Methods to Design Microstrip Antennas for Modern Applications

Full text

9
Methods to Design Microstrip Antennas for
Modern Applications
K. Siakavara
Aristotle University of Thessaloniki
Dept. of Physics, Radiocommunications Laboratory,
54124 Thessalonik,
Greece
1. Introduction
The evolution of modern wireless communications systems has increased dramatically the
demand for antennas, capable to be embedded in portable, or not, devices which serve a
wireless land mobile or terrestrial-satellite network. With time and requirements, these
devices become smaller in size and hence the antennas required for transmit and receive
signals have also to be smaller and lightweight. As a matter of fact, microstrip antennas can
meet these requirements. As they are lightweight and have low profile it is feasible them to
be structured conformally to the mounting hosts. Moreover, they are easy fabricated, have
low cost and are easy integrated into arrays or into microwave printed circuits. So, they are
attractive choices for the above mentioned type of applications.
For all that, the design of a microstrip antenna is not always an easy problem and the
antenna designer is faced with difficulties coming from a) the inherent disadvantages of a
printed resonant antenna element, for example the narrow impedance bandwidth, and b)
the various requirements of the specific applications, which concern the operation of the
radiating element, and can not be satisfied by a printed scheme with an ordinary
configuration. For example, it would be demanded, the microstrip element to have gain
characteristics that potentially incommensurate to its size or/and frequency bandwidth
greater than the element could give, taking into account that it operates as a resonant cavity.
Moreover, the rapid development in the field of Land Mobile Telephony as well as in the
field of Wireless Local Area Networks(WLANs) demands devices capable to operate in
more than one frequency bands. So the design of a printed antenna with intend to conform
to multiple communications protocols, for example the IEEE 802.11b/g, in the band of
2.4GHz, and the IEEE 802.11a at 5.3GHz and 5.8GHz, would be a difficult task but at the
same time a challenge for the designer. Counting in the above the possibility the device, and
so the antenna, to serve terrestrial and also satellite navigation systems the problem of the
antenna design is even more complicated.
In this chapter techniques will be analysed, to design microstrip antennas that combine the
attributes mentioned above which make them suitable for modern communications
applications. Specific examples will be also presented for every case.

Microstrip Antennas
174
2. Bandwidth enhancement and multiband operation
2.1 Bandwidth
Conventional microstrip antennas have a conducting patch printed on a grounded dielectric
substrate and operate as resonant cavity elements. This operation leads inherently to
narrow impedance bandwidth which is a barrier for microstrip antennas applications in
wireless communications. Moreover in many of these applications, as further requirement
would be a multi-frequency operation. So, the enhancement of the bandwidth and the
achievement of multifrequency operation are major challenges for the antenna designer and
many techniques have been proposed for this purpose.
The fundamental definition of the bandwidth of an antenna is the difference between the
upper and lower frequencies of operation( H
f and L
f respectively)
HL
b
wf f
=
− (1)
For all that, the spectrum managers often use a variety of different bandwidth definitions,
including fractional or percent bandwidth. These measures of relative bandwidth require
the calculation of a central frequency, which is either the arithmetic or geometric average of
the upper and lower frequencies. The center frequency is defined as the arithmetic average
of the upper and lower frequencies
HL
C
ff
f2
+
= (2)
An arithmetic average yields the central frequency when frequency is considered on a linear
scale. An alternate definition of center frequency involves the geometric average
CHL
fff= (3)
The geometric average yields the center frequency when frequency is considered on a
logarithmic scale and is less commonly used. So the arithmetic average should be assumed
unless otherwise is specified.
The fractional bandwidth of a system is the ratio of the bandwidth to the center
frequency(either the geometric or the arithmetic definition is used)
C
b
w
BW f
= (4)
Alternatively, fractional bandwidth may be defined on a percentage basis
C
bw
BW% 100%
f
= (5)
Since the geometric definition of center frequency always yields a frequency smaller than
the arithmetic average, fractional bandwidths calculated using the geometric definition are
always larger than the arithmetic ones. So, the designer of the antenna has to pay attention
on which definition is used. Although specific limit values of BW in order an antenna to be
considered as broadband do not exist , criteria to rank an antenna as ultra-wideband have
been enacted. In accordance with these criterions an antenna is characterized as UWB if its

Methods to Design Microstrip Antennas for Modern Applications
175
fractional bandwidth (based on the arithmetic central frequency) exceeds the value of
0.25(Defense Advanced Research Projects Agency_DRPA) or the value of 0.2(Federal
Communications Commission_FCC).
2.2 Bandwidth enhancement
The impedance frequency bandwidth of a microstrip antenna depends primarily on both the
thickness and the dielectric permittivity of the substrate. A thick substrate with a low
dielectric permittivity can increase the bandwidth of the printed patch. Both these selections
could be a solution of the problem of bandwidth enhancement if the thickness of the
substrate did not a)pose difficulties in integration of the antenna with other microwave
circuits, and b)cause some other problems such as the surface wave propagation and the
large inductive image part of the input impedance of the antenna, which makes its
resonance unfeasible. Thus, a reasonable thickness should be considered in the selection of
substrate and the bandwidth would be enhanced using additional techniques. The most
common and effective of them, are: a) the loading of the surface of the printed element with
slots of appropriate shape b)the texturing of narrow or wide slits at the boundary of the
microstrip patch . Other effective techniques used for the enhancement of the bandwidth is
the utilization of a)stacked, shorted or not patches, [1]-[3] and b) extra microstrip resonators
[4]. The technique of stacked patches is based on the fact that the bandwidth is in general
proportional to the antenna volume measured in wavelengths but at the same time a
relatively large volume is a disadvantage for many applications. The utilization of
additional parasitic patches of different size directly- or gap-coupled to the main patch is an
effective method but results to an increased antenna size which would also be undesired.
Superior to these methods are the techniques of slot loading or texturing the patches by slits
because they ensure the small size and the low profile of the antennas.
2.2.1 Slot loaded patches
The slot loading is made by two ways. By one of them, slots are embedded in the printed
patch. Their dimensions and positions are properly selected in order to the first two
broadside-radiation modes of the patch be perturbed such that their resonance frequencies
get close to each other to form a wide impedance bandwidth. The slots would be of various
shapes as, toothbrush(Fig. 1a)[5], double bend (Fig. 1b)[6] , cross (Fig. 1c) [7], or U-shape.
(a)
(b)
(c)
Fig. 1. Microstrip antennas loaded with slots of various shapes : a) ‘toothbrush’ b) ‘double
bend c)cross

Microstrip Antennas
176
The microstrip elements etched with a U-slot could be rectangular or triangle patches (Fig.
2)and are perhaps the most popular among the antennas of this type[8]-[15], as they have
been proved to be versatile radiating elements: they can be designed not only for wideband
applications but also for dual- and triple-band as well as for circular polarization operation.
It has been found that the U-slot loaded patch can provide impedance bandwidth in excess
of 30% for an air substrate thickness of 0.08λo and in excess of 20% for material substrate of
similar thickness [16].
(a)
(b)
(c)
Fig. 2. The geometry of the U-slot patch antenna.
Τhe U-slot is consisted of two parallel vertical rectangular slots and an horizontal
rectangular one. The parameters that affect the broadband performance of the patch antenna
are the slot length and width and the position of the slot. It is noted that the slot width
should be small relative to the slot length and the higher resonant mode is sensitive to the
length variation of the horizontal slot, whereas the lower resonant mode strongly depends
on the perimeter of the U-slot. Unfortunately mathematical formulas, even empirical, by
which the accurate geometry of the structure patch-U-slot could be find in order the antenna
to have a pre-specified operation, are not available yet. Empirical formulas introduced in the
past[10] are limited to electrically thin substrates which lead to small bandwidth. Thus the
antenna designers should adjust the dimensions and the position of the slot by iterative
trials or, potentially, by employment of an evolutionary technique of prediction and
optimization as in §4 is discussed.
An alternative version of the U-slot patch is the half U-slot patch Fig3a, which maintain an
impedance bandwidth similar to that of a full U-slot patch having the further advantage of
the smaller size[17]-[18]. This is due to the fact that the electrical current distribution is
symmetrical along the line of symmetry of the full U-slot printed element. So, removing half
of the patch does not appreciably affect the current paths and , hence, the resonant behavior
of the structure. Moreover, a shorting wall (Fig. 3b) can be integrated to reduce the size of
the full U-slot [19]-[20] or the half –U-slot patch antenna[17]- [21]. These small-size wide
bandwidth designs are particularly suitable for handset devices.
2.2.2 Slit loaded patches
By slit cutting at the boundaries, instead of slot cutting on the surface, of the patch, similar
broadband operation can be obtained. The resulting configuration is E-shaped or suchlike
and the design can be applied to the antennas with rectangular, circular or triangular
patches(Fig. 4)[22]- [24].

Methods to Design Microstrip Antennas for Modern Applications
177
(a)
(b)
Fig. 3. Patterns of a)Half U-slot loaded patch b) U-slot loaded patch with shorting wall
(a)
(b)
(c)
Fig. 4. The geometry of slit loaded microstrip antennas
The wideband performance of the slit loaded patch is based, similarly to the method of slot-
loading, on the excitation of more than one adjacent resonant modes. Moreover the presence
of the slits in the vicinity of the feeding probe could add a capacitive load at the input
impedance of the patch. This capacitive load could effectively contribute to the resonance of
the patch because can counteract the inductive part of the probe’s input impedance. It is
noticed that this inductive part would inevitably be large if a thick substrate is chosen for
wideband operation. So, the insertion of slits enhances by two ways the width of the
operation band, and it has been reported that bandwidth greater than 25% can be achieved.
The width of the frequency band of the antenna can be controlled by the slits’ length and
width and the slits’ position. The slits divide the patch in three or more parts and at each
one corresponds an equivalent circuit of resonance(Fig. 5)[25]-[26]. In order the three regions
of resonance to overlap, forming a unified wideband range of operation, the slits could
potentially have unequal size.

Microstrip Antennas
178
Fig. 5. The structure and the equivalent circuits of E-shaped printed antenna
A lightly different patch with as much bandwidth as the E-patch can be realized by
configuring the slits in a way that results in a printed element with meander shape (fig 6) [27].
Fig. 6. Top and side view of the meander shaped printed antenna
Another slit loading process to obtain broad band operation is to create several slits at the
boundary of the patch. This technique provides the patch with the additional advantage of a
size smaller than that of the previous configurations. As an example a patch with five-couple
staggered slits is presented in figure 7a [28]. A hybrid feed, inset feed combined with probe, is
used for easy impedance matching. A quantitative analysis can prove the possibility to expand
the operation bandwidth and reduce the antenna size. When the structure parameters of the
antenna are selected properly, two TM10 modes can be excited simultaneously. The simple
electrical current sketches of the two modes on the patch are shown in Fig. 7b. It can be found
that the current paths of the modes overlap at the top portion of the patch whereas they
occupy a different structure branch at the bottom portion of the patch.
In addition, the current path lengths are determined by the patch length and the slit depths.
The asymmetric slit loading technique leads to different resonance lengths of the two TM10
modes. The two modes can be matched by adjusting the depth of the feed slits and the patch
length because their input impedances vary with the inset feed position. Moreover due to
the slits, the currents of the modes flow on meandering paths that ensure the required
length for resonance at the modes’ frequencies, whereas the entire length of the patch is
smaller compared with that of ordinary rectangular microstrip antennas. Furthermore, the
current paths of the TM10 modes are along the same direction, thus the radiation
characteristics of the modes are similar, in accordance to the radiation theory. Due to the
above mechanism of operation, the bandwidth and the size of the antenna could be
controlled by the slit number and the geometry of the whole configuration.

