Journal of ELECTRICAL ENGINEERING, VOL 71(2020), NO5, 347–352

sciendo
PAPERS
Practical design of Ku band Vivaldi antenna array
Jakub Prouza, Zbyněk Raida1

In the paper, a practical design of a Ku-band Vivaldi antenna array for compact radar system is presented. The realized
gain, the beam width, polarization purity and the possible electronic beamforming in the horizontal plane were the most
important requirements. Since the array was requested to show an enhanced mechanical stability, a novel geometry of elements
was proposed. The Vivaldi slot was created on two substrates connected by metallic vias suppressing surface currents, and
the microstrip feeder was placed in-between those substrates. Simulations are based on special approach in CST Microwave
using infinite array of antenna elements, which should reduce computing time.
K e y w o r d s: antenna array, Ku band, Vivaldi antenna

1 Introduction In [5], authors discussed a linear array of antipodal el-

ements with an overlap broadening the impedance band-
The Vivaldi antenna array is requested to provide a width of the array. Due to the antipodal structure, polar-
high gain and an exactly given shape of the radiation ization purity was not sufficient.
pattern. Antenna elements for vertical and horizontal po- In [6], the exponentially tapered slot on the top side
larizations have to be separated, and the beam control in of the substrate was excited by a microstrip transmission
the horizontal plane has to be provided by relative phase line on the bottom side. Authors discussed practical issues
shifting between the adjacent elements. The beam width of the design and fabrication of a dual-polarized array
in the horizontal plane should be around 5◦ . As the ar- consisting of 128 elements. Attention was turned to the
ray will be used in compact radar system, there is also light weight and compact size.
request to be as small as possible with power dividing sys- In [7], an array of 4 × 4 Vivaldi elements was designed.
tem creating one block. This work is a part of research On sides next to the tapered slot, grooves on both sides
and development of a new radar system, so not all details of the exponentially tapered slot were etched to suppress
and results can be published as they are part of intellec- surface currents and improve both impedance and radia-
tual property. tion characteristics of the antenna.
Vivaldi antennas were firstly described by Peter J. The practical design presented in the paper was aimed
Gibson in 1979 [1]. These antennas belong to a group of to develop a Vivaldi array with high polarization purity
slot antennas with longitudinal radiation [2]. The Vivaldi (the antipodal structure was denied), suppressed surface
antenna is created by an exponentially tapered slot in a currents (metallic vias on side walls were used) and in-
metallic layer (on a surface of a microwave substrate). creased mechanical stability (two-substrate Vivaldi ele-
Both sides of the Vivaldi slot antenna can be created ments with a microstrip excitation in between were de-
by a single metallic layer [3] on one side of a microwave veloped). According to our knowledge, such a concept has
substrate, or the antenna can be antipodal (sides of the not been published in the open literature yet.
slot are placed on different surfaces of a microwave sub- The design of a single antenna element is described,
strate [2]). The antipodal design suffers from a higher and several configurations of the antenna array are dis-
cross-polarization. cussed, and the feeding network is presented before the
In order to reduce the cross-polarization, the balanced prototype of the array is shown.
antipodal Vivaldi antenna was designed [4] using an addi-
tional substrate. On the new substrate, the metallic side
of the slot from the top surface is mirrored. 2 Vivaldi antenna array
The low limit of the operation frequency is related to
the width of the aperture. The Vivaldi antenna is linearly The designed Vivaldi element is created by an expo-
polarized with electric field intensity vector parallel to the nentially tapered slot etched into a metallic film on one
slot plane. surface of a microwave substrate. The element is fed by
In the open literature, several papers on the design of a microstrip line and a terminating stub on the opposite
Ku-band Vivaldi arrays can be found. side of the substrate (see Fig. 1).

1 Fakulta elektrotechniky a komunikačnı́ch technologiı́ VUT v Brně, [email protected], [email protected]

DOI:10.2478/jee-2020–0047, Print (till 2015) ISSN 1335-3632, On-line ISSN 1339-309X

c This is an open access article licensed under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http: //creativecommons.org/licenses/by-nc-nd/3.0/).
348 J. Prouza, Z. Raida: PRACTICAL DESIGN OF Ku BAND VIVALDI ANTENNA ARRAY

: ) ƒ 63DUDPHWHU G%
 




/

'


% 

& 
    4 ƒ 
6

Fig. 1. Layout of Vivaldi element (left), top side with Vivaldi slot Fig. 2. Dependency of active reflection coefficient on frequency and
(right) bottom side with microstrip feeder scanning angle

