You are on page 1of 8

IET Microwaves, Antennas & Propagation

Research Article

Optimised 4 × 4 millimetre-wave antenna array ISSN 1751-8725


Received on 4th November 2016
Revised 11th April 2017
with DGS using hybrid ECFO-NM algorithm Accepted on 25th May 2017
E-First on 8th August 2017
for 5G mobile networks doi: 10.1049/iet-map.2016.0959
www.ietdl.org

Korany R. Mahmoud1,2 , Ahmed M. Montaser3


1Department of Electronics, Communications, and Computers, Faculty of Engineering, Helwan University, Cairo, Egypt
2National Telecommunications Regulatory Authority, Ministry of Communication and Information Technology, Giza, Egypt
3Electrical Engineering Department, Sohag University, Sohag, Egypt

E-mail: kurany_hameda@h-eng.helwan.edu.eg

Abstract: A novel optimised 4 × 4 microstrip patch antenna array operating in the 28−38 GHz frequency range for fifth
generation mobile networks is presented. The proposed structure is fed by the corporate feed network printed on the same side
of the substrate. To improve the array radiation characteristics, a defected ground plane structure (DGS), which acts as coupled
c-shaped in the ground plane, is used. The dimensions of the antenna array structure are optimised using a new hybrid
approach involving the enhanced central force optimisation (ECFO) and Nelder–Mead (NM) algorithms. To illustrate the
convergence capability of the ECFO-NM algorithm, the results are compared with those obtained using stand-alone ECFO and
particle swarm optimisation algorithms. The results show the effect of DGS on the antenna performance characteristics, i.e. the
radiation efficiency and return loss are improved, on average, by ∼17.14 and 69.2%, respectively. Furthermore, the realised
antenna gain was increased by 2.44 dBi with a negligible effect on the half-power beamwidth. The optimised antenna array
yielded a broadband impedance bandwidth higher than 32% for a reflection coefficient lower than −12 dB and a total realised
gain of up to 18.65 dBi, which makes it suitable for imaging applications and millimetre-wave wireless communications.

1 Introduction antenna array achieved an antenna gain of 14.71 dBi and −18.81 


dB side-lobe level (SLL) at 28 GHz. In [7, 8], the design of a four
In recent times, the discrepancy between spectrum shortage and element dual-band antenna/array operating at 28 and 38 GHz is
capacity requirements has become increasingly eminent due to the presented. In [7], the dual-band substrate integrated waveguide
explosive growth of mobile traffic demand. The bottleneck of antenna/array designed to achieve a maximum gain of 11.9 and
wireless bandwidth becomes a key problem of the next generation 11.2 dBi at 28 and 38 GHz, respectively. Similarly, in [8], the
of mobile networks (fifth generation – 5G). However, there is a results of the printed slot antenna array designed to show a stable
vast bandwidth available in the millimetre wave (mmWave) band gain with a value of 10.58 dBi in the first band at 28 GHz and
ranging from 30 to 300 GHz. Thus, mmWave communications 12.15 dBi in the second band at 38 GHz. However, these different
have been proposed to be an important part of the 5G mobile designs presented in [6–8] addressed only single or dual-band
network to provide multi-gigabit communication services [1, 2]. In antenna arrays. Therefore, the design of broadband and high gain
an effort to globally align the standardisation of mmWave four-element antenna arrays with an electromagnetic bandgap
frequencies, the International Telecommunication Union (ITU) (EBG) ground structure-feeding network was proposed in [9, 10].
released a list of the proposed globally applicable frequencies In [9], the proposed broadband four-element tapered slot antenna
between 24 and 86 GHz (24.25–27.5, 31.8–33.4, 37–40.5, and array exhibited stable end-fire radiation characteristics within the
40.5–42.5 GHz) [3]. In addition, the Federal Communications operating frequency band ranging from 28 to 38 GHz. The array
Commission (FCC) issued a notice of proposed rulemaking that design parameters were tuned to achieve a 20 dBi gain at 30 GHz,
proposed new flexible service rules among the 28, 37, 39, and 64– and an impedance bandwidth that covers the entire band but at the
71 GHz bands [4]. expense of increasing the design's dimensions and complexity.
The short wavelength at mmWave frequencies allows creating Consequently, a new dense dielectric (DD) patch array antenna
compact antenna structures and enables the use of a large number prototype with an EBG ground structure is presented in [10]. It was
of antenna elements in small dimensions, which is convenient for found that the measured impedance bandwidth of the DD patch
integration in smart phones, wireless LAN bridges, and tablets. In array antenna ranged from 27 GHz to more than 32 GHz for a
addition, high-gain mmWave antenna arrays enable multi- reflection coefficient of less than 10 dB. However, this array
gigabits/sec transmissions over exemplary indoor distances. There structure suffers from high design complexity, since the antenna
have been various reports in the open literature of mmWave array consists of multiple layers, which also decreases radiation
antennas achieving effective operation at one or more of the 5G efficiency. Designing an mmWave antenna array involves reducing
bands. In [5], broadband and 28/38 GHz dual-band printed slot/ the mutual coupling between adjacent radiating elements. Mutual
monopole antennas are presented. The designed broadband antenna coupling reduction accordingly increases the system's overall
had an impedance of −10 dB starting from 20 GHz up to >40 GHz performance, resulting in higher gain efficiency [11]. Defected
with an average gain of 4 dBi, whereas the dual band antenna had a ground structure (DGS) is one of the most common techniques that
gain of 3.5 and 4.5 dBi at 28 and 38 GHz, respectively. However, has been applied between the antenna array radiating elements to
to overcome the path loss caused by the atmospheric absorption of reduce the coupling effects [12]. DGS has been widely used in
electromagnetic waves at higher frequencies, there is a need for filters, coplanar waveguides, microwave amplifiers, and antennas
designing antenna arrays in order to achieve a higher gain. to improve their performance. It has the advantages of reducing the
Therefore, various works related to designing 5G antennas/arrays size of components, enhancing the operating bandwidth and gain,
have been recently published [6–10]. Recently, a 1 × 7 microstrip and suppressing higher order harmonics and unwanted cross-
linear antenna array with parasitic patches in the 28 GHz sub- polarisation (X-pol) [13, 14]. The defected ground structure has
millimetre-wave frequency band was designed [6]. The designed taken many proposed shapes and dimensions depending on the

IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523 1516
© The Institution of Engineering and Technology 2017
operating frequency, the bandwidth, and the required application Therefore, in [27], the DS size is adaptively reduced around the
[15–17]. In [15], the c-shaped slot was used in the partial ground probe's location with the best fitness after every kth step, in which
plane for its notch filtering characteristics at the 5–6 GHz the DS's boundary coordinates are halved in a coordinate-by-
frequency range. Various couplings of the c-shaped slot etched in coordinate basis.
ground plane structures were studied in [16] for the design of Recently, an effective technique for adaptively ‘compressing’
cross-coupled resonator filters. In addition, different external the DS named dynamic threshold optimisation (DTO) was
coupling mechanisms using a c-shaped slot in the ground plane presented in [28]. This approach was based on compressing the DS
were investigated in [17]. from below in the direction of the dependent variable using a series
In the last few decades, many optimisation techniques have of successively increasing ‘thresholds’, instead of shrinking the DS
been performed successfully in the field of antenna design, such as as in [24, 27]. Bounding the DS from below removes local
genetic algorithm [18, 19], particle swarm optimisation (PSO) [20– maxima, and as the threshold, or ‘floor’, is increased, more and
22], bacterial swarm optimisation (BSO) [23], and central force more local maxima are eliminated. In [28], DTO was applied to
optimisation (CFO) [24]. In [15], different optimisation techniques Schwefel's problem in 2 and 30 dimensions using CFO. The
such as PSO, BSO, and CFO were considered in order to design an DTO/CFO algorithm returned a best overall fitness with an error of
E-shaped patch antenna for Bluetooth/notched UWB applications. 2.688×10−5%. In DTO, the threshold value (T) is subsequently
To improve the convergence capability of these optimisation updated by applying the following auxiliary function:
techniques, many hybrid algorithms have been proposed to
synthesise both algorithms’ strengths [25–27]. The main idea is to g x = f x − Tk − 1 ⋅ U f x − Tk − 1 + Tk − 1 (1)
integrate the ability of exploitation of a global optimisation
algorithm with the ability of exploration of another algorithm. In where U[ . ] is the unit step function. Thus, for f x ≥ T,
[25, 26], a simple Nelder–Mead (NM) algorithm, which is a local g x = f x , whereas for f x < T, g x = T. DTO was
optimisation technique, was used along with a global BSO implemented with a P number of passes with a progressively
algorithm in order to explore the search space for the design of a increasing threshold until a user-specified termination criterion was
bow-tie antenna for 2.45 GHz and a hexa-band planar inverted F- met. On each successive pass, DTO changes the topology of the
antenna, respectively. In [27], a hybrid approach involving CFO decision space being searched as follows:
and NM algorithms for accurately determining the resonant
frequency and performing the feed-point calculation of rectangular Cthk
microstrip antenna elements with various dimensions and various T k = Fmin + Fbest − Fmin (2)
P
substrate thicknesses is presented. Recently, a novel deterministic
framework of CFO, called dynamic threshold optimisation (DTO) where k = 1, …, P (no threshold applied for k = 0), and Fbest and
was proposed in [28] to be an effective technique for adaptively
changing the topology of the decision space in a multidimensional Fmin are the best and worst overall fitnesses returned after pass k,
search and optimisation problem. respectively. The coefficient Cth was set to 0.98 [28]. In this paper,
In this study, a novel broadband millimetre-wave microstrip CFO exponents α = β = 2, the reposition factor Frep was set to 0.5,
patch antenna array is proposed to resonate from 28 to 38 GHz for and the gravitational constant Go = 1 [24, 27–29].
a matched input impedance (Zin) of 50 Ω. The array antenna Once the global optima are found using the ECFO algorithm,
dimensions were optimised using a new hybrid approach involving the NM local optimisation technique [30] is performed to increase
enhanced CFO (ECFO) based on the DTO approach and the NM exploitation ability. The NM method thus falls in the general class
algorithm. Our results will be compared with those obtained using of direct search methods. It is based on the comparison of the
stand-alone ECFO and a well-known PSO algorithm to show the function values at the (D + 1) vertices for D-dimensional decision
convergence capability of the proposed technique. In addition, an variables. The simplex method has essentially four possible steps
appropriate defected ground plane structure is introduced to during each iteration: reflection, contraction in one dimension,
enhance the resonant frequencies and bandwidth of the antenna contraction around the low vertex, and expansion. Four scalar
array. The proposed design was analysed completely using CST parameters must be specified to define a complete NM method:
Microwave Studio and linked with the ECFO-NM algorithm, coefficients of reflection (ρ), expansion (χ), contraction (γ), and
coded in MATLAB, to optimise the antenna dimensions. Finally, shrinkage (σ). The nearly universally selected values used in the
our results are compared with those obtained using the finite- standard NM algorithm are ρ = 1, χ = 2, γ = 0.5, and σ = 0.5 [30].
difference time-domain (FDTD) program written in MATLAB and The main steps of the ECFO-NM algorithm are summarised in the
with the measured results to verify the validity of the numerical flowchart shown in Fig. 1.
simulations.
The remaining of this paper is organised as follows. Section 2 3 Antenna array geometry and configuration
provides a brief literature overview of the proposed hybrid ECFO-
NM algorithm. In Section 3, the description of proposed antenna Fig. 2 shows a schematic of the structure of the proposed 4 × 4
array configuration is explained. In Section 4, the simulated and microstrip patch array antenna, in which each single element
measured results are presented and discussed. Lastly, Section 5 consists of three radial stub-like patches with a tapering towards
contains the general conclusions of this paper. the outer end. This geometry was designed and constructed on a
50 × 50 mm2 Rogers® Duroid™ RT5880 with a 0.508 mm
substrate thickness, relative permittivity ɛr = 2.2, and loss tangent
2 Hybrid ECFO-NM algorithm
tanδ = 0.0009. In the proposed design, the spacing between every
The goal of any optimisation technique is to formulate an efficient two elements is 0.75λ in both the x and y directions. Other array
method to navigate through a large parameter space in order to find antenna parameters will be optimised within the assigned decision
the best set of values. One of the approaches to improve the spaces, as summarised in Table 1. The inset impedance of the
performance of global optimisation techniques is to limit the antenna was chosen to be 50 Ω for constructing a corporate feed
parameter's decision space (DS) range. In [24], their developed network; this corporate feed network is utilised for power division
improved CFO algorithm includes an adaptive strategy that between antenna elements. In this paper, a microstrip transmission
periodically shrinks the DS around the probe with the best fitness line feed was chosen for integrating with millimetre-wave circuits
by making the independent variable's domain shrink from the [31].
‘sides’ to improve convergence rate. The study is based on Fig. 2b shows the bottom view of the proposed antenna array,
determining where along the diagonal of the orthogonal plane the illustrating the structure of the defected ground plane. The DGS
probe array should be placed (γbest) by selecting the best value of includes 12 unit cells that are arranged in a special manner mainly
parameter γ, which starts from 0 and can go up to 1 in increments below the main feeding line. These are used for reducing the
of 0.1. This means excessive runs should be performed first for mutual coupling between antenna array elements in addition to
each optimised problem to determine its corresponding γbest. reducing the losses caused by surface waves and unwanted

IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523 1517
© The Institution of Engineering and Technology 2017
Fig. 1  Flowchart showing the main steps of the hybrid ECFO-NM algorithm

responses. Each unit cell contains two opposed c-shaped slot 38 GHz
configurations, which concentrate the magnetic and electric fields Objective function = max ∑ [G( f ) + S11( f ) ] (3)
near the short-circuit ends and the opposite side, respectively. Both f = 28 GHz
positive and negative coupling coefficients can be realised easily.
This helps to reduce the mutual coupling between antenna array The antenna array dimensions were first optimised using the PSO
elements in the top side of antenna, increases the overall system and stand-alone ECFO algorithms with an evaluation number of
performance, results in higher gain and high efficiency, and makes 20,000. We found that slightly better results were obtained using
the structure more compact. The DGS dimensions have been the stand-alone ECFO algorithm when compared with the PSO
parametrically optimised for the frequency range of 28–37 GHz. algorithm in terms of return loss and realised gain, as depicted in
Table 2 contains the dimensions of one element of the DGS, in Fig. 3. Then, the hybrid ECFO-NM algorithm was used with a total
which the inter-element spacing among DGS elements is set to evaluation number of 10,750 (10,000 evaluations using the ECFO
1.32 mm. algorithm and the remaining 750 using the NM algorithm). We
The microstrip patch antenna array was analysed completely found that better results were obtained, since the average gain and
using CST Microwave Studio. Then, CST Microwave Studio was return loss across the entire assigned band were improved by 26.54
linked with the ECFO-NM algorithm, which was coded in and 37.88%, respectively, compared with stand-alone ECFO.
MATLAB, to optimise the dimensions of the proposed antenna Fig. 4 shows a comparison between the normalised objective
array. Interchanging information between CST Microwave Studio values of the PSO, ECFO, and hybrid ECFO-NM algorithms
and MATLAB is fully described in [32]. In order to validate the plotted against iterations number. It should be noted that the
numerical results obtained from the CST Simulator, the proposed ECFO-NM algorithm outperformed the stand-alone ECFO and
antenna array was simulated using the FDTD method written in classical PSO algorithms by 8.62 and 34.84%, respectively, in
MATLAB and the results were compared with the measurements. addition to requiring 46.25% less iterations. The comparison of the
simulations shows that the ECFO-NM optimisation technique has
the ability to enhance the global search capability of the original
4 Simulated and measured results ECFO method and effectively improves its convergence capability.
To design a broadband microstrip patch antenna array operating in The ECFO-NM method enhanced the search capability by 24.34%
the frequency range of 28–38 GHz, the applied objective function compared with the ECFO method, which stopped improvement
should focus on maximising antenna gain (G) and minimising after iteration 202, and also by 52.67% compared with the PSO
return loss (S11) as follows: method which stopped improvement at iteration 128.

1518 IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523
© The Institution of Engineering and Technology 2017
Fig. 3  Comparison of return loss and antenna gain of the proposed 4 × 4
patch array antenna using PSO, ECFO, and hybrid ECFO-NM algorithms
(a) Return loss, (b) Antenna gain

performance of each element in the microstrip patch array antenna


over a wide frequency band. Evidently, the surface current
Fig. 2  4 × 4 Patch antenna array layout distributions are fairly stable at the three frequencies, which
(a) Patch antenna array (top view), (b) DGS (bottom view) indicate that the overall performance of the elements in the
microstrip patch array varies slightly with frequency.
Fig. 5 displays the simulated excited surface currents over the The analysis results for surface current distribution shown
microstrip patch array antenna at 28, 33, and 38 GHz. The above indicate that the characteristics of metal surface current
simulated results were obtained using CST Microwave Studio. The distribution at the low frequency (28 GHz), medium frequency (33 
behaviour at these selected frequencies can properly illustrate the GHz), and high frequency (38 GHz) bands are different. The