Methods to Design Microstrip Antennas for Modern Applications
179
(a)
(b)
Fig. 7. a)Configuration of multi-slit loaded low profile microstrip antenna b)Current
sketches of the two modes on the patch
In all cases of the slit loading, the slits play an important role to control the wideband
behavior of the antenna. However to find the proper geometry of the textured patch is
usually a difficult process because there are no mathematical formulas, thus, no prediction
could be made. The whole process is iterative and is applied by simulations using a high
frequency electromagnetic field simulation software. The set up of the process is decided by
the designer. Usually the first step of the procedure is to select the dielectric substrate and
design the initial unmodified patch. Then slots are etched and their position and geometry
are iteratively adjusted to tune the required frequency and the bandwidth. The whole
strategy potentially includes steps of simultaneous small variations of the patch dimensions
and of the substrate with intend to control the frequency tuning and the bandwidth[24],
[27]. Besides this strategy, which is substantially a trial and error one, Artificial Network
Algorithms or Optimization techniques as Genetic Algorithms, Particle Swarm
Optimization(PSO) or Differential Evolution could be applied(See §4). By these methods all
the material and geometrical parameters of the antenna structure are simultaneously taken
into account in every step of the design process.
2.2.3 Ultra-Wideband microstrip antennas
There are two criteria available for identifying when an antenna may be considered ultra-
wideband(UWB). One definition(by Defense Advanced Research Projects Agency report)
requires a UWB to have fractional bandwidth greater than 0.25. An alternate and more
recent definition, by Federal Communications Commission(FCC) places the limit at 0.2.
Using fractional bandwidth (eq. 4), in mathematical form it is
{
ff 0.25 DARPA
HL
BW 2 0.2 FCC
ff
HL
−
==
+ (6)
In the past, different types of small antennas meeting the above requirement have been
investigated: resistively loaded pyramidal horn antennas, stacked patch antennas and
tapered slot antennas or spiral-shaped antennas. Among the existing designs of printed
structures, the ones that provide the required operational bandwidth while maintain

Microstrip Antennas
180
sufficient radiation efficiency, simple design structure and low fabrication cost are: a) the
spiral shaped antenna b)the tapered slot, especially the antipodal tapered, as the Vivaldi
antennas and c) the bowtie shaped antennas.
a. Spiral antennas
Spiral antennas have been the subject of research for several decades and have emerged as
leading candidates for applications requiring circularly polarized broadband antennas.
Spirals can have a single arm or multiple arms and have been realized in microstrip form
as well as in slot form. But spiral antennas have shortcomings as well. They require deeper
lossy cavities and special attention to impedance transformer network design. Recently
several designs of planar spiral antennas without complicated feeding networks were
proposed. They are planar, easy to fabricate and do not need a matching network. The spiral
could have annular(fig. 8a ) or rectangular(fig. 8b) shape. The microstrip Archimedean
annular spiral antenna can be achieved by the traditional Archimedean spiral concept with a
dielectric-substrate and a metal ground plane. The spiral configuration could increase the
gain of the antenna. However the main lobe splits at the higher frequency band.
The radiation principle of this kind of microstrip configuration is based on the mirror
concept. To reach the property of ultra-wide band, the height of the dielectric –slab and the
value of the dielectric constant is very important. At the lower part of frequency band, the
maximum radiation of the antenna can be found along the axis. By the rising of the
frequency the front to back ratio increases and one-sided radiation can be achieved. The
direction of the maximum radiation starts to deflect at a specific frequency, depending on
the structural characteristics of the spiral. As the frequency rises further the direction
deflects to both sides of the axis which is perpendicular to the antenna’s surface. The
decreasing of the radiation along the axis is explained as follows: The height of the
dielectric-slab is related to the central frequency of the antenna, being about a quarter of the
wavelength at this frequency, as it is defined inside the dielectric. So, at the higher frequency
band the height tends to be one-half of the wavelength at the respective frequencies. In this
case, the phase of the original signal, compared with that of the wave reflected by the the
ground plane is just the opposite when it reaches the spiral plane . After the superposition
of the signals, the radiation at the broadside decreases and the main lobe splits. A
modification of the substrate is proposed in this case. A cylinder with a certain radius
should be subtracted out from the center of the dielectric so as to improve the gain along the
axial direction at higher frequency band. The physical thickness of the substrate in the
hollow part corresponds to an electrical length smaller than that in the rest of the dielectric
slab. The parameters of the hollow have proper values in order this length to be equal to λ/4
(a)
(b)
Fig. 8. a) Structural pattern of the microstrip circular spiral printed antenna b)Sketch of the
rectangular spiral antenna

Methods to Design Microstrip Antennas for Modern Applications
181
instead of λ/2. So, the condition for maximum radiation toward the direction perpendicular
to the surface of the antenna is fulfilled. At the operation inside the wide frequency band,
the effective region of the antenna at each frequency includes the annuluses of the
Archimedean spiral whose perimeters are approximately equal to the respective
wavelength. By this modification, the proper radius of the hollow cylinder and the proper
dielectric constant of the substrate, we can lead the spiral configuration to a wide band
operation, broadside radiation and a gain greater than that of the ordinary spiral structure.
b. Vivaldi antenna
Vivaldi antenna was invented by Gibson in 1979, and has been widely investigated and
used in various applications due to its broad bandwidth, low cross polarization and high
directivity[30]-[31]. It is a kind of traveling-wave planar antenna, guiding the wave from a
narrower slotline to a wider slotline, which varies by exponential rule, where the wave is
radiated out from the horn opening. Due to the non-uniform shape of the printed area,
different parts are activated at different frequencies. In virtue of this performance the entire
scheme can operate in a wide band of frequencies. However it is difficult to select the best
parameters and variances to lead the antenna up to the best radiation performance.
The standard model of the Vivaldi antenna is shown in Fig. 9. There are more than ten
parameters to be designed. Furthermore, one important part of the antenna is the
microstrip-slotline transition feeding structure. This type of feeding is the most commonly
used. Microstrip and slotline are deposited in the different sides of the substrate. The
incident wave power can be coupled from the microstrip to slotline and then the power can
be transmitted to the air through the exponential tapered parts. The impedance match at the
coupling point is usually difficult. Three are the parameters that affect the radiation
performance of the antenna: the microstrip-slotline transition magnitudes of strip stub, the
slotline circle cavity and the gradual rates of the antenna. So, the design of the slotline
structure, and the calculation of the antenna’ s exponential tapered parts are the basic steps
of the design.
The configuration of the exponential tapered part is governed by three parameters: a) the
slotline hatch(H) b)the antenna tapered length(L) and c) the exponential gradual (α) , as
shown in Fig. 9. According to experience, the higher (FH) and lower FL cut-off frequencies
determine the slotline hatch. Generally, the widest hatch(H) is equal to 1.3λcL , where λcL
corresponds to the wavelength of the lower cut-off frequency; the narrowest hatch is equal
to 0.2λcH , where λcH corresponds to the wavelength of the higher cut-off frequency. The
antenna tapered length(L) is equal to 3~5 times the wavelength of the central operation
frequency.
The exponential gradual ratio can be determined by the exponential gradual curve
αx
12
yCe C
=
+ (7)
Where 1
C and 2
C can be calculated using 21
21
1αxαx
yy
Cee
−
=− and
21
21
αxαx
12
2αxαx
ye ye
Cee
−
=−
And 11
(x ,y ) and 22
(x , y ) are the begin point and the end point of the exponential gradual
curve.
A modified form of the traditional Vivaldi antenna is the dual exponentially tapered
antipodal antenna(DETASA)[31]. It is realized by exponentially tapering both the inner and

Microstrip Antennas
182
(a) (b)
Fig. 9. Printed Vivaldi antenna a) the model b) the manufactured structure
the outer edges of the radiating flare and placing the mirrored flares on both sides of the
substrate(Fig. 10a). It is a slow leaky end-fire traveling wave antenna. The electromagnetic
wave travels down the gradually curved paths of the flares. As the separation between the
flares increases, the wave becomes progressively weaker and radiates away. A wideband
performance can be achieved using DETASA, in virtue of its inherently simple wideband
transition from microstrip line to tapered slot flare through parallel strips.
The rule for the design of the radiating flare is described by equations 8a and 8b.
()
(
)
sft
inner s w s s inner
xcc/2cexpky
⎡
⎤
=± − + + ⋅
⎣
⎦ (8a)
()
(
)
sfo
outer s w w w outer
xcc/2cexpky
⎡
⎤
=± − + + ⋅
⎣
⎦ (8b)
Where inner
x and outer
x are the horizontal distances from the inner and outer exponential
edges to the center line of the antenna. inner
y and outer
y are the vertical distances from the
edges to the bottom line of the antenna. The other parameters are obtained by the trial-and
error optimization procedure using full-wave electromagnetic simulation or applying
optimization techniques as Genetic Algorithms, Particle Swarm Optimization, etc.
In order to further extend the operational bandwidth and he radiation performance, a
modification of DETASA is introduced by merging the exponential flare with a circular
tapering termination before the edges reach the end of the substrate(Fig. 10b). It creates a
longer path for the current flow making the antenna bandwidth greater and smoother. In
order to avoid sharp gradient change between the exponential edges and the circular
(a) (b)
Fig. 10. Two Configurations of the Dual Exponential Antipodal Vivaldi Antenna

Methods to Design Microstrip Antennas for Modern Applications
183
tapering, optimization is performed to search for appropriate tapering parameters along the
pre-fixed inner and outer exponential edges and to provide defining parameters for the
antenna configuration. Extra spacing is kept between the flares and the edge of the
substrate. This additional space helps to reduce ringing of the time-domain impulse
response.
With respect to the feeding method of this type of UWB antennas, the transition from the
feeding source to the radiation flares is designed by the high frequency transmission line
theory. Τhe exact theoretical analysis in detail, and the corresponding realization depend on
the Vivaldi antenna configurations(traditional or modified).
c) Bowtie printed antennas
Another patch configuration that drives the antenna system to multifrequency operation is the
bowtie shape. Bowtie microstrip antennas have become attractive candidates in the present
day communication systems due to their size, that is smaller than the size of a conventional
rectangular patch although they have similar characteristics and operation at the same
frequency. The fundamental shape of a bowtie microstrip antenna is shown in fig. 11.
Fig. 11. Geometry of the bow-tie microstrip antenna
The bowtie printed object comes, substantially, from a rectangular patch via modification
and the equations for the approximate calculation of the resonance frequency fr in terms of
the geometrical and material parameter values are,[32]
(
)
(
)
()()
c
r
eff
W2L W 2L
cL
f 1.152 2W2LS2L
2L
+Δ + +Δ
⎡
⎤
⎣
⎦
=+Δ +Δ
ε (9)
Where
()
()
c
eff
c
eff
WW
0.412h 0.3 0.262
2h
LWW
0.258 0.813
2h
+
⎛⎞
ε+ +
⎜⎟
⎝⎠
Δ=
⎡
+⎤
⎛⎞
ε− +
⎜⎟
⎢
⎥
⎝⎠
⎣
⎦
(10)
and
1/ 2
rr
eff
c
11
24h 1
22WW
−
⎛⎞
ε+ ε−
⎛⎞⎛⎞
ε= + +
⎜⎟
⎜⎟⎜⎟
+
⎝⎠⎝⎠
⎝⎠
(11)

Microstrip Antennas
184
By altering the dimensions c
W and L and keeping W constant it is possible to change the
fundamental frequency. The equations given above are a good starting point for the antenna
design. However for the best impedance matching, the coordinates of the feed point must be
found through iterative simulations.
Modified bowtie type antennas have also been proposed fig.12, [33]-[34]. By the appropriate
values of the geometrical parameters a bandwidth value that exceeds 90% can be obtained
whereas the antenna exhibits small size.
(a)
(b)
Fig. 12. Two configurations of the double sided wideband bow-tie antenna
The antenna of fig 12a consists of two identical printed bows, one on the top and one on the
bottom of the substrate material. The top and the bottom bows are connected to the
microstrip feedline and the ground plane through a stub and mitered transition to match the
bow-tie with the selected characteristic impedance of the feedline.
By further modification of the bowtie antenna, namely by rounding of the bow patches, as
shown in fig. 12b, wider bandwidth, higher co-polarization and lower cross-polarization for
the UWB range can be achieved[34].
2.3 Multiband microstrip antennas
The rapid advances in the wireless communication industry demand novel antenna designs
that could be used in more than one frequency bands and that will allow size reduction. For
example mobile telephony’s services require portable devices compatible with
GSM900/DCS1800/UMTS2000 technology and the same equipment should also connect the
users to WLAN networks based on 802.11 standards (2.5GH/5GHz). So, the design of small
antennas suitable for these devices is of great interest. Many techniques have been proposed
for the design of radiating elements of this type, the great majority of which are microstrip
antennas. The common characteristic, of almost the total, of the multiband printed elements
is that they usually come from an initial patch of ordinary shape which in the following is
perturbed. On the basis of the way of the shape perturbation, the multiband microstrip
antennas would be classified in categories a) printed elements with incorporated slits or
slots. [35]-[40] b) patches with more than one radiating elements conductively connected
or/and inductively coupled[41] c) patches of specific shape as conductively connected cross
dipoles[42], the bowtie[46], and the spiral [47]-[48] d) more than one stacked patches[49]. A
separate category includes the fractal microstrip antennas which come from an initial simple
printed element and are developed by a recursive process(§3.5).

