This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

High-Isolation 3.5-GHz 8-Antenna MIMO Array

Using Balanced Open Slot Antenna Element for
5G Smartphones
Yixin Li, Chow-Yen-Desmond Sim, Senior Member, IEEE, Yong Luo, and Guangli Yang, Member,
IEEE

 80
2G/3G/4G Ant. Metal strip on
Abstract—A high-isolation 8-antenna multi-input multi-output 25 Ant 1 the front surface
(MIMO) array operating in the 3.5-GHz band (3.4–3.6 GHz) for Ground plane on
future smartphones is proposed. Here, a novel balanced open slot the back surface
Ant 6 Ant 2
antenna is designed as an array antenna element, in which this Feeding point
antenna design can yield a balanced slot mode (with reduced
ground effects) that can enhance the isolation between two 19.5 50-ohm microstrip feed line
adjacent input ports. Furthermore, by meticulously arranging the FR4 substrate
positions of the eight antenna elements, desirable polarization Ant 7 Ant 3 150 × 80 × 0.8 mm3
diversity can also be successfully achieved, which further 150
mitigates the coupling between antenna elements. A prototype Unit: mm
was manufactured to validate the simulation. Good impedance 19 System ground plane
matching (return loss > 10 dB), high isolation (> 17.5 dB), high
total efficiency (> 62%) and low envelope correlation coefficient y
(ECC, < 0.05) were measured across the desired operation Ant 8 Ant 4
bandwidth. To verify the MIMO performance, ergodic channel
capacity using the Kronecker channel model was calculated. 25 Ant 5 z x
Effects of hand phantom were also studied. 2G/3G/4G Ant.
56.5 3
Index Terms—5G smartphone, MIMO antenna, balanced open (a)
slot antenna, decoupling, massive MIMO

I. INTRODUCTION

F OR the fifth-generation (5G) mobile communication,

massive multi-input multi-output (MIMO) can significantly
enhance the channel capacity and spectral efficiency, as
compared with the low-order MIMO schemes (2×2 or 4×4 (b)
Fig. 1. Geometry and dimensions of the proposed MIMO antenna array. (a)
MIMO) deployed in the fourth-generation (4G) Perspective view. (b) Detailed structure of the balanced open slot antenna
long-term-evolution (LTE) systems [1], [2]. To arouse diversity element (Ant 1).
and multiplexing potentials, effective array decoupling is In recent years, many decoupling techniques using external
required, in which presently, a 10-dB isolation level is decoupling structures have been proposed, including protruded
acceptable for 5G MIMO in sub-6 GHz spectrum. However, to ground [3], etched slot [4], neutralization line (NL) [5],
achieve better MIMO performances, the lower bound of parasitic scatterer [6], and electromagnetic bandgap structure
isolation should be improved to larger than 15 dB, which is [7]. Even though these decoupling structures are good in
presently a very challenging topic for 5G antenna design. enhancing isolation, for highly integrated 5G MIMO arrays,
some of these designs will usually incur additional efficiency
loss. In other words, isolation improvement is usually at the
This work was supported in part by the Shanghai Eastern Scholarship under
Grant SH03027. It is also partially sponsored by Ministry of Science and cost of antenna efficiency. For example, the multi-NL-based
Technology MOST107-2221-E-035-034. (Corresponding author: Guangli 5G arrays in [8] have shown undesirable total efficiencies (>
Yang.) 30%). Even though the multi-mode protruded strips of the
Y. Li, Y. Luo and G. Yang are with Shanghai Institute for Advanced
Communication and Data Science, Shanghai University, Shanghai 200444, 8-antenna array in [9] can enhance the isolation to larger than
China, and also with Key Laboratory of Specialty Fiber Optics and Optical 15 dB, the simulated lower bound of total efficiency has been
Access Networks, Shanghai University, Shanghai 200444, China (e-mail: reduced to 34% with these decoupling structures. In contrast,
[email protected]; [email protected]; [email protected]).
C. Y. D. Sim is with Department of Electrical Engineering, Feng Chia intrinsic decoupling refers to suppressing mutual coupling by
University, Taichung 40724, Taiwan (e-mail: [email protected]). manipulating radiation mechanisms, including reducing ground

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

effects [10], orthogonal chassis mode [11], and polarization and L, respectively, and they share the same width of 1.5 mm.
diversity [12]–[14]. If a MIMO antenna array is intrinsically TABLE I
LENGTH PARAMETERS OF EACH ANTENNA ELEMENT
decoupled without introducing lossy external decoupling
Parameters L (mm) L1 (mm) L2 (mm)
structures, a tradeoff between isolation and efficiency can be
Ants 1 and 5 9.5 6.5 6.5
obtained. Presently, it is still difficult to devise a novel Ants 2, 4, 6 and 8 13.5 6.7 6.1
methodology to intrinsically decouple a 5G MIMO array while Ants 3 and 7 13.5 6.8 5.9
minimizing additional efficiency loss.
In this paper, an 8-antenna MIMO array operating in the Each antenna element in the array is designed to work in the
3.5-GHz band (3.4–3.6 GHz, LTE band 42) for future possible 3.5-GHz band (3.4–3.6 GHz). Because of its small-size (3 mm
5G smartphone applications is proposed. The most important × 21.5 mm, which is 0.035λ0 × 0.25λ0 at 3.5 GHz) characteristic,
feature of this proposed antenna array is that it is able to achieve the proposed antenna element with a slot radiator is compatible
high isolation (> 17.5 dB) and high total efficiency (> 60%) with narrow-frame smartphones. Notably, to attain better
simultaneously, without employing external decoupling agreement of antenna performances, the length parameters L, L1
structures. To achieve good isolation between adjacent antenna and L2 of the eight antenna elements are fine-tuned with respect
elements, the proposed balanced open slot antenna element is to their respective positions. The detailed values of the three
devised with reduced ground current effects. Furthermore, all parameters are given in Table I.
the eight antenna elements are also meticulously arranged on 0.5 0.5
the ground plane (or chassis), two of which are with orthogonal
configurations to generate polarization diversity. The proposed 20.5 25
antenna array is simulated, fabricated and tested. Besides
S-parameters, radiation performances and MIMO
proposed ref 1
performances, user’s hand impacts are also investigated.
with branch slot without branch slot
Fig. 2. The proposed antenna element and its referential case (ref 1).
II. PROPOSED MIMO ANTENNA ARRAY Reflection coefficients (dB) 0 LTE band 42
A. Antenna Geometry
-5
The geometry of the proposed antenna array is depicted in single antenna
at the top right
Fig. 1(a). In this figure, eight antenna elements, namely, -10
corner
antennas (Ants) 1 to 8 are disposed along the four side edges of
-15
a double-sided copper-clad FR4 substrate (relative permittivity
= 4.4 and loss tangent = 0.02). The substrate has a dimension of -20
ref 1
150 mm × 80 mm × 0.8 mm, which is typically used in 6-inch proposed
-25
smartphones. Here, it is also considered as the printed circuit 3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz)
board (PCB), as the feeding strips (in yellow) and ground plane
(a)
(in blue) are printed on its front and back surfaces, respectively. 0 5000 10000 15000
V/m
Amid the eight antenna elements, Ants 1 and 5 are horizontally
arranged along the short edges, while the other six antennas E-field
(Ants 2, 3, 4, 6, 7 and 8) are vertically disposed along the long null proposed
edges. The two rectangular regions (in orange) at the two PCB magnetic
current
corners are considered for accommodating the two LTE main
antennas for 2×2 MIMO.
Fig. 1(b) exhibits the detailed structure of the antenna E-field
element (consider Ant 1 as an example). Different from the ref 1
well-known single-end open slot antenna [15] and closed slot
null magnetic current
antenna [16], an inverted-E shaped slot radiator with two open
ends is etched on the ground plane. Here, the slot radiator is
dashed arrow denotes magnetic current path
composed of three slot sections, namely, vertical open-end slot, longer arrow denotes stronger magnetic current
horizontal slot and branch slot. The two vertical open-end cross denotes null point
(b)
sections share the same length and width of 3 mm and 0.5 mm, 0 15 30 45
A/m (electric) surface
respectively. The two horizontal slots linked to the two vertical current density
open-end slots share the same length L1 and width of 1 mm. The proposed
branch slot laying between the two horizontal slots has an
extended width of d = 1.5 mm and length L2. It is noteworthy ref 1
that this extended branch slot can aid in reducing the length of
the slot radiator. The slot radiator is fed by an L-shaped (c)
Fig. 3. Simulated or estimated results of the two cases (proposed and ref 1), at
microstrip feeding line at the feeding point. The lengths of the 3.5 GHz. (a) Reflection coefficients. (b) Electric field (E-field) and equivalent
horizontal and vertical sections of this feeding line are 7.5 mm magnetic current distributions. (c) Surface electric current distributions.

