You are on page 1of 7

2784 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO.

11, NOVEMBER 2014

Microstrip Transmission Line Method for Broadband


Permittivity Measurement of Dielectric Substrates
Prasanth Moolakuzhy Narayanan

Abstract—This paper presents a novel microstrip transmission of T/R techniques have been developed over the years, such
line method for broadband relative permittivity measurement of as coaxial [5], planar [6]–[9], and waveguide [10], [11]. The
planar dielectric substrate materials. The method requires three theory of T/R methods based on coaxial and waveguide trans-
sets of -parameter measurements of the microstrip line together
with an obstacle in three equidistant positions over the line. The mission lines and various errors in the methods are described
measurement and simulation results for the broadband relative in [12]. In [13], a strip-line transmission line method has been
permittivity of high-frequency substrates, TLX-8, RF60A, CER10, proposed that enables simultaneous broadband measurement of
and the widely used FR4 are presented. The errors are calculated complex permittivity and permeability. However, the method
based on the manufacturers’ data sheet value. Both the simulation is sensitive to dimensional resonance and impedance mis-
and measurement results are found to be within 16% of the data
sheet values for CER10 and 10% for all other substrates. The pro- match. Moreover, the method requires a test chamber and test
posed method can minimize errors due to the nonreproducibility of samples with dimensions suitable to fit in the test chamber. In
connectors and impedance mismatch problems, prevalent in trans- [6], a microstrip-based method for complex permittivity and
mission line methods. However, the method is highly sensitive to the permeability measurement is presented. The method requires a
positioning of the obstacle, which can be overcome through the use microstrip line of known permittivity substrate, and the sample
of high accuracy obstacle positioning methods.
under test placed over the microstrip line. It needs calibration to
Index Terms—Broadband measurement, dielectric materials, eliminate the systematic errors due to network analyzer, cables,
material characterization, microwave measurement, permittivity,
and test device. In addition, the possibility of an air gap between
planar substrates, scattering parameters.
the sample substrate and microstrip line may cause errors. A
number of self-calibration methods that eliminate the effect of
I. INTRODUCTION impedance mismatch and calibration errors are reported in the
literature [5], [10], [11]. However, they use coaxial or wave-
guide transmission lines and they are not suitable for planar or

B ROADBAND characterization of dielectric substrates is


essential for its wideband and microwave applications.
Frequency-dependent variation in substrate permittivity causes
printed circuit board (PCB) substrate characterization.
In [7], two methods for measuring substrate permittivity
using microstrip and stripline have been proposed. The first
dispersion in transmission lines, which causes pulse distor- method requires two microstrip lines of different lengths. The
tion in ultra-wideband (UWB) systems [1], [2]. In addition, accuracy of this method is limited by the error in measuring
microwave integrated circuits and RF packaging materials the physical lengths of the two lines, as well as the nonideality
require a good understanding of the broadband permittivity of connectors. The second method consists of a stripline with
characteristics of the substrate materials for accurate design. different substrates on either side of the conductor. One of
Microwave permittivity measurement methods can be broadly them is a standard substrate with known permittivity and the
classified into resonant and nonresonant methods. Resonant other is a test substrate. Here the method can overcome the
methods are more accurate and suitable for characterization connector mismatch problem reported in other methods [4],
of low-loss dielectrics [3]; however, they can only provide [12]; however, possibility of air gap between the substrates of
measurement results at a single frequency or a number of the strip-line exists. In [8], a method for determining complex
discrete frequencies. In the nonresonant class, the transmis- permittivity and permeability of microstrip substrates based
sion/reflection (T/R) methods are relatively easy to implement on -parameter measurements has been proposed. However,
and enable broadband measurements; however, they are less the method is sensitive to impedance mismatch problems. In
accurate than the resonant methods [3]. The accuracy of T/R [9], a method for determining the causal dielectric model of a
methods are limited by calibration errors, nonreproducibility microstrip substrate was proposed. The method uses a multiline
of connectors, and impedance mismatch [4]. Various types through reflect line (TRL) calibration method for extraction of
the propagation constant. This method requires multiple line
standards for the measurement.
Manuscript received March 18, 2014; revised July 18, 2014 and August 19,
In this work, a microstrip transmission line method for mea-
2014; accepted August 23, 2014. Date of publication September 12, 2014; date
of current version November 03, 2014. suring the relative permittivity of planar dielectric substrates is
The author is with the Nimbus Centre, Department of Electronic Engineering, presented. The proposed method uses the sample under test as
Cork Institute of Technology, Cork, Ireland (e-mail: mnprasanth@gmail.com).
the substrate of microstrip line, and an obstacle that has to be
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. placed at different positions over the microstrip line for mea-
Digital Object Identifier 10.1109/TMTT.2014.2354595 surements. The method uses the line network network (LNN)

