122 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL.

6, 2007

Analysis of Mutual Coupling, Correlations, and

TARC in WiBro MIMO Array Antenna
Sung Ho Chae, Se-keun Oh, and Seong-Ook Park, Member, IEEE

Abstract—This paper presents the evaluation of multiple- tics, and enough wide bandwidth to cover 200 MHz, which is
input–multiple-output (MIMO) arrays by analyzing the mutual required for the WiBro system.
coupling, correlation coefficient, and total active reflection co- This paper investigates the mutual coupling, correlation, and
efficient. The meaning and analysis of the three parameters are
TARC in three different types of two-array antennas.
demonstrated in this paper. We show the effect of spatial and
pattern diversity on MIMO arrays. We also attempt to achieve
optimization of the array antenna’s allocation for wireless broad- II. THEORETICAL BACKGROUND
band MIMO applications by performance evaluation.
The MIMO system can be modeled as
Index Terms—Correlation coefficient, multiple-input–multiple-
output (MIMO), mutual coupling, radiation pattern, total active (1)
reflection coefficient (TARC).
where is an 1 vector, which represents the transmitted
signal by the antennas. Likewise, the received signal and ad-
I. INTRODUCTION ditive noise vector are represented by and vectors.
represents the path gain between transmitter and receiver,
which is induced by the MIMO channel.
The instananeous channel capacity of MIMO has been shown
I NFORMATION theory has shown that an upper limit exists
for the average spectral efficiency using a single transmitter
and single receiver [1]. Thus, multiple-input–multiple-output
to be [2]

(MIMO) antennas have received great attention for their ability (2)
to overcome the limits of single-input single-output channel
capacity [2]. or in terms of the eigenvalues of as
In order to study the performance of the MIMO antenna, some
parameters need to be considered. Mutual coupling is one of the
important factors because higher mutual coupling means lower (3)
antenna efficiency. The correlation coefficient between the two
antennas is another important parameter since it is associated
with the loss of spectral efficiency and degradation of perfor- where denotes the complex conjugate and
mance of a MIMO system [3]. Total active reflection coefficient .
(TARC) must also be considered. We use TARC rather than the Since is random, the capacity is randomly varying with .
traditional scattering matrix because the scattering matrix does In the correlated Rayleigh-fading case, the following covari-
not accurately characterize the radiating efficiency and band- ance structure is valid [5]:
width of an antenna array [4].
The proposed antenna arrays are composed of printed in- (4)
verted-F antenna (PIFA) with U-shaped slot. The antenna is
designed for wireless broadband (WiBro) application in the where represents transmit correlation between transmit an-
United States and Canada (2.5–2.7 GHz), which is based on the tenna and . Likewise, represents receive correlation be-
IEEE 802.16e technical standard. WiBro is a mobile broadband tween receive antenna and . These two correlations are in-
access technology being developed and deployed in South dependent of each other. Thus, can be factorized in form as
Korea. It provides not only large coverage and fast mobility follows:
but also data rates as high as wireless local-area network. The
proposed antenna has low profiles, good radiation characteris- (5)

where the entries of are independent identically distributed

Manuscript received October 3, 2006; revised February 4, 2007. This work (i.i.d.). From (5), if we assume that only receive antennas are
was supported by Intelligent Radio Engineering, Korea Science and Engi-
neering Foundation, Korea, under Contract R11-2006-029-02004-0 and the
correlated, the upper bound to ergordic capacity is expressed as
Electronics and Telecommunications Research Institute, Korea, under Contract follows [6]:
2006EG0700.
The authors are with the School of Engineering, Information and Commu-
nications University, Daejeon 305-714, Korea (e-mail: slapguitar@icu.ac.kr;
sopark@icu.ac.kr).
Digital Object Identifier 10.1109/LAWP.2007.893109 (6)
1536-1225/$25.00 © 2007 IEEE
CHAE et al.: MUTUAL COUPLING, CORRELATIONS, AND TARC IN WIBRO MIMO ARRAY ANTENNA 123

correlation coefficient can be calculated by S-parameters [8].

The antennas are assumed to be lossless

(11)

The mutual coupling between the antennas can be measured and

simultated from parameters.
TARC is defined as the ratio of the square root of total re-
flected power divided by the square root of total incident power
[4]. The TARC at the port antenna can be described as

(12)

where is the incident signal and is the reflected signal.

