BALANCED ANTIPODAL VIVALDI ANTENNA FOR BREAST

CANCER DETECTION
Jeremie Bourqui*, Michal Okoniewski †, Elise C. Fear †
*Department of Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada, Email:
bourquij@ucalgary.ca

† Department of Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada

Keywords: Breast cancer detection, balanced antipodal reasonably small antenna size (total length less than 100mm)
Vivaldi antenna (BAVA), ultra-wideband (UWB) antenna, is desired.
microwave imaging.
Different antennas have been designed and measured
Abstract including, for the TSAR system, the resistively-loaded dipole
[4] which has proven its imaging ability but suffers from low
A new ultra-wideband antenna for tissue sensing adaptive efficiency and directivity. A tapered slot antenna [8] has also
radar system is presented in this paper. The antenna is based demonstrated promising results, however we believe that
on a balanced antipodal Vivaldi design. The measurements there is still room for improvement. Other antennas for radar-
showed that the antenna has better that -10dB return loss based breast imaging include the ridged pyramidal horn
above 2.4GHz. Simulation indicated that the antennas is antenna with curved launching plane [11] used in the MIST
capable of sending very short pulses with fidelity above 0.96 system, the stacked patched antenna [6] developed at the
at a distance of 20mm and up to 0.994 further away from the University of Bristol and finally a planar tapered slot antenna
antenna aperture. 70% of the accepted energy is radiated in [1] operating in air.
the endfire direction with a beamwidth of 34mm by 44mm on
a plane 20mm away. In this paper the balanced antipodal Vivaldi antenna (BAVA)
design [7] is investigated as a potential antenna candidate for
1 Introduction TSAR. Its geometry, the design process and its performance
characterisation in the near-field are presented. Measurement
Microwave breast imaging has been proposed to assist in the results for S11 are also provided.
early detection of breast cancer [2]. Radar-based microwave
breast imaging approaches involve illuminating the breast External
Stacking
with an ultra-wideband pulse of microwaves and detecting Substrate (Ground)
reflections. The reflections are then processed to create Layers
images that indicate the presence and location of tumors in
the breast. Radar-based imaging systems include microwave
imaging via space time (MIST) beamforming [5], time-
reversal (TR) algorithm [9], and the method considered in this
paper, Tissue Sensing Adaptive Radar (TSAR) [4]. A key
component of these systems is the antenna that is used to
radiate and receive the ultra-wideband pulses.

For the TSAR system, the antenna must operate in matching

medium, such as canola oil, which exhibits a relative
permittivity of εr=2.5. Since the purpose of the antenna is to Central
transmit a short pulse in time it needs to be broadband, so a Supportive Stacking (Conductor)
reflection coefficient (S11) less than -10dB over the frequency Substrates Substrate Layer
range from 2 to 12GHz is required. Moreover, the signal
processing of TSAR requires that the shape of the emitted Figure 1: BAVA construction with the 3 copper layers and the
pulse be closely related to the excitation pulse and the 4 dielectric layers.
radiation pattern to be endfire-directed with a narrow
beamwidth. Given that the antenna is placed very close to the
breast, all of the radiation requirements need to be fulfilled in
the near field region. Finally, for practical reasons, a
2 Antenna Design Curve A P B C
Wts − Wg Wg
Et Pt 0 − At
2*(ePt *Lt − 1) 2
2.1 Construction and Geometry Wts
Ef Af Pf Lt+Lts − Af
The antenna consists of three copper layers; the two external 2
layers are connected to the feeding line ground planes and the Wts + Wa Wts
Ea Pa Lt+Lts − − Aa
central layer is connected to the signal conductor of the 2*(e Pa *La − 1) 2
feeding line (Figure 1). The copper layers are supported by
two dielectric substrates and two additional dielectric layers Table 1: Exponential curves parameters based on the system
are stacked on each side of the antenna. The feed is a gradual coordinate origin of Figure 2.
transition between a stripline to a tri-strip transmission line
(TL). The conductor width increases linearly while the 2.2 FDTD Simulation and Performance Metrics
ground width decreases exponentially to keep a constant
impedance along the transition. The tri-strip TL extends for a The antenna is simulated using SEMCAD X from SPEAG
short distance and then the grounds and conductor start to AG, Switzerland. A coaxial TL is connected to the stripline
flare in opposite directions with exponential curvatures to and represents the SMA to stripline transition used in
create the antenna aperture. practice. The modelled antenna is presented in Figure 3. The
total simulation space has a size of 4Mcells.
Lt Lts La
Y
X

