You are on page 1of 4


A Novel Antenna for Ultra Wide Band (UWB) Intelligent Antenna

Bruno Biscontini*, Sofian Hamid*, Franz Demmelt and Peter Russer*
*Institute for High Frequency Engineering,
Munich, Germany 80333, Email:
Telephone: (+49) 89 289 23373, Fax: (+49) 89 289 23365
tRohde & Schwarz, Munich, Germany

Abstract- We present a novel antenna for UWB intelligent y

antenna systems. The UWB antenna element uses a planar S'
antenna element with an impedance bandwidth from 3 -26
GHz. The antenna design is based on genetic algorithm (GA)
in connection with method of moments (MoM). To achieve
compact size and unidirectional and narrow beam radiation
characteristics the antenna is embedded into a dielectric
substrate-superstrates multilayered structure. We show that
a substrate-superstrate configuration allows to achieve an 0
v 1.
impedance bandwidth from 3 -26 GHz and unidirectional 0a) SurfaceS'.x 0 x
and narrow radiation characteristics. The space filtering (a) Surface S'. (b) Meshed surface S'.
properties of the antenna are also investigated. The results
are validated by comparing the computation with numerical Fig. 1. In this Figure an arbitrary perfectly electric conductive (PEC)
simulations with commercial CAD software. surface S' and its approximation with triangular meshes is shown.
Index Terms- UWB systems, intelligent antennas, planar
multilayered antennas.
In this work, a novel antenna for intelligent UWB
I. INTRODUCTION systems applications is presented. A planar antenna is
Intelligent antenna systems for ultra wide band (UWB) embedded in a dielectric substrate-superstrates structure.
applications are of growing importance in a multitude of Due to the small size requirement, the geometry of the
applications. antenna element is chosen to be planar. The synthesis and
In recent years, considerable research in communica- optimization of the antenna is performed using a genetic
tion and microwave engineering has been devoted to the algorithm (GA) in connection with method of moments
development (UWB) systems and technology [2], [4]. (MoM). To achieve directional radiation without using a
This technology is expected to offer high data rates over ground plane the substrate-superstrates structure is used.
short distances without suffering from the multi-path in- We show that the impedance bandwidth of the antenna
terferences [5]. Among many other possible applications, is in the range of 3 -26 GHz also with the substrate-
UWB antennas are also used in high resolution radar superstrates structure. The phase center of the antenna is
imaging, for secure transmission and in multi-user network investigated by a directional antenna transfer function in
applications [6]. frequency domain [3] for different elevation angles.
Although the technology allows the digital signal pro- II. THEORY
cessing part of the intelligent antennas to work in a very
broad frequency band, the real limitation of the system For the design of the antenna element of the radiating
seems to be the antenna itself [6]. Indeed there are several element the electric field integral equation (EFIE) ie used.
issues in designing intelligent UWB systems related with We consider an arbitrary perfect electric conductor (PEC)
the antenna properties. In particular also in the most surface S'. The scattered electric field ES, can be com-
general case of non-line-of-sight propagation an antenna puted from the surface current by
with high directivity allows space division multiplex trans-
mission [3]. This yields severe specification requirements ni (n A PS) jwfs,e(r r')js(r')
d (f3s e(r, r')djs (r'))
for the antenna design like a frequency independent phase -
center, frequency independent radiation pattern, and large where Js (r') is the surface density current differential
impedance bandwidth. Popular broadband antennas nowa- form and Qe (r, r') is the free space scalar Green's double
days like bow-tie, vivaldi, TEM horn antennas, etc., cannot one-form [1]. To solve the integral equation (1) we use
fulfill all those requirements [4]. MoM expanding the current into a series of vector basis

0-7803-9542-5/06/$20.00 C2006 IEEE


functions called RWG [7]. Since the RWG basis functions
are supported by a triangular mesh, an arbitrary surface S' -5
has to be meshed with triangles, Fig. l(b). -10



