF) techniques for their

practical implementation. To maximize the SDMA capacity or the level of frequency reuse, it is

desirable to generate a large number of independently steered high-gain mobile spot beams.

However, this is difficult to implement and get a good result using the analog beamforming technology

that is currently used in satellite communication antenna systems. Analog beamforming techniques

57
are subject to an upper limit in terms of their capacity for generating independently high-gain beams.

DBF is the only technology that would support the generation of many beams.

DBF is a marriage between antenna technology and digital technology [6]. A generic DBF antenna

system shown in Figure 3.1 consists of three major components: the antenna array, the digital

transceivers, and the digital signal processor.

Figure 3.1: A generic DBF antenna system

In a DBF antenna system, the received signals are detected and digitized at the element level. The RF

information is captured in the form of a digital stream, then DSP techniques and algorithms are used to

extract information from the spatial domain data. DBF is based on capturing the radio frequency (RF)

signals at each of the antenna elements and converting them into two streams of binary baseband

signals (i.e., in-phase (I) and quadrature-phase (Q) channels). The beamforming is carried out by

weighting these digital baseband signals, thereby adjusting their amplitudes and phases such that

when added together they form the desired beam. This process can be carried out using a special-

purpose DSP. That is, the function, which is usually carried out using an analog beamformer, is now

carried out using a digital processor. Thanks to this approach, the total information available at the

aperture is maintained, in contrast to an analog beamformer, which produces only the weighted sum of

these signals and thus reduces the signal dimensionality from M to1 (Figure 3.2).

58
Figure 3.2: An analog beamformer

The key to this technology is the accurate translation of the analog signals into the digital regime. This

is accomplished using complete heterodyne receivers, which must all be closely matched in amplitude

and phase. The receivers perform the following functions: frequency down-conversion, filtering, and

amplification so that signal levels are commensurate with the input requirements of analog-to-digital

converters (ADC). The main advantage to be gained from digital beamforming is greatly added

flexibility without any attendant degradation in SNR.

3.2.2. Narrowband Beamforming vs. Wideband beamforming

(a) Narrowband Beamforming

The narrowband assumption for array signal processing is that, the bandwidth of the signal is narrow

enough and that the array dimensions are small enough for the modulating function to stay almost

constant during the delay time (the time taken by a plane wave arriving from a source and measured

from one of the antenna elements to the origin) of the signal wavefront.

In a narrowband beamformer, as shown in Figure 3.3, signals from each element are multiplied by a

complex weight and summed to form the array output.

59
Figure 3.3: A narrowband (digital) beamforming structure

The array output is the linear combination of the data at the M array elements at time n is given by

M
y ( n) = ∑ wm* xm (n) (3.1)
m =1

where * denotes the complex conjugate, xm is the signal from the m element of the array, and wm is
th

the weight applied to xm .

The weightings of the beamformer are denoted as

w = [ w1 , w2 , ⋅⋅⋅, wM ]T (3.2)

and signals induced on all elements are denoted as

x( n) = [ x1 (n), x2 (n), ⋅⋅ ⋅, xM ( n)]T (3.3)

Array output is usually expressed in vector form as

y ( n) = w H x ( n ) (3.4)

60
where superscripts T and H , respectively, denote the transpose and complex conjugate transpose

of a vector or matrix. If xm (n) = Wm and wk = e


− jk β d sin θ
, y ( n) becomes equal to AF in (1.21).

However, one is able to do more than just the evaluation of the array factor. There are many ways and

criterias to select the weights of the beamformer depicted in Figure 3.2. We will just mention their

names and more information about criterias and algorithms for the weightings can be found in [4, 6,

30, 31]

• Some criterias or beamforming methods

1. Delay-and-sum

2. Null-steering

3. Minimum variance distortionless response (MVDR)

4. Mean squared error (MSE) minimization

5. Signal-to-interference ratio (SIR) maximation

• Some adaptive beamforming algorithms

1. Sample matrix inversion (SMI)

2. Least mean squares (LMS)

3. Recursive least squares (RLS)

4. Neural networks

5. Conjugate gradient

(b) Wideband Beamforming

The problem of designing of arrays with non-uniformly spaced elements for operation at a single

frequency f D was discussed previously. However, when a single frequency is used over a wide

bandwidth, the array performance degrades significantly: at frequencies below f D the bandwidth

increases, resulting in reduced spatial resolution; at frequencies above f D the beamwidth decreases

and grating lobes may be introduced into the beam pattern [32].

