252 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO.

1, FEBRUARY 2005

An Improved Method to Determine the

Antenna Factor
Wout Joseph and Luc Martens, Member, IEEE

Abstract—In this paper, we present an improved method to de- [5] and [6]. These methods make use of tabulated values of
termine the antenna factor of three antennas. Instead of using a the maximum field strength for frequencies below 1000 MHz.
reflecting ground plane we use absorbers. Destructive interference An advantage of our method is that it is applicable for both E-
between the direct beam and the residual reflected beam from the
absorbers is avoided by splitting the measured frequency range and H-field probes. For the calibration of loop antennas, two
in bands and changing the distance between the two antennas de- methods are described in [7]. The first method is based on cal-
pending on the frequency band. Furthermore, this method is ap- culation of the loop impedances. The second method is by gen-
plicable for both E- and H-field probes. Our method has also the erating a well-defined standard magnetic field. The first method
advantage of being low-cost: The method does not need to be per- cannot be used because the geometric shape of the split-shield
formed in an anechoic chamber to obtain high accuracy. To take the
residual reflections of the environment into account, we perform a loop probes is not simple. For the second method, the gener-
de-embedding procedure. We have developed two de-embedding ation of a well-defined standard magnetic field is necessary.
methods. Our method does not need such a standard magnetic field. The
Index Terms—Antenna factor, calibration, dipole antenna, split- method of Glimm et al. [8] is only applicable for directional an-
shield loop antenna. tennas and not for loop antennas.
Mostly, measurements to determine the antenna factor are
done in an anechoic chamber. However, this is expensive. If
I. INTRODUCTION
one wants to optimize the gain and the sensitivity of the mea-

D URING THE last decade, the use of mobile telephones

[especially global system for mobile communications
(GSM) phones] has increased enormously. At the same time,
surement antennas, several characterization cycles are needed.
Therefore, the use of a low-cost method is advised.
In this paper, we present a method based on S-parameter mea-
the question whether the fields of the base stations are harmful, surements. The method does not make use of tabulated values
becomes more and more important. So there is a need to as in [5] and [6]. The frequency range that we investigate is
measure electromagnetic fields at frequencies around 900 from 600 to 2000 MHz. This range contains the GSM900 and
(GSM900) and 1800 (GSM1800) MHz. Electromagnetic fields GSM1800 frequencies. With our method, the antenna calibra-
are also measured around sources for use in electromagnetic tion must not be performed in an anechoic chamber but is done
compatibility studies. Accurate measuring of electromagnetic in an indoor open-site surrounded by absorbers to minimize in-
fields can only be done if the electromagnetic-field probes are terference. This makes it a low-cost method delivering accurate
calibrated carefully. Antenna calibration involves the determi- results.
nation of the antenna factor (AF). The antenna factor is defined A limitation to our method is that it is only applicable for pairs
as follows: of omnidirectional antennas and when the antennas are carefully
aligned. By using appropriate stands, this alignment can be done
AF (1) properly.
In Section II, our method is explained. The results are
where is the incident electric field at the antenna to be cal- described in Section III. The conclusions are presented in
ibrated and is the voltage measured at the terminals of the Section IV.
antenna to be calibrated.
The principle of using three antennas for the calibration is in-
II. METHOD
troduced by Smith [1], [2]. In this method, the reflected beam
is taken into account theoretically. Reflections can largely dis- A. Configuration
turb the measurements. The theoretical site attenuation values The calibration setup is shown in Fig. 1. We use a well-ab-
in the ANSI C63.4 [3] are based on [1] and [2]. Antenna mea- sorbing material on the ground and name this an absorbing
surements on different types of open-area test sites (OATS) with ground plane.
different conducting planes are reported in [4]. In this paper, an We make sure that the far-field conditions are fulfilled
absorbing ground plane is used to reduce the reflections. Sev-
eral improvements to the method of Smith have been published
(2)

Manuscript received April 24, 2002; revised March 15, 2004. with the distance from the antenna, the maximum di-
The authors are with the Department of Information Technology, Ghent Uni-
versity, B-9000 Ghent, Belgium (e-mail: wout.joseph@intecUGent.be). mension of the antenna, and the wavelength at the operation
Digital Object Identifier 10.1109/TIM.2004.838116 frequency.
0018-9456/$20.00 © 2005 IEEE
JOSEPH AND MARTENS: AN IMPROVED METHOD TO DETERMINE THE ANTENNA FACTOR 253

