Access to this full-text is provided by Wiley.
Content available from IET Microwaves, Antennas & Propagation
This content is subject to copyright. Terms and conditions apply.
Received: 29 March 2021
-
Revised: 14 August 2021
-
Accepted: 16 August 2021
-
IET Microwaves, Antennas & Propagation
DOI: 10.1049/mia2.12194
ORIGINAL RESEARCH PAPER
Ultra‐wideband antenna system for in‐band full‐duplex
applications
Hadi Hijazi
1,2
|Marc Le Roy
2
|Raafat Lababidi
1
|Denis Le Jeune
1
|
Andre Pérennec
2
1
ENSTA Bretagne, Lab‐STICC, CNRS, Brest, France
2
Univ Brest, Lab‐STICC, CNRS, Brest, France
Correspondence
Hadi Hijazi, ENSTA Bretagne, Lab‐STICC, CNRS,
UMR 6285, F‐29200 Brest, France.
Email: hadi.hijazi@ensta-bretagne.org
Abstract
In this study, we present an ultra‐wideband full‐duplex system constituted of a wideband
4‐element Vivaldi array and wideband microstrip‐to‐slotline baluns. The proposed system
is characterised by its simplicity, high directivity, and high self‐interference cancellation
levels over a wide frequency bandwidth. The system is fabricated using basic printed‐
circuit board (PCB) technology and can provide at least 50 dB of self‐interference
cancellation over the bandwidth of operation, 4–40 GHz, with an average gain of
7.8 dBi. The system has a size of 8 cm 8 cm 9.4 cm and can be used as a high data
rate link between two distant wireless nodes. To assess the merits of the proposed system
and compare it to other published works, a new gure of merit (FoM
WFD
) dedicated to
wide‐band full‐duplex antenna topologies is introduced in this study.
KEYWORDS
baluns, microstrip transitions, ultra wideband antennas, Vivaldi antennas
1
|
INTRODUCTION
Estimates of future mobile trafc indicate that the global number
of mobile subscriptions could be 13.8 billion in 2025 and 17.1
billion in 2030 [1]. The increase in the number of mobile users
creates a spectrum congestion problem and elevates the pressure
on the available mobile infrastructure to keep up with the
increased demand for mobile services. These problems create
the need for new approaches to nd more spectrum resources by
improving traditional spectrum sharing techniques and/or har-
nessing new ones. Recently, in‐band full‐duplex (IBFD) tech-
nology emerged as a promising solution for the spectrum
congestion problem [2]. As opposed to the traditional out‐of‐
band full‐duplex (OBFD), which utilises orthogonal resources
(frequency or time) to establish a full‐duplex communication, in‐
band full‐duplex allows the simultaneous transmission and
reception between two communicating nodes at the same fre-
quency and the same time slot. This can only be achieved by
reducing the self‐interference signals that are coupled from the
transmitter of one node to its own receiver below the noise oor
level, allowing the receiver to receive the useful signal
transmitted by the other node. Nominally, 110 dB of self‐
interference cancellation is required to be able to establish an
in‐band full‐duplex communication [3].
Figure 1depicts a general block diagram of an in‐band full‐
duplex heterodyne transceiver where it can be seen that self‐
interference cancellation takes place at three stages or levels:
at the antenna level, at the analogue level, and at the digital
level. At the antenna level, the system can be monostatic where
a shared antenna(s) is used to transmit and receive, or it can be
bi‐static where separate antennas can be used. However, for
monostatic systems a duplexing device is required to isolate the
transmitted signal from the received signal, such as circulators
[4] or hybrid transformers [5]. While, for bistatic systems, the
transmit antennas can be placed or fed in a way to create null
planes at the positions of the receive antennas which can
reduce the coupling between them.
On the other hand, both the analogue and digital cancel-
lation circuitry benet from the fact that the transmit and
receive antennas are co‐located on the same board or platform,
and therefore, the receiver section has knowledge of the
originally transmitted signal and of the characteristics of the
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the
original work is properly cited, the use is non‐commercial and no modications or adaptations are made.
© 2021 The Authors. IET Microwaves, Antennas & Propagation published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.
IET Microw. Antennas Propag. 2021;15:1853–1865. wileyonlinelibrary.com/journal/mia2
-
1853
direct path coupling, which can be predicted and measured. So,
a copy of the transmitted signal is taken, and direct path losses
and phase changes are applied to it, then it gets subtracted
from the received signal [3]. At the analogue level, cancellation
can take place at one of three stages: at the RF stage [6] before
the ampliers, then at the intermediate frequency (IF) stage [7],
and at the base band (BB) stage [8] before the converters. Note
that, it is necessary to suppress the self‐interference signals, at
the antenna and analogue levels, by at least a certain amount
(typically 60 dB [3]) to prevent saturating the analogue‐to‐
digital converter.
The simultaneous transmit‐receive property of in‐band
full‐duplex technology can enable the introduction of novel
and efcient multiple access techniques [9], physical layer se-
curity protocols [10], relaying solutions [11], and can reduce air
interface delay [12]. Also, it might benet radar systems [13],
cognitive radios (CR) [14], and multiple‐input multiple output
(MIMO) arrays [15]. So, considering their appealing advan-
tages, it is highly desirable to implement wideband full‐duplex
systems for civil, military, and space applications. Available
publications on wideband full‐duplex systems indicate that
achieving 50 dB of self‐interference cancellation level over a
wide bandwidth is possible at the antenna level [16–18],
however, the analogue and digital parts fall far behind with
achieving wideband cancellation, mainly, due to circuitry
bandwidth limitations and cost. Nevertheless, implementing a
wideband full‐duplex antenna system remains very attractive as
it can be used as a universal transmit‐receive system suitable
for different standards, frequency bands, or bandwidths, and
even for software‐dened radios (SDR) or cognitive radios
(CR). In light of that, the scope of this work will be mainly
focussed on wideband self‐interference cancellation techniques
at the antenna level.