3DUDPHWHU9DOXH is given by the following equation (2) published for


example in [3]:
VS
y = C1 eR x + C2 . (2)

Here, C1 and C2 are given by (3) and (4) so the flare
\ goes through the both end points (x1 , y1 ) and (x2 , y2 ):

 y2
− y1
C1 = Rx
, (3)
− eRx1
e 2
y1 eRx2 − y2 eRx1
5 C2 = . (4)
eRx2 − eRx1
    2SWLPL]HU6WHS
In these equations, ending points (y1 , y2 ) also define
the length of the tapered slot.
Fig. 3. Converging of optimized parameters
• Slot aperture W : the parameter determines the lowest
operation frequency and is also given by the ending
When designing an array, mutual coupling between points (x1 , x2 ).
elements has to be considered, including the effect of • Spacing between elements SP : the parameter should
conductive connections among neighbouring antennas. be λ/4 approximately for the highest frequency. SP
In simulations, the following formula can be used to affects the virtual waveguide created by the grid of
calculate reflection coefficient of elements in the array [8] antenna elements.
The array of Vivaldi radiators was modelled in CST
X
N
an
Γact = Smn . (1) Microwave Studio and optimized by the default Trusted
n=1
am region method. In comparison with Particle swarm opti-
mization and Newton method, Trusted region was able
Here, m and n are numbers of elements in an array, to converge to the opti-mal solution in about 15 steps.
an is excitation coefficient and Smn is the coupling coef- Other two optimization methods needed more steps to
ficient. find optimal values. As an optimized parameter were cho-
Equation (1) can be extended by parameters of scan- sen parameters above. Below is a graph exported from
ning angles in case of a phased antenna array, and the CST Microwave showing converging of these pa-rameters
active reflection coefficient can be depicted in form of a with method of trusted region, see Fig. 3.
colour map (the value of coefficient is expressed depend- The unit cell the entire antenna array is formed from,
ing on frequency and scanning angle as shown in Fig. 2). consists of two identical antenna elements mutually ro-
Here, we can obtain reflection coefficient for the whole tated for 90◦ . That way, both required polarizations are
frequency band by the worst-case method. And the ex- obtained, see Fig. 4.
tracted reflection coefficient can be used as a parameter For a better mechanical stability of the array, the an-
of the objective function in optimization. tenna element was manufactured from two layers of a
• Opening rate R: the parameter of the exponentially microwave substrate. On each substrate, identical expo-
tapered slot influenc-ing matching in the whole fre- nential slots on the metalized surface were etched. The
quency band. It affects the shape of the slot, which feedline was designed on the common surface of layers,
 Journal of ELECTRICAL ENGINEERING 71(2020), NO5 349

Fig. 4. Cell of antenna array providing dual polarization Fig. 5. Antenna element consisting of two substrates.microstrip
feed placed in between substrates

*DFW G%
 ƒ
D ƒ ƒ
D

ƒ ƒ


*+]
ƒ ƒ
*+]


ƒ ƒ
*+]
ƒ ƒ
  I *+]  ƒ
*DFW G% ƒ

ƒ ƒ
E E

ƒ ƒ
*+]


ƒ ƒ

 *+]
ƒ *+] ƒ

ƒ ƒ
ƒ
  I *+] 

Fig. 6. Frequency response of reflection coefficient at the input of Fig. 7. Radiation pattern of 32 × 16 cells from single-layer perpen-
single-layer Vivaldi element (top) and double-layer one (bottom) dicular Vivaldi slots horizontal plane (top), vertical plane (bottom)

see Fig. 5. Due to the higher computational complexity one are depicted in Fig. 6. Simulations were performed for
of this model, the element was used in an antenna array scanning angles ±20◦ in the horizontal plane to verify the
with a reduced number of elements. possibility of the beam control.
The Vivaldi element is of the dimensions 26 mm ×
8 mm. Since the design is a part of intellectual property, 2.1 32 × 16 antenna array
all the dimensions of the basic antenna element are not An array consisting of 32 × 16 cells was created
allowed to be published. from single-layer perpendicular Vivaldi slots. This model
Simulated frequency responses of reflection coefficient should serve for a numerical analysis of the possibility to
at the input of a single-layer element and a double-layer obtain the required gain and beam width.
350 J. Prouza, Z. Raida: PRACTICAL DESIGN OF Ku BAND VIVALDI ANTENNA ARRAY

ƒ
ƒ ƒ


ƒ ƒ

*+]
ƒ ƒ
*+] *+]