Fig. 1. Measurement setup. Fig. 2. Geometry of the split-shield loop antenna with loading resistance of 0
or 50
.
Measurements are made with a network analyzer (Rohde and
Schwarz ZVR). The network analyzer is calibrated with a SOLT
calibration, using the standards short, open, load, and thru. By
connecting these standards with the coaxial connectors of the
measurement planes and using the SOLT-calibration formulas,
the reference planes of the measurements are moved to the in-
puts of the antennas as indicated on Fig. 1. The antenna factors
are determined between 600 and 2000 MHz.
The antennas used in the method are dipoles with lengths
15 cm ( at 900 MHz) and 7.5 cm ( at 1800 MHz), a
conical dipole antenna and an in-house made split-shield loop
antenna, one terminated in a 50 load and the other one in
a 0 load, shown in Fig. 2. The split-shield loop antenna is
designed to reject the contribution of the electric field to the
magnetic-field measurement [9]. Both split-shield loop antennas
have a diameter of 3.4 cm.

B. Theory of the Method

Fig. 3. (a) Antenna setup with indication of powers and (b) its S-parameter
We start from the Friis formula [10], [11] equivalent.

(3) Combining (3) and (4), we obtain

where is the received power delivered to the receiver load, (5)

is the available input power of the transmitting antenna,
is the realized (apparent) gain of the receiving antenna, and When we measure with three different antennas using identical
is the realized gain of the transmitting antenna. and con- setups characterized by , shown in Fig. 1, we obtain
tain the mismatch of the receiving and the transmitting antenna. the following equations:
Figs. 3(a) and (b) clarify the notations used in (3).
PLF stands for polarization loss factor. It is defined as (6)
where is the unit vector of the incident
field, is the unit vector of the polarization of the field of the
receiving antenna, and the angle between the two unit vectors. (7)
is the distance between both antennas.
If we measure the -parameter (with respect to 50 ) be- (8)
tween the calibration planes of the antennas, the attenuation is
with the realized gains of the three antennas. We
(4) position the antennas in such a way that the factors are
254 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005

approximated to one. Solving (6), (7), and (8), we obtain the

realized gains.
Using the same notations as in formula (3), the power density
at distance in the far field of the source antenna can be
derived

(9)

is the realized gain of the transmitting antenna, is the elec-

tric field (V/m). Considering (1) and combining formula (9) with
formula (3) also equal to (50 measurement system),
the antenna factor [dB(m )] is derived
AF (10)
with the frequency in MHz.

C. Improvements
Although we use an absorbing ground plane, destructive in- Fig. 4. Determination of d= for three values of d (h = 0:60 m and
h = 0:45 m). The black spots indicate the frequency points where destructive
terference between the direct beam and the reflected beam could interference occurs.
still appear, resulting in unreliable results. In Smith [1], [2], this
problem is avoided by scanning the height of the receiving an-
tenna until the maximum output voltage is measured. In our
method, we split the considered frequency range in bands. In
each frequency band, the distance between both antennas is
changed after determining the path length difference between
both beams. For horizontal polarization and using the notations
of Fig. 1 we get
(11)
If with the wavelength and , then
both beams will interfere destructively. Before measuring, we
choose the height of the transmitting antenna and the height
of the receiving antenna . This choice is made such that the
antennas are polarization matched (PLF is close to 1) and that
the antennas are positioned high enough above the absorbing
ground plane. This is because of the fact that the more orthog-
onal the beam hits the absorbers on the ground the better it will
be absorbed. We then determine R, the distance between both
antennas, in a frequency band using Fig. 4 and formula (11). To
produce this figure, the reflection coefficient of the ground was Fig. 5. De-embedding by placing an absorber in between the two antennas.
approximated to one in amplitude and 180 in phase. The points
where destructive interference occurs are indicated with black We take the reflections into account by subtracting
spots. The frequencies in the neighborhood of these points of from , where represents the actual measurement
destructive interference should be avoided because at these nulls
it will be impossible to solve (6)–(8) and the steep gradient of (12)
the field in the region of a null can result in large measurement
errors due to small errors in antenna positioning [1], [2]. with .
We have not yet taken into account residual reflections. To In the second de-embedding method, we first select the fre-
this end, we perform a de-embedding step. We describe two quency range where the antenna factors must be determined. An
methods of de-embedding the reflected beam: The first one is appropriate configuration of the antennas is selected to avoid
performing an additional measurement by placing an absorber destructive interference in the desired frequency range. To ob-
in between the two antennas. The second method uses the in- tain enough resolution in the time domain to distinguish the di-
verse fast Fourier transform (FFT) to obtain the time-domain rect and reflected beam, a measurement is performed in a much
signals and we then apply time-domain gating [12] and [13]. larger frequency range. Using a tenth-order Butterworth band-
The de-embedding step by placing an absorber in between the pass filter, the -parameters in the desired frequency range
two antennas is shown in Fig. 5. The absorption eliminates the are obtained. We then take the inverse FFT of the -param-
direct beam. This results in the measurement of all the reflec- eters to obtain . Next, we apply a time-domain gating tech-
tions by the surrounding environment, represented by . nique [12], [13] to eliminate the residual reflection: The latest
JOSEPH AND MARTENS: AN IMPROVED METHOD TO DETERMINE THE ANTENNA FACTOR 255