Several techniques were proposed to attain wideband self‐
interference cancellation at the antenna level. For monostatic
systems, some implementations based on wideband circula-
tors and hybrid transformers can be found in [4, 5], respec-
tively. Nevertheless, the limited isolation of the duplexing
devices limits the maximum achievable level of cancellation.
On the other hand, for bistatic systems, the basic technique is
to increase the separation between the Tx and Rx antennas
[19], however, this results in an increased system size. So,
instead, the beams of the antennas can be directed somewhat
in different directions to reduce the overlap between them
[19], nonetheless, not all applications can tolerate to transmit
and receive in different directions. Alternatively, using
orthogonal polarisations for the Tx and Rx antennas [20] can
increase the level of cancellation, but this requires the Tx
antennas of one node to be aligned with the Rx antennas of
the other node. Moreover, near‐eld cancellation [17] showed
the capability of achieving a decent level of cancellation with
a simple system design and compactness. Also, circularly
phased arrays [21–23], can provide a similar performance to
near‐eld cancellation with a quasi‐omnidirectional radiation
pattern, however, it requires a more complicated feeding
network. In addition to all the above, placing high impedance
structures between the antennas were considered in [16],
which can provide a high level of cancellation, however, the
size of such structures can drastically increase the overall size
of the system. Finally, based on the above techniques wide-
band full‐duplex arrays were also demonstrated in [24–27].
Note that, one or more techniques can be used in conjunc-
tion to achieve the maximum possible level of self‐
interference cancellation.
In the next section, our system specications, detailed
contributions, and the potential applications of the proposed
system will be presented. Then, the whole system with all the
individual devices will be presented in Section 3, where its
principle of operation will be explained. And in Section 4, the
nal system assembly, matching, self‐interference cancellation,
and far‐eld performance will be demonstrated both in sim-
ulations and in measurements. Also, in this section, the new
gure of merit will be introduced, and then the section will be
closed with a table that compares this work to previous works
in the literature, especially in terms of the newly introduced
gure of merit.
RF Cancellation
IF Cancellation
BB Cancellation
Digital Cancellation
Feeding Network
CDAANL
DACPA
Rx-M1 Rx-M2
Tx-M1 Tx-M2
Antenna(s)
Rx
Tx
Digital LevelAnalog LevelAntenna Level
Targeted Cancellation:
> 50 dB
Targeted Cancellation:
> 30 dB
Targeted Cancellation:
> 30 dB
FIGURE 1 General block diagram of an in‐band full‐duplex transceiver
1854
-
HIJAZI ET AL.
2
|
SCOPE OF WORK
In this work, we aim to design an ultra‐wideband in‐band full‐
duplex system with at least 50 dB of self‐interference cancel-
lation at the antenna level, that can transmit and receive in the
same direction with a fairly high directivity. And as it was
evident, from the published works, that bistatic techniques
have more wideband potential than monostatic techniques, in
particular, near‐eld cancellation which can satisfy the required
specications if, at the same time, orthogonal polarisations are
used for the transmit and receive antennas. And in addition to
that, near‐eld cancellation can generate the most compact
full‐duplex systems as compared to all other bistatic tech-
niques. However, it requires at least four antennas and two
180° out‐of‐phase power dividers (or baluns), and it can be
sensitive to the baluns' amplitude and phase imbalances, and,
also, to the antennas' placement.
Near‐eld cancellation has been implemented in several
previous works, mainly, with planar antennas operating at the
low frequency side of the spectrum (below the K‐band), and
over bandwidths not more than one or two octaves [17, 20,
28–30]. But, in this work, we aim to have a system that can
operate up to the Ka‐band while covering most of the lower
bands (starting from the C‐band). To achieve this goal, and to
satisfy the previously mentioned specications, we propose a
system that consists of a 4‐element Vivaldi array and two
microstrip‐to‐slotline baluns, as depicted in Figure 2.
Although the individual devices of the system are planar, yet
the system is three‐dimensional, which creates some vagueness
about the operation of the near‐eld cancellation technique,
taking into consideration that it has only been demonstrated
for planar systems. Thus, our work can be considered the rst
work to implement the near‐eld cancellation technique over
an extremely wide bandwidth with a three‐dimensional an-
tenna system, and our contributions can be summarised as
follows:
Achieving 50 dB of self‐interference cancellation, at the
antenna level, over an extremely wide bandwidth (4–
40 GHz).
Optimising the microstrip‐to‐slotline transition to operate
over the desired bandwidth. The transition will be used to
feed the antennas and it constitutes the basic building block
of the used baluns.
Proposing a way to align the antennas using a 3D‐printed
support.
Introducing a new gure of merit, the wideband full‐duplex
systems gure of merit (FoM
WFD
), to assess the perfor-
mance of wideband in‐band full‐duplex systems, and it ac-
counts for the achieved self‐interference cancellation level,
the gain, the bandwidth, and the size of the system.
The proposed system can be considered for diverse ap-
plications where capacity, low latency, and secrecy capabilities,
offered by in‐band full‐duplex, are required. That may concern,
for example, in cellular telecommunication systems, backhaul
microwave links between the core network and the radio access
network, or wireless links between two remote base stations
instead of using bre optic. More broadly, the ‘sensing‐as‐
transmitting’ property, along with all other characteristics of in‐
band full‐duplex, can be valuable for any wideband multi‐
standard communication operation especially for cognitive
radios which require continuously scanning a wide range of
frequencies while transmitting.
3
|
SYSTEM DESIGN
3.1
|
Principle of operation
The proposed in‐band full‐duplex system is depicted in Figure 2
where the substrates were omitted for better clarity. The system
is composed of four Vivaldi antennas which are placed on the
perimeter of a circle of a xed radius and sequentially rotated by
90° from one another, thus forming two diametrically opposite
pairs of antennas, where each pair is orthogonal to the other. One
pair is used to transmit, and the other is used to receive. Then
each pair of antennas is fed signals of equal amplitudes and 180°
out‐of‐phase by using wideband microstrip‐to‐slotline baluns.