ƒ ƒ

ƒ ƒ
ƒ

Fig. 8. Beam steering of 32 × 16 array by phase shifting Fig. 9. The array of 8 × 4 active elements surrounded by row of
passive elements

ƒ
ƒ ƒ central frequency 15 GHz. For better clarity of the shape
D of patterns, no gain values are given in figures.
ƒ ƒ The antenna array should provide the ability to control
the main beam in the horizontal plane by phase shifting
*+] *+] of neighbouring elements as demonstrated in Fig. 8. Here,
the main beam is steered by 20◦ in the horizontal plane.
ƒ ƒ
As shown in [3], the beam steering can cause the for-
mation of a parasitic side lobe on the opposite side of the
*+] main-lobe half-space. This effect can be suppressed by
ƒ ƒ the optimization of the slot length L . But in that case,
we can observe only the parasitic back lobe as in Fig. 7.
ƒ ƒ This back lobe can be further suppressed by an additional
ƒ shielding.

ƒ
ƒ ƒ ƒ
E ƒ ƒ
*+]
ƒ ƒ
ƒ ƒ

*+]
ƒ ƒ
ƒ ƒ
*+]

*+] *+]
ƒ ƒ
ƒ ƒ
*+]
ƒ ƒ
ƒ ƒ ƒ
ƒ

Fig. 10. Radiation pattern of 8 × 4 cells from double-layer perpen- Fig. 11. Beam steering of 8 × 4 array by phase shifting
dicular Vivaldi slots horizontal plane (top), vertical plane (bottom)

2.2 8 × 4 antenna array

Due to the size of the array, the distribution of am- To reduce the computational complexity, a smaller an-
plitudes on individual elements followed Taylor series to tenna array consisting of 8 × 4 cells was created from
suppress side lobes. The substrate Rogers RT5880 with double-layer perpendicular Vivaldi slots. The array was
0.254 mm thickness was used. simulated without a controlled amplitude distribution
Figure 7 shows radiation patterns in the horizontal to suppress side lobes. Figure 9 shows that there is an
plane (top) and the vertical planes (bottom) for three additional row of antenna elements terminated by 50 Ω
significant frequencies. A large number of side lobes are loads around the whole antenna array. These passive el-
visible. Their suppression is about 25 dB, the beam width ements help to minimalize boundary effects and ensure
5.4 ◦ in the horizontal plane and the gain is 27 dB at the a better suppression of side lobes. The substrate Rogers
 Journal of ELECTRICAL ENGINEERING 71(2020), NO5 351

Fig. 12. Implemented feeding network with on-chip power dividers Fig. 13. Complete antenna array measured in anechoic chamber

6 G% *UHO G%
 
*+]

+RUL]RQWDOHOHPHQWV 




*+]


*+]
9HUWLFDOHOHPHQWV

       4 ƒ  

I *+]

Fig. 14. Measured frequency response of reflection coefficeient: Fig. 15. Measured radiation pattern of vertical elements in hori-
vertical polarization (blue), horizontal polarization (red) zontal plane

RO4730JXR with 0.526 mm thickness was considered. (1024 active antenna elements for vertical and horizon-
The model was subsequently used to manufacture the first tal polarizations together). Hence, the design of the feed-
prototype. ing network is quite demanding. We used a model of the
The radiation pattern of the 8 × 4 array, see Fig. 10, multistage Wilkinson power divider and simulated it with
shows the expected decrease of the number of side lobes good results. These power dividers can be cascaded for all
as well as their worse suppression. The 13 dB side lobe antenna elements. But for the first prototype of the array
level agrees with theory - an antenna array of more than was decided to be fed by an conventional feeding network
eight elements cannot achieve better side lobe level than based on chip power dividers EPK2+ by Mini Circuits
13.26 dB [4]. Due to the smaller number of elements in the [9].
array, the beam width in the horizontal plane is 13.3◦ and
the gain at 15 GHz was decreased to 17.9 dB. Obviously
4 Prototype
worse parameters correspond to the smaller number of
elements and a uniform distribution of amplitudes. Radi-
In order to manufacture the prototype, the array of
ation pattern in Fig. 11 shows that the main beam steer-
8 × 4 elements was used. The array was completed by the
ing does not cause the extra parasitic side lobe mirrored
feeding system with two inputs for separate polarizations.
to the main beam. In order to implement the designed
The whole structure in the mounting bracket is shown in
arrays, a proper feeding network has to be developed.
Fig. 13. To reduce the costs, no main beam steering was
implemented in the first prototype.
3 Power dividing network
4.1 Measurements
The final array contains a high number of active an- The antenna array was measured in an anechoic cham-
tenna elements - 512 cells in case of the 32 × 16 array ber in the whole frequency band for both the polariza-
352 J. Prouza, Z. Raida: PRACTICAL DESIGN OF Ku BAND VIVALDI ANTENNA ARRAY

impedance matching, phase-shift steering and mechani-

*UHO G% cal stability. To meet the requirements, a double-layer

*+] antenna element with microstrip feeding in between lay-
ers was proposed. The antenna can be manufactured by
common etching of microwave substrates.

