New Approach in Antenna Design Automation

Applied to a Dual-Band GNSS Micro-Array

Lionel Rudant, Lotfi Batel, Antonio Clemente, Christophe Delaveaud
Univ. Grenoble Alpes, CEA, Leti, F-38000 Grenoble, France, [email protected]

Abstract—This article describes a new approach in antenna have limited prospects for integration in compact systems
design automation. We propose to use a micro-array of due to their cost.
coupled antennas loaded by impedances and to perform the A popular GNSS antenna architecture, useful for
optimization of these impedance values in order to reach
integration in low-cost devices, is the single-fed patch with
specific design goals. In particular, we present a demonstration
for dual-band GNSS that requires good RHCP gain (in the truncated corner or slit cut. The basic operation is a coupling
range -2 dBic to 0 dBic) and low LHCP gain (below -10 dBic). of two orthogonal radiating modes within the patch cavity. In
The proposed automated design process has been able to meet [8], the authors present a multi-band single-fed staked-patch
these requirements jointly in L2 (1215–1240 MHz) and L1 geometry. The mastering of the geometry optimization is a
frequency bands (1560–1610 MHz). A prototype has been challenge for the designers due to a very large number of
fabricated, and characterized in anechoic chamber to validate parameters. Consequently, the antenna remains an issue, as
the proposed design methodology.
its design is a long process and cannot be easily handled in
Index Terms—micro-array, parasitic radiating elements, the product design phase.
antenna coupling, flow diagram, GNSS, multi-band, antenna In this paper, we aim to demonstrate that design
design automation, optimization algorithm. automation tool is a promising technology in order to keep
antenna low-cost and to achieve the desired performances
I. INTRODUCTION (gain, polarization, bandwidth, etc.). The paper is organized
Multi-band Global Navigation Satellite Systems (GNSS) as follows. In Section II, we present the principles of our
is a key technology for future mobility services as it approach for antenna design automation. Then, in Section
improves accuracy and reliability of geolocation. Multi-band III, we present the results achieved in the case of the
GNSS antennas must meet specific requirements on radiation optimization of a dual-band GNSS antenna. In Section IV,
properties, such as optimized Right Hand Circular prototype validation results are discussed. In the conclusion,
Polarization (RHCP) gain towards the sky, as well as low Section V, we open discussion about our next steps and the
cross-polarization Left Hand Circular Polarization (LHCP). global objectives of our roadmap.
Designers must optimize these properties jointly on several
wide frequency bands. As real-time geolocation is critical for
all mobility services, multi-band GNSS antenna are expected
to be integrated in a wide range of devices and vehicles. This
generates a need of affordable and compact antenna solutions
for mass production.
The recent state-of-the-art of GNSS antennas presents a
large number of designs based on ceramic patch [1],
quadrifilar helix [2] and crossed-dipoles [3]. The basic
operation principle is based a simultaneous excitation of
complementary radiating modes with the control of the
relative phases in order to generate the circular polarization
[4]. The typical architecture consists of a 4-element radiating
structure and a feeding network with quadrature properties
[5]. For multiband operation, designers proposed multi-
resonant radiating elements such as stacked-patch geometry
or multi-length helix. They designed compact wideband
feeding network based on hybrid couplers and sophisticated
microstrip arrays [6]. In [7], the authors present a simple
small quadrifilar helix combined with a dual-resonant
feeding circuit. In this approach, the radiating elements
remain very simple but the design of a dual-band feeding
network cannot be avoided. These complex antenna solutions Fig. 1. Antenna design automation based on the optimization of
impedance loads – application to GNSS with a micro-array of IFAs.