Table 1 Decision space for each parameter of the proposed array antenna and the corresponding optimised value using PSO,
stand-alone ECFO, and the hybrid ECFO-NM algorithms
Variable, mm Decision space Initial value Optimised values
From To PSO Stand-alone ECFO Hybrid ECFO-NM
a 1 1.5 1.25 1.15 1.08 1.21
a1 2 8 4 4.82 5.26 3.838
a2 8 10 8.5 9.3 8.87 9.616
a3 0.25 1 0.25 0.34 0.56 0.5
a4 0.15 0.25 0.2 0.23 0.205 0.205
a5 1.5 2 1.5 1.53 1.85 1.746
a6 18 25 20 22.74 21.65 20.841
a7 0.25 0.75 0.5 0.358 0.674 0.498
a8 0.15 0.25 0.2 0.195 0.201 0.215
b 0.5 1 0.75 0.693 0.738 0.789
b1 2 3 2.5 2.46 2.63 2.497
b2 0.1 0.2 0.15 0.174 0.179 0.185
b3 1.5 2.5 2 2.16 1.739 1.825
b4 5.5 7.5 6.5 6.38 6.184 5.966
b5 0.25 0.75 0.5 0.379 0.418 0.498
b6 0.15 0.25 0.2 0.175 0.201 0.215
b7 2 5 3.5 3.02 2.96 2.44
b8 16 20 18 17.35 19.64 16.935

Table 2 Dimensions of the DGS in the ground plane


Parameter L1 L2 L3 L4 L5 W1 W2 W3 W4 L6
value, mm 4 0.25 1 0.5 2 6 0.5 3 1 0.65

IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523 1519
© The Institution of Engineering and Technology 2017
current is a travelling-wave current in the medium and high omnidirectional radiation characteristic in the H-plane and a low
frequency bands, while it is standing wave current in the low SLL in the E-plane. Thus, the plane perpendicular to the current
frequency band. As a consequence, and determined from empirical direction has relatively low radiation levels, while the SLL in the
analysis, when the travelling-wave current has an electrical length E-plane is relatively high. As a result of this, the radiation patterns
longer than the wavelength, the angle between the maximum begin to split up.
radiation direction and the current direction is relatively small Operating in the low-frequency region (at 28 GHz) can
(<90°). When the standing wave current has an electrical length guarantee the omnidirectional radiation characteristic of the
shorter than the wavelength, the maximum radiation direction is proposed antenna and the radiation patterns not splitting up.
perpendicular to the current direction. At the same time, it has However, when operating in the high frequency region (at 38 
GHz), the splitting phenomenon occurs in the radiation patterns.
Simulation results in far-field radiation patterns have also proved
this. Therefore, if a wide radiation pattern across a large bandwidth
is required, miniaturisation is necessary. This fact further implies
that the proposed array should have stable radiation attributes and
impedance characteristics over a wide frequency band. Fig. 6
shows the calculated E-plane and H-plane co-polarisation and X-
pol radiation patterns for the suggested patch array antenna at
various frequencies (28, 33, and 38 GHz). As anticipated, the
figure clearly depicts that the patch array has a directional radiation
pattern in the horizontal H-plane, whilst the E-plane has broadside
and approximately symmetric radiation patterns.
A photograph of the fabricated patch array antenna optimised
Fig. 4  Comparison of the normalised objective value for PSO, ECFO, and
using the ECFO-NM algorithm is shown in Fig. 7. The
hybrid ECFO-NM algorithm versus iterations number
measurements of S11 and the array antenna gain are presented in

Fig. 5  Surface current distribution on the microstrip patch array antenna


for various frequencies
(a) At 28 GHz
(b) At 33 GHz
(c) At 38 GHz Fig. 6  Continued

1520 IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523
© The Institution of Engineering and Technology 2017
Fig. 8  Comparison of the measured and simulated values of S11 and
antenna gain of the proposed 4 × 4 patch array antenna with and without
DGS
(a) Return loss, (b) Antenna gain