Et Z
Wg

Wa
W

o
Ea
Wts
Ws

Ef
Z
X
L
T

Y o
T1
T2

X
Figure 3: BAVA FDTD model. For clarity, only the bottom
stacking substrate layer is shown on this figure. The view is
Figure 2: Geometry and dimensions of the BAVA. Top and
flipped compared to Figures 1 and 2. Note that, in the
side view with the coordinate axis shown on left and origin
simulation, the coordinate system origin is positioned at the
labelled as o.
antenna aperture centre.
The geometry parameters are presented in Figure 2: W, L and
The performance of the antenna is evaluated in terms of: S11,
T are the overall width, length and thickness respectively, Ws
half energy beamwidth (HEBW), forward radiation efficiency
and Wg are the width of the stripline and corresponding
(FRE) and fidelity. The HEBW is defined on a plane
ground plane, Wts is the width of the tri-strip TL, Wa defines
orthogonal to the endfire direction (X+) situated at a given
the actual aperture width, Lt, Lts and La correspond to the
distance away from the antenna aperture. It corresponds to the
transition, tri-strip TL and aperture length respectively. The
width of the area where the energy is larger or equal to half of
dielectric thicknesses are defined by T1 for the supportive
the maximum value on the plane. The FRE is defined as the
substrates and T2 for the stacking layers. Finally, the
amount of energy radiated in the endfire direction in contrast
exponential curves for the transition, flare and aperture (Et,
with the total antenna input energy. To compute the FRE, a
Ef, Ea) are defined by the following equation:
large plane sensor is placed orthogonal to the endfire
direction at the antenna aperture (i.e. at x=0). The total
z = ± A * e P*( x − B ) + C (1) radiated energy passing through this plane sensor is
computed. The area of the sensor is established by
where A is the scaling factor, P the exponential rate, B the determining when a 20% enlargement of its edges results in
shifting value and C the offset. These parameters are defined an increase of the FRE that is less than 1%.
for each curve in Table 1.
HEBW and FRE are calculated using the energy flux density First, the material to be used as substrate is studied. The study
(EFD) which is computed in the time domain at a spatial shows that dielectric with permittivity slightly higher than the
point as follows: surrounding environment (εr=2.5) produces superior radiation
behaviour while the thickness of the supportive substrate (T1)
Tsim
 J  does not significantly influence the antenna performance. The
EFD(x, y, z) = ∫ E(x, y, z, t) × H(x, y, z, t) dt







 m2  (2) thickness T2 is not critical but a thicker substrate leads to
0  
better antenna operation.
where E and H are the field values in the time domain from
Once the dielectric material has been selected, the influence
the simulation and Tsim is the total simulation duration.
of increasing the aperture variables W, Wa and Pa has been
investigated, as summarised in Table 2. Increasing the
The fidelity measures how faithfully the excitation pulse is
aperture or total width improves the S11 performance but
transmitted or received by the antenna and therefore reflects
usually degrades the other metrics. A small exponential rate
the distortion due to the frequency band limitation and phase
(Pa) is preferred which actually translates to a longer aperture.
non-linearity [10]. Only the transmitting fidelity is of interest
in this design. Its value (F(x,y,z)) is calculated using Equation
Param. S11 HEBW FRE Fidelity
(3). It consists of maximising the cross correlation between a
component of the E-field measured at a specific position W + - o -
(E(x,y,z,t)) and a reference signal r(t) which is the time Wa + o o -
derivative of the excitation signal. Both signals are Pa o - o -
normalised as described in Equations (4) and (5). The z
component of the E-field dominates the radiated field Table 2: Influence of increasing the W, Wa and Pa parameters
therefore the fidelity is calculated using this component only. on the antenna performances. +: Improvement, o: no
significant change, -: deterioration.
 Tref 
F(x, y, z) = max  ∫ Eˆ z (x, y, z, t + τ)r(t)
ˆ  [1] (3) The flare parameters Af and Pf influence the matching and the
 