Methods to Design Microstrip Antennas for Modern Applications
185
At all the aforementioned categories the patches are produced starting from basic theoretical
concepts and the procedures of development focus to similar targets. However, it has to be
noticed that in no one of them a rigorous design process exists, namely a process which
would lead the designer to begin from the same initial concept and, aiming at specific
attributes of operation, to end in all cases in the same shape of patch. So, the design of a
multi-frequency patch is substantially an art and the ways of texturing the patch’ s surface
as well as the techniques, used for optimization of the configuration, are just useful tools.
The designing of a multi-band printed antenna requires that the efficient operational
features namely, gain greater than 0dB, almost uniform space distribution of the radiated
power, if necessary, circular polarization and mainly small reflection coefficient at the
feeding port must be ensured in all the frequency bands of operation. Furthermore all these
properties must not be obtained at the cost of a complex feeding network, a non compact
fabrication or an antenna arrangement of large size. Therefore the design of this type of
antennas is a difficult task and beyond the used standard techniques any novel concept or
modification of the existing methods would be useful and would lead to interesting antenna
schemes.
2.3.1 Slot loaded multiband microstrip antennas
The technique of slot cutting the surface of the printed antenna, besides the broadening of
the bandwidth has been proved effective in driving the patch to multi-frequency operation
[35]-[40]. Various slot shapes have been proposed for the texturing of the patch; some
indicative results are presented in figures 13 to 16. Cutting a rectangular patch with L-
shaped(Fig. 13a) or folded slit(fig. 13b) the antenna is considered to consist of two connected
resonators of different sizes. The shorting pins at the corner of the patch permit to reduce its
size, making this type of compact dual-band antenna suitable for applications in handset
mobile communication units. The dimensions of the larger and the smaller sub-patches(Fig.
13a) can be designed to roughly resonate as quarter-wavelength structures at the pre-
specified frequencies. In the alternative configuration(Fig. 13b), the smaller sub-patch, that
resonates at the higher frequency, starts from the feed point and extends into the center
portion of the rectangular patch. In this way it is encircled by the slit and is surrounded by
the outer larger sub-patch which resonates at the lower frequency.
Instead of using an L-shaped slit or a folded slit to obtain two separate sub-patches, an
embedded U-slot can be used(Fig. 13c). In this configuration the smaller rectangular, of
dimensions 22
LW×, resonates at the higher of the pre-specified frequencies and occupies
the central portion of the original rectangular patch of dimensions 11
LW
×
which resonates
at the lower frequency. The lower L
f and the upper H
f operating frequencies of this design
can be approximately determined from
L
11
c
f4(L W )
≅+ (12)
H
22
c
f4(L W )
≅+ (13)
Where c is the speed of light in free space

Microstrip Antennas
186
(a) (b)
(c)
Fig. 13. Geometries of shorted rectangular patch antenna with a) an L-shaped slit b) a folded
slit and c) U-slot, for dual frequency operation
In figures 14a and 14b alternate types of perturbation of the patch’s surface are shown[38].
The non radiating edges (with respect to the basic TM01 mode )of the initial rectangular
patch are modified by T-shaped notches in such a way that the current distribution of TM01
mode is similar to the current of the higher order TM21 mode (fig. 14b) driving the patch to
similar operational characteristics at both frequencies. A further modification of the patch by
two pairs of smaller T-notches with proper size results in triple frequency band
operation(Fig.14c).
(a) (b)
(c)
Fig. 14. (a), (c)Microstrip elements textured with single or multiple T-notches. (b) surface
current density at TM01 and TM21 modes.
The concept of texturing the surface of the patch by U-slots that can effectively lead to a
broadband operation can also be efficient in driving the element in multiband operation.
Two U-slots [15],of different size (fig. 15a) or unequal slits combined with tiny circular slots
(Fig15b)can ensure triple band operation[39].
Another configuration of microstrip antenna, loaded with slits is proposed in[40]. The patch
(fig. 16a) has annular ring shape, is etched on a two layered dielectric substrate and is

Methods to Design Microstrip Antennas for Modern Applications
187
(a)
(b)
Fig. 15. Configurations of tri-band microstrip antennas a) dually U-slot loaded patch b)patch
loaded with unequal slits and tiny circular slots
(a)
(b)
1.01.52.02.53.03.5
-35
-30
-25
-20
-15
-10
-5
0
|R| [dB]
frequency [GHz]
desired
desired by fitting
ANN
ANN (from ANN results)
(c)
Fig. 16. (a), (b) Top and side view of a dual-band microstrip structure with multilayered
substrate. (c) Reflection coefficient of the signal at the feeding point for a specific set of
geometrical and material parameter values.
covered by a dielectric superstrate (fig. 16b). The inherent advantage of a ring antenna is, the
property to resonate for a diameter less than λg/2 (λg is the guiding wavelength of the
equivalent linear microstrip line having width equal to that of the ring). This attribute
means a physical size smaller than the size of the respective circular disc resonating at the
same frequency. In the proposed configuration the total height of the substrate is high and
ensures the broadness of frequency bands(fig. 16c). Additionally the slits in the vicinity of
the position of the probe compensate efficiently the large inductive input impedance that
inevitably comes from the thickness of the substrate. Due to the large number of the
structural parameters of such an antenna an ordinary procedure, for example iterations with
gradual variations of the values of the parameters, could be laborious and perhaps
ineffective or unfeasible. So, a stochastic technique, which following an objective process ,
would yield a suitable antenna configuration.
The antenna of fig. 16a was designed using an Artificial Neural Network(ANN) algorithm(see
§4). The values of the structural parameters, as shown in figures 16a and 16b, were calculated
by a properly designed and trained ANN at the input of which, the plot (sampled) of the
scattering parameter of the signal at the feeding point was given. By this plot an antenna
capable to resonate at 1.572GHz(GPS)and at 2.45GHz(WLAN based on protocol 802.11) was
demanded from the ANN. The antenna implemented by the results of the output of the ANN,
exhibits operation that meet these requirements, as shown in fig. 16c.

Microstrip Antennas
188
2.3.2 Multiple patches
An alternative approach to the design of a multiband printed antenna is to use more than
one elements with different size that resonate at different frequencies. An indicative
structure [41] is shown in fig 17a. In detail, the antenna is an array of concentric microstrip
annular rings printed on a double dielectric layer. The multi-frequency performance comes
from the resonances of the individual rings and from the further modification made to the
rings. The entire configuration has two inherent advantages a) each annular ring resonates
having diameter less than λg/2, as explained for the antenna of fig 16a, b) a ring shaped
printed element allows other rings of smaller radius , resonating at different frequencies, to
be printed in the internal area, ensuring multi-frequency operation and at the same time
compactness.
(a)
(b)
Fig. 17. Top and side-view of a multi-frequency microstrip antenna with three conductively
connected rings
To find the frequencies of resonance of each ring it is necessary to solve the electromagnetic
problem of printed annular ring antennas fed by probes. The basic and rigorous method for
this solution, in spatial domain, is to determine the respective Green function [43]-[44]. The
final form of this Green function, expanded in series of orthonormal eigenfunctions, is:
nmn nmn nmn
nm Jn ΥnJn
nmn
Υnmn
11 1
G(ρ,φ|ρ,φ)[J(λρ)Υ(λρ)][ J ( λρ)
πCπCπC
1cos(nφ)cos(nφ)
Υ(λρ)]
πCλλ
′′
=
++
′
′
+⋅
−
∑∑
(14)
where ρ,φ
′′
are the coordinates of the feeding probe, a and b are the inner and outer radius
of the ring, mn
λ is the th
mn eigenvalue , n
J,n
Y are Bessel functions of the first and second
kind respectively and Jn
C and Yn
C are the constants produced via the ortho-normalization
process of the eigenfunctions
22
Jn n n
22
mn mn
Yn nn
CJ J
ba
[(λb)] [ ( λa)]
C22
YY
⎧⎫ ⎧⎫
′′
⎧⎫ ⎪⎪ ⎪⎪
=−
⎨ ⎬ ⎨⎬ ⎨⎬
⎩⎭ ′′
⎪⎪ ⎪⎪
⎩⎭ ⎩⎭
(15)
In order the orthonormal eigenfunctions to satisfy the electromagnetic boundary conditions
of the structure, the following equation must be satisfied

Methods to Design Microstrip Antennas for Modern Applications
189
n mn n mn n mn n mn
J(λa)Y ( λb) Y ( λa)J ( λb) 0
′′ ′′
+
= (16)
Equation 16 is the characteristic equation of the problem and its solution gives the
frequencies at which the ring resonates, that is
mn
mn
λ
f2πμε
= (17)
The aforementioned formulation can be used to solve the inverse problem, namely to find
the geometrical and material parameters of the ring antenna in order to resonate at the
desired frequencies.
On the basis of the above theory, an antenna configuration with three rings was
implemented in [41]. The rings were printed on the top side of a dielectric layer of 3.17mm
height( 2
h) and dielectric constant 2.15(RT/Duroid 5880). The layer was suspended at a
distance 4mm over the metallic ground plane, namely r1
ε1
=
and 1
h4mm=. The
frequencies of resonance of each one of the rings were calculated applying Eq. 17 and using
as εr the effective dielectric constant of the corresponding suspended microstrip line[45]
1
2
re 1 2
11
r2
hw1
ε[1 (c c l n )( 1)]
hh
ε
−
=+ − − (18)
where 44
22
12
11
hh
c (0.8621 0.125ln ) , c (0.4986 0.1397 ln )
hh
=− =−
The theoretically calculated frequencies of resonance are the frequencies at which each one
of the rings would resonate without the presence of any other element. Therefore, they are
just an approximate assessment of the expected resonances of the entire structure, due to the
mutual coupling between the rings and the additional modification made to the
construction. The modification was necessary because the input impedance of each ring was
large and caused strong reflection of the feeding signal. In order to reduce the input
impedance, narrow slits were created and the rings were conductively connected (Fig. 17a).
As a consequence the entire printed structure has the shape of a spiral. The width of the slits
and those of the printed interconnections of the rings, were found via repeated simulations,
the object of which was the achievement of better feed matching. This type of antenna
would also be designed using the techniques either of ANN or GA and PSO.
With respect to the feeding, two feeding probes were used and their positions were found
with the criterion of feed matching. The employment of two probes was necessary because a
single one could not excite all the theoretically predicted modes. Therefore at some
frequency ranges (e.g. at 1.39-1.5GHz and at 2.82-2.83GHz) both probes should be activated
while at the other frequencies the antenna should operate via one of them.
In Fig.18 the measured and simulated scattering coefficients at the input of both probes are
depicted. Due to the aforementioned reasons a difference between the theoretically
predicted resonate frequencies and the measured ones, appears (see Table 1). Moreover,
frequencies of operation smaller 1GHz exist. Their appearance is due to the fact that the path
of the current along the spiral is larger than the perimeter of even the larger of the rings.
These frequencies would not be predicted by the theoretical analysis, not even

Microstrip Antennas
190
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-30
-25
-20
-15
-10
-5
0
probe 1
simulated
measured
|Sin| [dB]
frequency [GHz]
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-45
-40
-35
-30
-25
-20
-15
-10
-5
0
|Sin| [dB]
frequency [GHz]
probe 2
simulated
measured
Fig. 18. Reflection coefficient of the signal at the feeding probes of the antenna shown in fig. 17
approximately, because in accordance to equations 16 and 17 the lower resonant frequency
of all three rings is the basic resonant frequency of the larger one and it was found equal to
1.26GHz, as shown in Table 1. The results obtained via simulation approximate the
respective measurements, in most of the frequency bands.
Table 1 includes the theoretical results for the frequencies of resonance and the respective
ones found experimentally.
Theoretical results
Frequency[GHz] mode Measured values [GHz]
1.26 n=0 m=1 (ring 1)
1.005-1.05 (probe1)
(probe1)
1.389-1.506 (probe2)
1.7 n=0 m=1(ring 2)
1.785 n=0 m=2(ring 1)
1.653-1.682 (probe2)
1.967-1.987 (probe1)
2.18 n=0 m=1(ring 3)
n=0 m=3(ring 1) 2.087-2.208 (probe2)
2.83 n=0 m=5(ring 1)
2.94 n=0 m=3(ring 2)
3.06 n=0 m=2(ring 3)
(probe2)
(probe2)
2.7-2.975 (probe1)
Table 1.
Due to the annular ring shape of the elements of the antenna, the components of the
radiated field do not differ by more than 5dB on both main planes, in all frequency bands.
Moreover the utilization of two probes offer the potential to increase the level of the gain
applying a proper phase shift between the probes. Indicative results of the field pattern at
2GHz are presented in fig. 19.
2.3.3 Bowtie patches loaded with slots
The basic principle of creating slots on the surface of a printed antenna in order to drive it in
multi-frequency operation can be applied to bowtie patches, as well. An indicative
configuration(see Fig. 20) is proposed in [46]. The first step of this design is to choose the
appropriate dimensions of the arms of the bowtie in order to ensure the first band of
frequency. The calculations can be made either using equations 9 to 11 or equations for the
design a triangular patch antenna.

Methods to Design Microstrip Antennas for Modern Applications
191
-60
-40
-20
0
030
60
90
120
150
180
210
240
270
300
330
-60
-40
-20
0 |G|,φ=90ο
|G|,φ=0ο
θ[deg]
(a)
-60
-40
-20
0
030
60
90
120
150
180
210
240
270
300
330
-60
-40
-20
0 |G|,φ=90Ο
θ[deg]
|G|,φ=0Ο
(b)
Fig. 19. Gain pattern at 2.0 GHz of the antenna of fig. 17: a) probe 1 is fed, and b) probes are
fed by 180o out of phase.
(a) (b)
(c)
(d)
Fig. 20. Bowtie microstrip antennas: a) the ordinary configuration, (b),(c),(d)modified
patches, by loading with slots of various shapes, for multiband operation.
The resonant frequency corresponding to the various modes TMmn of the equilateral
triangular microstrip antenna is [45]
22
r
r
2c
fmmnn
3
=++
αε (19)
Where c is the velocity of light in free space,
α
is the side length of the bow-tie patch and
r
ε is the dielectric constant of the substrate.