0018-9480 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
NARAYANAN: MICROSTRIP TRANSMISSION LINE METHOD FOR BROADBAND PERMITTIVITY MEASUREMENT OF DIELECTRIC SUBSTRATES 2785

[15]. The propagation constant of the transmission line can be


calculated from the measured transmission matrices using (5)
[11],

(5)

There are four solutions for (5). The first two solutions are
Fig. 1. Schematic diagram of the microstrip line with three different positions given as follows in (6) and (7), the remaining two solutions are
( 1, 2, and 3) of the obstacle, the step length is represented as . the same as the first two solutions and do not need to be consid-
, ).
ered:

[14] calibration procedure of vector network analyzers (VNAs) (6)


for the extraction of the propagation constant of the transmis-
sion line. It does not require reconnecting of connectors for each (7)
measurement: hence, avoids errors due to nonreproducibility of
connections. where is the step length (the step length is the distance between
Section II describes the method to extract the propagation the centers of two consecutive obstacle positions), is the prop-
constant of the transmission line and the calculation of relative agation constant of the microstrip line, and can be found using
permittivity. (8),

(8)
II. MICROSTRIP-BASED TRANSMISSION LINE METHOD
trace (9)
Fig. 1 shows the schematic diagram of the microstrip line with
trace (10)
three consecutive and equidistant positions of the obstacle. The
positions of the obstacle are labeled as 1, 2, and 3. The ob- The first solution of propagation constant will work up
stacle is placed consecutively at the three positions, and the three to the frequency given in (11). The equation provides the
sets of two-port -parameters of the microstrip line are mea- frequency corresponding to the line length that equals half of
sured together with the obstacle. Determination of the propaga- guided wavelength. The line length is given as twice the step-
tion constant of the microstrip line, from two-port -parameters length
based on the LNN calibration procedure [14], [15], is described
next. The transmission matrix ( ) of the microstrip line can be (11)
calculated from the two-port -parameters using (1)
is the speed of light in free space and is the effective per-
(1) mittivity of the microstrip line. The complex effective permit-
tivity can be calculated using the propagation constant
given as follows in (12) [8]:
The transmission matrix of the microstrip line with the ob-
stacle in position 1, 2, and 3 can be written as
(12)
(2)
where is the frequency. The propagation constant consists of
(3)
various losses such as conductor resistance and phase-shift
(4) constant . The relative permittivity of the substrate material
is derived from the effective permittivity and dimensions of the
, , and are the transmission matrices microstrip line using (13) [16],
of the microstrip line with the obstacle in position 1, 2, and
3, respectively. Here, is the transmission matrix of the ob- (13)
stacle and is the transmission matrix of the microstrip line.
Also, , and are the two-port transmission matrices (14)
representing the systematic errors of the VNA and measure-
ment setup. The systematic errors include the microstrip sec- (15)
tions outside the calibration reference planes ( 1 and 3) and
impedance mismatch of the ports. The error matrices are the where is the width of copper track, is the thickness of
same for all measurements and they will be eliminated during copper track, is the thickness of substrate, and is the
the calculation. Therefore, the method enables the permittivity effective microstrip width. Equation (13) was used in [16] for
measurement without the influence of various systematic errors. permittivity calculation at frequencies up to 35 GHz and good
Detailed information about the error networks can be found in accuracy was shown.
2786 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 11, NOVEMBER 2014

Fig. 4. Simulated transmission coefficient of various obstacles.

