Mutual Coupling in Antenna Arrays: Guest Editors: Hon Tat Hui, Marek E. Bialkowski, and Hoi Shun Lui
Mutual Coupling in Antenna Arrays: Guest Editors: Hon Tat Hui, Marek E. Bialkowski, and Hoi Shun Lui
Mutual Coupling in Antenna Arrays: Guest Editors: Hon Tat Hui, Marek E. Bialkowski, and Hoi Shun Lui
This is a special issue published in volume 2010 of “International Journal of Antennas and Propagation.” All articles are open access
articles distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
International Journal of Antennas and Propagation
Editorial Board
Mohammod Ali, USA Wei Hong, China Athanasios Panagopoulos, Greece
Charles Bunting, USA Hon Tat Hui, Singapore Matteo Pastorino, Italy
Christophe Caloz, Canada Tamer S. Ibrahim, USA Sadasiva M. Rao, USA
Dau-Chyrh Chang, Taiwan Shyh-Kang Jeng, Taiwan Stefano Selleri, Italy
Deb Chatterjee, USA Mandeep Jit Singh, Malaysia Krishnasamy T. Selvan, Malaysia
Zhi Ning Chen, Singapore A. A. Kishk, USA Zhongxiang Q. Shen, Singapore
Christos Christodoulou, USA Ju Hong Lee, Taiwan Seong-Youp Suh, USA
Shyh-Jong Chung, Taiwan Byungje Lee, Republic of Korea Parveen Wahid, USA
Tayeb A. Denidni, Canada L. Li, Singapore Yuanxun Ethan Wang, USA
Karu Esselle, Australia Yilong Lu, Singapore Jeffrey Ward, USA
Miguel Ferrando, Spain Andrea Massa, Italy Quan Xue, Hong Kong
Dejan Filipovic, USA Derek McNamara, Canada Tat Soon Yeo, Singapore
Lance Griffiths, USA Ananda S. Mohan, Australia Wenhua Yu, USA
Yang Hao, UK Pavel Nikitin, USA Lei Zhu, Singapore
Contents
Mutual Coupling in Antenna Arrays, Hon Tat Hui, Marek E. Bialkowski, and Hoi Shun Lui
Volume 2010, Article ID 736701, 2 pages
An Open-Source Code for the Calculation of the Effects of Mutual Coupling in Arrays of Wires and for
the ASM-MBF Method, Christophe Craeye, Belen Andrés Garcı́a, Enrique Garcı́a Muñoz, and Rémi Sarkis
Volume 2010, Article ID 964260, 11 pages
Bandwidth Enhancement of Antenna Arrays Utilizing Mutual Coupling between Antenna Elements,
Min Wang, Wen Wu, and Zhongxiang Shen
Volume 2010, Article ID 690713, 9 pages
Decoupling of Multifrequency Dipole Antenna Arrays for Microwave Imaging Applications, E. Saenz,
K. Guven, E. Ozbay, I. Ederra, and R. Gonzalo
Volume 2010, Article ID 843624, 8 pages
Compact Printed Arrays with Embedded Coupling Mitigation for Energy-Efficient Wireless Sensor
Networking, Constantine G. Kakoyiannis and Philip Constantinou
Volume 2010, Article ID 596291, 18 pages
Optimization of Training Signal Transmission for Estimating MIMO Channel under Antenna Mutual
Coupling Conditions, Xia Liu and Marek E. Bialkowski
Volume 2010, Article ID 213576, 10 pages
Effect of Antenna Mutual Coupling on MIMO Channel Estimation and Capacity, Xia Liu and
Marek E. Bialkowski
Volume 2010, Article ID 306173, 9 pages
The Effect of Mutual Coupling on a High Altitude Platform Diversity System Using Compact Antenna
Arrays, Tommy Hult and Abbas Mohammed
Volume 2010, Article ID 492915, 7 pages
Mutual Coupling Effects on Pattern Diversity Antennas for MIMO Femtocells, Yue Gao, Shihua Wang,
Oluyemi Falade, Xiadong Chen, Clive Parini, and Laurie Cuthbert
Volume 2010, Article ID 756848, 8 pages
Hindawi Publishing Corporation
International Journal of Antennas and Propagation
Volume 2010, Article ID 736701, 2 pages
doi:10.1155/2010/736701
Editorial
Mutual Coupling in Antenna Arrays
Mutual coupling is an electromagnetic phenomenon which described and their use is demonstrated in suitably tailored
exists in many antenna arrays. In most of the usual cases examples. This work provides an easy-to-use tool for the
it is detrimental to the antenna operation, although there general understanding of the mutual coupling phenomenon
are some examples in which its presence can be beneficial. in antenna arrays.
Because of its varying impact, it has attracted a large volume In the second paper “Mutual coupling compensation for
of research to study its causes and remedies. The effect direction-of-arrival estimations using the receiving-mutual-
of mutual coupling on array antennas stretches over many impedance method” by H. S. Lui and H. T. Hui, the
different areas from the conventional use of antennas to authors introduce a new idea in mutual coupling analysis by
their modern employment in such exotic areas as multiple- proposing a so-called receiving mutual impedance method
input multiple-output (MIMO) systems, diversity systems, (RMIM). This method, which is suggested to be specifically
medical imaging, and sonar and radar systems. Over the past used with receiving antenna arrays, provides an accurate
years, there have been many methods proposed to study the method for the analysis of the mutual coupling effect in
mutual coupling problem and many solutions have also been receiving antenna arrays. A new concept of receiving mutual
provided. The development of ever-decreasing size radio impedance is introduced and its significance and impact
transceivers has favored the emergence of small-size antenna on mutual coupling analysis is fully explained and demon-
arrays in recent years. Such new demands place the study strated. The receiving mutual impedance, which is funda-
of mutual coupling problem in an even more important mentally different from the traditionally defined mutual
priority. impedance (based on the open-circuit voltage concept) is
This special issue provides for the first time an inter- used in a direction-of-arrival (DOA) estimation problem
national forum for researchers in antenna mutual coupling to yield accurate DOA estimation results. Interested readers
research to disseminate their results and ideas. Nine dedi- are strongly suggested to find out more details from this
cated papers from eight groups of authors have contributed paper.
to this special issue. To help interested readers to have a The third paper, entitled “Bandwidth enhancement of
quick reference to the main themes of these papers, we briefly antenna arrays utilizing mutual coupling between antenna
introduce them below. elements” by M. Wang and W. Wu et al. introduces an
The first paper “An open-source code for the calculation interesting idea in mutual coupling analysis by relating
of the effects of mutual coupling in arrays of wires and for the the enhancement of an array’s bandwidth to a proper
ASM-MBF method” by C. Craeye et al. introduces a method manipulation of its input S parameters. The authors present
known as “array scanning method-maro basis function a formula relating the VSWR bandwidth with the so-called
(ASM-MBF)” to illustrate the fast numerical calculation frequency derivatives of the reflection coefficients, which in
and analysis of wire antenna arrays. Open-source codes in turn can be expressed in terms of the frequency derivatives
the Matlab language for both finite and infinite arrays are of the S parameters. They further point out that a mutual
2 International Journal of Antennas and Propagation
cancellation of some of the S parameters can increase the rigorously studied, leading to determination of the optimal
bandwidth of an antenna array. angle separations between transmitters and receivers that
In the fourth paper, “Decoupling of multi-frequency maximize the total mutual information of the HAP system.
dipole antenna arrays for microwave imaging applications,” In this paper, a novel communication system is considered
the authors E. Saenz and K. Guven et al. make use of employing multiple antenna channels.
a metasurface superstrate to reduce mutual coupling of a In the final ninth paper which entitled “Mutual coupling
printed dipoles array. The metasurface consists of three layers effects on pattern diversity antennas for MIMO femtocells,”
of printed metallic strips aiming to spread the radiated the authors Y. Gao and S. Wang et al. characterize the
field over a larger aperture, thus enhancing the final gain effect of mutual coupling on the performance of a diversity
of the dipole array laid on top of it. It is found that this antenna system operating in MIMO femtocells. This is
metasurface structure can reduce the mutual coupling by accomplished through experimental investigations. A suit-
3 dB to 20 dB, increase the boresight radiation, and reduce able patch antenna is designed with two excitation ports
the endfire radiation. to achieve pattern diversity. The dependence of the channel
In the fifth paper, “Compact printed arrays with embed- capacity on mutual coupling between the excitation ports is
ded coupling mitigation for energy-efficient wireless sensor determined by measurements. This is an application study
networking,” C. G. Kakoyiannis and P. Constantinou use with practical measurement values provided.
photonic bandgap (PBG) structures to reduce mutual cou- All papers appearing in this special issue have been
pling between compact printed antenna arrays used for the subject to a strict peer reviewing process. They are of
construction of wireless sensor networks. Their study shows high quality and address the mutual coupling problem
that an efficient reduction of mutual coupling effect by 15– from different angles and application perspectives. It is our
20 dB can be achieved by a proper design of the PBG struc- expectation that through this special issue, some valuable
ture. Furthermore, they show that impedance bandwidth can ideas and conclusions on mutual coupling research are
be maintained even with a substantial decrease in element provided and at the same time, we hope that more ideas and
separation. research can be encouraged to come out in the future.
The remaining four papers all deal with the antenna
mutual coupling effect in MIMO systems. In the sixth Hon Tat Hui
paper of “Optimization of training signal transmission for Marek E. Bialkowski
estimating MIMO channel under antenna mutual coupling Hoi Shun Lui
conditions,” X. Liu and M. E. Bialkowski investigate the effect
of antenna mutual coupling on the performance of training-
based channel estimation for MIMO systems. Two training-
based channel estimation methods are intensively studied:
the scaled least square (SLS) method and the minimum
mean square error (MMSE) method. They find that the
accuracy of MIMO channel estimation is governed by the
sum of eigenvalues of the channel correlation matrix which,
in turn, is affected by the mutual coupling in the transmitting
and receiving antennas. A water-filling-based procedure is
proposed to optimize the training signal transmission to
minimize the MIMO channel estimation errors.
In the seventh paper of “Effect of antenna mutual coupling
on MIMO channel estimation and capacity,” X. Liu and
M. E. Bialkowski further consider the antenna mutual
coupling effect on MIMO channel capacity with imperfect
channel estimations, that is, the case of lacking perfect
channel state information (CSI). Their study investigates the
effect of different antenna element separations on channel
estimations and the resulting MIMO channel capacities.
The eighth paper, entitled “The effect of mutual coupling
on an HAP diversity system using compact antenna arrays” by
T. Hult and A. Mohammed, investigates the effect of antenna
mutual coupling on MIMO system performance in a special
diversity system. This system is the high altitude platform
(HAP) diversity system proposed in Europe. In this HAP
system, signal repeaters equipped with multiple antennas are
deployed in the stratosphere layer to relay signals within a
given local area, forming an effective MIMO communication
system. The effect of mutual coupling which is present in
the MIMO compact antenna arrays on the system capacity is
Hindawi Publishing Corporation
International Journal of Antennas and Propagation
Volume 2010, Article ID 137903, 10 pages
doi:10.1155/2010/137903
Research Article
An Open-Source Code for the Calculation of the Effects of Mutual
Coupling in Arrays of Wires and for the ASM-MBF Method
Christophe Craeye,1 Belen Andrés Garcı́a,2 Enrique Garcı́a Muñoz,2 and Rémi Sarkis1
1 Communications and Remote Sensing Laboratory, Université Catholique de Louvain, Place du Levant, 2,
1348 Louvain-la-Neuve, Belgium
2 Department of Signal Theory and Communications, Carlos III University of Madrid, Avenida de la Universidad 30,
Copyright © 2010 Christophe Craeye et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
A general description of mutual coupling in finite and infinite antenna arrays is provided and an open-source code is described for
the analysis of mutual coupling in linear arrays of parallel dipoles. The ASM-MBF method is illustrated with the help of that code
and it is shown that the method also can handle situations where eigenmodes are supported by the array: single machine precision
is achieved with four Macro Basis Functions only.
through which the properties of a given radiator—mainly between element patterns and the active input impedance of
its input impedance and radiation pattern—are modified antennas in infinite periodic arrays is established.
when another object is brought in the near field of the Mutual coupling in large arrays is the subject of intense
radiator. In this particular (or rather “general”) perspective, efforts from the numerical point of view. In the following,
mutual coupling between several antennas is not so different details are provided for the case of dipole antennas, and
from mutual coupling between an antenna and a supporting we show how very high accuracy can be achieved when
element (the circuit to which the antenna is connected, the infinite-array solutions are incorporated in the analysis of
body of a person wearing the antenna, for instance), as is finite arrays. All explanations are accompanied by a short
sketched in Figure 1. A possible interpretation is that the introduction to the open-source codes provided in parallel
supporting element is part of the antenna, which therefore with this paper and allowing the reader to become familiar
has other radiation properties than the isolated antenna. It with the different aspects of mutual coupling.
is difficult to say which is the minimal distance from which
mutual coupling becomes negligible. In general, one will
assume distances comparable to the wavelength. The effect 3. MoM for the Array of Wires
of objects located further away will in general be accounted This section briefly recalls the Method of Moments for the
for in a different way, for instance with the help of ray-tracing analysis of linear arrays of parallel thin wires, of radius a
or multiple-scattering approaches. smaller than about a hundredth of the wavelength, spaced
The object (see cup in Figure 1(b)) located in the near by a distance d. This mehod has been devised several
field of the (original) antenna can change the radiation decades ago [12, 13], but it is recalled here, with the help
pattern, absorb a fraction of the radiated power and also of some details given in Appendix B, such that the reader
change the input impedance of the radiator. These phenom- can understand the open-source Matlab code [4] and
ena can be understood as resulting from the change of the adapt it by himself to other situations (configurations with
electric (and magnetic, if volume or surface equivalence are ground planes, planar arrays . . .). This section also allows
used) current distribution on the antenna itself and from us to introduce the concepts necessary to the understanding
the appearance of currents on the perturbating object. Both of the ASM-MBF method described in the next section.
changes in current distributions can be regarded as due to The traditional Pocklington method for thin wires is used.
the introduction of new boundary conditions, imposed by Currents are represented by triangular basis functions, Ji ,
the presence of the nearby object. In traditional antenna
located on the z axis of the wires: J(z) M i=1 xi Ji , where
terminology, these changes in current distribution will
the xi are the unknowns. Fields are tested on the external
impact the antenna input impedance, its radiation pattern
part of the wire, with the help of the same set of functions
and its radiation efficiency. As for the latter quantity, when
(displaced by the radius of the wire). The element of the
one involves the power absorbed by the nearby object in
MoM impedance matrix corresponding to basis function
the estimation of the efficiency of the antenna, one implicity
J j (z) and testing function Ji (z) can be written as:
assumes the object as part of the radiator. It seems reasonable
to say that mutual coupling can be neglected, or could at least
be treated with simplified multiple-scattering approaches, 1 dJi (z) dJ j (z )
Zi j = G(z, z ) k2 Ji (z)J j (z ) − dz dz
when the impact of the neighboring object on the antenna jω dz dz
input impedance is negligible. (1)
Let us consider now the case where the near-field objects
referred to above are made of one or several other radiators where ω is the radian frequency, k is the wavenumber,
(see second antenna in Figure 1(c)). Following the same is the permittivity of surrounding space and G =
philosophy as above, when considering the properties of a exp(−
given antenna in the array, the other antennas in the array, jkR)/(4πR) is the free-space Green’s function, with
R = (z − z )2 + a2 if basis and testing functions are located
terminated by the impedance of their transmitter or receiver,
can momentarily be considered as part of that particular on the same wire, and R = (z − z )2 + e2 if they are
antenna. The obtained radiation pattern then corresponds located on different -parallel- wires, with e = pd the
to the embedded element pattern and the input impedance distance between centers of the wires, where p is integer.
of the antenna in the array (sometimes called the passive The excitation vector v, that is, the right-hand side of the
impedance) can be substantially different from that of the system of equations, is zero everywhere, except for the entry
radiator taken in isolation. In the case of arrays of radiators, corresponding to the testing function that overlaps the delta-
obviously, coupling coefficients can also be defined, in terms gap source. That entry equals −V , where V is the excitation
of impedance, admittance or S-parameter matrices. It is voltage. When the excitation is represented by a voltage
important to notice, however, that all the effects of mutual source V followed by a series load ZL , the voltage jump at
coupling cannot in general be described with the help of the delta-gap level is equal to V1 = V − ZL xs , where xs
such matrices only. Minimum-scattering antennas form an is the current coefficient that multiplies the basis function
exception to this rule [7]. Nevertheless, several links can overlapping the source. Since this coefficient belongs to the
be established between the radiation properties and the N- unknowns, the corresponding term can be moved to the left-
port coupling coefficients. A link between S-parameters and hand side. This leads to subtraction of ZL from Zs,s . Finally,
radiation patterns is given in [1, 8], while in [9–11], the link the current coefficients x are obtained by solving the Zx = v
International Journal of Antennas and Propagation 3
∼ ∼ ∼
Figure 1: Mutual coupling between an antenna and an object located in its vicinity. Effects on radiation pattern and input impedance: (a)
isolated antenna; (b) antenna with object in its vicinity; (c) antenna and other antenna in its vicinity.
system of equations. The most important part of the work functions. For a given pair of antennas, such interactions
consists of computing the double integrals (1). are first stored in the z vector (Here, regularity along the
In the wiremom-finite.m code, this is carried out by wire is exploited.) and then reorganized into the zz matrix.
the inter and inner0 routines. The integrations are carried Such calculations are carried out for all relative positions
out with N points over the interval of length dl, which between pairs of antennas (Here, regularity of the array is
corresponds to the width of the basis function. For a given exploited.) and the zz matrices are accordingly organized
testing function, variables jr and djr represent the currents into the Ztot impedance matrix for the whole array. The
and their derivatives versus z, while they are named jr p and solution is then computed for excitation on successive
djr p for the basis functions. The integrals are carried out antennas to obtain, for each excitation, all the currents on
with the trapezoidal rule, the variables w and w p represent the antennas. Those currents, in turn, allow the computation
the corresponding weights. of the array impedance matrix and of all embedded element
It is well known that, for very thin wires, the integrals patterns. The patterns are very easy to compute in the plane
will not be accurate [12], because of the nearly-singular perpendicular to the antennas, where contributions from
(The function is not perflectly singular since R ≥ a.) different segments of a given antenna appear in phase. In this
behavior of the Green’s function. Hence, when basis and case, for an array of Na elements, the embedded pattern is
testing functions are very close to each other, it is advised to proportional to:
extract the singularity from the Green’s function and to carry
out separately the integrals for the singular part, 1/(4 π R),
Na
M
F(θ) = xn, j e jkd(n−1) cos θ (2)
of the Green’s function (in the inter routine, extract = 1 n=1 j =1
indicates that extraction is to be carried out; the criterion
is based on proximity between basis and testing functions). where θ is the angle with respect to the array axis and xn, j
The integration is carried out numerically over the testing refers to the coefficient that multiplies basis function j on
function and analytically over the basis function. In the antenna n. Figure 2 shows embedded element patterns for
wiremom-finite.m code, those operations are organized elements 1, 2, 4 and 8 in arrays of 17 elements made of wires
as follows. First, the numerical integration over the testing of 1 mm radius and with spacings of 15 cm. The first example
function is carried out in the same inter routine. The has termination impedances of 100 Ohm and the wire length
integration over the basis function is carried out in the and wavelength are 15 cm and 30 cm, respectively. The
inner1 routine, using the Green’s function from which the second example has 0 Ohm termination impedances, while
singularity has been extracted. Besides this, for the singular the wire length and wavelength are 30 cm and 66 cm, respec-
part, the analytical integration over the basis function is tively. The latter supports eigenmodes, which originates from
carried out with the help of the inner2 routine. Details mutual coupling between elements of the array and which
regarding this integration are given in Appendix B. In the leads to complex patterns, mainly oriented along the array
inter routine, the contributions from the regular and axis. In both cases, 21 triangular basis functions have been
singular parts of the inner integrals are represented by the used on each antenna. The impedance matrix (and from
int 1 and int 2 variables, respectively. there also, the scattering matrix) of the array can be obtained
Finally, it is interesting to notice that the main code of from currents calculated for all excitations. Let us denote by
wiremom-finite.m makes use of the regularity of the basis Yt the matrix whose column i contains the port currents
functions over a given antenna and of the regularity of the for excitation of port i with a unit voltage. Then, the array
array by exploiting the fact that the interactions (1) only impedance matrix is given by Z = Yt−1 − UZL where U in a
depend on the vector distances between basis and testing unit matrix and ZL is the termination impedance referred to
4 International Journal of Antennas and Propagation
90 0.1 90 0.6
120 60 120 60
0.4
0.05
150 30 150 30
0.2
180 0 180 0
Figure 2: Radiation patterns of elements 1(-), 2(:), 4(.-) and 8(- -) in arrays of 17 elements. (a) without eigenmodes, (b) with eigenmodes.
The horizontal axis corresponds to the array axis.
above (When terminations are not all the same, UZL should 17]. Indeed, as such, formulation (3) converges very slowly.
be replaced by a diagonal matrix with the ith diagonal entry However, here, the extra computation time required for the
equal to the ith termination impedance.). introduction of the infinite-array Green’s function is made
The validation of the code, illustrated in Appendix A, marginal by (i) using series accelerators, like the Levin-T
has been carried out in two different ways. At the element accelerator [18] and (ii) by tabulating the Green’s function
level, the input admittance has been compared with results before filling the MoM impedance matrix. In the case of
provided by the NEC2 free software as provided by [14], as arrays of parallel wires, the tabulation can be limited to a one-
well as with results shown in [15]. At the array level, the dimensional table, since the infinite-array Green’s function
effects of mutual coupling have been verified by establishing depends only on the z − z coordinate. To be more precise, it
the balance between power accepted by a given port and the is interesting to notice that, in the Green’s function above, the
sum of radiated and dissipated powers. The radiated power is n = 0 term corresponds to the case of the isolated dipole. The
obtained through integration over the unit sphere of the 3-D corresponding contribution to the infinite-array impedance
power density pattern. The dissipated power corresponds to matrix can hence be recuperated from the code referred to in
power delivered to the loads terminating the other elements the previous section. The remainder is then computed using
of the array; in dense arrays, it can correspond to a large the Green’s function (3) made non-singular by withdrawing
fraction of the delivered power. In all cases, the relative error the n = 0 term from it. The latter point also greatly facilitates
was less than 0.1 percent. the tabulation and interpolation steps.
Infinite-array simulations can be carried out with the
4. Fast Calculations with the ASM-MBF Method help of the wiremom-infinite.m code [4]. With respect
to the wiremom-finite.m code, the differences are the
4.1. Infinite-Array Simulations. For infinite regular arrays following. First, the solution is looked for in the unit cell only
with uniform amplitude and linear phase excitation, the and the result depends on phase shift ψ between consecutive
method above can be extended such that calculations can elements. Second, the infinite-array Green’s function is
be limited to a unit cell of the array, while involving all the tabulated in subroutine table1. It is computed as the
effects of mutual coupling. This is done with the help of explicit summation of contributions from successive dipoles.
the introduction of an infinite-array Green’s function. For The summation is accelerated with the help of the Levin-
a linear array, with inter-element phase shifts equal to ψ, it T accelarator [18]. It is important to notice that, for the
reads: latter to be efficient, the summation must be carried out
∞ independently for the two semi-infinite arrays that compose
exp − jkRn
G(R) ⇒ e− jnψ (3) the infinite array. The contribution from the closest dipole is
n=−∞ 4πRn excluded and added when necessary in the inner0 and inner1
routines. As in the code for finite arrays, those routines
where Rn = (a + d n)2 + (z − z )2 . For n =/ 0, in view of calculate the interactions over basis functions, for the case
the thin wire approximation, the term a can be omitted as when singularity is extracted and for the case where it is not
long as d a. Other formulations for this Green’s function, extracted, respectively. The Green’s function is interpolated
with faster convergence, can be found in the literature [16, from the table using a simple first-order approach.
International Journal of Antennas and Propagation 5
×10−3 ×10−3
6.6 12
10
6.4
8
Real part of port current
6 4
2
5.8
0
5.6
−2
5.4 −4
0 5 10 15 20 0 5 10 15 20
Element index Element index
(a) (b)
Figure 3: Real part of port currents in array of 17 dipoles excited in phase. Horizontal lines: infinite-array simulations. (a) without
eigenmodes; (b) with eigenmodes.
Table 1: For two types of arrays, with 8 elements each, power the x direction with periodic excitation at the feed-points
delivered to the array at port 1, radiated power, power disspiated in level, that is, with constant amplitudes and linear phase
loads of other elements and sum of radiated and dissipated powers. progression; the inter-element phase shift is given by ψ. If we
Units are milliWatts. denote a given field in the reference unit cell, which contains
−
→∞ →
Accepted Radiated Dissipated Rad. + Diss. the coordinate x = 0, by P (− r , ψ), then the field obtained
Type 1 1.2129 1.1542 0.0586 1.2128 in cell m when only the source in the reference unit cell is
Type 2 14.154 14.142 0 14.142 excited is given by:
2π
→0 −
− → 1 →∞ −
− →
Pm r = P r , ψ e− jmψ dψ (4)
A simulation example is provided in Figure 3 for the 2π 0
same two configurations as in Section 3, the second of where − →
r are coordinates relative to a reference point
which supports eigenmodes. All elements are excited, with connected to each antenna. A discretized version of this
constant amplitudes and constant phases. What is shown integral is:
is the real part of port currents. Important differences are
observed between finite and infinite-array solutions. This is N −1
→0 −
− → 1 − →∞ −
especially true for the array supporting eigenmodes. In the Pm r P → r , ψn e− jmψn for 0 ≤ m ≤ N − 1
N n=0
latter case, it is interesting to notice that eigenmodes affect
the solution of the finite array, even if, in the infinite array, (5)
the eigenmode solution appears only for other inter-element with ψn = n/N 2π. It is interesting to notice that this
phase shifts [19]. This is because the finite-array solution summation exactly has the same form as the FFT, as defined
actually involves a broad spectrum (in terms of inter-element in the Matlab language for instance. The result of the
phase shifts) of infinite-array solutions. This fact is explicitely discretization of integral (4) is that the obtained fields
exploited in the ASM-MBF method described below. actually correspond to those excited by sources located every
−
→0
N cells along the array. In other words, P m (− →
r ) becomes
4.2. Array Scanning Method. The Array Scanning Method
periodic along m with period N, as illustrated in Figure 5. To
(ASM) allows the determination of the behavior of an
avoid this aliasing problem, the solution may consist of using
active antenna (or any single-source excitation) in an infinite
adaptive integration techniques [21]. Another possiblity
passively terminated array from results obtained for fully
consists of artifically increasing the number of sampling
periodic array excitations, while accounting for all the effects
points by increasing N and by interpolating them where the
of mutual coupling. The ASM is connected with the Discrete −
→∞ →
Fourier Transform and has first been proposed by Munk function P (− r , ψ) can be assumed smooth. Among others,
and Burrell [20], although DFT-type links between single- particular attention (and hence, many explicitly computed
−
→∞ −
→∞
element excitation and periodic excitation were already values of fields P ) will be given to the estimation of P
reported in [2]. It will be recalled here for the case of near the limits of visible space (ψ = ±kd) and near some
linear arrays. We consider an array with periodicity d along possible values of ψ where the impedance matrix becomes
6 International Journal of Antennas and Propagation
singular, which corresponds to eigenmode solutions [19, appears natural to consider as MBFs the solutions obtained
22]. According to the properties of the Fourier Transform, on successive antennas when the ASM is used.
−
→∞ →
singular values of P (− r , ψ) in the reciprocal ψ domain In practice, the discretized form of the ASM, computed
will lead to widely spread solutions in the spatial domain. with N infinite-array solutions, will be used. It may be
A straightfoward implementation of the ASM is provided by expected that the aliasing that characterizes the discretized
routine asm.m. Simulation examples are shown in Figure 4 ASM (see Figure 5) may severely degrade the ability to
for the two configurations already considered previously. The represent current distributions in the finite array excited by
second case is dominated by eigenmodes, which propagate one element only. However, it is important to recall that, in a
over the whole structure and are reflected by the ends. first instance, what is looked for is just a representative basis
for current distributions. In this respect, the overlapping
between aliased solutions is not a major difficulty, since the
4.3. ASM-MBF. The Macro Basis Functions approach
different superimposed solutions are all physical. Of course,
(MBF) applied to periodic antenna arrays consists of assum-
for larger values of N in the ASM, more current distributions
ing that, whatever is the excitation law of the array, the
are produced and a higher accuracy can be achieved. Finally,
solution on a given unit cell can be obtained as the linear
since the ASM solutions are themselves superpositions of
superposition of a small number of pre-computed solutions
infinite-array solutions with constant-amplitude excitation
obtained from small problems [23–25].
with linear phase progression, the latter may equally well
Once MBFs have been defined on every sub-domain and
serve as MBFs. Hence, the ASM does not need to be explicitly
concatenated in matrix form (for sub-domain i, each column
calculated to form the MBFs. The main point learned from
of Qi corresponds to the list of coefficients of a given MBF
the reasoning based on the ASM is that, upon forming MBFs,
in terms of elementary basis functions), a reduced system of
it is good to have the phase shifts ψ of the infinite-array
equations is obtained by writing for every sub-domain:
solution uniformly distributed from 0 to 2π. This means
x i = Qi y i (6) that, as soon as the element spacing is smaller than half
a wavelength, solutions from outside the visible space (i.e.,
and by using as many Macro Testing Functions (MTFs) as with active impedances that are purely reactive) will be
there are MBFs. In this work, we will use Galerkin’s method necessary to form a good set of MBFs.
and consider that, on every sub-domain, the set of MTFs is
identical to the set of MBFs. In this case, for instance, the 4.4. Application to Finite Arrays of Wire Antennas. The
reduced system of equations for two sub-domains can be ASM-MBF technique described above will be applied to
written as: finite arrays of parallel antennas. This efficient and accurate
⎡ ⎤⎡ ⎤ ⎡ ⎤ approach is applicable for any excitation of the array and
Q1H Z11 Q1 Q1H Z12 Q2 y1 Q1H v1
⎣ ⎦⎣ ⎦=⎣ ⎦ (7) accounts for all the effects of mutual coupling. In view of
Q2H Z21 Q1 Q2H Z22 Q2 y2 Q2H v2 the symmetry of the infinite array with respect to a plane
perpendicular to the array axis and passing through the
The (H ) suffix denotes transposed conjugate. The conjuga- middle of an element, it is interesting to notice that infinite-
tion could be omitted: results would be slightly different, but array solutions will be identical for ψ and −ψ phase shifts.
numerical experience showed us that accuracy is similar. Therefore, only positive values of ψ need to be considered
A very popular technique for obtaining the MBFs in uppon forming the set of MBFs. Hence, if N infinite-array
antenna arrays consists of generating primary and secondary solution are considered to form the MBFs, the related phase
current distributions [26]. The primary solution is the shifts should be equal to ψn = (n − 1)/Nπ for 1 ≤ n ≤ N.
solution obtained by exciting an isolated element, while a The arrays-compare.m routine compares the “brute-force”
secondary is obtained by exciting a neighboring element solution (from Gaussian elimination in the original MoM
with the fields radiated by the primary MBF and then by system of equations) with the solution obtained with the
computing the induced current. Several secondaries can be help of the ASM-MBF method. The routine is made of four
defined and higher-order MBFs may also be generated [27]. parts. First, the brute-force solution is computed with the
In the present paper, we will adopt another approach based help of the wiremom-finite.m routine. Second, the Mr
on the array-scanning method, as demonstrated in [5] for infinite-array solutions are computed and stored in matrix
arrays of tapered-slot antennas. This approach is based on C (equivalent to Q matrix above). Then, the original MoM
the observation that the solutions obtained for an arbitrary impedance matrix and excitation vector for the array are
excitation law can be obtained from solutions computed for reduced. Finally, the reduced system of equations is solved
excitations at every single element of the otherwise passively and the solution is determined with the help of the MBFs.
terminated array. The case of excitation at a single element We will illustrate the accuracy of the ASM-MBF method
is similar to the case considered in the previous paragraph, for the two 17-elements arrays already considered in the
where the ASM is used, the difference being that the array previous section, one of which supports eigenmodes. The
is finite here. The difference with respect to the case of first element of the array is excited and the relative error in
infinite arrays may be regarded as currents reflected from the dB of the ASM-MBF solution, as compared to the brute-force
edges of the array [28]. Under the asumption that reflected solution, is obtained as 20 log10 (|Ia − Ie |/ |Ie |), where Ia is
current distributions are similar to those found in the direct a vector that contains all the coefficients of the ASM-MBF
wave (excited by a single element in the infinite array), it solution over successive antennas, while Ie is a similar vector
International Journal of Antennas and Propagation 7
×10−3
6 0.025
5
0.02
4
Current (A)
Current (A)
0.015
3
0.01
2
0.005
1
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Element index Element index
(a) (b)
Figure 4: Magnitude of port currents obtained for the array configurations referred to in the previous section, when element 5 only is
excited. Discrete-ASM solution obtained for N = 59. (a) without eigenmodes; (b) with eigenmodes.