Microstrip Antennas
192
The above equation is valid when the triangular resonator is enclosed by a perfect magnetic
wall. In the case it is not valid the replacement of side length
α
by an effective value e
α has
been suggested. An approximate expression for e
α
produced by curve fitting experimental
and theoretical results for the resonant frequency for TM10 mode is given [45], by
22
e
r
rr
hh hh 1h
[1 2.199 12.853 16.436 6.182( ) 9.802 ( ) ]α=α+ − + + −
ααεαα
αε ε (20)
where h is the thickness of the substrate.
In the second step of the design process circular or six sided polygon slots are incorporated
to the bow patches. This texturing drives the antenna to dual-frequency operation: the initial
frequency, obtained by the first step of design, and a second desired one that can be
obtained by properly choosing the size of the slots (Figs. 20b and 20c). A tri-band
performance can be achieved cutting, in the two parts of the antenna, slots either of different
size or of different shape, as shown in Fig. 20d.
In all configurations implemented in [46], the printed elements are fabricated one on the
bottom and the other on the top side of the dielectric substrate. The feeding method depends
on the selected bow-tie configuration and the parameter values of the feeding network are
calculated by the high frequency transmission microstrip line theory
2.3.4 Spiral antennas
Multi-frequency operation can also be achieved using spiral printed antennas. A suitable
modification of the ordinary shapes of the printed spirals, especially of the rectangularly
shaped, has been proved to be effective to multi-band performance. Three indicative proposed
structures are shown in figure 21. The design of all three is based on a modified spiral line –
technique model and on the use of shorting –pin method[47]-[48]. An advantage of these
antennas is that they have the ability to resonate in frequencies much smaller than the other
multiband anennas can, having at the same time very small size. The key for this performance
is the increased length of the current path along the conductor area due to the spiral shape of
this conductor. For the structures of figures 21a and 21b, two ways of folding the conductor in
order to maximize its length were used. One is the standard rectangular folding and the other
is the folding the conductor back onto itself in the two dimensions. Each configuration has
three dependent variables for a constant antenna area, namely the number of bends in the
conductor, the conductor width and the gap between the conductors. Attention has to be
drawn to the thickness of the conductor because it has a lower limit, set by the diameter of the
probe and the shorting pins. Using these configurations, in [47], dual band operation was
obtained at frequencies being both smaller than 1GHz.
A multi band operation can also be obtained by the configuration of fig. 21c. In this case a
small number of bends for the spiral shape are used and the corners are truncated to allow
smooth current flow. The frequencies of resonance and the corresponding bandwidths as well
as the matching, are controlled by the line’s width and length, the gaps between the lines, the
position of the shorting pin and the feed point. The utilization of a dual slab substrate, the
lower slab being of air or foam, may enhance the bandwidth. This spiral configuration due to
the relatively small length of the conductor is recommended for high frequency applications.
In [48] by proper selection of the parameter values a dual band operation was achieved. One
from 2.37GHz to 2.47GHz and another, of large width, from 5GHz to 6.3 GHz.

Methods to Design Microstrip Antennas for Modern Applications
193
(a) (b)
(c)
Fig. 21. Microstrip spiral configurations a) the ordinary shape b)the conductor is folded back
onto itself c) small number of bends and truncated corners.
2.3.5 Fractal antennas
Fractal technique has been applied to many areas of science including fractal
electrodynamics in which fractal concepts are combined with electromagnetic theory for the
purpose of investigating a new class of radiation, propagation and scattering problems.
Thus, fractal antennas are an innovative type of radiating elements coming from the
inspired combination of these two theories.
The fractal antenna engineering focuses in two promising research areas : the first deals with
the analysis and design of fractal radiating elements and the second, concerns to the
application of the fractal theory to the design of antenna arrays[50]-[54]. Both antenna types
have attributes, highly desirable in military as well as in commercial sectors. The majority of
fractal antenna elements, are printed configurations, have compact size, low profile and
cost, multi-band operation, easy feeding and, potentially, their operation could be optimized
by suitable modification of their shape. On the other hand the application of the fractal
technique to the design of an antenna array can produce radiating systems of large size, and
as a consequence of high gain, with frequency-independent or multi-band operation and
radiation patterns with low side-lobe level. Moreover, the elements of the array can be fed
one by one, thus operating as Direct Radiating Antennas(DRAs) and can function as phased
arrays[55]-[57].
Fractal objects can be classified in two categories: deterministic and random. Deterministic
are those that are generated of several scaled-and rotated copies of themselves using a
recursive algorithm. Random fractals also contain elements of randomness that allow
simulation of natural phenomena. Procedures and algorithms for generating fractals both
deterministic and random have been found. However most of the proposed fractal antennas
have been designed with a deterministic fractal process.
The fractal technique is based on the idea of realization the operational characteristics of the
antenna by repeating an initial radiating structure in arbitrary or regular scales. The self
similarity properties of the produced antenna configurations are translated into their
electromagnetic behavior and since the initial antenna was introduced, the higher stages of
fractaly developed geometries would have, for example, the attribute of multi-frequency
operation. The basic scheme of a fractaly designed antenna is a generating, termed also
initiating, radiating element. In particular, the entire antenna can be formed recursively
through repetitive application of the generating element under a specified scaling factor
which is one of the parameters of the problem. This process is realized following potentially

Microstrip Antennas
194
two different strategies: By one of them, the repetition of the generating antenna is made in
such a way that its entire size gets larger from stage to stage of fractal building. By the
second strategy, the entire area, which the final antenna is permitted to occupy, is defined a
priori. Then, by the process of the proper repetition of the generator, the available area is
filled by scaled replicas of the generator.
This space filling leads to antenna configurations with electrically large lengths although
they have small size thus, constituting substantially a miniaturization technique producing
elements efficiently packed into small areas, therefore suitable for installation in portable
telecommunication devices.
The theoretical analysis as well as the interpretation of the operation of a microstrip fractal
antenna is based on the underlying relation between a fractal resonator and a microstrip
antenna. The physical problem of the operation of fractal resonators is often referred to, in
the literature, as the determination of the ‘fractal drum’ vibration modes. The simplest
example of a surface fractal resonator is the fractal drum, where a vibrating membrane is
bounded by a fractal curve, its vibration modes named ‘fractions’. The vibration modes of
fractal drums exhibit some interesting properties such as the existence of localized modes.
In these modes the vibration is strongly localized in certain parts of the membrane. These
vibrational states are obtained after the solution of the Helmholtz equation with the
appropriate boundary condition. For surface fractals, a distinction is made between
Neumann and Dirichlet fractions according to the boundary condition that is applied.
On the other hand microstrip antennas can be modeled, in a first approximation, as a cavity.
For a microstrip patch antenna of electrically small height of substrate, the field distribution
can be found with very good accuracy from the eigenfuntions of the Helmholtz equation
subject to the Neumann boundary condition. Therefore, it is expected that fractal boundary
microstrip antennas will exhibit vibration modes similar to those ones of the fractal drum.
The most popular fractal geometries, referred in the literature, which were proved efficient
in antenna design are the Koch fractal, the Sierpinski gasket or carpet fractal, the Hilbert
fractal, the Minkowski and the Square Curve fractals. All of them were proved to yield
printed antennas, planar or conformal, with multi-frequency performance as well as easy
feeding, satisfactory polarization properties and gain, whereas they have small size. All
these attributes make them an attractive choice for portable telecommunication equipment.
2.3.5 a) Koch: direct and inverse fractal islands
The Koch fractal microstrip patches are commonly used in virtue of their attractive
properties: they have small size, a single feeding port is enough and their higher order
modes result in directive radiation patterns[58]-[60]. The basic geometry of the Koch island
is obtained by replacing the sides of an equilateral triangle by a Koch curve. The triangle is
the generator and the higher stages of fractal development follow the respective stages of
the Koch fractal curve, as shown in fig. 22.
The Koch fractal curve is constructed as follows: A straight line of length α, is the generator
and is first divided into three equal segments. The middle segment is removed and replaced
by two segments having the same length to generate an equilateral triangle. This 4-sides line
is the first fractal stage, its direct length from one end to the other is equal the length of the
generator but the summation of the lengths of all four segments is greater and is calculated
by the general expression
n
n
4
Lα3
⎛⎞
=⎜⎟
⎝⎠
, where n is the number of iteration. It is noticed

Methods to Design Microstrip Antennas for Modern Applications
195
that to use the Koch curve in order to build the Koch island, the length of the generator
should be equal to the side of the triangle. In the next stage the process is repeated for the 4
segments generated at the first iteration, leading to the drawing of fig. 22c for the curve, and
the 22g for the island. The process in the next steps is evident.
(a) generator
(e) generator
(b) 1st stage
(f) 1st stage
(c) 2nd stage
(g) 2nd stage
(d) 3rd stage
(h) 3rd stage
Fig. 22. (a)-(d) The generator and the first three stages of the Koch fractal curve. (e)-(h) The
generator and the first three stages of the Koch fractal island and potentially printed antenna.
A basic advantage of the Koch patch, as mentioned above, is its attribute to resonate
efficiently at frequencies lower than those of a regular patch of equal size. It has been
verified via electromagnetic simulation as well as experimentally. To estimate the
superiority of the Koch fractal to a conventional triangular patch a comparison between
their sizes is necessary. At each new iteration n the area of the Koch island increases. Let An
be the area at nth iteration , then the area of the next iteration can be computed as

Microstrip Antennas
196
n1
2
n1 n
34
AA α
12 9
−
+⎛⎞
=+⎜⎟
⎝⎠ (21)
Where α is the side of the initial triangle that has an area
2
0
3
Aα
4
= (22)
The geometry series given by (21) converges to
2
2
A3α
5
= (23)
All the iterations are circumscribed inside a circumference of radius r3α/3=. On the
other hand the perimeter increases at each new iteration. The overall perimeter for iteration
k is given by
n
n
4
l3α3
⎛⎞
=⎜⎟
⎝⎠
.
This evolution process leads to a fractal object with an infinite perimeter bounding a finite
area. Despite of the increasing irregularity of the boundary, the manufacturing process does
not become more complex at each new iteration. The patch can be manufactured by
standard photo-etching techniques. The fundamental limitation in building the antenna is
given by the resolution of the photo-etching process. When the number of iterations is
increased the new added details in the structure cannot be resolved and they are not
reproduced in the manufactured element.
In fig 23 the frequency dependence of the input impedance of microstrip Koch atennas for
first five iterations are presented. All cases concern to the fundamental mode and the results
were received by numerical analysis with the Method of Moments(MoM)[58].
Fig. 23. Indicative results of the input resistance and reactance of a Koch fractal microstrip
patch for several fractal iterations
The Koch patches were generated by an equilateral triangle whose side was118.2mm. They
were printed on dielectric substrate with dielectric constant εr=4.17 and thickness h= 1.6mm.
The printed metallic patch was placed 3.4 mm from the ground plane. So, a dual layer
substrate was created. The patches were fed by a coaxial probe and the feed point was at the
same location for all the patches and was placed at 15.5mm from the center of the patch. An

Methods to Design Microstrip Antennas for Modern Applications
197
interesting conclusion can be derived from the input impedance plot. The fundamental
resonant frequency decreases when the number of iterations increases. Nevertheless this
reduction tends to limit by the increment of iterations. So, the difference between the
resonant frequencies of stage 4 and stage 5 is very small. This performance proves the great
attribute of this type of antenna to resonate at small frequencies having at the same time
small size. For example the generator, namely the ordinary equilateral triangle, of the above
structure resonates approximately at 2.2 GHz. It can be verified if we use eq. (19) and set as
εr for the dual layer substrate an equivalent value resulting from the equation
22
i
ri
i1 i1 ri
h
h
==
ε≈ ε
∑∑
. So, the ratio of 2.2GHz over the ~1.15GHz at which resonates the Koch
island of the 5th stage is about 1.9. At the same time, in accordance to equations (21) and (22)
the ratio
th
Area of the 5 stage Koch island 1.6
Area of the triangle
. If , instead of using the 5th stage Koch
island, an ordinary equilateral triangular patch was used, the length of its side had to be 1.9
times the side of the initial triangle in order to resonate at ~1.15GHz . In this case the ratio of
the surfaces of the two triangles had to be equal to 2
1.9 3.61= instead of 1.6.
It is worthwhile to note that the concept Koch fractal can be used to build monopole
antennas, enlongated or loop formed, which exhibit multi-frequency operation. In [59] a
Koch loop monopole is proposed(fig. 24). It was designed to resonate at the bands of
2.45GHz and 5.25GHz therefore capable to serve a 802.11b/g WLAN system.
Fig. 24. The geometry of a planar monopole antenna with Koch fractal loop.
An alternative design of a fractal planar object, based on the Koch concept is shown in figure
25a. In this case, the generator is a square patch. The first stage of the fractal building can be
produced by dividing each side of the initial square element into three equal segments. Then
the middle segment of each size was removed and replaced by two segments of equal length
and generating an equilateral triangle. Up to this point the process is similar to that of the
Koch curve. In the following this triangle was directed inward to the initial square patch and
was subtracted from it.
This configuration has been proposed in [60] and was successfully used to build a tri-band
microstrip antenna configuration for GPS and terrestrial radio services. The initial square
patch had side length equal to 78.25mm and was printed on a dielectric substrate having
εr=1.046 and height 6mm. The patch of the first stage fed solely by one probe had small gain
and single band operation. The target was the antenna to operate in three frequency ranges