τ  0  FRE of the antenna. A sharp flare curve increases the endfire
radiation but at the expense of degrading the S11 performance.
E z (x, y, z, t)
Ê z (x, y, z, t) = 1 The transition dimensions Wg, Ws and Wts are defined to
 Tsim 2 (4) provide TLs with 50 Ohms characteristic impedance at both
 ∫ E z (x, y, z, t) dt 
2

 0  ends. Pt is selected to get the best transmission coefficient

(S21) when two transitions are placed back to back.
r(t)
r̂(t) = 1
2.4 Final Design and Performances
Tref
 2 (5)
 ∫ r(t) dt 
2 The selected dielectric is RT/duroid® 6002 from Rogers
 0  Corporation which has a relative permittivity of 2.94. The
thickness T1 is chosen to be 1.524mm to accommodate the
Tref is the time length of the reference signal and Tsim remains coax-stripline transition and T2=3.048mm which is the
the total simulation time. thickest material available. The lengths were already defined
but are listed here again: L=74mm, Lt=23mm, Lts=1mm,
S11 is simulated using the standard broadband excitation La=50mm. Finally the remaining parameters are: W=44mm,
implemented by SEMCAD X, however the simulations for Wa=34mm, Wg=10, Ws=2mm, Wts=2.24mm, T= ~9.2mm,
the EFD and fidelity are computed using a specific ultra- Pt=-0.15, Pf=0.4, Pa=0.05 and Af=0.1.
wideband pulse [3] of the form
For this design, S11 is below -10dB above 2.2GHz, as shown
− ( t − t 0 ) / τ2
in Figure 4. No upper limit has been found up to 17GHz
V(t) = V0 * ( t − t 0 ) *e (6) (simulation limit). The FRE is reasonably high with 70% of
the total input energy radiating toward the endfire direction.
where V0 is used to adjust the amplitude of the pulse, τ=62.5 The HEBW at 20mm away from the aperture has a size of
ps and t0=4τ. 44mm by 28mm in y and z directions, respectively. The total
EFD in the near field is presented in Figure 5.
2.3 Parametric Study and Design Process
To facilitate the design, a parametric study has been carried
out. Note that for practical reasons, the overall length (L) is
selected to be 74mm with Lt=23mm, Lts=1mm and
La=50mm.
 0 1

-5
0.98
-10

-15

Fidelity [1]
0.96
S11 [dB]

-20

-25 0.94

-30
0.92
-35

-40 0.9
2 4 6 8 10 12 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Frequency [GHz] x [m]

Figure 4: Simulated reflection coefficient (S11). Figure 6: Simulated fidelity of the z component E-field along
the x axis at the centre of the antenna aperture (y=0, z=0).

0.6
Simulation
Theory
0.4
Fidelity = 0.991
0.2
Ez [V/m]

0
(a)

-0.2

-0.4
0 0.5 1 1.5 2
Time [s] -9
x 10

Figure 7: Simulated z component E-field at 50mm away from

the antenna aperture centre compared to the theoretical
radiated pulse.
(b)
3 Antenna Implementation and Measurements
2
Figure 5: Simulated total EFD in dB normalised to 1 pJ/m , A photolithography process is used to pattern the copper layer
on the (a) Y Plane @ y=0 and (b) Z Plane @ z=0. and the bonding of the different layers is achieved using
Rogers 3001 bonding film. The feeding consists of a custom
Figure 6 shows the fidelity as a function of the distance from made SMA to stripline transition which has an overall length
the antenna aperture centre. The fidelity improves further of 6 mm (connector excluded). A picture of the final antenna
from the antenna and can be considered as nearly stable is presented in Figure 8. Its measured S11 matches reasonably
20mm away. the simulation, as shown in Figure 9. The cut-off frequency
increases from 2.2 to 2.4GHz, which remains acceptable. It
The z component E-field at x=50mm is presented in Figure 7. should be noted that the bonding film has a relative
The radiated pulse fits almost perfectly the excitation permittivity of 2.28 and was not included in the simulation
derivative without any significant late time ringing. model.
 Acknowledgements
The authors would like to acknowledge the technical support
of F. Hickli, B. Isenor and J. Shelley, all of the University of
Calgary, Calgary, AB, Canada.

