IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO.

5, MAY 2014 2767

A Coupled Resonator Decoupling Network

for Two-Element Compact Antenna Arrays
in Mobile Terminals
Luyu Zhao, Member, IEEE, Lap K. Yeung, Member, IEEE, and Ke-Li Wu, Fellow, IEEE

Abstract—A new concept for decoupling two coupled antenna el- and a receiver along with some complex signal processing
ements in a broad band using a coupled resonator decoupling net- techniques, MIMO technology enables the system to set up
work (CRDN) is proposed for the first time. A synthesis and design parallel data streams on a single channel, thereby increasing
theory of a CRDN is presented. Based on the admittance parame-
ters of a given antenna array, a set of required rational functions the channel capacity by several times [2]. Therefore, wideband
and, consequently, the coupling matrix for a second-order decou- transmitting and receiving multiple antenna arrays have become
pling network is obtained analytically. To prove the concept, two indispensable in an advanced wireless communication system.
prototypes using microstrip resonators are designed and experi- When a multiple antenna system is implemented in a mobile
mentally studied. Measurement results have demonstrated that an terminal, the array elements have to be contained in a com-
isolation improvement of more than 10 dB can be achieved within
more than 15% bandwidth in both examples. The benefits of using
pact volume, which results in large pattern/spatial correlation
a CRDN for different levels of isolation in a MIMO terminal are in- and strong mutual coupling between the elements. The channel
vestigated through experiments and simulations. The results have capacity decreases as the correlation increases. Additionally, if
shown that, as compared to the existing decoupling scheme using mutual coupling is strong, a large portion of the power fed into
a lumped element, the proposed CRDN scheme can significantly one port will be coupled to the other port rather than radiating
increase the radiation efficiency, reduce the correlation, improve
to free space, and thus reducing the signal-to-noise ratio, the ra-
the channel capacity, and above all enhance the throughput of a
MIMO terminal. The technique is general and can be applied to diation efficiency and eventually deteriorating the channel ca-
both symmetric and asymmetric arrays. pacity. All of these degradations diminish the benefits of a mul-
Index Terms—Compact antenna array, decoupling network,
tiple antenna system. Therefore, developing an effective decou-
multiple-input multiple-output (MIMO), mutual coupling, net- pling technology has become a focused research area in recent
work synthesis. years.
Various decoupling schemes have been proposed, which can
be divided into four major categories:
I. INTRODUCTION 1) Eigen-Mode Decomposition Scheme: Its principle is to

W ITH THE increasing demands on higher data rate, diagonalize the scattering matrix of a compact array using 90
greater spectrum efficiency, larger average throughput and/or 180 hybrids. The scheme can be applied to coupled
and shorter latency by wireless systems, many advanced monopoles [3], loop antennas [4], and mobile phone antennas
technologies have been developed to confront the chal- that use the chassis as the main radiator [5]. A systematic design
lenges. Among them, orthogonal frequency division multiplex procedure exists and can be applied to any symmetric array [6].
(OFDM) and multiple-input multiple-output (MIMO) are two Moreover, by making use of the orthogonal modes, mode-based
important enabling technologies that have received tremendous beam steering can be achieved [7]. These eigenmodes, however,
attention in both industry and academic community. OFDM have significantly different matching bandwidth and efficiency.
uses multiple sub-carriers that are multiplexed in a wide fre- In fact, some of the high modes may have an impractically
quency spectrum, whereas MIMO provides a way of utilizing narrow bandwidth.
the multiple signal paths that exist between a transmitter and a 2) The Inserted Component Scheme: In the 1970’s, Andersen
receiver to significantly improve the data throughput on a given [8] proposed a necessary condition for the occurrence of no cou-
radio channel [1]. By using multiple antennas at a transmitter pling between antennas and proved the concept by inserting a
section of transmission-line between the coupled antenna ports.
The limitation is that the antenna spacing has to be fixed to cer-
Manuscript received September 09, 2013; revised December 25, 2013; ac-
cepted February 07, 2014. Date of publication February 26, 2014; date of current
tain values. A lumped element connected in shunt with the cou-
version May 01, 2014. This work was supported in part by University Grants pled antennas can also achieve certain level of isolation [9],
Committee of Hong Kong under Grant AoE/P-04/08 and also in part by Devel- [10]. Such decoupling schemes overcome some major short-
opment and Reform Commission of Shenzhen Municipality under Grant Shen
Fa Gai (2013) 1673.
comings of [8] and can be considered as a decoupling network
The authors are with the Department of Electronic Engineering, The Chinese of zeroth-order with relatively narrow decoupling bandwidth,
University of Hong Kong, Shatin, NT, Hong Kong (e-mail: [email protected]. which causes high sensitivity to the surroundings of the an-
edu.hk; [email protected]; [email protected]).
Color versions of one or more of the figures in this paper are available online
tennas in close proximity such as human hands. Recently, a neu-
at http://ieeexplore.ieee.org. tralization line technique was proposed for decoupling two an-
Digital Object Identifier 10.1109/TAP.2014.2308547 tennas [11]–[13]. The line not only creates a second path of cur-

