IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. IM-32, NO.

2, JUNE 1983 385

ANA Calibration Method for Measurements of Dielectric

Properties
ANDRZEJ KRASZEWSKI, MARIA A. STUCHLY,
SENIOR MEMBER, IEEE, AND STANISLAW S. STUCHLY,
SENIOR MEMBER, IEEE
Abstract-Routine network analyzer calibration procedures in measure-
ments of the dielectric properties of materials using an open-ended coaxial line
probe are frequently inadequate and limit the accuracy of measurements. A
calibration method, which makes use of liquids whose properties are well known,
is proposed to alleviate this limitation. It is shown that even one liquid used as
a standard in place of a matched load greatly improves the accuracy of mea-
surements. Theoretical relationships and experimental results as well as some
practical suggestions related to the application of this method are given. Frequency in GHz
Fig. 1. Input impedance of an open 0.141-in teflon coaxial line immersed
INTRODUCTION in a test liquid versus frequency. The frequencies are shown in gigahertz.
All liquids at 20°C. (1 Methanol. (Z) Distilled water. () 0.70-N saline so-
A computerized network-analyzer-based method was used to lution. @ 1.37-N saline solution.
measure dielectric properties of biological tissues in vivo at radio and
microwave frequencies [1] -[3]. In this method, a probe consisting
of an open-ended coaxial line [2], [3] or a short monopole [1] is placed PRINCIPLE OF OPERATION
in contact with a measured sample, and the dielectric permittivity
is determined from measurements of the input impedance at the in- The principle of one-port error correction was previously described
terface of the probe/measured sample. The accuracy of measurement [4], [6], and various terminations in addition to the conventional case
of the permitivity is, to the greatest extent, determined by the accu- of short/open/matched load were considered [6]. The true reflection
racy with which the input impedance is measured. coefficient is related to the measured reflection coefficient as fol-
The accuracy of impedance measurements using a network ana- lows:
r= Pm -SlI
S22Pm + S I 2S2 1 -S I I S22
(1)
s _ Frr2p3(pI - P2) + FlF3P2(P3 - PI) + T'2P3PI(P2 - P3) (2)
F-F2(pI - P2) + FlF3(p3 - PI) + F2F3(P2 - P3)
lyzer can be significantly improved by employing a standard cali- IF(p2 - S11) + F2(S1I - PI) 2 (3)
bration procedure [4], [5]. In this procedure, three known termina- F1F2(P2 - PI)
tions, most frequently a short circuit, an open circuit, and a matched (PI - S1)(1 - S221)
load, are consecutively placed at the reference plane, the input re- (4)
flection coefficient is measured, and simple algorithms for corrections
are employed [4]. However, the application of this procedure in the where is the true complex reflection coefficient, Pm is the measured
case of permittivity measurements creates some complications, which reflection coefficient, S Ss 2, S21, S22, are the scattering parameters
I,

limit the accuracy. In particular, a standard matched load can only of the error network, FI, F2, r3 are the reflection coefficients of the
be placed at the probe-connector end instead of the probe-measuring three standard terminations used for calibration, and PI, P2, p3 are
end. In this arrangement, the imperfections of the probe connector the measured reflection coefficients corresponding to the calibration
and the probe section (reflections, deviations from the 50-4 line im- terminations, respectively.
pedance, etc.) remain uncompensated by the calibration procedure. In measurements of the permittivity using one-port techniques, for
Design of a special load which could be attached to the measuring end instance for in vivo tissue measurements, when probes such as an
of the probe is not only costly, but rather difficult when a low reflec- open-ended coaxial line, short monopole, stripline, or similar structure
tion performance in a wide frequency range is required. are used, the three terminations are a short, an open line, and a liquid
A simple calibration method for use in terminal (one-port) mea- of a known permittivity. In this case
surements of the permittivity was developed. In this method, the probe
is immersed in a liquid of well-known dielectric properties instead of
Fr = -1 (short circuit) (5)
using a matched load. Similarly, when difficulties are experienced F2 = exp[-j2arctan (WCTZO)]
in placing a short circuit at the probe end (e.g., for a short monopole), (open circuit modified by the
another standard liquid can be used. The modified correction algo- fringing capacitance) (6)
rithms and an example illustrating the selection of the liquids and the
improvement in the measurement accuracy are presented.
p3 = 1 - jwoCoZo0 jcojZo
(7)
1 + j]WCoZa + jwCfZo
Manuscript received August 26, 1981.
where CT = CO + Cf, Co is the fringing capacitatice due to the
A. Kraszewski and S. S. Stuchly are with the Department of Electrical fringing fields outside the coaxial line, Cf arises from the fringing
Engineering, University of Ottawa, Ottawa, Ont., KIN 6N5, Canada. fields inside the line (for more detailed explanation and methods of
M. A. Stuchly is with the Radiation Protection Bureau, Health and Welfare determination of C0 and Cf consult [2], [3]), = 2wf, f is the fre-
w