Figs. 8a and b, respectively, and compared with the values


produced using CST Microwave Studio and the FDTD method in
MATLAB. The simulations and experimental results are in good
agreement. It can be noted that the antenna array optimised with
the hybrid ECFO-NM method yielded a maximum gain of 18.65 
dBi and an average gain of 15.45 dBi within the entire bandwidth
ranging from 28 to 38 GHz, which satisfies the requisites for short-
range indoor wireless applications. The differences in the values
for the losses and gains obtained by the FDTD method and the
CST simulation come from the effect caused by the stability
adjustment in the FDTD method and the differences in the
mathematical mechanisms involved in the FDTD method and the
CST simulator (which is based on the finite integral method).
Sometimes the boundary conditions set by the perfectly matching
layer in the FDTD method does not absorb the full wave at high
frequencies, while the CST simulator completely absorbs the
waves of both high and low frequencies.
Fig. 6  Simulated E-plane and H-plane co-polarisation and X-pol Finally, the effects of DGS on antenna performance
radiation patterns for the suggested patch array antenna at different characteristics such as gain, SLL, front-to-back (F/B) ratio and X-
frequencies with and without the DGS pol levels, radiation and total efficiencies are presented and
(a) At 28 GHz without DGS, (b) At 33 GHz without DGS, (c) At 38 GHz without summarised in Table 3. As shown in the table, the total efficiency
DGS, (d) at 28 GHz with DGS, (e) At 33 GHz with DGS, (f) At 38 GHz with DGS at both frequencies was improved by ∼9% on average when DGS
is used. Moreover, the antenna gain was increased by 2.04 dBi at
28 GHz and 1.41 dBi at 38 GHz. In addition, the S11 was improved

Fig. 7  Photograph of the fabricated patch array antenna


(a) Front side, (b) Back side

IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523 1521
© The Institution of Engineering and Technology 2017
Table 3 Performance comparison of the suggested patch array antenna with and without DGS at 28 and 38 GHz
Patch antenna array
Without DGS With DGS
28 GHz 38 GHz 28 GHz 38 GHz
radiation efficiency, % 86.4 86.1 93 93.5
total efficiency, % 83 84.9 92.2 93.4
gain, dBi 15.9 16.2 17.94 17.61
S11, dB −13.2 −12.6 −17.77 −17.55
E-plane HPBW 30° 31.2° 30.1° 31°
SLL, dB −9.8 −10.2 −13 −12.3
X-pol level, dB 48.3 50.4 59.2 61.8
H-plane HPBW 11.9° 12° 12.8° 12.1°
SLL, dB −11.15 −10.6 −11.5 −10.9
X-pol level, dB 49.8 50.3 68.3 66.7