Most of the smart antennas proposed in literature use narrowband beamforming methods. However,

the use of wideband signals is a crucial solution to solve the requirement of very high data rates in

future wireless communication systems. We can distinguish narrowband signals from wideband

61
signals by checking fractional bandwidth (FB) of those signals. The fractional bandwidth of a signal is

the ratio of the bandwidth to the center frequency as follows:

fh − fl
FB = × 100% (3.5)
f h + fl 2

where f h and f l are the highest and the lowest frequencies of the signal, respectively. The

narrowband signals are assumed to have a fractional bandwidth (FB) of less than 1%. Wideband

arrays are designed for FB of up to 50% and ultra-wideband arrays are proposed for FB of 50 to 200%

[33].

The bandwidth of signals incident on the array has a significant effect on the ability of the array to

avoid interference. In [4], degrade of the carrier-to-interference-and-noise-ratio (CINR) is

demonstrated as a function of the bandwidth of signals for a narrowband array as shown in Table 3.1.

Table 3.1: Reduction in CINR as a result of increase in bandwidth of signals incident on the
narrowband array

Bandwidth of Desired and CINR


Interfering Signals
10 kHz 50.0 dB

100 kHz 49.8 dB

1 MHz 41.8 dB

10 MHz 22.5 dB

Wideband beamformers combine spatial filtering with temporal filtering. Wideband beamforming can

be accomplished based on two approaches: 1) Time-domain processing 2) Frequency-domain

processing. However, the important point is, to produce a frequency invariant (frequency independent)

array beam pattern for the wideband signals. In the following section, we will present some

conventional frequency invariant wideband beamformers.

62
3.3. Frequency-Invariant Wideband Beamforming

Conventional wideband beamforming methods are in time domain using tapped-delay-line (TDL) filters

or frequency dilation filters [34]. A common wideband beamformer using time-domain processing

approach is shown in Figure 3.4.

Figure 3.4: Time-domain processing in wideband beamforming with M antenna elements

In Figure 3.4 this figure, filters are designed such that the beampattern is frequency invariant over the

signal bandwidth. In Figure 3.5 a conventional wideband beamformer based on time-processing using

tapped-delay-line filters is depicted.

Figure 3.5: Time-domain processing in wideband beamforming using tapped-delay-line structure

63
In this figure, M antenna element feeds a tapped-delay-lines (transversal filters) and the filter outputs

are summed to produce the overall output on each branch of the array, a tapped-delay-line filter

allows each element to have a phase response that varies with frequency. As a result, the phase shifts

due to higher and lower frequencies are equalized by temporal signal processing. The bandwidth of

signals determine the length ( L ) of the tapped-delay-line. A longer filter is needed for processing

signals with a larger bandwidth leading larger computation requirement [35]. The tapped-delay-lines are

used to flatten the spatial response of the array as a function of frequency. However, it is not possible

to achieve an arbitrary antenna pattern across all frequencies.

Another approach for time-domain frequency-invariant beamformer (TDFIB) is accomplished by using

dilation filters as depicted in Figure 3.6.

Figure 3.6: Time-domain processing in wideband beamforming using dilation filters

Frequency responses of the dilation filters relate to array aperture distribution and relate to each other

by a frequency scaling [35]. For instance, in the above figure, the scaling factor for the fourth dilation

filter is x3 x1 . The scaling factor depends on the element positions relative to the element at the

origin. In this figure, the origin element resides at the position x1 .

