IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 46, NO.

2, APRIL 1997 515

A Broadband Free-Space Dielectric Properties

Measurement System at Millimeter Wavelengths
Gerhard L. Friedsam and Erwin M. Biebl, Senior Member, IEEE

Abstract— A free-space system operating in the 75 GHz to

95 GHz range for measurement of the complex permittivity
of matter is presented. The complex transmission coefficient of
a parallel polarized quasiplane wave through a planar sample
is determined for different incidence angles. The quasiplane
wave is realized by a collimated Gaussian beam. The complex
permittivity is subsequently calculated from the measured data
by applying a nonlinear least-squares method. Errors induced
by multiple reflections between the transmitting and receiving Fig. 1. Quasioptical setup for measuring the complex transmission coeffi-
horn antennas are removed by a time–domain gating technique. cient.
The design of the quasioptical measurement setup is based on
Gaussian beam optics. A dielectric lens collimates the beam of the
transmitting horn antenna for minimization of diffraction effects
at the edges of the sample. In order to verify the Gaussian beam
and thin lens approximations at millimeter wavelengths, the far-
field pattern of the horn antenna and the dielectric lens system
has been measured. The performance of the measurement setup
has been tested using glass samples.
Index Terms—Complex permittivity, dielectric properties, free-
space method, free-space transmission measurements, Gaussian
beam, millimeter wave frequencies, quasioptical method.

I. INTRODUCTION

I N THE millimeter wave range the complex permittivity

of matter can be determined by using either dielectric
waveguide, cavity resonators, open resonators, or free-space
techniques. Free-space methods are excellently suited for the
broadband measurement of the complex permittivity and per-
meability [1]–[3]. This technique is based on the measurement
of the transmission coefficient and/or the reflection coefficient
of planar samples for different polarization states, angles
of incidence, frequencies, and thicknesses of the sample.
Free-space methods are nondestructive and can be used for
measuring a wide range of materials including gases, liquids,
and solids.
The major disadvantages of the free-space methods are
diffraction effects around the sample edges and multiple re-
flections between the horn antennas. The diffraction effects at
the edges of the sample can be minimized by using horn-lens-
combinations as transmitters and receivers. With broadband
measurement systems it is possible to remove the multiple
reflections by means of a time domain gating technique.
In this paper, we present a broadband free-space measure-
ment system to determine the complex permittivity in the
Manuscript received June 20, 1996; revised October 1, 1996. This work
was supported by the Hanns-Seidel-Stiftung e.V. with assistance of the
Bundesministerium für Bildung und Wissenschaft.
The authors are with Lehrstuhl für Hochfrequenztechnik, Technische Uni-
versität München, 80333 München, Germany.
Publisher Item Identifier S 0018-9456(97)01623-9.
0018–9456/97$10.00  1997 IEEE
Fig. 2. 3-dB beamwidth 1=2 of the horn-lens-combination over 75–95 GHz:
theory (——) and measurement (- - - -).
75 GHz to 95 GHz range by measuring the transmission coef-
ficient for different angles of incidence. The application of the
method is demonstrated using glass samples. The influence of
the measurement errors of the complex transmission coefficient
and of the sample thickness on the accuracy of the proposed
method is discussed.
II. MEASUREMENT SET-UP
The design of the quasioptical measurement setup is based
on Gaussian beam optics [4], [5]. Fig. 1 shows the measure-
ment setup. The free-space transmission path consists of two
corrugated horn antennas and a dielectric lens. The aperture
radius and flare angle of the corrugated horn antennas are
13.5 mm and 10.11 , respectively. The ratio of focal distance
( 210 mm) to diameter of the lens ( 131 mm) is
equal to 1.6. In order to confirm the Gaussian beam as well
as the thin lens approximations, the -plane radiation pattern
of the lens-horn combination was measured at 21 frequencies.
From the measured data the far-field 3 dB beamwidth of the
electrical field distribution was determined. Fig. 2 shows the
measured (solid line) and the theoretical values (broken line) of