Fig. 7. Comparison of the antenna factors of the conical dipole antenna,

Fig. 6. Application of time-domain gating by filtering the reflected beam respectively, obtained by the Austrian Research Center in Seibersdorf and
arriving 2 ns later than the direct beam. obtained with our method without de-embedding.

arriving time-components (the reflected beam will arrive later

than the direct beam) of are suppressed by a tenth-order
Butterworth digital bandpass filter. This type of filter is selected
because of its flat passband and the absence of side lobes. The
difference of the propagation time between the direct and the
reflected beam can be determined using formula (11). After fil-
tering, we obtain . Finally we take the FFT to obtain
and use formula (13) to relate to the gains

(13)

with .
To further explain the de-embedding method based on the
inverse FFT we show as an example the determination of the
antenna factor of the conical dipole antenna in the frequency
range 1100–1400 MHz. We used as heights m and
m. Using Fig. 4 and formula (11), the distance be- Fig. 8. Comparison of the antenna factors of the conical dipole antenna,
respectively, obtained by the Austrian Research Center in Seibersdorf and
tween both antennas is chosen in such a way that no destructive obtained with our method with de-embedding by placing an absorber between
disturbance occurs in the considered frequency range. The dis- both antennas.
tance between both antennas is 0.60 m. For this configuration
the difference of the propagation time between the direct and the
calibration from 1400 to 2000 MHz. In between those frequency
reflected beam is 2 ns. To obtain a resolution smaller than 2 ns,
ranges, we can use both dipoles.
it is necessary to use for the FFT a 500 MHz.
The three measurements of each 801 points are performed from
300 kHz to 4 GHz, therefore, GHz is large enough. III. RESULTS
Fig. 6 shows how we filtered the reflected beam in the time do- To check the accuracy of our method, we compare the
main. The measurement from Fig. 6 is performed with the 15 cm antenna factor of the conical dipole antenna, determined with
dipole and the conical dipole antenna using the configuration de- our method, with the data obtained from the Austrian Research
scribed above. The results of both de-embedding methods and a Center Seibersdorf where the antenna calibration is performed
comparison between both methods will be shown in Section III. in an anechoic chamber. For our method, we used as heights
Finally, we had to take into account that the 15 cm dipole m and m. We show the results obtained
at 900 MHz) antenna is almost insensitive for frequencies without de-embedding and obtained with the two presented
higher than 1600 MHz, while the 7.5 cm dipole ( at 1800 de-embedding methods in Figs. 7, 8, and 9. Table I lists the
MHz) antenna becomes insensitive for frequencies lower than mean and maximum deviation. As can be noted there is already
800 MHz. So it is advisable to use the 15 cm dipole for the improvement by placing an absorber: The maximum deviation
calibration from 600 to 1100 MHz and the 7.5 cm dipole for the is reduced by 1.5 dB. However, we obtain the best results with
256 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005

Fig. 9. Comparison of the antenna factors of the conical dipole antenna, Fig. 10. Comparison of the antenna factor of the split-shield loop probes using
respectively, obtained by the Austrian Research Center in Seibersdorf and both de-embedding methods.
obtained with our method with de-embedding by taking the inverse FFT.