0018-926X © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
2768 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 5, MAY 2014

rent flow but also perturbs the radiated fields. Existing designs
are carried out intuitively and are antenna dependent.
3) Artificial Structure Decoupling Scheme: This scheme
involves the use of sub-wavelength EM structures such as
mushroom-like electromagnetic band gap (EBG) structures
[14], defected ground structures (DGS) [15], and magnetic
metamaterials [16]. These structures provide bandstop filtering
characteristics at certain extent. A common issue for these
structures is that they are all ad hoc to particular antennas
and their footprints are large. Another similar narrow band
approach is to use a parasitic scatter/radiator [17].
4) Coupled Resonator Decoupling Scheme: A preliminary
result of decoupling a pair of coupled elements using a second-
order coupled resonator network was reported in 2012 by the au- Fig. 1. Network representation of two coupled antennas in shunt with a CRDN.
thors [18], in which not only the sequential but also a cross cou-
plings are introduced for a broadband decoupling performance.
Nevertheless, a thorough investigation both theoretically and is connected in parallel to the pair of coupled antennas. The
experimentally is needed. two matched ports of the decoupled system are terminated by
This paper systematically presents the general concept of a two unit loads. Assuming the CRDN is lossless, the entries in
coupled resonator decoupling network (CRDN). Compared to its admittance matrix are all purely imaginary. Obviously,
the existing decoupling techniques, the proposed decoupling the total admittance is the sum of the two individual admittance
network possess the following unique features: 1) it provides matrices as
a higher-order and wideband decoupling solution; 2) it offers
a tradeoff between decoupling bandwidth and degree of iso- (1)
lation; 3) it is relatively antenna independent and can be real-
ized using a wide selection of manufacturing technologies; 4) it Notice that the overall network is reciprocal but not neces-
can be applied to multiple antenna problems [19]; and 5) it can sarily symmetric. It can be easily obtained from an S-to-Y trans-
be designed using existing filter design theory and realization formation that the two ports will be well isolated within a given
techniques. frequency range if
The synthesis and design theory of a general CRDN starts
from the decoupling and matching conditions of admittance pa- (2)
rameters. The detailed procedures for synthesizing the required
where is the radian frequency. Since is imaginary, (2)
polynomials that lead to the coupling coefficients of a CRDN
will be elaborated. For illustration, the design processes for implies the following two decoupling conditions:
two practical compact antenna arrays are presented. Although (3a)
the CRDNs used in these two examples are implemented by
microstrip resonators, they can also be implemented by various and
other means, such as low temperature co-fired ceramics (LTCC)
and silicon-based integrated passive device (IPD) technolo- (3b)
gies. The working mechanism of the CRDN is revealed by
Having had (3) satisfied, it is clear that
comparing simulated current distributions of a decoupled array
and its coupled counterpart. The antenna dependency issue is
(4)
also discussed in detail. To give a full picture of the CRDN,
several important figures of merits for mobile terminals, such as
isolation, radiation efficiency, envelope correlation coefficient from which the matching conditions can be obtained as
(ECC), channel capacity, and throughput are investigated. A
(5a)
decoupled array using a lumped element network [9] is also
studied for comparison purpose. The measurement results (5b)
demonstrated the promising potentials of the proposed tech-
nique for MIMO and other multiple antenna systems. B. Admittance Transformation
Supposing that the coupled antenna array is matched to a real
II. SYNTHESIS OF A CRDN reference load , condition (3a) can be realized at the center
frequency by inserting a piece of transmission line with elec-
A. Decoupling and Matching Conditions trical length and characteristic impedance at each antenna
It is assumed that a pair of coupled antennas is represented port. Assuming the of the original antenna array to
by a 2-by-2 admittance matrix with complex entries. For be a complex number in general, then [9]
the decoupling configuration shown in Fig. 1, the decoupling
network consisting of a series of coupled resonators (i.e. CRDN) (6)
ZHAO et al.: COUPLED RESONATOR DECOUPLING NETWORK FOR TWO-ELEMENT COMPACT ANTENNA ARRAYS 2769

C. Synthesis of Decoupling Network By substituting (8) and (11) into (3b) and (5b), parameters
, , , , and can be determined analytically and the
The concept of coupling matrix, which is developed for de-
transversal coupling matrix for the decoupling network
signing bandpass filters, can be applied to describe a CRDN
can then be synthesized by following the procedure given in
as shown in Fig. 1. The design process of this network starts
[20]. To obtain the coupling matrix for a folded coupling struc-
with synthesizing the rational functions for the admittance pa-
ture, a matrix similarity transformation with pivot (2, 3) can be
rameters using the requirements imposed by (3) and (5). For a
applied. The transformed coupling matrix can be analytically
second-order CRDN without finite transmission zeroes, the mu-
expressed by
tual admittance can be expressed as

