10

PATCH ANTENNAS

10.1. Introduction 10.2. Code Sequence

10.3. Model of the Probe Feed

10.4. Generation of the Antenna Structure

10.5. Input Impedance, Return Loss, and Radiation Pattern 10.6. Why Do We Need the Wide Patch?

10.7. A Practical Example

10.8. Dielectric Model

10.9. Accuracy of the Dielectric Model 10.10. Conclusions

References Problems

10.1. INTRODUCTION

Patch or microstrip antennas [1,2J are probably the most widely used class of antennas today. They are important in many commercial applications, such as mobile radio and wireless communications. These antennas are low-profile and conformable to planar and nonplanar surfaces ranging from aircraft and rocket shapes to human bodies. The modern patch arrays perhaps constitute the most sophisticated and intriguing electromagnetic journey today. A cornprehensive review of canonic patch/array shapes can be found in [2].

Compared to the other antenna shapes modeled in the preceding chapters, there are two new concepts that should be introduced when patch antennas

233

234 PATCH ANTENNAS

Metal patch

Metal ground plane

Coaxial connector

Figure 10.1. Typical probe feed for the microstrip antenna.

are studied. One concerns the supporting dielectric, and the other concerns the antenna feed. Patch antennas are very sensitive to the specific fceding method. Many of these methods exist [2].

Two groups of patch antennas may be considered: air-filled or 10wdielectric (SR - I) patch antennas and metal-dielectric antennas (SI( - 2-50). The presence of a high-s dielectric generally shifts the operation frequency to a considerably lower value and reduces its Q [1.3].

In this chapter we will study air-filled as well as 10w-£ finite-size patch antennas. The approximation of an electrically thin substrate [4] is used in the latter case. Such an approximation does not add much computational complexity to the MoM method used in the preceding chapters. Also the computational time remains nearly unchanged.

There are many configurations that can be used to feed patch antennas.

The most popular are microstrip line feed, probe feed, and a proximitycoupled feed [1,2]. In this chapter we will mostly study the coaxial probefeed patch antennas. This feed configuration is shown in Fig. 10.1. The way to model the probe feed is very similar to that for the base-driven monopole (Chapter 4).

The microstrip feed can be introduced in the algorithm as well. In this case the feeding edge should be the one of the strip edges. A printed dipole can be simulated in the same way. Then the feeding edge must the center edge of the dipole.

This chapter presents a simple patch antenna mesh generator that is used to create patch antenna structures in a matter of seconds. This generator employs the Matlab mouse function ginput to identify triangles belonging to the patch and the feed position. Multiple feeds and parasitic patches can be introduced.

10.2. CODE SEQUENCE

The code sequence is very similar to that of Chapter 7. The corresponding flowchart is shown in Fig. 10.2. To simplify the code, we omit 3D radiation

Patch antenna mesh from subdirectory mesh

mesh trnat

rwg1.m

rwg2.m

mesh2.mat

current.mat

current-mat

,--_sw_ee_p_l_ot_.m _ __.1 }

mesh2.mat I I

I-----~~ • rwg5single.~

current.mat ~ __J

mesh2.mat

efield3single.m

current mat '-- ~

CODESEOUENCE 235

l Creates RWG edge elements

Computes impedance matrix

} and solves MoM equations, Outputs impedance and current distribution

Plots input impedance and return loss

} Determines and visualizes surface currents at a single frequency

} Computes directivity

pattern(s) at a single frequency

Figure 10.2. Flowchart of the code sequence of Chapter 10. The script containing the frequency loop is grayed,

patterns. Significant changes are made in the script rwg3 . m where the code is modified to take the effect of dielectric into account. We will describe these changes in this chapter. The dielectric constant e (electric permittivity of the dielectric substrate) should be specified at the beginning of this script.

