INTERNATIONAL JOURNAL OF NUMERICAL MODELLING : ELECTRONIC NETWORKS, DEVICES AND FIELDS,

Vol. 7, 189-199 (1994)

ANALYSIS OF A THIN WIRE CIRCULAR LOOP ANTENNA

EDWARD K . N. YUNG AND ROGER S. K. WONG

Department of Electronic Engineering, City Polytechnic of Hong Kong, 83 Tat Chee Avenue, Hong Kong

SUMMARY
A rigorous technique for studying a thin wire circular loop antenna excited by a magnetic frill is developed.
The exact kernel is derived and used to improve the accuracy. The scheme is applicable to small as well as
large loops. Results obtained are nearly indistinguishable from those determined by the Fourier expansion
method for non-resonant loops. A disagreement of 2 per cent is generally observed when the loop is close
to resonant for pb > 2. Compared with the conventional numerical method where a loop is represented by
a polygon of straight wires, the present scheme is much faster and more accurate. Extension of this method
to cover an antenna system of arbitrarily oriented loops intermingled with straight wires can be easily
achieved.

1. INTRODUCTION

Thin wire loops and straight wires are the most extensively used elements in the design of wire
antennas. Owing to the abrupt change in geometry at both ends, waves are bouncing back and
forth on the surface of a straight wire antenna. As a result, the current distribution on the surface
of the antenna can only be determined by numerical means.1.2 On the other hand, the current
on a loop antenna is less complicated. When the wire is thin, Hallen3 and later Storer4 used the
method of Fourier series expansion to determine the current on a circular loop. By modifying
the kernel, Wu5 was able to express the coefficients in a series of special functions. For small
loops, good agreement between theoretical results and measurement is observed. Ever since these
elegant formulas were published, they have been used in nearly every study of loop antennas6*’
as if there were no alternative technique.
The major advantage in using the Fourier series method is speed. For small loops, the series
converges in less than 20 terms; thus, the analysis can be completed in fractions of a second.
Ironically, determining the coefficients by formulas is also its major handicap because for any
change in antenna configuration, all formulas must be re-derived from scratch, and a super
computer is of little help. Hence, extension of the Fourier method for an array of loops is seldom
found.6 Apparently, it is impossible to further extend the method for analysing a system of
arbitrarily oriented loops of various sizes intermingled with straight wires.
Analysis of a loop antenna by numerical techniques was first found in the 1 9 6 0 ~Using ~ a
reduced kernel and the Galerkin method, Harrington9 obtained a diagonal impedance matrix
whose elements are given in a form which is similar to Wu’s formulas. Compared with Wu’s
results, these numerically obtained data are less accurate. This is expected because these studies
were carried out during the teething days of the numerical method development. Little progress
on numerical analysis of a loop antenna was observed in past decades. However, significant
advancement was witnessed on the study of cylindrical antennas in the same period. In fact, the
development is so one-sided that a loop is now treated as a polygon of straight wires in many
CAD packages such as the NEC.l0 Although excellent results in the study of far-field scattering
can be assured by these indirect methods; results obtained in near-field studies and antenna
analyses are merely acceptable. This inconsistency is perhaps due to the use of the reduced kernel.
Attempts to rectify the problem by replacing the reduced kernel with the exact kernel of a
cylindrical wire have been made, and better results for a curvilinear antenna are observed.*l
Unfortunately, the error due to approximating a loop by a polygon is never addressed.

CCC 0894-3370/94/030189-11 Received 8 December 1992

@ 1994 by John Wiley & Sons, Ltd. Revised 1 June 1993
190 E. K . N . YUNG AND R. S. K . WONG

2. FORMULATION

The configuration of concern is a circular loop antenna of radius b as shown in Figure 1. It is

constructed from a perfectly conducting wire of radius a. The antenna is excited by a delta
function generator or a magnetic frill placed at 4 = 0. Harmonic analysis is used and the elor
factor is suppressed throughout this paper. Compared with wavelength, the wire is thin; while
the diameter of the loop is of the order of a wavelength. The loop is placed on the x-y plane
with its centre at the origin. Under these arrangements, the surface current on the loop is
predominantly in the azimuth direction, 4.
For easy reference, a local co-ordinate system (u,v,w) as shown in Figure 1 is defined such
that one of the base vectors, say w , coincides with the axis of the loop. Unit vectors in this local
coordinate system are dependent on 4; i.e.,