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

0 A/m

Reflection coefficients (dB)

30 Ant 8 Ant 7 Ant 6
-5
-10
Proposed 15
(electric) surface current density
-15 (d = 1.5 mm,
L2 = 6.5 mm) Ant 5 Ant 1
-20 d = 1.8 mm 0
d = 1.2 mm Ant 4 Ant 3 Ant 2
-25 L2 = 8.5 mm
L2 = 4.5 mm (b)
-30
3.0 3.2 3.4 3.6 3.8 4.0 Fig. 5. Surface electric current distributions on the ground when (a) Ant 1 is
Frequency (GHz) excited and (b) Ant 3 is excited.
Fig. 4. Simulated reflection coefficients as a function of the length L2 and width The simulated reflection coefficients when tuning the length
d of the branch slot.
L2 or width d of the branch slot are demonstrated in Fig. 4. It is
B. Design Process and Corresponding Analyses seen that increasing either parameter L2 or d (when the other
As depicted in Fig. 2 and Fig. 3(a), to reveal the design one remains constant) will allow the resonant mode to shift to
process, a referential case (ref 1, horizontal slot length = 25 mm, the lower frequency, and vice versa. Thus, the resonant mode of
without the branch slot) is compared with the proposed antenna the balanced open slot antenna element can be easily tuned by
element (identical with Ant 1, horizontal slot length = 20.5 mm). varying parameter L2 or d of the branch slot.
When the two antenna elements are individually placed at the To verify that the proposed balanced slot antenna element
top right corner (same position as Ant 1), similar resonances at has reduced ground effect (weak ground current distribution)
approximately 3.47 GHz are excited, while the proposed characteristic [10], Figs. 5(a) and 5(b) show the ground current
antenna element is 4.5-mm (18%) shorter than ref 1. Thus, the distributions when Ant 1 and Ant 3 are excited separately. As
antenna length is effectively reduced by etching the branch slot. shown in both figures, when either Ant 1 or Ant 3 is excited, its
The electric field (E-field) and equivalent magnetic current corresponding ground current does not spread towards or
distributions of the two cases that are plotted in Fig. 3(b) will highly affect its adjacent antenna port, indicating that the
explain the operating principle of the antenna element. For both coupling effect from the ground is weak.
cases, the E-field peaks (antinodes) are located at the two open Ant a1 Ant b1
ends, while the null (node) is located across the branch slot (or
the middle portion). The excited E-field behaves like two Ant a6 Ant a2 Ant b6 Ant b2
unbalanced open slot antennas joining at the common null case A case B
where the direction of E-field reverses, meaning that a
half-wavelength balanced slot mode is supported. As balanced Ant a7 Ant a3 Ant b7 Ant b3
y y
antenna yields weaker ground current effects, a higher isolation
z x z x
level can be expected [17]. Furthermore, as indicated by this Ant a8 Ant a4 Ant b8 Ant b4
field distribution, if two adjacent antennas are arranged
perpendicularly, orthogonal polarizations in the horizontal Ant a5 Ant b5
(x-axis) and vertical (y-axis) directions are generated, and the (a) (b)
mutual coupling between them will be weak.
The miniaturization principle of the branch slot can also be Ant c5 Ant c1 Ant d1
comprehended by observing Fig. 3(b). The slot etched near the Ant d6 Ant d2
E-field null can be regarded as inductive loading, leading to
Ant c6 Ant c2 case D
strong normal E-field extended to its rectangular periphery.
Therefore, a prolonged magnetic current path surrounding the case C Ant d7 Ant d3
branch slot is created to aid in length reduction [18]. From the y
Ant c7 Ant c3
electric current distributions presented in Fig. 3(c), one can see y z x
that strong surface current occurs at the middle portion, and Ant d8 Ant d4
z x
gradually decreases toward the open ends. Ant c8 Ant c4 Ant d5
A/m
30 Ant 8 Ant 7 Ant 6
(c) (d)
Fig. 6. Four referential array configurations. (a) Case A. (b) Case B. (c) Case C.
(d) Case D.
15
(electric) surface current density C. Array Configuration
Ant 5 Ant 1 It has been reported that for 5G MIMO antenna arrays, an
0 appropriate array configuration can bring considerable
Ant 4 Ant 3 Ant 2 improvement in impedance matching and isolation [8], [12],
(a) [13]. To emphasize the advantage of the proposed array
configuration, four referential array configurations, namely,