Fig. 2. Simulated relative permittivity of TLX8 substrate with obstacles of var-


ious reflectivity.

Fig. 5. Simulated relative permittivity of TLX8 substrate with different ob-


stacle width ( ) and length ( ).

Fig. 3. Simulated reflection coefficient of various obstacles. 1.5% over the frequency range of 4–10 GHz when compared
with the reference obstacle ( , mm). The reference
III. PERFORMANCE ANALYSIS THROUGH SIMULATIONS obstacle was selected based on best measurement result, after
In this section, the effect of various design parameters on the comparing the results of different obstacle sizes. However, an
performance of the proposed method is analyzed. The perfor- error of 7% is found for the obstacle with 12-mm length at fre-
mance of the method is evaluated in terms of accuracy and band- quency of about 11.5 GHz. This is most likely due to resonance
width. According to sensitivity simulations reported in [17], for in the obstacle because the obstacle length is about one-half of
optimum performance, the obstacle should exhibit a reflection the wavelength of the frequency. The results show that the ob-
and transmission loss within the range of 4–12 and 2–30 dB, stacle dimensions do not significantly affect the measurement
respectively. A section of CER-10 copper-clad PCB was used accuracy if the obstacle reflectivity is within the above speci-
as the obstacle for both experimental measurements and simu- fied range. However, long metallic obstacle will cause errors be-
lations. The obstacle was placed in such a way that the metal cause of the resonances. In addition, the accuracy of the method
side of the obstacle is in touch with the trace of the microstrip is independent of the obstacle’s position on the microstrip line
line. Several simulations have been conducted to study the effect as long as the positions are equidistant. Therefore, the three dif-
of the obstacle reflectivity and dimensions on the simulation re- ferent equidistant positions can be selected anywhere on the mi-
sults. Simulated relative permittivity of the TLX8 substrate with crostrip line.
the obstacles of various reflectivity is shown in Fig. 2. In addi- The bandwidth of the proposed method depends on the elec-
tion, the simulated reflection and transmission coefficients of the trical length of the line. The maximum frequency of operation
various obstacles are shown in Figs. 3 and 4, respectively. It can for the first solution of propagation constant is given in
be observed from Fig. 2 that the simulated permittivity results (11). From (11), it can be observed that the upper frequency limit
with the lower reflective obstacles of permittivity 2 and 5 are of is inversely related to the line length and square root of
erroneous and unstable. Highly reflective obstacles of permit- effective permittivity. For frequencies above , the second so-
tivity 20 and metal are demonstrating accurate results. The cur- lution of the propagation constant can be used. The upper
rent simulation results show that the reflection and transmission frequency limit of varies between different substrate mate-
loss required for the obstacle, reported in [17], can be modified rials. The breaking of is most likely due to the higher order
as 4–20 and 0.5–30 dB, respectively. Fig. 5 compares the sim- resonances in the microstrip line. Considering the lowest oper-
ulated permittivity of the TLX8 substrate over the 4–12-GHz ating frequency, the electrical length of the line should be long
frequency range with three different obstacle dimensions. The enough for accurate measurements at low frequencies. The sim-
lower frequency limit is selected as 4 GHz because the lowest ulation study reveals that the lowest operating frequency can be
operating frequency of the TLX8 substrate for 5-mm step length lowered by increasing the step length. Simulations were carried
is about 4 GHz. The upper frequency limit used for all simula- out with four different substrate materials using four different
tion and measurement experiments is 12 GHz or less. Results step lengths (equal electrical lengths for all substrates) in each
show that the maximum variation in permittivity is less than material. Simulation results show that while the average lowest
NARAYANAN: MICROSTRIP TRANSMISSION LINE METHOD FOR BROADBAND PERMITTIVITY MEASUREMENT OF DIELECTRIC SUBSTRATES 2787

Fig. 6. Simulated relative permittivity of the TLX8 substrate with different step
lengths.