Theory and design example of the wideband arrays using dilation filters can be found [32]. In this

work, a frequency invariant beampattern property for a theoretical continuous aperture was developed

and then a discrete aperture array was formed by approximating this continuous aperture. Trapezoid

integration method was used as the approximation method. It was shown that, a frequency invariant

64
wideband array is obtained for a 10:1 frequency range with a uniform aperture. Although traditional

TDFIB methods have some advantages like low data storage requirement and low computational

complexity, there is a newer approach, which comes out superior and is called frequency-domain

frequency-invariant beamformer (FDFIB).

3.4. Frequency-Domain Frequency-Invariant Beamformer (FDFIB)

Frequency-domain frequency-invariant beamformers (FDFIB) offer important advantages over time-

domain frequency-invariant beamformers (TDFIB). A new FDFIB method is proposed in [34] that can

overcome the drawbacks of time-domain processing methods. This method is suitable to operate with

low number of antenna elements within arbitrary geometry. Arbitrary geometry implies both sparse

arrays and multidimensional arrays in addition to one-dimensional and uniform arrays. Furthermore, it

doesn’t require additional filters. With the use of TDFIB, it is difficult to control the beam shape (main

beamwidth and sidelobe levels) since the amplitudes of the beamforming weights can influence the

frequency-invariant characteristic of the beampattern. However, with proposed method, the phases of

the weights control the invariance of the frequency, while the amplitudes of the weights are used to

control beampattern shape. This selection of the weights may be performed in parallel, leading to a

faster processing; therefore, it is well suited for VLSI implementation and is less sensitive to coefficient

quantization [34]. Frequency-domain structures are not sensitive to the sampling rate, and they can

decrease the effects of element malfunctioning on the beampattern. Therefore, there is no necessity

for high-rate A/D converters, which leads to a decrease in hardware cost. For instance, TDFIB method

requires a sampling rate that is typically about 5-10 times the Nyquist rate, while FDFIB method only

requires at the Nyquist sampling frequency rate [36].

A general structure of the frequency-domain processing beamfomer is depicted in Figure 3.7. With the

use of Fast Fourier Transform (FFT), wideband signals from each element are transformed into

frequency domain and each frequency bin is processed by a narrowband processing structure. The

weighted signals are then summed in order to produce an output at each bin. The weights are

selected by independently minimizing the mean output power at each frequency bin subject to

steering-direction constraints. Therefore, the weights required for each frequency bin are selected

independently and also parallelly, leading to a faster weight update. When adaptive algorithms are

used for weight update, a different step size may be used for each bin, leading to a faster convergence

[31].

65
Figure 3.7: Frequency-domain wideband beamforming with M antenna elements

3.4.1. Simulation Numerical Results

Wideband beamforming using FFT and IFFT is illustrated clearer in Figure 3.7. We consider an

arbitrary antenna array composed of m = 1, 2, ⋅⋅⋅, M identical elements, all antenna elements are

assumed to be omnidirectional and mutual coupling between elements is not taken into account.

Figure 3.8: Frequency-domain wideband beamforming with M antenna elements

66
The frequency range of the FFT processed wideband signals ( X 0,1,⋅⋅⋅, M ) are considered from ωl to ωh ,

where ωl and ωh are the lowest and highest frequencies, respectively. We modify the aperture

smoothing function given in (2.10) to vector form as

W (ωk ) = w H (ω k )bΩ
(ω k , b ) (3.6)

where b(ωk ) is often called steering vector and given by

ωk T ω ω
−j dΩ − j kd TmΩ − j kd TMΩ
(ωk , ) = [e , ⋅⋅⋅, e , ⋅⋅⋅, e
1 b b b
bΩ b
c c c
]T (3.7)

where vector d m = [ xm , ym , z m ] and denotes the coordinates of the m − th element in the array and
T

Ωb = [sin φb cosθ b ,sin φb cosθ b ,cos φ b ]T is the unit direction vector, where φb and θb are steered

azimuth and steered elevation at b − th direction, respectively, −180 ≤ φb ≤ 180 , 0 ≤ θb ≤ 90 and


o o o o

c = λ f is the propagation speed. In (3.6) w (ωk ) is the weighting vector and given by

w (ωk ) = [ w1 (ωk ), ⋅⋅ ⋅, wm (ω k ), ⋅⋅ ⋅, wM (ω k )]T (3.8)

where wm (ωk ) is the complex weight at frequency ωk of the beamformer at m − th antenna element.