ASM aliased source Finite array: direct and reflected waves ASM aliased source
Figure 5: Wave phenomenology for port currents in finite and infinite arrays with a single excited element. Repeated curves are due to
aliasing caused by numerical ASM integration.
for the brute-force solution. Figure 6 shows the relative error when the number N of MBFs is increased, even when the
for an increasing order N of the ASM-MBF method. It is array supports eigenmodes, which can strongly distort the
interesting to see that, with N = 4 MBFs per antenna, the embedded element patterns. If M is the number of elemen-
error is already well below −100 dB, even for the case where tary basis functions and N is the number of macro basis
the array supports eigenmodes. For the latter case, it can be functions, for large arrays, the solution time is proportional
seen that solutions with a single MBF per antenna is not to N 3 instead of M 3 . Knowing that, with the ASM, N is an
acceptable; however, as soon as more MBFs are used, the order of magnitude lower than M, very large time savings
accuracy of the solution becomes comparable to the one are achieved, while preserving an accuracy close to single
obtained with arrays not supporting eigenmodes. machine precison.
5. Conclusion Appendices
We described a simple open-source tool for the under- A. Validation Data
standing of mutual coupling in linear arrays of antennas.
Appendix C provides the general organization of the code. Figure 7 shows a comparison between results obtained with
We showed that finite-array and infinite-array approaches NEC2 [14] and with the proposed open-source code for
can be reconciled by using the solutions of the Array the example given in Figure 4.5 of [15] for a dipole with
Scanning Method (ASM) as Macro Basis Functions (MBFs). a ratio between length L and diameter equal to 74.2. In
In practice, the ASM must be computed with a limited both cases, 41 basis functions were considered. Values for
number N of infinite-array solutions. It is striking to see the conductance G are extremely close, while a slight drift for
how fast the solution converges toward the reference result the susceptance B appears for L/λ ratios larger than 0.6. It is
(obtained without reduction of the system of equations) interesting to notice, however, that results from [15] (shown
8 International Journal of Antennas and Propagation
0.012
0
0
0 20 40 60 80 100 120 140 160
−80 Basis function index
−100 Figure 8: Currents on four parallel dipoles. Solid: our code. Dotted:
NEC2 [14].
−120
−140
15 cm. For Type 2, λ = 0.66 m, load impedance is 0 Ohm and
−160 wire length is 30 cm.
1 2 3 4 5 6 7
For the Type 1 array, after considering individual excita-
Number of MBFs
tion at successive ports, the maximum relative error found is
Figure 6: Errors versus number of MBFs between ASM-MBF and 1.3 10−4 , for Type 2 array, it is equal to 12 10−4 . The latter has
brute-force solutions for a 17-dipoles array excited at first element. a zero load impedance (from there the zero dissipated power
Solid: without eigenmodes. Dashed: with eigenmodes. in the Table). If a 100 Ohm load is considered for the Type 2
case, the maximum relative error is very close to 10−4 .
15
B. Details About Singularity Extraction
For the sake of completeness, we provide here more details
10 about the singularity extraction procedure. It consists of
carrying out analytically the convolution integral over the
testing function for the singular part of the free-space Green’s
functions, that is, for Gs (−
→ −
r ,→
r ) = 1/(4πR) with R = |− →
G, B (ms)
r −
5 −
→
r |. We denote by z the height of the observation point
(on the surface of the cylindrical conductor) and, in a first
instance, we consider the half basis function with a triangular
0 shape equal to zero at z = zm and equal to one at z = z p =
zm + dl/2 where dl is the length of the domain over which
the complete basis function is defined. Then, within a factor
1/(4π), the inner integral in (1) can be decomposed into the
−5
0 0.5 1 1.5 2 two following terms:
L/λ zp
k2 Jt 1 z − zm
Figure 7: Comparison between admittances obtained with the Ia = dz
(dl/2) zm R
proposed code (thick lines) and those obtained with NEC2 (thin zp (B.1)
lines). Solid lines: conductance G. Dashed lines: susceptance B. dJ /dz1 1
Circles: MoM data from Figure 4.5 of [15].
Ib = − t dz
(dl/2) zm R
The Ia integral is decomposed into two terms using z − zm =
(z − z) + (z − zm ) and Ia and Ib then become:
with circles) are slightly closer to our results than to those
k2 Jt 1
obtained with NEC2. Also, for this isolated dipole, and for Ia = (B.2)
L/λ = 2, the power conservation check (see below) provided (dl/2)(I1 + (z − zm )I2 )
a relative error limited to 2 10−4 . Figure 8 compares port dJt /dz1
currents obtained at 1 GHz for an array of 4 such dipoles for Ib = − (B.3)
(dl/2)I2
L = λ/2 and spacing d = λ/2.
Regarding results obtained for arrays, it is interesting with
to verify that power accepted by a given antenna in the x=z p −z
array corresponds to the sum of radiated power and power I1 = x2 + a2 x=zm −z
dissipated in the loads of the other elements of the array. This x=z p −z (B.4)
verification is presented in Table 1 for the two different arrays I2 = ln(x + x2 + a2 ) x=zm −z
considered throughout the paper, with 8 elements each. The
wire radius is 1 mm, array spacing is 15 cm. For Type 1, A similar treatment is applied to the other half of the basis
λ = 0.3 m, load impedance is 100 Ohm and wire length is function. In that case, the differences are (i) an overall change
International Journal of Antennas and Propagation 9
of sign, (ii) zm is the previous z p and z p = zm + dl/2 and (iii) [10] D. M. Pozar, “The active element pattern,” IEEE Transactions
the use of z − z p instead of z − zm in (B.2). on Antennas and Propagation, vol. 42, no. 8, pp. 1176–1178,
1994.
[11] C. Craeye and M. Arts, “On the receiving cross section of an
C. Organization of Routines antenna in infinite linear and planar arrays,” Radio Science, vol.
Here, we briefly summarize the different routines of the 39, no. 2, Article ID RS2010, 8 pages, 2004.
open-source code available on [4] and we provide some [12] R. F. Harrington, “Matrix methods for field problems,”
guidelines on how to try them. Proceedings of the IEEE, vol. 55, pp. 136–149, 1967.
[13] W. C. Gibson, The Method of Moments in Electromagnetics,
(1) wiremom-finite.m gives the currents on a finite Chapman & Hall/CRC, 2008.
array when it is excited at successive ports. To see [14] 4NEC2, http://home.ict.nl/∼arivoors/.
how it can be called, one can have a look at the
[15] R. F. Harrington, Field Computations by Moment Methods,
patterns.m routine, which provides the embedded IEEE Press, New-York, NY, USA, 1993.
element patterns.
[16] R. J. Mailloux, “Phased array theory and technology,” Proceed-
(2) wiremom-infinite.m is a similar code, for infinite ings of the IEEE, vol. 70, no. 3, pp. 246–291, 1982.
arrays. To see how it can be called, one can have a look [17] G. Valerio, P. Baccarelli, P. Burghignoli, and A. Galli, “Com-
at asm.m, which corresponds to the exploitation of parative analysis of acceleration techniques for 2-D and 3-D
infinite-array solutions in the framework of the Array Green’s functions in periodic structures along one and two
Scanning Method. directions,” IEEE Transactions on Antennas and Propagation,
(3) fininf.m compares finite and infinite-array solu- vol. 55, no. 6, pp. 1630–1643, 2007.
tions for the case of periodic excitations with inter- [18] S. Singh and R. Singh, “On the use of Levin’s T-transform
element phase shift ψ; it also makes use of the in accelerating the summation of series representing the
free-space periodic Green’s functions,” IEEE Transactions on
previous codes).
Microwave Theory and Techniques, vol. 41, no. 5, pp. 884–886,
(4) arrays-compare.m provides the brute-force and 1993.
approximate currents for a finite array excited at ele- [19] D. S. Janning and B. A. Munk, “Effects of surface waves on the
ment shift (defined inside the code). “Approximate” currents of truncated periodic arrays,” IEEE Transactions on
means with the help of the ASM-MBF method. The Antennas and Propagation, vol. 50, no. 9, pp. 1254–1265, 2002.
sweep.m routine computes the errors versus element [20] B. A. Munk and G. A. Burrell, “Plane-wave expansion for
spacing and versus frequency. arrays of arbitrarily oriented piecewise linear elements and its
application in determining the impedance of a single linear
antenna in a lossy half-space,” IEEE Transactions on Antennas
References and Propagation, vol. 27, no. 3, pp. 331–343, 1979.
[1] S. Stein, “On cross coupling in multiple-beam antennas,” IRE [21] F. Capolino, D. R. Jackson, and D. R. Wilton, “Fundamental
Transactions on Antennas and Propagation, vol. 10, no. 5, pp. properties of the field at the interface between air and a
548–557, 1962. periodic artificial material excited by a line source,” IEEE
[2] R. C. Hansen, “Radiating elements and mutual coupling,” Transactions on Antennas and Propagation, vol. 53, no. 1, pp.
in Microwave Scanning Antennas, Vol. II, A. A. Oliner, Ed., 91–99, 2005.
chapter 2, Academic Press, New York, NY, USA, 1966. [22] C. Craeye, “Exploitation of infinite-array results for accurate
[3] F. Serracchioli and C. A. Levis, “The calculated phase velocity solution of finite wideband arrays,” in Proceedings of the
of long end-fire uniform dipole arrays,” IRE Transactions on URSI Electromagnetic Theory Symposium (EMTS ’07), Ottawa,
Antennas and Propagation, vol. 7, pp. 424–434, 1959. Canada, July 2007.
[4] http://www.mathworks.com/matlabcentral/fileexchange/ [23] E. Suter and J. R. Mosig, “Subdomain multilevel approach
26823-arraymutualcouplingasmmbf. for the efficient MoM analysis of large planar antennas,”
[5] C. Craeye and R. Sarkis, “Finite array analysis through combi- Microwave and Optical Technology Letters, vol. 26, no. 4, pp.
nation of macro basis functions and array scanning methods,” 270–277, 2000.
Applied Computational Electromagnetics Society Journal, vol.
[24] L. Matekovits, G. Vecchi, G. Dassano, and M. Orefice,
23, no. 3, pp. 255–261, 2008.
“Synthetic function analysis of large printed structures: the
[6] C. Craeye, B. Parvais, and X. Dardenne, “MoM simulation of
solution space sampling approach,” in Proceedings of IEEE
signal-to-noise patterns in infinite and finite receiving antenna
Antennas and Propagation Society International Symposium,
arrays,” IEEE Transactions on Antennas and Propagation, vol.
vol. 2, pp. 568–571, Boston, Mass, USA, July 2001.
52, no. 12, pp. 3245–3256, 2004.
[7] B. Clerckx, C. Craeye, D. Vanhoenacker-Janvier, and C. [25] V. V. S. Prakash and R. Mittra, “Characteristic basis function
Oestges, “Impact of antenna coupling on 2 × 2 MIMO method: a new technique for efficient solution of method of
communications,” IEEE Transactions on Vehicular Technology, moments matrix equations,” Microwave and Optical Technol-
vol. 56, no. 3, pp. 1009–1018, 2007. ogy Letters, vol. 36, no. 2, pp. 95–100, 2003.
[8] A. Stjernman, “Relationship between radiation pattern corre- [26] J. Yeo, V. V. S. Prakash, and R. Mittra, “Efficient analysis of
lation and scattering matrix of lossless and lossy antennas,” a class of microstrip antennas using the characteristic basis
Electronics Letters, vol. 41, no. 12, pp. 678–680, 2005. function method (CBFM),” Microwave and Optical Technology
[9] P. Hannan, “The element-gain paradox for a phased-array Letters, vol. 39, no. 6, pp. 456–464, 2003.
antenna,” IEEE Transactions on Antennas and Propagation, vol. [27] C. Craeye, Th. Gilles, and X. Dardenne, “Efficient full-wave
12, pp. 423–433, 1964. characterisation of arrays of antennas embedded in finite
10 International Journal of Antennas and Propagation
Research Article
Mutual Coupling Compensation for
Direction-of-Arrival Estimations Using
the Receiving-Mutual-Impedance Method
Copyright © 2010 H. S. Lui and H. T. Hui. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
A short review of the receiving-mutual-impedance method (RMIM) for mutual coupling compensation in direction finding
applications using linear array is conducted. The differences between the conventional-mutual-impedance method (CMIM) and
RMIM, as well as the three different determination methods for receiving mutual impedance (RMI), will be discussed in details. As
an example, direction finding with better accuracies is used for demonstrating the superiority of mutual coupling compensation
using RMIM.
(RMIM) [10–15] have been proposed and demonstrate antenna α and segment b of antenna β, ZL corresponds to the
better mutual coupling compensation performance over the load impedance of the antenna terminal. The elements Iiα and
CMIM method. A recent review of various mutual coupling Viα of the [Iα ] and [Vα ] matrices, respectively are the induced
methods can be found in [16]. Direct comparisons of these current and the excitation voltage to the corresponding ith
methods for the same DOA estimation problem can be found segment of the antenna α, where m = 1, 2, . . . , M. Expanding
in [17]. The superiority of these methods over the CMIM from (1), this gives
relies on the fact that the mutual coupling of the receiving ⎡ ⎤⎡ ⎤
11 11 11
array is considered, as compared to the CMIM which the z11 − ZL z12 · · · z1M I11
⎢ ⎥⎢ ⎥
CMIs are determined when the array is in transmitting ⎢ 11
z21 11
z22 11 ⎥⎢ 1 ⎥
· · · z2M
⎢ ⎥⎢ I2 ⎥
mode. The differences of the mutual coupling problems ⎢ ⎥⎢ ⎥
⎢ .. .. .. ⎥ ⎢ ⎥
for transmitting and receiving arrays have been covered in ⎢ .. ⎥⎢ .. ⎥
⎢ . . . . ⎥ ⎢ ⎥
Balanis [18] and Daniel [19]. The differences between the ⎣ ⎦⎣ . ⎦
11 11 11 1
CMIM, RMIM, full-wave method and calibration method zM1 zM2 · · · zMM IM
have recently been discussed in Lui et al. [20]. ⎡ 12 12 12 ⎤⎡ ⎤ ⎡ ⎤
In this paper, a review of the RMIM in the context z11 − ZL z12 · · · z1M I12 V11
⎢ ⎥⎢ ⎢ ⎥ ⎥
of DOA estimation using linear array is presented. The ⎢ 12
z21 12
z22 12 ⎥⎢ 2 ⎥
· · · z2M ⎢V1 ⎥
⎢ ⎥⎢ I2 ⎥ ⎢ 2⎥
concept of receiving mutual impedance (RMI) was first ⎢ ⎥⎢ ⎥ ⎢ ⎥
+⎢
⎢ .. .. .. .. ⎥ ⎢ ⎥=⎢ ⎥,
⎥⎢ .. ⎥ ⎢ .. ⎥
introduced by Hui [10, 11] and applied on uniform linear ⎢ . . . . ⎦⎣ . ⎦ ⎣ . ⎥
⎥ ⎢ ⎥ ⎢
array (ULA) for DOA estimations and adaptive nulling. A ⎣ ⎦
12 12 12 2 1
review of the RMIM will be given in details in the next zM1 zM2 · · · zMM IM VM
section. In particular, the formation of RMIM, the “single- ⎡ ⎤⎡ ⎤ (3)
21 21 21
z11 − ZL z12 · · · z1M I11
mode” approximation [10] as well as the difference between ⎢ ⎥⎢ ⎥
⎢ 21
z21 21
z22 21 ⎥⎢ 1 ⎥
· · · z2M
RMIM and CMIM will be discussed in details. Throughout ⎢ ⎥⎢ I2 ⎥
⎢ ⎥⎢ ⎥
the years, there have been three methods to determine the ⎢ .. .. .. ⎥⎢ . ⎥
⎢ .. ⎥⎢ . ⎥
RMI for linear arrays [12, 13, 15]. Discussions of these ⎢ . . . . ⎥⎢ . ⎥
⎣ ⎦⎣ ⎦
determination methods will be given in details in Section 3. 21 21 21 1
zM1 zM2 · · · zMM IM
A DOA estimation example using Matrix Pencil Method
(MPM) [9, 21, 22] with better accuracies demonstrate the ⎡ 22 22 22 ⎤⎡ ⎤ ⎡ ⎤
z11 − ZL z12 · · · z1M I12 V12
superiority of mutual coupling compensation of RMIM. ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ 22
z21 22
z22 22 ⎥⎢ 2 ⎥
· · · z2M ⎢ 2⎥
⎢ ⎥⎢ I2 ⎥ ⎢ V2 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
+⎢
⎢ .. .. .. .. ⎥ ⎢ ⎥=⎢
⎥⎢ .. ⎥ ⎢ .. ⎥
⎥.
2. Receiving Mutual Impedance ⎢ . . . . ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦⎣ . ⎦ ⎣ . ⎦
Consider a simple case of a two-element dipole array. 22
zM1 22
zM2 22
· · · zMM 2
IM VM 2
According to the standard MoM analysis, the entire electro-
magnetic problem of the antenna array with each antenna As we can see, the mutual coupling effect is due to the
discretized into M segments can be written as [10, 23] off-diagonal matrix blocks, that is, Zαβ with α = / β. If
⎡ ⎤⎡ ⎤ ⎡ ⎤ the entire current distribution of the antenna [Iα ] and
[Z11 ] [Z12 ] [I1 ] [V1 ] the full knowledge of the all elements in the impedance
⎣ ⎦⎣ ⎦=⎣ ⎦, (1)
[Z21 ] [Z22 ] [I2 ] [V2 ] matrices Zαβ are known, the mutual coupling effect can
be accurately quantified and removed [7–9]. However, this
where the matrix blocks are can only be done in the modeling domain. In practice, it is
⎡ ⎤ not possible to measure the entire current distribution of the
αβ αβ αβ
⎢z11 − ZL z12 · · · z1M ⎥ antenna, that is, all the values in the [Iα ] matrix. Only the
⎢ ⎥
⎢
z22 · · · z2M ⎥ terminal voltages of the antenna terminals, that is, Vtα = I1α ZL
αβ αβ αβ
⎢ z21 ⎥
⎢ ⎥ can be measured. As a result, Hui [10] has introduced the
=⎢
Zαβ ⎢ .. .. . . .. ⎥
⎥
,
RMI based on the “single-mode” assumption, that is, we treat
⎢ . . . . ⎥
⎢ ⎥
⎣ ⎦ the current induced on the each antenna element of the array
αβ αβ αβ
zM1 zM2 · · · zMM consist of a single-mode current based on the current at the
(2) antenna terminal. For a two-element receiving array, this can
⎡ ⎤ ⎡ ⎤
I1α V1α be written as
⎢ α⎥ ⎢ α⎥
⎢I ⎥ ⎢V ⎥
⎢ 2⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥
[Iα ] = ⎢ . ⎥, [Vα ] = ⎢ . ⎥,
⎢ . ⎥ ⎢ . ⎥ U1 + Zt12 I2 = V1 , (4a)
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦
α
IM α
VM U2 + Zt21 I1 = V2 , (4b)
respectively. α and β ranges from 1 and 2 for the case of where Zt12 and Zt21 are defined as the RMI between the two
two antenna elements. The elements in the Zαβ matrix, antennas, I1 = I11 and I2 = I12 are the induced current in
αβ
zab , correspond to the impedance between the segment a of the antenna terminals, V1 and V2 are the induced voltages
International Journal of Antennas and Propagation 3
in the antenna terminal 1 and 2, respectively, and U1 and With a priori knowledge of the RMIs together with the
U2 are the terminal voltages solely based on the incoming terminal voltages Vk s from the DOA problem, the uncoupled
signals without any mutual coupling effect. The subscript t voltages Uk s can thus be determined using (6) and the
denotes that the RMI is defined based on the current and undesirable mutual coupling effect is thus compensated.
voltage parameters at the antenna terminal.
It is apparent that under the “single-mode” assumption, 3. Comparision with the
the RMIs defined in (4a) and (4b) are no longer the Conventional-Mutual-Impedence Method
elements of the Zαβ matrix in the original integral equation
formulation in (1). As shown in (4a) and (4b), the RMI varies Other than the RMIM, another well establishing method, the
as the induced terminal current Itα and voltages Vtα of the CMIM has been well known in the literature. The differences
antennas, as well as the load impedance (if ZL in (3) changes, between the two methods have not been well discussed
the entire electromagnetic problem changes and so as the and we take this opportunity to give a clear explanation.
RMI). For omnidirectional antenna elements, the current According to [1], the relationship between the voltages at
distribution remains to be the same for any angle of azimuth terminal k with other antenna terminals can be written as
angle φ in the elevation plane of θ = 90◦ (horizontal plane)
V1 = I1 Zk,1 + I2 Zk,2 + · · · + Ii Zk,i + · · · + Ik Zk,k + · · ·
[13].
Now we extend the concept to antenna array with N + IN Zk,N + VOk ,
antenna elements and each of them are terminated with the (7)
same load impedance ZL . The measured voltage at antenna
terminal Vk can be written as where Zk,i is the CMI between antenna terminals k and j, Zk,k
corresponds to the self impedance of the antenna element
k, Ii corresponds to the induced current at the antenna
Vk = ZL Ik = Uk + Wk , (5a) terminal i and VOk is defined as the open-circuit voltage at
terminal k when all other antenna elements under open-
where circuit condition, that is, Ii = 0 for i = 1, 2, . . . , N. The
relationship between the terminal voltage and current can be
Wk = Ztk,1 I1 + Ztk,2 I2 + · · · + Ztk,k−1 Ik−1 + Ztk,k+1 Ik+1 + · · · given by
+ Ztk,N IN . Vit
(5b) I1t = − , (8)
ZL
Once again Uk is the measured terminal voltage due to the where ZL is the load impedance terminated at the feed. With
incoming signal alone, Wk is the voltage due to the mutual the open-circuit condition and substituting (8) into (7), the
coupling effect from other array elements, Ztk,i is the RMI relationship between the open-circuit voltages and terminal
between antenna elements k and i, Ii is the current induced at voltages can be written as
the antenna terminal, given by Ii = Vi /ZL for i = 1, 2, . . . , N. ⎡ ⎤
Z11 Z12 Z1N
Rearranging (5a) and (5b) the relationship between Uk s and ⎢1 + Z ··· ⎥ ⎡ ⎤
⎢ L ZL ZL ⎥ V1
Vk s can be written as ⎢ ⎥ ⎢ ⎥
⎢ Z21 Z Z2N ⎥ ⎢V ⎥
⎡ ⎤ ⎢ 1 + 22 · · · ⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥
Zt1,2 Zt1,N −1 Zt1,N ⎢ ZL
⎢
ZL ZL ⎥
⎥ ⎢ . ⎥
⎢ 1 − ··· − − ⎥ ⎢ .. .. .. .. ⎥ ⎢ . ⎥
⎢ ZL ZL ZL ⎥ ⎢ . ⎥ ⎢ . ⎥
⎢ ⎥ ⎢ . . . ⎥ ⎣ ⎦
⎢ ⎥ ⎢ ⎥
⎢ 2,N −1 2,N ⎥ ⎣ ZN1 ZNN ⎦ VN
⎢ Z 2,1
Z Z ⎥ ZN2 1×N
⎢ − t 1 ··· −
t
−
t ⎥ ··· 1 + (9)
⎢ ZL ⎥ ZL ZL ZL
⎢ ZL ZL ⎥ N ×N
⎢ .. .. .. .. ⎥ ⎡ ⎤
⎢ .. ⎥
⎢ . . . . . ⎥ VO1
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢V ⎥
⎢ Z N −1,1 Z N −1,2
Z N −1,N ⎥
⎢ O2 ⎥
⎢ t t t ⎥ ⎢ ⎥
⎢− − ··· 1 − ⎥ =⎢ . ⎥ .
⎢ Z Z Z ⎥ ⎢ . ⎥
⎢ L L L ⎥ ⎢ . ⎥
⎢ ⎥ ⎣ ⎦
⎢ N,N −1 ⎥ (6)
⎣ ZtN,1 Zt N,2
Zt ⎦
− − ··· − 1 VON 1×N
ZL ZL ZL N ×N
⎡ ⎤ ⎡ ⎤ Comparing CMIM and RMIM, there are three main differ-
V1 U1 ences between the two. They are:
⎢ ⎥ ⎢ ⎥
⎢ V2 ⎥ ⎢ U2 ⎥
⎢ ⎥ ⎢ ⎥ (1) The CMI is determined when the antenna array is
⎢ ⎥ ⎢ ⎥
⎢ . ⎥ ⎢ . ⎥ in transmitting mode while the RMI is determined
×⎢
⎢ .. ⎥
⎥ =⎢
⎢ .. ⎥
⎥ .
⎢ ⎥ ⎢ ⎥ when the array is in receiving mode. As discussed in
⎢ ⎥ ⎢ ⎥
⎢VN −1 ⎥ ⎢UN −1 ⎥ [18–20], the mutual coupling problems of transmit-
⎣ ⎦ ⎣ ⎦
ting and receiving arrays are two different problems
VN N ×1
UN N ×1 in general and they should be treated separately.
4 International Journal of Antennas and Propagation
(2) The CMI is defined when the antenna element is The main drawback of both method 1 and 2 is that
open circuit while the RMI is defined when the only two antenna elements are considered at a time and
antenna is loaded. The current distribution of the the scattering effect from other elements (if there are more
antenna varies with different load impedance. The than two elements in the array) are not taken into account.
antenna is usually terminated with load impedance in According to (5a) and (5b), the mutual coupling problem
practice. Strictly speaking, an open-circuit assump- for a N-element receiving array can be written as
tion provides only an approximate solution to the Vk = Uk + Ztk,1 I1 + Ztk,2 I2 + · · · + Ztk,k−1 Ik−1 + Ztk,k+1 Ik+1
mutual coupling characterization.
(3) The two mutual impedances are defined differently + · · · + Ztk,N IN .
and thus the decoupling matrices are different ((6) (11)
and (9)). In particular, notice that the knowledge of Consider all the voltages and current values are found as
self-impedance is not required for the RMIM. described above, there are N − 1 unknowns in (11) (N − 1
RMIs) which cannot be solved mathematically. As a result,
4. Determination Methods of only two antennas can be considered in one time in the
previous two methods and removal of “other” antenna
the Receiving Mutual Impedance elements is needed when N > 2. Strictly speaking, this
A discussion of the three determination methods of the method only provides an accurate solution for the case of two
RMI is covered. The motivations, differences and limitations antenna elements. When there are more than two antennas,
between these methods are discussed in details. the “extra” antennas have to be removed and thus the
electromagnetic problem has changed. The scattering effect
4.1. Method 1. In this method, the RMI is determined by from “other elements” has not been taken into consideration
first taking the current distribution of an antenna element and thus alternative solution is required to handle such
under plane-wave excitation. Here, a two-element dipole drawback.
array is considered and the incident wave is coming from
the elevation plane of θ = 90◦ (perpendicular to the dipole 4.3. Method 3. For antenna arrays with omnidirectional
element). Analytical expression of the current distribution antenna elements, the RMI remains to be a constant as
can be found in the literature [10]. In the computational the azimuth angle φ varies. Consider the terminal voltages
domain, such current distribution is then used to excite are measured at φn and repeat the measurement at N − 1
an antenna element in the array and the coupled induced incoming aspects, (11) can be written as
terminal currents Ii and voltages Vi in other antenna ⎡ φ φ
⎤
elements can then be obtained. The terminal voltages that Vk 1 − Uk 1
⎢ ⎥
⎢ ⎥
are solely under plane wave excitation Ui are determined by ⎢ V φ2 − Uk ⎥
φ2
⎢ k ⎥
considering the current distribution of an antenna element ⎢ ⎥
⎢ .. ⎥
with antenna k removed. The RMIs of the two antenna ⎢ . ⎥
⎢ ⎥
elements k and i can be calculated by rearranging (4a) or ⎢ φN −2 φN −2 ⎥
⎢V − Uk ⎥
⎢ k ⎥
(4b) as ⎣ ⎦
φ φN −1
V − Uk Vk − Uk Vk N −1 − Uk
Ztk,i = k = ZL . (10) (N −1)×1
Ii Vi ⎡ ⎤
φ φ1 φ φ
I1 1 · · · Ik−1 · · · IN
1 1
The limitation of this method is that the accuracy of the Ik+1
⎢ ⎥
determined RMI is strongly dependant on the accuracies ⎢ φ2 φ2 ⎥
⎢ I · · · Ik−1 Ik+1 · · · IN ⎥
φ2 φ2
of the approximated current distribution. In practice, the ⎢ 1 ⎥
⎢ ⎥
current distribution varies when are more than one signal ⎢ . .. .. .. ⎥
=⎢
⎢ . . ..
. . .
..