TABLE I
DIMENSIONS OF THE MICROSTRIP LINES USED FOR EXPERIMENTS

Fig. 7. Measurement setup with sample substrate and obstacle.

operating frequency of the four substrates is 2.5 GHz, the corre-


sponding guided wavelength is eight times the line length in all
substrates. However, the relation between lowest operating fre-
quency and line length found varies for different line lengths.
The simulated permittivity of the TLX8 substrate as function
of frequency for four different step lengths is shown in Fig. 6.
For the step lengths of 5, 10, 15, and 20 mm, the corresponding Fig. 8. Microstrip lines manufactured for the measurements with the sample
materials TLX8, FR4, RF60, and CER10.
lowest operating frequency was found as about 2.9, 1.9, 1.5, and
0.8 GHz, respectively. The maximum error at lowest frequency
is found less than 6.9%. connected to the two ports of the VNA using a coaxial cable. A
photograph of the microstrip lines with the four different sub-
IV. EXPERIMENTAL VERIFICATION strate materials is shown in Fig. 8.

Simulations and measurement experiments have been con-


A. Simulation and Measurement Results
ducted to verify the proposed method. Four different substrate
materials, the widely used FR4 and the high-frequency Taconic In Fig. 9, comparison of the simulated and measured permit-
substrates TLX8, RF60A, and CER10, have been characterized. tivity for the FR4 and TLX8 substrates over the frequency range
Table I shows the parameters of the microstrip lines used for the of 4–11 GHz is shown. Furthermore, Fig. 9 shows the mea-
experiments. All the microstrip lines are designed for a charac- surement results of the FR4 substrate over the frequency range
teristic impedance of 50 . In the case of unknown substrates, an of 4–7.5 GHz using an improved two-microstrip line method
approximate knowledge of the substrate permittivity is required [18]. Simulation experiments are free from most of the measure-
for designing the microstrip lines. Simulation experiments were ment errors such as obstacle positioning, geometry of the mi-
carried out with 3-D electromagnetic simulation software, CST crostirp line, and surface roughness of the conductor. However,
Microwave Studio. The simulation software requires the permit- the frequency-dependent dispersion model used for the simula-
tivity and loss tangent values at single frequency. The software tion may have small variation form the actual dispersion char-
uses a standard dispersion model such as the Debye model for acteristics of the microstrip line due to the possible errors in the
calculating the frequency-dependent loss and permittivity. Mea- model. The overall error in the simulations will be smaller than
surements were conducted with a HP8720C VNA. The compu- the measurements. Therefore, comparison between the simula-
tations were performed using a program written in MATLAB. tion and measurement results helps to identify most of the errors
All simulation and measurement experiments are performed in the measurement results. The simulated and measured results
using a 5-mm step length because it enables the measurements match well for both substrates. The largest difference between
over the UWB frequency range in all substrate materials. The simulation and measurement results for FR4 is about 8% of the
microstrip line was fixed on a test fixture (Anritsu 3680-20), actual permittivity at 9 GHz and for TLX8 it is about 6.6% of
which was then connected to the VNA. Fig. 7 shows the mea- the actual permittivity at 10.2 GHz. The largest difference be-
surement setup used for the experiments, where ports 1 and 2 are tween the measurement results of FR4 with the two-microstrip
2788 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 11, NOVEMBER 2014