Parameters of the proposed array were verified by com-
puter simulations and validated by a simple experiment.
 For measurements, an array of 8 × 4 active elements was
surrounded by a row of passive elements. The array was
*+] completed by a feeding network based on chip power di-
*+] viders. Due to the problems with soldering, measured re-

sults were significantly worse compared to measurements.
Even the prototype did not meet the given require-
ments, there is a good assumption to achieve the required
    4 ƒ  
parameters with proper modifications of the power divid-
ing network. Network optimization could be as demand-
Fig. 16. Measured radiation pattern of horizontal elements in hor- ing as the design of the antenna array.
izontal plane After solving problems, the antenna array can be com-
pleted by the system of distributed amplitudes for bet-
tions. Impedance characteristics were measured on both ter side lobe suppression and by phase shifters to obtain
the inputs of the antenna array for vertical and horizon- beam steering in the horizontal plane.
tal elements (see Fig. 14). Obviously, impedance match-
ing of the antenna array is not acceptable. In laboratory, Acknowledgments
we revealed problems with soldering chip power dividers The presented research was supported by the Inter-
causing these bad values. nal Grant Agency of Brno University of Technology, the
Considering the measured impedance matching, the project no. FEKT-S-20-6526.
gain was measured only at lower frequencies. As expected,
the obtained gain deviated from simulations. At 12 GHz,
the measured gain of vertical elements was 8.15 dBi com- References
pared to 16 dBi in simulation.
[1] P. J. Gibson, “The Vivaldi Aerial”, 9th European Microwave
Nevertheless, the measurement was useful. Fig. 15 Conference, pp. 101–105, 1979.
shows that we were able to obtain the main beam width [2] C. A. Balanis, Antenna theory: Analysis and design, 3rd ed.,
of 14 ◦ in the horizontal plane (the essential one) for ver- Hoboken: Wiley-Interscience, 2005.
tical elements. The side-lobe suppression at 12 GHz with [3] F. B. Gross, Ed. Frontiers antennas, New York, McGraw-Hill,
the value of 12.8 dBi is also satisfactory (the simulated 2011.
value was 13 dBi). The values of gain are related to the [4] J. D. S. Langley, P. S. Hall, and P. Newham, “Balanced Antipo-
dal Vivaldi Antenna for Wide Bandwidth Phased Arrays”, IEE
main lobe gain (8.15 dBi). Proceedings – Microwaves, Antennas and Propagation, vol. 143,
Similarly, Fig. 16 shows the gain in the horizontal no. 2, 1996.
plane for horizontal elements. The gain at 12 GHz was [5] M. Huang, L. Zhang,W. Wang, and Qiao, “An Ultrawideband
4.8 dBi which is insufficient. The side-lobe suppression Tightly Coupled Antipodal Vivaldi Antenna Array for UHF-Ku
at 12 GHz was 10 dB and the main beam width was Band Applications”, pp. 1323–1324, 2017.
[6] R. Hahnel and D. Plettemeier, “Dual-polarized Vivaldi Array for
17◦ showing a reasonable agreement with simulations. At
X- and Ku-Band”, Proceedings of the 2012 IEEE International
13.5 GHz, there is a noticeable high level of side lobes Symposium on Antennas and Propagation, pp. 1–2, 2012.
caused by impedance mismatch of central elements for [7] E. Hanbay and M. E. Aydemir, “High Gain Ultrawide Band Vi-
this frequency. valdi Antenna Design for Mini/Micro Satellite Synthetic Aper-
ture Radar Applications”, 9th International Conference on Re-
cent Advances Space Technologies (RAST), pp. 491–495, 2019.
5 Conclusion [8] R. C. Hansen, Phased array antennas, 2nd ed., Hoboken, N. J.,
Wiley, 2009.
[9] Power Splitter/Combiner EP2K+ Mini-Circuits.
In the paper, a concept of an antenna array based on
Vivaldi antenna elements was presented. The design was Received 20 April 2020
optimized for requested gain, beam width, sidelobe level,