The script rwgl . m introduces a new array EdgeIndicator that is used to identify edge clements belonging to the different parts of the patch antenna. The following values are available:

% 0 - metal ground plane
% 1 - feeding bottom edges
% 2 - connecting strip (probe feed)
% 3 ~ feeding top edges
% 4 - patch area Indexes 1 and 3 corresponds to the edge elements that have one triangle, which belongs to the probe feed, and another triangle, which belongs to the ground

236 PATCH ANTENNAS

plane or to the patch, respectively. Also all triangles of the structure now have the following identification numbers:

%t(4,:)=O triangles of the metal ground plane
%t(4,:)=1 triangles of the probe feed
%t(4,: )=2 triangles of the patch
%t(4,:)=3 triangles of the upper boundary of dielectric 10.3. MODEL OF THE PROBE FEED

The thin-strip model of the probe feed is based on the algorithm of Chapter 4, and it essentially reproduces the model for the bottom-fed monopole. Figure 10.3 shows the probe feed model adapted to R\VG edge elements. The starting point is a strip in Fig. 10.3a, which models the probe connector. In Fig. 1 0.3a, two RWG edge elements, having common triangle T, support the surface current J that is directed exactly along the strip axis. For a thin strip the radius of the equivalent cylindrical wire is given by Eq. (4.1), that is, aC<Jv = 0.258, where 8 is the strip width.

A model of the strip-plate junction is shown in Fig. 10.3b. There arc two edge elements that have the common edge 1m. Such a model can connect two and more plates together as demonstrated in Fig. 1O.3c.

FOI the base-driven probe, the feeding edge at junction 1m requires special treatment (Chapter 4). It is common to both the plate and the strip. Therefore two RWG elements correspond to the same edge. To separate these two elements it is convenient to "double" the junction edge, that is, just repeat it twice in the mesh code. The delta-function generator is then applied to edge I,ii' The input impedance of the antenna is given by ratio of the common voltage through edge IIJI (1 V) and the sum of the currents of two edge elements COIresponding to 1,1/'

The feed identification is made in the script rwg3 . m. This is done using the array EdgeIndicator introduced in Section 10.2. Since a bottom-driven

a)

b)

c)

Figure 10.3. Model of the probe feed. (a) Single strip supporting vertical current; (b) stripto-plate junction; (c) strip connecting two plates. 1, 2 indicate two possible positions of the feeding edge.

GENERATION OF THE ANTENNA STRUCTURE 237

probe is mostly considered (Edgelndica tor==l), the following code is used to find the antenna impedance

Index=find(Edgelndicator==l) V(Index)=l*EdgeLength(Index) j

GapCurrent(FF)=sum(I(Index) .*EdgeLength(Index) ') j GapVoltage(FF)=mean(V(Index) ./EdgeLength(Index)) j Impedance(FF)=GapVoltage(FF)/GapCurrent(FF)

The coaxial connector should be small enough compared to the antenna size. Otherwise, the model of the TEM magnetic frill should be employed [5] to properly model the coaxial line excitation.

10.4. GENERATION OF THE ANTENNA STRUCTURE

Before attempting the calculations for the patch antenna, we must first create the antenna structure. This is done in the scrip patchgenerator. m from the subdirectory mesh of the Matlab directory of this chapter, As an example we will execute that script and not change anything in the code. The Matlab plot is shown in Fig. 10Aa.

This figure shows the ground plane of a patch antenna. The other three figures follow the next three steps:

1. Use the Matlab mouse cross and the left mouse button to mark the white triangles shown in Fig. lO.4b, These triangles will belong to the patch(es).

2. Press return key after you are ready. The return key fixes the patch(es) structure. The picture will be updated as shown in Fig. 1OAc.

3. Use the Matlab mouse cross and the left mouse button to mark the white triangles shown in Fig. lOAd. The edge between these triangles will belong to the feed(s). Press return key.

After these steps are completed, the antenna structure is created. It appears on the screen as shown in Fig. 10.5, The structure is the probe-fed finite patch antenna lOx 5 x 1 em in size, with a rather narrow patch. This structure could be rotated to change the observation direction.