(7)

where , , and are real constants, and is the

complex frequency variable defined in the lowpass frequency
domain. Without loss of generality, it is assumed that . It (12)
is worth mentioning that for the lowpass prototype circuit (LPC)
of a coupled resonator network, the rational function is where , , and are the coupling coefficients be-
not necessarily a quotient of even to odd or odd to even poly- tween source to resonator 1, resonator 2 to load, and source to
nomials. This is because frequency invariant reactances (FIRs) load, respectively. Entries and are the self-couplings
and an extra inverter element are introduced in the second-order of resonator 1 and resonator 2, where that is
LPC [20]. It can be shown that with appropriate approximation proportional to the frequency shift of the resonator w.r.t.
in converting the rational function to a network realization, the the center frequency .
resultant network will be a quotient of even to odd or odd to even For a symmetrical decoupling network, and
polynomials. Furthermore, may not be positive real [21]. , the coupling matrix can be simplified to
The partial fractional expansion of (7) can be re-expressed as

(8)
(13)
where residues and . It will be shown later that parameter needs to be as large
Notice that the two residues of must be real but with as the manufacturing process allows. The sign of must be
opposite signs. In general, the residues of and opposite to that of the imaginary part of . It can be seen from
can be obtained through [20] (12) and (13) that to accommodate different couplings of cou-
pled antennas, only input/output couplings need to be adjusted.
(9)
D. Design of Decoupling Network
Based on the residues of , a proper sign in (9) can be In order to determine parameters , , , , and and ob-
chosen. For a realizable passive network, and tain the coupling coefficients, the admittance parameters of the
must be positive real [21]. Then to find the rational functions coupled antennas are transformed to the lowpass frequency do-
and from (9), two positive scaling coefficients main with reference to unit terminations and designated frac-
and need to be introduced so that tional bandwidth (FBW) by

(14)

where is the center frequency in the bandpass domain.

To illustrate the design procedure, a compact antenna array
(10) working at 2.6 GHz as shown in Fig. 2(a) is considered. The
array is composed of one cone-shaped monopole and one mean-
The absolute signs in (10) ensure and to be pos- dered monopole. The center-to-center distance between the two
itive real. Therefore, the rational functions and elements is 16.55 mm and the edge-to-edge distance
can now be expressed as is 3.9 mm . The simulated and the measured -pa-
rameters are shown in Fig. 2(b).
It has been found by simulation and measurement that for
many types of practical coupled antenna pairs, such as coupled
(11a) PIFAs, monopoles, or patches, the imaginary part of is al-
most constant over a frequency range in the vicinity of reso-
nance. In order to satisfy (3b), is desired to be relatively
(11b) constant within the band of interest. In other words, because the
2770 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 5, MAY 2014

that satisfies (3b) should be opposite to ,

that is:

(15)
In a general case where the imaginary part of could not
be approximated by a constant over a certain frequency band,
will result in a better decoupling bandwidth. An
alternative choice is to introduce two finite transmission zeros in
the CRDN such that the mutual coupling can be cancelled
out in a broadband sense [18].

E. Design of Matching Network

To achieve a broadband impedance matching,
should be the same as
provided that approaches to zero in a
broadband sense. From Fig. 2(d), it is observed that, in general,
and can be different in terms of their slopes
and zero locations. Therefore, to satisfy (5b) at the center
frequency for both ports, the following relations are
obtained by setting in (11a) and (11b):

(16a)

(16b)

Since and , it is obvious that

(17)

which suggests that the center frequency in (14) should be

chosen in the way that and have different signs.
Although this new is different from the original center fre-
quency of the antennas, two extra matching circuits can still be
added to each port to match the decoupled antennas at the orig-
inal center frequency. In this example, if is chosen
to be 2.55 GHz, (16a) and (16b) can be satisfied simultaneously
as shown in Fig. 2(d).
For a given antenna configuration, the slope of is an in-
herent characteristic depending on the value of the antenna.
In order to optimally fulfill the impedance matching condition
(5b), one needs to minimize the slope of at the resonant fre-
quency and to use this slope information to represent the slope
of within the decoupling frequency band. Closely exam-
ining the slope, one can find that it is inversely proportional to
. Therefore, a large results in a small slope of within
a frequency band and consequently good impedance matching
can be obtained providing that a proper is chosen to achieve
a good isolation according to (15). The same conclusion can be
Fig. 2. (a) Asymmetric two-element compact array. (b) Simulated and mea-
drawn for . Having had chosen as large as the network re-
sured -parameter of the array. (c) Simulated of the array and a CRDN. alization allows, parameters , , and can be determined by
(d) Simulated normalized and of the array and a CRDN. (15) and (16). Consequently, the coupling coefficients in (12)
can be obtained. It is also observed in Fig. 2(d) that when a
CRDN is added to the original coupled antenna pair, in addition
slope of at is proportional to according to to an improved isolation, its radiation efficiency increases and,
(7), it is necessary to have . One can then use consequently, the matching bandwidth decreases. Like other de-
the value of at to approximate the value of within coupling techniques, this decoupling technique trades matching
the decoupling band. As illustrated by Fig. 2(c), the optimum bandwidth for a better isolation and higher efficiency [22].
ZHAO et al.: COUPLED RESONATOR DECOUPLING NETWORK FOR TWO-ELEMENT COMPACT ANTENNA ARRAYS 2771