by ∼69.2% on average with a negligible effect on the half-power [3] Resolution Com6/20, Provisional Final Acts WRC-15. WRC-15 (pp.
424-426). Geneva: ITU. http://www.itu.int/dms_pub/itu-r/opb/act/R-ACT-
beamwidth (HPBW) in both the E-plane and H-plane. Furthermore, WRC.11-2015-PDF-E.pdf
the X-pol level was increased by 10.9 and 18.5 dB in the E-plane [4] Use of Spectrum Bands above 24 GHz for Mobile Radio Services, GN
and H-plane, respectively, at 28 GHz. The results obtained clearly Docket No. 14-177, Notice of Proposed Rulemaking, 15 FCC Record 138A1
show that the DGS reduced the mutual couplings between antenna (rel. Oct. 23, 2015)
[5] Haraz, O.M.: ‘Broadband and 28/38-GHz dual-band printed monopole/
array elements, in addition to reducing the losses due to surface elliptical slot ring antennas for the future 5G cellular communications’, J.
waves and unwanted responses. In these regards, the overall system Infrared Millim. Terahertz Waves, 2016, 37, (4), pp. 308–317
performance was improved. At the same time, an additional [6] Philip, A.D., Kwang-Seon, K., Woo-Jin, B., et al.: ‘Stacked microstrip linear
inductance is introduced because of the magnetic flux in the etched array with highly suppressed side-lobe levels and wide bandwidth’, IET
Microw. Antennas Propag., 2017, 11, (1), pp. 17–22
out aperture and the gap capacitance on the ground plane. [7] Nadeem, A., Osama, H., Muhammad, A.A., et al.: ‘28/38-GHz dual-band
Consequently, the surface impedance is changed because of the millimeter wave SIW array antenna with EBG structures for 5G applications’.
etched defect, which in turn alters the shield current distribution in Int. Conf. Information and Communication Technology Research (ICTRC),
the plane. The change in surface impedance changes the current May 2015, pp. 1–4
[8] Osama, M.H., Mohamed, M.M.A., Elboushi, A., et al.: ‘Four-element dual-
phase velocity, which in turn changes the apparent effective band printed slot antenna array for the future 5G mobile communication
permittivity. This change in the apparent effective permittivity networks’. IEEE Int. Symp. Antennas and Propagation & USNC/URSI
leads to the improvements of various parameters of the National Radio Science Meeting, July 2015, pp. 1–2
conventional planar antenna, such as enhancing the bandwidth of [9] Elboushi, A., Briqech, Z., Sebak, A.: ‘4-elements MMW array with EBG
feeding network’. Proc. IEEE Int. Symp. Antennas Propagation (APS/URSI),
the microstrip antenna, suppressing the X-pol level of the antenna, July 2013, pp. 162–163
and increasing the antenna gain and radiation efficiency. [10] Haraz, O.M., Elboushi, A., Saleh, A.A., et al.: ‘Dense dielectric patch array
antenna with improved radiation characteristics using EBG ground structure
and dielectric superstrate for future 5G cellular networks’, IEEE Access,
5 Conclusion 2014, 2, pp. 909–913
[11] Zhu, F.-G., Xu, J.-D., Xu, Q.: ‘Reduction of mutual coupling between closely
In this paper, a novel millimetre-wave 4 × 4 patch antenna array packed antenna elements using defected ground structure’, Electron. Lett.,
with defect slots in the ground plane was designed to operate 2009, 45, (12), pp. 601–602
efficiently within the frequency range of 28–38 GHz for 5G [12] Hammoodi, A.I., Al-Rizzo, H.M., Isaac, A.A.: ‘Mutual coupling reduction
wireless communications systems. The dimensions of the array between two monopole antennas using fractal based DGS’. IEEE Inter. Symp.
Antennas and Propagation and USNC/URSI National Radio Science Meeting,
antenna structure were optimised using a new hybrid ECFO-NM Vancouver, Canada, 19–25 July, 2015, pp. 416–417