Looking through the script patchgenera tor. rn, one can see that it employs the Matlab function delaunay in order to create the ground plane mesh. Next it separates the patch triangles using the method of images in the ground plane and lifts them up to a certain height. After that, the probe feed is introduced as a common edge to two triangles identified at step 3. The script patchgenerator.m allows the following parameters to be changed

Figure 10.4. Four stages of patch antenna generation. Mat!ab mouse cross is seen.

INPUT IMPEDANCE, RETURN LOSS, AND THE RADIATION PATTERN 239

Z, m o

-0.005

-0.01

-0.05

x, m

-0.01

y. m

0.05

Figure 10.5. Antenna structure after execution pa tchgenera tor .m.

%Separation distance between patch and ground plane h:::O.Ol;

L:::0.I0;

%plate length (along the x-axis)

T,['J=O. 05; %Plate width (along the y-axis)

Nx=17; %Discretization parameter (length)

Ny::: 9 ; %Discretization parameter (width)

%Number of rectangles of the feeding strip:

Number=3;

The script can introduce multiple patches and multiple feeds. The problems at the end of the chapter provide corresponding examples. The patch antenna's structure is not limited in size, but it is recommended that the total number of triangles does not exceed 4000.

Surprisingly the Matlab mouse's output has been found to be sensitive to the geometrical size of the structure. When the size of the structure is a small fraction of 1 (as it is usually the case), the mouse's output may stop working properly if the number of triangles exceeds 1000. Therefore it is necessary to rescale the antenna before the mesh generation.

Another problem with the script pa tchgenera tor. m is that the width of the feeding strip is limited by the discretization accuracy of the ground plane. The two patch generators considered below are free of this limitation.

10.5. INPUT IMPEDANCE, RETURN LOSS, AND THE RADIATION PATIERN

Although the primitive patch antenna model introduced in the previous section (Fig. I 0.5) is not very practical, it can readily be used to start the basic

240 PATCH ANTENNAS

E 500

£:

o

iii 400

o c:

ro

13

(II

l£

~

c:

ro 1i)

'w l£

- :l

0-

.s -100

Resistance-solid line; reactance-dashed line

600

b)

a)

300

200

100

o

:-

-

-200

. /.

.. I' ....

\' .

:/

.. ~

-300

a

6

-1

CIL2 '0

(t)

(t)

.Q

E -3

:J

Q5 0:::

-4

-5

-6

1

1.5

2.5

4.5

6

5

3

3.5

4

5.5

2

Frequency, GHz Figure 10.6. Input impedance (a) and return loss (b) of the simple patch antenna. The circle corresponds to the bandwidth area of the antenna.

WHY DO WE NEED A W1DE PATCH? 241

calculations. First, the script rwgl . m would be run and reproduce Fig. 10.5. Next, rwg2.m would be executed. After the RWG edge elements are created, the script rwg3 . m would be run. By default this script does a frequency loop of 21 points over the band from 1 to 6 GHz. The default dielectric constant (relative dielectric permittivity) of the substrate is 1. The execution time for the script rwg3 . m should be about 6 minutes on a Pentium IV processor with Intel™ motherboard.

After the script rwg3 . m is executed, the script sweeplot . m provides us with the input impedance and return loss of the antenna over the desired frequency band. Figure 10.6 shows the input impedance of the antenna and its return loss. The solid line denotes the resistance (impedance real part), and the dashed line corresponds to the reactance (impedance imaginary part).

The input impedance exhibits typical "resonance" behavior with the resistance peak [1, p. 763]. Such behavior is very common for other resonant antennas, as we learned in Chapter 7, and not only for the patch antenna. The return loss has a minimum of approximately -5 dB at 2.5 GHz. When a finer frequency step is employed, the minimum value drops to approximately -8dB. To ensure proper load matching, the antenna center frequency must therefore be close to 2.5GHz.