In order to keep the beampattern constant over the frequency band, the phases of the complex

weights are selected such that

W (ωkω
)W ω
(=)ω


∈ [ , l ]h
k
0
(3.9)

where ω0 is focusing frequency selected in [ωl , ωh ] . From (3.6) and (3.9), the weights for frequency-

invariant beamformer at each frequency are found as

ωk T ω
−j +

m bj d Tm
0
Ω
wm (ω = (3.10)
b

k ) a me
c c

where am is amplitude of the weight, m = 1, 2, ⋅⋅⋅, M .

67
In Figure 3.9, we show frequency-variant and frequency-invariant systems for a 6-element minimum

redundancy array (MRA). Figure 3.9(a) and Figure 3.9(c) depict the frequency-invariant beampatterns

at steered azimuth angle of 90° and steered elevation angle of 90° using the FDFIB illustrated and

described above. The normalized frequency band was chosen in the range of [ωl , ωh ] = [0.3,0.5] with

the focusing frequency ω0 = ωh = 0.5 . The frequencies are normalized over the sampling frequency.

In Figure 3.9(b) and Figure 3.9(d) illustrate the frequency-variant beampatterns at steered azimuth

angle of 90° and steered elevation angle of 90° within the same frequency band. The beampatterns of

all figures were calculated by using squared aperture smoothing function.

(a) Frequency-invariant beampattern for (b) Frequency-variant beampattern for


6-element MRA using FDFIB 6-element MRA

(c) Frequency-invariant: superposition of (d) Frequency-variant: superposition of


all 51 beampatterns all 51 beampatterns

Figure 3.9: Frequency-variant vs. frequency-invariant beampatterns for 6-element minimum


redundancy array in the normalized band [0.3,0.5]

68
Conclusion

The rapid growth of wireless communication systems due to both increased number of users as well

as high-data-rate services leads to a need for the solution of bottleneck in capacity and range.

Presently, smart antenna is one of the most promising techniques that can provide potential

improvements including capacity increase, range extension, data rate increase, interference

suppression, multipath diversity, and new services. Moreover, the high-data-rate channels can be

achieved only by using wideband signals in high-frequency bands. Therefore, smart antennas for

wideband signals (wideband smart antennas) will be the key solution for dependable high-data-rate

wireless systems.

Smart antenna techniques are mainly dependent on digital beamforming techniques for their practical

implementation. However, the requirement for a large number of elements (including antenna

elements, receiver modules, A/D converters, etc.) is one of the main challenges in digital

beamforming. The decrease in the number of antenna elements results in reduction in complexity and

cost of the system. Furthermore, with wideband signals, the shape of the beampattern will be

frequency-variant. To investigate a frequency-invariant beamforming scheme was our second

objective.

In the first part of this thesis, to reduce the number of antenna elements we used sparsely sampled

arrays. We introduced two array geometry approaches that approximate perfect arrays known as

minimum redundancy arrays (MRA) and minimum hole arrays (MHA). For six antenna elements, we

found the beampatterns of MRA and MHA geometries, and all the possible array geometries between

MRA and MHA geometries. We showed that the minimum redundancy and minimum hole arrays

provide a narrowed main beamwidth and optimal or close to optimal peak side lobe levels (SLL). We

compared uniform linear array (ULA), MRA and MHA for same number of elements, in terms of main

beamwidth and peak SLL. We showed that sparse arrays (MRA and MHA) result in incomparable

better resolution (narrowed main beamwidth) compared to ULA. However, the peak side lobe level is

better for ULA. SLL may not be a problem for sparse arrays in case of the usage of large number of

elements if there is no limitation in array antenna size. We also observed that, near perfect arrays,

which allow redundancies and holes may give lower side lobe levels. We gave a comparison of main

beamwidths and peak SLLs of all known MRA and MHA. The side lobe suppression can be achieved

with some weight optimisation methods. We showed that, side lobes can be suppressed within a

limited range of angles, but with a concession in main beamwidth. For some applications or

69
environments, two-dimensional arrays may be necessary due to their two-dimensional beam-forming

capability, in both azimuth and elevation angles. Finally, we extended our MRA and MHA geometries

to two-dimensional arrays and obtained beampatters. We observed same conclusions for two-

dimensional arrays as for one-dimensional arrays.