TABLE I
COUPLING COEFFICIENTS OF DESIGNED AND REALIZED DECOUPLING
NETWORKS FOR EXAMPLE 1 AND EXAMPLE 2 (DESIGNED )

found to be 0.7954 and 0.8895, respectively. Finally, with , ,

, and determined, the coupling coefficients for the decoupling
network are obtained from (12). The designed lowpass coupling
Fig. 3. Simulated and with different values.
matrix is listed in Table I with designed fractional bandwidth
.
The importance of choosing a large for broadband matching The coupling matrix is realized using two short-circuited and
is studied parametrically and is presented in Fig. 3, where the folded quarter-wavelength resonators with edge inter-resonator
circuit in Fig. 1 is used with the coupled antennas shown in coupling. The input/output couplings are realized by the tapped-
Fig. 2. The decoupling bandwidths with dB and line configuration. The resonant frequency of each resonator is
dB for , , and are 112, determined by the self-coupling. With appropriate layout de-
244, and 317 MHz, respectively. It is seen from (12) and (13) sign, the physical dimensions are tuned by comparing the ex-
that the maximum is stipulated by the maximum achievable tracted coupling coefficients to the synthesized ones until the
inter-resonator coupling when realizing a CRDN. There- overall responses match the desired one. Details on how to de-
fore, to obtain an optimum matching performance, it is advis- sign a microstrip resonator and required couplings can be found
able to maximize when designing the decoupling network. in many text books [24].
For different realization technologies, such as microstrip and The overall physical layout of the decoupled antenna array is
multilayer LTCC, the maximum realizable can be quite dif- shown in Fig. 4(a). Notice that two additional matching stubs
ferent. It is interesting to observe from Fig. 3 that a tradeoff be- are used to broaden the matching bandwidth. Fig. 4(b) shows
tween the isolation level and the decoupling bandwidth can be the simulated and measured magnitudes of -parameters of the
made by choosing an appropriate value of parameter . decoupled array. The realized coupling matrix is extracted using
the circuit model specified in Fig. 1 and is given in Table I
III. DESIGN EXAMPLES as well for comparison. It is seen from Fig. 4(b) that the de-
coupling bandwidth with dB is about 23% and
Two examples are presented in this section. In both examples,
the impedance matching bandwidth with dB and
printed monopole antenna elements are fabricated together with
dB is about 9.2%. For the commonly accepted 6
a CRDN on a double-sided FR4 substrate with relative dielectric
dB return loss requirement for mobile terminals, the matching
constant of 4.3 and thickness of 1.6 mm. The center frequency is
bandwidth of this decoupled array is about 15%.
chosen to be 2.6 GHz (LTE Band 7). Full-wave electromagnetic
To explain the working mechanism of this CRDN, EM sim-
simulations of these examples are performed by Agilent EMPro
ulated current distributions on the decoupled and the coupled
[23].
arrays with excitation at port 1 are plotted in Fig. 5(a) and (b),
A. Example 1: A CRDN for an Asymmetric Array respectively. It can be observed from Fig. 5(a) that the current
induced by the antenna-to-antenna coupling at the terminated
The first example is to synthesize and design a CRDN for
port (port 2) is well canceled by the current introduced through
an asymmetric two-element array as shown in Fig. 2(a). The
the CRDN.
simulated and measured -parameter of the original coupled
but matched array is given in Fig. 2(b), in which an isolation
B. Example 2: A CRDN for a Symmetric Array
of poorer than 8 dB at 2.6 GHz is observed.
A section of transmission line with its electric length deter- The decoupling theory is applied to a symmetric array in this
mined by (6) is first introduced at each antenna port so that example, in which a pair of symmetric cone-shaped monopole
at the center frequency. Fig. 2(c) shows the sim- antennas is considered. The ground size and substrate thickness
ulated in lowpass domain. Due to the in-house fabrication are the same as those of example 1. The edge to edge spacing
constraints, the minimum line width and spacing are limited to between the two elements is 9.8 mm .
0.2 mm. In this example is designed to be 3. It can be seen For the symmetric array, a symmetric decoupling network
from Fig. 2(c) that the imaginary part of is nearly a con- can be synthesized and designed according to (13). The synthe-
stant of 0.51 over a wide frequency band. According to (15), sized lowpass coupling matrix is listed in Table I. According
is found to be 4.59. to the coupling matrix, the physical dimensions of the res-
As shown in Fig. 2(d), the calculated onators are determined and depicted in Fig. 6(a). Two extra
and . Therefore, using (16), and are matching stubs are added. Simulation and measurement results
2772 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 5, MAY 2014

Fig. 4. (a) Layout of the decoupled asymmetric antenna array. (b) Simulated
and measured responses of the decoupled array.
Fig. 6. (a) Layout of the decoupled symmetric antenna array. (b) Simulated
and measured responses of the decoupled array.