The radiation patterns are obtained using the script efield3 single. m.

They are shown in Fig. 10.7. It is seen that the present patch antenna performs rather poorly. In particular, the zero radiated field occurs on the antenna's axis. This is a weak design since the microstrip patch should have its maximum radiation normal to the patch (broadside radiator) [1].

There arc t\VO reasons why the broadside radiation is missing. First, the structure is symmetric. Since the currents on the "dipole patch" are oppositely directed, their contributions cancel each other in the far field. Figure 10.8 shows the surface current distribution on the patch obtained using the script rwgSsingle.m at 2.5GHz.

Second, the patch itself is too narrow to radiate effectively. In conventional patch antennas it is not the patch but the regions along the patch's edges that radiate the electromagnetic signal [1]. In these regions the E-field lines arc fringed. The region beneath the patch is basically a high-Q cavity resonator with an approximately uniform electric field. So, in order to achieve better radiation, we have to make the patch wider and enlarge its radiating ci rcumference.

10.6. WHY DO WE NEED A WIDE PATCH?

In this section we modify the preceding example in order to make the patch antenna work properly. The new antenna structure is shown in Fig. 10.9. The structure is again obtained running the script pa tchgenera tor. m. We do not have to change anything in the code. However, in contrast to the previous example, we increase the patch size (the patch now contains 9 x 5 rectangles).

242 PATCH ANTENNAS

a) a-plane

b) yz-plane

.. j
I
/
!
330
300 y

Figure 10.7. Radiation patterns of the dipole patch antenna at 2.5 GHz.

-0.02

y, m

Z, m o

-0.005

-0.01

0.05

Figure 10.8. Surface current distribution of the dipole patch antenna at 2.5GHz. The white color corresponds to larger magnitudes.

WHY DO WE NEED A WIDE PATCH? 243
a) y, m
0,025
0.02
0.015
0.01
0.005
0
-0.005
·0.01
-0.015
-0.02
-0,025
0 0.01 0.02 0.03 0.04 0.05
x, m ·0.05 -0.04 -0.03 -0.02 -0.01

o ·0.005 -0.01

-0.05

0.05 Figure 10.9. Patch antenna structure after execution patchger.erator.r:l.

Also the antenna feed is shifted toward one of the patch edges. Both of these changes are critical for the antenna's performance. The interested reader should follow exactly the steps of Section lOA to create the patch antenna shown in Fig. 10.9.

The main code sequence rwgl . ill, rwg2 . m, and rwg3 . m is executed. The result for the return loss is shown in Fig. 1 0,1 Oa. We see that the minimum of return loss again occurs at 2.5 GHz. This is the center frequency of the antenna. The load matching at 2.5 GHz is excellent, so the return loss drops to -16dB.

Two radiation patterns of the antenna are shown in Fig 1O.10h. Clearly, the present patch antenna performs very well as a broadside radiator with approximately a 600 beamwidth.

V.le know from the literature [I ,2J that real input impedances are obtained when the patch dimensions are roughly )J2 on a side. In Fig, 10.9, the patch

244 PATCH ANTENNAS

·4

·5

ttl

: ·6 _Q

·12

·14

8 F,equency. GHz

xz-plsne

b)

vz-plane

z

z

~..---~ ,~( ....

f'·.. . ·10.:

! ." .. - ... : ..... :

( :' ".: . '.,

~ '· J(;k

2~ ' .... ; ... ,

10 dB

10 ca

30

GO

I

" .. I X I I

240

300

270

Figure 10.10. Return loss and radiation patterns of the "wide" patch antenna at 2.5GHz.

A PRACTICAL EXAMPLE 245

length is approximately 5.6 em, which corresponds to A = 11.2cm and a center frequency off = 2.7 GRz. This value is in a good agreement with the frequency f= 2.5 GRz which corresponds to the minimum of the return loss in Fig. lO.lOa.