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

cases A to D [see Figs. 6(a)–6(d)] are simulated. D) cannot yield good antenna performances as compared with
From Fig. 1(b), it is notable that the proposed antenna the proposed one. For case A, as shown in Fig. 7(c), because the
element is symmetrical with respect to its central axis. This open ends of the feeding strips are too close to the PCB edge,
means that if it is fed by the L-shaped feeding strip, the feeding the two horizontal antennas (Ants a1 and a5) are unable to
strip can be placed at either side of the central axis to excite the cover the 3.5-GHz band. Furthermore, the S21 and S54 in this
resonant mode. Thus, cases A and B are simulated to study the case are increased to –13.5 dB. For case B, as demonstrated in
effects of the location of the feeding strip. In case A, the feeding Fig. 7(d), although the antenna elements can completely cover
strips of Ants 1 and 5 are reversed and pointing at the PCB the operation bandwidth, the isolation between Ant 2 and Ant 3
corners. In case B, the feeding strips of Ants 2, 4, 6 and 8 are (11 dB) is relatively poor. From these two cases, it is interesting
reversed. In addition, cases C and D are investigated to to note that when the feeding strips (open ends) of two adjacent
comprehend the decoupling effects of orthogonal antenna antennas are pointing at each other, undesirable isolation is
arrangements and reduced ground effects. In case C, all the generated. This behavior can be comprehended by observing
eight antennas are placed vertically to attain consistent the ground current distributions in Fig. 3(c). As can be seen,
polarizations. Here, the dimensions of Ants c1 and c2 are when the feeding strip is placed on the left side, stronger current
identical with Ants 2 and 3 in the proposed antenna array. The is also concentrated around the left half slot. Consequently, if
other six antennas are the mirror images of Ants c1 and c2. In the open ends of two adjacent feeding strips are pointing at each
case D, the proposed balanced antenna elements are replaced other, poorer isolation can be realized due to the following two
by eight unbalanced open slot antennas whose resonant modes main reasons: (1) the two strong ground current regions will be
are well designed at about 3.5 GHz. nearer to each other, and (2) the separation distance between the
0 -10
two feeding ports is reduced. Therefore, a design guideline for
Transmission coefficients (dB)

LTE band 42 LTE band 42 S21 S32

Reflection coefficients (dB)

proposed proposed S43 S54 the proposed antenna array can be drawn, that is, the two
-5 S61 S85
-15 face-to-face feeding strips (open ends) of adjacent antennas
-10 2:1 VSWR –17.5 dB
-20 should be arranged farther (such as Ants 1 and 6, or Ants 5 and
-15 S11
S22 8 of the proposed antenna array), so as to achieve a balance of
S33 -25
-20
S44 isolations.
S55
-25 -30
3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz) Frequency (GHz)
(a) (b)
0 0
S11 S22 S33 S21 S11 S22 S33
S32 S43 -5 S21 S32 S43
-5
S-parameters (dB)

S54
S-parameters (dB)

-10 S54 -10 S61

S61 –11 dB S85
-15 –13.5 dB S85 -15

-20 -20

-25 -25

-30 case A -30 case B

3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz) Frequency (GHz)
(c) (d)
0 S11 S22 S21 S32 0 S11 S22 S33 S21
-5 S51 S62 S32 S43 S54
-5
S-parameters (dB)

S-parameters (dB)

-10 S61
–11 dB -10
-15 –11.5 dB S85
-15
-20
-20
-25
-30 -25
-35 case C -30 case D
3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz) Frequency (GHz)
(e) (f) Fig. 8. Photographs (front view and back view) of the fabricated prototype.
Fig. 7. Simulated results. (a) Reflection coefficients and (b) transmission
coefficients of the proposed antenna array. S-parameters of (c) case A, (d) case
B, (e) case C and (f) case D. For case C, because orthogonal antenna polarizations no
The simulated S-parameters of the proposed antenna array longer exist, deteriorated isolation (11 dB) between Ant 1 and
and cases A to D are given in Fig. 7. As depicted in Figs. 7(a) Ant 2 is shown in Fig. 7(e). For case D, as depicted in Fig. 7(f),
and 7(b), the proposed MIMO antenna array can well cover the even though the open slot antennas can fully cover the 3.5-GHz
desired 3.5-GHz band (3.4–3.6 GHz) with reflection coefficient band, the isolations between adjacent ports have dropped down
less than –10 dB (2:1 voltage standing wave ratio, VSWR). to 11.5 dB, which is not desirable enough for 5G MIMO
Because of the symmetrical antenna placements, the reflection implementation. Nonetheless, with the aid of orthogonal
coefficients of Ants 6 to 8 are omitted here, for brevity. Owing antenna arrangements and reduced ground effects, the isolation
to the two intrinsic decoupling techniques (orthogonal antenna level can be improved by approximately 6 dB. Therefore, it can
arrangements and reduced ground effects), very desirable be summarized that the two intrinsic decoupling techniques
isolation of larger than 17.5 dB is acquired across the operation proposed in this work are very effective for enhancing the
bandwidth. Nevertheless, the four referential arrays (cases A to overall performances of the proposed antenna array.

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

0 LTE band 42 A. S-Parameters

Reflection coefficients (dB)

Measured
-5 Fig. 9 exhibits the measured S-parameters. Apparently, the
simulation and measurement validate well with each other.
-10 2:1 VSWR
However, slight discrepancies are still observed, which is
-15 S11 probably due to the fabrication tolerances and measurement
S22 errors. As shown in Fig. 9(a), for Ants 1 to 5, the measured
-20 S33
S44 stacked impedance bandwidth defined by 2:1 VSWR or 10-dB
-25
S55 return loss was 3.39–3.67 GHz (7.93% relative bandwidth),
3.0 3.2 3.4 3.6 3.8 4.0 which is wide enough for supporting the desired 3.5-GHz band
Frequency (GHz)
(a) (3.4–3.6 GHz, 5.71% relative bandwidth). As depicted in Fig.
-10 9(b), the measured transmission coefficients between adjacent
Transmission coefficients (dB)

LTE band 42 S21 S32

Measured S43 S54 ports were less than –17.5 dB, which is advantageous for
-15 S61 S85 achieving good diversity and multiplexing performances.
–17.5 dB
0 0+x
+x
-20 5 330 30 5 330 30
-5 -5
300 60 300 60
-25 -15 -15
-25 -25
–y +y –y +y

dB

dB
-35 270 90 -35 270 90
-30
3.4 3.6 3.0
3.8 4.03.2 -25 -25
Frequency (GHz)
-15 -15
(b) 240 120 240 120
-5 Ant 1 -5 Ant 5
Fig. 9. Measured S-parameters of the proposed antenna array. (a) Reflection
5 210 150 5 210 150
coefficients. (b) Transmission coefficients. –x
180
–x
180
(a) (b)
0+x 0+x
III. RESULTS AND DISCUSSION 5 330 30 5 330 30
-5 -5
The proposed antenna array was fabricated, and its front and 300 60 300 60
-15 -15
back photographs are depicted in Fig. 8. The simulated results -25 -25
of the proposed antenna array were obtained through CST –y +y –y +y

dB
dB

-35 270 90 -35 270 90

Microwave Studio Version 17, and its S-parameters were -25 -25
measured by an Agilent E5071C vector network analyzer. The -15 -15
240 120 240 120
far-field results were obtained by using an ETS-Lindgren -5 Ant 2 -5 Ant 4
microwave anechoic chamber, with the excited port connected 5 210 –x 150 5 210 –x
180
150
180
to the testing cable, and the irrelevant ports terminated to (c) (d)
0+x
50-ohm broadband matched loads. The corresponding results 5 330 30
will be presented and discussed in the following subsections. -5
300 60
-15
100 LTE band 42 -25
Gain-theta (mea.)
Ant 1
Simulated –y +y Gain-phi (mea.)
Ant 2
dB