. . ⎥
⎥
or when the incoming elevation angle deviates from θ = 90◦ . ⎢ ⎥
⎢ φN −2 ⎥
Furthermore, another limitation is that the RMI can only be ⎢I · · ·
φN −2 φN −2
· · ·
φN −2 ⎥
⎢1 I k−1 I k+1 I N ⎥
determined via numerical modeling. Alternative solution is ⎣ ⎦
φN −1 φN −1 φN −1 φN −1
thus required if one would like to measure the RMI. I1 · · · Ik−1 Ik+1 · · · IN (N −1)×(N −1)
Table 1: DOA estimation using MPM using a ULA with 8 antenna elements with the element separation of Δ = 0.3λ. DOA estimation is
based on the decoupled voltages using RMI determined based on methods 2 and 3.
φ3 = 100◦ Z
dH
φ2 = 35◦
φ1 = 20◦
Δ y
h
Y X
Antenna elements
x Figure 2: The NMHA with L = 25 mm, dH = 2.6 mm, h = 3 mm
Figure 1: The top view of the DOA estimation problems. The ULA and dw = 0.6 mm.
is lined in the x axis and the incoming signals are coming from the
azimuth angles of φ1 = 20◦ , φ2 = 35◦ and φ3 = 100◦ . The antenna
separation is denoted as Δ.
again. Normal-Mode Helical Antenna (NMHA) element
shown in Figure 2 is considered. Three signals coming from
the azimuth directions of φ1 = 20◦ , φ2 = 35◦ , and φ3 = 100◦
where the superscripts φn on the voltages and currents terms (all from the elevation angle of θ = 90◦ ) shown in Figure 1
correspond to the voltages and currents measured under are considered. The incoming signals are contaminated with
the plane-wave illumination from the azimuth angle of φn , Additive Gaussian White Noise (AGWN) with a Signal-to-
respectively. The RMI Ztk,i s can be determined by solving Noise Ratio (SNR) of 10 dB. The entire simulations are
(11), where i = 1, 2, . . . , k − 1, k + 1, . . . , N. done in full-wave electromagnetic solver FEKO [24] so that
As previously mentioned, the first advantage this method the scattering effect from all antenna elements is taken
does not require the removal of antenna elements such that into account. The signals are first compensated using the
the scattering effect from all antenna elements is taken into RMIM based on the RMI determined using method 2 and
consideration in the calculation when we calculate/measure a Matrix Pencil Method (MPM) algorithm is used for DOA
Vk . Preliminary results in [15] have first demonstrated that estimations. The results out of 1000 simulations are tabulated
better accuracies can be obtained for DOA estimations using in Tables 1 and 2. For the case with an element separation
the RMI determined based on method 3 as compared to of 0.3λ, there is not significant differences for the estimated
method 2. More importantly, as the element separation of the incoming directions when the RMI is determined using
antenna elements becomes smaller, the scattering effect from the two methods. However, as the element separation is
other elements becomes significant and RMI determined reduced to 0.15λ, as shown in Table 2, it is found that
using method 2 may fail to compensate the mutual coupling the estimated directions of signal 1 and signal 2 becomes
effect. 26◦ and 80◦ , respectively, which are not correct. Using the
To demonstrate the superiority of method 3, a DOA esti- RMI determined using method 3, three incoming signals
mation example using two 8-element uniform linear arrays can be clearly resolved with less than 0.5◦ of bias, standard
with element separations of 0.3λ and 0.15λ is considered. deviations, root-mean-square errors (RMSEs).
Decoupling using CMIM is not included here as it has Further investigations about the solution of the RMI
previously been demonstrated [10, 11, 14, 15] that RMIM determined using different number of incidence, can be
is better than CMIM and we are not going to repeat it here found in our recent publication [25]. The only limitation of
6 International Journal of Antennas and Propagation
Table 2: DOA estimation using MPM and a ULA with 8 antenna elements with the element separation of Δ = 0.15λ. DOA estimation is
based on the decoupled voltages using the RMI determined based on methods 2 and 3.
this method is that it is only applicable to omnidirectional [4] B. Friedlander and A. J. Weiss, “Direction finding in the
antenna elements. Alternative solutions are thus required for presence of mutual coupling,” IEEE Transactions on Antennas
nonomnidirectional antenna elements [26, 27]. and Propagation, vol. 39, no. 3, pp. 273–284, 1991.
[5] K. R. Dandekar, H. Ling, and G. Xu, “Experimental study
of mutual coupling compensation in smart antenna applica-
5. Conclusions tions,” IEEE Transactions on Wireless Communications, vol. 1,
no. 3, pp. 480–487, 2002.
A review of the RMIM for mutual coupling compensation [6] R. Hossa and M. Bialkowski, “Mutual coupling compensation
in the scope of DOA estimations using ULA is presented. in narrowband small linear-antenna arrays,” Microwave and
In particular, the definition of RMI, the “single-mode” Optical Technology Letters, vol. 40, no. 5, pp. 391–396, 2004.
approximation, as well as the differences between CMIM [7] K. M. Pasala and E. M. Friel, “Mutual coupling effects and
and RMIM are discussed in details. The advantages and their reduction in wideband direction of arrival estimation,”
limitations of the three determination methods of RMI have IEEE Transactions on Aerospace and Electronic Systems, vol. 30,
also covered followed by a DOA estimation example which no. 4, pp. 1116–1122, 1994.
demonstrates the feasibilities and superiorities of mutual [8] R. S. Adve and T. K. Sarkar, “Compensation for the effects of
coupling compensation using RMI. In addition to theoretical mutual coupling on direct data domain adaptive algorithm,”
studies, we have recently demonstrated experimentally that IEEE Transactions on Antennas and Propagation, vol. 48, no. 1,
the RMIM can be used for mutual coupling compensation pp. 86–94, 2000.
[28]. [9] C. K. E. Lau, R. S. Adve, and T. K. Sarkar, “Minimum norm
The RMIM has also been applied to mutual coupling mutual coupling compensation with applications in direction
compensation of DOA estimations using circular array [29] of arrival estimation,” IEEE Transactions on Antennas and
and phase array in Magnetic Resonance Imaging (MRI) Propagation, vol. 52, no. 8, pp. 2034–2041, 2004.
[30–32]. Recently, the RMI has been adopted to model the [10] H. T. Hui, “A practical approach to compensate for the mutual
mutual coupling effect of Multiple-Input Multiple-Output coupling effect in an adaptive dipole array,” IEEE Transactions
(MIMO) system and new models for channel capacities of on Antennas and Propagation, vol. 52, no. 5, pp. 1262–1269,
2004.
the MIMO systems have been developed [33, 34]. It is no
doubt that the RMIM is an important model for mutual [11] H. T. Hui, “Improved compensation for the mutual coupling
effect in a dipole array for direction finding,” IEEE Transactions
coupling characterization and it is not difficult to see that
on Antennas and Propagation, vol. 51, no. 9, pp. 2498–2503,
such model can be used in other radar, biomedical, and 2003.
communication applications.
[12] H. T. Hui, “A new definition of mutual impedance for
application in dipole receiving antenna arrays,” IEEE Antennas
Acknowledgments & Wireless Propagation Letters, vol. 3, no. 1, pp. 364–367, 2004.
[13] H. T. Hui, H. P. Low, T. T. Zhang, and Y. L. Lu, “Receiving
This work was supported in part by the National University mutual impedance between two normal-mode helical anten-
of Singapore (NUS) under Grant no. R-263-000-469-112 and nas (NMHAs),” IEEE Antennas and Propagation Magazine, vol.
in part by the US ONR research fund under the Project no. 48, no. 4, pp. 92–96, 2006.
PR. NO. 09PR03332-01. [14] T. T. Zhang, H. T. Hui, and Y. L. Lu, “Compensation
for the mutual coupling effect in the ESPRIT direction
finding algorithm by using a more effective method,” IEEE
References Transactions on Antennas and Propagation, vol. 53, no. 4, pp.
[1] I. J. Gupta and A. A. Ksienski, “EEffect of mutual coupling 1552–1555, 2005.
on the performance of adaptive arrays,” IEEE Transactions on [15] H.-S. Lui, H. T. Hui, and M. S. Leong, “A new calculation
Antennas and Propagation, vol. 31, no. 5, pp. 785–791, 1983. method of the receiving mutual impedance for linear antenna
[2] R. E. Collin and F. J. Zucker, Eds., Antenna Theory, McGraw- array,” in Proceedings of Asia Pacific Microwave Conference
Hill, New York, NY, USA, 1968. (APMC ’08), Macau, China, December 2008.
[3] C.-C. Yeh, M.-L. Leou, and D. R. Ucci, “Bearing estimations [16] H. T. Hui, “Decoupling methods for the mutual coupling
with mutual coupling present,” IEEE Transactions on Antennas effect in antenna arrays: a review,” Recent Patents on Engineer-
and Propagation, vol. 37, no. 10, pp. 1332–1335, 1989. ing, vol. 1, pp. 187–193, 2007.
International Journal of Antennas and Propagation 7
[17] Y. Wu and Z. Nie, “On the improvement of the mutual cou- [34] S. Lu, H. T. Hui, and M. Bialkowski, “Performance analy-
pling compensation in DOA estimation,” Journal of Systems sis of multiple-input multiple-output orthogonal frequency
Engineering and Electronics, vol. 19, no. 1, pp. 1–6, 2008. division multiplexing systems under the influence of antenna
[18] C. A. Balanis, Antenna Theory and Design, John Wiley & Sons, mutual coupling effect,” IET Microwaves, Antennas and Prop-
New York, NY, USA, 2nd edition, 1997. agation, vol. 3, no. 2, pp. 288–295, 2009.
[19] J. Daniel, “Mutual coupling between antennas for emission
or reception—application to passive and active dipoles,” IEEE
Transactions on Antennas and Propagation, vol. 22, no. 2, pp.
347–349, 1974.
[20] H. S. Lui, H. T. Hui, and M. S. Leong, “A note on the mutual
coupling problems in transmitting and receiving antenna
array,” to appear in IIEEE Antennas and Propagation Magazine.
[21] T. K. Sarkar and O. Pereira, “Using the matrix pencil method
to estimate the parameters of a sum of complex exponentials,”
IEEE Antennas and Propagation Magazine, vol. 37, no. 1, pp.
48–55, 1995.
[22] T. K. Sarkar, M. C. Wicks, M. Salazar-Palma, and R. J.
Bonneau, Smart Antennas, John Wiley & Sons, New York, NY,
USA, 3rd edition, 2003.
[23] R. F. Harrington, Field Computation by Moment Methods, IEEE
Press, New York, NY, USA, 1993.
[24] FEKO EM Software & Systems S. A., (Pty) Ltd, 32 Techno
Lane, Technopark, Stellenbosch, 7600, South Africa.
[25] H. S. Lui and H. T. Hui, “Effective mutual coupling compensa-
tion for direction-of-arrival estimations using a new, accurate
determination method for the receiving mutual impedance,”
Journal of Electromagnetic Waves and Applications, vol. 24, no.
2-3, pp. 271–281, 2010.
[26] Y. Yu, H. T. Hui, and M. S. Leong, “Receiving mutual
impedance of a compact patch antenna array with two
elements in a rich multipath environment,” in Proceedings
of Asia Pacific Microwave Conference (APMC ’09), Singapore,
December 2009.
[27] W. M. Lee, Y. Yap, and H. T. Hui, “Receiving mutual
impedance between two printed dipole antennas,” Microwave
and Optical Technology Letters, vol. 49, no. 4, pp. 892–896,
2007.
[28] Lui H. S., Y. Yu, H. T. Hui, and M. S. Leong, “Experimental
study of mutual coupling compensation in direction finding
using a compact antenna array,” in Proceeding of the IEEE Asia
Pacific Symposium on Electromagnetic Compatibilities, Beijing,
China, April 2010.
[29] T. T. Zhang, Y. L. Lu, and H. T. Hui, “Compensation for the
mutual coupling effect in uniform circular arrays for 2D DOA
estimations employing the maximum likelihood technique,”
IEEE Transactions on Aerospace and Electronic Systems, vol. 44,
no. 3, pp. 1215–1221, 2008.
[30] H. T. Hui, “An effective compensation method for the mutual
coupling effect in phased arrays for magnetic resonance
imaging,” IEEE Transactions on Antennas and Propagation, vol.
53, no. 11, pp. 3576–3583, 2005.
[31] H. T. Hui, B. K. Li, and S. Crozier, “A new decoupling method
for quadrature coils in magnetic resonance imaging,” IEEE
Transactions on Biomedical Engineering, vol. 53, no. 10, pp.
2114–2116, 2006.
[32] B. K. Li, H. T. Hui, C. H. Yang, and S. Crozier, “A new
decoupling method for phased arrays in magnetic resonance
imaging: an experimental approach,” IEE Proceedings, Science,
Measurement and Technology, vol. 2, no. 5, pp. 317–325, 2008.
[33] S. Lu, H. T. Hui, and M. Bialkowski, “Optimizing transmission
strategy in single-user MIMO systems under the influence
of antenna mutual coupling,” IEEE Antennas And Wireless
Propagation Letters, vol. 7, pp. 287–290, 2008.
Hindawi Publishing Corporation
International Journal of Antennas and Propagation
Volume 2010, Article ID 690713, 9 pages
doi:10.1155/2010/690713
Research Article
Bandwidth Enhancement of Antenna Arrays Utilizing Mutual
Coupling between Antenna Elements
Copyright © 2010 Min Wang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The mutual coupling effect between antenna elements on an array’s bandwidth is investigated using scattering parameters instead
of the mutual impedance. First, an approximate expression is derived for matched voltage standing wave ratio (VSWR) bandwidth
of a tuned antenna, which reveals that the bandwidth is inversely proportional to the magnitude |Γ0 (ω0 )| of the frequency derivative
of the reflection coefficient. Next, considering linear antenna arrays with corporate feed as an example, closed-form expressions of
the reflection coefficient are derived at the input port of the feeding network, which shows that the active reflection coefficient of an
array is the linear superposition of elements’ passive reflection coefficient S11 and the mutual coupling coefficient S12 from adjacent
elements. The VSWR bandwidth expressions for an array imply that bandwidth enhancement of the overall array can be achieved
when the element passive reflection coefficient S11 and mutual coupling S12 are cancelled, as well as the frequency derivatives S11
and S12 also cancel each other. Slot arrays and a two-element Vivaldi array are investigated to verify the validity of our theoretical
analysis. Numerical and experimental results are presented to successfully demonstrate the bandwidth enhancement of antenna
arrays utilizing mutual coupling effect.
network is assumed to be constructed by an equal-power where the subscript notifies parameters for an antenna
divider with equal phases. Due to mutual coupling between matched at ω0 .
array elements, conversion of energy implies unavoidable The reflection coefficient of the antenna is then
existence of cross-coupling between the feed lines [7]. Taking
Z0 (ω) − Zch
the array system as a cascaded connection of the array and the Γ0 (ω) = . (2)
feeding network, closed-form expressions for the reflection Z0 (ω) + Zch
coefficient at the feed input of the array can be derived. For a The matched VSWR bandwidth for an antenna tuned at
two-element array, the reflection coefficient is approximately a frequency ω0 is defined as the difference between the two
the linear superposition of scattering coefficients S11 and S12 frequencies on either side of ω0 at which the VSWR equals
of the array when treated as a two-port network. For a linear a constant s or, equivalently, at which |Γ0 (ω)|2 equals α =
array with all elements excited simultaneously, the active (s − 1)2 /(s+1)2 . It is derived in [5] that the fractional matched
reflection coefficient is the linear superposition of elements’ VSWR bandwidth FBWV (ω0 ) takes the form
passive reflection coefficient and the mutual coupling from
adjacent elements, which is similar to that of the array 4 βR0 (ω0 )
,
α s−1
with two identical elements. Therefore, the mutual coupling FBWV (ω0 ) ≈ β= = √ ≤1
effect on array bandwidth can be expressed in terms of S- ω0 Z (ω0 )
0 1−α 2 s
parameters. The bandwidth expressions for an array imply (3)
that bandwidth enhancement of the overall array can be which holds for tuned antennas under the sufficient condi-
achieved when the element passive reflection coefficient tions that X0 (ω0 ) and R0 (ω0 ) do not change greatly over the
S11 and mutual coupling S12 are cancelled, as well as the bandwidth (conditions that hold if FBWV (ω0 ) 1, which
frequency derivatives S11 and S12 also cancel each other. can always be satisfied if β is chosen small enough).
Finally, numerical simulation and experimental verifica- In this paper, we would like to analyze mutual coupling
tion are carried out to verify the bandwidth enhancement. effect on the bandwidth enhancement of antenna arrays. It
Printed slot antennas have extensive applications in phased is known that mutual coupling effect in an antenna array is
arrays, satellite communication systems, and airborne sys- commonly modeled as a change in the driving impedance
tems due to their compact size and high efficiency [8–11] and of the elements and it is usually referred to as mutual
are therefore considered here. Two- and four-element slot impedance variation [4]. For the simplest array that consists
arrays are used to study their mutual coupling and to demon- of two identical elements with equal excitation, we can easily
strate the bandwidth enhancement by invoking suitable obtain the elements’ driving-point impedances as follows:
mutual coupling. Experimental prototypes are fabricated
and measurement results verify the validity of our theoretical Z1d = Z11 + Z12 , (4)
analysis. Furthermore, a two-element Vivaldi array has
been investigated based on our formulation, which has a where Z1d , Z11 , and Z12 represent the element’s driving-point
combination of two or more resonances and antiresonances. impedance, the element’s self-impedance, and the mutual
It is well known that performances of isolated Vivaldi anten- impedance, respectively.
nas and elements in an array configuration differ greatly For an array with all elements excited simultaneously,
due to strong mutual coupling effect [12–15]. Particular it is not the element’s self-impedance but the driving-point
parametric optimizations are conducted to obtain suitable impedance that should be matched. If the array is tuned at
mutual coupling and element’s passive reflection coefficient a frequency ω 0 (normally, ω
0 is not equal to the resonant
for the array. Simulated results further demonstrate that or antiresonant frequency ω0 of elements due to mutual
appropriate mutual coupling between elements results in coupling effect), the conductance X1d0 (ω 0 ) equals zero and
bandwidth enhancement and also provides a useful indicator the driving impedance is
for parametric optimization of Vivaldi arrays.
0 ) = Z110 (ω
Z1d0 (ω 0 ) + Z120 (ω
0 ) = R1d0 (ω
0 ). (5)
2. Theoretical Analysis Substituting (5) into (3), we obtain the fractional matched
VSWR bandwidth FBWV (ω 0 ) for the array as
2.1. Matched VSWR Bandwidth. Yaghjian and Best investi-
gated the bandwidth and Q of a general single-feed (one- 0)
4 βRd 0 (ω 0)
4 βRd 0 (ω
port) lossy or lossless linear antenna tuned to resonance 0) ≈
FBWV (ω = .
0 Z110
ω
0 ) + Z120
(ω 0 )
(ω
or antiresonance and gave the formula for the fractional 0 Zd0
ω 0 )
(ω
matched VSWR bandwidth FBWV (ω0 ) [5]. Considering a (6)
general transmitting antenna composed of electromagneti-
cally linear materials and fed by a feed line with characteristic It is shown that from (3) and (6), provided we match
impedance Zch that carried just one propagating mode and 0 ) ≈ R0 (ω0 ) and appropriately tune
the array with R1d0 (ω
tuned at a frequency ω0 , the tuned complex input impedance impedances with Z110 0 ) ≈ −Z120
(ω 0 ), the matched
(ω
can be written as VSWR bandwidth FBWV (ω 0 ) of the array would be greatly
enhanced.
Through above analysis, we show that mutual coupling
Z0 (ω) = R0 (ω) + jX0 (ω), (1) effect on array bandwidth could be analyzed in the view of
International Journal of Antennas and Propagation 3
the mutual impedance. However, the analysis associated with For a lossless 3-dB T-branch power divider [15], we know
mutual impedance is usually quite complicated, especially that
for a larger array. The expressions of element’s open-circuit ⎡
1 1 ⎤
voltage are usually needed to calculate the mutual impedance 0 √ √
⎡ ⎤ ⎢ ⎢ 2 2⎥⎥⎡ a ⎤
while array element’s port voltage is difficult to be measured b ⎢
for most practical configurations or even to be defined in ⎢ ⎥ ⎢ √1 1 1 ⎥ ⎥⎢ ⎥
⎣a1 ⎦ = ⎢ − ⎥⎣b1 ⎦. (11)
some situations. On the other hand, S-parameters are a ⎢ 2 2 2 ⎥
a2 ⎢ ⎥ b2
⎣ 1 1 1⎦
versatile tool to analyze microwave networks. Furthermore, √ −
the matched VSWR bandwidth is naturally defined using the 2 2 2
reflection coefficient. Therefore, we would rather investigate
the mutual coupling effect on an array’s bandwidth in For the antenna array, we consider only passive and identical
the view of S-parameters. Therefore, we go back to the elements, and then we have
definition of matched VSWR bandwidth, (ω+ − ω− ), which
b1 S11 S12 a1
is determined by = . (12)
b2 S12 S 11 a2
2
|Γ0 (ω± )| = α. (7) Due to the symmetry of the array and the power divider, it
can be easily derived from (11) and (12) that
Expanding the left-hand side of (7) in a Taylor series
1
aboutω0 , we find a1 = a2 = √ a,
2
(13)
Γ (ω0 )2 (Δω± )2 ≈ α (8) 1
0 b1 = b2 = √ a(S11 + S12 ).
2
under the condition Γ0 (ω0 ) ≈ 0 and the assumption that Finally, we obtain the reflection coefficient of the overall
the O[(Δω± )3 ] terms are negligible, where Γ0 (ω) denotes network at the input port of the divider as
the frequency derivative of the reflection coefficient. This
assumption is generally satisfied if |Δω± /ω| 1. The b
Γ= = S11 + S12 . (14)
solutions to (8) for Δω± are a
√ Equation (14) shows that the reflection coefficient Γ of the
α
Δω± ≈ ±
Γ (ω0 ) , (9) overall network is the superposition of S11 and S12 for the
0 two-antenna array.
We further discuss a linear array with 4 identical elements
which results in the fractional matched VSWR bandwidth fed through a 4-way power divider which is shown in
FBWV (ω0) Figure 2, with the incident and reflected waves labeled at
the ports. Similarly, considering the system as the cascaded
√
ω+ − ω− Δω+ − Δω− 2 α connection of a five-port network (power divider) with a
FBWV (ω0 ) = = ≈ .
ω0 ω0 ω0 Γ0 (ω0 ) four-port network, we can write the relationship of those
(10) input and output signals as
⎡ 1 1 1 1 ⎤
Equation (10) holds true for tuned antennas under the ⎢0 2 2 2 2 ⎥
⎢ ⎥
sufficient condition that Γ0 (ω) does not change greatly over ⎡ ⎤ ⎢ 1 ⎥
⎢1 3 1 1 ⎥⎡ ⎤
the bandwidth. It also reveals that the fractional matched b ⎢ − ⎥ a
⎢a ⎥ ⎢ 2 4 4 4 4 ⎥⎥⎢ ⎥
VSWR bandwidth for a tuned antenna is inversely propor- ⎢ 1⎥ ⎢ ⎢ b1 ⎥
⎢ ⎥ ⎢ ⎢ 1 1 3 1 1 ⎥⎥⎢ ⎥
tional to the absolute value of the derivative of the reflection ⎢a2 ⎥ = ⎢ − ⎢ b2 ⎥ , (15)
⎢ ⎥ ⎢2 4
⎣a3 ⎦ ⎢ 4 4 4 ⎥⎥⎢ ⎥
coefficient with respect to ω. Note that (10) is a general ⎥ ⎣ b3 ⎦
⎢1 1 1 3 ⎥
1 ⎥ b
equation for antennas that can be used to investigate mutual a4 ⎢ − 4
⎢2 4 4 4 4 ⎥
coupling effect on bandwidth in terms of S-parameters. ⎢ ⎥
⎣1 1 1 1 3⎦
−
2.2. Active Reflection Coefficient of Arrays. Figure 1 shows 2 4 4 4 4
⎡ ⎤ ⎡ ⎤⎡ ⎤
a two-element array fed through a 3-dB T-branch power b1 S11 S12 0 0 a1
divider, which can be considered as the cascaded connection ⎢b ⎥ ⎢S S S ⎥⎢ ⎥
⎢ 2 ⎥ ⎢ 12 22 23 0 ⎥⎢a2 ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥. (16)
of a three-port network (T-branch power divider) with a ⎣b3 ⎦ ⎣ 0 S23 S22 S12 ⎦⎣a3 ⎦
two-port network. The incident and reflected waves are b4 0 0 S12 S11 a4
labelled at the ports as a, a1 , a2 , b, b1 , and b2 . The
relationships between incident and reflected waves can be The fact that the array is symmetric and array elements are
established by the S-matrix of these two networks, so that identical is considered in (16), and only the mutual coupling
the reflection coefficient of the overall network is derived. between adjacent elements is taken into account.
4 International Journal of Antennas and Propagation
50 mm d d
1 mm
21 mm
(a) Single slot antenna 50 mm (b) Array no. 1 (c) Array no. 2
80 mm
d d
1 mm
50 mm
21 mm
for the whole array. Firstly, the element passive reflection 3.1. Slot Antenna Array. We consider a single-slot antenna
coefficient S11 and mutual coupling S12 should be out of and an array of two identical slots shown in Figure 3. Their
phase and be cancelled as shown in (23), so that the array resonant frequencies are about 5.8 GHz. The substrate used is
can be tuned. Furthermore, the frequency derivatives S11 and Rogers RO4230 (εr = 3) of thickness 0.762 mm. The slots are
S12 should be out of phase to cancel each other. We will cut on the ground plane of the substrate and the microstrip
conduct numerical simulation and experimental verification feed lines are printed on the top side. Array no. 1, which is fed
to demonstrate this bandwidth enhancement. through two independent microstrip lines, is used to obtain
S-parameters of the array. Array no. 2, which is fed through
a T-branch power divider, is the one we hope to investigate.
3. Simulated and Measured Results
Obviously, it is the element spacing d that mainly
The closed-form expressions derived in the previous section influences the mutual coupling between elements. So the
imply that the bandwidth enhancement of the overall array phase differences between S11 and S12 of array no. 1
can be achieved by invoking appropriate mutual coupling with different spacings are firstly calculated and shown in
between elements. In this section, numerical simulation Figure 4. Note that the element spacing should be less than
and experimental verification are conducted to investigate one wavelength to avoid grating lobes and evoke mutual
mutual coupling effect on array bandwidth. Two types of coupling strong enough to achieve bandwidth enhancement.
antenna are underanalysed. Firstly, we consider slot antenna Figure 4 indicates that the phase difference between S11
arrays, which are normally narrowband and only have one and S12 changes with the element spacing and has a sharp
antiresonant frequency. Then, a broadband two-element turn at the resonant frequency. Therefore, we can evaluate
Vivaldi array is investigated, which has a combination of the spacing to achieve appropriate amplitudes and phase
two or more resonances and antiresonances that are so difference of ±π for S parameters approximately at a
close together to build a broad bandwidth. To optimize the small deviation from the resonant frequency, meanwhile,
array bandwidth, we will vary element spacing d to match physical parameters are optimized to ensure proper phase
the active array with appropriate amplitudes and phase relation for their derivatives with respect to frequency. In
differences for S11 and S12 as shown in (23). Then, for a given particular, we found the optimized spacing d = 14 mm
element spacing d, particular parametric optimizations are (about a quarter wavelengths at the resonant frequency),
conducted to achieve bandwidth enhancement for the arrays as well as the optimized physical parameters, with which
by cancelling the derivatives of S-parameters as shown in the bandwidth of the antenna array obviously improved.
(24). The corresponding simulated results for phase difference
6 International Journal of Antennas and Propagation
−40 250
−60 200
−80
−100 100
−120 50
−140 0
−160 −50
−180 −100
−200 −150
−220 −200
4 4.5 5 5.5 6 6.5 7 4 4.5 5 5.5 6 6.5 7
Frequency (GHz) Frequency (GHz)
0 −5
Return loss (dB)
−5 −10
Magnitude of S-parameters (dB)
−10 −15
−15 −20
−20 −25
4 4.5 5 5.5 6 6.5 7 7.5 8
Frequency (GHz)
−25
4 4.5 5 5.5 6 6.5 7 Simulated single slot Measured single slot
Frequency (GHz) Simulated array #2 Measured array #2
S11
Figure 7: Simulated and measured return loss results of single-slot
S12 and two-slot array no. 2 (d = 14 mm).
8 10
9
7
8
6
5 6
5
4
4
3 3
2
2
1
5 5.5 6 6.5 7 5 5.5 6 6.5 7
Frequency (GHz) Frequency (GHz)
Figure 8: Simulated and measured gain results of single-slot and Figure 10: Simulated and measured gain results of single-slot and
two-slot array no. 2 (d = 14 mm). four-slot array no. 4 (d = 14 mm).
0
of array no. 4 is improved from 0.50 GHz (6.00 GHz–
−5 5.50 GHz) to 1.09 GHz (6.69 GHz–5.60 GHz) according to
simulated results and from 0.40 GHz (5.95 GHz–5.55 GHz)
−10 to 1.38 GHz (6.78 GHz–5.40 GHz) according to measured
Return loss (dB)
ones. The realized gain results are also provided in Figure 10,
−15 which show that moderately improved gain results have been
achieved as well as the bandwidth enhancement.
−20
−25 3.2. Vivaldi Antenna Array. The tapered slot antenna (TSA)
was initially introduced as an array element by Lewis et al. in
−30 1974 [11]. Gibson [12] later named the exponentially tapered
slot antenna as Vivaldi antenna. Because of its potential for
−35 wide-band and wide-scan arrays, ease of fabrication, and
4 4.5 5 5.5 6 6.5 7 7.5 8 convenient feed techniques, Vivaldi antenna array is a prime
Frequency (GHz) candidate for high-performance phased-array systems. In
Simulated single slot Measured single slot
this paper, Vivaldi array is chosen as an example for
Simulated array #4 Measured array #4 broadband arrays having a combination of two or more
resonances and antiresonances ω1 , ω2 , . . . , ωn , . . . , ωN (N ≥
Figure 9: Simulated and measured return loss results of single-slot 2). The bandwidth enhancement can be achieved at these
and four-slot array no. 4 (d = 14 mm). frequencies by utilizing mutual coupling between array
elements, so that the total bandwidth can be improved for
a broadband array.