Microstrip Antennas
198
(GPS, DCS1800 and 2.7GHz). The desired performance was obtained by the configuration
shown in Fig. 25b. It came from the combination of four patches of the first stage, lightly
overlapped. Two probes were used and the operation was enhanced by two pins. The
suitable position of the probes and the pins was found by a number of simulations whereas
the relatively large size of the entire printed scheme, due to the four fractal patch , ensured
the satisfactory high gain. For best performance, the presence of the probes and pins was
exploited. A phase difference between the feeding probes was inserted and on-off switching
of the pins, depending on frequency area, was imposed. These situations are shown in Table 2.
(a)
(b)
Fig. 25. Microstrip patch antenna for GPS and terrestrial radio services: a) A square patch
modified via the first stage inverse van Koch fractal curve b) The compact radiating
structure produced by the union of four fractal elements.
GPS
(1.575GHz)
DCS
1.78GHz –1.9 GHz 2.67GHz – 2.78GHz
Current of probe 1 1 1 1
Current of probe 2 j/2
1e
π
1 2
1
j
/
eπ
pins off switched to ground switched to ground
Table 2.
The results for the scattering coefficient of the signal at the feed input are depicted in figure
26 and show the tri-band operation.
The gain patterns of the antenna are illustrated in figures 27 and 28. At 1.575GHz(fig. 27),
the system appears a broadside radiation suitable for a GPS link when the antenna is
horizontally positioned and the total maximum gain is 6.8 dB.
The distribution of the radiated power in the DCS frequency range is shown in fig. 28. In
this case the field is linearly polarized on yz-plane having a maximum gain value 4.9dB,
approximately constant in a range o
40±about the broadside direction. In xz-plane it is also
linearly polarized and the Gθ component has maximum value(4dB) at the broadside and
minimum(-23dB) at o
36± apart from the broadside whereas the Gφ component has
maximum (4dB) at a direction 30o apart from the broadside).

Methods to Design Microstrip Antennas for Modern Applications
199
1,20 1,25 1,30 1,35 1,40 1,45 1,50 1,55 1,60 1,65 1,70
-20,0
-17,5
-15,0
-12,5
-10,0
-7,5
-5,0
-2,5
0,0
|S11| [dB]
frequency[GHz]
1,6 1,8 2,0 2,2 2,4 2,6 2,8 3,0 3,2
-25
-20
-15
-10
-5
0
|S11| [dB]
frequency[GHz]
Fig. 26. Operation of the antenna of fig. 25b a) Scattering coefficient at the input of each
probe in the range1.2GHz to1.7GHz and b) Scattering coefficient at the input of each probe
in the range1.6GHz to 3GHz
0
30
60
90
120
150
180
210
240
270
300
330
-100
-80
-60
-40
-20
0
-100
-80
-60
-40
-20
0
|Gθ,φ(θ)| [dB]
θ
Gθ
Gφ
xz-plane
f=1.575GHz
(a)
0
30
60
90
120
150
180
210
240
270
300
330
-100
-80
-60
-40
-20
0
-100
-80
-60
-40
-20
0
|Gθ,φ(θ)| [dB]
θ
Gθ
Gφ
yz-plane
f=1.575GHz
(b)
Fig. 27. Radiation patterns of the antenna of fig. 25 : the power gain components at
1.575GHz on xz- and yz-plane
0
30
60
90
120
150
180
210
240
270
300
330
-100
-80
-60
-40
-20
0
-100
-80
-60
-40
-20
0
|Gθ,φ(θ)| [dB]
θ
Gθ
Gφ
xz-plane
f=1.8GHz
0
30
60
90
120
150
180
210
240
270
300
330
-100
-80
-60
-40
-20
0
-100
-80
-60
-40
-20
0
|Gθ,φ(θ)| [dB]
θ
Gθ
Gφ
yz-plane
f=1.8GHz
Fig. 28. Radiation patterns of the antenna of fig. 25 : the power gain components at 1.8 GHz
on xz- and yz-plane

Microstrip Antennas
200
2.3.5 b) Sierpinski fractals
Another fractal concept widely used for the design microstrip antennas is the Sierpinski
fractal[61]-[69]. Various Sierpinski fractal objects have been proposed: The Sierpinski
Gasket(or Triangle), the Sierpinski Carpet ( or rectangle), the Sierpinski Pentagon and the
Sierpinski Hexagon. Judging from the literature the most efficient shapes for antenna
applications are the carpet and especially the gasket. Monopole or dipole gasket fractal
microstrip schemes have been proposed as multifrequency antennas.
Although the Sierpinski objects are based on different geometrical basis, they share the same
construction principle. The geometrical construction of the popular Sierpinski gasket begins
with an equilateral triangle which is considered as generator(fig. 29a). The next step in the
construction process is to remove the central triangle, namely the one with vertices that are
located at the midpoints of the sides of the original triangle. After the substruction, three
equal triangles remain on the structure, each one being half of the size of the original
one(fig. 29b). This process is then repeated for the three remaining triangles etc(figures 29c,
29d). If the iteration is carried out an infinite number of times the ideal fractal Sierpinski
gasket is obtained. In each stage of the fractal building each one of the three main parts of
the produced structure is exactly similar to the whole object, but scaled by a factor. Thus the
Sierpinski gasket, as well as the other Sierpinski objects, are characteristic examples of self
similar schemes.
It has to be pointed out that from an antenna engineering point of view the black triangular
areas represent a metallic conductor whereas the white triangular represent regions where
metal has been removed.
generator
1st order
2nd order
3rd order
Fig. 29. The generator and the first three stages of the Sierpinski fractal gasket.
Figure 30 shows a Sierpinski gasket monopole printed antenna. Typically such antennas
exhibit a log-periodic spacing of resonant frequencies as well as an increase in the
impedance bandwidth at higher bands. It is interesting to note that the band number n and
the iteration k are interchangeable. For example the band zero and the 0th iteration
correspond to the fundamental resonance of the antenna. The first band and the first fractal
iteration correspond to the first log-periodic resonant frequency. Therefore after the first
fractal iteration two resonant frequencies are available : the fundamental and the first log
periodic frequency. This is valid for other higher fractal iterations.
The specific positions of the frequency bands depend on the geometry of the generator and
the parameter values of the dielectric substrate. It has to be noticed that the generator would
potentially be not an equilateral triangle, namely the angle(flare) that corresponds to the
vertex at which the feeding is applied would be not equal to 60o. Many such configurations
have been proposed. The potential to select another value for this angle is an advantage
because there are two geometrical parameters to control the frequencies of resonance. The
height of the triangle and the flare angle. Indicative configurations are shown in fig.31a, and
the respective input impedance diagrams are depicted in fig. 31b [64].

Methods to Design Microstrip Antennas for Modern Applications
201
(a)
(b)
Fig. 30. The Sierpinski monopole(a) and dipole(b) printed antenna
From the design point of view, mathematical expressions for the calculation of the frequencies
of resonance of the Sierpinki gasket, are necessary. The most recent available formula[65] in
terms of the structural parameters and the order of iteration, for a monopole(fig. 30a) with
flare angle equal to 60o is that of eq. (23). This expression includes the parameters of the
geometry of the gasket as well as the thickness and the dielectric constant of the substrate.
()
ρξ
δ
n
1
e
n
e
c
(0.15345 0.34 x) for n 0
h
rc
0.26 for n 0
h
f
−
+
=
>
≅ (23)
Where n
n1
h
ξh+
= is the ratio of the height of the gasket in the nth iteration to that in the
(n+1)th iteration, δ=1/ξ is the scale factor ρξ0.230735
=
− and 0, n 0
1, n 0
x
=
>
=.
Moreover
e
e
s3
h2
= (24a)
0.5
er
sst(ε)
−
=+ (24b)
where s is the length of the side of the gasket and t and r
ε are the thickness and the
dielectric constant of the substrate. The above equation is valid even in cases where the
geometry is perturbed to get different scale factors
In practice the given parameter value is the frequency of resonance and the values of t and εr
are selected by the designer. So, for a specific value of n, the required parameters are those
of the geometry of the gasket. For these calculations the side length of the generating
triangle of the gasket is given by the expression
()
ρξ ε
δε
n
1
rr
n
rr
1ct
(0.3069 0.68 x ) for n 0
f
3
0.52 c t for n 0
f
3
s
−
+
−=
−>
≅ (25)

Microstrip Antennas
202
(a)
(b)
Fig. 31. a)Sierpinski gasket antennas with different flare angle b) Indicative results of their
performance: Real and imaginary part of the input impedance for specific geometrical and
material parameter values
It is worthwhile to mention that by additional modification of the Sierpinski gasket as
proposed in[62] or in [69] (see fig. 32), the bands of resonance could be further controlled in
order to meet the technical requirements of the applications for which the antenna is
designed.
The Sierpinski carpet is another Sierpinski fractal configuration reported in antenna
applications. Sierpinski carpet dipole antennas are shown in figures 33 and 34. The study of
these configurations guide to the conclusion that no multiband performance can be

Methods to Design Microstrip Antennas for Modern Applications
203
obtained. It is due to the fact that the fractal iterations do not perturb the active current
carrying region . So, their performance is similar to that of a simple square patch.
(a)
(b)
Fig. 32. Printed multiband antenna monopoles a)The self similar ordinary Sierpinski gasket.
b) Modified Sierpinski gaskets.
Fig. 33. Sierpinski carpet dipole antennas: the generator and the first three fractal orders
The negative version of the above fractal scheme is shown in Fig. 34. The geometrical
method to design this carpet is the following: The generator is that of fig. 34a. In the first
iteration the area around the central patch is divided in nine sub-areas of equal size and at
the center of each sub-area a rectangular patch with side length three times smaller than the
initial central patch is located. The same process is applied in the next iteration. In this
antenna only the central element is driven and the energy of the other smaller patches is
coupled parasitically from the driven patch.
Fig. 34. Sierpinski carpet fractal antennas: the generator and the first two orders.
This fractal microstip configuration exhibits multifrequency performance, Fig. 35, but it was
found[63] that the results come from the driven element, not from the parasitic ones.
2.3.5 c)Hilbert fractals
The properties of the Hilbert curve make them attractive candidates for use in the design of
fractal antennas. These curves apart from being self similar have the additional property of
approximately filling a plane and this attribute is exploited in realizing a ‘small’ resonant
antenna. Hilbert fractal antennas with size smaller than λ/10 are capable to resonate, with
performance comparable to that of a dipole whose resonant length is close to λ/2.

Microstrip Antennas
204
Fig. 35. The reflection coefficients of Sierpinski carpet microstrip antenna(fig. 34) in different
iterations
The generator of the Hilbert curve has the form of a rectangular U as shown in fig. 36a. The
Hilbert curves for the first several iterations are shown in figures 36b-36d. The construction
at a stage is obtained by putting together four copies of the previous iteration connected by
additional line segments.
The
generator(order=0)
1st order
2nd order
3rd order
Fig. 36. The Hilbert fractal printed antennas of various stages.
It would be interesting to identify the fractal properties of this geometry. The space-filling
nature is evident by comparing the first few iterations shown in figure 36. It may however
be mentioned that this geometry is not strictly self similar since additional connection
segments are required when an extra iteration order is added to an existing one. But the
contribution of this additional length is small compared to the overall length of the
geometry, especially when the order of the iteration is large. Hence, this small length can be
disregarded which makes the geometry self similar. Moreover the curve is almost filling a
plane. In other words the total length, if sum the line segments, tends to be extremely large.
This could lead to a significant advantage, since the resonant frequency can be reduced
considerably for a given area by increasing the fractal iteration order. Thus, this approach
strives to overcome one of the fundamental limitations of antenna engineering with regard
to small antennas.