90 -35 270 90
Total efficiency (%)

Gain-theta (sim.)
Ant 3
-25 Gain-phi (sim.)
80 Ant 4
Ant 5 -15
240 120
70 -5 Ant 3
5 210 –x 150
60 180
(e)
50 Fig. 11. Simulated and measured radiation patterns for (a) Ant 1, (b) Ant 5, (c)
40 Ant 2, (d) Ant 4, and (e) Ant 3, at 3.5 GHz.
3.3 3.4 3.5 3.6 3.7
Frequency (GHz) B. Radiation Performances
(a) Fig. 10 exhibits the simulated and measured total efficiencies,
100 LTE band 42 Ant 1 which have included the effects of mismatching and coupling
Measured Ant 2
90
Total efficiency (%)

Ant 3 losses. Here, it is seen that the measured results also agreed
80 Ant 4 with the simulated ones. The measured total efficiencies were
Ant 5
70 62%–76% in the operation bandwidth, which is approximately
5% lower than the simulated ones. Nevertheless, all the
60
measured results are higher than 60% with isolation larger than
50 17.5 dB, indicating that a proper balance between the isolation
40 and antenna efficiency is achieved.
3.5 3.6
3.3 3.73.4 The simulated and measured radiation patterns in the xy
Frequency (GHz)
(b) plane for Ants 1 to 5 working in the 3.5-GHz band are plotted in
Fig. 10. (a) Simulated and (b) measured total efficiencies of Ants 1 to 5. Fig. 11. The results of Ants 6 to 8 are omitted here, because

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

their radiation patterns are almost the mirror images of Ants 2 elements of the proposed antenna array act as the receiving
to 4. As shown in Figs. 11(a) and 11(b), horizontal Ants 1 and 5 antennas. In the figure, the ergodic channel capacity of the
have yielded strong radiation in the phi = 135° and phi = 225° proposed antenna array varies approximately between 39.5
directions, respectively. In comparison, the maximum radiating bps/Hz and 40.8 bps/Hz. The peak achievable channel capacity
directions of vertical Ants 2 to 4, as shown in Fig. 11(c)–11(e), (40.8 bps/Hz) is 88.7% of the upper limit for 8×8 MIMO (46
are in the phi = 0° (+x axis) orientation. Desirable pattern bps/Hz), and is 354.8% of the upper limit for 2×2 MIMO (11.5
diversity feature is demonstrated, because the radiation patterns bps/Hz), thereby showing good multiplexing capability.
of the two horizontal antennas or the three vertical antennas are TABLE II
PERFORMANCE COMPARISON OF 5G MIMO HANDSET ANTENNA ARRAYS
distinctly complementary, and their respective main radiating
References BW (GHz) Sij (dB)/ECC TE (%) PCC#
directions are scattering away from the PCB.
3.4–3.6 62–76 40.8
Proposed < –17.5/0.05
(–10 dB) (mea.) (8×8)
Envelope correlation coefficient

0.10 Ant 1 and Ant 2 Ant 2 and Ant 3 3.4–3.6 62–78 39

[5]^ < –10/0.2
Ant 3 and Ant 4 Ant 4 and Ant 5 (–10 dB) (mea.) (8×8)
Ant 1 and Ant 6 Ant 5 and Ant 8 3.4–3.6 30–52 37
[8]* < –10/0.3
Ant 2 and Ant 6 Ant 3 and Ant 7 (–6 dB) (mea.) (8×8)
3.4–3.6 34–65
LTE band 42 [9] < –15/NG NG
0.05 (–10 dB) (sim.)
2.55–2.65 48–63 40
[12] < –12.5/0.15
(–10 dB) (mea.) (8×8)
2.55–2.65 48–58 39
[13] < –13/0.2
(–10 dB) (mea.) (8×8)
0.00 3.4–3.6 50–68 57
[14] < –12/0.2
(–6 dB) (mea.) (12×12)
3.3 3.4 3.5 3.6 3.7
Frequency (GHz) 3.4–3.6 40–52 36
[21] < –10/0.15
(–6 dB) (mea.) (8×8)
(a)
3.4–3.6 59–72 15
Ergodic channel capacity (bps/Hz)

[22] NG/NG
55 (–6 dB) (sim.) (2×8)
50 46 bps/Hz, upper limit for 8×8 MIMO 3.4–3.6 45–58 37
[23] < –12/0.1
45 (–6 dB) (mea.) (8×8)
40 3.4–3.8 40–62 47
[24] < –10/0.1
35 (–6 dB) (mea.) (10×10)
Proposed 8×8 MIMO antenna array
30 Abbreviations: BW=Bandwidth, TE=Total Efficiency, PCC=Peak Channel
25 LTE band 42 Capacity, sim.=simulation, mea.=measurement, NG=Not Given.
#
20 All the channel capacities are obtained by assuming SNR = 20 dB.
15 11.5 bps/Hz, upper limit for 2×2 MIMO ^Only 5G antennas are considered for comparison.
10 *Without loss of generality, only array B is considered for comparison.
5
3.5 3.6 3.3
3.7 3.4
Frequency (GHz)
Ant 8 Ant 7 Ant 6
(b) Ant 1
Fig. 12. Calculated (a) envelope correlation coefficient (ECC) and (b) ergodic
channel capacity from the measured results. Ant 6 Ant 2 Ant 4 Ant 3 Ant 2