Next, we consider arrays consisting of four identical We consider a two-element Vivaldi array shown in
slots shown in Figure 3 to further demonstrate our analysis. Figure 11. For each element, the stripline feed is coupled to
Array no. 3 that is fed by four independent microstrip bilateral slotline cut on the ground plane. Element parame-
lines is used to obtain the S-parameters of the array. Array ters can be subdivided into the stripline/slotline transition,
no. 4 fed through T-branch power dividers is the one the tapered slot, and the stripline stub and slotline cavity
we hope to investigate. The same substrate and element [13]. The bilateral slotline is terminated at one end with a
dimension are adopted as the two-element slot array. Par- circular slot cavity of diameter DSL , and at the other end it
ticularly, we evaluate the optimized spacing d = 14 mm opens into an exponentially tapered slotline with an opening
again for approximately achieving optimal S parameters rate of R. The stripline feed is terminated in a radial line stub
and their derivatives. The comparison of return loss results with radius RST and angle θ. The stripline/slotline transition
between a single-slot antenna and the four-slot array no. is specified by WST (stripline width) and WSL (slotline
4 is shown in Figure 9. It is seen that the bandwidth width). Distances from the transition to the slotline cavity
8 International Journal of Antennas and Propagation
10 mm 0
10.7 GHz
−5
−10
S-parameters (dB)
−15
−20
−25
z1 L z2
5.5 GHz
Figure 11: Schematic graph of Vivaldi antenna array. −5
−10
and the taper can be properly adjusted to accommodate the
transition. Broadband behavior has been obtained by appro-
Γ (dB)
−15
priately designed WST , WSL , DSL , RST , and θ with coupling
from the dominant stripline mode to the traveling wave in
tapered slot. The broadband characteristic of antenna itself −20
mainly attributes to the exponential taper profile, which is
defined by the opening rate R and two points P1 (z1 , y1 ) and −25
P2 (z2 , y2 )
−30
y = c1 eRz + c2 , (25) 4 6 8 10 12 14 16 18
Frequency (GHz)
where
R = 0.15 mm−1
y2 − y1 y1 eRz2 − y2 eRz1 R = 0.3 mm−1
c1 = , c2 = . (26)
eRz2 − eRz1 eRz2 − eRz1 R = 0.45 mm−1
The taper length L is z2 − z1 and the aperture height H is
Figure 13: The active reflection coefficient Γ with various opening
2(y2 − y1 ) + WSL .
rates R.
The Vivaldi antenna element has theoretically unlimited
bandwidth, and the upper operating frequency is mainly
limited by the transition. However, for a Vivaldi array,
the upper operating frequency is limited by the onset of RO4230 (εr = 3) of thickness 1.524 mm. After some
grating lobes, which is determined by the element spacing. optimization following [14, 15], the physical parameters are
Therefore, we usually prescribe the element spacing and chosen: H = 8 mm, WST = WSL = 0.5 mm, DSL = RST =
optimize antenna parameters to achieve desired bandwidth 2.6 mm, θ = 70◦ , L = 22 mm, and R = 0.3 mm−1 . To
of the array. When the element spacing is prescribed, the clarify the relationship between S11 and S12 , the variations
mutual coupling between elements S12 varies slightly and the of S-parameters with respect to different opening rates R
bandwidth enhancement is mainly achieved by optimizing are provided and are shown in Figure 12. The corresponding
the elements’ passive reflection coefficient S11 to be well active reflection coefficients Γ are shown in Figure 13. It is
cancelled with S12 . shown that with R = 0.3 mm−1 , the minimum operating
Here, we prescribe the element spacing d = 10 mm, frequency is about 5.5 GHz while the frequency of one
which is also the element width. The substrate used is element is about 10.7 GHz. For a specified upper operating
International Journal of Antennas and Propagation 9
100 coefficient S11 and mutual coupling S12 are well cancelled, as
well as the frequency derivatives S11 and S12 also cancel each
50
other. Both numerical simulated and experimental results
0 successfully demonstrate our analysis.
−50
−100 References
−150 [1] A. C. Ludwig, “Mutual coupling, gain, and directivity of
−200 an array of two identical antennas,” IEEE Transactions on
Antennas and Propagation, vol. 24, no. 6, pp. 837–841, 1976.
−250 [2] I. J. Gupta and A. A. Ksienski, “Effect of mutual coupling
4 6 8 10 12 14 16 18
on the performance of adaptive arrays,” IEEE Transactions on
Frequency (GHz)
Antennas and Propagation, vol. 31, no. 5, pp. 785–791, 1983.
Figure 14: The phase difference between S11 and S12 with R = [3] J. A. G. Malherbe, “Analysis of a linear antenna array
0.3 mm−1 . including the effects of mutual coupling,” IEEE Transactions
on Education, vol. 32, no. 1, pp. 29–34, 1989.
[4] C. A. Balanis, Antenna Theory-Analysis and Design, John Wiley
300
& Sons, New York, NY, USA, 3rd edition, 2005.
[5] A. D. Yaghjian and S. R. Best, “Impedance, bandwidth, and Q
200 of antennas,” IEEE Transactions on Antennas and Propagation,
vol. 53, no. 4, pp. 1298–1324, 2005.
Phase(S11 )−phase(S12 ) (◦ )
Application Article
Decoupling of Multifrequency Dipole Antenna Arrays for
Microwave Imaging Applications
Copyright © 2010 E. Saenz et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The mutual coupling between elements of a multifrequency dipole antenna array is experimentally investigated by S-parameter
measurements and planar near-field scanning of the radiated field. A multifrequency array with six dipoles is analyzed. In order to
reduce the coupling between dipoles, a planar metasurface is placed atop the array acting as superstrate. The mutual coupling of the
antenna elements in the absence and presence of the superstrate is presented comparatively. Between 3 and 20 dB mutual coupling
reduction is achieved when the superstrate is used. By scanning the field radiated by the antennas and far-field measurements of
the radiation pattern, it is observed that the superstrate confines the radiated power, increases the boresight radiation, and reduces
the endfire radiation.
Dh
h h E k
1 2 3
H
Dv
ld ld l
l t t
D
D
2h
t D
t D/2
(a) (b) (c)
Figure 1: (a) Schematic representation of the uniform metasurface. (b) Unit cell geometry: top and lateral views. (c) MF superstrate, gray
strips resonating at a low resonant frequency and black strips resonating at a high resonant frequency.
coupling coefficients, spherical near-field measurements, and resonant frequency of the metasurface, the radiated field is
far-field radiation patterns of the array were reported in spread over a larger radiating aperture enhancing its gain.
[9]. In order to design an MF metasurface, a combination
In the present paper, a novel multifrequency antenna of two unit cells tuned to two different resonant frequen-
design using metasurfaces as follows from the previous stud- cies in a chessboard-like layout was created. A schematic
ies [5–9] is experimentally investigated. The mutual coupling representation of the metasurface is shown in Figure 1(c).
coefficients, when the dipoles are radiating in free space The gray strips are tuned to the so-called Low Resonant
and with the metasurface, are calculated by S-parameters Frequency (LRF) and the black ones to the High Resonant
measurements and by planar near-field (NF) inspection of Frequency (HRF). The dimensions of both of unit cells are the
the radiated field. The design of the individual antennas was following.we have LRF: ld =6.60, l =7.32, t =0.58, D =2.30,
optimized for best impedance matching, maximum gain, and h =0.5, and εr =4; HRF: ld =5.30, l =6.17, t =0.50, D =2.30,
minimum coupling between them in the presence of the h =0.5, and εr =4. The distance between centers of cells
metasurface. In the last section, far-field measurements of (HRF to LRF) is Dh =12.71 mm and Dv =20.95 mm.
the radiation pattern with and without the superstrate are To find the transmission window of the LRF and HRF
presented to prove the benefits of the metasurface. unit cells, the transmission (T) and reflection (R) coefficients
were calculated with Ansoft-HFSS under normal incident
plane wave excitation. The unit cell excitation is depicted in
2. Metasurface the inset of Figure 2. Two wave ports were defined on the
top and bottom faces of the unit cell and Periodic Boundary
The metasurface used in this work as superstrate of the Conditions were applied on the sides in order to simulate an
planar antenna is schematically depicted in Figure 1(a). A infinite slab. The results are presented in Figure 2. Both of
detailed explanation of the physical phenomena behind this unit cells exhibit a transmission window from 9 to 11 GHz in
structure can be found in [6]. The metasurface is formed by the case of the LRF cell and from 11 to 13.6 GHz in the case
three layers. The first and third grids consist of parallel short of the HRF.
metal strips, while the second one consists of continuous Once the passband of the cells was known, dipole
wires. At a certain frequency every pair of strips in grids 1 antennas tuned to the transmission window of the cells were
and 3 (see Figure 1(a)) shows a magnetic resonant mode. The designed in order to make an MF dipole antenna array.
currents induced on individual paired metal strips cancel
each other, allowing the incident wave to propagate through
the grids. These magnetic dipole moments are local, but 3. Multifrequency Dipole Antenna Array
every strip pair in the superstrate will radiate principally
in a similar manner leading to an approximately uniform In [9], the mutual coupling between dipole antennas with
aperture phase distribution on the outer superstrate surface. this kind of metasurface as superstrate and the effect in
The key idea behind this configuration is that when the the radiation properties (e.g., radiation pattern, radiation
metasurface is illuminated by a primary source tuned to the efficiency) was analyzed for a square array formed by 2 × 2
International Journal of Antennas and Propagation 3
−30 |S21 |2
Ex |C |2 = . (1)
Hy 1 − |S11 |2 1 − |S22 |2
−35
kz
Port 2 Using the measured S-parameters of every two-port
−40
8 9 10 11 12 13 14 15 combination, the input impedance matching (Sii i = 1, 2),
Frequency (GHz) H-plane coupling (CH ), E-plane coupling (CE ), and cross-
coupling (Cc ) (coupling between the top-left and bottom-
RLRF RHRF right dipoles) were derived. The results, when the dipoles
TLRF THRF
are radiating in free space (without superstrate) and in the
presence of the superstrate are presented in Figure 4. The first
Figure 2: Transmission and reflection coefficients of the LRF and
column corresponds to the case of dipoles radiating in free
HRF manufactured in FR4 under normal incident plane wave
excitation. Inset: unit cell excitation. space and the second one in the presence of the metasurface.
The first row represents the coupling coefficients between
HRF dipoles, the second one between LRF dipoles, and the
third one cross-frequency coupling between HRF and LRF
ones. Due to the geometry of the array, the E, H, and cross-
dipoles operating at a certain frequency. Two different coupling for the HRF were measured. However, only the H-
approaches were used: S-parameters measurements, which plane coupling can be measured for the LRF dipoles.
allow evaluating the coupling coefficients, and near-field When the fed dipole is radiating in free space, the H-
measurements of the surface illumination. The radiation plane coupling with its counterpart dipoles, that is, the ones
coming from the balun was also investigated and it was operating at the same frequency as in Figures 4(a) and
concluded that in the presence of the superstrate the 4(c), is the most significant, with coupling values around
configuration operates better when the antennas are flipped CH = −20 dB at the resonant frequency. The cross and E-
upside down, so that the baluns are on the bottom face and plane couplings are around −35 dB. The input impedance
their ground plane is facing upwards. matching of the dipoles is S11 = −13 dB at 12.15 GHz for
A similar analysis to the one carried out for the 2 × 2 the HRF and S22 = −18 dB at 10.15 GHz for the LRF.
dipole antenna array has been performed for an MF array. When the metasurface is placed on top of the array
The array is formed by 6 dipoles (see Figure 3(a)) fed by a (Figures 4(b) and 4(d)), a frequency shift is observed due to
microstrip-to-coplanar stripline (CPS) balun to match the the loading effect of the metasurface. The input impedance
SMA connector to the CPS [10]. The dipoles in the middle matching is S11 =−40 dB at 13 GHz for the HRF dipoles and
resonate at the LRF and the 4 dipoles at the corners resonate S22 = −17 dB at 10.80 GHz for the LRF ones. In the case of
at the HRF. Afterwards, the nonuniform metasurface was the HRF dipoles, a second minimum of the S11 is observed
placed on top of the array with the dipoles operating at around 10.5 GHz, which is due to a resonance in the feeding
the resonant frequency of the group of cells on top of balun. This resonance can be mitigated by surrounding the
them (see Figure 3(b)). As it was observed in the single- balun with absorbing material, avoiding the distortion in the
frequency array [9], the resonant frequency of the dipoles radiation pattern. As far as the coupling is concerned, the
shifts when the metasurface is placed on top of array. H-plane coupling decreases to −40 dB, which means 20 dB
Therefore, LRF and HRF will be used to distinguish between coupling reduction with respect to free space. The E and
the two type of dipoles, rather than specifying the resonant cross-coupling remain at a similar level: Cc = −45 dB and
frequency, since it depends on the presence or absence of CE = −35 dB.
the superstrate. The dielectric material used as substrate was Although the dipole antenna was optimized to have good
FR4. Pictures of the manufactured multifrequency antenna matching, coupling reduction, and gain enhancement in the
array (MFAA), when the dipoles are radiating in free space presence of the superstrate, an input impedance matching
and with the superstrate, are shown in Figures 3(a) and better than −13 dB was obtained in both cases, with and
3(b), respectively. The dipole antennas were designed for without the superstrate.
good input impedance matching and gain enhancement In the case of the cross-frequency coupling, that is,
around the transmission window in the presence of the coupling to the LRF dipoles when the HRF one is fed
metasurface, that is, taking into account the loading effect and vice versa (Figures 4(e) and 4(f)), the E-plane and
of the superstrate. cross-coupling values were calculated. When the dipoles are
4 International Journal of Antennas and Propagation
0.8λ0
0.7λ0
0.7λ0
(a) (b)
Figure 3: Multifrequency dipole antenna arrays. (a) Printed dipoles without superstrate. (b) Dipoles with superstrate. Distance between
antennas: E-plane distance = 0.7λ0 ; H-plane distance = 0.8λ0 at 10 GHz.
radiating in free space (Figure 4(e)), the E-plane and cross- λ0 /2π [11]. The step of the acquisition points Δx and Δ y was
coupling are around −25 dB. However, in the presence of fixed to 2 mm so that the sampling criteria [11, 12]
the superstrate (Figure 4(e)), the cross-frequency coupling
λ0
coefficients decrease to −40 dB, except for the E-coupling Δx , Δ y = (2)
at LRF that remains around −27 dB. These results are 2 1 + (λ0 /z0 )2
summarized in Table 1.
for measurement distances z0 smaller than λ0 was satisfied for
the HRF and LRF. Since the E field radiated by the antenna
3.2. Near-Field Measurements. The mutual coupling phe- is polarized along the metal strips, the probe was oriented
nomenon was also investigated by NF measurements of parallel to the strips and only the copolar polarization was
the array. One of the dipoles was fed and the others were taken into account.
matched with 50 Ω loads. By this measurement, the surface The results of the scanned field when the HRF and
illumination and the field that reaches the neighboring LRF dipoles were radiating are shown in Figure 5. When
dipoles were detected. The frequencies for the near-field the metasurface is placed atop the array (Figures 5(b) and
measurement were the resonant frequency of the configu- 5(d)), the power radiated by the driven dipole is confined
rations with and without superstrate, respectively. Although by the metasurface area that resonates at the same frequency
the operational frequency is different, the aim of this analysis and radiates in the boresight direction, reducing the endfire
is to compare two configurations operating at their optimum radiation and, therefore, the mutual coupling between
frequency, that is, at the resonance. Moreover, the variation elements. As expected from the S-parameters measurements,
with frequency of the coupling coefficients in air is very a clear reduction in H-plane coupling is observed. The
small, as shown in Figure 4. The vector network analyzer E-plane, cross-coupling, and cross-frequency coupling also
HP8510C provided the microwave signal to feed the antenna decrease, showing overall a good level of isolation between
and measured the field detected by the probe. A custom antennas.
LabView program coordinates the motion of the scan with
the data acquisition of the VNA. 3.3. Radiation Pattern. In order to check the effect of
A monopole antenna was used as a probe. It was the mutual coupling on the radiation performance of the
mounted over a 2D scanning system and placed λ0 /4 away antenna, the radiated far field was measured in the DTU-
from the antenna. The distance from the antenna to the ESA Antenna Test Facility with and without the metasurface
probe, z0 , was chosen so that the probe was close enough at the resonant frequency. Only one dipole was fed and
to measure the surface illumination but remains out of the the others were matched with 50 Ω loads. The normalized
reactive near field of the antenna, which is usually taken as measurements of the E- and H-plane cuts of the radiation
International Journal of Antennas and Propagation 5
HRF HRF
0 0
−10 −10
Magnitude (dB)
Magnitude (dB)
−20 −20
−30 −30
−40 −40
−50 −50
−60 −60
9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14
Frequency (GHz) Frequency (GHz)
S11 Cc f11 S11 Cc f11
CH f11 CE f11 CH f11 CE f11
(a) (b)
LRF LRF
0 0
−10 −10
Magnitude (dB)
Magnitude (dB)
−20 −20
−30 −30
−40 −40
−50 −50
−60 −60
9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14
Frequency (GHz) Frequency (GHz)
S22 S22
CH f22 CH f22
(c) (d)
Cross-frequency Cross-frequency
0 0
−10 −10
S-parameter (dB)
Magnitude (dB)
−20 −20
−30 −30
−40 −40
−50 −50
−60 −60
9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14
Frequency (GHz) Frequency (GHz)
S11 Cc f12 S11 Cc f12
S22 CE f12 S22 CE f12
(e) (f)
Figure 4: Coupling coefficients calculated from the S-parameters measurements: Left column without superstrate (free space) and right
column with superstrate. (a), (b) Coupling between dipoles operating at f 1 = HRF. (c), (d) Coupling between dipoles operating at f2 =
LRF. (e), (f) Coupling between dipoles operating at different frequencies f12 (cross-frequency coupling).
6 International Journal of Antennas and Propagation
40 40
−5 −5
30 30
20 20
−10 −10
10 10
y (mm)
y (mm)
0 −15 0 −15
−10 −10
−20 −20
−20 −20
−30 −30
−25 −25
−40 −40
40 40
−5 −5
30 30
20 20
−10 −10
10 10
y (mm)
y (mm)
0 −15 0 −15
−10 −10
−20 −20
−20 −20
−30 −30
−25 −25
−40 −40
Figure 5: Near-field measurements of the surface illumination when the dipoles resonating are radiating in free space(left column) and with
the superstrate (right column). (a), (b) Resonant frequency at fHRF . (c), (d) Resonant frequency at fLRF .
pattern in the absence and presence of the metasurface for the in the E-plane is confirmed with the radiation pattern; very
HRF and LRF dipoles are shown in Figure 6. The radiation is low radiation is observed in the endfire direction and the
along the boresight direction. radiation pattern is almost symmetrical in both of half
When the HRF or LRF dipoles are radiating in free space spaces.
at the resonant frequency (Figure 6(a)), due to the strong H- When the superstrate is placed on top of the array
plane coupling between the elements and the presence of the (Figure 6(b)), the power radiated by the antenna, is picked
ground plane in this plane, the radiation pattern is strongly up by the superstrate, confined over a larger area on top of
distorted (dashed line). Instead of a uniform circular shape, the antenna and reradiated producing an enhancement of the
several peaks and nulls appear. The low coupling measured boresight radiation and a reduction of the back radiation.
International Journal of Antennas and Propagation 7
0 0
330 30 330 30
300 60 300 60
270 90 270 90
−20 −15 −10 −5 0 −20 −15 −10 −5 0
300 60 300 60
270 90 270 90
−20 −15 −10 −5 0 −20 −15 −10 −5 0
Figure 6: Normalized E- and H-plane cuts of the radiation pattern. (a) HRF dipole without superstrate at 12 GHz. (b) HRF dipole with
superstrate at 13.1 GHz. (c) LRF dipole without superstrate at 10.3 GHz. (d) LRF dipole with superstrate at 11 GHz.
Due to the power confinement and the H-plane coupling coupling was demonstrated by S-parameters and near-field
reduction, the H-plane endfire radiation is strongly reduced. measurements of the radiated field. In the case of the H-
Notice that although the E-plane coupling increases in the plane coupling, 20 dB coupling reduction was observed due
presence of the superstrate, the E-plane cut of the radiation to the presence of the superstrate. Although the E-plane
pattern is not distorted in the boresight direction, but only coupling increases with respect to the free-space case due to
an increase of the E-plane endfire radiation is observed. the guiding of the fields along the strips, it remains below
−20 dB.
respective neighbors, which is required for imaging array [11] A. Yaghjian, “An overview of near-field antenna measure-
applications. Apart from the MF capability, polarization ments,” IEEE Transactions on Antennas and Propagation, vol.
distinction could be added to the array by rotating one of the 34, no. 1, pp. 30–45, 1986.
sets of dipoles (HRF or LRF) 90◦ . This would increase the [12] Y. T. Looand and S. W. Lee, in Antenna Handbook, J. Appel-
capacity and compactness of the array and its performance Hansen, Ed., Van Nostrand Reinhold, New York, NY, USA,
1998.
for imaging array applications.
Acknowledgments
This research has been financially supported by METAMOR-
PHOSE NoE funded by E. C. under contract no. NMP3-
CT-2004-50252 and Spanish Government under project
no. TEC2006-13248-C04-03/TCM. One of the authors (K.
Guven) acknowledges support of TUBITAK under the
Project no. 106E198. The authors would also like to thank
Professor O. Brenjerg and Dr. S. Pivnenko from DTU for the
assistance with the radiation pattern measurements.
References
[1] T.-H. Chu and N. H. Farhat, “Frequency-swept microwave
imaging of dielectric objects,” IEEE Transactions on Microwave
Theory and Techniques, vol. 36, no. 3, pp. 489–493, 1988.
[2] V. Semenchik, V. Pahomov, and S. Kurilo, “Multifrequency
microwave imaging,” in Proceedings of the 5th International
Kharkov Symposium on Physics and Engineering of Microwaves,
Millimeter, and Submillimeter Waves (MSMW ’04), vol. 1, pp.
193–195, Kharkiv, Ukraine, June 2004.
[3] R. Gonzalo, P. de Maagt, and M. Sorolla, “Enhanced
patch-antenna performance by suppressing surface waves
using photonic-bandgap substrates,” IEEE Transactions on
Microwave Theory and Techniques, vol. 47, no. 11, pp. 2131–
2138, 1999.
[4] I. Ederra, L. Azcona, B. E. J. Alderman, et al., “A 250 GHz
subharmonic mixer design using EBG technology,” IEEE
Transactions on Antennas and Propagation, vol. 55, no. 11, part
1, pp. 2974–2982, 2007.
[5] E. Saenz, R. Gonzalo, I. Ederra, J. C. Vardaxoglou, and P.
de Maagt, “Resonant meta-surface superstrate for single and
multifrequency dipole antenna arrays,” IEEE Transactions on
Antennas and Propagation, vol. 56, no. 4, pp. 951–960, 2008.
[6] P. M. T. Ikonen, E. Saenz, R. Gonzalo, and S. A. Tretyakov,
“Modeling and analysis of composite antenna superstrates
consisting on grids of loaded wires,” IEEE Transactions on
Antennas and Propagation, vol. 55, no. 10, pp. 2692–2700,
2007.
[7] E. Saenz, I. Ederra, P. De Maagt, and R. Gonzalo, “Highly
efficient dipole antenna with planar meta-surface,” Electronics
Letters, vol. 43, no. 16, pp. 850–851, 2007.
[8] E. Saenz, K. Guven, I. Ederra, E. Ozbay, P. De Maagt, and
R. Gonzalo, “Near-field measurement of a planar meta-
surface illuminated by dipole antennas,” in Proceedings of the
Loughborough Antennas and Propagation Conference (LAPC
’08), pp. 221–224, Loughborough, UK, March 2008.
[9] E. Saenz, I. Ederra, R. Gonzalo, S. Pivnenko, O. Breinbjerg,
and P. de Maagt, “Coupling reduction between dipole antenna
elements by using a planar meta-surface,” IEEE Transactions on
Antennas and Propagation, vol. 57, no. 2, pp. 383–394, 2009.
[10] Y. Qian, W. R. Deal, N. Kaneda, and T. Itoh, “Microstrip-fed
quasi-Yagi antenna with broadband characteristics,” Electron-
ics Letters, vol. 34, no. 23, pp. 2194–2196, 1998.
Hindawi Publishing Corporation
International Journal of Antennas and Propagation
Volume 2010, Article ID 596291, 18 pages
doi:10.1155/2010/596291
Research Article
Compact Printed Arrays with Embedded Coupling Mitigation for
Energy-Efficient Wireless Sensor Networking
Copyright © 2010 C. G. Kakoyiannis and P. Constantinou. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
Wireless sensors emerged as narrowband, resource-constrained devices to provide monitoring services over a wide life span. Future
applications of sensor networks are multimedia-driven and include sensor mobility. Thus, sensors must combine small size, large
bandwidth, and diversity capabilities. Compact arrays, offering transmit/receive diversity, suffer from strong mutual coupling
(MC), which causes lower antenna efficiency, loss of bandwidth, and signal correlation. An efficient technique to reduce coupling
in compact arrays is described herein: a defect was inserted in the ground plane (GNDP) area between each pair of elements. The
defect disturbed the GNDP currents and offered multidecibel coupling suppression, bandwidth recovery, and reduction of in-band
correlation. Minimal pattern distortion was estimated. Computational results were supported by measurements. The bandwidth
of unloaded arrays degraded gracefully from 38% to 28% with decreasing interelement distance (0.25λ to 0.10λ). Defect-loaded
arrays exhibited active impedance bandwidths 37–45%, respectively. Measured coupling was reduced by 15–20 dB.
Antennas are the most neglected circuits-and-systems robustness (diversity gain) and not through capacity
aspect of sensor nodes and networks. This has brought maximization.
about difficulties in integrating antennas into microsensors, (iv) Spatial multiplexing techniques work well only above
and it was suggested that either frequencies beyond 10 GHz a threshold SNR value at the receiver, for example,
or inefficient chip antennas are used [2]. Despite the lack SNRmin = 20 dB. Such high SNR values are not
of studies on WSN-targeted antennas, wireless sensors are guaranteed in severe NLOS conditions. Below the
cross-layer entities and thus optimization of algorithmic and given threshold, only diversity gain can be sought.
networking aspects has taken place under the assumption
that the antenna system possesses certain properties. For
example, energy-efficient medium access solutions have 1.2. Mutual Coupling Reduction in Compact Arrays. In what
been theorized under the assumption of antenna directivity, follows, antenna spacing refers to the distance of the
either switched-beam or beamformed [7, 8], implemented feed points, not edge-to-edge. Important dimensions are
as uniform linear or circular arrays. The same holds for expressed in terms of free-space wavelength λ.
routing algorithms [9, 10]. Information-theoretic studies Yang and Rahmat-Samii [17] presented an early study
have utilized elements of antenna systems to extend the of mutual coupling suppression based on electromagnetic
connectivity and lifetime of WSNs [11]. Ultra-wideband band-gap (EBG) structures. Two patch antennas were
implementations of WSNs for localization at millimeter- printed on a thick substrate and spaced at d = 0.75λ. A 14 × 4
wave frequencies have been proposed again under the matrix of mushroom-like patches was printed between the
assumption of antenna directivity [12]. Still, little has been antennas. The technique produced a coupling reduction of
done to address the challenges and complexities of antenna 8.8 dB and occupied an area 1.07λ × 0.30λ.
implementation for WSNs, and especially arrays; two recent Dossche et al. [18] suggested the use of a combination
efforts are described in [13, 14]. By WSN standards, both of matching and decoupling network to suppress the cor-
studies resulted in bulky prototypes, attesting the fact that relation coefficient. The achieved resonance was very deep
a lot of ground needs to be covered in this area. but also narrowband (1.6%), since the circuit was based on
To the best of the authors’ knowledge, the work by transmission line sections. The required PCB area measured
Abbosh and Thiel [15] is the only existing study on the 1.43λ × 1.34λ.
effect of mutual coupling on wireless sensors. The authors A different type of ground defect was studied and
developed a theoretical model that relates the energy expen- presented very well in [19]. The authors built two planar
diture per transmitted bit to the presence of mutual coupling inverted-F antennas (IFAs) on a square PCB and spaced them
between the elements of uniform circular arrays in 2 × 2 and at d = 0.12λ. Two arrays of slits where then etched away on
4 × 4 MIMO systems. Mutual coupling was incorporated in opposite sides on the GND plane, between the two antennas.
circuit terms into the channel matrix and lead to a modified The disturbance caused to the ground currents offered a
MIMO channel matrix. Line-of-sight (LOS) propagation coupling reduction in the range 8–16 dB for various antenna
conditions were assumed, and the study opted for capacity scenarios. The slits occupied a PCB area 0.33λ × 0.07λ.
maximization. To this end, full knowledge of channel state Kokkinos et al. [20] built two planar IFAs on a long
information (CSI) was assumed at the transmitter, and PCB and spaced them at d = 0.62λ. Two slits where then
waterfilling adjusted properly the power transmitted from etched away on opposite sides on the GND plane, between
each antenna element. The authors concluded that mutual the two antennas. The authors optimized the dimensions and
coupling is beneficial for such a configuration, since it leads spacing of the slits, to make them resonate like a magnetic
to a decorrelated modified channel matrix and decreases loop. In the band of interest, coupling reduction ranged
the required energy per bit. The work presented herein between 10 and 40 dB. The slits occupied an area 0.25λ ×
diverges significantly from [15] in terms of fundamental 0.14λ. A single slit was also applied in [21] to decouple two
assumptions. planar IFAs. The difference was that the slit was not etched
between the elements, but further down the GND plane. The
(i) The complexity introduced by the multiple receive size of the slit was 0.18λ × 0.12λ.
and transmit RF chains is unrealistic, at least by The authors in [22] also studied the application of
current CMOS standards and implementation costs. electromagnetic band-gap (EBG) structures. Two horizontal
(ii) CSI availability at the transmitter introduces over- dipole antennas were mounted on a substrate and spaced at
head (feedback) and severe computational com- d = 0.17λ. A 9 × 3 matrix of mushroom-like patches was
plexity, because of the need for real-time matrix printed between the antennas. The technique produced an
inversions [16]. Full CSI knowledge is a common 8-dB coupling reduction and occupied an area 0.56λ × 0.11λ.
assumption made in algorithmic MIMO studies, but EBG was also the method of choice for Michailidis et al.
in every other aspect, it is very difficult to be hard- [23]. Two patch antennas were printed on a substrate and
coded at the FPGA or DSP level. spaced at d = 0.67λ. A 9×3 matrix of mushroom-like patches
(iii) In both indoor and outdoor WSN implementations, was printed between the antennas. The periodic structure
LOS propagation conditions are not given; rather, occupied an area 0.60λ × 0.20λ.
NLOS conditions should be expected. Consider, for
example, deployment in office floors and in forests. 1.3. Scope and Outline of the Paper. This work focuses
Thus, energy efficiency will be obtained through link on the straightforward and low-cost implementation of
International Journal of Antennas and Propagation 3
Figure 5: The top layer of three fabricated compact arrays. From left
to right, the interelement distance is 30 mm (0.25λ), 18 mm (0.15λ),
and 12 mm (0.10λ) respectively.