In the second part, we summarized traditional frequency-invariant beam-forming methods in time-

domain and then introduced a new frequency-domain frequency-invariant beamformer (FDFIB)

method proposed in [34]. This FDFIB method is suitable to operate with arbitrary antenna arrays

Therefore, we applied the method to the sparse array geometries, which have been proposed in the

first part. We illustrated a comparison of frequency-invariant beampattern using this technique and

frequency-variant beampattern in the normalized frequency band [0.3,0.5] for 6-element MRA. The

method has several advantages over time-domain frequency-invariant beamforming (TDFIB) methods.

It is well suited for VLSI implementation and is less sensitive to coefficient quantization. It is insensitive

to the sampling rate, and it can decrease the effects of element malfunctioning on the beampattern.

Therefore, there is no necessity for high-rate A/D converters, which leads to a decrease in hardware

cost. The computational time is also low compared with other methods.

70
References

71
1

[] A. Jacobsen, Smart Antennas for Dummies, Technical report, Telenor R&D, January 2001.

2
[] L. C. Godara, “Applications of antenna arrays to mobile communications, part I: performance improvement,
feasibility, and system considerations”, Proceedings of the IEEE, vol. 85, no. 7, pp. 1031-1060, July 1997.

3
[] P. H. Lehne and M. Pettersen, “An overview of smart antennas technology for mobile communications
systems”, IEEE Communications Surveys, vol. 2, no. 4, pp. 2-13, 4th quarter 1999.

4
[] J. C. Liberti and T.S. Rappaport, Smart Antennas for Wireless Communications: IS-95 and Third-Generation
CDMA Applications, Englewood Cliffs, NJ: Prentice-Hall, 1999.

5
[] Smart Antenna Systems, Web ProForum Tutorials, The International Engineering Consortium,
http://www.iec.org/.

6
[] J. Litva and T. Kwok-Yeung Lo, Digital Beamforming in Wireless Communications, Artech House Publishers,
1999.

7
[] G. V. Tsoulos, “Smart antennas for mobile communication systems: benefits and challenges”, Electronics and
Communication Engineering Journal, Vol. 11, no. 2, pp. 84-94, April 1999.

8
[] H. L. Bachman, Intelligent antennas – an emerging technology, Southcon/96 Conference Record, pp. 56-59,
25-27 June 1996.

9
[] J. H. Winters, “Smart antennas for wireless systems”, IEEE Personal Communications Magazine, pp 23-27,
February 1998.

10
[] K. Sheikh, D. Gesbert, D. Gore, and A. Paulraj, “Smart antennas for broadband wireless access networks”,
IEEE Communication Magazine, vol. 37, pp. 134-142, November 1999.

11
[] A. J. Paulraj, C. B. Papadias, “Space-time processing for wireless communications”, IEEE Signal Processing
Magazine, vol. 14, pp. 49-83, November 1997.

12
[] D. Gesbert, M. Shafi, D. Shiu, P.J. Smith, and A. Naguib, “From theory to practice: an overview of MIMO
space-time coded wireless systems”, IEEE Journal on Selected Areas in Communications, vol. 21, no. 3, April
2003.

13
[] C. A. Balanis, Antenna Theory, John Wiley & Sons Inc, 1997.
14
[] Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Theory, Applications and Design, Van Nostrand Reinhold
Company, New York, 1988.

15
[] B. D. Steinberg, Principles of aperture and array system design, John Wiley & Sons, New York, 1976.

16
[] D. G. Manolakis, V. K. Ingle, S. M. Kogon, Statistical and Adaptive Signal Processing, McGraw-Hill, 2000.