TABLE I
COMPARISON OF THE ANTENNA FACTOR OF THE CONICAL DIPOLE ANTENNA
OBTAINED BY THE AUSTRIAN RESEARCH CENTER IN SEIBERSDORF WITH THE
RESULT OBTAINED WITH OUR METHOD WITH AND WITHOUT DE-EMBEDDING

the de-embedding by taking the inverse FFT: The deviation is

maximal 0.84 dB and the average deviation is only 0.39 dB.
We can conclude that performing a de-embedding step largely
improves the determination of the antenna factor. The Austrian
Research Center Seibersdorf specified the antenna factor with Fig. 11. Consistency check by comparison of the antenna factor of the conical
dipole antenna using our three-antenna method with different antennas.
an uncertainty of dB. So our result is lying within this
uncertainty interval.
Putting an absorber in between the two antennas will intro- walls, will not be eliminated using this method in contrary
duce diffracted waves. This is a reason why the results of this to the method using the inverse FFT.
method are worse than using the method with the inverse FFT. The antenna factors of the split-shield loop probes (used
The size of the absorber we used is 60 60 3 cm . To investi- for measuring the magnetic field) determined by means of our
gate the diffraction, we performed a worst-case finite difference method are shown in Fig. 10. For this figure, we used an ab-
time domain (FDTD) electromagnetic simulation with a per- sorbing ground plane and we performed the calibration for both
fectly conducting plate in between the two antennas and with a de-embedding methods. A disadvantage of the 0 loop antenna
separation of 70 cm between the antennas. At 900 MHz and with is the appearance of the resonance peak at 1000 MHz. The
the dimensions of the plate equal to the absorber, we obtained resonance peak is smaller using the de-embedding by taking
that the electric fields at the receiving antenna due to diffrac- the inverse FFT because of the filtering in the time domain
tion are six times lower than the direct wave when no absorber what comes down to a convolution and, thus, a smoothening in
is present. Using an absorber of dimensions 120 120 3 cm the frequency domain.
( at 900 MHz) delivers diffracted fields that To check the consistency of our method, we performed our
are more than 140 times lower. The method with an absorber in method with different antennas and we compared the antenna
between the two antennas will deliver better results when using factor of the conical dipole antenna. We used the de-embedding
much larger absorbers, but this will be more expensive. The by taking the inverse FFT. For the first measurement, the two
dimensions of the absorber are determined by the lowest fre- other antennas were the two dipoles and the 0 split-shield
quency under consideration. Residual reflections from ceiling, loop antenna. For the second measurement, we used a 50
JOSEPH AND MARTENS: AN IMPROVED METHOD TO DETERMINE THE ANTENNA FACTOR 257