Methods to Design Microstrip Antennas for Modern Applications
205
For an accurate study of the operational features of a Hilbert fractal printed antenna
information about its geometric parameters are necessary. It is obvious that as the iteration
order increases, the total length of the line segments is increased in almost geometric
progression if the outer dimension is kept fixed. Thus, for a Hilbert curve antenna with side
dimension L and order n, the total segment length S(n) is calculated by the formula
2n
n
21
S(n ) L
21
−
=− (26)
and the length of each line segment is given by
n
L
d21
=
−
(27)
A theoretical approach for the calculation of the resonant frequencies of the antenna
considers the turns of the Hilbert curve as short circuited parallel-two-wire lines and begins
with the calculation of the inductance of these lines[70], [71]. This approach is illustrated in
figure 37. The self inductance of a straight line connecting all these turns is then added to the
above, inductance multiplied by the number of shorted lines, to get the total inductance. To
find the frequencies of resonance, the total inductance is compared with the inductance of a
regular half wavelength dipole.
Fig. 37. The 2nd order of fractal building. The segments used to connect the geometry of the
previous iteration are shown in dashed lines
In detail for a Hilbert curve fractal antenna with outer dimension of L and order of fractal
iteration n, there are n1
m4
−
=
short circuited parallel wire connections each of length d.
Moreover the segments not forming the parallel wire sections amount to a total length of
() ()
2n 1 2n 1
n
L
sd21 21
21
−−
=
−= −
− (28)
The characteristic impedance of a parallel wire transmission line consisting of wires with
diameter b , spacing d , are given by
o
2d
Z120log
b
⎛⎞
=⎜⎟
⎝⎠
(29)

Microstrip Antennas
206
The above expression can be used to calculate the input impedance at the end of the each
line section , which is purely inductive
()
oo
in in n
ZZ
L
Z L tan d tan 21
⎛⎞
== β= β
⎜⎟
ωω
−
⎝⎠
(30)
It is noticed that at the nth stage of fractal building there are n1
m4
−
=
such sections.
The self inductance due to a straight line of length s is
o
s
8s
Lslog1
b
μ⎛⎞
=
−
⎜⎟
π⎝⎠
(31)
So, the total inductance is
oo
Ts in n
Z
8s L
LLmL slog 1mtan
b21
μ⎛⎞ ⎛⎞
=+ = −+ β
⎜⎟ ⎜⎟
πω
−
⎝⎠ ⎝⎠
(32)
To find the resonant frequency of the antenna, this total inductance is equated with that of a
resonant half-wave dipole antenna with approximate length equal to λ/2. Taking into
account that regular dipole antennas also resonate when the arm length is a multiple of
quarter wavelength we can obtain the resonant frequencies of the multi-band Hilbert curve
fractal antenna by the expression
()
o o
8k 120 2d 8s
klog 1m logtand slog 1
4b4 b b
μμ
λ
⎡λ⎤ ⎡ ⎤
⎛⎞ ⎛⎞
−
=
β
+−
⎜⎟ ⎜⎟
⎢
⎥⎢⎥
πωπ
⎝⎠ ⎝⎠
⎣
⎦⎣⎦
(33)
where k is an odd integer. It is noticed that this expression does not account for higher order
effects and hence may not be accurate at higher resonant modes.
At these antennas the feeding point is located at a place of symmetry or at one end of the
curve, thus driving the structure to operate as a monopole antenna. It is noticed that the
bandwidth at resonances is generally small, whereas the positions of resonant frequencies
can be controlled by perturbing the fractal geometry.
In the basis of the above theory, several applications of this type of fractal antenna have
been reported. Antennas that can efficiently operate in the range of UHF, as well as in
multiple bands, at 2.43GHz and 5.35GHz, serving Wireless Local Area Networks [71]-[73].
2.3.5 d)Square Curve fractals
The design of microstrip antennas by the square curve fractal algorithm can yield radiating
structures with multiband operation. The generator of this type of fractal objects is a
rectangular ring and as a consequence the curves of the various stages are closed curves.
The square curve fractals do not belong to the category of the space filling curves. However
the increment of their total length from stage to stage is not significant, thus permitting the
antennas to meet the requirement of the small size and at the same time to exhibit an
increasing gain in virtue of their increasing length.
The staring point of the construction process is the selection of the size of the generator
which is a rectangular ring with side length L(Fig. 38a). At the next step of the recursive
process, the four corners of the square ring are used as the center of four smaller squares

Methods to Design Microstrip Antennas for Modern Applications
207
each having side, half that of the main square. Overlapping areas are eliminated. The curve
produced by this first iteration is shown in Fig.38b. Following the same algorithm the
second stage of the fractal antenna can be derived(Fig. 38c). The building of the higher
stages is evident.
The total length of the curve is calculated as follow.
a. generator ring: the perimeter is 4L
⋅
b. 1st stage : each side of the generator is divided in four segments of equal length. Two
segments are removed, at each corner and they are replace by smaller squares with side
length equal to L/2. So, the length of the curve is equal to the sum of segments,
common between generator square and the first recursion, plus the length of the
newly added segments. The total length of the two segments removed at each corner, is
L2, so the total removed is (L 2) 4
⋅
. Looking only at the added segments the length
increase of the curve is inc
st1
L(L2)8
=
⋅.
c. 2nd stage: On the second iteration, the corners of the four small squares added at the
first iteration are replaced by four even smaller squares with side length L/4. Here the
length of the segments removed at each square corner is equal to L4 and the length of
the smaller squares added is equal to (L 4) 3
⋅
. Taking into account only the added
segments the length increase is inc
st1
L(L4)24
=
⋅.
The general formula for the length increase is inc n ( n 1)
L(L/2)83
−
=⋅⋅ where n is the iteration
number .
The total curve length is n
1
L((r 8) 4)
⋅
⋅− , where r is the ratio of the length increase between
two sequential iterations.
(a)
(b)
(c)
Fig. 38. The Square Curve fractal a) the generator and (b)-(c) the lower two stages
In[75] a microstrip fractal structure designed with the aforementioned algorithm is
proposed. It was printed on a dielectric substrate with r1.046
ε
=( Rohacell 51HF, Northern
Fiber Glass Service, Inc.) and height h1mm
=
. A value for r equal to 1.5 and the second
stage of development gave an object with outer dimensions 8.4cm x 8.4cm. A fundamental
parameter of the structure is the width of the printed strip which forms the curve. Attention
must be paid to the proper selection of the values of the strip’s width because there is a
trade off between this value and the input impedance of the antenna. A narrow strip
guides to high input impedance and inserts difficulties to the matching of the antenna. On

Microstrip Antennas
208
the other side a wider strip would yield input impedance suitable for direct matching but
could produce difficulties related with the space filling during the process of the fractal
expanding. More over, the keys to drive this antenna in multi-band operation are the proper
number and positions of the feeding points. The incorporation of a pin can also enhance the
performance of the antenna.
In figure 39, results received using three different feedings are depicted. Figure 39a shows
the variation of the scattering coefficient at the feeding input using one probe, positioned at
a point on an axis of symmetry. This choice is common at many fractal antennas. It is
observed that only two frequency bands give scattering coefficient lower than -10dB. It is
due the high input impedance of the antenna, as shown in figure 39b. A better performance
with seven frequency bands is obtained with two probes(Fig. 39b); and an even satisfactory
operation is achieved when a shorting pin is installed between the probes. The pin
(a)
4,0 4,5 5,0 5,5 6,0 6,5 7,0 7,5 8,0 8,5 9,0
0
100
200
300
400
500
600
|Zin|[Ohms]
frequency [GHz]
(b)
4,04,55,05,56,06,57,07,58,08,59,0
-25
-20
-15
-10
-5
0
|Sij|[dB]
frequency [GHz]
(c)
4,0 4,5 5,0 5,5 6,0 6,5 7,0 7, 5 8,0 8,5 9,0
-40
-35
-30
-25
-20
-15
-10
-5
0
|Sii|[dB]
frequency [GHz]
(d)
Fig. 39. a) Scattering coefficient at the input of the second stage fractal antenna fed with one
probe and b) the respective input impedance. c) Scattering coefficient when fed with two
probes and d)fed with two probes and loaded with one pin.

Methods to Design Microstrip Antennas for Modern Applications
209
acts as a short circuit between the trace and the ground plane, reflecting the wave produced
by the probe. So, two complex impedances combined in parallel appear at the point of the
feed. The one is due to the line between the probe and the pin and the other to the
remaining trace of the structure. These two parallel impedances involve a lower total
impedance which would be suitable for direct match to an 75Ohm probe. Moreover, this
matching is attainable in wide ranges around the frequencies of resonance. The results of fig
39c, for the scattering coefficient, show a multiband - and at the same time wideband
operation.
3. Electromagnetic Bandgap Structures (EBG) in antenna applications
3.1 The EBG structure and properties
Electromagnetic Band Gap (EBG) structures constitute a specific class of recently discovered
microwave objects that, due to their special electromagnetic behavior, reveal promising
solutions to several microwave problems, especially in the area of communications[76],[77].
The EBG structures are generally defined as ‘artificial periodic or non periodic objects that
prevent the propagation of electromagnetic waves in a specified band of frequency for all
incident angles and all polarization states’. They can be categorized into three groups
according to their geometric configuration: Three dimensional volumetric structures, two-
dimensional planar surfaces and one dimensional transmission lines. Among these three
categories the planar EBG objects are the most commonly used in antenna systems. They
consist of a two dimensional lattice of metal plates conductively connected to a ground
plane by metal-plated vias, as shown in figure 40, and are easily fabricated using printed –
circuit board technology.
The effective application of EBG surfaces to the antenna design is based on the exploitation
of their distinctive electromagnetic properties with respect to the type of the incident
electromagnetic waves:
i. When a plane wave impinges on an EBG surface it is reflected with a phase that varies
with frequency as shown in figure 41. At a certain frequency the reflection phase is zero
degrees. The value of this frequency depends on the structural parameters of the EBG
object. This performance resembles a perfect magnetic conductor that does not exist in
the nature.
ii. When the incident wave is a surface wave the EBG structures show a frequency band
gap through which the surface wave cannot propagate for any incident angles and
polarization states.
Both the above attributes contribute to the enhancement of the performance of printed or
not printed antenna elements or arrays of elements. For example in the case of a microstrip
antenna integrated with EBG structures, the suppression of the surface waves could reduce
the mutual coupling between the antenna elements, if the antenna is an array, and also
prevent the wave to reach the boundaries of the configuration and be diffracted. This
prevention involves lower radiation towards the back space of the antenna, an attribute that
would ensure low interference with adjacent microwave elements or low radiation towards
the user of the equipment that hosts the antenna. Moreover, due to the property of an EBG
cell to work as a resonator, enhancement of the antenna gain could be obtained and also an
easier and effective matching of the system to the feeding probe. Furthermore an EBG
object could drive the microstrip antenna to a dual frequency operation modifying the
higher order radiation patterns, thus making them similar to those of the basic mode. On the

Microstrip Antennas
210
other hand an EBG surface is a unique object to obtain low profile antennas if the radiating
element is not a microstrip antenna but a wire dipole , thus providing simple and effective
radiation systems.
Fig. 40. Geometry of the mushroom-like EBG structure
Fig. 41. The phase of a plane wave reflected by the EBG surface
A theoretical analysis of an EBG structure, via various models, gets an insight into the way
by which they perform and can yield mathematical expressions for their operational
parameters. The analysis would explain the mechanisms by which the EBG objects enhance
the performance of the antennas and would give to the antenna designer the ability to
properly exploit all the EBG properties, potentially useful in a specific antenna application.
3.2 Theoretical analysis
3.2.1 Low and high impedance surfaces
Flat metal sheets have low surface impedance and are used in many antennas as a reflector
or a ground plane. The boundary conditions on these surfaces impose that the tangential
component of the electric field intensity has to be equal to zero and this requirement
involves that the metal sheet reflects an impinging wave, shifting the phase of its intensity
by an amount of π. Moreover the metal sheet redirects one-half of the radiation into the
opposite direction improving the antenna gain by 3dB and partially shielding objects on the
other side. However, if the antenna is too close to the conductive surface, the out of phase