In the following the connection of the electromagnetic 0 -20 fo.c

problem with the genetic algorithms (GA) will be de- =
CC -25
scribed. To synthesize and optimize the UWB radiating
element we use genetic algorithms (GA) [9], [10]. We
exploit the GA via the geometry of the radiating element in -35
the following way. The RWG basis function are described -40'
by a pair of triangles that form one rectangular cell, 0 3 6 9 12 15 18 21
Frequency [GHz]
24 27
Fig. l(b). We map the cells numbering in each row into
a bit representation. In the GA algorithm terminology Fig. 2. The best structure its return loss.
these bits are called genes. The combinations of genes
form chromosomes. These chromosomes represent the
equivalent bit description for all the rows and therefore until the geometry fulfilling the wished requirements is
the antenna geometry. The genetic algorithm will shape not generated.
the geometry in the way to reach the required antenna
specifications. The specifications are related to the antenna B. Radiating Element Synthesis
design and the antenna geometry. To synthesize the radiating element we follow the pro-
cedure that we have described in the previous section. As
A. Genetic algorithm process first step we impose general condition of the geometry
In the following we will describe the genetic process itself. Since we deal with a 2D radiating element, we
to generate the antenna structure. The GA process starts define the largest and smallest possible structures that
with the initialization a of random chromosomes. Once may be examined by the GA. This is done specifying the
the geometry of the antenna is determined by means of maximum and the minimum area.
chromosomes, the electromagnetic field is computed solv- The number of cells used in this simulation is N
ing the integral equation (1) with MoM and the antenna R x C = 42 x C cells, where R is the fixed number of
parameters, input impedance, return loss, radiation pattern, row while C is the variable number of cells in every row,
etc., are computed. In this case, for example, as design ranging from 1 to 23, or C C {1: 23}. Therefore, each
parameter we have chosen to minimize the return loss, at gene (number of cells in a row) is represented by 5 bits
least below- 12 dB in the frequency range of 3- 26 GHz. since the maximum possible number of cells in a row is
To achieve that, a scoring scheme is developed to evaluate 23.
every structure generated randomly (at the beginning), or The two rows that intersect the feeding edge location are
genetically (for the following generation). The scoring set to have only one cell. In this way, the maximum number
system will be treated in more details in the next section. of cells is (42 x 23) + 2 = 968 cells while the minimum
The next step is to rank all structures to determine number of cells is 42 cells. Exploiting the symmetry of the
the quality of the structure in order to decide if the GA structure, the number of parameters (cells in each row) to
process has to be continued or ended. If the decision is to be optimized from the genetic algorithm are reduced to
continue the process, the next step is to select the parents, half, i.e., 42/2 = 21. Since the feeding point is in the
i.e., the best selected chromosomes, and perform mating center of the structure, for symmetry reasons, the number
to obtain new chromosomes (children). To open new of freely changing row cells must be odd. This yields
possibilities and enrich diversity, mutation is performed. to 12 row cells of variable position. Before running the
Mutation means to change randomly or accordingly to GA, some settings and operators must be decided. We
some criteria the bits sequence characterizing each gene chose the population number, i.e., the maximum number
and therefore the cells positions in the structure. In this of examinable structures, Npop = 16. We note that a
way, the solution is prevented to reach the goal too fast or pure random search would lead to 1221 = 4.6 x 1022
stop in a local minima. Moreover, to ensure that the best possible structures. The number of iterations or generation
chromosomes of each generation are better or at least the number is 40. The mutation per iteration is set to 10% of
same quality (score) of the previous one, the mutation is the total bits in all chromosomes (population). Random
not performed for the best chromosome of each generation. single point crossover is used here to mate the selected
The same process is repeated for the new generations parents. The selection strategy for parents is based on
TABLE I 2025
-4, ~L4
Return Loss Score 3, pt3
iiiiiiiiiiiiii F-2, t2
-4 < RL < 0 4000 F- , 1t
-8 < RL < -4 500
-10 < RL < -8 100
o' N2
-12 < RL < -10 15 -.
-15 < RL < -12 3 C x
-20 < RL < -15 1
RL < -20 O Fig. 3. Antenna element embedded in a dielectric multilayer structure.

proportionate selection (roulette-wheel selection). In this a substrate of dielectric constant E1 = 2.2. To account
way the probability of each chromosome to be selected for the effect of the dielectric, we reduce the size of
is proportional to its score value. The scoring scheme antenna element inside the dielectric by 35%. Additional
is chosen according to the return loss. In particular, we superstrate layers with permittivities E2 = 75, E3 = 25 and
score the return loss assigning a weight-score (a positive E4 = 9, respectively are used to maximize the radiation
number) to every frequency point, as it is shown in Table I. efficiency in the upper half-plane (Fig. 3). The thicknesses
The total score for each structure is calculated by summing are chosen in a way that the excitation of surface waves is
the score at every frequency point. To higher scores, is minimized. The overall behavior of the surface waves is
assigned a low probability to be chosen as a parent. similar to the single layer case. The lower dominant mode
According to this criterion, the GA process is guided to is the TM1. The lower higher order modes are the TE,
converge to a structure that has a return loss below -12 dB and TM2 [1].
or -15 dB or -20 dB. Indeed, the score for those values of
return loss are very small (between 0 and 3). V. RESULTS AND VALIDATION
One of the most severe problems of this approach is
In the following, we discuss the numerical computations
that the computation time is proportional to the number
of the input impendence, the radiation pattern and the
of frequency points. In order to reduce the number of
frequency points, the structure is first optimized at the phase center of the antenna.
range 1.5 GHz to 11.5 GHz using 21 frequency points,
A. Input impedance
or 500 MHz spacing. The size of the rectangle cell is
0.3 cm x 0.3 cm. This size corresponds to 0.015A and The input impedance is computed by MoM and the
0.115A at the lowest and highest frequency optimization, results are compared with these obtained with CST Mi-
respectively. After achieving the results, considering the crowave Studio. In Fig. 4, we can see that the real and
scaling properties of antennas, the optimized result will the imaginary part of the input impedance are centered
be scaled down by a factor of two and examined for the around 80 Q and 0 Q respectively within a frequency band
range 3-26 GHz. Fig. 2 shows the best generated structure of 3 -26 GHz. A good agreement of the results obtained
and its return loss. with both methods is observed. The computation time with
the MoM is by a factor 50 lower than with CST microwave
For the use of the generated antenna element in UWB
arrays, an important challenge is to achieve a unidirec- B. Antenna phase center
tional radiation pattern (+y) without ground plane as
shown in Fig. 3. It is well known, that the presence The stability of the phase center is investigate by means
of the ground plane deteriorates the performance of the of directional transfer functions [3]. The directional trans-
antenna at higher frequencies. In this work we propose fer function also depends on the elevation angle 0 (Fig. 5).
the application of a substrate-superstrate configuration to The phase is plotted as function of 0 and the frequency.
achieve a unidirectional radiation pattern. The antenna A linear behavior can be observed up to approximately
element is embedded into a dielectric multilayer structure 25 GHz.
as shown in Fig. 3.
It is already understood [8] that choosing properly the C. Radiation Pattern
value of the dielectric permittivity or permeability and the Fig. 6 shows a 3D view of the antenna radiation pattern.
thicknesses one can enhance gain and improve radiation As we can see the influence of the substrate-supertestrates
efficiency. In this case, to achieve a directional radiation tends to generate a unidirectional radiation characteristic
pattern, we place the radiating element in the middle of intermediate and high frequencies.

