PROCEEDINGS OF THE IEEE, VOL. 62, NO.

1, JANUARY 1974 33

--
Automatic Measurement of Complex Dielectric Constant
andPermeability a t Microwave Frequencies
WILLIAM B. WEIR, =mER, IEEE

Abstractwith the advent of the computer and automatic test CALIBRATIONREFERENCE

PLANES
equipment, new techniques for measuring complex dielectric constant
( a ) andpermeability Q can be considered. Such atechnique is !/
I I

described where a system is employed that automaticallymeasures 6-n-

the complex reflection andtransmission coe5cients that result when
a sampleof material is inserted in waveguide or a TEM transmission
line. Measurement results of c and @ for two common materials are
presented.
INTRODUCTION
I

T
I
H E M E A S U R E M E N T of complex dielectric constant I
I
I

I I I

and complex permeability is required not only for sci- REFLECTION i 2 TRANSMISSION
PORT (A) PORT IBI
entific but alsoforindustrialapplications.Forex-
ample, areas inwhich knowledge of the properties of materials Fig. 1. Transmission line section Containing dielectric material.
a t microwave frequencies (as described bye and p ) are applica-
tions of microwave heating, biological effects of microwaves, sample holder. This is to insure that the higher order evanes-
and nondestructive testing.' centmodesduetothecoaxial-to-waveguideadaptersare
Numerousmeasurementmethodssuitablefordifferent significantly attenuated prior to reaching the sample under
ranges of the numerical values of e and p have been given in test.
the books edited by Von Hippel [2], [3] and in publications of Under computer control, network analyzer system calibra-
the American Society for Testing and Materials. I t is possible,tion and measurement are obtained at the referenceplanes
however, to rapidly make measurements over the frequency indicated in Fig. 1. This is done over a number of predeter-
range from 100 MHz to 18 GHz with a computer-controlled mined frequencies a t which the complex values of e and p for
network analyzer such as the Hewlett-Packard Model 8540
the material are to be determined. The normalized scattering
series, and by means of appropriate data processing, to deter- parameters (Sij) of the transmission line section containing
mine the complex values of e and p for materials. the material are measured a t t h e calibration reference planes
Using themethoddescribedinthispaper,thecomplex at ports A and B and corrected for system errors included in
values of E and p are determined from measurements made the calibration data. The measured scattering parameters are
directlyinthefrequencydomain. A somewhatanalogous normalized to the characteristic impedance (20)of the trans-
method has been developed [7] where measurements are made mission line section. The reflection coefficient ( S I ) at the air-
in the time domainof the transient response to subnanosecond to-dielectricinterface,andthetransmission coefficient (SX)
pulses from a dielectric material. With the time-domain mea- through the material, are found at reference planes 1 and 2.
surement approach, a Fourier transformation is required to These coefficients are found directly from the measured scat-
determine E and p from the measuredtransientresponse. tering parameters after the appropriate phase corrections have
Furthermore, with this approach, the frequencies a t which e been applied to account for the shift in the reference planes
and p values are obtained are band-limited, dependingon the from ports A and B to the material interfaces (i.e., planes 1
time response of the pulse and its repetition frequency. Using and 2, Fig. 1).
thesystem describedinthispaper,discretefrequenciesin From the complex reflection and transmission coefficients,
less than 20-kHz steps may be selected anywhere within the thecomputerassociatedwiththenetworkanalyzerdeter-
entire 100-MHz to l8-GHz band. mines the real and imaginary parts of the dielectric constant
and permeability, the loss tangent, and the attenuation per
AUTOMATIC MEASUREMENT SYSTEM
unit length of material. In addition, data takena t several fre-
A computer-controlled network analyzer is used to mea- quencies are used to find the average group delay through the
sure the parameters of a network consisting of a section of sample. Average group delay, in turn, is used to automatically
transmission line containing the sample of material. The trans- resolve phase ambiguities t h a t result when the sample length
mission line section may either be waveguide or a TEM trans- of material is greater than a wavelength in the dielectric. The
missionline. The network is shown schematically in Fig. 1. complex e and p , loss tangent, and attenuation data are auto-
If coaxial-to-waveguideadaptersarerequired,additional maticallylistedby a teleprinter, or rapidlyplotted on an
lengths of waveguide areinsertedbetweenthemandthe X- Y recorder, or both.
DATAPROCESSING TECHNIQUES
Manuscript received April 17, 1973; revised July 9, 1973.
The author is with Stanford Research Institute, Menlo Park, Calif. The equations used to compute complex e and p from the
94025.
For an example of a system that cdn be used for nondestructive test-
measuredreflection (SI') andtransmission (&I) coefficients
ing see [I]. are presented below together with the equations that relatee
Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 03,2024 at 11:03:22 UTC from IEEE Xplore. Restrictions apply.
34 PROCEEDINGS OF THE IEEE, JANUARY 1974