Canada, Ottawa, Ont., Canada, and with the Department of Electrical En- quency, Zo is the characteristic impedance of the coaxial line, and
gineering, University of Ottawa, Ottawa, Ont., KIN 6N5, Canada. E is the complex permittivity of the standard liquid.

0018-9456/83/0600-0385$01.00 (© 1983 IEEE

386 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. IM-32, NO. 2, JUNE 1983

80 As mentioned previously, it is advantageous to have the reflection

00
.%. x- -..i X, 0 coefficients of the calibration terminations significantly different from

0
~~~~~~00 0 0 X>S each other. In the frequency range of interest (0.1 -10 GHz) any two
0
w>> saline solutions shown in Fig. 1 can be selected, as well as any solution
0 00
of an intermediate salinity. Water is a good standard liquid at
0
0 0 0
00 { 0 0

0
0
frequencies between 0.5 and 10 GHz, while methanol is good only
7c~~ _____~ between 1 and 10 GHz. At lower frequencies, the input reflection
coefficients for water and methanol are too close to that of the open
circuit. Other polar liquids such as ethyl alcohol, propanol, ethylene
0.1 0.2 0.5 1.0 2.0 5.0 10 glycol may be considered depending on the line dimensions and the
Frequency [GHz] operating frequency.
(a)
EXPERIMENTAL RESULTS
Fig. 2 shows the dielectric constant E', and the loss factor e", of
distilled and deionized water as a function of frequency. Two sets of
experimental data obtained using the same measurement system, but
different calibration procedures are illustrated. The measurement
system was based on an HP 8410B network analyzer controlled by
an HP 9825S desktop computer. As a probe, a section of the 0.141-in
teflon coaxial line with an OSM connector was used.
It is evident from Fig. 2, that when the standard calibration pro-
cedure is used, a relatively wide spread of data points results, and the
results differ significantly from the reference data [ 11 ]. For instance,
the error in the dielectric constant at some frequencies is up to 7
0.1 0.2 0.5 1.0 2.0 5.0 10
percent. When a 0.7-N saline solution is used for calibration instead
Frequency [GHz] of a matched load, the difference between the reference data for water
(b) and the results of this work is less than 1 percent in the dielectric
Fig. 2. Permittivity of distilled and deionized water versus frequency at 250C constant. A similar improvement in the accuracy of measurements
measured using the conventional calibration method (0) as compared with of the loss factor is observed.
the calibration using 0.7-N saline solution (X). Solid line from [I I ].

The permittivity of the standard liquid should be selected such that
P3 is substantially different from IF and F2. This minimizes cali-
bration errors resulting from uncertainties in e [6]. Liquids that meet
this requirement are, in general, polar and exhibit significant con-
ductivity, and their permittivity is a function of frequency. It is con-
venient to select liquids whose permittivity changes with frequency
can be described analytically. For many polar liquids such as water
and alcohols, the permittivity is described by the Debye equation
1 + (U27rf)1 (8)
where sc, cs, are constants characterizing the liquid. When salt
T, a wave frequencies," IEEE Trans. Microwave Theory Tech., vol.,
is added to such a liquid, then
2irtffo
1 + (j2wfT)1 a
where is the liquid conductivity, co = 8.854 X 10-12 F/im.
a
EXAMPLE OF APPLICATION
An open-ended coaxial line and a computerized network analyzer
were used to measure the permittivity of biological tissues in vivo [2],
[3]. In the frequency range 0.1 -10 GHz, a 0.141-in teflon line was
employed [7]. This line was terminated by an OSM connector, while
an APC-7 connector was fitted to the network analyzer. When a
matched load placed at the network analyzer terminal was used for
calibration the effect of the uncorrected mismatch of the connectors
and adapters was apparent. Therefore, the proposed calibration
method provided a natural solution.
Fig. 1 shows the input impedance versus frequency (0.1 -10 GHz)
for water, methanol, and two saline solutions of different concen-