To explain the principle of operation of the proposed system,
two approaches are adopted [17]: an inter‐port approach that
describes the power coupling between the input and output
ports of the Tx and Rxbaluns, and a eld approach that describes
the electric eld behaviour in the far‐eld region.
Inter‐Port Approach: in order to develop this approach,
we nd that it is most suitable to work with power waves
[31], where andenotes the power wave incident at port n
and bndenotes the power wave reected from that port,
and accordingly, the total powers carried by each wave are
equal to janj2and jbnj2, respectively. The ow of power
waves in the system is depicted in Figure 2b assuming ideal
conditions where no mismatch exists between the system
components.
Now, the output power (Po) at the output of the Rx balun
can be expressed in terms of power waves coming from the
receive antennas 1 and 3 as follows:
Po¼ jboj2¼ao
0þa00
oejπ2¼b1þb3ejπ2ð1Þ
Also b1and b3can be expressed in terms of the power
waves incident on the transmit antennas 2 and 4 as follows:
b1¼S12a2þS14 a4ð2Þ
b3¼S32a2þS34 a4ð3Þ
Similarly, a2and a4can be written in terms of the power
wave aiincident at the input of the Tx balun:
a2¼bi
0¼ffiffiffi
2
p
2aið4Þ
a4¼b00
i¼ffiffiffi
2
p
2aiejπ ð5Þ
HIJAZI ET AL.
-
1855
By substituting (4) and (5) in (2) and (3) we obtain:
b1¼ffiffiffi
2
p
2S12aiþffiffiffi
2
p
2S14aiejπ ¼ffiffiffi2
p
2aiS12 þS14ejπ ð6Þ
b3¼ffiffiffi
2
p
2S32aiþffiffiffi
2
p
2S34aiejπ ¼ffiffiffi2
p
2aiS32 þS34ejπ ð7Þ
Finally, by substituting (6) and (7) in (1) we can obtain the
output power (Po) in terms of the input power (Pi):
Po¼ffiffiffi2
p
2aiS12 þS14ejπ þffiffiffi
2
p
2aiS32 þS34ejπ ejπ
2
Po¼ffiffiffi2
p
2jaij2jS12 þS14ejπ þS32 ejπ þS34 ej2πj2
Po¼1
2PijS12 −S14 −S32 þS34j2ð8Þ
Equation (8) implies that if S12 ¼S14 and S32 ¼S34,
then the power coupled from the input port of the system
to the output port will be equal to zero. This means that if
the Rx antennas are placed along the perpendicular bisector
of the Tx antennas, then innite isolation is, theoretically,
obtained between them. This conclusion holds only if the
antennas are placed and aligned precisely and if the balun
operates ideally, that is if the two output signals of the
balun are of equal amplitudes and 180° out‐of‐phase.
Nonetheless, in practice, there will be slight misplacements
of antennas and imbalances in the amplitude and phase of
the balun's output signals, thus the isolation between the
input and the output ports of the system is expected to
decrease.
Field Approach: Assuming that the antennas are not con-
nected to the baluns, then each individual antenna element
radiates a linearly polarised electric eld of magnitude E0which
is oriented in the opposite direction of the microstrip stub
orientation, Figure 2b. So, the individual electric elds can be
expressed as follows:
E
→
1¼E0ejπ y
→ð9Þ
E
→
2¼E0x
→ð10Þ
E
→
3¼E0y
→ð11Þ
E
→
4¼E0ejπ x
→ð12Þ
Now, if the baluns are connected to the antennas, then the
Tx and Rx electric elds can be described as follows:
E
→
Tx ¼E
→
2þE
→
4ejπ ¼2E0x
→ð13Þ
E
→
Rx ¼E
→
1þE
→
3ejπ ¼2E0ejπ y
→¼−2E0y
→ð14Þ
Equation (13) implies that the electric elds of the two
transmit antennas will combine constructively in the far‐eld
region and Equation (14) implies the same observation for
the receive antennas. However, this conclusion only holds if
each two opposite antennas are symmetrically rotated with
respect to the centre of symmetry of the system, that is, their
(a) (b)
FIGURE 2 (a) An illustration of the proposed wideband in‐band full‐duplex system with 4‐element Vivaldi array and two microstrip‐to‐slotline baluns, and
in the background a transparent view of the Vivaldi array showing the position and orientation of the microstrip feed lines. (b) A schematic of the system
showing antenna placement and numbering, ow of power waves, and the original electric elds radiated by each single antenna before connecting them to the
baluns
1856
-
HIJAZI ET AL.
feeding microstrip lines are oriented in opposite directions (see
the background image in Figure 2a), and if the antennas are
fed signals that are 180° out‐of‐phase. The antenna rotation
(or feeding line opposite orientation) condition is especially
critical, otherwise, electric elds will combine destructively in
the far‐eld, which creates a far‐eld null, and thus the system
will not be able to transmit or receive in the direction of this
null.
Note that if the system is not symmetric, that is, if the
number of transmit and receive antennas is not the same
or if they are fed differently, differences between the
transmit and receive far‐eld radiation patterns will emerge,
thus it is extremely favourable to maintain high symmetry
in the system. The antennas can be tangent or normal to
the circle or could form any angle with the circle in
general and can still achieve the same level of self‐
interference cancellation, however, this affects the total
size of the system and the orientation of the radiation
pattern. In addition to that, signicant grating lobes can be
observed in the far‐eld radiation pattern if the separation
between the opposite antennas is greater than a half‐
wavelength at a certain frequency.