Fig. 1. Geometries and dimensions of the PIFA: (a) top view and (b) side view. In the case of 2 2 antenna arrays, the scattering matrix can
be described as

(13)

We assume the reflected signal will be unity magnitude but

randomly phased with i.i.d. Gaussian random variable because
MIMO channels are assumed as Gaussian and multipath spread
in the propagation channel. Since sum or subtraction of inde-
pendent Gaussian random variables is also Gaussian, TARC is
described as follows:

(14)
Fig. 2. Measured and simulated return losses of single PIFA.
Using (14), TARC for the 2 2 antenna array can be directly
calculated from the scattering matrix.
where
and is a minor determinant of , which III. TWO ANTENNA ARRAYS
implies a determinant of the matrix lying in the This paper proposes a novel PIFA with U-slot. The antenna
rows and columns of [7]. volume measures 15 15 4 mm . The structure of the an-
For the case of (2 2 case), (6) becomes tenna is described in Fig. 1. The single antenna with ground size
of 45 85 mm has resonance at 2.6 GHz covering the band-
(7) width of 200 MHz in voltage standing-wave ratio 2:1. The pro-
posed antenna is an electrically small antenna , so it
is suitable for MIMO implementation in handsets. CST Micro
and for high SNR case Wave Studio is employed to simulate the scattering parameters
and radiation patterns. The comparison results between the mea-
(8) sured and simulated return losses are shown in Fig. 2. The mea-
sured and simulated copolarized and cross-polarized radiation
where is a 2 2 correlation matrix. patterns at 2.6 GHz are plotted in Fig. 3.
As shown in (7) and (8), the capacity loss increases due to In order to analyze the effect of parameters with variation of
spatial correlation between receive antennas. The elements of separation distance (spatial diversity) and the pattern of the an-
correlation matrix can be obtained by the following equations tennas (pattern diversity), we propose three different structures
[8]: of two antenna elements as shown in Fig. 4. By comparing Type
2 1 with Type 2 2, we can find the effect of pattern diversity.
(9) And, from Type 2 1 and Type 2 3, the influence of spatial di-
(10) versity can be discovered.
As shown in Fig. 5, in all cases, dual array covers 200 MHz
for or . bandwidth. The center frequency of Type 2 1 is slightly in-
The correlation coefficient is usually calculated from the creased compared to the original antenna. In the case of Type
three-dimensional radiation patterns. However, this process 2 2, the distance is identical with Type 2 1 but one of the an-
requires complex and advanced calculation. Recent studies tennas is rotated by 90 so that the return loss characteristic of
show that in some cases, such as a uniform random field, the the rotated antenna is slightly different from the original one.
124 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 6, 2007

Fig. 3. Radiation patterns of single antenna at 2.6 GHz. (a) x 0 y plane,

0 0
(b) y z plane, and (c) x z plane.

Fig. 5. Scattering parameters of three dual arrays. (a) Type 2 1. (b) Type 2 2.
(c) Type 2 2.

pattern even though the distance between antennas is similar.

It can be explained by Fig. 3. As shown in Fig. 3(b), the radi-
ating field in the left side of the ground is less than that in the
Fig. 4. Proposed three different types of dual arrays. (a) Type 2 1. (b) Type right side. Because of the antenna’s position in Type 2 1, we can
2 2. (c) Type 2 3. intuitively deduce that mutual coupling and correlation will be
lower than Type 2 2. Also, by comparing Type 2 1 with Type
2 2, we can notice that mutual coupling does not merely depend
Type 2 2 provides the maximum separation distance among the on the separation distance.
proposed cases. Thus, has the lowest value at resonance fre-
quency because it can provide independent resonance. B. Correlation Coefficient and Capacity Loss
Theoretically, the capacity of the MIMO system grows lin-
IV. PERFORMANCE EVALUATION
early with the number of antennas. However, the correlation in
the MIMO channel induces a loss of capacity. The correlation
A. Mutual Coupling
coefficients of each dual antenna array are shown in Fig. 7.
Mutual coupling can be obtained directly from . As shown In the case of high SNR, capacity loss can be derived from
in Fig. 6, much lower mutual coupling occurred in the Type 2 1 (8), and the result is shown in [6]
case. By comparing Type 2 1 with Type 2 2, we can infer that
mutual coupling differs according to each antenna’s radiation loss (15)
CHAE et al.: MUTUAL COUPLING, CORRELATIONS, AND TARC IN WIBRO MIMO ARRAY ANTENNA 125