algorithm. To illustrate the convergence capability of the ECFO- [13] Biswas, S., Guha, D., Kumar, C.: ‘Control of higher harmonics and their
NM algorithm, the results were compared with those obtained radiation in microstrip antennas using compact defected ground structures’,
using stand-alone ECFO and PSO algorithms. We found that the IEEE Trans. Antennas Propag., 2013, 61, (6), pp. 3349–3354
[14] Zulkifli, F.Y., Rahardjo, E.T., Hartanto, D.: ‘Radiation properties
ECFO-NM algorithm enhanced the search capability by 24.34 and enhancement of triangular patch microstrip antenna array using hexagonal
52.6%, compared with the stand-alone ECFO and PSO algorithms, defected ground structure’, Prog. Electromagn. Res. M, 2008, 5, pp. 101–109
respectively, in addition to saving the required evaluation steps by [15] Montaser, A.M., Mahmoud, K.R., Elmikati, H.A.: ‘Integration of an
46.25%. The simulated results were compared with the optimized E-shaped patch antenna into laptop structure for bluetooth and
notched-UWB standards using optimization techniques’, Appl. Comput.
measurements and with other simulated results obtained using an Electromagn. Soc. (ACES) J., 2012, 27, (10), pp. 786–794
FDTD program coded in MATLAB. The measurements of realised [16] Ali, A.R., Abdel-Rahman, A., Amari, S., et al.: ‘Direct and cross-coupled
gain and return loss are in good agreement with the results obtained resonator filters using defected ground structure (DGS) resonators’. 2005
from the simulations using the CST and FDTD programs. The European Microwave Conf., 4–6 October 2005, vol. 2, pp. 3
[17] Abdel-Rahman, A., Ali, A.R., Amari, S., et al.: ‘Compact bandpass filters
optimised array antenna achieved a maximum gain of 18.65 dBi using defected ground structure (DGS) coupled resonators’. IEEE MTT-S Int.
and an average of 15.45 dBi within the entire bandwidth for a Microwave Symp. Digest, 2005, June 2005, pp. 12–17
reflection coefficient lower than −12 dB, which satisfies the [18] Altshuler, E.E., Linden, D.S.: ‘Wire antenna designs using genetic
requisites for short-range indoor wireless applications. algorithms’, IEEE Antennas Propag. Mag., 1997, 39, (2), pp. 33–43
[19] Rahmat-Samii, Y., Michielssen, E.: ‘Electromagnetic optimization by genetic
algorithms’, Microw. J., 1999, 42, (11), pp. 232–232
6 Acknowledgment [20] Robinson, J., Rahmat-Samii, Y.: ‘Particle swarm optimization in
electromagnetics’, IEEE Trans. Antennas Propag., 2004, 52, (2), pp. 397–407
The authors acknowledge the Electronics Research Institute (ERI), [21] Liu, W.: ‘Design of multiband CPW-fed monopole antenna using a particle
Microstrip Department for the support and cooperation during swarm optimization approach’, IEEE Trans. Antennas Propag., 2005, 53,
(10), pp. 3273–3279
simulation of this research. [22] Zainud-Deen, S.H., Mahmoud, K.R., El-Adawy, M., et al.: ‘Design of Yagi-
Uda antenna and electromagnetically coupled curl antenna using particle
7 References swarm optimization algorithm’. 22th National Radio Science Conf. (NRSC
2005), March 2005, pp. 115–124
[1] Elkashlan, M., Duong, T.Q., Chen, H.-H.: ‘Millimeter-wave communications [23] Zainud-Deen, S.H., Ahmed, S., Mahmoud, K.R.: ‘Design of a notched ultra-
for 5G: fundamentals: Part I [Guest Editorial]’, IEEE Commun. Mag., 2014, wideband antenna with irregular radiator shape using bacterial swarm
52, (9), pp. 52–54 optimization’. 30th National Radio Science Conf. (NRSC 2013), April 2013,
[2] Elkashlan, M., Duong, T.Q., Chen, H.-H.: ‘Millimeter-wave communications pp. 109–116
for 5G – Part 2: applications’, IEEE Commun. Mag., 2015, 53, (1), pp. 166– [24] Formato, R.A.: ‘Improved CFO algorithm for antenna optimization’, Prog.
167 Electromagn. Res. B, 2010, 19, pp. 405–425