3.2
|
Wideband balun design
The initial design of the wideband balun used here was pro-
posed by Rhode et al. [32] and it consists of three microstrip‐
to‐slotline transitions terminated with radial stubs as in
Figure 3. The balun operates as follows: the input power is rst
injected in the middle microstrip line and is then coupled to the
slotline at its centre. After that, in the slotline, the power will be
divided equally between the left and right paths, and at the end
of each path, the power will be coupled again to the
(a) (b)
(c) (d)
FIGURE 3 (a) Illustrative drawing of the microstrip‐to‐slotline balun with (b) the parameters of the design. (c) Simulated and measured input matching and
insertion losses, and (d) amplitude imbalances (AI) and phase imbalances (PI) between the output ports. {L=40 mm, W=40 mm, Wm =0.4 mm,
R
m=1.73 mm, θm=165°, Ws=0.1 mm, Ls=8 mm, Rs=1.93 mm, θs=150°}
HIJAZI ET AL.
-
1857
corresponding output microstrip line. Since the two output
microstrip lines are oriented in opposite directions, the two
output signals will be inverted with respect to one another over
a wide bandwidth. The used balun was designed and optimised,
in CST Microwave Studio, based on the parametric analysis
performed by us in [33], after that, two baluns were fabricated
on a 203.2 μmthick RO4003C substrate (εr=3.55), and they
occupy an area of 40 mm 40 mm.
The measured S‐parameters in Figure 3c show that the
balun is matched from 4 to 50 GHz, where a relatively good
agreement with simulated results is observed below 30 GHz,
while some unexpected jumps in the matching response start
to appear beyond 30 GHz. This might probably come from the
limitations of the high frequency cables and/or connectors
used to conduct the measurements, nevertheless, none of these
jumps go higher than −10 dB. Also, the measured insertion
loss attains a value of −5 dB at 4 GHz and reaches a value of
−11 dB at 40 GHz, however, 3 dB of the insertion loss come
from the power division taking place and are not actually lost
in the balun. The measured insertion loss drastically increases
with frequency as compared to the simulated one, this means
that in practice more power is being lost or absorbed by the
substrate. This may come from a slight degradation of the
substrate characteristics during the fabrication process which
led to an increase in the value of its loss tangent (tanδ).
Moreover Figure 3d shows that the amplitude and phase
imbalances between the output ports of the balun are less than
1 dB and 7° respectively over the 4–40 GHz band, where
more uctuations in the measured results are observed as
compared to the simulated results, which might also be attrib-
uted to fabrication imperfections. In addition to that, poor levels
of matching and isolation between output ports were observed
(but not plotted here), which are considered normal as a result
of the fact that a 3‐port device cannot be lossless, reciprocal, and
matched at all ports at the same time. However, simulations
proved that low or high levels of isolation between output ports
of the balun do not affect the level of self‐interference cancel-
lation obtained in the in‐band full‐duplex system.
3.3
|
Wideband Vivaldi antenna design
The Vivaldi antenna, Figure 4, is a uniplanar exponentially
tapered slot antenna which is characterised by its wide
FIGURE 4 Illustrative drawing of the Vivaldi antenna {L=70 mm, W=40 mm, L
t¼10 mm, Lv=60 mm, Wv=20 mm, Wm=0.4 mm,
Rm=1.87 mm, θm=80°, Ws=0.1 mm, Rs=1.85 mm, θs=170°}
1858
-
HIJAZI ET AL.
bandwidth, simplicity of its design, its low cost, and its highly
directive far‐eld radiation pattern. The antenna is formed of
a microstrip‐to‐slotline transition acting as a feeding mech-
anism, and an exponentially tapered slotline acting as a
radiating element. The microstrip‐to‐slotline transition was
designed, optimised, and used, previously, in the balun design
with a matching bandwidth from below 4 GHz to beyond
40 GHz. So the same transition will be used to feed the
Vivaldi antenna, however, note that, a microstrip radial stub
with half the size of that used in the balun is adopted here.
The size reduction of the microstrip stub reduces the overlap
with the slotline stub and, consequently, reduces the power
leaked by the transition [33], which means that more power
will be fed to the antenna, and this will enhance its
efciency.
On the other hand, the exponential taper can be dened by
the following set of equations:
y¼AeαxþBð15Þ
A¼0:5ðWv−WsÞ
eαL−eαLtð16Þ
B¼0:5WseαL−WveαLt
eαL−eαLtð17Þ
where αis the growth rate of the exponential taper which can
be tuned to obtain different antenna characteristics (band-
width, beamwidth, and gain). In this design, αwas set to 0.05
for maximum matching bandwidth. Also, in order for the
radiating element to be matched in the same bandwidth as that
of the transition, the length of the exponential taper (Lv) was
set to approximately one wavelength at 4 GHz, and the aper-
ture width (Wv) was set to approximately a half‐wavelength at
the central frequency (22 GHz).
The antennas were also fabricated on the same RO4003C
substrate (εr=3.55 and h=203.2 μm) where the total size
of each antenna element is 70 mm 40 mm. Figure 5a
conrms that the antenna is matched, as expected, from 4 to
47 GHz. And Figure 5b shows that the maximum realised
gain of the antenna is better than 5 dBi over the entire
matching bandwidth (4–47 GHz) and better than 10 dBi
between 7 and 40 GHz in measurements. The Vivaldi gain
measurements were performed using three reference antennas
to cover the entire matching bandwidth: the rst antenna
operates in the 2–18 GHz band, the second antenna operates
in the 18–33 GHz band, and the third antenna covers the 33–
50 GHz band. The three measured gains are depicted in
Figure 5b in three different colours: red, blue, and green
respectively, while the simulated gain is plotted in black. It is
observed that the measured gain is slightly shifted towards
the lower frequency side as compared to the simulated gain,
which results from fabrication imperfections, however, the
shape of the measured curve is highly similar to the simulated
curve.
4
|
SYSTEM ASSEMBLY AND
PERFORMANCE
4.1
|
System assembly
To achieve self‐interference cancellation and ensure proper
system performance, it is of great importance to preserve a
high symmetry when assembling the system, that is, the
antennas should be placed exactly at the same distance away
from the centre of symmetry. While it is easy to place the
antennas precisely and symmetrically in a simulation envi-
ronment, yet, in practice it cannot be guaranteed. Moreover,
it was noticed that the fabricated antennas tend to bend
naturally due to the low thickness of the used substrate
(203.2 μm), which also contributes to the asymmetry of the
system. Also, the thickness of the substrate makes it very
fragile and prone to fracturing, mainly, due to the tension
exerted by the feeding cables and connectors. To resolve
those mechanical issues, a 3D‐printed support is designed
and fabricated to ensure precise antenna placement, to
reduce antenna bending and to hold it fully erect, and to
absorb mechanical tension from the feeding cables and
connectors.