Methods to Design Microstrip Antennas for Modern Applications
211
image currents ‘cancel’ the currents in the antenna, resulting in poor radiation efficiency.
This problem is often addressed by positioning the radiating element at a quarter-
wavelength distance from the ground plane but this arrangement requires minimum
thickness of λ/4.
By incorporating a special texture on a conducting surface it is possible to alter its radio-
frequency electromagnetic properties. A proper modification (see for example fig. 40)would
yield a specific surface with high surface impedance. On these textured surfaces the
tangential component of the magnetic field intensity tends to zero and this condition means
that the surface reflects an incident wave with an almost zero phase shift. This minimization
of the magnetic field is due to the minimization of the surface currents, that inevitably
comes from the cutting up of the metallic surface, into small patches. It is noticed that
although the magnetic field intensity is very small, the electric field may have a large value
due to the high voltage induced between the edges of the adjacent patches of the modified
sheet. The edges and the narrow gaps between them realize capacitors. The ratio of the high
electric field intensity, which is high, over the low magnetic field intensity defines the
impedance of the surface which, in this case, is obviously very high. The almost zero
tangential magnetic field permit us to term the surface as an artificial magnetic conductor.
This unusual boundary condition involves that the image currents are in phase rather than
out of phase, allowing radiating elements to lie in very close proximity to the surface while
still radiating efficiently.
As a consequence of the performance described above, this type of high impedance surface
can function as a new type of ground plane for low profile antennas. For example a dipole
positioned in parallel to a high impedance ground plane is not shorted out as it would be on
an ordinary metal sheet. In addition to their unusual reflection-phase properties, these
textured structures have a surface wave bandgap, within which they do not support bound
surface waves of either transverse magnetic (TM) or transverse electric(TE) polarization.
They may be considered as a kind of electromagnetic bandgap structures or photonic crystals
for surface waves. It is noticed that although bound surface waves are not supported, leaky
TE waves can propagate within the bandgap, and they are useful for certain applications.
The theoretical basis to explain the electromagnetic behavior of an EBG structure is the
general theory of surface waves and the useful tool for the description of its performance is
the surface impedance of the EBG object along with equivalent electric circuit of this
impedance.
3.2.2 Surface waves
Surface waves can occur on the interface between two dissimilar materials, for example
metal and free space[76]-[78]. They are bound to the interface and decay exponentially into
the surrounding materials although at radio frequencies the fields associated with these
waves can extend thousands of wavelengths into the surrounding space. To describe
theoretically and by a simple manner their physical entity, let us suppose an interface
parallel to yz plane as shown in figure 42.
Assume a wave that is bound to the surface in the +x direction with decay constant α , and
in –x direction with a decay constant γ. Τhe wave propagates in the z-direction with
propagation constant k. For a TM polarized surface wave y
E0
=
. The electric field in the
upper half-space has the following form

Microstrip Antennas
212
Fig. 42. A surface wave is bounded around the interface of two media and decays
exponentially into the surrounding space
()
jωtjkzαx
11x1z
ˆˆ
ExEzEe
−
−
=+
G
(34)
In the lower half-space the electric field is
()
jωtjkzγx
22x2z
ˆˆ
ExEzEe
−
−
=+
G
(35)
The solution of Maxwell equations yields the following results for the propagation constants
ωε
kc1ε
=
+
(36a)
ω1
αc1ε
−
=
+
(36b)
2
ωε
γc1ε
−
=
+
(36c)
If ε is real and positive, then α and γ are imaginary and the waves do not decay with
distance from the surface and propagate through the dielectric interface. Thus TM surface
waves do not exist on nonconductive dielectric materials. On the other hand if ε is less than -
1, or if it is imaginary or complex, the solution describes a wave that is bound to the surface.
These TM surface waves can occur on metals or other materials with non-positive dielectric
constants.
The complex permittivity of a metal can be expressed in the following form
r
r
1j
οο
⎛⎞
σ
ε=ε ε −
⎜⎟
ω
εε
⎝⎠
 (37)
where σ is the conductivity of the metal, ε0 is the permittivity of the free space and εris the
dielectric constant.
For relatively low frequencies, including the microwave spectrum, the conductivity is
primarily real and much greater than unity, thus, the permittivity is a large imaginary
number which, supposing that εr1
=
, is calculated by σ
εωε0
j≈
 and in accordance to (36a)

Methods to Design Microstrip Antennas for Modern Applications
213
the dispersion relation for surface waves at radio frequencies is approximately ω
kc
≈.
Thus, surface waves propagate at nearly the speed of light in the vacuum and they travel for
many wavelengths along the metal surface with little attenuation.
The decay constant of the fields into the surrounding space is derived by inserting (37) into
(36b). It is easy to ascertain, by an arithmetic application at microwave frequencies, that the
surface waves extend a great distance into the surrounding space.
Constant γ ,the inverse of which is related to the wave penetration depth into the metal, is
approximately
()
(
)
01j
ωμ σ
γ1j 2δ
+
≈+ = (38)
where δ is the skin depth. The surface currents penetrate only a very small distance into
the metal. They can be expressed in terms of the skin depth, the conductivity and the electric
field intensity and can be used for the determination of the magnetic field intensity. Then,
the surface impedance is derived by the electric and magnetic field intensities as follows
z
s
y
E1j
ZHσδ
+
== (39)
The above results show that the surface impedance has equal positive real and positive
imaginary parts, so the small surface resistance of the metal surface is accompanied by an
equal amount of surface inductance.
By texturing the metal surface we can alter its surface impedance and thereby change its
surface-wave properties. Thus the surface impedance would appear inductive or capacitive
imaginary part, depending on the frequency. The derivation of the impedance is made
considering that a wave decays exponentially away from the boundary with a decay
constant α , whereas the boundary is taken into account by its surface impedance. In this
case it has been proved that TM waves occur on an inductive surface, in which the surface
impedance is given by the following expression
TM
S
jα
Ζωε
= (40)
Correspondingly, TE waves can be excited on a capacitive surface. In this case the surface
impedance is
TE
S
jωμ
Ζα
−
= (41)
The wave vector k, in terms of the frequency and the decay constant α, helps to get an
insight into the behavior of the surface
222
00
kμεω α
=
+ (42)

Microstrip Antennas
214
Combining equation (42) with the equations (40) and (41) we can derive the dispersion
relation for TM waves
()
2
TM
S
TM
2
Ζ
ω
k1
cη
=− (43)
And for TE waves
()
2
TE
2
TE
S
ωη
k1
cΖ
=− (44)
where 0
0
μ
ηε
=is the impedance of the free space.
The complete dispersion diagrams of k versus frequency (Fig. 43), show that for a specific
textured surface there is a certain frequency, termed resonance frequency, below of which
TM surface waves are supported. At frequencies much lower than the resonance frequency
the dispersion curve is very near to the light line, revealing the propagation of the wave,
thus the field extends many wavelengths along the surface, as does on a metal sheet and the
surface is considered of low impedance. Near the resonance the surface waves are tightly
bound to the surface and, as the slope of the curve is very small, the group velocity is low.
All these mean the surface impedance of the structure becomes high. Above the resonance
frequency, the surface appears a capacitive behavior and TE waves are supported. Near the
resonance frequency the TE waves are weakly bound to the surface extending far into the
surrounding space. As the frequency increases the curve bends away from the light line and
it means that the waves are more tightly bound to the surface.
Fig. 43. The predicted surface wave dispersion diagram
3.2.3 Equivalent circuit model
They are exactly the above features of the propagating waves that permit us to approach the
electromagnetic behavior of the EBG surface using an equivalent parallel resonant circuit
which can be tuned to exhibit high impedance over a pre-specified frequency band. From
the physical side of view this equivalence can be explained as follows: as the EBG interacts

Methods to Design Microstrip Antennas for Modern Applications
215
with electromagnetic waves, currents are induced in the top metal plates. A voltage applied
parallel to the top surface causes charges to concentrate around and on the ends of the plates
which can be considered as a capacitance. As the charges move back and forth, they flow
around a long path through the vias and the bottom plate(Fig. 44). Associated with these
currents is a magnetic field and, thus, an inductance.
Fig. 44.
By this circuit model, the surface impedance, assigned to the sheet, corresponds to the
impedance of a parallel resonant circuit, consisting of the sheet capacitance and the sheet
inductance
2
jωL
Ζ
1ωLC
=− (45)
The surface is inductive at low frequencies, capacitive at high frequencies and the
impedance is very high near the resonance frequency 0
ω,
0
1
ωLC
= (46)
This high impedance is associated with a forbidden frequency band. In the 2D geometry of
the lattice of the EBG patches the capacitors are formed by the fringing electric fields
between adjacent metal patches and the inductance is fixed by the thickness of the structure.
An approximate expression for the capacity [79], in the case of the geometry shown in fig.
40, is equation (47), [79]
(
)
12 1
fringe
wεε α
Ccosh
πg
−
+⎛⎞
≈⎜⎟
⎝⎠
(47)
In the above expression g is the gap between the plates, w is the width of the plates, α is the
lattice constant namely αgw
=
+, and 1
ε and 2
ε are , respectively, the dielectric constants
of the substrate of the EBG and the material surrounding the surface which may be free
space. Equation (47), although approximate, is adequate for first order designs. The
inductance of a high –impedance surface is determined entirely by its thickness. This can be
understood by considering a solenoid of current that includes two rows of plates and their
associated vias. Current flows up one row of vias across the capacitors and down the next
set of vias to return through the ground plane. The length and width of the solenoid are
canceled to obtain the sheet inductance, L
Lμh
=
(48)

Microstrip Antennas
216
For the effective surface impedance approximation to be valid the lattice constant should be
small compared to the wavelength.
Another, also approximate, equation for C is proposed in [80]. It was produced considering
normally-incident plane waves and the vias conductors, connecting patch to the ground,
not excited. In this case also square patches, small compared with the wavelength, and gaps
between the patches, very narrow compared with the length side of the patches were
considered. In the used model the higher order Floquet modes generated by the periodical
mesh were neglected. Assuming only the fundamental mode plane waves between the
lattice of patches and the ground, the equivalent surface impedance was considered as the
parallel connection of two impedances: the impedance of the lattice and the input
impedance of a TEM line section of length h. The expression of the surface impedance,
normalized to the free space impedance was found
()
r0 0
r
sn
r00
r0 0
r
1
jtan(ωhεεμ )
ε
Z
(ε1) ωα ε μ 2α
1 log tan(ωhεεμ )
πg
πε
=+⎛⎞
−⎜⎟
⎝⎠
(49)
Where α and g were defined previously. The equation for the capacitance C is
(
)
0r
Dεε 12α
Clog()
ππg
+
= (50)
3.2.4 Reflection phase and Bandwidth
Besides their unusual surface wave features, the high impedance surfaces also have unusual
reflection phase properties. Using the effective surface impedance model, described
previously, we can determine the reflection phase for the resonant textured surface. For a
normally incident wave, the reflection phase of the surface is given as
s
s
Zη
ΦIm ln Zη
⎡
⎤
⎛⎞
−
=
⎢
⎥
⎜⎟
−
⎢
⎥
⎝⎠
⎣
⎦ (51)
where s
Z is the surface impedance and η120π
=
is the free space impedance. At very low
frequencies the reflection phase is π (see fig. 41) and the structure behaves like a smooth
metal surface. At higher frequencies the reflection phase slopes downward and crosses
through zero at the resonance frequency. This situation corresponds to the artificial
magnetic conductor behavior. Above the resonance frequency the phase converges to –π.
The phase varies between π/2 and –π/2 when the magnitude of the surface impedance
exceeds the impedance of the free space. It is worth noting that for a wide range of
geometries the bounds of the surface’ s wave band gap, occur at the same frequencies where
the reflection phase crosses through π/2 and –π/2.
To determine the bandwidth of an antenna positioned in parallel to an EBG surface , its
radiation is modeled as a resistor, valued the impedance of free space, in parallel to the LC
equivalent circuit of the EBG lattice. The amount of power dissipated in the resistor is a

Methods to Design Microstrip Antennas for Modern Applications
217
measure of the radiation efficiency of the antenna. The maximum radiation efficiency occurs
at the LC resonance frequency of the EBG surface, where the surface reactance is infinite. At
very low or at very high frequencies, the radiated power is reduced. It can be shown that the
frequencies where the radiation become equal to half of its maximum value, occur when the
magnitude of the surface impedance is equal to the impedance of free space. It is easily
proved that the frequencies at the bounds of the operating band are approximately given
by the expressions
0
1L
ω1ΖC
⎛⎞
≈±
⎜⎟
⎜⎟
⎝⎠
(52)
The two frequencies designated by the plus and minus signs delimit the frequency range
over which an antenna would radiate efficiently on such a surface. The relative bandwidth,
defined as the ratio of Δω over the frequency of resonance, is approximately equal to
0
00
Lε
Δω
BW ωCμ
=≈ (53)
It is the bandwidth, over which the phase of the reflection coefficient falls between π/2 and -
π/2 and image currents are more in phase than out of phase. As noted in the previous
section, this range often coincides with the surface wave band-gap.
The geometry of the EBG and the parameters of the materials can be adjusted to provide the
desired resonance frequency and the bandwidth and it is usually done via simulations.
However it is useful to have an initial solution to more rapidly converge on the correct
design. So, to design an EBG surface for pre-specified frequency of resonance ω0 and
bandwidth BW, in accordance to the aforementioned analysis the thickness of the substrate
is determined by the equation
0
cBW
hω
= (54)
and the equation for the required sheet capacitance is
0
1
CωηBW
= (55)
It is worth noting the there is a relation between the bandwidth and the variation of the
phase reflection. The rate by which the reflection phase varies, conditions the frequency
bandwidth of the mushroom type EBG and depends on the length of the side of the patch,
the width of the gap between the patches, the dielectric constant and the thickness of the
substrate[81].
Diagrams illustrating this performance are depicted in figure 45. In fig. 45a is shown that the
increment of the side length reduces the frequency of resonance, as expected, but also
reduces the bandwidth. From fig. 45b it is concluded that when the gap width is increased,
both the frequency band position and its bandwidth increase. Fig. 45c shows that when the
substrate permittivity εr is increased the frequency band position and the bandwidth