TABLE II
MEASURED, SIMULATED, AND ACTUAL PERMITTIVITY VALUES

ufacturers’ data sheet at frequencies at which the error is


calculated. The errors are calculated based on the manufac-
turers’ data sheet as reference values. The manufacturer is using
Fig. 9. Measurement and simulation results for the relative permittivity of FR4 the IPC-TM-650 standard for the permittivity measurements,
and TLX8 substrates. which uses a resonator method. Simulations show an error
of 4% and 7% at 10 GHz for the substrates TLX8 and FR4,
respectively. The measurement error is found as 4.7% for the
TLX8 at 10 GHz. The permittivity value of FR4 provided by
the manufacturer was only at low frequency and not suitable for
comparison. Therefore, the measured permittivity error of the
FR4 is calculated in comparison with the measurement results,
using the resonator method, provided in [19]. The measured
relative permittivity of the FR4 at 10 GHz is about 4.5 in [19].
Therefore, FR4 shows an error of 1.5% compared to the results
in [19].
The simulation and measurement errors of substrates RF60
and CER10 are calculated at frequency point that shows max-
imum errors, therefore the frequency points used for the error
Fig. 10. Measurement and simulation results for the relative permittivity of
estimation are different between simulation and measurement.
RF60 and CER10 substrates.
The largest error in simulation and measurement for RF60 is
line method and the current method is about 6.6% of the actual found as 8.4% at 3 GHz and 9.4% at 3.1 GHz, respectively.
permittivity at 7 GHz. Similarly, the largest error in simulation and measurement for
Fig. 10 shows the simulation and measurement results for CER10 is found as 10% at 3.2 GHz and 16% at 6.5 GHz, re-
the RF60 and CER10 substrates over the frequency range of spectively. The measurement results of the permittivity of the
3–8 GHz. Furthermore, Fig. 10 shows the measurement results substrates TLX8, RF60, and CER10 using broadband transmis-
of the CER10 substrate reported in [19] based on a resonator sion line methods have not been reported in the literature. The
method over the frequency range of 3–8 GHz. The largest dif- accuracy of the broadband microstrip transmission line method
ference between simulation and measurement results is 9% for reported in [6] is 5% over the frequency range of 0.5–10 GHz.
RF60 at 3 GHz and 14% for CER10 at 8 GHz. The largest differ- The two-microstrip-line method reported in [7] shows an error
ence between the measurement results of CER10 with the res- of 0.4% in the permittivity for 0.1-mm error in the line length.
onator method and the current method is about 12% of the actual A broadband permittivity measurement method based on the
permittivity at 6.5 GHz. The differences between the simula- time-domain measurements reported in [20] shows repeatability
tion and measurement are mainly due to the following reasons. within 0.1 for the permittivity range of 2–3 and 0.5 for the
Firstly, the small errors in the dispersion model of the simula- permittivity range of 12–15. The error calculated for the current
tion software cause error in the simulated permittivity. In ad- method is the largest possible error; therefore the error is higher
dition, various errors in the measurements such as obstacle po- for the substrates with large tolerance range. The tolerance of
sitioning error can cause error in the measurement results. The the CER10 substrate is 0.5, which is about 10.5% of the nom-
errors in the simulation and measurement results cause the dif- inal substrate permittivity. Therefore, the actual error in the cur-
ferences between the simulation and measurement results. The rent results may be less than the values calculated. For example,
effect of obstacle positioning error will be higher in high-per- the measurement error of the CER10 at 6.5 GHz is calculated
mittivity substrates due to the large electrical length. Therefore, as 16% based on the permittivity value of 10, but the error is
the difference between the simulation and measurement results 6.6% based on the permittivity value of 9. The error of CER10
is larger in the high-permittivity substrates such as CER10. Var- is 11.6% based on the nominal substrate permittivity of 9.5. The
ious measurement errors are discussed in Section V. current simulation and measurement results of substrates TLX8,
FR4, and RF60 show an error of 5% based on the nominal sub-
B. Simulation and Measurement Errors strate permittivity, which is similar to the error reported in other
Table II shows the permittivity values of the four different broadband methods [6], [20]. However, the nominal error of
substrates obtained from simulation, measurement, and man- the high-permittivity substrate, CER10, is found as higher than
NARAYANAN: MICROSTRIP TRANSMISSION LINE METHOD FOR BROADBAND PERMITTIVITY MEASUREMENT OF DIELECTRIC SUBSTRATES 2789