1522 IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523
© The Institution of Engineering and Technology 2017
[25] Mahmoud, K.R.: ‘Design optimization of a bow-tie antenna for 2.45 GHz
RFID readers using a hybrid BSO-NM algorithm’, Prog. Electromagn. Res.
PIER, 2010, 100, pp. 105–117
[26] Mahmoud, K.R., Montaser, A.M., Elmikati, H.A.: ‘Design of hexa-band
planar inverted-F antenna using hybrid BSO-NM algorithm for mobile phone
communications’, Int. J. RF Microw Comput Aided Eng., 2013, 23, (1), pp.
99–110
[27] Mahmoud, K.R.: ‘Central force optimization: Nelder-Mead hybrid algorithm
for rectangular microstrip antenna design’, Electromagnetics, 2011, 31, pp.
578–592
[28] Formato, R.A.: ‘Dynamic threshold optimization-a new approach?’, arXiv
preprint arXiv:1206.0414, 2012, pp. 1–17
[29] Mahmoud, K.R.: ‘Synthesis of unequally-spaced linear array using modified
central force optimisation algorithm’, IET Microw. Antennas Propag., 2016,
10, (10), pp. 1011–1021
[30] Nelder, J.A., Mead, R.: ‘A simplex method for function minimization’,
Comput. J., 1965, 7, pp. 308–313
[31] Ely, L., Gabi, M., David, T.: ‘A study of microstrip array antennas with the
feed network’, IEEE Trans. Antennas Propag., 1989, 37, (4), pp. 426–434
[32] Montaser, A.M., Mahmoud, K.R., Elmikati, H.A.: ‘An Interaction study
between PIFAs handset antenna and a human hand-head in personal
communications’, Prog. Electromagn. Res. B, 2012, 37, pp. 21–42

IET Microw. Antennas Propag., 2017, Vol. 11 Iss. 11, pp. 1516-1523 1523
© The Institution of Engineering and Technology 2017

You might also like