and p to attenuation and loss tangent. As will be seen, an X, cutoff wavelength of the transmission
line
section
infinite number of roots exist in the solution for the equations (X,= a for a T E M line),
E and p , but the correct root is related to the length of the
sample in termsof wavelength within the material. The means and
of determining the correct solutionis shown.
Referring to Fig. 1, thepropagationfactor for awave
propagating through the material is defined as [4]
p = e-7t = e-(Q+jB)l Equation (8) has aninfinitenumber of rootssincethe
(1)
imaginary part of the logarithm of a complex q u a n t i t y [ P i n
where (8)] is equal to the angleof the complex value plus2 m , where
n is equal to the integer of (Z/&). Equation (8) is ambiguous
y propagationconstant,
because the phase of the propagation factorP does not change
cy attenuationconstant,
when the length of the material is increased by a multiple of
B phaseconstant. wavelength.However,thedelaythroughthematerial is
A time factor of e*t is not explicitly shown in (1).The phase strictly a function of the total length of the material and can
constant is equal to be used to resolve the ambiguity.
The phase ambiguity is resolved by finding a solution for
2r E and cc from which a value of group delay is computed that
@=-
X, corresponds to the value determined from measured data at
two or more frequencies. For this method to work, the discrete
where X, is the transmission line guide wavelength. Consistent frequency steps at which measurements are obtained must be
with the definition of the propagation factor [(l)],e and p are small enough so that the phase of the propagation factor (P)
defined in terms of their real and imaginary parts as follows: changes less than 360" from one measurement frequency to the
e = E d 0 = (e,' -je,")eo (3) next. With the use of an automatic measurement system as
described in this paper, discrete frequency steps, small enough
cc = /+PO = ( / + I - j P r 1 l ) c c 0 . (4) to meet this requirement, can easily be selected. The group
delay at each frequency may be computed for each solution
The reflection coefficient (I?) at the interface between the air- of E and p assuming that the changes in E and p are negligible
filled transmission line and dielectric-filled line when the ma- over very small increments of frequency:
terialsampleisinfiniteinlengthmay be foundfromthe
measuredreflection (Sll) andtransmission (Ssl) coefficients
for a sample of finite length ( 1 ) : z

where f is the frequency in hertz andrgnis the group delay in

secondsforthenthsolution of (8) and (9). The measured
where
group delay is determined from the slope of the phase of the

X =
S112 - SZ12 +1 (6)
propagation factor versus frequency:

S l l

The propagation factor P can, in turn, be foundfrom S11,

Stl, and I':
where 4 is the phase in radians of P. Accuracyin the de-

P =
sll+ szl- r (7)
termination of r, may be increasedbyapplyingnumerical

1 - (sll+ s21)r
differentiationtechniques [SI where the slope is computed
using d a t a for three or more frequencies. The correct root,
The complex dielectric constant and permeability can be de- n = k, is found when
termined from P and r: Tgk - 7 9 'v 0.
Once the correct values of e and have been found at each
of the measurement frequencies, theloss tangent and attenua-
tion per unit length may be determined. In general, for both
a n electrically and magneticallylossy material, a loss'tangent
Pr=
1 1 is defined bya
*(1- PI'/- --
6,"
tan 6 = -
where 6,'
X0 freespacewavelength, where

* In terms of the air-filled line characteristic impedance ZOand the

dielectric-filled line characteristic impedance Za, r = (Zd-Ze)/(Zd+Zo).
Note that S I I =when ~ I is infinite. 3 Note that for the case when k " = O , tan 6=+"/t'.

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 03,2024 at 11:03:22 UTC from IEEE Xplore. Restrictions apply.
WEIR: MEASUREMENT OF DIELECTRIC CONSTANT AND PERMEABILITY 35

MEASURED DATA POINTS '30 O V

~ IMAGINARY PART k t 1
. MEASURED DATA mlm
8 9 10 11 12 10
FREOUENCY - GHz
I 1 I 1 1
la1
8 8 10 11 12 12.4
FREQUENCY --%Hz
Id
- Jl,= 1.0
. MEASURED DATA POINTS
1.0 I 1 1
t/----q
REAL PART ( 4 1 I I 1