1.2 = i cos( 4) + 9 sin( +)

$ = R sin($) - 9 cos(+)
$=f

Since the loop is circular, a point on the surface of the loop is completely specified by two
parameters, namely a and 4. The position vector of a point on the surface of the loop r is given
by

r = R[b cos(+) + a cos(4) cos(a)]

+ 9[b sin(4) + a sin(4) cos(a)]+ fa sin(a)
Based on the thin wire approximation, there is no loss of generality by fixing the observation
point at a = 7/2. It then follows that the Euclidean distance, R, between the observation point
( d 2 , 4) and the source point (a’,+’)is

R = { [ bCOS(+)- b C O S ( + ’ ) - u COS(+’)C O S ( ~ ’ ) ] ~
+ [b sin(4) - b sin(4’) - a sin(4’) cos(a‘)I2
+ [ a - a .~in(a’)]’}”~ (3)
As the wire is of infinite conductivity, currents and charges reside on the cylindrical surface only.
Thus, the current induced on the loop is better expressed in terms of the surface current density,
J,. Again, by invoking the thin wire approximation, the current density is circumferentially
independent; i.e.,

Figure 1. A loop antenna and the local co-ordinate system

 THIN WIRE CIRCULAR LOOP ANTENNA 191

where Z is the total loop current. Note that the unit vector in the above equation is also dependent
on 4‘:

The equivalence principle is applied to determine the total current and the line charge on the
loop. Accordingly, the induced currents and charges are treated as impressed sources radiating
in a homogeneous space. The correspondent electromagnetic fields are determined via the magnetic
vector potential A. In terms of the total loop current, A is

In the above equation, K is the exact kernel which is defined by

where p = o d p 0 c o
Similarly, the charge distribution on the loop is better described by the total line charge Q ,
and the electric scalar potential 0 is given in terms of the line charge; i.e.,

The unknown current and charge are determined by enforcing that the tangential electric field
is zero on the surface on the antenna. On the surface of the loop, the only component of interest
is E,. Thus, one scalar wave equation is sufficient to determine the total electromagnetic field;
i.e.,

After some algebraic manipulations, A, can be expressed as

For a loop excited by a delta function generator of voltage V0 across a gap of width 26 (6 0),
the boundary condition is

If the antenna is excited by a frill generator as shown in Figure 2, the above boundary condition
is replaced by
192 E. K. N . YUNG AND R . S. K . WONG

Figure 2. The magnetic frill generator

where u, is the outer radius of the magnetic current frill and R, is a shorthand notation for

Subject to the boundary condition given in (11) or (12), the unknown currently and charges
can be determined by solving (9). Similar to its counterpart in the study of cylindrical wires, this
equation can only be solved by numerical means.

3. NUMERICAL TECHNIQUES

The method of moments (MOM)12 is adopted to solve (9). The current is. expanded into a series
of N + 1 pulses:

where f,, are the coefficients of expansion to be determined and P,, are pulse functions. Since
the current is continuous at the driving point, f, = I N ; thus, the number of unknowns is reduced
to N. Pulses of equal width are used except the pulses on either sides of the generator. The
interior pulses are twice as wide as the end pulses as depicted in Figure 3. The width of the end
pulse A+ is

Current
Expansion Charge
h Expanslon

q)* 2a

Figure 3. Current and charge expansion pulses

 THIN WIRE CIRCULAR LOOP ANTENNA 193

IT
A+=- (14)
N

The centre of the nth pulse, +n is

+, = 2nA+ n = 1 , 2 , ..., N - 1

In terms of A+ and +,, the pulse functions are

1 , O<+<A+
0 , otherwise

pn(+> = 0 { 1 7 +n-A+<+<+n+++
, otherwise n = 1 , 2 , ..., N - 1 ;
1 , 2~-A+<+<2n
0 , otherwise

In order to better represent the phenomenon that the current is proportional to the rate of change
of the charge accumulation and for a faster rate of convergence, the line charge is expanded in
overlapping pulses as illustrated in Figure 3; i.e.,

where Q, are expansion coefficients and the pulse functions are

Pn(+) = 0 {1 ,, otherwise
+n<+<+n+l
n = l , 2 ,...,N .