1) In general, the isolation increases as the element spacing

increases, however, the coupled arrays with the small
ground (SG) are with better isolation as compared to those
with the large ground (LG).
2) Both the real and imaginary parts of the mutual admittance
are affected by the form factors that have to be considered
Fig. 5. (a) Current distributions of two antennas with a CRDN and (b) current
distribution of the same antennas without a CRDN, when port 1 is excited. in a design of a CRDN.
3) The larger the mutual coupling, the steeper the slope of the
real part of mutual admittance versus frequency and the
larger the imaginary part of the mutual admittance.
are shown in Fig. 6(b) with the corresponding extracted cou-
As revealed by (12) and (13), the same CRDN can be used to
pling matrix listed in Table I. The decoupling bandwidth with
decouple a coupled array with different array form factors by
dB is about 15% and the matching bandwidth
keeping the same self-couplings for the resonators and the same
with dB is about 19%.
inter-resonator coupling . Only input/output couplings,
i.e., and , need to be adjusted to accommodate the
C. Antenna Dependency and “One-Fit-All” Feature
change in the form factors. For illustration, the circuit models
In practice, the array form factors, such as inter-element of four CRDNs are designed to decouple the arrays with el-
spacing and ground plane size, affect the mutual admittance. ement spacing (S) of 8 and 22 mm, and the ground sizes of
Most of existing decoupling techniques require a complete SG and LG, respectively. The designed coupling coefficients
redesign if any of the form factors are changed. However, a are listed in Table II. It can be seen that only the Input/Output
useful and practical feature of the CRDN technique is that only couplings of a CRDN depend on the spacing and the ground
the input/output coupling coefficients and require to size. Therefore, a “one-fit-all” module can be provided by
be modified to accommodate the change. consolidating all the circuitry parts except the I/O couplings,
To demonstrate this feature, the properties of the mutual ad- which can be adjusted as external components. This feature
mittance of the symmetric array used in Example 2 are inten- makes the CRDN technique very attractive.
sively investigated by varying the spacing from 4 to 112
mm and setting ground plane sizes to mm (SG) and IV. PERFORMANCE COMPARISON IN PRACTICAL CONDITIONS
mm (LG) using EM simulations. Three important To have a thorough understanding of the performance of a
conclusions can be drawn. CRDN in a practical implementation, several important figures
ZHAO et al.: COUPLED RESONATOR DECOUPLING NETWORK FOR TWO-ELEMENT COMPACT ANTENNA ARRAYS 2773

TABLE II
COUPLING COEFFICIENTS OF CRDNS FOR ARRAYS 1 4

Fig. 7. (a) Hand phantom arrangement and measurements setup. (b) Four pairs
of antennas under investigation.

of merit such as isolation, radiation patterns and radiation effi-

ciencies, ECC and throughput are investigated both in free space
and in the presence of a hand phantom.
The SPEAG hand phantom for mono-block phones SHO
V2RB/LB is used1 (see Fig. 7(a)). Four pairs of antennas are
compared, including the symmetrical coupled cone-shaped
antennas (Pair A), the same antennas decoupled by a lumped
element designed by the procedure given in [9] (Pair B), the
same antennas decoupled by a CRDN with wideband isolation
of 20 dB (Pair C, CRDN 1) and the same antennas decoupled
by a CRDN with isolation of 30 dB (Pair D, CRDN 2). All the
prototypes are shown in Fig. 7(b). The radiation characteristics
of these four pairs of antennas with and without hand phantom
are measured using the in-house SATIMO SG128 spherical
near-field scanner2 in an ISO17025 accredited laboratory.

A. Matching and Isolation

The reflection and isolation coefficients of antenna pairs A,
B, C, D with and without the presence of the hand phantom
are examined and the results are displayed in Fig. 8. Since the
hand phantom acts as a lossy dielectric medium in close prox-
imity of the antennas, the resonant frequency of all the antenna Fig. 8. Measured return loss for (a) antenna pair A, (b) antenna pair B, (c) an-
tenna pair C, and (d) antenna pair D. (e) Measured isolation for antenna pairs
pairs shifts to a lower value. However, the impact of the hand A, B, C, and D.
phantom on the isolation for the four antenna pairs seems to be
1[Online]. Available: http://www.speag.com/ quite different. For antenna pairs B, C, and D, the isolation in-
2[Online]. Available: http://www.satimo.com/ creases in certain frequencies while it decreases in other
2774 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 5, MAY 2014