-r 0.75

0
8 9 10
FREOUENCY
Ibl

Fig. 2. Measured dielectric constant and permeability of Teflon in the

(b) Relative per-
X-band region. (a) Relative dielectric constant (6).
- GHz
11 12

0
8
a MEASURED DATA POINTS

,
9 10
, ,
11
I

12
4 12.4
meability k?). FREOUENCY - GHr
lbi

Fig. 4. Measured attenuation and loss tangent of Eccmorb SF-5.5 in

16 I 1 I the X-band region. (a) Attenuation (a).(b) Loss tangent [tan (a)1.

*-1:[-f:q ability b)is 1. T h e measured data shown in Fig. 2 correspond

closely to these known properties of Teflon.
4
Somenoticeablevariationsinthe 9.2- to9.8-GHzfre-
0 quency range are indicated in Fig. 2, however, from the aver-
8 11 9 10 12 12.4 age values of E and p . Furthermore, values (missing in Fig. 2)
FREOUENCY - GHz obtained at three frequencieswere erroneous. These errors are
(ai
attributed to the fact that the network analyzer signal source
was not frequency stabilized a t the time measurements were
3
obtained on theTeflonmaterial.'Without a frequency-
.MEASURED DATA POINTS stabilized signal source, the frequencies a t the time of mea-
surement do not exactly repeat those at the time of calibra-
'IMAGINARYPART (p;) tion, which produces errors in the phase of the measured scat-
0 I tering parameters. The accuracy of the values of complex E
8 9 10 11 12 12.4
and p depends significantly on the accuracy of the measured
FREOUENCY - GHz
I bJ scattering parameter phase data.
T h e Eccosorb SF-5.5 is an absorber type material which
Fig. 3. Measured dielectric constant and permeability of Eccosorb SF-
5.5 in theX-band region. (a) Relative dielectric constant(+). (b) Rela- resonates at 5.5 GHz. Thebasic composition of the material is
tive permeability &). silicone rubber presumably loaded with lossy magnetic parti-
cles. When backed with a metallic surface, the material is re-
ported to be effective in absorbing a normally incident wave
and
a t 5.5 GHz, although some absorption occurs at other fre-
6;' = /A;€;' + p;'€;. quencies and other angles of incidence as well. T h e require-
The attenuation in nepers per length is found from the equa- ment of a metallic surface backing suggests that the absorber
tion is a magnetically lossy material which absorbs the high cur-
rents existing along the metallic surface excited by the inci-
dent wave. That the material is magneticallylossy is con-
firmed by Fig. 3 which indicates a value of pr" greater than
zero over X band. Fig. 3 also indicates that the material is not
electrically lossy ( + " N O over the band) but does have a rela-
MEASUREDRESULTS
tivelyhighdielectricconstant ( ~ ~ ' E 1 3 . over
7 theband).
T h e results of measuring thecomplex e and p of Teflon and Furthermore, p7' is greater than 1.0, varying from about 1.9
Emerson and Cuming Eccosorb SF-5.5 are shown in Figs. 2 t o 1.4 over the band. Attenuation andloss tangent5 are shown
through 4. Measurements of bothmaterials were obtained in Fig. 4 for the Eccosorb material. Attenuation varies from
with a sample inserted in X-band (RG-52) waveguide and at about 24 to 4 0 dB/cm and loss tangent from about0.7 to 0.9
an ambient temperature of 22°C. T h e Teflon sample length over X band.
was 1 in; the Eccosorb sample length was 0.090 in.
Teflon has a relatively low E: (2.03 [ 6 ] )in the microwave
4 A frequency-stabilized signalsource was used, however, for the
region and has a very low loss at X band (tan 6 <0.0004 [2]). Eccosorb SF-5.5 measurements.
I t is a nonmagneticmaterialandhence its relativeperme- 5 Loas tangent (tan 6) =p,"/p', since E,"= 0.

