Figure 5 Measurement errors in the real and imaginary parts of

the permittivity of the PVC with a sample length of L s 120 mm and

L s 0.1 mm

From these values, an excellent agreement is observed

with those reported previously in the literature, also plotted
in Figure 4 as reference values w7x. The measurement accuracy also has been calculated using the procedure described
in Eq. 7., and is plotted in Figure 5. From this figure, it can
be seen that the error in the determination of the dielectric
constant is lower than 0.5%, while for the loss factor, this
value is increased to approximately 10%. Both values are in
good concordance with the theoretical analyses presented in
this paper, which may be reduced by considering longer
samples, as previously outlined.
IV. CONCLUSION

A fast and simple method for determining the complex dielectric properties of materials without ambiguity has been
developed based on one-port reflection measurements made
by an ANA. The developed algorithm solves the inherent
ambiguity of the problem under analysis, and eliminates the
need for any initial estimation of the dielectric properties, as
required by other reflection methods.
The error has been analyzed taking into account the
typical uncertainty of the measurement equipment and the
unascertained sample dimensions. Using these uncertainties,
the sample length has been identified as a very important
parameter to obtain high accuracy in the dielectric characterization. For low-loss materials, the accuracy of the method
may be limited if very long samples are not considered. In
this case, resonant cavity methods might be used to increment the accuracy.
Although the application of this method is useful for all
materials, it is particularly recommended for liquids, powders,
and granular materials.

4. S.O. Nelson, C.W. Schlaphoff, and L.E. Stetson, A computer

program for short circuited waveguide dielectric properties measurements on high or low loss materials, J Microwave Power 8
1973., 1322.
5. S. Travelsi, A.V. Kraszewsky, and S.O. Nelson, Use of measurements at two frequencies for phase-shift determination, IEEE
MTT-S Int Microwave Symp Workshop on Electromag Wave
Interaction with Water and Moist Substances, 1996, summaries,
pp. 129130.
6. K. Baek, H. Sung, and W.S. Park, A 3-position transmissionrreflection method for measuring the permittivity of low loss materials, IEEE Microwave Guided Wave Lett 5 1997., 35.
7. C. Wan, B. Nauwelaers, W. De Raedt, and M. Van Rossum, Two
new measurement methods for explicit determination of complex
permittivity, IEEE Trans Microwave Theory Tech 46 1998.,
16141619.
8. W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery,
Numerical recipes in C. The art of scientific computing, Cambridge University Press, Cambridge, England, 1992, 2nd ed.
A new model
9. S. Lefrancois, D. Pasquet, and G. Maze-Merceur,

for microwave characterization of composite materials in

guided-wave medium, IEEE Trans Microwave Theory Tech 44
1996., 15571562.
10. J. Barker-Jarvis, E.J. Vanzura, and W.A. Kissick, Improved technique for determining complex permittivity with the transmissionrreflection method, IEEE Trans Microwave Theory Tech 38
1990., 10961103.
2000 John Wiley & Sons, Inc.

OPTIMAL REDUCTION OF THE

INFLUENCE OF RF FEED CABLES
IN SMALL ANTENNA MEASUREMENTS
C. Icheln,1 M. Popov, 2 P. Vainikainen,1 and S. He 2
IRC / Radio Laboratory
Helsinki University of Technology
FIN-02015 HUT, Finland
2
Department of Electromagnetic Theory
Royal Institute of Technology
S-10044 Stockholm, Sweden

Recei ed 2 No ember 1999

ABSTRACT: A genetic algorithm is applied to optimize the size and
location of a slee e, which reduces parasitic currents on the surface of
the feed cable of a mobile handset antenna, so that the influence of the
RF cable on the radiation pattern is minimized. The pattern is calculated
by the electric-field integral equation (EFIE) method in each iteration of
the genetic algorithm. 2000 John Wiley & Sons, Inc. Microwave Opt
Technol Lett 25: 194196, 2000
Key words: small antenna measurement; quarter-wa e slee e;
parasitic currents on coaxial RF cable; genetic algorithm;
minimization of radiation pattern perturbation
INTRODUCTION

REFERENCES
1. R.C. Jain and W.A.G. Voss, Dielectric measurements methods
for industrial scientific and medical applications in microwave
frequency range, IETE Tech Rev 11 1994., 297311.
2. R.H. Voelker, G. Lei, G. Pan, and B.K. Gilbert, Determination
of complex permittivity of low-loss dielectrics, IEEE Trans Microwave Theory Tech 45 1997., 19551960.
3. S. Roberts and A. Von Hippel, A new method for measuring
dielectric constant and loss in the range of centimeter waves,
J Appl Phys 17 1946., 610616.