10.7. A PRACTICAL EXAMPLE

Now we consider a practical example of the dual-band E-shaped patch antenna studied in [7]. The air-filled patch antenna operates at 1.9 as well as 2.4 GRz [7, sec. III]. The patch height is 0.015 m. The antenna is fed by a coaxial probe. The probe location corresponds to the center of the common edge of two small white triangLes in Fig. lO.lla. A large but finite ground plane is assumed according to Ref. [7, fig. 1].

The antenna has a large ground plane and a slotted patch (Fig. 10.11).

Therefore the mesh of equal triangles created by patchgenerator. m will be too large. It is not appropriate for our purposes. We need a nonuniform mesh that has the finest resolution close to the patch slots as well as to the feed.

The initial 2D antenna template is created using the mesh generator of the Matlab PDE toolbox as shown in Fig. 1 0.11a. The corresponding POE file is eshape . m in subdirectory mesh. If you have the POE Toolbox installed on your machine, you can run this file using the Matlab command prompt to see the template. The large rectangle created by rectangle tool corresponds to the ground plane. An E-shaped polygon inside the rectangle is created using polygon tool, and it corresponds to the patch. A small rectangle within the patch is also created to refine the mesh close to the feed.

After we export the mesh to the main Matlab workspace (use export mesh and then press ok button) and save the result as eshape. rna t, the array of triangles will have in its last row the following display:

t (4, : ) t (4, : )

:-::1

:-::2 (or 3)

triangles of the metal ground plane triangles of the patch

The Matlab script pachgeneratorl.m in subdirectory mesh is intended to work with the PDE toolbox. It reads the binary file eshape .mat and creates the 30 patch antenna based on the existing triangle indication. It is not necessary to identify the patch shape using mouse. However, the mouse input still exists to identify the probe feed by marking two white triangles somewhere on the patch. The output of the script is shown in Fig. 1O.l1b. The width of the feeding strip is 0.0015 m. The structure has totally 1395 edge elements.

Scripts patchgenerator. m and patchgeneratorl.rn are two alternatives to the patch antenna mesh generation. Whereas the first script is rather a "quick and dirty" mesh generator, the second allows more precise adjustments in the equivalent size of the probe feed.

246 PATCH ANTENNAS

L"O.070; W"'O.050; Ls=O.040; Ws=0.006 m a)

0.1135m

y, m

0.06

0.04

0.02

o

-0.02

-0.04

-0.06

-0.030 m

0.014 m

0.04 x, m Figure 10.11. Patch antenna from Ref. [7] to scale. (a) Output of the POE toolbox; (b) output ~patchgeneratorl.

Next, the main code sequence rwgl . m, rwg2 . m, and rwg3 . m is executed, Calculations are done over the band 1.5 to 3.0GHz. Two models are considered: (1) feeding edge at the bottom edge of the strip (bottom feed 1 in Fig, HUe); and (2) feeding edge at the middle edge of the strip (center feed 2 in Fig. 10.3c). The results for the return loss are presented in Fig. 10.12, The experimental data [7] (dashed curve) and the HP-HFSS simulation (stars) performed in [7] are also given.

A PRACTICAL EXAMPLE 247

o -~:-\--------'------'-----'-------~)"""-l

'* *--

-5

-e-

U) ·10 .

00 00

.s

E

.2 -15 ru

a::

-20

if J I I

I I

* I

J

, I

, I

\ .... «,

-25

O~~~~----,-------,----

-5

~ -10 -

'" if)

o

E

~ -15

a::

-20

b)

I * I I

I

I f

.. i I

\ !

\ * l

'* !

I I

, ,

* \

\

I ' I

I, I,

1.5

-25L----------L--

2

2.5

3 Frequency, GHz

Figure 10.12. Return loss of the E-shaped antenna as a function of frequency. Dashed line: experiment [7]; stars: HFSS simulation [7]; solid line: present simulation. (a) Bottom feeding edge; (b) center feeding edge.