(a)
(b)
FIGURE 5 Simulated and measured (a) matching and (b) broadside
maximum realised gain of the designed Vivaldi antenna
HIJAZI ET AL.
-
1859
Figure 6b depicts the 3D‐printed support which was
made of Polyvinyl‐Chloride (PVC, εr≈3), and the different
parts of the support were xed together using PTFE screws
(εr≈2:1). The support is designed such that it only grasps
the edges of the antennas and is kept sufciently below the
top of the antenna. This ensures that the support will not
affect the antennas' matching or performance. But before
installing the antennas inside the 3D‐printed support,
Southwest 2.92 mm connectors were mounted on all devices,
and four Keysight N5448B phase‐paired cables (5 ps skew)
were used to connect the antennas to the baluns. However,
note here that the fabricated baluns were modied slightly to
have their output ports at the top (as shown in Figure 6a),
which makes it easier to connect them directly to the antennas
as shown in Figure 6c and 6d. Finally, the measurements were
carried out by using Rhode&Schwarz ZVA67 vector network
analyser.
4.2
|
System performance
The simulated and measured characteristics of the system are
depicted in Figure 7. Note that simulations were done taking
into account the PTFE screws and the 3D‐printed support,
which was modelled as a lossless dielectric material having
εr≈3. Also, simulations take into account the extra separa-
tion distance between the antennas as per the fabricated
design in Figure 6b. Moreover, in simulation, the phase‐
paired cables were modelled as straight pieces of coaxial
lines (25 cm long) formed of aluminium inner and outer
conductors and PTFE dielectric stufng. In addition to that,
the 2.92 mm connectors were not considered in simulations.
And nally, note that the baluns that were considered in
simulations are the simulated microstrip‐to‐slotline baluns
and not some ideal baluns, which means that the signals
entering the antennas are not ideal, that is they are not
perfectly of equal amplitudes and 180° out‐of‐phase, but
rather they are subject to the phase and amplitude imbalances
of the simulated baluns.
4.2.1
|
System matching
Figure 7a shows that the assembled in‐band full‐duplex
system has a return loss better than 10 dB over an ultra‐
wide frequency band (3.5–49.35 GHz). The measured
matching seems to be slightly better than the simulated one
around some frequencies, and this can be attributed to the
higher insertion losses of the fabricated baluns, and also to
the power losses resulting from the phase‐paired cables and
the 2.92 mm connectors. While these losses reect positively
on the system's matching, yet they will reect negatively on
the measured gain of the system, which is depicted in
Figure 7c.
FIGURE 6 Pictures of (a) the fabricated balun, (b) antenna assembly inside the 3D‐printed support, (c) whole system assembly and (d) measurements inside
the anechoic chamber
1860
-
HIJAZI ET AL.
4.2.2
|
Self‐interference cancellation
Figure 7b depicts the inter‐port coupling (or the insertion loss)
between the different ports of the system. Note that self‐
interference cancellation is equal to the negative of the inter‐
port coupling (in dB). In fact, inter‐port coupling and self‐
interference cancellation express similar meanings in opposite
senses. Inter‐port coupling tells how much of the power incident
at the input port is transmitted to the output port, while self‐
interference cancellation tells by how much the power trans-
mitted to the output port is weaker than the power incident at the
input port, or inversely, it tells by how much the power incident at
the input port is stronger than the power transmitted to the
output port. The different plots in Figure 7b are explained below:
“S21” is the coupling (or cross‐polarisation) between the two
orthogonal antennas 1 and 2 in Figure 2b if they are excited
individually without a balun. And the aim of this measure-
ment is to see how much cancellation is initially obtained
from the fact that the transmit and receive antennas are
orthogonally polarised.
“SAB Sim:” is the simulated coupling between the input port
of the Tx balun (Port A) and the output port of the Rx
balun (Port B).
“SAB Chamber” is the measured coupling between the input
port of the Tx balun (Port A) and the output port of the Rx
balun (Port B) inside the anechoic chamber.
“SAB Room” is the measured coupling between the input
port of the Tx balun (Port A) and the output port of the Rx
balun (Port B) in a normal room that has no electromagnetic
insulation and that contains many objects (chairs, tables,
instruments, etc.) which might cause multipath reections.
The purpose of this measurement is to compare its result to
the one obtained in the anechoic chamber, and to see if the
reections from the surrounding environment have any
effect on the performance of the system.
Now, the measured level of self‐interference cancellation
for the whole system in the anechoic chamber is about 50 dB
at 4 GHz and reaches about 70 dB at 40 GHz, while in
simulation it starts with 80 dB at 4 GHz and continues to
maintain approximately the same value until it reaches 40 GHz.
The lower level of cancellation in measurements was expected
as a result of fabrication imperfections, antenna misplacement
in the 3D‐printed support, and the higher imbalances in the
phases and amplitudes of the baluns. Also, if we compare the
system cancellation, which was obtained in the anechoic
chamber, to the measured cancellation between the two
orthogonal antennas 1 and 2, which is 30.5 dB at 4 GHz and
50.5 at 40 GHz, we can notice that the used cancellation
technique, with differential feeding, can provide, at least, an
extra 20 dB of cancellation, on average, on the top of the
cancellation obtained from cross‐polarisation. Finally, by
comparing the system cancellation in the anechoic chamber to
the system cancellation in a normal room, we can notice that
the difference between the two is almost negligible. This in-
dicates that the reections from the surrounding environment
are less signicant than the residual self‐interference, and
therefore, have a negligible effect on the cancellation perfor-
mance of the proposed system.