D. Channel Capacity and Throughput

The channel capacities for an ideal 2-by-2 MIMO system

using these four antenna arrays under different SNRs are calcu-
lated and are superposed in Fig. 12. The channel capacity calcu-
lation follows the method given in [27], which takes the antenna
array efficiency and correlation into account. The transmitter an-
tennas are assumed to be uncorrelated while all the arrays under
investigation are placed at the receiver end. A larger enhance-
ment of channel capacity is observed for cases having a hand
phantom. An averaged improvement of around 1 bps/Hz is ob-
served for Pair C and D as compared to Pair A.
Fig. 9. Total efficiencies of four antenna pairs. (The efficiencies are averaged According to COST Action 2100, throughput is suggested to
values of the two antennas for each pair.) be an important system level figure of merit [28]. However, to
assess the throughput of a MIMO system in a given multipath
environment is always challenging because it depends not
frequencies due to the detuning of resonance frequencies. only on the MIMO antenna array itself, but also on spatial
Nevertheless, both pairs C and D achieve 20 dB isolation with a and temporal characteristics of the radio channel as well as
frequency band that is as wide as more than three times of that
the space-time data processing algorithm. In this study, the
of pair B, with and without the presence of a hand phantom.
two-stage method is adopted, which calculates the throughput
using the passive antenna far-field pattern measurements and
B. Radiation Efficiency
a MIMO channel model with appropriate parameters. Specif-
The total efficiencies of the antenna pairs are shown in ically, the WINNER II channel model in Agilent’s W1715
Fig. 9. Under the free-space condition, the total efficiency for MIMO Channel Builder of SystemVue is used [29]. The model
the coupled array (Pair A) is no more than 73% from 2.4 to 2.7 is stochastic and geometry-based. The channel parameters for
GHz, whereas for the decoupled array (Pair B), the efficiency individual snapshots are based on the statistical distributions
is around 76% for the same band. For Pair C and Pair D extracted from the channel measurement. The channel scenario
using CRDNs, the typical efficiency has improved to 78% and B1 (Urban Micro-cell) is used in the throughput comparison of
83%, respectively. Further improvements in efficiency can be the four antenna pairs. The downlink transmitter is assumed to
expected if substrate with lower loss is used. use a two-element linear array with omni-directional elements
Due to the lossy nature of the hand phantom, the total effi- and half-wavelength spacing. The four antenna pairs under
ciencies for all the four pairs of arrays have a huge degradation investigation are placed horizontally at the receiver,
when the hand phantom is presented. Nevertheless, the Pair D facing the transmitter array in the broadside direction. The
still shows a significant efficiency improvement from 17% to patterns of the antenna pairs are imported from measured
26% as compared to Pair A. results at 2.6 GHz. Other parameters are set to be the default
values in the model. The fractional data throughputs, which
C. Envelope Correlation Coefficients (ECC) are normalized with respect to their corresponding maximum
achievable values, are superposed in Fig. 13.
The power patterns of the coupled array (Pair A) and one of
the decoupled arrays (Pair C) are shown in Fig. 10. In the mea-
surements, one of the antennas is excited while the other one is E. Summary of the Performance Comparison
terminated by a 50 load. Using the measured far-field vector
phasor radiation patterns, the ECC of the each antenna pairs can It can be concluded from the above performance evaluations
be calculated by the method introduced in [25]. It is known that and comparisons that the decoupled antenna arrays using dif-
a lower envelope correlation leads to a higher channel capacity ferent decoupling methods behave very differently with and
and a better diversity gain. For the decoupled antenna Pairs B, C, without human interferences. There are huge degradations in ef-
and D under the free space condition, an improvement in ECC ficiency, ECC, channel capacity and throughput in the presence
from around 0.05 to smaller than 0.01 is seen in Fig. 11. But of a hand phantom. However, despite the performance varia-
for the cases with the hand phantom, a much higher ECC is ob- tions, the arrays with a CRDN outperform the original coupled
tained for all the four pairs. This phenomenon is consistent with array and the one with a lumped decoupling element. Moreover,
the discovery in [4] and [26]. For Pair A, the ECC increases the array with the CRDN that provides an isolation of 30 dB has
to 0.17 0.27 within the band of interest. Although the ECCs the best performance among all the arrays under investigation.
for Pair B, C, and D are also increased, the ECC for Pair D is Therefore, it is worthwhile to design a compact array with a
below 0.025 from 2.4 to 2.7 GHz due to its wide band decou- high isolated level, especially for mobile terminals in a realistic
pling attribute. environment.
ZHAO et al.: COUPLED RESONATOR DECOUPLING NETWORK FOR TWO-ELEMENT COMPACT ANTENNA ARRAYS 2775

Fig. 10. Radiation patterns of the decoupled array-Pair C and the coupled array-Pair A in the plane at 2.6 GHz.

Fig. 13. Throughput fraction for the four antenna pairs without and with hand
phantom in WINNER II channel model B1.

Fig. 11. Calculated envelope correlation coefficients for the four antenna pairs.
of a pair of coupled antennas, the admittance functions of a cou-
pled resonator decoupling network can be synthesized, which
in turn leads to a 4-by-4 coupling matrix for a realizable circuit
model. Several design examples using microstrip resonators on
an FR4 substrate are also presented. More compact decoupling
networks can be fabricated using advanced manufacturing
technologies.
Simulated and experimental results have demonstrated that
both port decoupling and matching can be achieved simultane-
ously over a relatively wide bandwidth by using a CRDN. It has
also been shown through the measured radiation patterns that
the decoupled antennas have a significant improvement in ra-
diation efficiency, correlation coefficient, channel capacity, and
above all throughput, both in free space and in the presence of
a hand phantom. The examples have demonstrated a great po-
Fig. 12. Channel capacity for the four antenna arrays together with an ideal tential of the proposed CRDN concept for mobile terminals that
2-by-2 MIMO array.
use MIMO or diversity antenna technology.