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 03,2024 at 11:03:22 UTC from IEEE Xplore. Restrictions apply.
36
Complex E and p data have been presented for two com-
mercially available dielectric materials as examples of results
obtained using the automatic measurement system described
in this paper. This system, however, has also been found use-
ful in determining the dielectric properties of rock, soil, and
other nonstandard materials.
ACKNOWLEDGMENT
Theauthor wishes tothank Dr. L. Youngand L. A.
Robinson for many helpful suggestions and encouragements
during the development of the automatic measurement sys-
tem and related data processing techniques.
Location andRecognition of Discontinuities in Dielectric
Media UsingSynthetic
Absfracf-RF pulses are used in radar and sonar to detect and
locate targets in extended media. Short RF pulses (and impulses)
can be used to lind buried objects or voids by echo sounding, or can
be used to probe snow fields or the depths of the earth. Althopgh
similar to radar in principle, there are important differencesin these
applications that can lead to significant variations in the design ap-
proach.
A novel system is described, with a computer-programmed tun-
able RF source, that tunes rapidly through all the frequencies of a
nanosecond pulse spectrum,makes individual “CW” measurements,
stores them, andfinally computes asynthetic echo. Target signatures
can be “recognizedp9bycalculatingcorrelationfunctions.Experi-
mental results are presented.
I.INTRODUCTION
P ULSE SYSTEMS based on related technologies have
beenused for widelydifferentapplications. The two
main types of echo sounding systems are 1) radar and
sonar, where in general long R F pulses (‘long” in terms of R F
cycles) are used to locate moving objects in space hundredsor
thousands of pulse lengths away, and 2) time-domain reflec-
tometry ( T D R ) , where a step function or impulse is used in a
closed system to locate and diagnose a discontinuity.
I t is possible to replace the pulsed transmitter by a swept
frequency generator as in FM radar or in pulse compression
systems.
Fromtimetotimeunconventionalapplicationscome
Manuscript received April 7, 1973. This work was sponsored by the
MineDetectionDivision, Counter-Mine/Counter-Intrusion Department,
U. S . Army MobilityEquipment Research and Development Center, Ft.
Belvoir, Va. 22060.
L. A. Robinson and W. B. Weir are withStanford Research Institute, generate the frequencies of the Pulse.spectrum. After appropri-
Menlo Park, Calif. 94025.
L. Young was with Stanford Research Institute, M e d o Park, Calif.
94025. He is now with Naval Research Laboratory, Electronics Division,
Washington, D . C. 20375. synthetic-spectrum
technique.
the
Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 03,2024 at 11:03:22 UTC from IEEE Xplore. Restrictions apply.
PROCEEDINGS OF THE IEEE, VOL. 62, NO. 1, JANUARY 1974
REFERENCES
[ l ] L. A. Robinson, W. B. Weir, and L. Young, “Location and recogni-
tion of discontinuities in dielectric media using synthetic RF pulses,
this issue, pp. 36-44.
[2] A. R. Von Hipple, Ed., DielccfricMaterialsandApplications. New
York: Wiley, 1954.
[3] A. R. Von Hipple, Dielccfric and Waves. New York: Wiley, 1954.
[4] S. Ram0 and J. R. Whinnery, Fields and Waves inModem Radio, 2nd.
ed. New York: Wiley, 1953, ch. 8.
[5] W. E. Milne, Numerical Calculus. Princeton, N: J.: Princeton Univ.
Press, 1949, ch. IV.
[6] S. B. Cohn, “Confusion and misconceptions in microwave engineer-
ing.” Minowave J . ; vol. 11, p. 20, Sept. 1968.
[7] A. M. Nicolson and G . F. R w , ‘Measurement of the intrinsic proper-
ties of materials by time-domain techniques,” ZEEE Trans. Znsfrum.
MUS., V O ~ .IM-19, pp. 377-382, NOV.1970.
RF Pulses
along. Pulse radar concepts have been applied or to considered
for nondestructive testing, such as finding faults in dielectric
bodies [l], finding old sewers or buried cables or abandoned
coal mines in the ground, or finding burial chambers in ancient
pyramids [2]. Radioglaciology is concerned with making R F
soundings in glaciers [SI. R F waves have been used to deter-
mine thewatercontent of snow fields, and considered for
probing lunar and planetary surfaces[4]. Marine applications
are more likely to use sound than electromagnetic waves, but
the principles of echo sounding remain the same. In medicine,
ultrasonic waves again have better penetration combined with
resolution in biological tissue than electromagnetic waves.
What many of these applications have in common over the
conventional radar scenario is, firstof all, that the targetdoes
not move (or moves slowly), and second, a resolution of the
order of one wavelength is often required or desirable. The
system tobe described in this paper makesCW measurements
at many selected frequencies covering a wide spectrum. I t uses
a computer to control the sequenceof measurements, to store
the measuredparameters,andto process them.Finally,a
synthetic pulse echo is displayed on an oscilloscope or plotted
on a chart recorder. Earlier work on this system has been re-
ported briefly elsewhere [SI.
11. PRINCIPLE OF OPERATION
A . Reasons for Choice of Synthetic-Spectrum Method
T o copewith the problems of generatingawide-band R F
pulse, separateb u t coherent cw oscil~ators can be to
ate amplitude and phase adjustments, these signals can be
combined to form a pulse train. hi^ approach is referred to
as