Substituting Z in (10) by its expansion, A,(+) can be expressed as a series of integrals of the
kernel K. Similarly, replacing Q in (8) by its expansion, @(+) can also be written in terms of
integrals of K.
Two difficulties, namely, evaluations of the kernel and its derivative, must be resolved. Note
that when the variable of integration +'
is equal to the observation point +,
K is singular. It then
follows that both A, and @ are given in terms of singular integrals. A technique is developed to
determine these singular integrals efficiently and accurately. However, for smoother reading,
details of this are presented in the Appendix.
On the other hand, the problem of evaluating the derivative of a singular integral is handled
indirectly. To this end, a set of N triangular functions are chosen:

m = 1,2, ..., N -1

Multiplying (9) by one of these testing functions, and integrating the resultant product over the
range of interest, a set of N linearly independent algebraic equations is obtained. By the process
of integration by parts, the difficulty in evaluating d@(+)/d+ can be obliterated.
The number of unknown coefficients can be halved by invoking the principle of charge conser-
vation. Accordingly, the current and charge expansion coefficients are related by
194 E. K. N. YUNG AND R. S. K. WONG

Thus the reformulation of (9) for numerical analysis is completed. The unknown current coef-
ficients can be determined by a standard matrix solution technique such as the Crout reduction
method.13

4. RESULTS

A circular loop of b = 47.75 mm and a = 2-013 mm is used to highlight various characteristics

of the present scheme. The frequency of operation is 1 GHz; therefore pb = 1, pa = 0.042, and
the normalized thickness parameter = 21n(2nb/a) is 10. The current distribution on the loop
induced by a magnetic frill generator is sketched in Figure 4. The rate of convergence is very
fast where satisfactory results can be obtained for N = 12. This rate is comparable to that of the
analysis of cylindrical antennas where a minimum of eight pulses is required to depict the variation
of current on a half wavelength dipole. Also shown in Figure 4 is the current determined by the
Fourier series expansion method revised by Zhou.' Nearly perfect agreement between both sets
of results is observed. As Zhou claimed that his results are practically indistinguishable from
experimental data, we conclude that the numerically obtained currents are as accurate. For
comparison, a modified NEC where the reduced kernel is replaced by the exact kernel for
cylindrical antenna studies is employed to analyse the loop. It appears that the deviation between
the results obtained by NEC and MOM can never be reduced even though as many as 100 straight
wires have been used. At first glance, the deviation will disappear should each wire segment be
infinitesimally short. It is not true because an infinitesimal segment contradicts the thin wire
assumption where the radius of a dipole is negligible compared with its length.
The currents on the loop as illuminated by both magnetic frill and delta function generators
are sketched in Figure 5. Excellent agreement between the currents due to both excitation models
is seen except at points very close to the driving point. On the one hand, the input conductance
of a delta function driven loop converges rapidly to a value which is very close to its counterpart
obtained by a magnetic frill. On the other hand, the input susceptance apparently does not
converge at all as illustrated in Fig. 6 where the imaginary part of the current in the vicinity of
the driving point is sketched for N = 20,40, 100 and 200. This type of irregularity is also observed
in analogous studies of cylindrical antennas as it is a consequence of the unrealistic excitation

10
 THIN WIRE CIRCULAR LOOP ANTENNA 195

0 45 90 135 180
+ (degree)

Figure 5. Comparison of current distributions as obtained by delta function generator and magnetic frill

6.0

5.5

5.0

4.5
v
E
$ 4.0
C
0
+
23.5
V
VI

2 3.0
2.5

2.0

1.5
0 5 10 15 20 25 30 35
+ (degree)

Figure 6. Convergence of the input susceptance

model. As seen in Figure 6, the rate of convergence can be substantially improved if the impulse
source is replaced by a source with a finite gap 6. In this example, by choosing 6 = 1 mm (1.25"),
the input susceptances are 4.9 mS, 5 mS, 4.9 mS, 4.8 mS for N = 20, 40, 100, 200 respectively
(4.7 mS by measurement). The magnetic frill is used in all subsequent studies because it is difficult
to choose a 6 without the knowledge of experimental results.
Current distributions on the loop (a = 10) for pb = 1.5, 2-0, and 2.5 are sketched in Figure
7. Nearly perfect sinusoidal distributions are observed; therefore it is believed that a faster rate
of convergence can be obtained by expanding the current by entire domain functions. However,
pulse expansion is preferred because of its simplicity and easiness in extension to cover more
196 E. K. N. W N G AND R. S. K. WONG