Microstrip Antennas
218
decrease. The results of fig. 45d prove that when the substrate thickness is increased the
frequency band position decreases while its bandwidth increases.
(a)
(b)
(c)
(d)
Fig. 45. EBG parameters analysis. Reflection phase variation versus frequency. Effects of a)
patch width b)gap between the patches of the EBG lattice c) substrate permittivity
d)substrate thickness.
3.3 EBG structures in antenna applications
Although the mushroom type cell is the most commonly used, various EBG types of cells
(fig. 46) have been proposed for the enhancement of the antennas they host.
Whatever the type of the used EBG is, the determination of its proper configuration for the
improvement of the antenna performance is not an easy task. The band-gap of an EBG,
designed in advance, cannot guarantee the effective operation of the antenna that will be
incorporated in the following, because complicated interactions occur between the antenna
and the EBG lattice. Due to these interactions the improvement of the operation of the
system antenna-EBG, as well as the satisfactory feed matching of the antenna, are problems
that must be simultaneously solved. An efficient strategy for this purpose is the parametric
study of the entire configuration, which is supported by simulations via a high frequency

Methods to Design Microstrip Antennas for Modern Applications
219
electromagnetic software, before the manufacturing the antenna system. Alternative ways to
solve the problem is the employment of deterministic or stochastic algorithms, as will be
discussed in the next paragraph.
Honeycomb
Winding Cross
Rectangular patches conductively connected with
meander strip lines
Spiral slotted rectangular patches
Fig. 46. Various printed schemes suitable for EBG implementation
Many articles have reported the contribution of the EBG surfaces to the enhancement of the
operation of antennas incorporated in their structure. Some typical applications are referred
in the following.
In [82] a structure consisted of a ‘diamond dipole’ of two square patches over a mushroom
type EBG was studied(Fig. 42a). The suitable parameter values of the scheme yield the
results shown in fig. 47b. The presence of the EBG drives the dipole to operate effectively
within a bandwidth much wider than that of the ordinary metallic ground sheet.
The configurations shown in figures 48 and 49 concern in antenna array arrangements, in
which reduction of mutual coupling between their elements was obtained, via EBG
schemes.
Fig. 48b depicts the scattering parameter S12 between the printed elements in the
arrangement shown in fig. 48a. The significant reduction of S12 in the band of EBG
resonance, proves the effectiveness of the EBG lattice, embedded between them, in reducing
their mutual coupling[83].
In [84] an antenna array of microstrip elements with high directivity was designed. A large
directivity can be obtained by an array with large size and it means that the distance
between the elements has to be large. However the large distance would involve
undesirably high side lobes. So, to obtain high directivity the designers chose to keep the
inter-element distance small and to increase the size of the patch. Small inter-element
distance causes strong mutual coupling between the elements of the array. This problem
was confronted inserting an EBG configuration between adjacent elements as shown in

Microstrip Antennas
220
(a)
(b)
Fig. 47. a)Top and side view of a diamond planar dipole antenna over a mushroom type
EBG b) the broadband performance of the structure: scattering coefficient of the signal at the
feeding probe.
(a)
(b)
Fig. 48. a)Microstrip antenna separated by the mushroom-like EBG b) Mutual coupling
between the patches with and without EBG
Fig.49a. The increasing of the patch’s size was achieved using a dual layer dielectric substrate.
A large permittivity was selected for the lower layer while the upper layer was the air. The
additional advantage of this configuration is that a substrate with large thickness and low
effective permittivity was created and as a result the bandwidth was enhanced. Moreover the
patches of the EBG printed on the top of the lower layer with high permittivity had size
enough small, to be placed between the antenna radiating patches. In the proposed array, by
proper selection of the structural parameter values, a reduction of mutual coupling about 10dB
or 15dB, depending on the frequency bandwidth, was obtained.
Another representative example of mutual coupling reduction using EBG in a microstrip
antenna array, was recently proposed in [85]. In this configuration (fig. 49b) the thickness of
the substrate was large and this fact permitted to embed inside it an elongated mushroom
EBG lattice(fig. 49c). By this novel configuration, isolation exceeding 12dB, was obtained
between the antenna array’s elements.

Methods to Design Microstrip Antennas for Modern Applications
221
(b)
(a)
(c)
Fig. 49. Printed antennas integrated with EBG Structurs for mutual coupling reduction:
a) Array of two patches structured in a dual-layer substrate. The EBG embedded between
the patches reduces the mutual coupling b) a novel type of compact elongated mushroom
EBG (EM-EBG)structure c) Linear array of four microstrip antenna elements with isolating
(EM-EBG) objects
EBG structures are also effective to the design of dual frequency microstrip antennas. It has
been proved, in [86], that a properly designed EBG lattice can drive a printed antenna to
dual frequency operation, exploiting the higher order modes of the printed element.
Generally speaking, a microstrip antenna would be used as dual frequency band radiating
system when operates at the fundamental and at one of the higher order resonance
frequencies. However, the distributions of the radiated fields, at these two frequencies, are
not similar. So, the two bands can not be used for the same type of services. In [86], this
problem was confronted integrating the antenna with an EBG surface and dual frequency
operation with similar radiation patterns was obtained. The key of this performance is that
the antenna works at the two bands with different dielectric layers.
As an application example in[86], results for a ‘diamond’ antenna consisted of a pair
triangular patches were presented(fig. 50). The EBG structure was formed of equilateral
triangular microstrip patches.
In figure 51a the variation of the input impedance of the twin triangles without the EBG is
illustrated. The triangle patches of the diamond were designed for the TM10 mode at
3.17GHz. At the figure is shown that the second order mode, namely the TM11 ,appears at
5.6GHz. In figures 51b and 51c the radiation patterns of both modes are depicted. Obviously
the patterns are different, whereas the disadvantage of the pattern of TM11 , is the deep
minimum at the broadside.

Microstrip Antennas
222
(a)
(b)
Fig. 50. Top and side view of a triangular probe fed dual patch antenna over an EBG lattice
of triangular printed elements.
234567
-40
0
40
80
120
Input impedance[Ohms]
frequency [GHz]
real
imaginary
(a)
030
60
90
120
150
180
210
240
270
300
330
-20
-10
0
10
-20
-10
0
10 3.17GHz
-xz-
-yz-
(b)
-30
-15
0
030
60
90
120
150
180
210
240
270
300
330
-30
-15
0
5.6GHz
-xz-
-yz-
(c)
Fig. 51. (a) Input impedance of the patch antenna of fig 50, without EBG. (b), (c) the xz-and
yz-plane directivity patterns of the TM10 (3.17GHz) and TM11 (5.6GHz) modes.
In figure 52 the respective results for the diamond antenna integrated with the EBG are
presented. The EBG was designed to resonate in a band around 5.6GHz, namely in the band
of the second order mode of the diamond. In fig 52a a new resonance appears at 4.3GHz. It
is due to the fact that this frequency is out of the EBG band , thereby the EBG performs as an
electric conductor, namely as an ordinary metallic grounded sheet. So, the entire structure
operates as a microstrip antenna with a single dielectric layer, which is the upper dielectric
layer, operating at TM10. It can be also verified using the expressions for the frequency of
resonance of a triangular microstrip patch.
The radiation pattern at 4.3GHz is depicted in fig. 52c and is a pattern similar to respective
ones of microstrip antennas when operate at the fundamental mode. The radiation pattern
at 5.6GHz is depicted in figure 52d. Around this frequency the EBG is in resonance.
Comparing the figures 51c and 52d it is ascertained that the diamond exhibits a different
performance in the presence of the EBG. The pattern of 52d is similar to that of the
fundamental frequency rather, than to that of the ordinary second order mode. It is noted
that at both frequency ranges the diamond is driven only by its feeding probes and the
matching is satisfactory, as shown in figure 52b. So, the hybrid structure, microstrip
diamond-EBG, operates at two frequencies with similar radiation features.
The electric current distribution gives an explanation of the structure’s performance. In Fig.
53a the current at 3.17GHz, namely the TM10 mode of the antenna without the EBG, is
shown. This distribution is similar to that of the antenna with the EBG at 4.3GHz (Fig.53c).

Methods to Design Microstrip Antennas for Modern Applications
223
3456
-50
-25
0
25
50
75
100
125
Input impedance[Ohms]
frequency [GHz]
real
imaginary
(a)
23456
-20
-15
-10
-5
0
Reflection coefficient[dB]
frequency [GHz]
(b)
030
60
90
120
150
180
210
240
270
300
330
-30
-15
0
-30
-15
0
4.3GHz
-xz-
-yz-
(c)
030
60
90
120
150
180
210
240
270
300
330
-30
-15
0
-30
-15
0
5.6 GHz
-xz-
-yz-
(d)
Fig. 52. (a) Input impedance of the dual triangular patch antenna with EBG (b) The
reflection coefficient at the feed ports. (c), (d) the xz-and yz-plane directivity patterns of the
TM10 (4.3GHz) and TM11(5.6GHz) modes.
In this case, current does not exist on the EBG surface patches. That’s why the patterns of
directivity shown in Figures 51b and 52c, are similar. Comparing the current patterns of the
TM11 mode without and with EBG(Figs. 53b, 53d), we see that the current distribution is
modified. Intense currents flow on the EBG which resonates at this frequency. This
performance leads to the field depicted in Fig. 52d.
(a)
(b)
(c ) (d)
Fig. 53. (a),(b) Current distribution, at 3.17 GHz(TM10) and 5.6GHz(TM11), on the surface of
the triangular dual patch antenna without EBG. (c) and (d) The current distribution, at
4.3GHz and 5.6GHz when the EBG was embedded in the structure. In all cases the phase
difference between the probes is 180o.
It is worth to mention that the above described performance would lead to a design
procedure for various pairs of operation frequencies and various shapes of antenna and
EBG patches. The pair of frequencies have been given, the steps of the design are
summarized as follows:

Microstrip Antennas
224
a. The smaller frequency of the pair is defined as the f10 of the radiating patch and the larger
one as the frequency fmn of one of the higher modes as well as of the resonance of the EBG.
b. Taking into account that at f10 the EBG will perform as a perfect electric conductor and
the patch will work solely with the upper dielectric layer, we select the values of the
structural parameters of this layer and calculate the size of the patch.
c. At the selected fmn the EBG resonates, so the patch works with both dielectric layers and
we determine the characteristics of the lower layer in order the patch, with the already
calculated size, to resonate at fmn with the double dielectric substrate.
d. We calculate the size of the EBG patches to resonate at fmn in the determined dielectric
environment.
For example, suppose that 900MHz(GSM) and ~2GHz(UMTS) are the parts of the frequency
pair. We set f10=900MHz and f11=2GHz. Then selecting for the upper layer, r2 4.4ε= and
h2=0.5mm, the length of the side of the antenna has to be ap=105mm , in order to resonate at
900MHz. The calculations are made by equations 19 and 20. In order the patch, with side
length ap=105mm, to resonate at 2GHz with the TM11 mode, in the presence of the double
dielectric layer, the lower layer must have r1 2.2
ε
= and h1=1.5mm. The calculations are
made by equations 19, 20 and the expression
(
)
22
ri iri
i1 i1
hh
==
ε
=ε
∑∑ . The side length of the
EBG patch has to be a=65mm in order to resonate at 2GHz, if structured between the
determined dielectric layers. These values are of first order approximation. The simulation
of the configuration, in virtue of the interaction between the antenna and the EBG, is
expected to show that a small perturbation of the size of the elements is necessary to
resonate at the pre-specified frequencies.
4. Stochastic algorithms for antenna design and optimization
As in the previous sections has been discussed, the design of a printed antenna with
advanced characteristics of operation for multi-frequency applications is often a laborious
and difficult procedure. To obtain such a performance, the techniques of texturing the
surface of the printed patches, the utilization of fractal design procedure or the
incorporation of EBG objects are efficient tools for the design, but they are not always
enough. Moreover they produce complicated antenna schemes inserting additional
parameters that have to be taken into account during the design process. Besides all these,
there is a trade off between the antenna’s properties which are imposed by the technical
requirements. For example, the small size opposes the high gain; the small thickness of the
substrate reduces the surface waves and, in this way, the mutual coupling, but also reduces
the bandwidth; multi-frequency operation is, in general, not difficult to be achieved but it is
not always easy to obtain the frequencies specified by a particular application. On the top of
all, there is the difficulty that in most of the cases there are not mathematical expressions in
closed form that could help the designer of the antenna. Solution to the above problems
would be given by evolutionary techniques, most of which are inspired from the nature and
the biology. The Neural Network(NN) method, the Genetic Algorithms(GA), The Particle
Swarm Optimization(PSO) are some of the most commonly used methods which were
proved efficient in antenna design and more generally in many electromagnetic problems.
In the following a brief discussion and a very simple analysis of the NNs and GAs and will
be done. The detailed study is beyond the scope of this chapter and would be the content of
a separate self-existent book.