Authorized licensed use limited to: IEEE Xplore. Downloaded on December 8, 2008 at 23:05 from IEEE Xplore. Restrictions apply.
516
the 3 dB beamwidth. The deviation between the theoretical and
the measured values is less then 0.5 . The 3 dB beamwidth
of the collimated beam is smaller than 1.7 over the whole
frequency range from 75 GHz to 95 GHz. By the collimated
Gaussian beam a plane wave is approximated. In order to
minimize diffraction effects the diameter of the sample is
larger than eight times the radius of the collimated Gaussian
beam and the sample is placed near the beam waist. The
radius of the collimated Gaussian beam at the position of
the sample holder is about 12 over the entire frequency
range. The sample holder is mounted on a precision motor
driven rotary stage. In order to avoid extraneous reflections
into the measurement path, the area around the transmission
path is covered with absorbing material. An automatic network
analyzer HP8510 is used to analyze the received signals.
The measurement set-up and acquisition of measured data
is controlled by a computer program. With this set-up the
complex transmission coefficient of planar samples can be
measured at different angles of incidence and for both parallel
and perpendicular polarization in the 75 GHz to 95 GHz range.
III. THEORY
The transmission coefficient of a parallel polarized plane
electromagnetic wave through an infinitely wide slab with the
complex material parameters and
and thickness for an incidence angle is given by [6]
where
and
is the reflection coefficient for parallel polarization of a plane
electromagnetic wave incident on a dielectric boundary.
the free-space wavelength.
The measured transmission coefficient is normalized with
respect to the amplitude and phase readings obtained by the
calibration procedure. The resulting phase shift must be taken
into account by
Finally, we obtain the theoretical transmission coefficient
for the measurement system
The complex permittivity cannot be expressed
explicitly in terms of , , and . It is determined by fitting
the theoretical model (5) to the measured values of magnitude
and phase of the transmission coefficient using a nonlinear
Authorized licensed use limited to: IEEE Xplore. Downloaded on December 8, 2008 at 23:05 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 46, NO. 2, APRIL 1997
least-squares method. The error function is defined as
(6)
where , and , are magnitude and phase of
the theoretical and measured transmission coefficient for an
incidence angle , respectively. The terms and are
the residuals for each angle. By investigating the distribution
of the residuals the consistency of the measurement data can
be controlled.
Now a similar nonlinear least-squares estimate is calculated
by using another error function
(7)
which takes only the measured magnitudes into account. By
comparing the complex permittivity values obtained by the
two nonlinear least-squares estimates, the consistency of the
results can be checked. Particularly systematic errors of the
measurement set-up can be identified.
Uncertainties of the complex permittivity measurement de-
pend on the thickness of the sample. From computer sim-
ulations using (1)–(6) it can be shown, that the minimum
measurement error corresponds to and the maxi-
(1) mum measurement error to , respectively,
where is the wavelength in the sample and
is the number of half wavelengths in the sample.
(2)
IV. RESULTS
The measurement system has been tested using a glass
(3) sample with a thickness of mm. For
calibration the sample is removed from the transmission path.
This calibration procedure eliminates only the errors due to
the frequency response. Using a time domain gating technique
is the influence of multiple reflections between the transmitting
and receiving horns and between the sample and the receiving
horn can be minimized. Fig. 3 shows the measured complex
transmission coefficient for the incidence angles 10 , 20 , 30 ,
and 40 . The uncertainty in the magnitude and in the phase is
less than 0.025 dB and 0.6 , respectively, due to residual
load and source mismatch.
(4) The complex transmission coefficient has been measured at
122 different angles of incidence for the parallel polarization
in the 75 GHz to 95 GHz range at 401 frequency points. In
order to minimize the influence of the wave reflected from the
sample and the diffraction effects at the edges of the sample
holder on the measured data, the measurements have been
(5) limited to angles of incidence within the range 10
40 . Fig. 4 shows the measured complex permittivity for a
glass sample. In order to identify systematic measurement
errors, we compare the values of the complex permittivity
calculated from the magnitude and phase information of the
measured data (solid line) with those values calculated only
FRIEDSAM AND BIEBL: BROADBAND FREE-SPACE DIELECTRIC PROPERTIES
(a)
(b)
Fig. 3. Measured transmission coefficient over 75 GHz to 95 GHz: (a)
magnitude and (b) phase.
from the magnitude information (broken line). Computer sim-
ulations showed that the disagreement between these two
methods is due to systematic errors in the magnitude of
the transmission coefficient: Assuming a complex permittivity
of we predicted amplitude and phase
of the transmission coefficient for various frequencies and
various angles of incidence. Subsequently we increased all
values of the amplitude by 0.005, i.e., we simulated a small
systematic error of the amplitude measurement. From this set
of amplitude and phase values we calculated the complex
permittivity and obtained curves similar to Fig. 4. Within a
small frequency range significant deviations from the correct
value of the permittivity are observed (see peaks in Fig. 4),