In Fig. lO.12a we see that the bottom feed provides good agreement between the present results and experiments/simulations done in [7]. At the same time the center feed model (Fig. lO.12b) fails in almost the entire frequency domain. Very similar results were obtained for other orientations and sizes of the feeding edge on the patch plane. Thus the bottom feeding edge with two adjacent edge elements reasonably well describes the probe excitation of the patch antenna. The coaxial line diameter should be small enough

248 PATCH ANTENNAS

compared to the antenna size. Otherwise, the model of the TEM magnetic frill should be employed [5] for coaxial line excitation.

10.8. DIELECTRIC MODEL

The full-wave analysis of patch antennas via MoM involves substantial computational complexity. For infinite substrates, the Green's function of the grounded dielectric slab has to be evaluated carefully [8-10]. For finite ground planes, different sets of basis functions are necessary either for the volume/surface formulation or for the electric and magnetic currents of the EFIE [11,12].

The simple moment method solution presented here must be viewed as an approximate solution valid for electrically thin dielectric slabs. It originates from the paper of Newman and Tulyathan [4]. The finite dielectric slab is removed and replaced by the equivalent volume polarization currents J (A/m2),

J = jWe(eu -1)E

(10.1)

where £ is dielectric permittivity of vacuum, eR is relative permittivity, and E is the actual electric field in the slab. The dielectric volume is divided into triangular volume elements as shown in Fig. 10.13. Triangles of the surface mesh are used for that purpose. The electric field within each volume element is assumed to have only one significant vertical component. This component is defined by the surface charges at the top patch and the bottom patch of every volume element

{~ 1 io; _)

Zz P,,' -P" E;;=; e

i_I (p~ -0) 2£

between ground plane and patch (Fig, 1O.13a)

(10.2)

between ground plane and vacuum (Fig.lO.13b)

a) b) c)
z Ps- 0
r
E UJj11lE J

p+ p+
S S Figure 10.13. Ouasl-static electric field distribution in a dielectric substrate.

ACCURACY OF THE DIELECTRIC MODEL 249

where p~ is the surface charge density on the metal ground plane and the metal patch. Equation (10.2) is the well-known quasi-static relation for the parallelplate capacitor. The surface charges are those for the RWG elements on the metal surface.

The volume polarization currents (10.1) alter the electric field on the metal surface. They also interact with each other through the radiated electric field. The radiated electric field of a single volume element shown in Fig. 1 O.13c is obtained using the dipole model (see Chapter 3). The dipole model replaces the current in the volume element by a finite-length vertical dipole of height h. For better accuracy, several dipoles can be considered that are uniformly distributed over the cross section S of the volume element.

The modified moment equations take the form (index S refers to metal and index D to dielectric)

I z.; LZsf)

-z.; ][I_\,] IV]

Z{){) - J If) - L 0

(10.3)

where In is the vector of unknown volume currents. Matrix ZDS is filled using the dipole model. Its mn-element is proportional to the electric field on R\VG element m. due to volume element n. Function point_. m is used to calculate the electric field of a finite-length dipole. Matrix ZSf) is filled using the capacitor model (10.2). Its mn-element is proportional to the electric field in volume element m due to RWG clement n. Square matrix ZIJJ) (self-interaction matrix) is filled using the dipole model as well.

These changes are incorporated into the script rwg3 . m where the matrixes Z.IS, ZlJs. Zsn. and Zvn are calculated separately. The total execution time is approximately 100% higher than the execution time of the pure metallic model. The execution time can be reduced using the code vectorization.

To pursue the full-wave solution, the lateral electric current in the dielectric must be supported. This can be done by doubling the ground layer of RWG edge elements and lifting it up by h12, where h is the patch height. Such an intermediate layer is capable of supporting any lateral current distribution in the dielectric. The full-wave dielectric solution for the patch antennas using RWG edge elements will be announced on the book website.