Ant 5 Ant 1
C. Diversity and Multiplexing Performances
Ant 7 Ant 3 DHM
To verify the MIMO potentials of the proposed antenna array,
the corresponding envelope correlation coefficient (ECC) and Ant 8 Ant 4
channel capacity are studied in this subsection. The ECCs
shown in Fig. 12(a) are calculated from the measured Ant 5

three-dimensional complex E-field patterns, under the SHM

(a) (b)
hypothesis that the propagation scenario is uniform, with Fig. 13. Two representative application scenarios of the hand-held smartphone.
isotropic incident wave distributed in the theta-polarized and (a) Single-hand mode (SHM). (b) Dual-hand mode (DHM).
phi-polarized directions [19]. It is observed that the calculated
ECCs were all lower than 0.05. Across the operation bandwidth, D. Comprehensive Comparison of Antenna Performances
the largest ECC (between Ant 1 and Ant 2, at 3.4 GHz) is
Table II shows the performance comparison between the
merely 0.036, which is far less than the acceptable criterion
proposed work and those that have been reported. In order to
(ECC < 0.5) for 5G MIMO antenna arrays.
further understand the merits of the proposed work, more recent
As shown in Fig. 12(b), using the Kronecker channel model,
works with 8- or 10-antenna arrays [21]–[24] are also added
the ergodic channel capacity of the proposed antenna array is
into this Table. From the comprehensive comparison, it is
calculated by averaging 100,000 independent and identically
observed that the proposed antenna array has shown higher
distributed (i.i.d.) Rayleigh fading channel realizations [20].
isolation (> 17.5 dB) with a lesser sacrifice in total efficiency (>
Here, no channel state information in the transmitter and a
62%). This unique feature can make the proposed antenna array
20-dB signal-to-noise ratio (SNR) in the receiver are assumed.
more promising for future 5G MIMO operation. However,
In the 8×8 MIMO system, the transmitting antennas are
most referential 5G MIMO arrays in the table cannot achieve a
assumed to be uncorrelated and lossless, while the antenna
high isolation level of > 15 dB. Even though the work in [9] has

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

-10

Transmission coefficients (dB)

presented a high-isolation 8-antenna array using external S21 S32 S43 S54
decoupling strips, a deteriorated antenna efficiency of 34%– -15 S61 S85 S76 S87
65% has been observed, owing to the additional loss –18 dB
-20
engendered by the external decoupling structures. Although
polarization diversity has been applied to the antenna arrays in -25
[12]–[14], the proposed work still shows much better overall
-30
performances, because it has successfully combined two
intrinsic decoupling techniques. -35
0 DHM
LTE band 42 -40
Reflection coefficients (dB)

3.0 3.2 3.4 3.6 3.8 4.0

-5 Frequency (GHz)
(b)
-10 2:1 VSWR 80
LTE band 42
70

Total efficiency (%)

-15 60
S11 S22 Ant 1
-20 S33 S44 50 Ant 2
S55 S66 40 Ant 3
SHM S77 S88
-25 Ant 4
30
3.0 3.2 3.4 3.6 3.8 4.0 Ant 5
Frequency (GHz) 20 Ant 6
(a) 10 Ant 7
-10
Transmission coefficients (dB)

DHM Ant 8
S21 S32 S43 S54 0
-15 S61 S85 S76 S87 3.3 3.4 3.5 3.6 3.7
–17 dB Frequency (GHz)
-20 (c)
Fig. 15. Simulated performance metrics under dual-hand mode (DHM). (a)
-25 Reflection coefficients. (b) Transmission coefficients. (c) Total efficiencies.

-30
IV. PRACTICAL APPLICATION ANALYSES
-35
SHM
A. Impacts of User’s Hand
-40 The impacts of hand phantom are studied in this subsection.
3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz) Two typical application scenarios, namely, single-hand mode
(b) (SHM) and dual-hand mode (DHM) are depicted in Fig. 13 [25].
80
LTE band 42 The corresponding simulated performance metrics for the two
70 modes are plotted in Figs. 14 (for SHM) and 15 (for DHM).
Total efficiency (%)

60 Notably, the impacts of the user’s head are not included here,
Ant 1
50 Ant 2 because multi-port MIMO operation in sub-6 GHz is only to be
40 Ant 3 applied to data transmission, rather than talk mode.
30 Ant 4 For SHM, Ants 2, 6 and 7 are directly contacted by the
Ant 5
20 Ant 6 fingers. Because of this arrangement, as demonstrated in Fig.
10 Ant 7 14(a), the reflection coefficients of Ants 2, 6 and 7 are highly
SHM Ant 8
0 affected. Here, Ant 6 has also shown undesirable impedance
3.3 3.4 3.5 3.6 3.7 matching. In contrast, minor variations in the resonant modes of
Frequency (GHz)
(c) Ants 1, 4, 5 and 8 are observed, because they are farther away
Fig. 14. Simulated performance metrics under single-hand mode (SHM). (a) from the fingers. Nonetheless, as illustrated in Fig. 14(b), high
Reflection coefficients. (b) Transmission coefficients. (c) Total efficiencies. isolation level (> 17 dB) can still be maintained across the
0
LTE band 42 operation bandwidth. Because of the power absorption of hand
Reflection coefficients (dB)

-5 tissue, the total efficiencies of Ants 2, 6 and 7 have been

S11 reduced to below 30%. As for Ant 3, because it is located
-10 2:1 VSWR S22 within a close proximity to the palm, the power absorption by
S33 the palm has also decreased its efficiency to about 10%.
-15 S44
S55 However, as the other four antennas (Ants 1, 4, 5 and 8) are
-20 S66 farther away from the hand, Ants 1 and 4 have shown total
S77 efficiencies of higher than 40% and 35%, respectively, while
DHM S88
-25 Ants 5 and 8 have demonstrated higher total efficiencies of
3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz) larger than 55%.
(a) For DHM, as Ants 1 and 5 are covered by the thumbs, their
resonant modes are shifted to the higher frequency, and a slight
deterioration in their reflection coefficients is observed, as

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

shown in Fig. 15(a). In addition to that, because Ants 2 and 4 the six vertical antennas (Ants 2, 3, 4, 6, 7 and 8) are now
are lying above the two little fingers, they have also playing a major role, as seen in Fig. 16(b). The main lobes of
demonstrated poorer impedance matchings. In contrast, the the vertical antennas are directed away from the PCB and
other four antenna elements (Ants 3, 6, 7 and 8) are much scattering toward different directions, which signifies distinct
farther away from the hand phantom, thus their resonant complementarity.