This paper's copyright is held by the author(s). It is published in these proceedings and included in any archive
such as IEEE Xplore under the license granted by the "Agreement Granting EurAAP Rights Related to
Publication of Scholarly
Authorized licensed use limitedWork."
to: CEA GRENOBLE. Downloaded on November 22,2021 at 16:36:17 UTC from IEEE Xplore. Restrictions apply.
 II. NEW APPROACH IN ANTENNA DESIGN AUTOMATION electromagnetic simulation or measurement of the micro-
A new generation of automated design tools appear in all array. Then the effects of the impedance loads can be solved
engineering domains thanks to the democratization of by classic resolution of flow diagram presented in [12].
optimization algorithms and machine learning. So far, the Using this model, we can compute the micro-array properties
automation of antenna encounters serious limitations due to with respect to the impedance values without any extra
difficult prediction of the effects of several geometrical simulation.
parameters on the antenna properties and behavior. When
based on complex 3D geometries and with the integration B. Automated optimization process
effects, antenna systems miss easy-to-compute closed-form The use of a coupled micro-array allows handling an
models. Consequently, time-consuming simulations are optimization process on impedance values with reasonable
required to generate the initial information prior to any time of computing. The optimization process consists in
machine learning or optimization process. Recently, finding the impedance values that minimize a specific cost
researchers reviewed software solutions for antenna design function.
automation [9]-[11] demonstrating strong expectations from For GNSS applications, the radiation properties of the
the community. antenna must meet requirements regarding gain and
polarization in the 3D space. Definition of a cost function
A. Modeling of coupled micro-array based on 3D complex information is few useable. Spherical
In this work, we use a particular antenna geometry based Wave Expansion (SWE) of the radiation pattern allows
on a micro-array of Inverted-F Antennas (IFA) (see the Fig. describing the full radiation properties by means of a
1). With closed integration of the IFAs, the coupling between complex distribution of radiating modes. In [14], we
the different elements is very high with deep impact on demonstrated the convenience of SWE for a relevant
impedance and radiation properties of the micro-array. description of compact antennas’ radiation as well as the
The use of coupled antennas allows the control of the description of the effects of integration.
directivity [12]. One typical example of coupled-antenna The typical radiation pattern of a dual-fed GPS patch
array is the Yagi-Uda dipole array, very popular solution to antenna is presented in Fig. 3. The SWE of this reference
achieve high-gain in the UHF band. The optimization of the radiation pattern is presented in Table I. We have a full
radiation properties relies on multiple dipoles’ lengths – description of this radiation pattern with two radiating modes
longer dipoles providing a reflector effect and shorter dipoles combined with equal amplitude and opposite phase. SWE
providing a director effect. In this work, all IFAs have the gives a convenient way to define the cost function for the
same geometry but they are loaded with different impedance optimization process. The cost function is the difference
values Zi (see Fig. 1). Both impedance and radiation between the reference SWE presented in Table I and the
properties of the final micro-array depend on these SWE of the micro-array at each step of the optimization
impedance values. process.

Fig. 2. Modeling of coupled micro-array by means of flow diagram [9].

In Fig. 2, we present the flow diagram of a two-element

micro array [12]. Here, u is the signal at the generator
(antenna in transmitting mode). The terms ai are the Fig. 3. Reference radiation pattern for GNSS (as a function of elevation
incoming signals to the micro-array and bi are the outgoing angle θ), blue cruve is for RHCP and red is for LHCP (not visible here).
signals from the micro-array. ai and bi are related to the
scattering matrix coefficients Sij. Ei is the active element
pattern with respect to the port i [13]. Ei is defined as the TABLE I. SWE OF THE REFERENCE RADIATION PATTERN
radiation pattern of the micro-array when the port i is excited Radiating Mode Amplitude Phase
and all the other ports are loaded with 50 Ω. Γi is the
TE-11 1 0°
reflection coefficient at the port i depending of the
impedance value loading on the port i. Scattering matrix and TM-11 1 180°
active patterns can be extracted with a preliminary full-wave The normalized power of the radiating modes is 0.5 W.

This paper's copyright is held by the author(s). It is published in these proceedings and included in any archive
such as IEEE Xplore under the license granted by the "Agreement Granting EurAAP Rights Related to
Publication of Scholarly
Authorized licensed use limitedWork."
to: CEA GRENOBLE. Downloaded on November 22,2021 at 16:36:17 UTC from IEEE Xplore. Restrictions apply.
C. Sum-up of the full process of optimization seven footprints for the loading of IFA elements by means of
In this part, we summarize the main components of our surface-mounted components. A metallized via connects one
approach in antenna design automation. We use a coupled IFA to the opposite side of the PCB where a 50 Ω SMA
micro-array of IFAs. Each of the IFAs is loaded with a connector is mounted. The full dimension of the antenna
specific optimized impedance load. The effect of these geometry is 35 mm  35 mm  9 mm.
impedance loads can be efficiently computed based on the We perform electromagnetic simulations with multiple
flow diagram resolution without any electromagnetic excitations at the load footprints in order to bring all active
simulation. Consequently, we can use an optimization patterns to the model presented in Section II-A. We also
algorithm (e.g. convex optimization, etc.) to find the export the scattering matrix to model all coupling effects.
impedance values that give the better radiation pattern for
GNSS. The cost function in the optimization algorithm is B. Results from design automation
based on SWE, which is a convenient mathematic to describe We perform the optimization of the seven impedance
radiation properties. values that load the dual-band micro-array. The algorithm is
After optimization, we integrate surface-mounted programmed with classical convex optimization functions. In
components into the antenna geometry according to the this work, we allow the optimization algorithm to use a
optimized impedance values. resistive part in the optimized impedances. We define the
cost function as the difference between the reference SWE
III. PROOF OF CONCEPT DESIGN (see Table I) and the SWE of the micro-array within both L2
and L1 bands. We define the convergence criteria according
In this paper, we present the results from a proof-of-
to the typical requirements presented in Table II. A negative
concept prototype. We aim to demonstrate that our approach
gain is targeted in low-band L2 in order to keep a
in antenna design automation is suitable for compact dual-
demonstration with compact antenna (minimum sphere
band GNSS antennas.
diameter d = 0.6λ/π at L2 band). Gain requirements are
A. Antenna geometry for dual-band GNSS defined as expected value at boreside (zenith) direction
For dual-band operation, we brought two circular 4-IFA within the full frequency range.
micro-arrays together in one 8-element system as presented
in Fig. 4. The dual-band micro-array is placed over a limited- TABLE II. DESIGN GOALS FOR DUAL-BAND GNSS
sized ground plane.
Requirements Target