−25
2.3. Array Parametric Study and Numerical Results. Interele-
ment distance varied from 0.25λ (= 30 mm) down to 0.10λ
(= 12 mm). Under rich scattering conditions, diversity gain
−30
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 starts to deteriorate when IED drops below 0.2λ [24], hence
Frequency (GHz) the choice of the range of d. Figure 6 depicts the broadband
input matching. This is actually the active input impedance
i=2 i=4 of the array, since it corresponds to one port being excited
i=3 i=5
while the other is match terminated at 50 Ω [32]. By
Figure 4: Measured reflection coefficient and impedance band- comparison to the single-element simulation results in [25–
width of the four sinusoidal iterations shown in Figure 3. The
28], it is obvious that the upper adjacent resonance has
inductive behavior was attested by Smith charts of S11 ( jω) during
measurements. vanished, while the first resonances shifted lower. This was
due to the wider GNDP, which does not benefit sinusoidal
antennas. However, large bandwidths were obtained in the
2.2. Modeling the Compact Array. The following sections 2-3 GHz range; they ranged from 0.5 GHz to 0.7 GHz. The
describe the design and performance of a two-element respective fractional bandwidths ranged from 20% to 28%.
printed array built with sinusoidal elements and having The resonances were in the range 2.32 < fres < 2.48 GHz,
variable interelement distance (IED), denoted by d [26]. whereas the center frequencies were in the range 2.39 < fc <
Two third-iteration sinusoidal monopoles (i = 3) were 2.51 GHz.
placed on a common PCB carrier; so their feed lines shared Moreover, the results in Figure 7 show the changes in
the same GNDP and the same substrate. Figure 5 depicts broadband mutual coupling given by the scattering param-
three such arrays that were fabricated in this context. The eter S21 ( jω). Worst-case coupling reached Smax
21 = −4.5 dB
third-order sinusoidal element measures 10 mm × 10 mm. for the smallest interelement distance. Such strong coupling
The optimal GNDP dimensions were calculated at 26 mm × makes the suitability of any antenna array questionable
18 mm. The width of the sinusoidal copper trace equals for diversity/MIMO systems. Only at λ/4 spacing was the
6 International Journal of Antennas and Propagation
0 0
−5
Input reflection coefficient S11 (dB)
−5
−15
−15
−20
−20
−25
−25
−30
−30 −35
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (GHz) Frequency (GHz)
d = 30 mm d = 18 mm d = 30 mm d = 18 mm
d = 27 mm d = 15 mm d = 27 mm d = 15 mm
d = 24 mm d = 12 mm d = 24 mm d = 12 mm
d = 21 mm d = 21 mm
Figure 6: Parametric study of the broadband active input Figure 7: Parametric study of the broadband mutual coupling for
impedance for different values of interelement distance (0.25λ to different values of interelement distance (0.25λ to 0.10λ).
0.10λ).
0.5
0.45
coupling reduced down to −10 dB. To test for diversity
Envelope correlation coefficient
0 0
−5
−5
−10
−10
−15
−15
−20
−20
−25
−25 −30
−30 −35
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (GHz) Frequency (GHz)
d = 30 mm d = 30 mm
d = 18 mm d = 18 mm
d = 12 mm d = 12 mm
Figure 9: Measured active reflection coefficient and impedance Figure 10: Measured mutual coupling between the elements of
bandwidth of the three 2-element arrays shown in Figure 5. the arrays shown in Figure 5. Coupling was expressed in terms of
|S21 ( jω)|.
Indeed this response was recorded in Figure 9 and Table 2: Active input impedance bandwidths of the three measured
Table 2. All three resonances occurred below 2.5 GHz, arrays.
contrary to what Figure 4 shows. At the largest IED, the
active impedance bandwidth correlated well with the BWV d (mm) Center frequency (GHz) BWV (GHz) FBWV (%)
of the nonarrayed element. Partial detuning occurred with 30 2.632 0.995 37.8
decreasing IED. The measured mutual coupling in Figure 10 18 2.537 0.712 28.1
showed peak coupling that was stronger by 2 dB compared to 12 2.452 0.680 27.7
numerically anticipated values.
3. A Simple Technique That Reduces Mutual cell can be seen in Figures 11, 12, and 13. The proposed
method is a decoupling technique aiming to reduce coupling,
Coupling and Envelope Correlation rather than completely cancel it; moreover, it does so in a
There exist certain scenarios where it would be desirable or broadband manner. Seen from a different perspective, the
even mandatory to reduce coupling between the elements of DGS decoupler attempts to maintain a given level of array
the array. These cases include the following. efficiency and correlation (and thus diversity gain), despite
the addition of more antennas into the WSN terminal.
(i) The estimation of ρe via the S-parameters might
prove to be inaccurate due to low antenna efficiency. 3.1. Defected Ground Structures. Defected ground planes are
Although not mentioned in the original work [33], an evolution of photonic bandgap (PBG) structures from
the expression derived by the authors is based on the the optical regime towards microwave frequencies [35].
power balance of impinging, coupled and radiated Contrary to PBG structures, they are usually nonperiodic
fields; therefore, it assumes total radiation efficiency [36]. The DGS unit cell can be modeled near the resonance
equal to one [24]. frequency as a parallel RLC circuit in series with and between
(ii) The estimation of ρe via the S-parameters might two segments of transmission line. Park [37, 38] suggested
prove to be inaccurate due to the presence of a slightly more complex equivalent circuit by adding shunt
scattering objects near the sensor node. resistances and capacitances at the two reference planes of the
(iii) The feed points of the elements may need to be defect. That work also included a way to extract the values of
brought even closer than λ/10. the circuit elements from the ABCD-parameters of the two-
port defect. The ABCD-parameters can be extracted from
Since both array elements are strongly coupled with the S-parameters of the structure, which are obtained after
their common GNDP, the significant current distribution simulation in E/M solvers or measurement of actual hard-
on the GNDP is the dominant coupling factor, compared ware. Karmakar et al. provided a quasistatic analysis of the
to feed separation distance and near-field coupling [24]. most frequently occurred defect, that is, the dumbbell DGS
In this context, a simple way to achieve MC reduction is [39]. Their analysis attempted to quantify the contribution
to use a defected ground structure (DGS). A single DGS from every part of the defect in the frequency response. Their
8 International Journal of Antennas and Propagation
66.1
25.7
21.7
18.2
15.2
12.6
10.4
(A/m)
8.51
6.87
5.45
4.23
3.18
2.28
1.5
0.834
0
Figure 11: The proposed prefractal DGS cell acting on the wave
launched down a microstrip line. The SCD was calculated at the Figure 13: A prefractal dumbbell DGS etched on the Bottom layer
resonance frequency, where the DGS is reflective and creats a of a microstrip circuit. The board measures 80 mm × 60 mm, while
standing wave pattern. Only copper parts are shown here; the the size of the defect is 19.2 mm × 8.2 mm.
dielectric substrate is hidden. The SCD plot shows concurrent
maxima, which are physically unrealizable due to phase shifting.
0
−5
Scattering parameters (dB)
−10
−15
−20
−25
−30
0.5 1 1.5 2 2.5 3 3.5 4 4.5
Frequency (GHz)
Figure 12: A standard dumbbell DGS etched on the Bottom layer of
a microstrip circuit. The board measures 80 mm × 60 mm, whereas S11
the size of the defect is 29.8 mm × 13.5 mm. The microstrip runs on S21
the Top layer between the two connectors. The DGS comprises two
Figure 14: Measured S-parameters of the DGS-loaded microstrip
square apertures connected with a thin slot (0.3 mm). Microstrip
line shown in Figure 13. The proposed DGS resonates at 2.6 GHz,
width equals the distance of the two apertures.
causing high reflection and little signal transmission around that
frequency.
findings confirmed earlier simulation studies in that the
length of the current path around the structure determines
the distributed inductance, whereas the capacitance is mainly and R/L/C denote total distributed resistance, inductance,
controlled by the width of the narrow slot that connects the and capacitance at resonance.
two larger etched areas. The proposed DGS cell is shown in the model of
DGSs have been studied extensively for the past decade, Figure 11, where the defect has been etched below a
and they have been successfully applied in the design of microstrip line (the substrate has been removed for clarity).
filters and amplifiers. Their versatility has also found use A picture of the actual circuit is shown in Figure 13. It is
in the antenna field [40], though not nearly as much as in based on the dumbbell DGS, but it was eventually given a
microstrip filters and amplifiers. The properties of the DGS prefractal shape; it is a first-order Sierpinski carpet. Its overall
as a resonator are given by: dimensions are 19 mm × 8 mm. The measured electrical
performance of the DGS is shown in Figure 14, where a
R wide stopband is formed around 2.6 GHz. In terms of signal
Q= = ω0 RC, (4)
ω0 L transmission, the defect acts mainly as an open-circuit,
1 which creates a standing-wave pattern across the line. It also
FBW3 dB = , (5) acts partially as an antenna (wavetrap). The resonance is
Q
not very deep; this is a low-Q structure. The quality factor
where ω0 is the resonance frequency, Q is the quality factor, was estimated from the 3-dB fractional BW at Q = 1.3.
FBW3 dB is the 3-dB fractional bandwidth of the stopband, Nevertheless, this is just a blessing in disguise: the 30-dB
International Journal of Antennas and Propagation 9
−10
−15
−20
−25
−30
Figure 15: Perspective view of the 2-element compact array with 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
embedded DGS. The FR4 substrate is transparent so that the defect Frequency (GHz)
on the GNDP is visible. In this snapshot, the interelement distance
d = 20 mm d = 14 mm
is 18 mm (0.15λ).
d = 18 mm d = 12 mm
d = 16 mm
rejection is more than enough for the reduction of mutual Figure 16: Parametric study of the broadband input matching of
coupling, whereas the low Q-factor of the resonator makes the DGS-loaded array for different values of interelement distance
the stopband wide enough to match the BW of the antenna. (0.17λ to 0.10λ).
This is actually the reason why the defect was shaped after
a prefractal: for a wider stopband, larger inductance and leads to a working antenna with a 1-GHz operational BW.
less capacitance are needed. A longer current path increases The interelement distance affects the input impedance, and
inductance significantly, thus enabling a wider connecting this can be handled by resizing the elements and rematching
slot. Fractal shapes are a well-known quasideterministic at the feed point, but this is hardly the dominant factor here:
method to increase wire length in a confined area and have the 2D surface current flowing across the GNDP is perturbed
been extensively applied to antennas [41]. by the defect and more paths that are resonant are created.
A common misconception in DGS-related literature is This is the reason behind the bandwidth expansion (or
that the frequency response relates to the area of the defect, bandwidth recovery) seen in Figure 16 compared to Figure 6.
for example, that resonance frequency is inversely propor- Nevertheless, the BW augment is just a pleasant side
tional to area. This notion is repeated even in recent surveys effect of the insertion of the defect. The design procedure
on the topic [42]. The studied DGS models indicate that this opted for another sort of impact. Indeed, the current
is not so. It is the perimeter of the defect, rather than its area, disturbance and the resonant behavior of the DGS offer a
that affects the frequency response. This is demonstrated multidecibel drop in mutual coupling measured in terms of
briefly through a simple example. The prefractal defect in S21 ( jω) in Figure 17. For the 0.17λ distance the drop is 10 dB,
Figure 11 was transformed into a dumbbell-shaped defect while for the 0.10λ distance more than 15 dB of coupling
that resonated at the same frequency (2.5 GHz). All other reduction were gained.
parameters of the model remained the same as before. The This reduction was also studied in terms of envelope
area of the new defect was 24% larger, moving from 136 to correlation coefficient, which is an established metric for
169 mm2 . At the same time, the perimeter changed only by multielement antennas. The results in Figure 18 depict the
4%, increasing from 50 to 52 mm. broadband envelope correlation caused by the DGS when
the two elements are spaced λ/10 apart. The DGS that was
3.2. DGS-Loaded Compact Antenna Array. The concept applied is shown in Figure 19. The width of the long slot
behind the proposed technique is to insert a properly (nominally set at 0.6 mm) is one of the crucial dimensions
resonating DGS cell between the two elements of the of the defect, because it controls the distributed capacitance.
array and estimate the reduction in mutual coupling. As Three different values of slot width were used. The study
shown in Figure 15, the prefractal DGS is inserted (etched shows that as the slot widens, the defect becomes more
away) perpendicular to the flow of current between the “wideband”; that is, its stopband widens. Eventually, when
two elements. It attempts to block the ground currents the 0.6-mm wide slot was used, the envelope correlation
that contaminate the signal of the other antenna (signal was practically eliminated inside the operating BWV of the
bleeding). The PCB area occupied by the DGS decoupler is antenna element. In this way, the DGS bandgap is translated
17.3 mm × 7.3 mm, or 0.14λ × 0.06λ. The DGS cell is placed into a mutual coupling bandgap and consequently into a
1 mm away from the GNDP edge, where the return currents correlation bandgap for the compact array.
are stronger and increased self-inductance occurs. Getting into the details of the mechanism behind cou-
The frequency response of the defected array in Figure 16 pling suppression, Figure 20 shows three snapshots of the
shows that the disturbance of the current caused by the DGS current distribution on the ground plane. The snapshots
10 International Journal of Antennas and Propagation
−5
17.3
Mutual coupling S21 (dB)
−10
−15
7.3
0.6
−20
2.7 2.3
−25 2.5
−30
−35
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Figure 19: A closer look at the proposed PBG structure. The FR4
Frequency (GHz) substrate is transparent so that the defect on the GNDP is visible. All
dimensions are in millimeters. The overall PCB real estate occupied
d = 20 mm d = 14 mm by the defect is 17.3 mm × 7.3 mm.
d = 18 mm d = 12 mm
d = 16 mm
Figure 17: Parametric study of the broadband mutual coupling of Rahmat-Samii [17], it is instructive to compare the DGS
the DGS-loaded array for different values of interelement distance
decoupling technique with other simple and low-cost generic
(0.17λ to 0.10λ).
techniques, such as,
0.6
(i) the removal of substrate between antennas,
0.5 (ii) cavity-backed printed antennas.
Envelope correlation coefficient
118
91.9
77.4
65
54.3
45.1
37.2
(A/m)
30.4
24.5
19.5
15.1
11.4
8.14
5.36
2.97
0
2.1 GHz (outside bandgap) 2.7 GHz (inside bandgap) 3.1 GHz (outside bandgap)
Figure 20: Port 1 is excited (left), while Port 2 is terminated at 50 Ω (right). The surface current distribution on the conductive parts
demonstrates the suppression of mutual coupling at 2.7 GHz.
−5
Mutual coupling S21 (dB)
−10
−15
−20
−25
−30
−35
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 21: A two-element array that uses cavity backing instead of Frequency (GHz)
ground defect for coupling reduction.
Remove MSub
Cavity-backed
DGS loaded
a substrate with lower dielectric constant than the one Figure 22: Comparison of mutual coupling reduction using
assumed in simulations is obvious here: the defect resonated different decoupling techniques.
at 3 GHz instead of 2.6 GHz. However, by comparison with
the simulated data in Figure 16, it turns out that the substrate
practically had no adverse effect on the input impedance active impedance bandwidth. A secondary advantage is
results shown in Figure 24. The reason is that Zin ( jω) is the adjustable depth of resonance, which is controlled by
mostly influenced by the element-GNDP coupling, that is, balancing distributed inductance and capacitance. Regarding
by the shapes, sizes, and SCDs of these two objects. the occupied area, the results listed in Table 4 attest that the
Finally, the envelope correlation coefficient shown in DGS decoupler requires much less real estate compared to
Figure 26, which was calculated from the measured complex previously proposed decouplers; in fact, the difference is such
S-parameters of the array with IED d = 18 mm, remained that the DGS method reduces the smallest area required by
below ρe = 0.2 for all frequencies. This is even more pro- other methods by another 60% [19, 21].
nounced in Figure 27, where for λ/10 spacing the coefficient Regarding the two bandwidths, a few comments are
stays below ρe = 0.3. The formation of a correlation bandgap in order. First, decoupling methods are applied at either
in the 2-3 GHz range is apparent. the circuit level or the antenna level. Circuit-level decou-
By comparison with the results reported in recent litera- pling, which tries to solve the multiport-conjugate-matching
ture [17–24], it is concluded that the three most important problem [24], is well outside the scope of this paper. An
merits of the DGS decoupler are the small occupied PCB example circuit decoupler is the transmission line network
area, the width of the stopband (bandgap), and the resulting presented in [18]. Antenna-level decoupling can be done
12 International Journal of Antennas and Propagation
−5
−15
−20
−25
−30
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 23: The bottom layer of three fabricated DGS-loaded Frequency (GHz)
compact arrays. From left to right, the interelement distance is d = 30 mm
30 mm (0.25λ), 18 mm (0.15λ), and 12 mm (0.10λ), respectively. d = 18 mm
d = 12 mm
Table 3: Active input impedance bandwidths of the three measured Figure 24: Measured active reflection coefficient and impedance
DGS-loaded arrays. bandwidth of the three 2-element arrays shown in Figure 23.
d (mm) Center frequency (GHz) BWV (GHz) FBWV (%)
30 2.575 0.952 37.0 0
18 2.568 1.027 40.0
12 2.573 1.145 44.5 −5
Transmission parameter S21 (dB)
−10
Table 4: Comparison of occupied PCB real estate between pro-
posed decoupling methods.
−15
Ref. # Size Area (λ2 )
−20
[17] 1.07λ × 0.30λ 0.3210
[18] 1.43λ × 1.34λ 1.9162 −25
[19] 0.33λ × 0.07λ 0.0231
[20] 0.25λ × 0.14λ 0.0350 −30
[21] 0.18λ × 0.12λ 0.0216
[22] 0.56λ × 0.11λ 0.0616 −35
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
[23] 0.60λ × 0.20λ 0.1200 Frequency (GHz)
This work 0.14λ × 0.06λ 0.0084
d = 30 mm
d = 18 mm
d = 12 mm
in three ways: modified ground plane, neutralization line,
and parasitic scatterer [24]. All three techniques require that Figure 25: Measured mutual coupling between the elements of
the structure of the array be modified. With the exception the arrays shown in Figure 23. Coupling was expressed in terms of
of [18], all other methods listed in Table 4 are modified- |S21 ( jω)|.
ground-plane decouplers. Modification of the GNDP has the
obvious drawback that all subcircuits share ground and thus
significant changes may not be feasible. As it was pointed out 4. Radiation Properties of the Two-Element
in [24], antenna-level decoupling methods generally suffer Compact Array
from two side effects: (a) tens of decibels of MC suppression
are achievable, but only for a small fractional bandwidth, and This section investigates the effect that both the problems
(b) as the array elements draw closer, the active impedance discussed earlier and their solutions had on the radiation of
bandwidth of each element diminishes. By looking at the the sinusoidal elements of the array. This was carried out
reflection and coupling results in Figures 16, 17, 24, and 25, by calculating the degradation of total efficiency due to the
it occurs that the DGS decoupler does not face the above coupling, and by examination of the far-field patterns. To
restrictions; hence, it provides for a promising antenna-level facilitate the evaluation of these results, the far-field of a
decoupling technique, once it is perfected. single third-iteration monopole is quickly examined first.
International Journal of Antennas and Propagation 13
0.4 2.61
z
θ 2.14
1.82
Envelope correlation coefficient
0.3 1.51
1.19
0.871
0.554
0.2 0.238
(dB)
0
−4.15
−7.47
0.1
x −10.8
φ y −14.1
−17.4
0 −20.7
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 −24.1
Frequency (GHz) −27.4
0.3
scattering environment, where incoming waves arrive from
all directions in space; that is, they have a large angular spread
[25, 29–31]. In this case, terminals need to be able to receive
efficiently from all directions. The 3D pattern shows that
0.2 reception problems occur only along the x-axis.
Starting with the initial (unloaded) array, the results in
Figure 29 reveal the gradual degradation of total efficiency
0.1 as the elements were packed closer. The data were extracted
by feeding the input port of one of the elements while
terminating the other port at 50 Ω. In this configuration (6)
0
estimates the total efficiency:
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (GHz) ntotal = nrad 1 − |S11 |2 − |S21 |2 . (6)
DGS-loaded ρe When the elements were spaced at dmax = λ/4, there
Unloaded ρe was moderate mutual coupling, and thus total efficiency
Figure 27: Broadband envelope correlation coefficient calculated stayed above 70%. When the spacing dropped to dmin = λ/8,
from the measured complex S-matrix of the arrays with spacing d = efficiency dropped below 60% and even as low as 45%. The
12 mm = 0.10λ. loss in antenna efficiency is particularly problematic for the
energy efficiency of a sensor node. Figure 30 depicts one of
the active element patterns of the array [32]. It is evident
The numerical 3D gain pattern (IEEE) shown in that the directions of maximum gain have been rotated due
Figure 28 has been calculated at resonance and indicates that to the presence of the parasitic second element and the
the printed structure radiates in dipole mode. It exhibits wider GNDP. Reception from all directions in space was
the well-known toroidal pattern, which, in this case, shows maintained. The elements are too close (d = 15 mm = λ/8)
stronger radiation towards the backside of the PCB, because to provide increased directivity in any direction.
of the presence of the ground plane. The radiation pattern By inserting the resonating defect between the two
actually corresponds to a thick asymmetric dipole. The elements, the situation with total efficiency improved signifi-
achievable gain was estimated at Gmax = 2.6 dBi, reaching cantly. This is evidenced by the efficiency curves in Figure 31.
up to 0.5 dB higher than the λ/2-dipole. Part of this excess At every distance, and within a certain band, efficiency was
gain comes from the strong currents on the GNDP; the largely recovered. The DGS cell provided ntotal > 60% even
SCD occupies a larger volume inside the circumscribing for λ/10 spacing. The defect can be tailored to the designer’s
sphere compared to a thin wire dipole. The rest is due to the needs. A change in dimensions would alter the resonance
14 International Journal of Antennas and Propagation
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
2 2.2 2.4 2.6 2.8 3 2 2.2 2.4 2.6 2.8 3 3.2
Frequency (GHz) Frequency (GHz)
d = 15 mm d = 21 mm d = 12 mm d = 21 mm
d = 18 mm d = 30 mm d = 15 mm d = 30 mm
d = 18 mm
Figure 29: Variation in total efficiency with variable IED in the 2-
3 GHz range for the initial compact antenna array. Figure 31: Variation in total efficiency with variable IED in the 2-
3 GHz range for the DGS-loaded compact antenna array.
2.59
z 2.7
θ 2.28
1.98 z 2.21
1.67 θ 1.88
1.37
1.55
1.06
1.23
0.761
0.899
0.456
0.572
(dB)
0
0.245
−6.6 0
(dB)
−11
−5.65
−15.4
−10.2
x φ y −19.8
−24.2 −14.7
x y −19.2
−28.6 φ
−23.7
−33
−28.3
−37.4
−32.8
Figure 30: The transparent 3D active element pattern of the initial −37.3
array embedded around the sensor node. The pattern was calculated
at 2.5 GHz and corresponds to excitation of the right element. Figure 32: The transparent 3D active element pattern of the DGS-
loaded array embedded around the sensor node. The pattern was
calculated at 2.5 GHz and corresponds to excitation of the right
element.
and thus improve the efficiency in any specific desired band.
Another open problem is the possible distortion of the
radiation pattern in the far-field: the defect acts partly as a typical ad hoc networks. The goal of WSN system design is
wave trap and radiates. However, as shown in Figure 32, the to maximize the lifetime of the network, that is, to maintain
ground defect is an inefficient “slot antenna” and practically a reasonable level of connectivity for the largest possible
does not affect the pattern, which shows very little change duration. Wireless communication circuits tend to dominate
when compared with the case of no DGS loading. The the system energy budget [2]. The antenna is the spearhead
directions of maximum radiation rotate slightly towards the of the RF front-end; thus WSN-targeted antennas should be
θ = 0 axis, and maximum gain values increased by δGmax = designed with efficiency in mind.
+0.1 dB. Sinusoidal monopoles integrate well with portable
devices. Sinusoids function well with relatively small GND
5. Conclusions planes (∼ λ/4 × λ/6) and thus provide for building compact
printed arrays. Some detuning occurred, although not
Energy efficiency is an important topic in wireless sensor severe and easily compensated by following a few guidelines
networks and an important metric that sets them apart from (see Appendix A). The frequency response showed graceful
International Journal of Antennas and Propagation 15
subW
5 mm
hardware. Inaccurate assumptions regarding the properties z x
of available FR-4 substrates caused testing of devices on InductL W
stub
materials with lower permittivity at 2.5 GHz. Still, the
Wstrip
measured antennas displayed a propensity for large VSWR
bandwidths despite their small size. Even in their current
implementation, sinusoids offer the ability to trade excess Figure 33: The top layer of the fully parameterized sinusoidal
bandwidth for further size reduction. “system model”. The element is a “second iteration sinusoidal
Eventually CMOS-integrated circuits will advance to monopole” (i = 2). The substrate material is transparent so that
the point of enabling diversity (and perhaps even MIMO) the ground plane on the Bottom layer is visible. Element size is
2A × 2A mm2 .
applications on tiny sensor nodes. Antenna technology can
keep up with this trend so long as PCB real estate and size
are judiciously exploited. In any case, sensor radios face a
a half-period (Lstub = P/2), which is adequate but by no
challenging environment; so architectures with robustness
means optimal; this is a subject of future work. The design
to deep fading are required [2]. To this end, the insertion
is very repeatable, since it is based on an analytical curve.
of a single DGS cell in the ground plane of a generic 2-
The copper segments that produce the antenna element are
element array reduced mutual coupling inside the band of
formed by (x, y) points obeying the parametric description
interest and formed a correlation bandgap. The proposed
of (A.1), in which ξ is the independent parameter:
technique is cost-effective since it is compatible with PCB
printing processes and offers multidecibel gains in coupling x = ξ,
and correlation reduction. Finally, it was shown that the
2π P subL (A.1)
defect is an inefficient slot antenna, and so the distortion of y = −A sin ξ− − .
the radiation pattern was avoided. P 2 2
Taking a number of well-performing single antenna Numerical studies revealed that the input impedance
elements and displacing them as far as possible will not Zin ( jω) of the sinusoidal antenna is capacitive, as is typical
suffice for future compact multi-element antennas. A more of meander-line structures. This effect was exploited by
rigorous and holistic approach is needed to achieve good narrowing the last segment of the microstrip, which served
overall array characteristics. This is a very challenging task as an inductor in series with Zin ( jω); it is the simplest
when the combination of elements and GNDP becomes matching network. The length of this “lumped element”
electrically small [24]. So, further work is needed on the is 18◦ at 2.5 GHz. The main point in the design guide of
defect itself: it has to be made electrically smaller to enable these antennas is that the series inductor can be combined
closer antenna packing. Dielectric loading with a material with the width of the element Wstub to shape the input
of higher dielectric constant must be avoided, because it reflection coefficient in the frequency domain and thus tailor
would compromise antenna efficiency. It would also be very the bandwidth to the designer’s needs.
useful to discover ways to make its stopband even more It is a trivial optimization task to show that minimum
wideband. To these ends, other prefractal and Euclidean “antenna spread” on the PCB is achieved when the outline,
topologies will be investigated. The compensation of array or “envelope”, of the element becomes square. Since sinusoids
element detuning is yet to be demonstrated; it was omitted are folded monopoles, their unfolded length is constrained
here for brevity. Finally, an extension of this technique to to be somewhat greater than λ/4. This constraint imposes
four-element compact arrays is planned, which would be an equivalent constraint on the sum of the two sides of
targeted at more complex systems such as WiFi routers. Four- the envelope, because of spatial uniformity. Turning the
element arrays are much more appealing to MIMO system envelope into a square produces the envelope with the
applications, because they can extract the full rank of a shortest diagonal, hence the smallest spread on the PCB. The
realistic environment that is rich in scatterers [43]. constraint in (A.2) was applied to the dimensions of the
sinusoidal antenna to enforce the square spatial envelope:
Appendices
P
2A = (i + 1). (A.2)
A. Printed Sinusoidal Antenna Design Guide 2
The procedure to obtain a working sinusoidal antenna of
Figure 33 depicts all significant parameters that control the
given iteration i is summarized in the following design guide.
design of the antenna. The sinusoidal curve is characterized
by its amplitude A and spatial period P. The initial straight (i) Starting from a reasonable value of Wstub (1-2 mm)
segment is crucial to the operation of the antenna: it drives and zero series inductance, the simulator calculates
the first horizontal segment away from the GNDP and the spatial amplitude A that produces a lower reso-
improves radiation efficiency. Its length was set equal to nance at the desired frequency f0 . This calculation
16 International Journal of Antennas and Propagation
can be either a parametric sweep or an optimiza- and magnetic flow, a magnetic wall was placed across
tion. The optimization is single-parameter/single- the plane of symmetry. This boundary condition cut the
goal; thus the optimizer should take a reasonable computational burden in half, because only half of the struc-
amount of time to converge. An initial value for the tured needed solving. The complexity of the models ranged
spatial amplitude would be A = λ/(10i). between 250, 000 and 650, 000 Yee cells. Complexity depends
(ii) The value of inductW is altered to optimize the upon the size of the GND plane and the level of detail
input reflection coefficient in the desired manner. exhibited by the antenna element. Narrow copper traces and
For example, the input capacitance can be cancelled narrow object spacing contribute greatly to complexity.
up to the point where the Γin curve is distributed To reduce the computational complexity, a squared
evenly about the real axis of the Smith Chart near the equivalent connector replaced the realistic model of the
resonance. cylindrical end-launch SMA [25–28]. It is important to
model the feed structure as accurately as possible without
(iii) The value of Wstub is decreased to the point where
severely compromising simulation speed; the presence of the
the antenna cannot provided the desired impedance
SMA submodel is necessary to get an accurate simulated
bandwidth. If the trace is sufficiently wide, two dom-
surface current distribution. Thus, the SMA submodel helps
inant current paths exist, producing a dual-mode
predict the input impedance of the antenna with increased
antenna. Below a certain Wstub threshold, only one
accuracy.
mode is supported and bandwidth drops. Parameter
Wstub is a miniaturization technique embedded in
the antenna itself: decreasing Wstub leads to decreased References
amplitude A for a given resonance, and thus to
smaller element size, albeit with reduced bandwidth. [1] B. Warneke, M. Last, B. Liebowitz, and K. S. J. Pister, “Smart
dust: communicating with a cubic-millimeter computer,”
(iv) If the choice of Wstub leads to an electrically smaller Computer, pp. 44–51, 2001.
element (lower f0 ), amplitude A can be readjusted. [2] B. W. Cook, S. Lanzisera, and K. S. J. Pister, “SoC issues for RF
smart dust,” Proceedings of the IEEE, vol. 94, no. 6, pp. 1177–
B. Simulation Setup 1196, 2006.
[3] P. Basset, F. Alfaro, D. Novosel, A. De La Plaza, D. Stancil,
The antennas were designed and simulated in a reliable and G. K. Fedder, “Chip-size antennas for implantable sensors
Transient Solver [44], which exhibited good correlation and smart dust,” in Proceedings of the 13th International Con-
between simulated and measured results in prior studies ference on Solid-State Sensors and Actuators and Microsystems
(TRANSDUCERS ’05), pp. 457–460, Seoul, South Korea, June
[29–31]. The Solver is part of a full-wave electromagnetic
2005.
simulator that applies the Finite Integration Technique (FIT) [4] M. Haenggi and D. Puccinelli, “Routing in ad hoc networks: a
to reformulate Maxwell’s integral equations into the so-called case for long hops,” IEEE Communications Magazine, vol. 43,
“Maxwell Grid Equations”. In the time domain, by applying no. 10, pp. 93–101, 2005.