17
[] R. T. Hoctor and S. A. Kassam, “The unifying role of the coarray in aperture synthesis for coherent and
incoherent imaging”, Proc. IEEE, vol. 78, pp. 735-752, April 1990.

18
[] D. H. Johnson and D. E. Dudgeon, Array signal processing: Concepts and techniques, Englewood Cliffs, NJ:
Prentice-Hall, 1993.

19
[] S. Holm, A. Austeng, K. Iranpour, J. -F. Hopperstadt, Sparse sampling in array processing, Chapter 19 in
Sampling theory and practice, F. Marvasti Ed., Planum, New York, 2001.

20
[] J. O. Erstad and S. Holm, “An approach to the design of sparse array systems”, in Proc. IEEE Ultrason.
Symp., (Cannes, France), pp. 1507–1510, 1994.

21
[] A. T. Moffet, “Minimum-redundancy linear arrays”, IEEE Trans. Antennas and Propagation, vol. AP-16, no. 2,
pp. 172-175, March 1968.

22
[] M. Ishiguro, “ Minimum redundancy linear arrays for a large number of antennas”, Radio Sci., vol. 15, pp.
1163-1170, November-December 1980.

23
[] J.-F. Hopperstad and S. Holm, “ The coarray of sparse arrays with minimum sidelobe level”, in Proc. IEEE
NORSIG-98, (Vigs, Denmark), pp. 137-140, June 1998.

24
[] E. Vertatschitsch and S. Haykin, “Nonredundant arrays”, Proc. IEEE,74:217, January 1986.

25
[] G. S. Bloom and S.W. Golomb, “Application of numbered undirected graphs”, Proceedings of the IEEE, vol.
65, pp. 562–570, April 1977.

26
[] S. De Graaf and D. Johnson, “Optimal linear arrays for narrowband beamforming”, in Proceedings IEEE
ICASSP-84, vol. 7, pp. 40.8.1–40.8.4, IEEE, March 1984.

27
[] D. A. Linebarger, I. H. Sudborough, and I. G. Tollis, “Difference bases and sparse sensor arrays”, IEEE
Trans. Inf. Theory, vol. 39, pp. 716-721, March 1993.
28
[] A. Dollas, W. T. Rankin, and D. McCracken, “A new algorithm for Golomb ruler derivation and proof of the 19
mark ruler”, IEEE Trans. Inf. Theory, vol. 44, no. 1, 1998.

29
[] C. –Y. Tseng, L. J. Griffiths, “ Sidelobe suppression in minimum redundancy linear arrays”, IEEE Sixth SP
Workshop, pp. 288-291, 7-9 October 1992.

30
[] B. D. Van Veen and K. M. Buckley, “Beamforming: A Versatile Approach to Spatial Filtering”, IEEE ASSP
Magazine, pp. 4-24, April 1988.

31
[] L. C. Godara, “Applications of antenna arrays to mobile communications, part II: Beamforming and direction-
of-arrival consideration”, Proceedings of the IEEE, vol. 85, no. 8, pp. 1195-1245, August 1997.

32
[] D. B. Ward, R. A. Kennedy, R. C. Williamson, “Theory and design of broadband sensor arrays with frequency
invariant far-field beam patterns”, J. Acoustical Society of America, vol. 97, no. 2, pp. 1023-1034, February 1995.

33
[] M. Ghavami, “Wideband smart antennas theory using rectangular array structures”, IEEE Trans. Signal
Processing, vol. 50, no. 9, pp. 2143-2151, September 2002.

34
[] T. Do-Hong, P. Russer, “Signal Processing for Wideband Smart Antenna Array Applications”, IEEE
Microwave Mag., pp. 57-67, March 2004.

35
[] T. Do-Hong, “Wideband Direction-of-Arrival Estimation and Wideband Beamforming for Smart Antenna
Systems”, Lehrstuhl für Hochfrequenztechnik der Technischen Universität München, 2004.

36
[] R. A. Mucci, “A Comparison of Efficient Beamforming Algorithms”, IEEE Trans. on Acoust., Speech, Signal
Processing, Vol. 32, No. 3, pp. 548-558, June 1984.

Vous aimerez peut-être aussi