Envelope correlation coefficient

characteristics are nearly the same as before. Interestingly, as 0.10 Ant 1 and Ant 4 Ant 4 and Ant 5
depicted in Fig. 15(b), the isolation is higher than 18 dB over Ant 1 and Ant 8 Ant 5 and Ant 8

the operation bandwidth. As shown in Fig. 15(c), the total

efficiencies of Ants 1 and 5 (covered by the thumbs) have been LTE band 42
0.05
reduced tremendously to less than 25%, whereas Ants 2 and 4
(lying above the little fingers) have demonstrated slightly better
total efficiencies of larger than 35%. The other four antennas
(Ants 3, 6, 7 and 8) have exhibited acceptable total efficiencies 0.00 SHM
of larger than 50%. Considering the branch power balance, the 3.3 3.4 3.5 3.6 3.7
Frequency (GHz)
proposed antenna array is still competent for 4×4 (Ants 1, 4, 5 (a)
and 8) and 6×6 MIMO (Ants 2, 3, 4, 6, 7 and 8) operations

Envelope correlation coefficient

0.10 Ant 2 and Ant 3 Ant 3 and Ant 4
under the SHM and DHM conditions. Ant 6 and Ant 7 Ant 7 and Ant 8

LTE band 42
0.05

0.00 DHM
3.3 3.4 3.5 3.6 3.7
Frequency (GHz)
(b)
Fig. 17. Simulated ECCs for (a) SHM and (b) DHM.
Figs. 17 depicts the simulated ECCs for SHM and DHM. The
simulated results are still lower than 0.05, which shows
desirable diversity performance. Based on the above
S-parameters, antenna efficiencies, radiation patterns and ECCs
(a) results, it can be summarized that the proposed antenna array
has good robustness and resistance under hand-grip conditions.

(b)
Fig. 16. Simulated three-dimensional radiation patterns at 3.5 GHz, at (a) SHM
and (b) DHM. Fig. 18. Simulation model for the proposed antenna array with the LCD
As can be seen in Fig. 16, the radiation patterns of the module.
0 -10
Transmission coefficients (dB)

LTE band 42
corresponding antenna elements at the two hand-grip modes
Reflection coefficients (dB)

LTE band 42 S21 S32

S43 S54
-5
still demonstrate complementarity and orthogonality, which is 3:1 VSWR -15 S61 S85

-10
advantageous to spatial diversity. At SHM, the fingers are -20 –18.5 dB
mainly in contact with the two long edges, thus the two -15 S11
S22
-25
horizontal antennas (Ants 1 and 5) arranged at the PCB corners -20 S33
S44
yield relatively strong radiation. Besides, the two vertical -25
S55
-30
3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
antennas (Ants 4 and 8) have also contributed to achieving Frequency (GHz) Frequency (GHz)
diversity, because the radiating directions of horizontal and (a) (b)
Fig. 19. Simulated (a) reflection coefficients and (b) transmission coefficients
vertical antennas are quasi-orthogonal. At DHM, the two of the LCD integrated model.
horizontal antennas are covered by the fingers, and therefore

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

0 0 -10

Transmission coefficients (dB)

LTE band 42

Reflection coefficients (dB)

LTE band 42 LTE band 42 S21 S32
S43 S54
-5
-15 S61 S85
-5
-10 2:1 VSWR –17.5 dB

S11 (dB)
-20
-10 -15 S11
S22
S33 -25
-15 without tuning -20
S44
(t = 7.5, L2 = 6.5) S55
-25 -30
-20 with tuning 3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
(t = 8, L2 = 7.1) Frequency (GHz) Frequency (GHz)

-25
unit: mm (a) (b)
3.0 3.2 3.4 3.6 3.8 4.0 Fig. 22. Simulated (a) reflection coefficients and (b) transmission coefficients
Frequency (GHz) of the battery integrated model.
t L2
C. Impacts of Battery
Fig. 21 demonstrates the simulation model for the proposed
antenna array loaded with a battery. In this investigation, a
metal block (assumed as the battery) with a size of 118 mm ×
40 mm × 4 mm is placed on the front surface of the PCB, and it
Fig. 20. Comparison of simulated S11 (by tuning the parameters t and L2). is electrically connected to the bottom ground plane via 16
B. Impacts of Liquid Crystal Display (LCD) Module shorting pins. The results in Fig. 22 have illustrated that the
battery has negligible impacts on the antennas. The return
Fig. 18 shows the simulation model for the proposed antenna losses and isolations are greater than 10 dB and 17.5 dB,
array with the LCD module, and its corresponding simulated respectively.
results are plotted in Fig. 19. The LCD module is composed of 1 2
1: PCB front surface
two parts, namely, the LCD shield (which adheres to the ground Ant 1 2: PCB back surface
(ground plane)
plane and is made of stainless steel) and the LCD panel (which
Ant 6 Ant 2
adheres to the LCD shield and is made of glass, relative
permittivity = 7 and loss tangent = 0.02). The LCD shield and
Plastic frame
the LCD panel have the same dimension of 144 mm × 74 mm ×
Ant 7 Ant 3
1 mm. The LCD module is extended to the edges of the slot
radiators, and because it is electrically connected to the ground Front view
plane of the PCB, it can be regarded as the ground in real
smartphones. As depicted in Fig. 19, the LCD module has Ant 8 Ant 4 Side view

relatively large impacts on the antenna array, because it is quite Ant 5

1 mm
close to the slot radiators. Although slight frequency shifts have
been observed, as depicted in Fig. 19(a), the array can still Fig. 23. Simulation model for the proposed antenna array with the frame.
0 -10
Transmission coefficients (dB)
achieve a return loss level of > 6 dB, which is currently LTE band 42 LTE band 42 S21 S32
Reflection coefficients (dB)

S43 S54
-5
recognized as an acceptable criterion for most mobile handset 3:1 VSWR -15 S61 S85
–17.5 dB
antennas. By further observing Fig. 19(b), high isolation of > -10
-20
18.5 dB can still be obtained in this case. -15 S11
S22
-25
Even though the LCD module has undesired effects on -20 S33
S44
antennas, it should be noted that the slot resonant mode can be -25
S55
-30
3.0 3.2 3.4 3.6 3.8 4.0 3.0 3.2 3.4 3.6 3.8 4.0
controlled by adjusting the feeding line and the radiator. As Frequency (GHz) Frequency (GHz)

shown in Fig. 20, by fine tuning parameters t and L2, the (a) (b)
Fig. 24. Simulated (a) reflection coefficients and (b) transmission coefficients
matching and frequency shift of Ant 1 can be improved. of the frame integrated model.
0

-5
S11 (dB)

-10
Shorting pin
-15 without tuning
(d = 1.5)
-20 with tuning
Battery (d = 1.3)
unit: mm
-25
3.0 3.2 3.4 3.6 3.8 4.0
Frequency (GHz)
d

Fig. 21. Simulation model for the proposed antenna array with the battery. Fig. 25. Comparison of simulated S11 (by tuning the parameters d).