Fig. 6. Measured mutual coupling for dual antenna array. Fig. 9. Measured TARC for dual antenna array.

coupling and the phase of incident wave. Thus, we can conclude

that the meaning of TARC can be defined as the return loss of
the whole MIMO antenna array.
Likewise in the previous two parameters, Type 2 1 shows the
best performance with respect to bandwidth and return loss.

V. CONCLUSION
This paper investigates two antenna arrays in handsets
by considering mutual coupling, correlation coefficient, and
TARC.
The results show that these three parameters are not merely
dependent on antenna spacing. However, they are directly af-
Fig. 7. Measured correlation coefficient for dual antenna array. fected by radiation patterns. From these points, we can conclude
that the antenna pattern is a more considerable factor than the
separation distance when the ground size is small and the two
antennas are located in the array. Thus, it is better for us to re-
alize pattern diversity rather than spatial diversity to reduce the
correlation in the case of compact antenna arrays.
Since the variation of is much higher than , mutual
coupling determines the correlation and TARC.
Evaluating array antenna performance in terms of the MIMO
communication environment is important. This paper attempts
to optimally allocate the two PIFA elements in the MIMO
antenna array. The performance evaluation methodology can
apply any type of antenna arrays.

REFERENCES
Fig. 8. Measured capacity loss for dual antenna array. [1] C. E. Shannon, “A mathematical theory of communication,” Bell Syst.
Tech. J., vol. 27, pp. 379–423, Jul. 1948, Bell Syst. Tech. J., vol. 27, pp.
623–656, Oct. 1948.
[2] G. J. Foschini and M. J. Gans, “On limits of wireless communications
Fig. 8 shows the capacity loss of three different types of an- in a fading environment when using multiple antennas,” Wireless Per-
tenna arrays. We focus on the frequency range near and at the sonal Commun., no. 6, pp. 311–335, 1998.
[3] D. S. Shiu, G. J. Foschini, and M. J. Kahn, “Fading correlation and its
resonance frequency in order to get a clear comparison. The ca- effect on the capacity of multielement antenna systems,” IEEE Trans.
pacity loss has the same tendency as scattering parameters. As Commun., vol. 48, pp. 502–513, Mar. 2000.
shown in Fig. 8, Type 2 1 shows the lowest loss of capacity be- [4] M. Manteghi and Y. Rahmat-Samii, “Multiport characteristics of a
wide-band cavity backed annular patch antenna for multipolarization
cause it has the lowest correlation coefficient. We can infer the operations,” IEEE Trans. Antennas Propag., vol. 53, pp. 466–474, Jan.
effect of correlation between antennas on channel capacity by 2005.
Fig. 8. [5] C.-N. Chuah, D.N.C. Tse, J. M. Khan, and R. A. Valenzuela, “Ca-
pacity scaling in MIMO wireless systems under correlated fading,”
IEEE Trans. Inf. Theory, vol. 48, pp. 637–650, Mar. 2002.
C. TARC [6] H. Shin and J. H. Lee, “Capacity of multiple-antenna fading channel:
Spatial fading correlation, double scattering, and keyhole,” IEEE Trans.
The calculated average TARC with 20 excitation vectors for Inf. Theory, vol. 49, pp. 2636–2647, Oct. 2003.
Type 2 1, Type 2 2, and Type 2 2 are represented in Fig. 9. As [7] E. T. Browne, Introduction to the Theory of Deteminants and Ma-
shown in Fig. 9, TARC values retain the original behavior of trices. Chapel Hill, NC: North Carolina Press, 1958.
[8] S. Blanch, J. Romeu, and I. Corbella, “Exact representation of antenna
a single antenna characteristic. However, bandwidth and return system diversity performance from input parameter description,” Elec-
loss are changed because TARC contains the effect of mutual tron. Lett., vol. 39, pp. 705–707, May 2003.