L1 1560-1610 MHz
Frequency bands
L2 1215-1240 MHz
L1 0 dBic
RHCP gain (boreside)
L2 -2 dBic

LHCP gain (boreside) -10 dBic

TABLE III. OPTIMIZED VALUES FOR IMPEDANCE LOADS

IFA/load n. Optimized value

Z1 8.2 pF
Z2 0.4 pF
Z3 16 pF + 3.9 Ω
Z4 82 pF
Z5 0.2 pF
Z6 1.4 pF
Z7 1.4 pF
Fig. 4. Geometry of the coupled micro-array for dual-band GNSS.

The IFA elements are printed on small circuits (low-loss We present the optimized values in Table II. The set of
substrate ROGERS RO4003, thickness 0.8 mm). We have capacitance values do not correspond to any expected
two different elements for L1 and L2 with different lengths. scheme. Surprisingly, the automated optimization calls for a
We build the full micro-array structure by crossing the small resistor at IFA n. 3 only. We show the optimized gain
IFA circuits and soldering them to the ground-plane PCB performances in Fig. 5. The RHCP and LHCP gains meet the
(substrate FR-4, thickness 1.6 mm). Each of the IFAs has requirements i.e. -2 dBic in the L2 band and 0 dBic in the L1
one end connected to the ground and a feed point for band. We can observe that the optimization of the impedance
connection to an impedance load. The ground plane PCB has

This paper's copyright is held by the author(s). It is published in these proceedings and included in any archive
such as IEEE Xplore under the license granted by the "Agreement Granting EurAAP Rights Related to
Publication of Scholarly
Authorized licensed use limitedWork."
to: CEA GRENOBLE. Downloaded on November 22,2021 at 16:36:17 UTC from IEEE Xplore. Restrictions apply.
values succeeds in reaching the requirements in the targeted Fig. 7. Prototype of a micro-array for dual-band GNSS – surface-mounted
components for final configuration of the antenna.
frequency bands.
The micro-array prototype built for the demonstration is
illustrated in Fig. 6 and Fig. 7. We mount the set of discrete
components from the previous optimization without any
change. We use surface-mounted components in 0402
format, which it is compatible with the compact size of the
antenna. We perform radiation measurements at the CEA-
Leti anechoic chamber. As the antenna is compact, we use an
optic fiber to RF convertor, as shown in Fig. 6, in order to
avoid any cable effect on radiation [15].
We present the measurement results in Fig. 8 and Fig. 9
for L2 and L1 bands respectively. Simulation and
measurement are in acceptable agreement. We observe a
decrease (0.6-1.4 dB) in L2 gain, for both RHCP and LHCP
due to the small electrical size of the antenna that makes it
more sensitive to mounting, fabrication and component
tolerances. The agreement in L1 band is very good, but we
Fig. 5. Results from the automated design process – RHCP and LHCP
gains at boreside direction versus frequency. can observe some differences in the upper frequency range.
These results demonstrate that optimization of impedance
loads provides promising results for dual-band GNSS
IV. PROTOTYPE DEMONSTRATION applications.

Fig. 8. Gain versus frequency in boreside direction – comparison between

Fig. 6. Prototype of a micro-array for dual-band GNSS – measurement simulation and measurement in the L2 frequency band.
setup in anechoic chamber [15].

Fig. 9. Gain versus frequency in boreside direction – comparison between

simulation and measurement in the L1 frequency band.