split-shield loop antenna in combination with the dipoles. In [7] A. Aykan, “Calibration of circular loop antennas,” IEEE Trans. Instrum.
Fig. 11, we compare the obtained antenna factor of the conical Meas., vol. 47, no. 2, pp. 446–452, Apr. 1998.
[8] J. Glimm, R. Harms, K. Münter, M. Spitzer, and R. Pape, “A single-
dipole antenna. The deviation of the obtained antenna factors of antenna method for traceable antenna gain measurement,” IEEE Trans.
the conical dipole antenna is very small, despite the fact that the Electromag. Compat., vol. 41, no. 4, pp. 436–439, Nov. 1999.
antenna factor of the 0 split-shield loop antenna is totally dif- [9] C. Carobbi, L. Millanta, and L. Chiosi, “The high-frequency behavior of
the shield in the magnetic-field probes,” in IEEE Intl. Symp. Electromag.
ferent to the one of the 50 split-shield loop antenna in the fre- Compat., vol. 1, Washington, DC, Aug. 2000, pp. 35–40.
quency range of 900–1200 MHz. The average difference of both [10] C. A. Balanis, Antenna Theory Analysis and Design. New York: Wiley,
antenna factors for the whole frequency range is only 0.15 dB; 1982, pp. 100–163.
[11] S. A. Schelkunoff and H. T. Friis, Antennas, Theory, and Prac-
the maximum deviation is 0.49 dB. This shows that our method tice. New York: Wiley, 1952.
is consistent. [12] E. K. Miller, Time-Domain Measurements in Electromagnetics. New
York: Van Nostrand, 1986, pp. 416–455.
[13] R. Yagüe, A. Ibars, and L. Martinez, “Analysis and reduction of the dis-
IV. CONCLUSION tortions induced by time-domain filtering techniques in network ana-
lyzers,” IEEE Trans. Instrum. Meas., vol. 47, no. 4, pp. 930–934, Aug.
We have presented an improved method to determine the an- 1998.
tenna factor. From the results shown in this paper, we can con-
clude that our calibration method is accurate and consistent. By
using antennas with totally different antenna factors, the mean Wout Joseph was born in Ostend, Belgium, on
deviation is only 0.15 dB and the maximum deviation is only October 21, 1977. He received the M.Sc. degree in
0.49 dB for the antenna factor of a conical dipole antenna. The electrical engineering from Ghent University, Ghent,
Belgium, in 2000 and is currently pursuing the Ph.D.
use of an absorbing ground plane and performing a de-embed- degree in the Department of Information Technology
ding step eliminate undesired reflections. Furthermore, the cal- (INTEC), Ghent University.
ibration is easy to perform with a network analyzer, and this Since September 2000, he has been involved in
research on measuring and modeling of electro-
method does not need to be performed in an anechoic chamber magnetic fields around base stations for mobile
to be accurate, resulting in a low-cost method. communications related to the health effects of
the exposure to electromagnetic radiation. He is
interested in electromagnetic field measurements, antennas, and calibration.
REFERENCES
[1] A. A. Smith JR., “Standard-site method for determining antenna fac-
tors,” IEEE Trans. Electromag. Compat., vol. 24, no. 3, pp. 316–322,
Aug. 1982. Luc Martens (M’92) was born in Gent, Belgium, on
[2] A. A. Smith JR, F. German, and B. Pate, “Calculation of site attenuation May 14, 1963. He received the M.Sc. degree in elec-
from antenna factors,” IEEE Trans. Electromag. Compat., vol. 24, no. 3, trical engineering from Ghent University, Belgium,
pp. 301–316, Aug. 1982. in July 1986.
[3] American National Standard for Electromagnetic Compatibility—Ra- From September 1986 to December 1990, he was a
diated Emission Measurements in Electromagnetic Interference (EMI) Research Assistant at the Department of Information
Control—Calibration of Antennas (9 kHz to 40 GHz). ANSI C63.4- Technology (INTEC), Ghent University. During this
1998. period, his scientific work was focused on the phys-
[4] T. Moroika and K. Komiyama, “Measurements of antenna character- ical aspects of hyperthermic cancer therapy. His re-
istics above different conducting planes,” IEEE Trans. Instrum. Meas., search work dealt with electromagnetic and thermal
vol. 50, no. 2, pp. 393–396, Apr. 2001. modeling and with the development of measurement
[5] R. McConnell, “A method of determining free space antenna factor on systems for that application. This work led to a Ph.D. degree in December 1990.
an open area test site,” in IEEE Intl. Symp. Electromag. Compat., vol. 2, Since January 1991, he has been a member of the permanent staff of the In-
Washington, DC, Aug. 2000, pp. 499–504. teruniversity MicroElectronics Centre (IMEC), Ghent, and is responsible for the
[6] Z. Chen and M. Foegelle, “An improved method for determining nor- research on experimental characterization of the physical layer of telecommu-
malized site attenuation using log periodic dipole arrays,” in IEEE Intl. nication systems at INTEC. His group also studies topics related to the health
Symp. Electromag. Compat., vol. 2, Washington, DC, Aug. 2000, pp. effects of wireless communication devices. Since April 1993, he has been a Pro-
511–516. fessor in electrical applications of electromagnetism at Ghent University.