4.2.3
|
Fareld performance
The gain of the system is depicted in Figure 7c where it can be
seen that there are some differences between the simulated and
measured results. In fact, the simulated gain starts with a value
of 4.5 dBi at 4 GHz and then starts to increase with frequency.
It reaches a value of 10 dBi around 10 GHz and continues to
maintain almost the same value throughout the rest of the
bandwidth. Finally, it starts to drop again around 40 GHz. On
the other hand, the measured gain starts with a value of 5 dBi
(a)
(b)
(c)
FIGURE 7 The proposed in‐band full‐duplex system (a) matching,
(b) inter‐port coupling, and (c) broadside maximum realised gain
HIJAZI ET AL.
-
1861
at 4 GHz and then it continues to increase with frequency and
reaches its peak value (about 10 dBi) around 10 GHz. From
that point, the measured gain starts to drop with frequency
until it hits 5 dBi at 40 GHz. The drop in the measured gain
can be attributed to the higher insertion losses of the fabricated
balun and also to the power losses resulting from the phase‐
paired cables and the 2.92 mm connectors, which were not
considered in simulations. However the measured gain remains
higher or equal to 5 dBi in the bandwidth of matching (4–
40 GHz), and the average gain in this bandwidth is
7.8 2.6 dBi.
In addition to that, Figure 8a, 8b, 8c and 8d depict some
sample radiation pattern plots at 5, 15, 25, and 35 GHz,
respectively, where good agreement between simulated and
measured results is observed. The radiation pattern plots reveal
that a directive main lobe is present in the yz‐plane, and that at
least two grating lobes are present alongside the main lobe in
the xz‐plane. And it seems that the number of grating lobes
increases as we go higher in frequency. In fact, at least two
grating lobes start to appear in the radiation pattern when the
separation distance between the two opposite antennas is
higher than a half‐wavelength at the frequency of measure-
ment, and their number increases when the separation be-
comes higher than a multiple of half‐wavelength. Similar
results were observed in other publications on near‐eld
cancellation [17], and in publications on Vivaldi arrays [34].
This phenomenon can be considered as a major drawback of
the system, though it cannot be avoided unless the separation
distance between the antennas is reduced. However, for the
presented Vivaldi array, this cannot be feasible because the
separation distance is conditioned by the width of the single
antenna element, which is usually equal or greater than a
wavelength at the central frequency. This means, for such
wideband system, whatever the antenna width is, there will
always be a portion of the bandwidth (the higher frequency
range) where the grating lobes are present.
4.3
|
Figure of merit
To assess the merits of the proposed system a new gure of
merit, wideband full‐duplex gure of merit (FoM
WFD
), is
introduced, and it accounts for the achieved self‐interference
cancellation level, the gain, the bandwidth, and the normal-
ised size of the system. Firstly, both the values of the self‐
interference cancellation and the gain of the system should
be linear and not in dB, however, note that the dB value of the
self‐interference cancellation should be positive before con-
verting it to a linear scale. Secondly, concerning the size of the
system, some systems have a two‐dimensional geometry, such
as systems fabricated on printed‐circuit boards (PCBs) where
the thickness of the substrate is almost negligible, while other
systems possess a three‐dimensional geometry, which is the
case for our system, also, some systems have their feeding
(a) (b)
(c) (d)
FIGURE 8 Radiation pattern plots at (a) 5 GHz, (b) 15 GHz, (c) 25 GHz, and (d) 35 GHz, respectively
1862
-
HIJAZI ET AL.
network directly integrated on the same board or platform
while other systems have their feeding networks built on
separate boards and connected with cables. These differences
make it difcult to nd a common and fair way to compare
sizes of different systems, so here we propose a method which
is based on three points: (a) we let the system be inscribed in a
sphere of radius R, this radius will be used in the gure of
merit to represent the size of the system, (b) if the feeding
network is integrated with the antennas on the same board
then it will be accounted for in the size calculation, and,
otherwise, it will be disregarded, (c) after calculating the radius
of the sphere it will be normalised by the wavelength at the
centre frequency. Based on all the above, the gure of merit is
expressed as follows:
FoMWFD ¼log 10SIC Gain FBW
R=λcð18Þ
FBW ¼fu−fl
fcð19Þ
R¼1
2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l2þw2þh2
qð20Þ
SIC: average self‐interference cancellation level.
Gain: average system gain.
FBW: fractional bandwidth.
fu: upper frequency.
fl: lower frequency.
fc: central frequency.
λc: wavelength at the central frequency.
R: radius of the sphere circumscribing the full‐duplex
system.
l: length of the full‐duplex system.
w: width of the full‐duplex system.
h: height of the full‐duplex system.
The proposed gure of merit is an initial attempt to combine
different parameters of in‐band full‐duplex antenna congura-
tions or topologies in an expressive way to evaluate the perfor-
mances of the proposed systems and compare them to each
other. However, depending on the targeted application, it might
be modied to incorporate other metrics related to the eld of
application, such as the type of polarisation, the number of
polarisations (singular or dual), number of antennas used, an-
tenna efciency and level of side lobes, and the performance of
the feeding network. Finally, Table 1compares this work to other
in‐band full‐duplex antenna topologies previously published in
the literature, especially in terms of bandwidth, level of self‐
interference cancellation, system gain and size, and the newly
introduced gure of merit where it can be noticed that our
proposed system achieves the highest score. Also, up to our
knowledge, it works over the widest bandwidth (a decade
bandwidth) and reaches the highest upper frequency (40 GHz)
which is conditioned by a minimum gain level of 5 dBi, however,
otherwise, the highest matching frequency can go up to 49 GHz.