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10

D. Impacts of Plastic Frame [9] H. Xu, H. Zhou, S. Gao, H. Wang, and Y. Cheng, “Multimode decoupling
technique with independent tuning characteristic for mobile terminals,”
As depicted in Fig. 23, the proposed antenna array is IEEE Trans. Antennas Propag., vol. 65, no. 12, pp. 6739–6751, Dec.
surrounded by a 6-mm high plastic frame (relative permittivity 2017.
= 3 and loss tangent = 0.01). The results in Fig. 24 has shown [10] C. Y. D. Sim, C. C. Chen, and Y. L. Lee, “A dual-band antenna design for
MIMO LTE applications with reduced ground effects,” Int. J. RF and
that the resonant modes of the slot antennas are shifted to the Microwave Comp. Aid. Eng., vol. 26, no. 1, pp. 80–87, Jan. 2016.
lower frequency, which is mainly due to the dielectric loading [11] H. Li, B. K. Lau, Z. Ying and S. He, “Decoupling of multiple antennas in
of the frame. Nevertheless, the return loss can be > 6 dB, and terminals with chassis excitation using polarization diversity, angle
diversity and current control,” IEEE Trans. Antennas Propag., vol. 60, no.
the frame has minor effects on the isolation. The isolation level 12, pp. 5947–5957, Dec. 2012.
can be > 17.5 dB with the presence of the frame. Similar to the [12] M. Y. Li et al., “Eight-port orthogonally dual-polarized antenna array for
LCD integrated model, the resonant mode can be adjusted to 5G smartphone applications,” IEEE Trans. Antennas Propag., vol. 64, no.
9, pp. 3820–3830, Sep. 2016.
cover the desired bandwidth by fine tuning parameter d. It is [13] M. Y. Li, Z. Q. Xu, Y. L. Ban, C. Y. D. Sim, and Z. F. Yu, “Eight-port
observed in Fig. 25 that good 10-dB impedance bandwidth orthogonally dual-polarised MIMO antennas using loop structures for 5G
across the LTE band 42 is achieved by reducing parameter d. smartphone,” IET Microwaves Antennas Propag., vol. 11, no. 12, pp.
1810–1816, Oct. 2017.
Lastly, it is noteworthy that the impacts of this proposed [14] M. Y. Li, Y. L. Ban, Z. Q. Xu, J. Guo, and Z. F. Yu, “Tri-polarized
8-antenna MIMO array for future 5G mobile terminal in 12-antenna MIMO array for future 5G smartphone applications,” IEEE
different propagation scenarios [26] are not investigated in this Access, vol. 6, pp. 6160–6170, 2018.
[15] H. D. Chen, H. W. Yang and C. Y. D. Sim, “Single open-slot antenna for
work. LTE/WWAN smartphone application,” IEEE Trans. Antennas Propag.,
vol. 65, no. 8, pp. 4278–4282, Aug. 2017.
V. CONCLUSION [16] K. L. Wong and W. C. Wu, “Very-low-profile hybrid
open-slot/closed-slot/inverted-F antenna for the LTE smartphone,”
This work has reported an 8-element high-isolation antenna Microw. Opt. Technol. Lett., vol. 58, no. 7, pp. 1572–1577, Jul. 2016.
array for 5G MIMO operation in mobile handsets. A new 5G [17] M. Zheng, H. Wang, and Y. Hao, “Internal hexa-band folded
antenna design methodology is proposed for enhancing monopole/dipole/loop antenna with four resonances for mobile device,”
IEEE Trans. Antennas Propag., vol. 60, no. 6, pp. 2880–2885, Jun. 2012.
isolation as well as reducing efficiency loss. Here, high [18] H. Wang, D. Zhou, L. Xue, S. Gao, and H. Xu, “Modal analysis and
isolation can be realized by simply arranging balanced antenna excitation of wideband slot antennas,” IET Microwaves Antennas
elements into a proper configuration to generate polarization Propag., vol. 11, no. 13, pp. 1887–1891, Oct. 2017.
[19] S. Zhang, A. A. Glazunov, Z. Ying, and S. He, “Reduction of the
diversity. The proposed antenna array successfully covers the envelope correlation coefficient with improved total efficiency for mobile
3.5-GHz band (3.4–3.6 GHz) with suppressed mutual coupling LTE MIMO antenna arrays: Mutual scattering mode,” IEEE Trans.
(< –17.5 dB), desirable antenna efficiency (> 62%) and low Antennas Propag., vol. 61, no. 6, pp. 3280–3291, Jun. 2013.
[20] R. Tian, B. K. Lau, and Z. Ying, “Multiplexing efficiency of MIMO
ECC (< 0.05). The peak ergodic channel capacity in the 8×8 antennas,” IEEE Antennas Wireless Propag. Lett., vol. 10, pp. 183–186,
MIMO scheme (with a 20-dB SNR) can reach up to 354.8% of 2011.
the upper limit for 2×2 MIMO. The proposed antenna array has [21] K. L. Wong, C. Y. Tsai and J. Y. Lu, “Two asymmetrically mirrored
gap-coupled loop antennas as a compact building block for eight-antenna
also exhibited good robustness with the user’s hand, LCD MIMO array in the future smartphone,” IEEE Trans. Antennas Propag.,
module, battery and frame. vol. 65, no. 4, pp. 1765–1778, Apr. 2017.
[22] A. A. Al-Hadi, J. Ilvonen, R. Valkonen, and V. Viikari, “Eight-element
antenna array for diversity and MIMO mobile terminal in LTE 3500 MHz
REFERENCES
band,” Microw. Opt. Technol. Lett., vol. 56, no. 6, pp. 1323–1327, Jun.
[1] W. Hong, “Solving the 5G mobile antenna puzzle: Assessing future 2014.
directions for the 5G mobile antenna paradigm shift,” IEEE Microw. [23] J. Y. Lu, K. L. Wong and W. Y. Li, “Compact eight-antenna array in the
Mag., vol. 18, no. 7, pp. 86–102, Nov.–Dec. 2017. smartphone for the 3.5-GHz LTE 8 × 8 MIMO operation,” in Proc. IEEE
[2] Y. Li, C. Y. D. Sim, Y. Luo, and G. Yang, “12-port 5G massive MIMO Asia-Pac. Conf. Antennas Propag., Kaohsiung, Taiwan, 2016, pp. 323–
antenna array in sub-6GHz mobile handset for LTE bands 42/43/46 324.
applications,” IEEE Access, vol. 6, pp. 344–354, 2018. [24] K. L. Wong and J. Y. Lu, “3.6-GHz 10-antenna array for MIMO
[3] S. Zhang, Z. Ying, J. Xiong, and S. He, “Ultrawideband MIMO/diversity operation in the smartphone,” Microw. Opt. Technol. Lett., vol. 57, no. 7,
antennas with a tree-like structure to enhance wideband isolation,” IEEE pp. 1699–1704, Jul. 2015.
Antennas Wireless Propag. Lett., vol. 8, pp. 1279–1282, 2009. [25] S. Zhang, K. Zhao, Z. Ying, and S. He, “Adaptive quad-element
[4] S. Zhang, B. K. Lau, Y. Tan, Z. Ying, and S. He, “Mutual coupling multi-wideband antenna array for user-effective LTE MIMO mobile
reduction of two PIFAs with a T-shape slot impedance transformer for terminals,” IEEE Trans. Antennas Propag., vol. 61, no. 8, pp. 4275–4283,
MIMO mobile terminals,” IEEE Trans. Antennas Propag., vol. 60, no. 3, Aug. 2013.
pp. 1521–1531, Mar. 2012. [26] K. Zhao, E. Bengtsson, Z. Ying and S. He, “Multiplexing efficiency of
[5] Y. L. Ban, C. Li, C. Y. D. Sim, G. Wu, and K. L. Wong, “4G/5G multiple high order MIMO in mobile terminal in different propagation scenarios,”
antennas for future multi-mode smartphone applications,” IEEE Access, 2016 10th European Conference on Antennas and Propagation (EuCAP),
vol. 4, pp. 2981–2988, Jul. 2016. Davos, 2016, pp. 1–4.
[6] Z. Li, Z. Du, M. Takahashi, K. Saito, and K. Ito, “Reducing mutual
coupling of MIMO antennas with parasitic elements for mobile terminals,”
IEEE Trans. Antennas Propag., vol. 60, no. 2, pp. 473–481, Feb. 2012.
[7] J. Y. Lee, S. H. Kim, and J. H. Jang, “Reduction of mutual coupling in
planar multiple antenna by using 1-D EBG and SRR structures,” IEEE
Trans. Antennas Propag., vol. 63, no. 9, pp. 4194–4198, Sep. 2015.
[8] K. L. Wong, J. Y. Lu, L. Y. Chen, W. Y. Li, and Y. L. Ban, “8-antenna
and 16-antenna arrays using the quad-antenna linear array as a building
block for the 3.5-GHz LTE MIMO operation in the smartphone,” Microw.
Opt. Technol. Lett., vol. 58, no. 1, pp. 174–181, Jan. 2016.