-2 1 \
\
\
\
., /
,I'

/
/
/
7-

-4 I , I , , I , , , , ,

1
0 45 90 135 180
+ (degree)
Figure 7. Current distribution on a loop antenna of various sizes. (a) Real. (b) Imaginary

complex antenna systems. Moreover, it provides us a better mechanism to compare our results
with those obtained by NEC.
The input admittance of the loop (a = 10) is plotted as a function of electrical size in Figure 8,
Results obtained by the Fourier expansion method are also included for comparison. Admittances
obtained by Wu's formulas and the modified formulas by Zhou' are shown. For small loops,
three results are nearly indistinguishable from one another. Note that, results obtained by Wu's
formula are based on the first 20 terms. If more than 20 terms are used, worse results will be
obtained. This type of finding is also observed by Reference 7. Moreover, when pb is larger than
1, results obtained by Zhou's formulas begin to deviate from the numerical solution. In order to
further investigate the limit of Zhou's model, it is used to study very large loops. Comparison of
input admittances obtained numerically and by Zhou's formulas for various large loops (a = 10)
is shown in Figure 9. Except at points of resonance, good agreement is observed. A disagreement
 THIN WIRE CIRCULAR LOOP ANTENNA 197

Zhou's Formula
0 0 Q 0 0
* = = Wu's Formula
x

__ MOM - magnetic frill

. . .b . . . . . . . . . . . . . . . . . . .
0.0 0.5 1 .o 1.5 2.0
Bb
Figure 8. Variation of the input admittance of a loop antenna with its size

iv
- 1 I
1 2
*

3
Zhou's Formula
__ MOM - magic frill
" , , , , , , , , , , , , , ' , , ,
4
Bb
Figure 9. Variation of the input admittance of large loop antennas

of 2 per cent is generally witnessed when the loop is close to resonant for pb > 2 and the deviation
widens with the increase in pb.
At this point, it is worthwhile to compare and contrast the exact kernel and the modified kernel
used by Wu. According to Wu, the distance between the observation point and the source point
is approximated by R,:

R, =
J 4 2 sin2(T) + 4b2sin2
where T is a dummy variable of integration. It is observed that R , is very close to R , as given
198 E. K. N. YUNG AND R. S. K. WONG

in (A3). For small T, sin(T) can be approximated by T. Then, by comparing (21) and (A3) term
by term, it is found that the modification introduced by Wu is the omission of the second term
in (A3). Although this assumption is valid when a + b, the authors are confused as the second
term is ignored while the smaller first term is retained. Should both terms be ignored, the so-
called reduced kernel is obtained which implies the enforcement of the wave equation at points
on the axis of the wire and there is no singularity. A kernel with an appropriate singularity is
needed for the evaluation of the near field. It is especially critical in determining the input
impedance of an antenna. For 14 - 4’1 + 0, Wu’s kernel exhibits the following singularity charac-
teristics

By recognizing that bl4 - +‘I is analogous to (z - 2’1, Wu’s kernel and the exact kernel in the
study of straight wires share identical singular characteristics. This is expected because when
4’ + 4, an incremental section of the loop resembles a straight cylinder. On the other hand,
comparing it with the exact singularity given in (A4), the difference is a small constant! In other
words, Wu’s kernel provides a nearly perfect singular characteristic while the reduced kernel
cannot.

5. CONCLUSIONS

An accurate method is developed for the study of loop antennas. The exact kernel is derived
and used to enhance the accuracy of the numerical scheme. It is valid for small as well as large
loops. The present algorithm is extremely computationally efficient because many properties of
symmetry have been utilized to speed up the process. It is at least comparable to the summation
of special functions. Note that to sum the Fourier series, several special functions and their
integrals are needed to be computed and, as a result, summing the Fourier series is as time-
consuming as enumerating its singular integrals in MOM. This applies to small as well as large
loops because the series converge much more slowly for large loops. In addition, the memory
required is small; therefore the scheme can be easily implemented in a personal computer.
Although only two models of excitations are used in the present study, it is foreseen that
modification to accommodate any new excitation model can be readily achieved. There exists no
doubt that this technique can easily be extended to cover antenna systems of arbitrarily oriented
loops intermingled with straight wires. Moreover, all techniques are readily applicable in the
study of wave scattering, and therefore they can be used to enhance the speed and accuracy of
existing codes.