Yee’s spatial discretization and time-stepping scheme, FIT [5] N. Heo and P. K. Varshney, “Energy-efficient deployment
results in the same set of equations as FDTD. of intelligent mobile sensor networks,” IEEE Transactions on
A wideband Gaussian pulse excited the structures (DC- Systems, Man, and Cybernetics, Part A, vol. 35, no. 1, pp. 78–
5 GHz). A spatially adaptive hexahedral mesh discretized the 92, 2005.
objects. Finer meshing was applied inside the substrate to [6] S. Jain, R. C. Shah, W. Brunette, G. Borriello, and S. Roy,
capture the large gradients of the E-field; the same applies “Exploiting mobility for energy efficient data collection in
across the microstrip feed, on the sinusoidal element, and wireless sensor networks,” Mobile Networks and Applications,
vol. 11, no. 3, pp. 327–339, 2006.
inside the dielectric of the SMA connector. The simulator
[7] G. Manes, R. Fantacci, F. Chiti, et al., “Efficient MAC
stopped when the initial system energy decayed by 50 dB.
protocols for wireless sensor networks endowed with directive
This was a good trade-off between simulation speed and antennas: a cross-layer solution,” EURASIP Journal on Wireless
truncation error in the FFT engine that translates the results Communications and Networking, vol. 2007, Article ID 37910,
from the time- to the frequency-domain. The maximum cell 9 pages, 2007.
size at the maximum frequency (smallest wavelength) was [8] J. Dunlop and J. Cortes, “Co-design of efficient contention
set at λ5-GHz /25. The solvable space was terminated at several MAC with directional antennas in wireless sensor networks,”
Berenger PML layers. in Proceedings of the 13th International Wireless Communi-
cations and Mobile Computing Conference (IWCMC ’08), pp.
(i) During the initial design stages, four layers were 383–388, Crete Island, Greece, August 2008.
used to speed up the design cycle, with distance-to- [9] D. Li, Z. Li, and L. Liu, “Energy efficient broadcast routing
boundary equal to λ2.5-GHz /8. in ad hoc sensor networks with directional antennas,” in
(ii) For the final simulations, six layers were used, Proceedings of the Water and Sewerage Authority (WASA ’08),
with distance-to-boundary equal to λ2.5-GHz /4. By vol. 5258 of Lecture Notes in Computer Science, pp. 29–39,
increasing the distance to λ2.5-GHz /2 the results did 2008.
not improve further. [10] Z. Li and D. Li, “Minimum energy broadcast routing in ad hoc
and sensor networks with directional antennas,” in Proceedings
Whenever a model featured topological symmetry and of the Computations in Commutative Algebra, vol. 5573 of
satisfied the appropriate boundary conditions for electric Lecture Notes in Computer Science, pp. 507–518, 2009.
International Journal of Antennas and Propagation 17
[11] S. Shankar and D. Kundur, “Towards improved connectivity [25] C. G. Kakoyiannis, P. Gika, and P. Constantinou, “Small
with hybrid uni/omni-directional antennas in wireless sensor printed sinusoidal antennas: a simple design guide for smooth
networks,” in Proceedings of the IEEE INFOCOM Workshops, meander-line structures with augmented bandwidth,” in
Phoenix, Ariz, USA, April 2008. Proceedings of the IEEE International Workshop on Antenna
[12] A. Lecointre, D. Dragomirecu, P. Berthou, and R. Plana, Technology (iWAT ’09), Santa Monica, Calif, USA, March 2009.
“A reconfigurable IR-UWB radio interface, with directional [26] C. G. Kakoyiannis, P. Gika, and P. Constantinou, “Compact
antennas and localization capability, for wireless sensor multi-element antennas of sinusoidal printed monopoles
networks,” in Proceedings of the 14th IEEE International for sensors and portable devices,” in Proceedings of the 3rd
Conference on Electronics, Circuits and Systems (ICECS ’07), pp. European Conference on Antennas and Propagation (EuCAP
1055–1058, Marrakech, Morocco, December 2007. ’09), pp. 3423–3427, Berlin, Germany, March 2009.
[13] G. Giorgetti, A. Cidronali, S. K. S. Gupta, and G. Manes, [27] C. G. Kakoyiannis and P. Constantinou, “Coupling and
“Exploiting low-cost directional antennas in 2.4 GHz IEEE correlation reduction in compact arrays for WSN nodes via
802.15.4 wireless sensor networks,” in Proceedings of the 10th pre-fractal defected ground plane,” in Proceedings of the 3rd
European Conference on Wireless Technologies (ECWT ’07), pp. International Conference on Sensor Technologies and Applica-
217–220, Munich, Germany, October 2007. tions (SENSORCOMM ’09), pp. 175–180, Athens, Greece, June
[14] M. Cabedo, M. Gallo, E. Antonino, M. Ferrando, and M. 2009.
Bozzetti, “Modal analysis of a MIMO antenna for sensor [28] C. G. Kakoyiannis and P. Constantinou, “Reducing coupling in
networks,” in Proceedings of the IEEE Antenna and Propagation compact arrays for WSN nodes via pre-fractal defected ground
Society International Symposium (AP-S ’08), San Diego, Calif, structures,” in Proceedings of the 39th European Microwave
USA, July 2008. Conference (EuMC ’09), pp. 846–849, Rome, Italy, September-
[15] A. M. Abbosh and D. V. Thiel, “Effect of mutual coupling October 2009.
on the energy requirement of MIMO-based wireless sensor [29] C. G. Kakoyiannis, G. Stamatiou, and P. Constantinou, “Small
networks,” in Proceedings of the IEEE International Conference square meander-line antennas with reduced ground plane
on Signal Processing and Communications (ICSPC ’07), pp. size for multimedia WSN nodes,” in Proceedings of the 3rd
436–439, Dubai, UAE, November 2007. European Conference on Antennas and Propagation (EuCAP
[16] A. Marousis and P. Constantinou, “Performance of a low ’09), pp. 2411–2415, Berlin, Germany, March 2009.
complexity channel estimator for MC-CDMA systems over [30] C. G. Kakoyiannis and P. Constantinou, “Ground plane
realistic MIMO channels,” in Proceedings of the 6th Conference considerations for printed monopole antennas integrated into
Telecommunications, Peniche, Portugal, May 2007. wireless sensor devices,” in Proceedings of the International
Wireless Communications and Mobile Computing Conference
[17] F. Yang and Y. Rahmat-Samii, “Microstrip antennas inte-
(IWCMC ’08), pp. 738–743, Crete Island, Greece, August 2008.
grated with electromagnetic band-gap (EBG) structures: a
[31] C. G. Kakoyiannis and P. Constantinou, “Co-design of
low mutual coupling design for array applications,” IEEE
antenna element and ground plane for printed monopoles
Transactions on Antennas and Propagation, vol. 51, no. 10, pp.
embedded in wireless sensors,” in Proceedings of the 2nd
2936–2946, 2003.
International Conference on Sensor Technologies and Applica-
[18] S. Dossche, S. Blanch, and J. Romeu, “Optimum antenna
tions (SENSORCOMM ’08), pp. 413–418, Cap Esterel, France,
matching to minimise signal correlation on a two-port
August 2008.
antenna diversity system,” Electronics Letters, vol. 40, no. 19,
[32] D. M. Pozar, “A relation between the active input impedance
pp. 1164–1165, 2004.
and the active element pattern of a phased array,” IEEE
[19] C.-Y. Chiu, C.-H. Cheng, R. D. Murch, and C. R. Row- Transactions on Antennas and Propagation, vol. 51, no. 9, pp.
ell, “Reduction of mutual coupling between closely-packed 2486–2489, 2003.
antenna elements,” IEEE Transactions on Antennas and Prop- [33] S. Blanch, J. Romeu, and I. Corbella, “Exact representation of
agation, vol. 55, no. 6, pp. 1732–1738, 2007. antenna system diversity performance from input parameter
[20] T. Kokkinos, E. Liakou, and A. P. Feresidis, “Decoupling description,” Electronics Letters, vol. 39, no. 9, pp. 705–707,
antenna elements of PIFA arrays on handheld devices,” 2003.
Electronics Letters, vol. 44, no. 25, pp. 1442–1444, 2008. [34] J. Thaysen and K. B. Jakobsen, “Design considerations for low
[21] T. Ohishi, N. Oodachi, S. Sekine, and H. Shoki, “A method to antenna correlation and mutual coupling reduction in multi
improve the correlation coefficient and the mutual coupling antenna terminals,” European Transactions on Telecommunica-
for diversity antenna,” in Proceedings of the IEEE Antennas tions, vol. 18, no. 3, pp. 319–326, 2007.
and Propagation Society International Symposium, pp. 507– [35] C.-S. Kim, J.-S. Park, D. Ahn, and J.-B. Lim, “A novel 1-D
510, Washington, DC, USA, July 2005. periodic defected ground structure for planar circuits,” IEEE
[22] M. Schühler, R. Wansch, and M. A. Hein, “Reduced mutual Microwave and Guided Wave Letters, vol. 10, no. 4, pp. 131–
coupling in a compact antenna array using periodic struc- 133, 2000.
tures,” in Proceedings of the Loughborough Antennas and Propa- [36] D. Ahn, J.-S. Park, C.-S. Kim, J. Kim, Y. Qian, and T. Itoh,
gation Conference (LAPC ’08), pp. 93–96, Loughborough, UK, “A design of the low-pass filter using the novel microstrip
March 2008. defected ground structure,” IEEE Transactions on Microwave
[23] E. Michailidis, C. Tsimenidis, and G. Chester, “Mutual cou- Theory and Techniques, vol. 49, no. 1, pp. 86–93, 2001.
pling reduction in a linear two element patch array and its [37] J.-S. Park, “An equivalent circuit and modeling method for
effect on theoretical MIMO capacity,” in Proceedings of the defected ground structure and its application to the design of
Loughborough Antennas and Propagation Conference (LAPC microwave circuits,” Microwave Journal, vol. 46, no. 11, pp. 22–
’08), pp. 457–460, Loughborough, UK, March 2008. 44, 2003.
[24] B. K. Lau, “Multiple antenna terminals,” in MIMO: From [38] M.-S. Joung, J.-S. Park, and H.-S. Kim, “A novel modeling
Theory to Implementation, C. Oestges, A. Sibille, and A. method for defected ground structure using adaptive fre-
Zanella, Eds., Elsevier, New York, NY, USA, 2010. quency sampling and its application to microwave oscillator
18 International Journal of Antennas and Propagation
Research Article
Optimization of Training Signal Transmission for Estimating
MIMO Channel under Antenna Mutual Coupling Conditions
Copyright © 2010 X. Liu and M. E. Bialkowski. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
This paper reports investigations on the effect of antenna mutual coupling on performance of training-based Multiple-Input
Multiple-Output (MIMO) channel estimation. The influence of mutual coupling is assessed for two training-based channel
estimation methods, Scaled Least Square (SLS) and Minimum Mean Square Error (MMSE). It is shown that the accuracy of
MIMO channel estimation is governed by the sum of eigenvalues of channel correlation matrix which in turn is influenced by
the mutual coupling in transmitting and receiving array antennas. A water-filling-based procedure is proposed to optimize the
training signal transmission to minimize the MIMO channel estimation errors.
1. Introduction out that when the transmitted SPNR and the number of
antenna elements are fixed, the SLS and MMSE methods
In recent years, there has been a growing interest in offer better performance than the LS method. This can be
multiple-input multiple-output (MIMO) wireless commu- explained by the fact that the SLS and MMSE methods utilize
nication systems as they can significantly increase data the channel correlation in the estimator cost function while
throughput (capacity) without the need for extra operational the LS estimator does not take into account the channel
frequency bandwidth. In order to explore the advantages of properties.
MIMO technique, precise channel state information (CSI) It is worthwhile to note that the channel properties
is required at the receiver. The reason is that without CSI, are governed by a signal propagation environment, which
decoding of the received signal is impossible [1–5]. In in turn affects the spatial correlation (SC) observed at
turn, an inaccurate CSI leads to an increased bit error rate the input/output ports of the MIMO system. The spatial
(BER) that translates into a degraded capacity of the system correlation is dependent on an antenna configuration and
[6–8]. a distribution of scattering objects that are present in the
Obtaining accurate CSI can be accomplished using path between transmitter and receiver. In particular it is
suitable channel estimation methods. The methods based influenced by the finite antenna spacing in array antennas.
on the use of training sequences, known as the training- This finite antenna spacing is also responsible for mutual
based channel estimation methods, are the most popular. In coupling which adversely affects signal power transmission
[9, 10], several training-based methods including the least and reception. The mutual coupling effect is especially
square (LS) method, the scaled least square (SLS) method pronounced in tightly spaced arrays. Because there is a
and the minimum mean square error (MMSE) method have considerable demand for compact size Mobile Station (MS)
been investigated. It has been shown that the accuracy of terminals, the effect of mutual coupling cannot be neglected
these training-based estimation methods is influenced by and thus has to be taken into account while assessing the
the transmit signal power to noise ratio (SPNR) in the MIMO link performance. The problem of mutual coupling
training mode, and the number of antenna elements at the in MIMO systems for the case of peer-to-peer communica-
transmitter and receiver. In particular, it has been pointed tion has been addressed via simulations and measurements
2 International Journal of Antennas and Propagation
in [10–18]. It has been shown that the mutual coupling may over L time slots in the training mode is constrained by the
improve the channel capacity when the spacing of antenna following expression
elements in the array is between 0.2λ to 0.4λ (where λ is the
carrier wavelength) [15–18] . P 2F = P, (2)
In this paper, the focus is on the effect of mutual coupling
on MIMO channel estimation. The problem of mutual where P is a given constant representing the total power and
coupling and its effect on the MMSE method of channel · 2F stands for the Frobenius norm.
estimation have been investigated in [19, 20]. The presented The training matrices are assumed to be orthogonal [9,
results have been limited to simulations providing general 10] and the transmitted signal power to noise ratio (SPNR)
trends without any further mathematical insight. in the training mode is set to ρ equal to P/σn2 .
In this paper, we present a mathematical analysis explain- Using the LS method, the estimated channel matrix can
ing the mutual coupling effects on the SLS and MMSE be written as [19]
methods for channel estimation. It is shown that for a
fixed transmit SPNR and for a given number of transceiver HLS = Y P † , (3)
antenna elements the accuracy of SLS and MMSE methods is
determined by the sum of eigenvalues of channel correlation where {·}† stands for the pseudo-inverse operation and P † =
−1
matrix characterizing the signal propagation conditions. P H (PP H ) . The mean square error (MSE) of the estimated
The presence of mutual coupling in array antennas causes channel matrix in the LS method is given as
variations of the sum of these eigenvalues, which in turn 2
affects the accuracy of estimation. In order to reduce the
MSELS = E H − HLS
F
subject to P 2F = P. (4)
channel estimation errors, a water-filling based optimization
method is proposed to find an optimal training matrix
According to [9, 10], the optimal training sequence satisfying
minimizing the channel estimation errors when the MIMO
(4) is given as
system operates under the condition in which the mutual
coupling cannot be neglected. P
The rest of the paper is organized as follows. In Section 2, PP H = I. (5)
Mt
a MIMO system model is introduced. Also, LS, SLS, and
MMSE channel estimation methods are described followed The minimum value of MSE for the LS method is
by an analysis of channel estimation accuracy. Section 3
describes the method for modeling mutual coupling effects. Mt2 Mr
MSELS = . (6)
Section 4 presents the optimization of a training signal ρ
transmission when the MIMO system operates under mutual
coupling conditions. Section 5 describes computer simula- From (6) one can see that the optimal performance of the LS
tion results. Section 6 concludes the paper. estimator is influenced by the number of antenna elements at
the transmitter and the receiver. However, the channel matrix
has no effect on the minimum value of MSE.
2. System Description and Training-Based The SLS method further reduces the estimation error that
MIMO Channel Estimation is obtained in the LS method. The improvement is given by a
scaling factor γ which uses the channel estimator as given by
In this paper, a narrow band block fading MIMO channel
is assumed. The number of transmitting and receiving H SLS = γH LS . (7)
antennas is denoted as Mt and Mr , respectively. The channel
is described by the Mr ×Mt complex matrix H with entries hi j The resulting estimation error is given by (8)
representing the response between the ith receiving antenna 2
and the jth transmitting antenna.
MSESLS = E H − γH LS F subject to P 2F = P. (8)
2.1. Training-Based MIMO Channel Estimation. For the Expression (8) can be rewritten as (9)
training-based channel estimation method, the relationship
H
between the received signal and the training sequence is given MSESLS = E tr (H − γH LS ) H − γH LS
by
2
= E (1 − γ) tr{RH } + γ2 MSELS
Y = HP + V , (1)
2
tr{RH }
= E (MSELS + tr{RH }) γ −
MSELS + tr{RH }
where P represents the Mt × L complex training matrix and
L is the length of the training sequence. MSELS tr{RH }
+ ,
The goal is to estimate the complex channel matrix H MSELS + tr{RH }
from the knowledge of Y and P. The transmitted power (9)
International Journal of Antennas and Propagation 3
in which RH is the channel correlation matrix defined as Thus the optimal ψ can be represented by
RH = {H H H }.
−1
From (9), it is apparent that in order to minimize the ψ opt = (P H RH P + σn2 Mr I) P H RH . (19)
MSE of SLS, an optimal value of γ as given by (10) has to
be chosen The estimated channel matrix can be derived as,
tr{RH } −1
γ= . (10) H MMSE = Y ψ opt = Y (P H RH P + σn2 Mr I) P H RH . (20)
MSELS + tr{RH }
The minimized value of MSE can be written as [9, 10] By defining the channel estimation error as
MSELS tr{RH } e = H − H MMSE , (21)
MSESLS = E < MSELS . (11)
MSELS + tr{RH }
the MSE for the MMSE method can be expressed as
By substituting (6) into (11), the minimum value of MSE is
expressed as, MSEMMSE = E{tr{RE }}
1 = E tr E eeH
MSESLS = E . (12)
1/ tr{RH } + ρ/Mt2 Mr
−1
−1 1
= E tr (RH + 2 PP H )
The estimated channel matrix as given by SLS method can be σn M r
obtained using (7) and is given as −1
= E tr Λ−1 + σn−2 Mr−1 QH PP H Q .
H SLS = γH LS (22)
(13)
tr{RH }
= Y p P† . In (22), Q is a unitary eigenvector matrix of RH and Λ is
MSELS + tr{RH }
a diagonal matrix with eigenvalues of RH , which are given
In practice, RH can be obtained using the channel matrix through the eigenvalue decomposition of RH as
estimated by the LS method. Therefore
RH = H H H = QΛQH . (23)
tr RH = tr HLS
H
HLS . (14)
Here, to construct the channel matrix H, the Kronecker
channel model is postulated [22, 23]. In this model, the
In the MMSE method, the estimated channel matrix is given transmitter and receiver correlation matrices are assumed to
as [21], be separable and the channel matrix is represented as:
HMMSE = Y ψ. (15) H = R1/2 1/2
R GH RT , (24)
The objective is to find the optimal ψ to minimize the MSE. where GH is the matrix including identical independent
This task can be expressed as distributed (i.i.d) Gaussian entries with a zero mean and
2 a unit variance, and RR and RT are the spatial correlation
ψ opt = arg min H − H MMSE F matrices at the receiver and transmitter, respectively.
(16) The superiority of SLS and MMSE methods over LS
2
= arg min H − Y ψ F . method is due to the fact that SLS and MMSE methods
ψ utilize information of the channel correlation RH . Here, it
is assumed that the RH is perfectly obtained before channel
The mean-square error (MSE) for the MMSE method is estimation.
given as, In further considerations, it is assumed that the trans-
2 mitting and receiving array antennas are formed by vertically
MSEMMSE = E H − Y ψ F , polarized wire dipole antennas which are surrounded by
scattering objects. Assuming that these scattering objects are
= E tr{RH } − tr RH Pψ − tr ψ H P H RH (17) uniformly distributed within circles surrounding the array
antennas, the spatial correlation matrix elements can be
+ tr ψ H P H RH P + σn2 Mr I ψ . obtained using the Clark’s model as given by.
The optimal ψ minimizing MSE satisfies (18), ρi,R(T)
j = J0 κdi, j . (25)
the uniform array antenna. The correlation matrices RT and The expression (22) for the minimized value of MSE can
RR can be generated using (25) as be rewritten using the orthogonality properties of a training
⎡ ⎤ sequence P and the unitary matrix Q, as shown by
ρR(T) · · · ρ1,M R(T)
⎢ 1,1 r(t)
⎥ MSEMMSE
⎢ . .. .. ⎥
RR(T) = ⎢ .. . . ⎥. (26)
⎣ ⎦ −1
R(T) R(T)
ρMr(t) , j · · · ρMr(t) ,Mr(t) = E tr (Λ−1 + ρMr−1 I)
⎧ ⎧ ⎡ −1 ⎤⎫ ⎫
Having determined RT and RR , the channel matrix can be ⎪
⎪ ⎪
⎪ λ1−1 + ρMr−1 0 ··· 0 ⎪⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ ⎢ ⎥⎪⎪⎪⎪
⎪
⎨ ⎪
⎨⎢ −1 .. . ⎥⎪⎬⎪⎬
calculated using (24). ⎢ 0 λ2−1 + ρMr−1 . .
. ⎥
= E tr ⎢ ⎥
⎪
⎪ ⎪
⎪ ⎢ ⎥⎪ ⎪
⎪
⎪
⎪
⎪
⎪ ⎣ .
. .. .. ⎦⎪⎪
⎪
⎪
⎪
⎪
2.2. Estimation Accuracy Analysis. By taking into account ⎩ ⎪⎩ . . .
$
0
% −1
⎪
⎭ ⎪
⎭
0 0 ··· λn−1 + ρMr−1
where n = min(Mr , Mt ) and λi is the ith eigenvalue of the The expression (30) shows that similarly as in the SLS
channel correlation matrix RH . ρ is the transmit SPNR and method, a smaller sum of eigenvalues of the channel
equal to P/σn2 . P is the transmitted training sequence total correlation matrix RH leads to a smaller estimation error in
power. If power P and the number of transmit and receive the MMSE method. Similarly as in the SLS method, the MSE
antennas are fixed then the following relationship holds of MMSE is influenced by the number of transmitting and
⎧⎡ ⎤−1 ⎫ receiving antennas.
⎪ ⎛ ⎞−1 ⎪ ⎧ ⎫
⎪
⎨ ! n ⎪
⎬ ⎨!n ⎬ Through the above mathematical analysis, one can see
⎢⎝ ⎠ + ρ ⎥
MSESLS =E ⎣ λ ⎦ < E λ . (28) that if the total transmitted power P as well as the number
⎪
⎪
i 2
Mt Mr ⎪
⎪ ⎩ i ⎭
⎩ i ⎭ i of antenna elements at the transmitter and the receiver is
fixed, the accuracy of the training-based MIMO channel
It can be seen from (28) that MSESLS is smaller than the estimation is governed by the sum of eigenvalues of the
sum of the eigenvalues of the channel correlation matrix channel correlation matrix RH . The smaller is the sum, the
RH . Therefore, the sum of the eigenvalues of the channel more accurate channel estimation is achieved.
correlation matrix RH is the upper bound of MSE in the SLS Decreasing the sum of eigenvalues of channel correlation
method. is equivalent to decreasing the effective degree of freedom
As observed from expression (28), MSE decreases when (EDOF) of the MIMO channel [24]. This is caused by an
the sum of eignvalues of RH decreases. The same expression increased spatial correlation [25]. Hence, it is apparent that
shows that the MSE in the SLS method is influenced by an increased spatial correlation helps improving the channel
the number of antenna elements at the transmitter and estimation accuracy.
receiver. When the number of antenna elements on the
two sides of communication link drops to one, the system 3. Mutual Coupling Effects
becomes the conventional SISO system. In this case, the
channel estimation using a fixed-length training sequence Mutual coupling in an array of collinear side-by-side wire
becomes most accurate. This confirms the expectation that it dipoles can be modeled using the approach described in [26].
is easier to estimate the SISO channel which is characterized Assuming the array is formed by Mrt wire dipoles, the mutual
by a single transfer coefficient between single transmitting matrix can be calculated using the following relationship
and receiving antennas than the MIMO channel which is with the impedance matrix
described by a matrix of transfer coefficients between many
$ %−1
antenna elements. C = (ZA + ZT ) Z + ZT IMrt , (31)
The derived expression also shows that when the number
of transmit and receive antennas and the total transmitted where ZA is the element impedance in isolation, for example,
power P is fixed, MSE can be minimized by minimizing the when the wire dipole is λ/2, its value is ZA = 73 + j42.5[Ω];
sum of eigenvalues of RH . ZT is impedance of the receiver at each element chosen as
International Journal of Antennas and Propagation 5
eigenvalue
Z A + ZT Z12 ··· Z1Mrt
⎢ Z Z A + ZT ··· Z2Mrt ⎥ 18
⎢ 21 ⎥
Z=⎢
⎢ .. .. .. .. ⎥.
⎥ (32)
⎣ . . . . ⎦ 17
ZMrt 1 ZMrt 2 · · · Z A + ZT
16
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Note that this expression provides the circuit representation dt/λ
for the mutual coupling in array antennas. It is valid for
antennas operating in a single mode. Wire dipoles fall into (a) Sum of Channel Correlation Eigenvalues versus Transmitter Antenna
Spacing
this category.
eigenvalue
18
Zmn
⎧ 17
⎪
⎨30[0.5772+ln(2κl) − Ci (2κl)]+ j[30Si (2κl)],
⎪ m = n,
= 30[2Ci (u0 ) − Ci (u1 ) − Ci (u2 )] 16
⎪
⎪ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
⎩ − j[30(2S (u ) − S (u ) − S (u ))], m=
i 0 i 1 i 2 / n, dr/λ
(33)
Taking MC into account
where κ is the wave number equal to 2π/λ, Without taking MC into account
(b) Sum of Channel Correlation Eigenvalues versus Receiver Antenna
u0 = κdh , Spacing
&
u1 = κ dh 2
+ l2 +l , Figure 1: Mutual coupling effects on the sum of channel correlation
(34) matrix eigenvalues. (a) Sum of eigenvalues versus transmitter
& antenna element spacing; (b) Sum of eigenvalues versus receiver
u2 = κ dh 2 + l2 − l , antenna element spacing.
can see that when the antenna elements spacing is between 0.05
0.1λ to 0.5λ, there is an apparent difference between the sum 0.045
of channel correlation eigenvalues with and without taking 0.04
mutual coupling effects into account. The ones with mutual
MSE
0.035
coupling effects are higher than the ones without. When the 0.03
spacing is increased and exceeds 0.5λ, the difference becomes
0.025
smaller and the blue and red curves overlap. As a result, at the
antenna spacing between 0.1λ and 0.5λ, the mutual coupling 0.02
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
increases the sum of eigenvalues of the channel correlation dt/λ
matrix. This is because within this spacing range the mutual
(a) MMSE MSE versus Transmitter Antenna Spacing
coupling decreases the spatial correlation level. Therefore the
channel estimation accuracy is adversely affected. 0.774
0.773
0.772
4. MMSE Method Performance Optimization 0.771
MSE
Under Mutual Coupling Conditions 0.77
0.769
In order to optimize the performance of MMSE training- 0.768
based estimation method, we have to find the optimal 0.767
training matrix. This optimization problem can be expressed 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
as, dt/λ
−1
−1
MSEmc
MMSE = min tr (Λmc + σn−2 Mr−1 Qmc
H
PP H Qmc ) , MSE taking MC into account
2
P F =P MSE without taking MC into account
(b) SLS MSE versus Transmitter Antenna Spacing
(41)
−1 Figure 2: MMSE and SLS MSE performance versus transmitter
−1
P opt = arg min tr Λmc + σn−2 Mr−1 Qmc
H
PP H Qmc , antenna element spacing.
p
subject to P 2F = P.
To minimize MSE in (44), a water-filling based algorithm
(42) can be applied to optimize the power allocation to training
By denoting the combined training matrix as symbols. The symbols in the optimal training matrix are
given as [9, 10]
P = σn−1 Mr−1/2 Qmc
H
P, (43) &
−1 +
expression (41) can be rewritten as pi = (μ − λi,mc ) , (48)
−1
MSEmc −1
Λmc
H in which (x)+ = max(x, 0) and μ needs to be found to satisfy
MMSE = min tr +PP . (44)
2
P F =P P 2F = P.
Finally, the optimal training matrix is given as
Based on properties of the unitary eigenvector matrix Qmc ,
H
P P has a diagonal structure as given by P opt = σn−2 Mr−1 Qmc ([μI − Λmc
−1 + 1/2
] ) . (49)
⎡ ⎤
p12
⎢ p22 ⎥
⎢ ⎥ 5. Simulation Results
=⎢ ⎥,
H
PP ⎢ .. ⎥ (45)
⎣ . ⎦
2
This section reports the computer simulations aiming at the
pM t assessment of effects of mutual coupling on the training-
where pi2 is the power of each combined training symbol. based MIMO channel estimation methods. In the under-
When no optimization is applied, the transmit power is taken investigation, the transmitter and receiver are assumed
equally distributed into the training symbols as given by to be equipped with 4-element uniform array antennas. The
elements are wire dipoles having length of 0.5λ. The antenna
pi2 = σn−2 Mr−1 Mt−1 P. (46) element spacing at the receiver is fixed to 1λ while the spacing
at the transmitter varies from 0.1λ to 1λ. Equations (38)
Next (44) can be rewritten as, and (24) are used to construct the MIMO channel with and
!
Mt
−1
without taking mutual coupling effects into account. The
−1
MSEmc
MMSE = min
2
(λi,mc + pi2 ) transmit SPNR is fixed at 25 dB.