REFERENCES
V. CONCLUSION
[1] G. L. Stuber, J. R. Barry, S. W. McLaughlin, Y. Li, M. A. Ingram, and
This paper proposed a new concept for decoupling two T. G. Pratt, “Broadband MIMO-OFDM wireless communications,”
tightly coupled antennas using a coupled resonator network. Proc. IEEE, vol. 92, no. 2, pp. 271–294, Feb. 2004.
[2] M. A. Jensen and J. W. Wallace, “A review of antennas and propagation
The synthesis and design theory for the CRDN are also pre- for MIMO wireless communications,” IEEE Trans. Antennas Propag.,
sented in detail. Starting from the measured admittance matrix vol. 52, no. 11, pp. 2810–2824, Nov. 2004.
2776 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 62, NO. 5, MAY 2014

[3] J. C. Coetzee and Y. Yu, “Port decoupling for small arrays by means of [27] J. X. Yun and R. G. Vaughan, “Multiple element antenna efficiency and
an eigenmode feed network,” IEEE Trans. Antennas Propag., vol. 56, impact on diversity and capacity,” IEEE Trans. Antennas Propag., vol.
no. 6, pp. 1587–1593, Jun. 2008. 60, no. 2, pt. I, pp. 529–539, Feb. 2012.
[4] S. Zuo, Y.-Z. Yin, Y. Zhang, W.-J. Wu, and J.-J. Xie, “Eigenmode de- [28] Pervasive Mobile and Ambient Wireless Communications: COST Ac-
coupling for MIMO loop-antenna based on 180 coupler,” Progr. Elec- tion 2100, R. Verdone and A. Zanella, Eds. New York, NY, USA:
tromagn. Res. Lett., vol. 26, pp. 11–20, 2011. Springer, 2012.
[5] S. K. Chaudhury, H. J. Chaloupka, and A. Ziroff, “Multiport an- [29] Agilent Technologies, Inc., Santa Clara, CA, USA, “SystemVue Elec-
tenna Systems for MIMO and Diversity,” presented at the EUCAP, tronic System-Level (ESL) Design Software,” Version 2013.01.
Barcelona, Spain, Apr. 2010.
[6] C. Volmer, J. Weber, R. Stephan, K. Blau, and M. A. Hein, “An eigen-
analysis of compact antenna arrays and its application to port decou- Luyu Zhao (S’09–M’14) was born in Xi’an, China,
pling,” IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 360–370, in 1984. He received the B.Eng. degree from Xidian
Feb. 2008. University, Xi’an, China, in 2007, and is currently
[7] L. K. Yeung and Y. E. Wang, “Mode-based beamforming arrays for working toward the Ph.D. degree from The Chinese
miniaturized platforms,” IEEE Trans. Microw. Theory Tech., vol. 57, University of Hong Kong, Shatin, Hong Kong.
no. 1, pp. 45–52, Jan. 2009. From 2007 to 2009, he was with the National Key
[8] J. B. Andersen and H. H. Rasmussen, “Decoupling and descattering Laboratory of Antennas and Microwave Technology,
networks for antennas,” IEEE Trans. Antennas Propag., vol. 24, no. Xidian University, as a Research Assistant. His cur-
AP-6, pp. 841–846, Nov. 1976. rent research interests include design and application
[9] S. Chang, Y.-S. Wang, and S.-J. Chung, “A decoupling technique for of multiple antennas for next generation mobile com-
increasing the port isolation between strongly coupled antennas,” IEEE munication systems and passive RF and microwave
Trans. Antennas Propag., vol. 56, no. 12, pp. 3650–3658, Dec. 2008. components, circuits and systems.
[10] C.-Y. Lui, Y.-S. Wang, and S.-J. Chung, “Two nearby dual-band an-
tennas with high port isolation,” presented at the IEEE Int. Symp. An-
tennas Propag., San Diego, CA, USA, Jul. 2008.
[11] A. Diallo, C. Luxey, P. L. Thuc, R. Staraj, and G. Kossiavas, “Study
and reduction of the mutual coupling between two mobile phone PIFAs Lap K. Yeung (S’00–M’02) received the B.Eng.
operating in the DCS1800 and UMTS bands,” IEEE Trans. Antennas degree in electrical and information engineering
Propag., vol. 54, no. 11, pp. 3063–3073, Nov. 2006. from University of Sydney, Sydney, Australia, in
[12] S.-W. Su, C.-T. Lee, and F.-S. Chang, “Printed MIMO-antenna system 1998, the M.Eng. degree in electronic engineering
using neutralization-line technique for wireless USB-dongle applica- from The Chinese University of Hong Kong, Shatin,
tions,” IEEE Trans. Antennas Propag., vol. 60, no. 2, pp. 456–463, Hong Kong, in 2002, and the Ph.D. degree in
Feb. 2012. electrical engineering from University of California
[13] C. Luxey, “Design of multi-antenna systems for UTMS mobile at Los Angeles, Los Angeles, CA, USA, in 2009.
phones,” in Proc. Loughborough Antennas Propag. Conf., Nov. 2009, During 1999, he was with the Commonwealth
pp. 57–64. Scientific and Industrial Research Organization
[14] F. Yang and Y. R. Samii, “Microstrip antennas integrated with electro- (CSIRO), Sydney, Australia, where he was a re-