References
[1] A. M. Abbosh, H. K. Kan, and M. E. Bialkowski,
“Compact Ultra-Wideband Planar Tapered Slot Antenna
for Use in a Microwave Imaging System”, Microw. Opt.
Tech. Letters, vol. 48, no 11, pp 2212-2216, (2006).
[2] E. C. Fear, S. C. Hagness, P. M. Meaney, M.
Okoniewski, and M. A. Stuchly, “Enhancing Breast
Tumor Detection with Near-Field Imaging,” IEEE
Microw. Magazine, vol. 3, no. 1, pp. 48–56, (2002).
[3] E. C. Fear, X. Li, S. C. Hagness and M. A. Stuchly,
Figure 8: Picture of the implemented antenna. “Confocal Microwave Imaging for Breast Cancer
Detection: Localization of Tumors in Three
0 Dimensions”, IEEE Trans. Biomed. Eng., vol. 49, no. 8,
Measurement pp 812-822, (2002).
-5 Simulation [4] J.M. Sill and E.C. Fear, “Tissue Sensing Adaptive Radar
-10
for Breast Cancer Detection: Experimental Investigation
of Simple Tumour Models,” IEEE Trans. Microw.
-15 Theory Tech., vol. 53, pp. 3312-3319, (2005).
S11 [dB]

[5] E. J. Bond, X. Li, S. C. Hagness and B. D. Van Veen,

-20 “Microwave Imaging via Space Time Beamforming for
-25 Early Detection of Breast Cancer,” IEEE Trans.
Antennas Propag., vol. 51, no. 8, pp. 1690-1705, (2003).
-30 [6] I. Craddock, “Wideband Antennas for Biomedical
Imaging”, chap. 20 in “Ultra-widband Antennas and
-35 Propagation for Communications, Radar and Imaging”,
-40 ed. B. Allen and al., London: Wiley, (2006).
2 4 6 8 10 12 [7] J. D. S. Langley, P. S. Hall and P. Newham, “Novel
Frequency [GHz]
ultrawide-bandwidth Vivaldi antenna with low
crosspolarisation”, Electron. Letters, vol. 29, no. 23, pp
Figure 9: Measured reflection coefficient (S11) compared with 2004-2005, (1993).
simulated result. [8] M. H. Shenouda and E. C. Fear, “Ultrawideband
Antenna Design for Breast Tumor Detection," Abstracts
4 Conclusion of the 2006 URSI North American Radio Science
Meeting, 1 pp,(2006).
A BAVA with a very promising performance has been [9] P. Kosmas and C. M. Rappaport, “FDTD-Based Time
designed and constructed for the TSAR system. The Reversal for Microwave Breast Cancer Detection—
measurements show S11 below -10dB from 2.4GHz to at least Localization in Three Dimensions”, IEEE Trans.
12GHz, and simulations indicate low S11 up to 17GHz. The Microw. Theory Tech., vol. 54, no. 4, pp. 1921-1927,
antenna’s forward radiating efficiency is computed as 70%. (2006).
The beamwidth of the antenna is 34mm by 44mm wide at [10] T. P. Montoya and G. S. Smith, “A Study of Pulse
20mm away from the antenna aperture. The fidelity is above Radiation from Several Broad-Band Loaded
0.96 at distances of 20mm and further. The sensor has a total Monopoles”, IEEE Trans. Antennas Propag., vol. 44, no
size of 80mm long by 44mm wide. The lower limit of the 8, pp 1172-1182, (1996).
desired frequency bandwidth (2-12GHz) was not reached, [11] X. Li, S. C, Hagness, M. K. Choi and D. W. van der
however the antenna still shows better than -7dB reflection at Weide, “Numerical and Experimental Investigation of
that frequency. The fidelity calculations indicate that this is an Ultrawideband Ridged Pyramidal Horn Antenna with
sufficient for the proper operation of the antenna. Therefore, Curved Launching Plane for Pulse Radiation,” IEEE
the BAVA antenna meets the design requirements for a Ant. Wireless Propag. Letters. vol. 2, pp. 259-262,
TSAR sensor. Future work includes scanning breast (2003).
phantoms and testing imaging capabilities with this sensor.