trations. Both water and methanol exhibit the Debye behavior. The
reflection coefficients for water at 20°C were calculated using (8)
and taking c. = 4.23, e = 80.4, X = 9.31 ps, = 0.013 [8], while for
methanol taking E = 4.5, es = 34.8, T = 56 ps, = 0 [9]. For saline
e., and are dependent on the solution molarity and the
solutions c, T
conductivity term has to be added [10, eq. (9)].
CONCLUSIONS
A simple calibration method of the network analyzer used to
measure the permittivity of materials with an open-ended coaxial line
is proposed. The calibration method makes use of liquids whose
electrical properties (permittivity) are well known. A substantial
improvement in the accuracy of measurements is obtained.
REFERENCES
[1] E. C. Burdette, F. L. Cain, and J. Seals, "In-vivo probe microwave
technique for determining dieletric properties at VHF through micro-
MTT-28, pp. 414-427, 1980.
[2] T. W. Athey, M. A. Stuchly, and S. S. Stuchly, "Dielectric properties
of biological substances at radio frequencies, Part 1: Measurement
method," IEEE Trans. Microwave Theory Tech., vol. MTT-29,
1981.
[3] M. A. Stuchly, T. W. Athey, G. H. Samaras, and G. E. Taylor, "Di-
electric properties of biological substances at radio frequencies, Part
11: Experimental results," IEEE Trans. Microwave Theory Tech., vol.
MTT-29, 1981.
[4] Hewlett-Packard, "Automating the HP 8410 B Microwave Network
Analyzer," Application Note 221A, Hewlett-Packard, June 1980.
[5] J. K. Fitzpatrick, "Automatic network analyzer accuracy: How to get
it, lose it, then regain it," Microwave Syst. News, vol. 10, pp. 77-93,
1980.
[6] E. F. de Silva and M. K. McPhun, "Calibration techniques for one port
measurement," Microwave J., pp. 97- 100, June 1978.
[7] M. A. Stuchly and S. S. Stuchly, "Coaxial line reflection method for
measuring dielectric properties of biological substances at radio and
microwave frequencies-A review," IEEE Trans. Instrum. Meas., vol.
IM-29, pp. 176-183, 1980.
[8] A. M. Bottreau, J. M. Moreau, J. M. Laurent, and C. Marzat, "A
method analysing the dielectric relation spectra of mixtures. Application
to the study of the water spectrum," J. Chem. Phys., vol. 62, pp.
a
360-365, 1975.
a
[9] B. P. Jordan, R. J. Sheppard, and S. Szwarnowski, "The dielectric
properties of formamide, ethonedial and methanol," J. Phys. D: Appl.
Phys., vol. I 1, pp. 695 -701, 1978.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL.
[10] A. Stogryn, "Equations for calculating the dielectric constant of saline
water," IEEE Trans. Microwave Theory Tech., vol. MTT-19, pp.
733-736, 1971.
[11 ] H. P. Schwan, R. J. Sheppard, and E. H. Grant, "Complex permittivity
of water at 25°C," J. Chem. Phys., vol. 64, pp. 2257-2258, 1976.
A Photosensor-Based Measuring System for Precision Mechanics
GIOVANNI MAROLA AND MARCO SCHIAFFINO
Abstract-In this short paper we describe a general measurement method
for quality control of curved surfaces, obtained by means of machine-tools.
Such control can be performed by measuring the surfaces in correspondence
to a discrete number of adjacent points.
Our system is based on a matrix of analog position transducers using com-
mercially available photocouplers, microprocessor controlled by appropriate
programs.
To avoid the time variability of the displacement voltage transfer function,
the characteristics of the photocouplers are stored in a digital memory by
measuring a convenient number of reference elements or surfaces (initial cali-
bration). These stored characteristics are then utilized for the evaluation of the
unknown surface (measurement).
As a result, we obtain with economical components an instrument having a
remarkable precision which depends on the photocoupler stability and on the
time interval between calibration and measurement.
I. INTRODUCTION
The problem solved with this measurement system typically arises
in the field of precision mechanics, in conjunction with the quality
control of a great number of curved surfaces that should have a pre-
determined form.
The main difficulties in these measurements result from the ne-