P F =P i=1 (47) Figure 2 presents the MSE performance of MMSE and
assume Mt = Mr , SLS methods versus transmitter antenna elements spacing. In
this figure, the blue and red curves indicate with and without
in which λi,mc is the ith eigenvalue in Λ. taking into account mutual coupling cases, respectively. For
International Journal of Antennas and Propagation 7
MMSE MSE versus antenna element spacing MSE versus antenna element spacing
taking MC into account without taking MC into account
0.05 0.05
0.05 0.05 0.045
0.045
0.04 0.04 0.04
MSE
MSE
1 0.05 1 0.05
0.8 0.045
0.8 0.045 0.04
dr/λ
0.6
dr/λ
Figure 3: MMSE MSE versus transmitter and receiver antenna Figure 4: MMSE MSE versus transmitter and receiver antenna
element spacing taking MC into account. element spacing without taking MC into account.
both MMSE and SLS methods, an apparent difference occurs the optimization does not narrow the difference between the
between the MSE performance with and without taking two sets of results with and without taking mutual coupling
mutual coupling into account when the transmitter antenna into account when the antenna element spacing is between
element spacing varies from 0.1λ to 0.5λ. The MSE obtained 0.1λ to 1.0λ. In the case of a compact multiple antenna
for taking into account mutual coupling effects is worse mobile communication device whose antenna element spac-
than the MSE obtained without taking into account the ing is between 0.1λ to 0.5λ, the antenna decoupling is a
mutual coupling. When the antenna element spacing is larger required to achieve a better channel estimation performance.
than 0.5λ, the difference disappears. This finding agrees with Figure 8 further demonstrates the advantages of the
the prediction based on the sum of channel correlation optimization method. For the undertaken simulation, the
eigenvalues that shows that for the spacing between 0.1λ antenna element spacing at the receiver is fixed to 1λ while
to 0.5λ the mutual coupling undermines the estimation the spacing at the transmitter is made 0.2λ. From the pre-
accuracy. sented results one can see that the performances of MSE with
Further investigations are done by varying the antenna optimization are always better than the ones without it. At a
element spacing in both transmitter and receiver arrays. The lower transmit SPNR, the results for MSE with optimization
results are given in Figures 3, 4, 5 and 6. In each figure, there are significantly improved. The results also show that when
are two subfigures showing three dimensional (3D) and two the transmitter antenna element spacing is within 0.1λ to
dimensional (2D) views. Figure 3 shows the MMSE MSE 0.5λ, the MSE taking into account mutual coupling is worse
performance under the impact of mutual coupling. One can than the one without taking into account mutual coupling.
see that the worst estimation error occurs when the antenna This is irrespective from whether optimization is used or not.
element spacing in the transmitter and receiver is between
0.3λ and 0.4λ. Figure 3 shows the MMSE MSE performance 6. Conclusion
without taking into account the mutual coupling. When the
antenna element spacing in the transmitter and receiver is This paper has reported the investigations on the training-
between 0.5λ and 0.7λ, MSE achieves its best performance. based channel estimation methods for a narrowband MIMO
Similar results are also obtained for the SLS method, as system, in which both the transmitter and the receiver are
presented in Figures 5 and 6. equipped with multiple element antennas in the form of
Figure 7 shows the results for the MMSE estimation wire dipoles. The investigations have included the effect of
method when the training sequence is optimized. For mutual coupling in addition to spatial correlation that is
comparison purposes, the results for nonoptimized MMSE present at the transmitting and receiving array antennas due
are also plotted. One can see that with the optimized training to surrounding scattering objects. The spatial correlation
symbols, the channel estimation error decreases. However, has been taken into account using the Kronecker model, in
8 International Journal of Antennas and Propagation
SLS MSE versus antenna element spacing MMSE MSE versus transmitter antenna spacing
taking MC into account 0.05
0.778 0.045
0.785
0.78 0.776 0.04
MSE
0.775 0.774
0.035
0.77 0.772
MSE
0.765 0.03
1 0.77
0.5 0. 6 0.8 1 0.025
dr /
λ 0 0 0.2 0.4 0.768
dt/λ 0.02
(a) 3D view
0.015
1
0.778 0.01
0.8
0.776 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
dr/λ
0.77
0.772 1
0.765
1 0.77
1 0.8
0.5 0.6 0.8
dr /
λ 0 0 0.2 0.4 0.768
0.6
dt/λ
0.4
(a) 3D view
0.2
1
0.778 0
0.8
0.776 0 5 10 15
dr/λ
(b) 2D view Figure 8: MMSE MSE for versus transmit SPNR with and without
taking mutual coupling effects into account for optimized and
Figure 6: SLS MSE versus transmitter and receiver antenna element nonoptimized cases.
spacing without taking MC into account.
which correlation matrices for the transmitting and receiving carried out for the case when the transmitting and receiving
sides are separated. The mutual coupling has been included sides are equipped with four-element uniform half-wave
using the closed-form expressions for impedance matrices of dipole arrays surrounded by circles of uniformly distributed
parallel side-to-side wire dipoles. Two cases of nonoptimized scattering objects. The obtained simulation results have
and optimized training sequence transmission schemes have shown that at the antenna (dipole) spacing between 0.1λ to
been considered. The computer simulations have been 0.5λ, mutual coupling decreases the spatial correlation and
International Journal of Antennas and Propagation 9
adversely affects the channel estimation offered by MMSE [11] T. Svantesson and A. Ranheim, “Mutual coupling effects on
and SLS methods. When the spacing is larger than 0.5λ the capacity of multi-element antenna systems,” in Proceedings
the MSE is slightly different when the mutual coupling of the IEEE International Conference on Acoustics, Speech, and
is taken into account or neglected. Next considerations Signal Processing (ICASSP ’01), pp. 2485–2488, Las Vegas, Nev,
have focused on optimization of training signals for MIMO USA, April 2001.
[12] J. W. Wallace and M. A. Jensen, “The capacity of MIMO
channel estimation. When the optimization of the training
wireless systems with mutual coupling,” in Proceedings of the
sequence is applied, an improvement in MSE has been
IEEE Vehicular Technology Conference (VTC ’02), pp. 696–700,
demonstrated irrespectively whether the mutual coupling Vancouver, Canada, 2002.
effects are neglected or taken into account. This improve- [13] J. W. Wallace and M. A. Jensen, “Mutual coupling in MIMO
ment is more significant at a lower transmit SPNR. The wireless systems: a rigorous network theory analysis,” IEEE
difference in MSE for with and without taking into account Transactions on Wireless Communications, vol. 3, no. 4, pp.
mutual coupling effects is unchanged when the antenna 1317–1325, 2004.
element spacing is within 0.1λ to 0.5λ. This means that [14] P. N. Fletcher, M. Dean, and A. R. Nix, “Mutual coupling
for a compact multiple antenna mobile communication in multi-element array antennas and its influence on MIMO
device the improvement of MSE requires decoupling of the channel capacity,” Electronics Letters, vol. 39, no. 4, pp. 342–
antenna elements to achieve a better estimation of MIMO 344, 2003.
channel. [15] S. Lu, H. T. Hui, M. Bialkowski, and X. Liu, “Optimizing
transmission strategy in single-user MIMO systems under the
influence of antenna mutual coupling,” ECTI Transactions on
References Electrical Engineering Electronics, and Communications, vol. 6,
no. 2, pp. 73–78, 2008.
[1] E. Telatar, “Capacity of multi-antenna Gaussian channels,” [16] S. Lu, X. Liu, H. T. Hui, M. Bialkowski, N. Seman, and
European Transactions on Telecommunications, vol. 10, no. 6, H. Y. Zhang, “Power allocation strategy for compact MIMO
pp. 585–595, 1999. transmitters with Uniform Circular Arrays under the influence
[2] T. L. Marzetta and B. M. Hochwald, “Capacity of a mobile of antenna mutual coupling,” in Proceedings of the IEEE
multiple-antenna communication link in Rayleigh flat fading,” Asia Pacific Microwave Conference (APMC ’07), pp. 11–14,
IEEE Transactions on Information Theory, vol. 45, no. 1, pp. Bangkok, Thailand, December 2007.
139–157, 1999. [17] S. Lu, H. T. Hui, M. Bialkowski, and X. Liu, “Optimizing
[3] A. Narula, M. J. Lopez, M. D. Trott, and G. W. Wornell, transmission strategy in single-user MIMO systems under
“Efficient use of side information in multiple-antenna data the influence of antenna mutual coupling,” in Proceedings of
transmission over fading channels,” IEEE Journal on Selected the IEEE International Symposium on Communications and
Areas in Communications, vol. 16, no. 8, pp. 1423–1436, 1998. Information Technologies (ISCIT ’07), pp. 370–374, Sydney,
[4] C. Budianu and L. Tong, “Channel estimation for space- Australia, October 2007.
time orthogonal block codes,” IEEE Transactions on Signal [18] M. E. Bialkowski, P. Uthansakul, K. Bialkowski, and S.
Processing, vol. 50, no. 10, pp. 2515–2528, 2002. Durrani, “Investigating the performance of MIMO systems
[5] A. Grant, “Joint decoding and channel estimation for linear from an electromagnetic perspective,” Microwave and Optical
MIMO channels,” in Proceedings of the IEEE Wireless Commu- Technology Letters, vol. 48, no. 7, pp. 1233–1238, 2006.
nications and Networking Conference, pp. 1009–1012, Chicago, [19] X. Liu, S. Lu, M. E. Bialkowski, and H. T. Hui, “MMSE
Ill, USA, September 2000. channel estimation for MIMO system with receiver equipped
[6] A. S. Kyung, R. W. Heath, and B. K. Heung, “Shannon with a circular array antenna,” in Proceedings of the Asia-
capacity and symbol error rate of space-time block codes Pacific Microwave Conference (APMC ’07), Bangkok, Thailand,
in MIMO Rayleigh channels with channel estimation error,” December 2007.
IEEE Transactions on Wireless Communications, vol. 7, no. 1, [20] S. Lu, H. T. Hui, M. E. Bialkowski, X. Liu, H. S. Lui, and
pp. 324–333, 2008. N. V. Shuley, “The effect of antenna mutual coupling on
[7] X. Zhang and B. Ottersten, “Performance analysis of V-BLAST channel estimation of MIMO-OFDM systems,” in Proceedings
structure with channel estimation errors,” in Proceedings of the of the IEEE Antennas and Propagation Society International
4th IEEE Workshop on Signal Processing Advances in Wireless Symposium (APS ’07), pp. 2945–2948, Honolulu, Hawaii,
Communications (SPAWC ’03), pp. 487–491, Rome, Italy, June USA, June 2007.
2003. [21] S. M. Kay, Fundamentals of Statistical Signal Processing:
[8] P. Layec, P. Piantanida, R. Visoz, and A. O. Berthet, “Capacity Estimation Theory, Prentice-Hall, Upper Saddle River, NJ,
bounds for MIMO multiple access channel with imperfect USA, 1993.
channel state information,” in Proceedings of the IEEE Informa- [22] E. G. Larsson and P. Stoica, Space-Time Block Coding for Wire-
tion Theory Workshop (ITW ’08), pp. 21–25, Porto, Portugal, less Communication, Cambridge University Press, Cambridge,
May 2008. UK, 2003.
[9] M. Biguesh and A. B. Gershman, “MIMO channel estimation: [23] C.-N. Chuah, D. N. C. Tse, J. M. Kahn, and R. A. Valenzuela,
optimal training and tradeoffs between estimation tech- “Capacity scaling in MIMO wireless systems under correlated
niques,” in Proceedings of the IEEE International Conference fading,” IEEE Transactions on Information Theory, vol. 48, no.
on Communications (ICC ’04), vol. 5, pp. 2658–2662, Paris, 3, pp. 637–650, 2002.
France, June 2004. [24] S. Da-Shan, G. J. Foschini, M. J. Gans, et al., “Fading corre-
[10] M. Biguesh and A. B. Gershman, “Training-based MIMO lation and its effect on the capacity of multi-element antenna
channel estimation: a study of estimator tradeoffs and optimal systems,” in Proceedings of the IEEE International Conference
training signals,” IEEE Transactions on Signal Processing, vol. on Universal Personal Communications (ICUPC ’98), Florence,
54, no. 3, pp. 884–893, 2006. Italy, 1998.
10 International Journal of Antennas and Propagation
Research Article
Effect of Antenna Mutual Coupling on
MIMO Channel Estimation and Capacity
Copyright © 2010 X. Liu and M. E. Bialkowski. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
This paper describes investigations into the antenna mutual coupling (MC) effect on channel estimation and capacity of a multiple-
input multiple-output (MIMO) wireless communication system. The presented investigations close the gap existing in the previous
works which assessed the effect of mutual coupling on MIMO capacity under the assumption of availability of perfect channel state
information (CSI) at the receiver. The new approach assumes that the perfect CSI is not available due to channel estimation errors.
The investigations are carried out for different spacing between array antenna elements producing a varying effect of mutual
coupling on the channel estimation and the resulting MIMO channel capacity.
properly, the MIMO system has to estimate properties of the the input signal S and output signal Y . Assuming that S is a
channel before decoding the received data. set of signals, p(s) is the probability of S given s, the entropy
Obtaining an accurate CSI can be accomplished using or a measure of uncertainty of S can be written as,
suitable channel estimation methods. The methods based
on the use of training sequences, known as the training- H(S) = E{I(s)} = − p(s) log p(s). (1)
based channel estimation methods, are the most popular. s∈S
In [16, 17], several training-based methods including least
square (LS) method, scaled least square (SLS) method and in which I(s) is the self-information which is the entropy
minimum mean square error (MMSE) method have been contribution of an individual message; E{·} stands for the
investigated. It has been shown that the accuracy of the expected value.
investigated training-based channel estimation methods is Similarly, the entropy of Y given y can be expressed as,
influenced by the transmitted signal to noise ratio (SNR)
in the training mode, and the number of antenna elements H(Y ) = E I y =− p y log p y . (2)
at the transmitter and receiver. In particular, it has been y ∈Y
pointed out that when the transmitted SNR and the number
The joint entropy of S and Y is given as,
of antenna elements are fixed, the SLS and MMSE methods
offer better performance than the LS method. In [18–20],
H(S, Y ) = E − log p s, y = − p s, y log p s, y . (3)
it has been shown that if the training-based SLS or MMSE s,y
channel estimation methods are used to estimate CSI, their
performance can be improved when the channels exhibit a The conditional entropy (or conditional uncertainty) of S
higher spatial correlation level. According to [9–15], when given Y is expressed as,
the antenna element spacing is within 0.2λ and 0.4λ, the
mutual coupling can decrease the spatial correlation level. H(S | Y ) = E H S | y =− p y p s | y log p s | y
Therefore in such cases, the accuracy of training-based y ∈Y s∈S
channel estimation methods is compromised. From these
p s, y
considerations one can see a tradeoff between the accuracy = − p s, y log .
of channel estimation and the resulting capacity when the s,y p y
MIMO system operates under mutual coupling conditions. (4)
When the correlation increases, the channel estimation error
becomes smaller. However, at the same time the increased The conditional entropy gives the uncertainty about the
correlation is responsible for decreasing the effective degree channel input after the channel output has been observed.
of freedom thus the MIMO channel capacity. In this case, the uncertainty of the channel input can be
In this paper, the overall effect of antenna mutual expressed as H(X) − H(X | Y ). This is commonly known
coupling on performance of an MIMO wireless system is as the mutual information of the channel which is given by,
investigated. The channel capacity that provides the over-
all measure of MIMO system performance is determined p y|s
I(S; Y ) = H(X) − H(X | Y ) = p s, y log2 .
assuming an imperfect CSI is available at the receiver. A y s p y
closed-form mathematical expression for the MIMO channel (5)
capacity that includes channel estimation errors is derived.
The effects of mutual coupling on channel estimation and The mutual information that can be optimized in terms
resulting capacity are investigated via computer simulations of p(s) is independent of the channel. Because the wireless
using a suitable MIMO channel model. communication channel capacity is defined as the maximum
The rest of the paper is organized as follows. Section 2 mutual information, then the channel capacity can be
introduces MIMO channel and its capacity. Also described determined from,
in this section are the training-based MIMO channel esti-
mation methods, expressions for MIMO capacity including C = max I(S; Y ). (6)
p(s)
channel estimation errors, and the Kronecker model [21, 22]
of MIMO channel. Section 3 describes the mathematical In a narrow band MIMO wireless communication system,
model for the antenna mutual coupling. Section 4 describes the channel is described by a complex matrix H. The
computer simulation results for the MIMO system capacity relationship between the input signal S and the output signal
under the assumption of perfect CSI, as well as when an Y can be expressed as,
imperfect CSI due to estimation errors is available at the
receiver. Section 5 concludes the paper. Y = HS + N, (7)
The upper bond mutual information of MIMO can be The goal is to estimate the complex channel matrix H
written as, from the knowledge of Y and P. The transmitted power in
the training mode is constrained by the following expression
I(S; Y ) ≤ I(S; Y | H)
P 2F = P p , (14)
= E log2 det πe HQH H + RN − log2 det(πeRN )
where P p is a given power constant and · 2F stands for the
det HQH H + RN Frobenius norm. The training matrices are assumed to be
= E log2
det(RN ) orthogonal [16, 17] and the transmitted signal to noise ratio
(SNR) in the training mode is set to ρ equal to P p /σn2 .
Q Using the LS method, the estimated channel matrix can
= E log2 det IMr + 2 HH H .
σn be written in the form [23]
(8)
HLS = Y P † , (15)
Assuming the number of transmitting and receiving anten-
−1
nas is the same (Mt equal to Mr ) and the transmitted power where {·}† stands for pseudoinverse and P † = P H (PP H ) .
is equally allocated to the transmitting antennas, (8) turns The mean square error (MSE) of the estimated channel
into matrix in the LS method is given as
P 2
I(S; Y | H) = E log2 det IMr + s 2 HH H MSELS = E H − H LS F subject to : P 2F = P p (16)
Nt σn
(9)
ρ According to [16, 17], the optimal training sequence satisfy-
= E log2 det IMr + HH H .
Nt ing (17) is given by
Similar derivations apply for the MMSE channel estimation The channel capacity of MIMO system with an imperfectly
method. Using this method, the estimated channel matrix known H at the receiver is defined as the maximum mutual
can be obtained from, information between Y and S:
−1
H MMSE = Y P H RH P + σn2 Mr I P H RH . (23) C = max I S; Y , H . (32)
tr{Q}≤Ps
By denoting the channel estimation error as,
If the transmitter does not have any knowledge of the
e = H − H MMSE (24) channel, the mutual information in (32) can be written as,
C
I S; Y | H = h S | H − h S − uY | Y , H . (40)
pH H H
= E log2 det IMr +
pRe + σn2 IMr
For the case of S | H and (S − uY | Y , H)having a Gaussian
distribution, (40) can be expressed as [24–26], (P /M )H H H 1
= E log2 det IMr + s t2
σn IMr + Ps /Mt σn2 Re
I S; Y | H ρH H H 1
= E log2 det IMr +
Mt IMr + ρ/Mt Re
≥ E log2 [det(πeQ)]
1
= E log2 det IMr + H H H .
H
− E log2 det πeE (S − uY )(S − uY ) | H Mt /ρ IMr + Re
(44)
pH H H
≥ E log2 det IMr + .
IMr pσe2 + σn2 Equation (44) indicates that for a fixed value of ρ, the
(41) capacity is a function of the estimated channel matrix H
and the channel estimation error correlation Re . As a result,
the channel matrix and the quality of channel estimation
The lower bound of the ergodic channel capacity is then influence the MIMO capacity.
given as
2.4. MIMO Channel Model. In the undertaken investiga-
C tions, the Kronecker channel model [21, 22] is postulated
to construct the channel matrix H. In this model, the
pH H H transmitter and receiver correlation matrices are assumed to
= E log2 det IMr + 2 be separable and the channel matrix is represented by:
pσe + σn2
(P /M )H H H 1 H = R1/2 1/2
R GH RT , (45)
= E log2 det IMr + s t 2
σn 1+(Ps /Mt ) σe2 /σn2
where GH is the matrix including identical independent
distributed (i.i.d) Gaussian entries with the zero mean and
ρH H H 1
= E log2 det IMr + the unit variance, and RR and RT are the spatial correlation
Mt 1 + ρ/Mt (MSE/Mr )
matrices at the receiver and transmitter, respectively.
Here, it is assumed that the transmitting and receiving
1
= E log2 det IMr + H H H . sides of MIMO system are equipped with vertically polarized
Mt /ρ + MSE/Mr
wire dipole antennas. The scattering environment is repre-
(42)
sented by circles of uniformly distributed scattering objects
surrounding the transmitting and receiving nodes. In this
Equation (42) indicates that for a fixed value of ρ, the case, the spatial correlation matrix elements can be obtained
capacity is a function of the estimated channel matrix H and using the Clark’s model as given by,
the channel estimation MSE.
An uncorrelated Rayleigh channel is ideal. In practical ρiR(T)
j = J0 κdi j (46)
cases, an MIMO channel is usually correlated due to
the propagation environment. Consequently, the estimation where J0 stands for the zero-order Bessel function, κ is a wave
error matrix is correlated rather than i.i.d. In this case, the number and di j is the distance between elements i and j of
variances of elements in the error matrix are not equal to the uniform array antenna.
each other and the error correlation matrix is not diagonal. The required correlation matrices RT and RR can then be
Therefore, (30) and (31) are not true anymore. Thus, (37) generated using (46)
needs to be modified as, ⎡ ⎤
ρR(T) · · · ρ1,M
R(T)
⎢ 1,1 rt ⎥
⎢ ⎥
⎢ . .. .. ⎥
QH H RR(T) = ⎢ .. . . ⎥ (47)
u= ⎢ ⎥
⎣ ⎦
H
HQ H + QRe IMr + σn2 IMr R(T) R(T)
ρM rt , j
· · · ρMrt ,Mrt
(43)
pH H
= . Having determined RT and RR , the channel matrix can be
pH H H + pRe IMr + σn2 IMr calculated using (45).
6 International Journal of Antennas and Propagation
Capacity
16
[27]. Assuming the array is formed by Mrt wire dipoles,
the mutual matrix can be calculated using the following 14
relationship involving the impedance matrix
12
−1
C = (ZA + ZT ) Z + ZT IMrt (48)
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
where ZA is the element impedance in isolation, for example,
d/λ
when the wire dipole is λ/2, its value is ZA = 73 + j42.5[Ω];
ZT is impedance of the receiver at each element chosen as (a) Capacity versus transmitter antenna spacing
the complex conjugate of ZA to obtain the impedance match. 20
The mutual impedance matrix Z is given by
⎡ ⎤ 18
Z A + ZT Z12 ··· Z1Mrt
Capacity
⎢ ⎥ 16
⎢ Z21 Z A + ZT · · · Z2Mrt ⎥
⎢ ⎥
⎢ ⎥ 14
Z=⎢
⎢ .. .. .. .. ⎥
⎥
(49)
⎢ . . . . ⎥ 12
⎣ ⎦
ZMrt 1 ZMrt 2 · · · Z A + ZT 10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Note that this expression provides the circuit representation d/λ
for mutual coupling in array antennas. It is valid for single
Capacity taking MC into account
mode antennas. The assumed here wire dipoles fall into
Capacity without taking MC into account
this category. For a side-by-side array configuration of wire
dipoles having length l equal to 0.5λ, the expressions for (b) Capacity versus receiver antenna spacing
{Zmn } can be adapted from [27, 28] and are rewritten here
Figure 1: A 4 × 4 MIMO channel capacity considering mutual
as, coupling effect versus transmitter or receiver antenna spacing
Zmn at SNR equal to 20 dB. (a) Capacity versus transmitter antenna
spacing, receiver antenna spacing is set to 1λ. (b) Capacity versus
⎧
⎪
⎪ 30[0.5772+ln(2κl) − Ci (2κl)]+ j[30Si (2κl)], m = n, receiver antenna spacing, transmitter antenna spacing is set to 1λ.
⎪
⎪
⎨
= 30[2Ci (u0 ) − Ci (u1 ) − Ci (u2 )]
⎪
⎪
⎪
⎪
⎩ − j[30(2S (u ) − S (u ) − S (u ))], m= The modified transmit and receive correlation matrices
i 0 i 1 i 2 / n,
(including mutual coupling effects) can be expressed by
(50)
where κ is the wave number equal to 2π/λ, RRmu = CR R1/2
R ,
(54)
u0 = κdh , RTmu = R1/2
T CT .
'
Using (54), the expression shown in (45) can be rewritten as
u1 = κ dh2 + l2 + l (51) (55) and (56), respectively,
'
u2 = κ dh2 + l2 − l Hmc = RRmu GH RTmc , (55)
0.05 35
0.045
0.04 30
MSE
0.035
Channel capacity
25
0.03
0.025 20
0.02
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
15
dt/λ
(a) MMSE MSE versus transmitter antenna spacing 10
0.774
0.773 5
10 12 14 16 18 20 22 24 26 28 30
0.772
SNR(ρ) (dB)
0.771
MSE
10 12 14 16 18 20 22 24 26 28 30
CSI is available at the receiver. In the undertaken computer
simulations, both transmitter and receiver are assumed to SNR (ρ) (dB)
be equipped with a 4-element uniform array antenna. The MMSE MSE taking MC into account
elements are assumed to be wire dipoles having length of MMSE MSE without taking MC into account
0.5λ. Equations (45) and (55) are used to model the MIMO
(b) MMSE channel estimation MSE versus SNR
channel. Both cases when mutual coupling effects are taken
into account or disregarded are considered. The transmit Figure 3: (a) Channel capacity considering MMSE estimation error
SNR is fixed at 20 dB. In the figure, the blue curves represent for a 4 × 4 MIMO under mutual coupling effect versus SNR.
the case of taking into account mutual coupling effect while (b) MMSE MSE performance for a 4 × 4 MIMO under mutual
the blue curves represent the case when the mutual coupling coupling effect versus SNR; transmitter antenna spacing is set to 0.
effect is not taken into account. From the results presented 1λ, receiver antenna spacing is set to 1λ.
in Figure 1, one can see that when the antenna spacing is
smaller than 0.4λ, the capacity with mutual coupling effect
is larger than the one without taking mutual coupling effect the MMSE shows a superior performance. This agrees with
into account. When the spacing is larger than 0.4λ, the the findings in publications [16, 17]. On the other hand,
channel capacity when mutual coupling effect is taken into for both MMSE and SLS methods, an apparent difference
account is smaller than when the mutual coupling effect is in MSE occurs when mutual coupling effects are taken
disregarded. The cross-point occurs at the antenna spacing into account or disregarded when the transmitter antenna
equal to 0.4λ. These results show that when the antenna element spacing varies from 0.1λ to 0.5λ. It is apparent that
element spacing is within 0.2λ and 0.4λ, the mutual coupling MSE when mutual coupling effects are taken into account is
helps to increase the channel capacity when a perfect CSI is worse than the one when the mutual coupling is disregarded
assumed to be available at the receiver. (not taken effect into account). When the antenna element
Figure 2 presents the MSE performance of the MMSE spacing is larger than 0.5λ, the difference disappears. This
and SLS channel estimation methods versus the transmitter finding agrees with the prediction that when the antenna
antenna element spacing. One can see that MSE for the element spacing is between 0.2λ to 0.4λ the mutual coupling
MMSE method is smaller than for the SLS method. Therefore compromises the accuracy of MIMO channel estimation.
8 International Journal of Antennas and Propagation
4.2. Imperfect CSI due to Channel Estimation Errors. This IEEE Transactions on Information Theory, vol. 45, no. 1, pp.
time, computer simulations for MIMO channel capacity 139–157, 1999.
include both mutual coupling effects as well as channel [4] A. Narula, M. J. Lopez, M. D. Trott, and G. W. Wornell,
estimation errors. MMSE method is assumed to be used “Efficient use of side information in multiple-antenna data
to estimate the channel matrix (CSI). Again, both the transmission over fading channels,” IEEE Journal on Selected
transmitter and receiver are assumed to be equipped with Areas in Communications, vol. 16, no. 8, pp. 1423–1436, 1998.
4-element uniform array antennas. The antenna elements [5] A. Grant, “Joint decoding and channel estimation for linear
are assumed to be wire dipoles having length of 0.5λ. The MIMO channels,” in Proveedings of the IEEE Wireless Commu-
antenna spacing at the receiver is set to 1λ while the one at nications and Networking Conference, pp. 1009–1012, Chicago,
III, USA, September 2000.
transmitter is assumed to be 0.2λ.
[6] A. S. Kyung, R. W. Heath, and H. K. Baik, “Shannon
Figure 3(a) shows the results for the 4 × 4 MIMO
capacity and symbol error rate of space-time block codes in
channel capacity assuming perfect and imperfect CSI at the
MIMO rayleigh channels with channel estimation error,” IEEE
receiver. One can see that when the channel estimation error Transactions on Wireless Communications, vol. 7, no. 1, pp.
is taken into account, the channel capacity is lower than 324–333, 2008.
the one assuming a perfect CSI at the receiver. Also, at [7] X. Zhang and B. Ottersten, “Performance analysis of V-BLAST
the transmitter antenna element spacing equal to 0.2λ, the structure with channel estimation errors,” in Proceedings of the
channel capacity taking into account the mutual coupling IEEE 4th Workshop on Signal Processing Advances in Wireless
effect is always higher than when the mutual coupling is Communications (SPAWC ’03), pp. 487–491, June 2003.
neglected. Figure 3(b) shows the results for MSE of MMSE [8] P. Layec, P. Piantanida, R. Visoz, and A. O. Berthet, “Capacity
method for the 4 × 4 MIMO system when mutual coupling bounds for MIMO multiple access channel with imperfect
effect is included or disregarded. One can see that for both channel state information,” in Proceedings of the IEEE Infor-
cases, the MSE decreases as SNR becomes larger. One can mation Theory Workshop (ITW ’08), pp. 21–25, May 2008.
see that MSE becomes worse when mutual coupling effect [9] T. Svantesson and A. Ranheim, “Mutual coupling effects on
is included. The results presented in Figures 3(a) and 3(b) the capacity of multi-element antenna systems,” in Proceedings
show that although the mutual coupling undermines the of the Acoustics, Speech, and Signal Processing (ICASSP ’01), vol.
channel estimation accuracy, it does not adversely affect 4, pp. 2485–2488, April 2001.
the channel capacity. For the presented cases, the mutual [10] J. Wallace and M. Jensen, “The capacity of MIMO wireless
coupling decreasing the spatial correlation level improves systems with mutual coupling,” in Proceedings of the IEEE
the signal transmission quality. Therefore, the worsening Vehicular Technology Conference (VCT ’02), pp. 696–700,
Vancouver, Canada, 2002.
channel estimation has only an insignificant impact on
[11] J. Wallace and M. Jensen, “Mutual coupling in MIMO
capacity.
wireless systems: a rigorous network theory analysis,” IEEE
Transactions on Wireless Communications, vol. 3, no. 4, pp.
1317–1325, 2004.
5. Conclusion
[12] P. N. Fletcher, M. Dean, and A. R. Nix, “Mutual coupling
In this paper, the effect of mutual coupling on MIMO in multielement array antennas and its influence on MIMO
channel estimation accuracy and the resulting channel channel capacity,” IEEE Electronics Letters, vol. 39, no. 4, pp.
capacity has been investigated. The mathematical analysis 342–344, 2003.
and simulation results have shown that when the antenna [13] S. Lu, H. T. Hui, M. E. Bialkowski, and X. Liu, “Optimizing
signal transmission in a MIMO system influenced by antenna
element spacing at either transmitter or receiver is within
mutual coupling,” The ECTI Transactions on Electrical Engi-
0.2λ and 0.4λ, the mutual coupling decreases the spatial
neering, Electronics, and Communications, vol. 6, no. 2, pp. 73–
correlation level and undermines the estimation accuracy 78, 2008.
of the MIMO channel. However, the reduced accuracy of [14] S. Lu, X. Liu, H. T. Hui, M. Bialkowski, N. Seman, and
channel estimation does not have a significant impact on H. Y. Zhang, “Power allocation strategy for compact MIMO
the channel capacity. The carried out computer simulations transmitters with uniform circular arrays under the influence
have shown that the decreased spatial correlation improves of antenna mutual coupling,” in Priceedings of the Asia-Pacific
the capacity of MIMO system despite having an adverse effect Microwave Conference Proceedings (APMC ’07), December
on the channel estimation. 2007.
[15] S. Lu, H. T. Hui, M. Bialkowski, and X. Liu, “Optimizing
transmission strategy in single-user MIMO systems under the
References influence of antenna mutual coupling,” in Proceedings of the
International Symposium on Communications and Information
[1] E. Telatar, “Capacity of multi-antenna Gaussian channels,” Technologies (ISCIT ’07), pp. 370–374, October 2007.