if the permittivity is calculated only from the magnitude.
If we use both magnitude and phase, much more accurate
permittivity values are obtained.
The nonlinear least-squares estimate gives the mean value
of the complex permittivity for all angles of incidence. In
order to obtain an uncertainty we calculate the complex
permittivity values from the measurement for each single angle
of incidence. From this set of complex permittivity values
the standard error of estimate is determined. The accuracy
of the proposed method depends on the sample thickness. In
the 90 GHz to 94 GHz range the standard error is less 0.06
Authorized licensed use limited to: IEEE Xplore. Downloaded on December 8, 2008 at 23:05 from IEEE Xplore. Restrictions apply.
517
(a)
(b)
Fig. 4. Calculated permittivity for glass over 75 GHz to 95 GHz, using the
measured magnitude and phase (—— ) or only the magnitude (- - - -): (a)
real part and (b) imaginary part.
TABLE I
MEASURED PERMITTIVITY AT 77, 80, 85, 91, 92, 93, AND 94 GHZ
and 0.3% for the real part and imaginary part of the complex
permittivity, respectively. The sample thickness normalized to
the wavelength in the dielectric is about . In the 75
GHz to 80 GHz range the standard error is less than 0.2
and 2.3% for the real part and imaginary part of the complex
permittivity, respectively. The sample thickness normalized to
the wavelength in the dielectric is about . Table I gives
the measured complex permittivity at different frequencies.
Now, the residuals between the measured transmission co-
efficient and the calculated transmission coefficient according
to (6) are considered. The average residuals are less than 0.001
in the magnitude and 0.250 in the phase over the whole
frequency range. The maximum residuals are 0.003 in the
518 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 46, NO. 2, APRIL 1997
magnitude and 0.6 in the phase. The measured transmission
coefficient is consistent with the theoretical model over the
whole frequency range.
V. CONCLUSION
By using the proposed method the complex permittivity was
measured with high accuracy in the millimeter wave range.
The sample preparation is very easy and nondestructive. The
consistency of the measurement data is tested by varying
the error function and by comparison of the residuals. For
a glass sample with a thickness of an uncertainty less
than 0.06 and 0.3% for the real part and imaginary part
of the complex permittivity has been achieved, respectively.
By use of an optimized set-up, time gating and calibration
techniques we minimized the systematic uncertainties to a
great extent. Thus, we estimate that the total uncertainty of
our measurement technique is less than 0.25 and 4% for
the real part and imaginary part of the complex permittivity,
respectively, in the whole frequency range from 75 GHz to
95 GHz.
REFERENCES
[1] P. K. Kadaba, “Simultaneous measurement of complex permittivity and
permeability in the millimeter region by a frequency-domain technique,”
IEEE Trans. Instrum. Meas., vol. IM-33, pp. 336–340, Dec. 1984.
[2] F. I. Shimabukuro, S. Lazar, M. R. Chernick, and H .B. Dyson,
“A quasioptical method for measuring the complex permittivity of
materials,” IEEE Trans. Microwave Theory Tech., vol. MTT-32, pp.
659–665, July 1984.
[3] R. G. Nitsche and E. M. Biebl, “A free-space technique for measuring
the complex permittivity and permeability in the millimeter wave range,”
in 1994 IEEE MTT-S Microwave Symp. Dig., pp. 1465–1468.
[4] P. F. Goldsmith, “Quasioptical techniques at millimeter and submil-
limeter wavelengths,” in Infrared and Millimeter Waves. New York:
Academic, 1982, vol. 6, pp. 277–343.
Authorized licensed use limited to: IEEE Xplore. Downloaded on December 8, 2008 at 23:05 from IEEE Xplore. Restrictions apply.
[5] D. H. Martin and J. W. Bowen, “Long-wave optics,” Trans. Microwave
Theory Tech., vol. MTT-41, pp. 1676–1696, Oct. 1993.
[6] E. Hecht, Optik. Bonn/München, Germany: Addison-Wesley, 1989.
Gerhard L. Friedsam was born in Freyung,
Germany, in 1968. He received the Dipl.-Phys.
degree in 1994 from the Technical University,
Munich, Germany. He is currently pursuing the
Dr.-Ing. degree.
He is with the Lehrstuhl für Hochfrequenztech-
nik, Technical University, Munich. His research
interests include field-theoretical analysis of planar
antennas and quasioptical measurement techniques.
Mr. Friedsam has been awarded a Graduate
Scholarship of the Hanns-Seidel-Stiftung e.V.,
Germany, for the academic years 1994–1996. He was co-recipient of the ITG
award in 1996. He is a member of the Informationstechnische Gesellschaft
(ITG) in the Verband Deutscher Elektrotechniker (VDE).
Erwin M. Biebl (S’88–M’91–SM’96) was born in
Munich, Germany, in 1959. He received the Dipl.-
Ing., Dr.-Ing., and habilitation degrees in 1986,
1990, and 1993, respectively, all from the Technical
University, Munich, Germany.
In 1986, he joined Rohde & Schwarz, Munich,
where he was involved in the development of mo-
bile radio communication test sets. Since 1988, he
has been with the Institut für Hochfrequenztechnik,
Technical University, Munich, where he is now
University Lecturer and Head of the millimeterwave
devices group. He has been engaged in research on optical communications
and integrated optics. His current interests include field-theoretical analysis
of planar resonators and antennas, quasioptical measurement techniques, and
design and characterization of integrated millimeterwave devices.
Dr. Biebl was the recipient of the Dr. Georg Spinner award in 1991. In
1996, he was co-recipient of the ITG award. He is a member of the Informa-
tionstechnische Gesellschaft (ITG) in the Verband Deutscher Elektrotechniker
(VDE), Germany.