5
|
CONCLUSION
In this study, an ultra‐wideband full‐duplex system consisting
of four Vivaldi antennas and two microstrip‐to‐slotline baluns
is presented. The prototype was built on RO4003C substrate
(εr¼3:55, tan δ=0.0027, h=203.2 μm). The proposed
system can achieve 50 dB of self‐interference cancella-
tion at the antenna level and a gain better than 5 dBi over the
4–40 GHz frequency range with a highly directive radiation
pattern, which proves that in‐band full‐duplex can be extended
to ultra‐wideband operation. The proposed system is a general‐
purpose tool that can be used in various applications especially
those that require high data‐rate links, or recongurability over
a wide bandwidth. The performance of the proposed system
can be further enhanced by considering the following
perspectives:
▪ Due to the relatively high loss tangent of the used substrate
and due to fabrication imperfections, the fabricated baluns
suffer from high insertion losses, especially at high fre-
quencies, which decreases the total gain of the system. To
resolve these issues the baluns and antennas can be designed
and fabricated on a different substrate with lower tangent
loss, RT/Duroid 5880 (εr¼2:2, tan δ=0.0009) for
example. Our primary simulations conrm that the current
matching bandwidth (4–40 GHz) can still be maintained
using the latter substrate with a thickness of 0.254 mm.
▪ Also, integrating the baluns and antennas on the same board
is currently under consideration, which allows us to get rid
of the phase‐paired cables and the 2.92 mm connectors, and
that reduces the losses in the system.
TABLE 1A table comparing several wideband antenna systems for in‐band full‐duplex applications
Reference Frequency range (GHz) FBW SIC (dB) Gain (dBi) l wh (cm £cm £cm) R(cm) FoM
WFD
[4]
a
4–8 0.67 45 25 40 £40 £25.3 30.984 4.531
[22]
a
0.5–2 1.2 45 5 20 £20 £12.5 15.462 4.77
[23]
a
0.8–1.7 0.72 40 4 60 £60 £30 45 3.784
[16]
a
6–19 1.04 60 12 38 £13 £19 22.215 4.651
This work 4–40 1.64 64 7.8 8 £8£9.4 7.55 5.072
a
Some parameters' values were extracted from gures or graphs.
HIJAZI ET AL.
-
1863
▪ Moreover, the low substrate thickness is necessary for the
wideband performance of the fabricated devices, however,
this comes at the expense of the fragility of these devices. To
encounter this problem, we designed a 3D‐printed support
to hold the antennas and absorb any external mechanical
tensions. However, as an alternative solution, one can in-
crease the thickness of the substrate while sacricing a
portion of the obtained bandwidth.
▪ In addition to the above, other than in‐band full‐duplex
applications, the whole system can be used as one dual‐
polarised antenna array, and the way it is implemented,
and fed can greatly enhance the level of cross‐polarisation
over a very wide bandwidth.
▪ Finally, the whole system can be used to generate circularly
polarised waves by modifying the feeding network. The
circular polarisation can be obtained by feeding the four
antennas signals having equal amplitudes and having phases
of 0°, 90°, 180°, and 270°, consecutively.
ACKNOWLEDGEMENTS
This project has been supported by The French Directorate
General of Armaments (DGA), the European Regional
Development Fund (ERDF) of the European Union, the
Brittany Region (France), the Departmental Council of Fini-
stère and Brest Métropole as part of the Cyber‐SSI project
within the framework of the Brittany 2015–2020 State‐Region
Contract (CPER).
CONFLICT OF INTEREST
The authors declare that there is no conict of interest.
PERMISSION TO REPRODUCE MATERIALS
FROM OTHER SOURCES
None.
DATA AVAILABILITY STATEMENT
The data that support the ndings of this study are available
from the corresponding author upon reasonable request.
ORCID
Hadi Hijazi
https://orcid.org/0000-0001-7337-452X
REFERENCES
1. IT Union: IMT trafc estimates for the years 2020 to 2030. Report ITU
(2015)
2. Sabharwal, A., et al.: In‐band full‐duplex wireless: challenges and op-
portunities. IEEE J. Sel. Area Commun. 32(9), 1637–1652 (2014)
3. Bharadia, D., McMilin, E., Katti, S.: Full duplex radios. In: Proceedings of
the ACM SIGCOMM 2013 Conference on SIGCOMM, pp. 375–386
(2013)
4. Prasannakumar, P.V., Elmansouri, M.A., Filipovic, D.S.: Broadband
monostatic simultaneous transmit and receive reector antenna system.
In: 2017 IEEE International Symposium on Antennas and Propagation
and USNC/URSI National Radio Science Meeting, pp. 553–554. IEEE
(2017)
5. Etellisi, E.A., Elmansouri, M.A., Filipovic, D.S.: Wideband monostatic
co‐polarized co‐channel simultaneous transmit and receive broadside
circular array antenna. IEEE Trans. Antenn. Propag. 67(2), 843–852
(2018)
6. Venkatakrishnan, S.B., Alwan, E.A., Volakis, J.L.: Wideband RF self‐
interference cancellation circuit for phased array simultaneous transmit
and receive systems. IEEE Access. 6, 3425–3432 (2018)
7. Van den Broek, D.J., Klumperink, E.A., Nauta, B.: An in‐band full‐
duplex radio receiver with a passive vector modulator downmixer for
self‐interference cancellation. IEEE J. Solid. State Circ. 50(12),
3003–3014 (2015)
8. Liu, Y., et al.: Digitally assisted analog interference cancellation for in‐
band full‐duplex radios. IEEE Commun. Lett. 21(5), 1079–1082 (2017)
9. Amin, M., Hossain, M., Atiquzzaman, M.: In‐band full duplex wireless
LANs: medium access control protocols, design issues and their chal-
lenges. J. Inf. 11(4), 216 (2020)
10. Van Nguyen, B., Jung, H., Kim, K.: Physical layer security schemes for
full‐duplex cooperative systems: state of the art and beyond. IEEE
Commun. Mag. 56(11), 131–137 (2018)
11. Liu, G., et al.: In‐band full‐duplex relaying: a survey, research issues and
challenges. Commun. Surv. Tutorials, IEEE. 17(2), 500–524 (2015)
12. Debaillie, B., et al.: In‐band full‐duplex transceiver technology for 5G
mobile networks. In: ESSCIRC Conference 2015‐41st European Solid‐
State Circuits Conference (ESSCIRC), pp. 84–87. IEEE (2015)
13. Stove, A.G.: Linear FMCW radar techniques. IEE Proceedings F (Radar
and Signal Processing), vol. 139, pp. 343–350. IET (1992)
14. Darabkh, K.A., et al.: A–Z overview of the in‐band full‐duplex cognitive
radio networks. Comput. Commun. 145, 66–95 (2019)
15. Du, X., et al.: MIMO broadcast channel with continuous feedback using
full‐duplex radios. In: 2014 48th Asilomar Conference on Signals, Sys-
tems and Computers, pp. 1701–1705. IEEE (2014)
16. Prasannakumar, P.V., Elmansouri, M.A., Filipovic, D.S.: Wideband
decoupling techniques for dual‐polarized Bi‐static simultaneous transmit
and receive antenna subsystem. IEEE Trans. Antenn. Propag. 65(10),
4991–5001 (2017)
17. Wu, J., Li, M., Behdad, N.: A wideband, unidirectional circularly polarized
antenna for full‐duplex applications. IEEE Trans. Antenn. Propag. 66(3),
1559–1563 (2018)
18. Prasannakumar, P.V., et al.: Wideband quasi‐monostatic simultaneous
transmit and receive reector antenna. IEEE Trans. Antenn. Propag.