cessity to evaluate length variations in a small range (in the order of
100 g.m) and to test a large number of points (in order of hun-
dreds).
Moreover, because the points to be tested are very near each other,
the traditional methods using differential transformers (LVDT's),
capacitive transducers, etc., [1] are impracticable. Nevertheless, a
simple and even economical solution could be obtained using, as a
sensor, a light-emitting diode/phototransistor combination [2], which
allows the implementation of a high-speed linear-position transducer
having very small dimensions.
These photocouplers, available in a wide range of shapes and di-
mensions, can be placed on a suitable reading head (as shown in Fig.
1) to satisfy many requirements of precision mechanics.
However, measurement systems utilizing optocouplers are not at
present of practical feasibility, owing to the variability with time of
the current transfer ratio (CTR) (degradation) [4]. The above
variability changes the device transfer characteristics remarkably,
not allowing accurate results to be obtained when the time interval
between two measurements is long.
Thus we propose a new method to overcome this crucial limitation.
In fact, storing at the beginning of each measurement, the charac-
teristic data of each sensor and utilizing them immediately after,
during the measurement, it is possible to avoid the intrinsic instability
with time of the device.
Manuscript received March 3, 1982; revised February 28, 1983.
G. Marola is with the Istituto di Elettronica, UniversitA di Pisa, 56100 Pisa,
Italy.
M. Schiaffino is with the Istituto di Scienza dell'Informazione, Universita
di Pisa, 56100 Pisa, Italy.
0018-9456/83/0600-0387$01.00
IM-32, NO. 2, JUNE 1983 387
ad j u at
Fig. 1. Reading head using an array of slotted photocouplers.
LE0 PHOTOTRANSISTOR
(a) (b) (c)
Fig. 2. (a) Physical structure of the transmissive optical displacement
sensor. (b) Electrical circuit. (c) Transfer curve.
The calibration procedure is performed periodically by using a set
of reference elements and storing in a digital memory the corre-
sponding values of each photocoupler. In other words, the micro-
processor carries out, in an orderly way, the acquisition of the initial
transfer curve of each photocoupler which is then utilized for the
calculation of the height of the corresponding point tested.
II. THE MEASURING HEAD
The use of the solid-state optocoupler as a simple, robust dis-
placement transducer that is unaffected by shock forces has been
proposed in previous works [2], [3].
The basic structure of a transmissive displacement optical sensor,
as well as its electrical circuit and voltage displacement characteristic,
are shown in Fig. 2(a), (b), and (c).
The voltage V is a function of the amount of flux coupled into the
receiver transistor and, therefore, it gives the position of the me-
chanical element to be tested. Furthermore, the linearity of the curve
of Fig. 2(c) can be improved by utilizing a well-suited diode function
generator [2].
In a particular application, a set of 525 of these elements has been
arranged in a matrix with 15 rows and 35 columns, to implement a
flat 30 X 10 cm reading head.
The reading head just described is simple, compact, and of rugged
construction.
Nevertheless, the well-known degradation [4] of each optocoupler
is a serious drawback for the implementation of precise and stable
displacement measuring instruments.
In fact, even if it is not easy to assign an order of magnitude to this
degradation, which depends on the actual operating conditions, we
may assume as acceptable a degradation of CTR equal to 10 percent
every 100 h [4].
This characteristic degradation should imply, therefore, an un-
acceptable variation of the instrument's long-term accuracy.
III. THE MICROPROCESSOR-BASED CONTROL SYSTEM
The block diagram of the control system is shown in Fig. 3. All the
photosensor collectors are wired together, permitting the use of a
single A/D converter for the collection of the output voltages. of the
1983 IEEE