magnetic band-gap EBG structures: A low mutual coupling design for search engineer involved in the numerical modeling of different antenna
array applications,” IEEE Trans. Antennas Propag., vol. 51, no. 10, pp. structures. From 2003 to 2006, he was with The Chinese University of Hong
2936–2946, Oct. 2003. Kong, where he is involved in various LTCC multichip module (MCM)
[15] C. Y. Chiu, C. H. Cheng, R. D. Murch, and C. R. Rowell, “Reduction designs and the development of numerical algorithms for analyzing multilayer
of mutual coupling between closely-packed antenna element,” IEEE embedded RF modules.
Trans. Antennas Propag., vol. 55, no. 6, pp. 1732–1738, Jun. 2007.
[16] M. M. Bait-Suwailam, M. S. Boybay, and O. M. Ramahi, “Electro-
magnetic coupling reduction in high-profile monopole antennas using
single-negative magnetic metamaterials for MIMO applications,” Ke-Li Wu (M’90–SM’96–F’11) received the B.S.
IEEE Trans. Antennas Propag., vol. 58, no. 9, pp. 2894–2902, Sep. and M.Eng. degrees from the Nanjing University of
2010. Science and Technology, Nanjing, China, in 1982
[17] B. K. Lau and J. B. Andersen, “Simple and efficient decoupling of com- and 1985, respectively, and the Ph.D. degree from
pact arrays with parasitic scatterers,” IEEE Trans. Antennas Propag., Laval University, Quebec, QC, Canada, in 1989.
vol. 60, no. 2, pp. 464–472, Feb. 2012. From 1989 to 1993, he was with the Communi-
[18] L. Zhao, L. K. Yeung, and K.-L. Wu, “A novel second-order decoupling cations Research Laboratory, McMaster University,
network for two-element compact antenna arrays,” in Proc. Asia-Pa- as a Research Engineer and a Group Manager. In
cific Microwave Conf., Dec. 2012, pp. 1172–1174. March 1993, he joined the Corporate R&D Division,
[19] L. Zhao and K.-L. Wu, “A broadband coupled resonator decoupling COM DEV International, the largest Canadian space
network for a three-element compact array,” presented at the IEEE equipment manufacturer, where he was a Principal
MTT-S Int. Microw. Symp., Seattle, WA, USA, Jun. 2013. Member of Technical Staff. Since October 1999, he has been with The Chinese
[20] R. J. Cameron, C. M. Kudsia, and R. R. Mansour, Microwave Filters University of Hong Kong, Hong Kong, where he is a Professor and the
for Communication systems. New York, NY, USA: Wiley, 2007, ch. Director of the Radiofrequency Radiation Research Laboratory (R3L). He has
6–8. authored or coauthored numerous publications in the areas of EM modeling
[21] M. E. van Valkenburg, Network Analysis. Englewood Cliffs, NJ, and microwave passive components, microwave filter and antenna engineering.
USA: Prentice-Hall, 1955. His current research interests include partial element equivalent circuit (PEEC)
[22] G. Shaker, S. Safavi-Naeini, and N. Sangary, “Q-Bandwidth relations and derived physically expressive circuit (DPEC) EM modeling of high speed
for the design of coupled multi-element antennas,” in Proc. IEEE An- circuits, RF and microwave passive circuits and systems, synthesis theory and
tennas Propag. Soc. Int. Symp., Jun. 1–5, 2009, pp. 1–4. practices of microwave filters, antennas for wireless terminals, LTCC-based
[23] Agilent Technologies, Inc., Santa Clara, CA, USA, “EMPro 3D EM multichip modules (MCMs), and RF identification (RFID) technologies. His
Simulation Software,” Version 2012.09. research group is the main workforce in various active RFID research and
[24] J.-S. Hong and M. J. Lancaster, Microstrip Filters for RF/Microwave applications in Hong Kong.
Applications, 2nd ed. New York, NY, USA: Wiley, 2011, ch. 7–10. Prof. Wu is a member of IEEE MTT-8 subcommittee (Filters and Passive
[25] R. G. Vaughan and J. B. Andersen, “Antenna diversity in mobile Components) and also serves as a TPC member for many prestigious inter-
communications,” IEEE Trans. Veh. Technol., vol. VT-36, no. 4, pp. national conferences including International Microwave Symposium. He was
147–172, Nov. 1987. an Associate Editor of IEEE TRANSACTIONS ON MICROWAVE THEORY AND
[26] B. Yanakiev, J. Ødum Nielsen, M. Christensen, and G. F. Pedersen, TECHNIQUES from 2006 to 2009. He was a recipient of the 1998 COM DEV
“On small terminal antenna correlation and impact on MIMO channel Achievement Award for the development of exact EM design software of
capacity,” IEEE Trans. Antennas Propag., vol. 60, no. 2, pp. 689–699, microwave filters and multiplexers and Asia Pacific Microwave Conference
Feb. 2012. Prize in 2008 and 2012.