68(4), 2630–2637 (2019)
19. Everett, E., Sahai, A., Sabharwal, A.: Passive self‐interference suppres-
sion for full‐duplex infrastructure nodes. IEEE Trans. Wireless Com-
mun. 13(2), 680–694 (2014)
20. Ha, J., et al.: Monostatic Co‐polarized full‐duplex antenna with left‐or
right‐hand circular polarization. IEEE Trans. Antenn. Propag. 65(10),
5103–5111 (2017)
21. Moulder, W.F., Perry, B.T., Herd, J.S.: Wideband Antenna array for
simultaneous transmit and receive (STAR) applications. In: 2014 IEEE
Antennas and Propagation Society International Symposium (APSURSI),
pp. 243–244. IEEE (2014)
22. Lian, R., et al.: A high‐isolation, ultra‐wideband simultaneous transmit
and receive antenna with monopole‐like radiation characteristics. IEEE
Trans. Antenn. Propag. 66(2), 1002–1007 (2017)
23. Etellisi, E.A., Elmansouri, M.A., Filipovic, D.: Broadband full‐duplex
monostatic circular‐antenna arrays: circular arrays reaching simulta-
neous transmit and receive operation. IEEE Antenn. Propag. Mag. 60(5),
62–77 (2018)
24. Alwan, E.A., Hovsepian, A., Volakis, J.L.: Ultra‐wideband dual polari-
zation arrays with collocated elements for high isolation simultaneous
transmit and receive systems. In: 2016 IEEE International Symposium
on Phased Array Systems and Technology (PAST), pp. 1–3. IEEE (2016)
25. Hovsepian, A., Alwan, E.A., Volakis, J.L.: Wideband scanning array of
spiral antennas for simultaneous transmit and receive (STAR). In: 2017
IEEE International Symposium on Antennas and Propagation and
USNC/URSI National Radio Science Meeting, pp. 487–488. IEEE
(2017)
26. Zhong, J., Alwan, E.A., Volakis, J.L.: Ultra‐wideband dual‐linear polarized
phased array with 60° scanning for simultaneous transmit and receive
systems. In: 2017 International Workshop on Antenna Technology: Small
Antennas, Innovative Structures, and Applications (iWAT), pp. 140–141.
IEEE (2017)
1864
-
HIJAZI ET AL.
27. Elmansouri, M.A., Kee, A.J., Filipovic, D.S.: Wideband antenna array for
simultaneous transmit and receive (STAR) applications. IEEE Antenn.
Wireless Propag. Lett. 16, 1277–1280 (2016)
28. Etellisi, E.A., Elmansouri, M.A., Filipovic, D.S.: Wideband monostatic
simultaneous transmit and receive (STAR) antenna. IEEE Trans. Antenn.
Propag. 64(1), 6–15 (2015)
29. Etellisi, E.A., Elmansouri, M.A., Filipovic, D.S.: Wideband monostatic
spiral array for full‐duplex applications. In: 2017 IEEE International
Symposium on Antennas and Propagation and USNC/URSI National
Radio Science Meeting, pp. 1101–1102. IEEE (2017)
30. Filipovic, D.S., Elmansouri, M., Etellisi, E.A.: On wideband simultaneous
transmit and receive (STAR) with a single aperture. In: 2016 IEEE In-
ternational Symposium on Antennas and Propagation (APSURSI),
pp. 1075–1076. IEEE (2016)
31. Kurokawa, K.: Power waves and the scattering matrix. IEEE Trans.
Microw. Theor. Tech. 13(2), 194–202 (1965)
32. Rohde, U., et al.: Ultra wide band balun/180° power divider using
microstrip‐slotline‐microstrip transition. In: 2015 IEEE MTT‐S
International Microwave and RF Conference (IMaRC), pp. 400–404.
IEEE (2015)
33. Hijazi, H., et al.: 4‐40 GHz in‐phase/180° out‐of‐phase power dividers
with enhanced isolation. In: 14th European Conference on Antennas and
Propagation–EuCAP 2020, pp. 1–4 (2020)
34. Chan, K.K.M., Tan, A.E.C., Rambabu, K.: Decade bandwidth circularly
polarized antenna array. IEEE Trans. Antenn. Propag. 61(11), 5435–5443
(2013)
How to cite this article: Hijazi, H., et al.: Ultra‐
wideband antenna system for in‐band full‐duplex
applications. IET Microw. Antennas Propag. 15(15),
1853–1865 (2021). https://doi.org/10.1049/mia2.
12194
HIJAZI ET AL.
-
1865
Content uploaded by Marc Le Roy
Author content
All content in this area was uploaded by Marc Le Roy on Nov 17, 2021
Content may be subject to copyright.