0018-926X (c) 2018 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAP.2019.2902751, IEEE
Transactions on Antennas and Propagation
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 11

Yixin Li was born in Hunan, China, in 1994. He Guangli Yang (M’06) received B.S. degree in
received the B.S. degree in communication Physics from Beijing University of Science and
engineering from Shanghai University (SHU), Technology, China in 1997. He received Ph.D. degree
Shanghai, China, in 2016. He is currently pursuing his in Electrical Engineering from University of South
M.S. degree in electromagnetic fields and microwave Carolina, USA, in 2005. He was mostly with antenna
technology with the School of Communication and and RF research group at Motorola Inc. in USA from
Information Engineering, Shanghai University. His 2005 to 2013 started as Senior, Senior Staff and
current research interests include electrically small promoted as Principal Engineer before left the
antennas, MIMO antennas, and 5G antennas. He is company. After that, he had been a professor at
now serving as a reviewer for IEEE Access and Shanghai University since March 2014. He is author
Applied Computational Electromagnetics Society (ACES) Journal. or co-author of more than 60 publications and has
over 21 patents filed or issued. His research interests
include smart antennas, antenna miniaturization and configurability, digital
Chow-Yen-Desmond Sim (M’07–SM’13) was born beam-forming system, and microwave circuits. Dr. Yang is the Director of the
in Singapore in 1971. He received the B.Sc. degree RF Research Group at Shanghai University and the recipient of Shanghai
from the Engineering Department, University of “Eastern Scholarship” Awards in 2013.
Leicester, U.K., in 1998, and the Ph.D. degree from
the Radio System Group, Engineering Department,
University of Leicester, in 2003. From 2003 to 2007,
he was an Assistant Professor with the Department of
Computer and Communication Engineering,
Chienkuo Technology University, Changhua, Taiwan.
In 2007, he joined the Department of Electrical
Engineering, Feng Chia University (FCU), Taichung,
Taiwan, as an Associate Professor, where he became
a Full Professor in 2012 and as a Distinguish Professor in 2017. He has served
as the Executive Officer of Master’s Program with the College of Information
and Electrical Engineering (Industrial Research and Development), the
Director of Intelligent IoT Industrial Ph.D. Program between August 2015 and
July 2018. He is now serving as the Head of Department of Electrical
Engineering and Director of the Antennas and Microwave Circuits Innovation
Research Center in Feng Chia University. He has authored or coauthored over
130 SCI papers. His current research interests include antenna design,
VHF/UHF tropospheric propagation, and RFID applications. He is a Fellow of
the Institute of Engineering and Technology (FIET), a Senior Member of the
IEEE Antennas and Propagation Society, and a Life Member of the IAET. He
served as the TPC Member of the APMC 2012, the APCAP 2015, IMWS-Bio
2015, CSQRWC 2016, ICCEM 2017, APCAP 2018, and CIAP 2018. He has
also served as the TPC Sub-Committee Chair (Antenna) of the ISAP 2014 and
PIERS 2017. He was invited as the Workshop/Tutorial Speaker in APEMC
2015, iAIM 2017, InCAP 2018, and Invited Speaker of TDAT 2015, iWAT
2018, and APCAP 2018. He was the Keynote Speaker of SOLI 2018. He has
served as the Advisory Committee of InCAP 2018, and has also served as the
TPC Chair of the APCAP 2016, and the Chapter Chair of the IEEE AP-Society,
Taipei Chapter (from 01/2016 to 12/2017). He is the founding Chapter Chair of
the IEEE Council of RFID, Taipei Chapter (from 10/2017). He is now serving
as the Associate Editor of IEEE AWPL, IEEE Access, IEEE Journal of RFID
and (Wiley) International Journal of RF and Microwave Computer-Aided
Engineering. Since October 2016, he has been serving as the technical
consultant of SAG (Securitag Assembly Group), which is one of the largest
RFID tag manufacturers in Taiwan. He was the recipient of the IEEE Antennas
and Propagation Society Top 10 Outstanding Reviewer Award (IEEE
Transaction Antennas and Propagation) for five consecutive years between
2014 and 2018. He has also received the Outstanding Associate Editor Award
from the IEEE Antennas Wireless and Propagation Letters in July 2018.

Yong Luo currently works in Shanghai University

(SHU) as an assistant professor in the Department of
Electronic and Information Engineering, School of
Communication and Information Engineering. Before
joining SHU, He had worked in Prof. Sievenpiper’s
Applied Electromagnetic Group as a postdoc in the
field of non-linear meta-surfaces using diodes in 2016.
From the year 2012 to 2015, he studied in the
University of Tokyo (Utokyo) for the Ph.D., where
his work focused on active metamaterials for
scanning radiation beams by using micro-machining fabrication process to
monolithically integrate antenna with MEMS. During 2010 to 2012, Yong
worked in Huawei as an electrical engineer in phased array antennas for base
stations. His research interests include active antennas, non-linear metamaterial,
and RF-MEMS.

0018-926X (c) 2018 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.