194

During the development of a mobile handset antenna, the

complex transmission coefficient between a well-known measurement antenna and the mounted handset antenna under
test AUT. is usually measured with the help of a network
analyzer. Due to the RF coaxial measurement cable attached
to the handset antenna, this setup does not resemble the
standard operation setup of the mobile handset.
Contract grant sponsor: Nordisk Forskerutdanningsakademi NorFA .
Contract grant sponsor: Academy of Finland

MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 25, No. 3, May 5 2000

Errors in the measured radiation pattern or antenna gain

can be expected due to the RF cable acting as: 1. a secondary
radiator driven by the surface currents on the conducting
surface of the handset, and 2. a reflector that reradiates part
of the arriving fields. It was found w1, 2x that a feed cable
vertically positioned. could introduce large errors in the
radiation characteristics of measured antennas with vertical
polarization. Ferrite chokes around the cables reduce the
cable-related effect significantly. However, the disadvantage
of generally lossy ferrite beads is their effect of partly absorbing the energy propagating along the RF cable, which reduces
the total radiated power of the antenna. In w3x, the use of
cap-like sleeves on the feed cable, located near the handset,
was proposed as an effective, lossless, although band limited,
means to reduce the effect of the RF cable on the antenna
measurement Fig. 1..
The aim of this work is to optimize the size and position of
the sleeve in order to obtain a good design for the prototype.
The finite element method code that was used in the previous
work w3x is too time consuming when it comes to frequently
changing a model of the measurement setup. Also, the calculation time increased considerably when the set-up was modeled with a longer RF cable than in w3x. A cable length of
0.51 m is considered to be a sufficient description of the real
case with an arbitrarily long measurement cable.
METHOD-OF-MOMENTS CODE

A method-of-moments MoM. code was developed using the

electric-field integral equation EFIE.. The code is based on
w46x. The handset, including the cable and sleeve, is modeled
by 360 planar triangular surface patches see Fig. 2.. A set of
basis functions is used, where each basis function is defined
on a pair of adjacent triangular patches. The excitation of the
model is done by a current dipole point source, located where
the tip of a typical helix antenna is located on a real handset,
i.e., 15 mm from the upper end of the chassis see Fig. 2.. The
initial boundary conditions for the MoM code are given by
the currents on the surface of the object, which are directly
induced by the current dipole.

Figure 1 Mobile handset with feed cable, sleeve, and current

dipole at location of antenna

Figure 2 Triangular mesh of the handset model simulated at 1 GHz.

The mesh consists of 360 triangles. The cable continues to the left at
a length of 600 mm

The handset model is characterized by conducting surfaces. The model includes the rectangular box of the handset
chassis with the same dimensions as used in w3x: l s 100 mm,
w s 40 mm, and t s 10 mm. The RF feeding cable is 600 mm
long, and the coaxial sleeve on the cable has a variable size
and position. The sleeve is defined as an infinitely thin
conducting sheet.
The MoM code is used to calculate the current distribution, on the triangular patches. From the current distribution,
the radiated far-field pattern at any point in space can be
obtained by superposition of the radiated fields from all
current sources on the model surface.
However, one drawback of the MoM code is obvious if we
look at the long end of the feed cable. A typical feed cable
during a measurement is several meters long, and after some
12 m from the device under test it is situated in an arbitrary
orientation, possibly lying on or behind absorbing material,
and finally, it reaches a feedthrough leading out of the
shielded room. In the MoM calculation a straight cable with
a certain length has to be chosen. Only a length of less than
1 m can be simulated with a reasonable use of resources and
time for a PC. For any length of cable, the termination of the
cable stub is inherently an open end, as the implementation
of lossy material is not possible in this EFIE code. Therefore,
the reflection coefficient at the cable end is < < s 1, causing
resonances and standing waves between the cable end and
the sleeve. Such a resonance can result in an especially high
radiation. Thus, a good position of the sleeve might be
rejected because that specific length of the cable behind the
sleeve happens to support a resonance in the surface current
on the cable. A solution to this problem is described in the
next section.
GENETIC ALGORITHM

We want to formulate a design problem for the position,

length, and diameter of the sleeve, which provide the best
match between the perturbed radiation pattern calculated
with an attached cable and the reference field calculated
without any cable attached to the handset. The optimization
of the sleeve should also consider its desired independence of
the RF cable length and location, the desired effective bandwidth, or the kind of handset antenna used,, which leaves us
with a large number of variables. Therefore, it seems reasonable to use a global optimization method like a genetic
algorithm see, e.g., w7, 8x..
Genetic algorithms are based on Charles Darwins theory
of natural selection. The basic building blocks in a genetic
algorithm are genes. A gene is one bit in a binary chromosome encoding the parameters. In our case, each chromosome encodes the position, length, and diameter of the sleeve.
Each chromosome is associated and ranked with a value from
the objective functional. In each iteration, a new set of
chromosomes is created that basically consist of parts of the
highest ranked chromosomes, with a few occasionally mutated genes. After ranking, selection, crossover, and mutation
of the chromosomes in each iteration, the global minimum is
obtained after a number of iterations w79x.
As the first step in the optimization algorithm, a suitable
objective functional is introduced to be able to rank the
chromosomes. The MoM code is used to calculate the radiation pattern in one significant E-plane cut. The cut through
the xz-plane is sufficient as it includes the whole dynamic
range of the handset radiation. The objective functional is
the mean-square error between the calculated perturbed

MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 25, No. 3, May 5 2000

195

pattern and the unperturbed reference pattern. The pattern

is calculated every 3. When evaluating the far-field radiation
pattern of the modeled handset, the parameter sets are
ranked according to their objective functional values. The
genetic algorithm finds the parameter set with the minimal
objective functional value.
The genetic algorithm requires a discretization of the
search space, i.e., of the position and dimensions of the
sleeve. Therefore, an exact absolute minimum is generally
not obtained, unless a very fine discretization is chosen. The
present optimization case does not require a very fine discretization, as the changes in the radiation pattern are rather
smooth. 2 mm was chosen for the discretization step for the
length and location of the sleeve, and 1 mm for that of the
diameter. If a more accurate design for the parameters is
required, a gradient method can be employed to make a finer
search.
The above-mentioned effect of cable-length-related resonances and standing waves has been reduced by the following
method. Each evaluation of the objective functional includes
the solution of the MoM code for five cases instead of one,
i.e., the radiation patterns for five different cable lengths are
calculated: the original cable length of c 0 s 600 mm, and
four varied cable lengths c n s c 0 q n = r10, where n s
"1, "2. The corresponding five complex radiation patterns
are added up in the objective functional before averaging the
differences from the reference case. Therefore, possible
cable-length-dependent narrowband resonances do not affect
the results as much.
RESULTS

After the optimization of the sleeve design, the optimal

diameter of the sleeve is d s 13.8 mm, and the optimal
length of the sleeve l s 63.7 mm. The optimal distance of the
open side of the sleeve to the bottom of the chassis is
p s 28.1 mm. This means that the distance between the short
cut end of the sleeve and the chassis of the handset is
l q p s 91.8 mm, i.e., sigificantly more than r4 s 75 mm.,
which was the first guess used in w3x, where the proposed
dimensions were l 0 s 60 mm, d 0 s 16 mm, p 0 s 15. That
previous case results in a worse agreement with the reference
case see Fig. 3. as no optimization had been used.
CONCLUSIONS

During a radiation pattern measurement of a mobile handset,

one can observe perturbations in the far-field pattern of a
handset caused by parasitic currents on the surface of the
coaxial RF feed cable. A sleeve on the coaxial feed cable of a
mobile handset can reduce the perturbations. A genetic algorithm is used to find a sleeve design resulting in a minimum
perturbation of the far-field pattern of a mobile handset.
To calculate the far-field pattern, the electric-field integralequation EFIE. method is used in each iteration step of the
genetic algorithm. The three parameters length l, diameter d,
and location p of the sleeve have been optimized for an
overall cable length of c 0 s 600 mm. The results show that
the genetic algorithm is very suitable for optimizing this
design problem. The method can be generalized to include
several more design parameters, such as the required bandwidth of the handset antenna, and the location and size of a
possible second sleeve along the first 2030 cm of the feed
cable. The final aim is to build a short piece of modified
coaxial cable, which can be attached to any handset to

196

Figure 3 The complex sum of the radiation pattern with five

slightly different cable lengths is indicated by the solid line, the
reference case without any cable attached is denoted by dots, and the
far-field pattern with a 600 mm long cable attached and no sleeve is
marked with circles. Additionally, the radiation pattern with the
earlier proposed dimensions w3x is shown by triangles

increase the accuracy of far-field measurements in a certain

frequency range.
ACKNOWLEDGMENT

The authors thank Dr. Jiasu Cao for making the EFIE core
code available in Matlab format.
REFERENCES
1. J. DeMarinis, The antenna cable as a source of error in EMI
measurements, Proc 1988 IEEE Int Symp EMC, Seattle, WA,
1988, pp. 914.
2. S. Saario, D.V. Thiel, J.W. Lu, and S.G. OKeefe, An assessment
of cable radiation effects on mobile communications antenna
measurements, Proc 1997 IEEE Int Symp Antennas Propagat,
Montreal, Canada, 1997, pp. 550553.
3. C. Icheln, J. Ollikainen, and P. Vainikainen, Reducing the influence of feed cables on small antenna measurements, Electronics
Lett 35 1999., 12121214.
4. S.M. Rao, D.R. Wilton, and A.W. Glisson, Electromagnetic scattering by surfaces of arbitrary shape, IEEE Trans Antennas
Propagat, AP30, 1982., 409418.
5. S.M. Rao, Electromagnetic scattering and radiation of arbitrarily
shaped surfaces by triangular patch modelling, Ph.D. dissertation,
Univ. Mississippi, Aug. 1980.
6. S. He, J. Cao, Scattering from a bi-isotropic object of arbitrary
shape, J Electromag Waves Appl 12 1998., 15471574.
7. D.E. Goldberg, Genetic algorithms in search, optimization and
machine learning, Addison-Wesley, Reading, MA, 1989.
8. D.S. Weile, and E. Michielssen, Genetic algorithms optimization
applied to electromagnetics: A review, IEEE Trans Antennas
Propagat 45 1997., 343353.
9. R.L. Haupt, An introduction to genetic algorithms for electromagnetics, IEEE Antennas and Propagat Mag 37 1995., 715.
2000 John Wiley & Sons, Inc.

MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 25, No. 3, May 5 2000