"0 5 10 15 20 25 30 35
Frequency [GHz]

f. . . . .
r, n e
6 GHz
0 -
-50- (a) Radiation pattern at (b) Radiation pattern
N + MoM
_1 -
E -00
6 GHz. at 12 GHz.
.4 En_
0 5 10 15 20 25 30 35
Frequency [GHz]

Fig. 4. Real and imaginary part of the input impedance.

4 4
f = 1GHz
-500 f = 3.9GHz
a (c) Radiation pattern at (d) Radiation pattern at
" 1000. f = 6.8GHz 18 GHz. 21 GHz.
f = 9.7GHz

-f = 12.6GHz Fig. 6. In figure a 3D view of the radiation pattern as function of the

frequency is shown.
.-2000. - e -f = 15.5GHz

R -2500
-0-f = 18.4GHz
0~ f = 21.3GHz
X= -3000
f = 24.2GHz [4] H. G. Schantz, "Introduction to ultra wideband antennas," IEEE
------- -- ---- -4---
-4-f = 27.1GHz UWBST Conference Proceedings, 2003.
-3500. [5] Z. N. Chen, X. H. Wu, H. F. Li, and M. Y W. Chia, "Consideration
- b-f = 30GHz
for source pulses and antennas in UWB radio systems," IEEE Trans.
-4000L0 10 20 30 40 50 Antennas Propagat., vol. 52, No. 7, pp. 1739-1748, Jul. 2004.
Elevation Angle 0 [degree] [6] T. Do-Hong and P. Russer, " Wideband Smart Antenna Array
Applications ," IEEE Microwave Magazine, pp. 57-67, March 2004.
Fig. 5. In figure the phase behavior of the antenna in space and frequency [7] S. M. Rao, D. R. Wilton, and A. W. Glisson, "Electromagnetic
is shown. scattering by surfaces of arbitrary shape," IEEE Trans. Antennas
Propagat., vol.AP-30, pp. 409-418, May 1982.
[8] N.G. Alexopoulos and D. R. Jackson, "Fundamental superstrate
(cover) effects on printed circuit," IEEE Trans. Antennas and Prop-
VI. CONCLUSIONS agat., vol. AP-32, pp.807-816, Aug. 1984.
[9] D. R. Jackson and N.G. Alexopoulos, "Gain enhancement methods
A novel antenna for UWB intelligent antenna systems for printed circuit antennas," IEEE Trans. Antennas and Propagat.,
has been presented. The planar antenna element has been vol. AP-33, No. 9, pp. 976-987, Sept. 1985.
designed by the combination of a genetic algorithm and [10] F. J. Villegas, T. Cwik, Y Rahmat-Samii and M. Manteghi, "A
parallel electromagnetic genetic-algorithm (EGO) application for
the electric field integral equation method. We have shown patch antenna design," IEEE Trans. Antennas Propagat., vol.52, No.
that a UWB antenna element embedded in a dielectric 9, pp. 2424-2435, Sept. 2004.
substrate-superstrates structure exhibits impedance band-
width, unidirectional radiation characteristics and constant
phase center from 3 to 25 GHz.
The authors wish to acknowledge Rohde & Schwarz
company for the support to the work.

[1] P. Russer, Electromagnetics, Microwave Circuit and Antenna Design
for Communications Engineering, Artech House, Boston, 2003.
[2] R. A. Scholtz, D. M. Pozar and W. Namgoong,"Ultra-wideband
radio," EURASIP Journal on Applied Signal Processing, vol. 3, pp.
252-272, 2005.
[3] W. Sorgel and W. Wiesbeck, "Influence of the antennas on the
ultra-wideband transmission," EURASIP Journal on Applied Signal
Processing, pp. 296-305, 2005.