10.9. ACCURACY OF THE DIELECTRIC MODEL

Figure 10.14 shows a dielectric patch antenna setup to scale. This setup is a modified project DEMO-366 of"WIPL-D software l5J. The modifications imply that (I) patch thickness is reduced to minimum, (2) the feeding point is moved from the patch boundary to the inner point (0.1 0.166) of the patch, and (3) the patch height is increased to {l.U5 m. The probe radius 0.004 m is chosen, and it corresponds to a strip width of 0.016 m. WI PL-D presumabJy uses "theorem of transfer of excitation" [5, p. 58] and the center-fed wire segment to model

250 PATCH ANTENNAS

L=0.40; W=O.20 m 0.40 m

a)

y. m

0.3

O. 2 I'--"'""""-;<----,l~-; 0.166 m;":"_--\-+-+-i'--I-f--r--r-

01

~0.1

-0.3

-0.4 "-----"'--"--"----"'--"---'----"'---'

-0.2

-0.1

o

0.1

0.2

x, m

b)

Figure 10.14. Dielectric-filled patch antenna setup to scale. (a) Intermediate result after identifying the feed position; (b) output of patcl:ge:l.erator2.

the probe feed of a patch antenna. Therefore, for the purposes of comparison, the center feed model of Fig HUe is used here. It is programmed in the script rwg3 . m just below the code for the base-driven probe.

The initial uniform 2D mesh is created with the help of the script pa tchgenera tor2 . m in the subdirectory mesh. This script essentially repeats the script pa tchgenera tor. m. However, it manually introduces additional vertexes in order to refine the uniform mesh X, Y in the vicinity of the probe feed

CONCLUSIONS 251

and control the probe thickness. This is done by adding four points to the existing vertexes. The addition to the code of pa tchgenera tor. m may have the following form:

%Identify probe feed edge x= [0. 092 O. 108] ;

y=[O.166 0.166];

X= [X x];

y= [Y,y] ;

x1=[mean(x) mean(x)];

yl=mean(y} +2* [max(x)-mean(x) min(x)-mean(x)]; X=[X xl];

Y=[Y,y1];

The rest of the script remains unchanged. The result is seen in Fig. 1O.14b. The total number of RWG edge elements (359) matches the number of unknowns in the corresponding WIPL-D model (variable from 310 to 474).

Antenna calculations are done over the band 200 to 400MHz. The results for the return loss are presented in Fig. 1O.15a to c for three different values of the dielectric constant En = 1.0, 1.5, 2.6. The solid line corresponds to the present code, whereas stars denote the WIPL-D results.

Reasonably good agreement is observed in Fig. 10.15. We can track the evolution and decay of the fundamental antenna resonance shown by black arrows as Eli increases. Also noticeable is a significant disagreement [or the pure metallic model in Fig. 1 0.15a. We believe that this disagreement is due to insufficient discretization accuracy close to the antenna feed.

10.10. CONCLUSIONS

We have presented the code sequence for the air-filled patch antennas and for the patch antennas in the approximation of the electrically thin substrate. These codes were tested for a few examples.

The patch antenna structure can be created using one of the scripts pa t c hgenera tor /patchgenera torl/pa tchgenera tor2 . ill in the subdirectory mesh. These scripts are described in the text. To enable faster structure processing, all generator scripts include mouse output. Multiple feeds and a patch antenna array can be created using these scripts.

The total execution time of the model is approximately 100% higher than the execution time of the pure metallic model. To pursue the full-wave solution, the lateral electric current in the dielectric must be supported. This can be done by doubling the ground layer of RWG edge e1ements and lifting it up by h12, where h is the patch height. Such an intermediate layer is capable of supporting any lateral current distribution in the dielectric. No other surface or volume discretization becomes necessary. The interested reader can modify

-2

-7

'"

cg -3 _.