European Transactions on Telecommunications, vol. 10, no. 6,
[16] M. Biguesh and A. B. Gershman, “MIMO channel estimation:
pp. 585–595, 1999.
Optimal training and tradeoffs between estimation tech-
[2] C. Budianu and L. Tong, “Channel estimation for space- niques,” in Proceedings of the IEEE International Conference
time orthogonal block codes,” IEEE Transactions on Signal on Communications (ICC ’04), vol. 5, pp. 2658–2662, Paris,
Processing, vol. 50, no. 10, pp. 2515–2528, 2002. France, June 2004.
[3] T. L. Marzetta and B. M. Hochwald, “Capacity of a mobile [17] M. Biguesh and A. B. Gershman, “Training-based MIMO
multiple-antenna communication link in rayleigh flat fading,” channel estimation: a study of estimator tradeoffs and optimal
International Journal of Antennas and Propagation 9
Research Article
The Effect of Mutual Coupling on a High Altitude Platform
Diversity System Using Compact Antenna Arrays
Copyright © 2010 T. Hult and A. Mohammed. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
We analyze the destructive effects of mutual coupling and spatial correlation between the separate antenna elements on a combined
diversity system consisting of multiple HAPs (High-Altitude Platforms) employing various compact MIMO (Multiple-Input
Multiple-Output) antenna array configurations, in order to enhance the mutual information in HAP communication links. In
addition, we assess the influence of the separation angle between HAPs on system performance, and determine the optimal
separation angles that maximize the total mutual information of the system for various compact MIMO antennas. Simulation
results show that although the mutual information is degraded by mutual coupling and spatial correlation, the proposed HAP
diversity system still provides better performance compared to a nondiversity system for all tested scenarios.
Figure 1: The MIMO-HAP diversity system with three HAPs and where η is the intrinsic impedance of vacuum and under-
the channel paths from the transmitter (Tx) antenna array to the scoring denotes a field vector. The base functions Xlm (ϕ, θ)
receiver (Rx) antenna array.
represent mode (l, m) of the electromagnetic field in spheri-
cal coordinates when the far-field of the antenna is projected
onto the virtual sphere. The functions gl and fl in (3)
their characteristics, and modelling are introduced. The are spherical Hankel functions of order l representing an
analysis of mutual coupling and spatial correlation effects on outgoing (transmitted) or incoming (received) wave. The
these channels is explained in Section 5. Simulation results scalar weights aE and aM are the corresponding coefficients
and comparisons are presented in Section 6. Finally the which will give the gain of each orthogonal function (mode)
conclusions are presented in Section 7. for a particular electromagnetic far-field pattern, as shown
by
2. The Proposed Multiple HAP Diversity System
1 k ∗
aE (l, m) ≈ · fl (kr) Ylm θ, ϕ r · Ed3 r,
In this paper, we propose an application for high data rate η0 l(l + 1)
(4)
transmissions using a system employing multiple HAPs. k ∗
This system consists of a number of uncorrelated MIMO aM (l, m) ≈ − gl (kr) Ylm θ, ϕ r · Hd3 r.
l(l + 1)
channels created in the space/polarization domain using
HAP diversity in combination with a special type of compact Using (4) we can calculate which modes are active
MIMO antenna arrays [7–9] as discussed in Section 4. on any arbitrary antenna enveloped by the virtual sphere
Figure 1 shows the location diversity setup for the case of by knowing the electric and magnetic field patterns of
three HAPs separated by the angle θ. the antenna. These modes are theoretically orthogonal to
The propagation channel Hch from the transmitter to the each other and therefore represent independent ports of
receiver can be seen as a simple transformation that contains the antenna. The transmitting antenna-channel transform
the distance dependent-free space gain of the signal being matrix Mtx is then assumed to be the linear transformation
transmitted of the input signals x into the mode domain atx according to
c atx = Mtx x, and for the receiving channel-antenna, we have
Hmn r, f = , (1) a similar transformation matrix from the mode domain arx
4π f |rm − rn |
into the output signals y of the system following y = Mrx arx ,
where |rm − rn | is the distance along the path between where atx and arx are vectors containing the mode gains for a
transmitting antenna element n and receiving antenna specific antenna type.
element m. There are no atmospheric interferences and the
noise in the system is modelled as uncorrelated Gaussian 4. The Compact MIMO Antenna Array
noise. The total MIMO channel can then be assembled as
Configurations
H = Mrx · Hch · Mtx , (2) Each transmit and receive antenna arrangement of the
proposed system consists of a special compact MIMO
where Mrx and Mtx are transformations of the transmitter antenna array configuration, which possesses a structure of
and receiver antenna signals into and out of the mode different designs and complexities as shown in Figure 2.
domain as described in Section 3. The first compact MIMO antenna array is known as the
MIMO-Cube [10], which consists of twelve electric dipoles
3. The Space-Polarization Channel positioned on the twelve edges of a cube. It can be seen
from Figure 3 that this antenna array theoretically has twelve
The polarization and radiation pattern of the electromag- independent antenna ports.
netic field can be expressed as a multipole expansion [10] The second compact antenna that we have investigated
of the spherical wave field emanating from a virtual sphere is known as the MIMO-Tetrahedron, which has six electric
encompassing the antenna array that is being analyzed. dipoles placed at the six edges of a tetrahedron (see Figure 2),
This expansion consists of weighted orthogonal vector base and so theoretically has six independent ports as can be seen
functions on the surface of the virtual sphere that allow for in Figure 4.
International Journal of Antennas and Propagation 3
1
0.5
0
0 5 10 15 20 25 30
Mode number
1
0.5
0
0 5 10 15 20 25 30
Mode number
1
0.5
(a) (b) 0
0 5 10 15 20 25 30
Figure 2: The structure of the two compact MIMO antenna array Mode number
configurations: (a) MIMO-cube and (b) MIMO-tetrahedron. 1
0.5
0
100 0 5 10 15 20 25 30
Mode number
1
0.5
0
50 0 5 10 15 20 25 30
Elevation θ (degrees)
Mode number
1
0.5
0 0
0 5 10 15 20 25 30
Mode number
−100
−200 −100 0 100 200
In Figure 5 we show an example of the activation of the
Azimuth φ (degrees) first 30 modes calculated by (4) for the MIMO-Tetrahedron
antenna array.
Figure 3: The theoretical antenna pattern when all twelve channels In order to separate the modes in Figure 5 into inde-
of the MIMO-Cube antenna are activated. The four antenna beams pendent MIMO channels, we employ a singular value
at elevation angle 0 degrees have vertical linear polarization and decomposition (SVD) operation
the eight beams at elevation ±45 degrees have horizontal linear
polarization. H = UΣVH , (5)
y = UH H(Vx) + UH v, (6)
−50
where y is a vector containing the received and decoded
signal from the independent MIMO subchannels. Inserting
(5) into (6) then orthogonalizes the channels into
−100
−200 −100 0 100 200
y = Σx + w, (7)
Azimuth φ (degrees)
where w = UH v is the received noise. Figure 6 shows
Figure 4: The theoretical antenna pattern when all six channels of the orthogonalized subchannels for the MIMO-Cube and
the MIMO-Tetrahedron antenna are activated. The three antenna MIMO-Tetrahedron antenna arrays, respectively. It can
beams at elevation angle 30 degrees have vertical linear polarization
be clearly seen from this figure that the MIMO-Cube
and the three beams at elevation −45 degrees have horizontal linear
polarization. antenna provides superior channel gain levels compared to
the MIMO-Tetrahedron antenna. This translates to higher
4 International Journal of Antennas and Propagation
0 0
−10 −5
−20 −10
Relative channel gain (dB)
−80 −35
−90 −40
−100 −45
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6
Channel number Channel number
(a) (b)
Figure 6: The orthogonalized 12 subchannels of the MIMO-Cube antenna (a), and the 6 subchannels of the MIMO-Tetrahedron antenna
(b).
system mutual information as will be shown in the results singular value decomposition (SVD) technique, we can now
section. write the channel frequency response matrix H(k) as
Through the combination MIMO-OFDM (Orthogonal
Frequency Division Multiplexing), we also achieve orthog- H(k) = U(k)Σ(k)VH (k), (11)
onality over the frequency domain by using the Discrete where Σ(k) is an M × N matrix containing singular values
Fourier Transform (DFT) with a Cyclic Prefix (CP). Assum- that are larger than zero σ1 (k) ≥ σ2 (k) ≥ · · · ≥ σr (k) > 0,
ing a cyclic prefix CP > τd · fs , where fs = 1/Ts and Ts is the where r is the rank of the matrix H(k), and the M × M matrix
symbol time, we can write the received signals as a circular U(k) and the N × N matrix V(k) are unitary and contain
convolution over the DFT-frame the corresponding eigenvectors as matrix column vectors. To
obtain a diagonalized system, we then define
N
K −1
rm (t) = hmn (t) ⊗ sn (t) + vm (t), t ∈ 0, , (8)
n=1
fs Y(k) = Σ(k)X(k) + N(k), (12)
where
where K is the size of the DFT, m = {1, . . . , M } is receiving
antennas, n = {1, . . . , N } is transmitting antennas, and Y(k) = UH (k)R(k),
vm (t) is AWGN. If we have a MIMO antenna system with S(k) = V(k)X(k), (13)
N transmitting antennas and M receiving antennas, we can
then write the signal for subchannel k in the frequency N(k) = UH (k)W(k).
domain between any pair of transmitting and receiving Since the MIMO-OFDM channels in (12) are uncorrelated
antennas as and the correlation of the noise N(k) is E{N(k)NH (k)} = σn2 ·
N IM , then we can write the mutual information of the system
Rm (k) = Hmn (k)Sn (k) + Wmn (k), k ∈ [0, K − 1], (9) [11] as
n=1
−1
K r
σ 2 (k)
C= log2 1 + σx2m (k) m 2 , (14)
where k denotes each separate subcarrier. If we write (9) in σn
k=0 m=1
vector notation, we get
where σx2m (k) is the variance of the separate uncorrelated
R(k) = H(k)S(k) + W(k), (10) input signals in vector X(k) and r is the rank of the system.
The mutual information in (14) is constrained by the total
where R(k) is an M × 1 vector and S(k) is an N × 1 vector of radiated power from the transmitting antennas, defined as
the received and transmitted signals, respectively. H(k) is the
−1
K r
M × N frequency response matrix of the channel between N
transmitters and M receivers. The noise in the system W(k) is Prad = σx2m (k). (15)
k=0 m=1
an M × 1 vector assumed to be additive white Gaussian noise.
The correlation matrix of the noise vector W(k) is thereby To maximize the total sum of mutual information in all
E{W(k)WH (k)} = σn2 · IM , where σn2 is the variance of the the subchannels, we use the so-called “water-filling” tech-
noise and IM is the M × M identity matrix. Since we are using nique in which we allocate more power to the subchannels
International Journal of Antennas and Propagation 5
3 where Ztx and Zrx are the impedance matrices of the antenna
array. Ztx and Zrx have been calculated numerically by using
2 the finite element method (FEM) for the antenna arrays
shown in Figure 2. An infinite approximation of the array
1 is assumed by not taking the edge effects of the finite array
into account. The terminal voltages are then associated with
0
5 10 15 20 25 30 the source voltages through the source impedances ZS , as
Average signal to noise ratio SNR (dB) described in [13],
Ideal MIMO-cube
utx = ZTX u(TX)
S ,
MIMO-cube with mutual coupling and 6 HAPs (19)
Single HAP system urx = ZRX u(RX)
S ,
Vect. elem. ant. or ideal MIMO-tetrahedron. 3 HAPs
MIMO-tetrahedron with mutual coupling. 3 HAPs
Mutual coupling and spatial correlation. 3 HAPs where u(TX)
S is the source voltage at the transmitter and
MIMO-cube with mutual coupling and u(RX)
S is the source voltage at the receiver generated by the
spatial correlation, 6 HAPs impinging electromagnetic wave. The impedance coupling
matrices are denoted as ZTX and ZRX , and are defined by
Figure 7: The mutual information versus average SNR for differ-
ent configurations of the MIMO-Cube and MIMO-Tetrahedron
Ztx (Ztx + ZS )−1
antenna arrays in comparison to the ideal vector element antenna ZTX = ,
array and a single HAP system. The elevation angle is 10 degrees Ctx
(20)
and the separation angle between HAPs is 20 degrees. Z (Z + ZL )−1
ZRX = L rx ,
Crx
with high eigenvalues. The optimal “water-filling” solution where Ztx and Zrx are the impedance matrices of the
[12] is then given by transmitting and receiving compact antenna arrays, ZS
and ZL are diagonal matrices containing the source and
⎧
⎪
⎪ σn2 σn2 load impedances which have been chosen as the complex
⎪
⎨γ − if γ − > 0,
σm2 (k) σm2 (k) conjugate of the self-impedance of the antenna given by the
σx2m (k) = ⎪ σ 2 (16) diagonal [Ztx ]ii and [Zrx ]ii of the impedance matrices Ztx
⎪
⎪ if γ − 2 n > 0,
⎩0 and Zrx , and Ctx = [Ztx ]11 /([Ztx ]11 + [Z∗tx ]11 ) and Crx =
σm (k)
[Zrx ]11 /([Zrx ]11 + [Z∗rx ]11 ) are normalizing factors [13].
where k = {0, . . . , K − 1}, m = {1, . . . , r }, σn2 is the variance In (8) it was shown that the complex envelope of the
of the noise, and γ is a predefined threshold level of the received signal at the receiving compact array after matched
signal-to-noise ratio in the system and it is dependent on the filtering can be expressed in vector notation as
total transmitted power Prad in (15). The average signal-to-
noise ratio SNRavg in the simulations is calculated as r = Hs + v, (21)
⎛ ⎞
−1
K r where s is the N × 1 vector containing the signals from the
1 σm2 (k) ⎠
SNRavg = 10 log10 ⎝ σx2m (k) · , (17) N transmitting antenna elements, v is the M × 1 zero-mean
Nactive k=0 m=1 σn2 AWGN vector, and H is the normalized M × N channel
matrix that, assuming separable transmit and receive spatial
where Nactive is the number of active channels used for which correlation matrices, is modeled using the Kronecker model
the variance of the input signals σx2m (k) > 0. [18, 19],
5. Mutual Coupling and Spatial Correlation H = ZRX (Rrx )1/2 H0 (Rtx )1/2 ZTX , (22)
The interaction between the elements of the compact where H0 is an i.i.d. channel response without spatial cor-
antenna array will cause mutual interference in the radiation relation and mutual coupling. Rrx and Rtx are the separable
patterns of the array elements, and this effect is known as spatial correlation matrices on the receiving and transmitting
“mutual coupling” [13–17]. The mutual coupling between sides, respectively.
6 International Journal of Antennas and Propagation
5 simulations.
It is evident from Figure 7 that the HAP diversity system
4.5 provides superior performance as compared to the single
4 HAP case, and the MIMO-Cube antenna provides more
mutual information than the MIMO-Tetrahedron antenna
3.5
which is the result of the higher number of acquired HAP
3 platforms. Simulation results also show that although the
mutual information is degraded by correlation and mutual
2.5
coupling, we still achieve significant gain compared to the
2 single HAP case. From Figure 7 it is also clear that for
example, for an average SNR of 20 dB we get a 110%
1.5
0 5 10 15 20 25 30 35 more mutual information from the ideal MIMO-Cube
Separation angle θ (degrees) antenna than from the ideal MIMO-Tetrahedron antenna,
and a 228% more mutual information compared with the
MIMO-cube ant. 6 HAPs nondiversity single HAP system. If we take into account
MIMO-tetrahedron ant. 3 HAPs
the correlation and the mutual coupling, the MIMO-Cube
Figure 8: The total system mutual information versus separation antenna still has a 108% more mutual information than
angle between HAPs for an average SNR of 20 dB for the two the MIMO-Tetrahedron antenna and a 195% more mutual
compact MIMO antenna array configurations. information than the ideal nondiversity single HAP system.
In Figure 8, we show the effect of separation angles
between HAPs on the mutual information of the two
The spatial correlation matrices Rtx and Rrx are calculated compact MIMO antenna arrays. The separation angle θ
between the antenna element positions and polarization (see Figure 2) is varied from 0 to 35 degrees in steps of 5
states according to [20] degrees (due to computational limitations) and plotted for
an average SNR of 20 dB. It is clear from this figure that
ϕ θ ζ κ αq (Θ)α p (Θ) sin(θ)p(Θ)dκ dζ
∗
dθ dϕ the separation angle between HAPs has a great impact on
ρ p,q =
2
the mutual information of the system. It is also evident that
ϕ θ ζ κ αq (Θ) sin(θ)p(Θ)dκ dζ dθ dϕ the optimal separation angle that maximizes the total mutual
1 information of the system at this SNR is found to be approx-
× ,
2
imately 10 degrees for the MIMO-Tetrahedron antenna and
ϕ θ ζ κ α p (Θ)
sin(θ)p(Θ)dκ dζ dθ dϕ approximately 15 degrees for the MIMO-Cube antenna. The
(23) degradation of mutual information performance at wider
separation angles is due to the increase in distance between
where element ρ p,q of matrix Rtx or Rrx is the spatial the HAPs and the ground stations. Figure 8 also suggests that
correlation between antenna elements p and q. ϕ and θ the vector element antenna array is superior to the MIMO-
are the spherical coordinates expressing the spatial domain Tetrahedron array at wide separation angles even though
and ζ and κ are the polarization angle and phase difference, they access the same number of independent subchannels.
respectively, that account for the effects in the polarization
domain [20]. In (23), α(Θ) is the steering vector and
p(Θ) is the joint probability distribution function of the
7. Conclusions
parameter vector Θ = [θ ϕ ζ κ]T , where T denotes a
transpose operator. It is assumed that all the parameters are In this paper, we have investigated the degradation in
independent and that p(ζ) = u{0, π } and p(κ) = u{−π, π } mutual information, due to spatial correlation and mutual
are uniformly distributed. coupling, when using different compact MIMO antenna
arrays in a multiple HAP system employing multiple spatial-
6. Simulation Results polarization modes. Simulation results show that the multi-
ple HAP diversity systems utilizing compact MIMO antenna
In this section, we investigate the mutual information arrays provide significant mutual information enhancement,
improvement resulting from the use of multiple HAPs system even though it can also be noticed that small degradation
employing compact MIMO antenna arrays and compare the in mutual information is resulted due to the effects of
results to that obtained from a system without diversity. In spatial correlation and mutual coupling between the sep-
addition, we will also show the effects of mutual coupling arate antenna elements of the MIMO-Cube and MIMO-
and spatial-polarization correlation (calculated by the Finite Tetrahedron arrays. It was also shown that the performance
Element Method) on the mutual information of the system. of the compact MIMO-Cube antenna is superior to the
The mutual information results shown in Figure 7 are MIMO-Tetrahedron antenna since it has twice the number
International Journal of Antennas and Propagation 7
of independent channels which will result in more system [14] C. Waldschmidt, S. Schulteis, and W. Wiesbeck, “Complete RF
mutual information. Finally, we have also shown the effects system model for analysis of compact MIMO arrays,” IEEE
of the separation angle between HAPs on system perfor- Transactions on Vehicular Technology, vol. 53, no. 3, pp. 579–
mance, and determined that the optimal separation angle 586, 2004.
that maximizes the total mutual information of the system [15] J. W. Wallace and M. A. Jensen, “Mutual coupling in MIMO
wireless systems: a rigorous network theory analysis,” IEEE
is within a narrow sector of approximately 20 degrees for the
Transactions on Wireless Communications, vol. 3, no. 4, pp.
compact MIMO antenna arrays investigated in this paper.
1317–1325, 2004.
[16] Y. Yang, R. S. Blum, and S. Sfar, “Antenna selection for MIMO
References systems with closely spaced antennas,” EURASIP Journal on
Wireless Communications and Networking, vol. 2009, Article ID
[1] A. Mohammed, S. Arnon, D. Grace, M. Mondin, and R. 739828, 11 pages, 2009.
Miura, “Advanced communication techniques and appli- [17] T. Svantesson and A. Ranheim, “Mutual coupling effects on
cations for high-altitude platforms,” EURASIP Journal on the capacity of multielement antenna systems,” in Proceedings
Wireless Communications and Networking, vol. 2008, Article ID of IEEE International Conference on Acoustics, Speech and
934837, 3 pages, 2008. Signal Processing (ICASSP ’01), vol. 4, pp. 2485–2488, Salt Lake
[2] T. Hult, D. Grace, and A. Mohammed, “WCDMA uplink City, Utah, USA, May 2001.
interference assessment from multiple high altitude platform [18] D.-S. Shiu, G. J. Foschini, M. J. Gans, and J. M. Kahn, “Fading
configurations,” EURASIP Journal on Wireless Communica- correlation and its effect on the capacity of multielement
tions and Networking, vol. 2008, Article ID 182042, 7 pages, antenna systems,” IEEE Transactions on Communications, vol.
2008. 48, no. 3, pp. 502–513, 2000.
[3] A. Widiawan, R. Tafazolli, B. Evans, V. Milas, and P. Con- [19] J. P. Kermoal, L. Schumacher, K. I. Pedersen, P. E. Mogensen,
stantinou, “Coexistence of high altitude platform station, and F. Frederiksen, “A stochastic MIMO radio channel model
satellite, and terrestial systems for fixed and mobile services,” with experimental validation,” IEEE Journal on Selected Areas
in Proceedings of International Workshop on High Altitude in Communications, vol. 20, no. 6, pp. 1211–1226, 2002.
Platform Systems, Athens, Greece, September 2005. [20] S. K. Yong and J. S. Thompson, “Three-dimensional spatial
[4] V. F. Milas and P. Constantinou, “Interference environment fading correlation models for compact MIMO receivers,” IEEE
between high altitude platform networks (HAPN), geostation- Transactions on Wireless Communications, vol. 4, no. 6, pp.
ary (GEO) satellite and wireless terrestrial systems,” Wireless 2856–2868, 2005.
Personal Communications, vol. 32, no. 3-4, pp. 257–274, 2005.
[5] G. Aranti, A. Iera, and A. Molinaro, “The role of HAPs
in supporting multimedia broadcast and multicast services
in terrestrial-satellite integrated systems,” Wireless Personal
Communications, vol. 32, no. 3-4, pp. 195–213, 2005.
[6] G. Chen, D. Grace, and T. C. Tozer, “Performance of
multiple high altitude platforms using directive HAP and user
antennas,” Wireless Personal Communications, vol. 32, no. 3-4,
pp. 275–299, 2005.
[7] S. Nordebo, A. Mohammed, and J. Lundbäck, “On the use
of polarization channels in satellite communications,” in
Proceeding of Radio Vetenskap och Kommunication, Stockholm,
Sweden, June 2002.
[8] M. R. Andrews, P. P. Mitra, and R. Decarvalho, “Tripling the
capacity of wireless communications using electromagnetic
polarization,” Nature, vol. 409, no. 6818, pp. 316–318, 2001.
[9] J. Andersen and B. Getu, “The MIMO cube—a compact
MIMO antenna,” in Proceedings of the 5th IEEE International
Symposium on Wireless Personal Multimedia Communications,
vol. 1, pp. 112–114, 2002.
[10] R. Fitzpatrick, Lecture Notes: Advanced Classical Electromag-
netism, University of Texas, Austin, Tex, USA, 1996.
[11] M. A. Khalighi, J.-M. Brossier, G. Jourdain, and K. Raoof,
“Water filling capacity of Rayleigh MIMO channels,” in
Proceedings of the 12th IEEE International Symposium on
Personal, Indoor and Mobile Radio Communications (PIMRC
’01), vol. 1, pp. A155–A158, San Diego, Calif, USA, September
2001.
[12] G. J. Foschini and M. J. Gans, “On limits of wireless com-
munications in a fading environment when using multiple
antennas,” Wireless Personal Communications, vol. 6, no. 3, pp.
311–335, 1998.
[13] R. Janaswamy, “Effect of element mutual coupling on the
capacity of fixed length linear arrays,” IEEE Antennas and
Wireless Propagation Letters, vol. 1, pp. 157–160, 2002.
Hindawi Publishing Corporation
International Journal of Antennas and Propagation
Volume 2010, Article ID 756848, 8 pages
doi:10.1155/2010/756848
Research Article
Mutual Coupling Effects on Pattern Diversity Antennas for
MIMO Femtocells
Yue Gao, Shihua Wang, Oluyemi Falade, Xiadong Chen, Clive Parini, and Laurie Cuthbert
School of Electronic Engineering and Computer Science, Queen Mary, University of London, London E1 4NS, UK
Copyright © 2010 Yue Gao et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Diversity antennas play an important role in wireless communications. However, mutual coupling between multiple ports of a
diversity antenna has significant effects on wireless radio links and channel capacity. In this paper, dual-port pattern diversity
antennas for femtocell applications are proposed to cover GSM1800, UMTS, and WLAN frequency bands. The channel capacities
of the proposed antennas and two ideal dipoles with different mutual coupling levels are investigated in an indoor environment.
The relation between mutual coupling and channel capacity is observed through investigations of these antennas.
Radiator
160
Shorting post 97
Ground plane
Radiating patch Shorting posts
160
12.5 Probe to hybrid ring
Probe to hybrid ring
z 31.5
19.5
Hybrid ring
Port 4
Figure 1: Geometry of the proposed diversity antenna (Design 1): (a) top view, (b) side view, and (c) back view (Units in mm).
and easy fabrication [5, 6]. This paper presents a modified (FIT) for electromagnetic computation [7]. As discussed
circular microstrip L-probe wideband patch antenna to in [6], the performance of the antenna is dependent on a
achieve diversity operations for FAPs based on a previous number of design parameters, such as antenna height, radius
study [6]. The circular patch with a diameter of 97 mm is of the radiator, shorting posts, and hybrid ring. Based on
built on a square ground plane with a size of 160 mm by simulations using CST Microwave studio, a prototype of this
160 mm, as shown in Figure 1(a). The patch is supported by antenna has been fabricated in the Antenna Measurement
four shorting posts, as shown in Figure 1(b). It is fed by an Laboratory at Queen Mary University of London (QMUL).
L-shaped probe for the conical mode and a hybrid ring (e.g., The S-parameters of the antenna were measured by an
Port 4) for the broadside mode, as shown in Figure 1(c). The HP8720ES vector network analyzer and compared to the
hybrid ring is built on an FR4 PCB (Printed Circuit Board) simulated results in Figure 2. The impedance bandwidth
with a substrate thickness of 1.6 mm and permittivity of 3. at −10 dB for both measured and simulated S11 and S22
Ports 2 and 3 of the hybrid ring extend two L-probes to the operates from 1.7 GHz to 2.3 GHz, while at −6 dB it operates
radiator, and Port 1 is terminated by a 50Ω chip resistor. from 1.68 GHz to 2.5 GHz which covers GSM1800, UMTS,
The detailed dimensions and feeding structure are shown in LTE, and WLAN. It can be seen in Figure 2 that measured
Figure 1(b). S11 and S22 have the same number of resonances as those of
Computer Aided Design (CAD) was carried out by using the simulated values, and both measurement and simulation
the Computer Simulation Technology (CST) Microwave show similar amount of bandwidth at −10 dB. However, the
Studio package, which utilizes the Finite Integral Technique main resonance of the measured S11 and S22 are shifted from
International Journal of Antennas and Propagation 3
0 0
330 20 30
−5 10
0
−10 300 −10
60
Amplitude (dB)
−20
−15
−30
−20 270 90
−25
240 120
−30
−35
210 150
180
−40
1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 (a)
Frequency (GHz)
0
Measured S11 Measured S22 330 20 30
Simulated S11 Simulated S22 10
0
Figure 2: Simulated and measured return losses of the antenna 300 −10
60
(Design 1). −20
−30
270 90
0
240 120
−5
210 150
Amplitude (dB)
180
−10 (b)
0
−15 330 10 30
0
300 −10 60
−20
−20
−30
1.5 2 2.5 270 90
Frequency (GHz)
Figure 3: Isolation between the conical and broadside modes. 210 150
180
(c)
0 0
330 20 30 330 20 30
10 10
0 0
300 −10
60 300 −10
60
−20 −20
−30 −30
270 90 270 90
Figure 5: Measured (+) and simulated (−) cross-polarized radia- Figure 6: Measured (+) and simulated (−) copolarized radiation
tion patterns for broadside mode (S11) in XZ-plane at (a) 1.7 GHz, patterns for conical mode (S22) in XZ plane at (a) 1.7 GHz, (b)
(b) 1.85 GHz, and (c) 2.1 GHz. 1.85 GHz, and (c) 2.1 GHz.
plotted in Figures 6 and 7. The measured and simulated between multiple antennas should be minimised to maintain
patterns are almost identical. In comparison with the high efficiency of the multiple antennas [9, 10]. Further
broadside mode, the radiation patterns at conical mode have investigations were conducted using CST Microwave Studio
a similar trend, that is the copolarization is higher than to minimise the mutual coupling between the broadside and
cross-polarization. However, the dominant copolarizations conical modes. As discussed earlier, the length of the hybrid
have nulls at the upward direction (0 degree) and stronger ring has significant effects on the antenna performance,
radiations at 50 degree elevation angles as shown in Figure 6. and there is tradeoff between the bandwidth and isolation.
These patterns are complementary to those in the broadside Design 1 with the hybrid ring length of (3/2) λ0 achieved a
mode in Figure 4, which indicates good pattern diversity wider bandwidth but with a 12 dB isolation; by tuning the
characteristics. length of the hybrid ring to 2 λ0 , an isolation better than
In the case of multiantenna applications such as a diver- 30 dB over the frequency bandwidth is achieved, as shown
sity antenna array, it is required that the mutual coupling in Figure 8. However, the impedance bandwidth is narrower
International Journal of Antennas and Propagation 5
0 0
330 20 30
10 −5
0 −10
300 −10
60
Amplitude (dB)
−20 −15
−30 −20
270 90
−25
−30
240 120
−35
20.7
210 150 26
180 2.3
(b) Port 4
0 46.1 12
330 20 30
10 Figure 9: The geometry of the hybrid ring for better isolation in
0 Design 2.
300 −10
60
−20
−30
270 90
Room A
Rx2
16 metres
Rx1
FAP
RoomB
Rx3
Rx4 45 metres
X
Receivers Corridor
FAP Offices and labs
Figure 10: Floor plan of the second floor of Electronic Engineering Building at QMUL.
1 1
0.9 0.9
0.8 0.8
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0 1 2 3 4 5 6 7
Channel capacity (bit/s/Hz) Channel capacity (bit/s/Hz)
(a) (a)
1 1
0.9 0.9
0.8 0.8
0.7 0.7
Prob. x < abscissa
Prob. x < abscissa
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 2
0 1 2 3 4 5 6 7 2 2.5 3 3.5 4 4.5 5
Channel capacity (bit/s/Hz) Channel capacity (bit/s/Hz)
Figure 11: CDFs for different receiver locations at Room A: (a) Rx1 Figure 12: CDF for different receiver locations at Room B: (a) Rx3
and (b) Rx2. and (b) Rx4.
International Journal of Antennas and Propagation 7