This paper's copyright is held by the author(s). It is published in these proceedings and included in any archive
such as IEEE Xplore under the license granted by the "Agreement Granting EurAAP Rights Related to
Publication of Scholarly
Authorized licensed use limitedWork."
to: CEA GRENOBLE. Downloaded on November 22,2021 at 16:36:17 UTC from IEEE Xplore. Restrictions apply.
 V. CONCLUSION [12] A. Clemente, M. Pigeon, L. Rudant, and C. Delaveaud, “Design of a
super directive four-element compact antenna array using spherical
We propose a new approach in antenna design wave expansion,” IEEE Trans. Antennas Propag., vol. 63, n. 11, pp.
automation consisting in the use of coupled micro-array 4715-4722, Nov. 2015.
loaded with impedances that are automatically optimized in [13] D. M. Pozar, “The active element pattern,” IEEE Trans. Antennas
Propag., vol. 42, n. 8, pp. 1176 - 1178, Aug. 1994.
order to reach the required design goals. The use of SWE
[14] K. Belmkaddem, L. Rudant and T.-P. Vuong, “Analysis of open-slot
makes this process suitable for a matching of radiation antenna radiation pattern using spherical wave expansion,” IET
properties. With an application to dual-band GNSS, we Microw. Antennas Propag., vol. 9, n. 13, pp. 1407-1411, Oct. 2015.
demonstrate that this approach can handle optimization in [15] Schmid & Partner Engineering AG, speag.swiss.
several frequency bands. In this paper, we also present a
prototype of compact antenna showing that the automated
design process is able to achieve promising performance and
miniaturization in a concrete manner.
In the next steps, we aim to apply this automated design
for vehicular applications. A micro-array geometry will be
designed for close integration in cars and a GNSS test
campaign will be performed in the field. In future works, we
aim to design different micro-arrays for others applications
e.g. satellite machine-to-machine connectivity.

REFERENCES
[1] Z. Wang, S. Fang, S. Fu, and S. Lu, “Dual-band probe-fed stacked
patch antenna for GNSS applications, ” IEEE Antennas and Wireless
Propagation Letters, vol. 8, pp. 100-103, Jan. 2009.
[2] P. G. Elliot, E. N. Rosario, and R. J. Davis, “Novel Quadrifilar Helix
Antenna combining GNSS, Iridium, and a UHF communications
monopole, ” IEEE Military Communications Conference (MILCOM),
pp. 1-6, Oct. 2012.
[3] K. E. Kedze, and I. Park, “Folded Crossed-Dipole Antenna for
Circular Polarization, ” in Proc. International Workshop on Antenna
Technology (iWAT), Feb. 2020
[4] E. R. Gafarov, A. A. Erokhin, A. V. Stankovsky, and Y. P.
Salomatov, “Multiband three-layer GNSS microstrip antenna,” in
Proc. 2019 Intern. Siberian Conf. Control Comm. (SIBCON 2019),
Tomsk, Russia, Apr. 2019.
[5] W.-I. Son, W.-G. Lim, M.-Q. Lee, S.-B. Min, and J.-W. Yu, “Design
of compact quadruple inverted-F antenna with circular polarization
for GPS receiver,” IEEE Trans. Antennas Propag., vol. 58, n. 5, pp.
1503-1510, May 2010.
[6] Y. Zhou, C.-C. Chen, and J. L. Volakis, “Proximity-coupled stacked
patch antenna for tri-band GPS applications,” in Proc. 2006 IEEE
Antennas Propag. Society Int. Symp. (AP-S/URSI), Albuquerque, NM,
USA, Jul. 2006.
[7] H. Sakamoto, T. Yanagi, T. Fukasawa, and H. Miyashita, “A small
quadrifilar helical antenna with parallel resonance circuit for dual-
band application,” in Proc. 2016 Int. Symp. Antennas Propag. (ISAP
2016), Okinawa, Japan, Oct. 2016.
[8] O. P. Falada, M. U. Rehman, Y. (F.) Gao, X. Chen, and C. G. Parini,
“Single feed stacked patch circular polarized antenna for triple band
GPS receivers,” IEEE Trans. Antennas Propag., vol. 60, n. 10, pp.
4479-4484, Oct. 2012.
[9] H. M. El Misilmani, T. Naous, and S. K. Al Khatib, “A review on the
design and optimization of antennas using machine learning
algorithmes and techniques,” Wiley Int. J. RF Microw. Computer-
Aided Eng., vol. 30, n. 10, Oct. 2020.
[10] V. Grout, M. O. Akinsolu, B. Liu, P. I. Lazaridis, K. K. Mistry, and
Z. D. Zaharis, “Software solutions for antenna design exploration: a
comparison of packages, tools, techniques, and algorithms for various
design challenges,” IEEE Antennas Propag. Mag., vol. 61, n. 3, pp.
48-59, Jun. 2019.
[11] Contributions by Altair, ANSYS, Candence/AWR, Keysight, and
Mathworks, “Artificial intelligence and machine learning add new
capabilities to traditional RF EDA tools,” Microw. J., , pp. 20-36, Jul.
2020.

This paper's copyright is held by the author(s). It is published in these proceedings and included in any archive
such as IEEE Xplore under the license granted by the "Agreement Granting EurAAP Rights Related to
Publication of Scholarly
Authorized licensed use limitedWork."
to: CEA GRENOBLE. Downloaded on November 22,2021 at 16:36:17 UTC from IEEE Xplore. Restrictions apply.