~ -4

<Jl <Jl

.3 -5

<:

:;

d) -6 a:

-8

~ .. j

b)

...

ell -2 "

f.

~ -3,·

..

'" a:

.3 -4

c

:s -5

~

-6

...

..

* .. '

.. ..

-7

-7

-9

200 220 240 260 280

300 320 340 360 380

400

Frequency, MHz

Figure 10.15. Return loss as a function of frequency for the tested patch antenna. Stars: WIPLo simulation [5J; solid line: present model. The arrows show the evolution of the fundamental resonance as the dielectric constant increases.

REFERENCES 253

the source code accordingly. The full-wave dielectric solution for the patch antennas using RWG edge elements will be announced on the book's website.

REFERENCES

1. C. A. Balanis. Antenna Theory: Analysis and Design, 2nd ed., ch. 14. Wiley, New York,1997.

2. 1. R James and P. S. Hall. Handbook of Microstrip Antennas, Vols, 1, 2. Peter Peregrinus Ltd., London, 1989.

3. J-F. Zucher and F. E. GardioL Broadband Patch Antennas, Artech House, Norwood, ?>.1A, 1995.

4. E. H. Newman and P. Tulyathan. Analysis of microstrip antennas using moment method. IEEE Trans. Antennas and Propagation, 29 (1); 47~53, 1981.

5. B. M. Kolundzija, J. S. Ognjanovic, aod T K. Sarkar. WIPL-D: Electromagnetic Modeling a/Composite Metallic and Dielectric Structures. Artech House, Norwood, MA,2000.

6. D. M. Pozar, Microwave Engineering, 2nd ed. Wiley, New' York, 1998, p. 67.

7. F. Yang. X.-X. Zhang, X. Ye, and Y. Rahmat-Sarnii, Wide-band E-shaped patch antennas for 'wireless communications. IEEE Trans. Antennas and Propagation, 49 (7); 1094-1100,2001.

8. M. Marin, S. Barkeshli, and P. H. Pathak. Efficient analysis of planar mictrostrip geometries using a closed-form asymptotic representation of the grounded dielectric slab Green's function. IEEE Trans. Microwave Theory and Techniques, 37 (4): 669~679, 1989.

9. R. Kipp and C. H. Chan. Triangular-domain basis functions for full-wave analysis of microstrip discontinuities. IEEE Trans. Microwave Theory and Techniques, 4"1 (6/7): 1187-1194,1993.

10. L Tarricone, M. Mongiardo, and F. Cervelli, A quasi-onedimensional integration technique for the analysis of planar microstrip circuits via MPIE/MoM. IEEE Trans. Microwave Theory and Techniques, 49 (3); 517-523,2001.

11. T K. Sarkar, S. M. Rao, and A. R Djordjevic. Electromagnetic scattering and radiation from finite microstrip structures. IEEE Trans. Microwave Theory and Techniques, 38 (11): 1568~1575, 1990.

12. B. G. Salman and A. McCowen. The CFJE technique applied to finite-size planar and non-planar microstrip antenna. Computation in Electromagnetics (Conf. Publ, Ko. 420), pp. 338---341,1996.

13. D. Sievenpiper, L. Zhang, R. F. Jimenez Broas, N. G. Alexopolous, and Eli Yablonovitch. High impedance electromagnetic surfaces with a forbidden frequency band. IEEE Trans. Microwave Theory and Techniques, 47 (11): 2059~2074, 1999,

14. D. Sievenpiper, H. P. Hsu, 1. Schaffner, G. Tangonan, R Garcia, and S. Ontiveros.

Low-profile, four sector diversity antenna on high-impedance ground plane. Electronics Letters, 36 (16): 1343-1345,2000.

IS. R F. Jimenez Broas, D. Sievenpiper, and Eli Yablonovitch. A high-impedance ground plane applied to a cell phone handset geometry. IEEE Trans. Microwave Theory and Techniques, 49 (7) 1262-1265,2001.