APPENDIX

For the evaluation of the kernel given in (7) when I+ - +’I + 0, a new variable of integration
is introduced to replace a‘;i.e.,

a’ Tl
-T=---
2 4

It then follows that R can be rewritten as

RZ
- = a2 + 62 - a6 sin(2~)- a* COS(27)
2
- [b2- ab s i n ( 2 ~ ) ] ~ 0 ~-( (+’I)
+

It is observed that R vanishes when both I+ - +‘I and T are approaching zero. Denote the limiting
value of R by R,:
 THIN WIRE CIRCULAR LOOP ANTENNA 199

(A3)

To evaluate the kernel accurately and efficiently, K is partitioned into

With the singularity removed, the first integral on the RHS of the above equation can be evaluated
by a standard numerical integration method such as the Gaussian quadrature method.13 On the
other hand, the second integral, denoted by K,, can be evaluated analytically; i.e.,

REFERENCES
1. D. R. Wilton, ‘Current status of antenna modeling using the method of moments’, Journkes Internutionales de sur
les Antennes, Nice, France, November 1990, pp. 17-21.
2. A. J. Julian, J. C. Logan, J. W. Rockway and D. R. Wilton, ‘MININEC, A mini-numerical electromagnetic code’,
ZEEE Antennas and Propagation Symposium 1983, Vol. 1, 1983, pp. 220-223.
3. E. Hallen, ‘Theoretical investigations into the transmitting and receiving qualities of antennae’, Nova Acta Regine
SOC. Sci. Upsaliensis, ser. IV, 11, no. 4, 1-4 (1938).
4. J. E. Storer, ‘Impedance of thin wire loop antennas’, Amer. Znst. Elec. Eng. Trans., 75, pt. 1, 606-619 (1956).
5. T. T. Wu, ‘Theory of thin circular loop ai-tennas’, 1. Marh. Phys., 3, 1301-1304 (1962).
6. T. Korekado, K. Okuno and S. Kurazono, ‘Design method of Yagi-Uda two stacked circular loop array antennas’,
ZEEE Trans. Antennas Propagat., AP-39, 1112-1118 (1991).
7. G. Zhou and G. S. Smith, ‘An accurate theoretical modei for thin wire circular half loop antenna’, IEEE Trans.
Antennas Propagat., AP-39, 1167-1177 (1991).
8. A. Baghdasarian and D. J. Angelakos, ‘Scattering from conducting loops and solution of circular loop antennas by
numerical methods’, Proc. of ZEEE, August, 818-822 (1%5).
9. R. F. Harrington and J. Mautz, ‘Electromagnetic behaviour of circular wire loops with arbitrary excitation and
loading’, Proc. ZEE, 115, no. 1, 68-77 (1968).
10. G. J. Burke and A. J. Poggio, ‘Numerical electromagnetic code (NEC)-method of moments’, Naval Electronic
System Command, Naval Ocean System Center, San Diego, CA, Tech. Doc. 116, AWL-TR-76-320, 1977.
11. N. J. Champagne 11, J. T. Williams, R. M. Sharpe, S. U.Hwu and D. R. Wilton, ‘Numerical modelling of impedance
loaded mJlti-arm Archimedean spiral antennas’, ZEEE Trans. Antennas Propagat., 40, no. 1, 102-108 (1992).
12. R. F. Harrington, Field Computation by Moment Methods, Macmillan, New York, 1968.
13. F. B. Hildebrand, Introduction to Numerical Anafysis, McGraw-Hill, New York, 1974.

Authors’ biographies:
Edward K. N. Yung was born in Hong Kong in 1947. He received the B.Sc. with
special distinction, M.Sc and Ph.D. degrees from the University of Mississippi, all in
electrical engineering, in 1972,1974, and 1977, respectively. After graduation, he worked
briefly at the Electromagnetic Laboratory, University of Illinois at Urbana-Champagne.
In 1984 he joined the newly established City Polytechnic of Hong Kong. He is currently
a Reader in the Department of Electronic Engineering and Head of the Teleeommuni-
cation Research Center. His research interest is in electromagnetic theory, numerical
analysis and antenna design.

Roger S. K. Wong was born in 1969. He received the B.Eng. degree in computer
engineering from the City Polytechnic of Hong Kong in 1991. He stays with his Alma
Mater where he is currently a graduate student in the Department of Electronic Engin-
eering. His research interests include telecommunication systems, computer languages
and antenna design.