TABLE III reduction in line length. In Case 2, obstacles 1 and 3 moved


STEP LENGTHS CORRESPONDING TO VARIOUS SIMULATED ERROR CASES 0.1 mm away from the center, which introduced an error of 2%
increment in line length. Cases 1 and 2 caused an average of
4.2% and 4.6% errors in permittivity, respectively. The results
of Cases 1 and 2 show that equal error in step lengths cause uni-
form error in permittivity over frequency. For Cases 3–6, the
error in step lengths was implemented using obstacle positions
1 and 3, whereas the center position (position 2) was kept fixed.
In Cases 3 and 4, the step lengths are unequal, but the line length
is correct because in Case 3 decreased by 0.1 mm when in-
creased by 0.1 mm. Cases 3 and 4 show that unequal step lengths
will cause large permittivity error at low frequency (an error of
2% in step lengths caused an error of more than 10% in per-
mittivity, for frequencies below 4 GHz), but there is no error at
high frequency, above 9 GHz. The high error at low frequency
is most likely due to the fact that the electrical length of the line
is smaller as compared to the wavelength of the signal, which
makes the small error in the step lengths become significant.
At higher frequencies, the line length is larger as compared to
the wavelength of the signal, which makes the small mismatch
in the step lengths become insignificant. Cases 5 and 6 have un-
equal step lengths and an error in line length, which caused large
error at low frequencies (an error of 1% in line length caused
an error of above 10% in permittivity, for frequencies below
Fig. 11. Simulated errors in relative permittivity, in percentage, of TLX8 sub-
strate corresponding to the various step-length errors.
3 GHz) and a uniform error at higher frequencies (1% error in
line length caused a uniform error of about 2% at frequencies
the other broadband measurement methods. The higher error in above 9 GHz). The results of Cases 3–6 show that the lowest
CER10 is due to the following reasons. Firstly, because of the operating frequency is reduces in accordance with an increase in
high permittivity, the small error in the obstacle positioning will mismatch between the step lengths. In addition, the results show
cause large error in the electrical length of the line. Secondly, that an error in the line length causes a permittivity error at all
the calculated largest error will be higher due to the large toler- frequencies. In real measurements, the error reported in Cases 5
ance of the material. The main source of measurement error is and 6 is most likely to occur due to possible error in all three po-
the error in obstacle positioning during the manual placement, sitions. The error in step lengths for Cases 7 and 8 is the same as
which is discussed in Section V. in Cases 3 and 4; however, for Cases 7 and 8, the error occurred
in the position of the central obstacle. The error in permittivity
V. SENSITIVITY ANALYSIS for Cases 3, 4, 7, and 8 can be found as about the same because
The obstacle positioning error causes error in the step lengths. their errors in the step lengths are the same. However, the errors
The error in step length will introduce error in the line length, in step lengths are introduced by the obstacle at different posi-
as the line length is twice the step length. The incorrect step tions for different cases. The results show that the position of the
lengths or line length will cause error in the propagation co- obstacle that makes the error is not important as long as the error
efficient , which leads to error in measurement results. The in the step lengths is the same. Therefore, it is not required to
human hand exhibits various errors in the placement of obsta- simulate all combinations of positioning errors that make sim-
cles. Various papers quantified the attainable accuracy in hand ilar error in the step lengths. The sensitivity of positioning error
manipulation [21], [22]. In [21], the root mean square (rms) am- will reduce with the increase in line length. Therefore, the effect
plitude of hand-positioning errors in a stationary task was re- of positioning error can be minimized by increasing the total
ported as 526 m. Simulations have been conducted to study line length. Furthermore, accurate positioning methods such as
the effect of different possible errors in step lengths on the mea- motor positioning or a fixture with slots [11] for placement of
surement results. Various simulated step-length errors using the the obstacle can be used to minimize positioning errors. In ad-
TLX8 substrate are shown in Table III. When first and second dition to the obstacle positioning error, poor reflectivity of the
step lengths are represented as and , the error in each case obstacle because of the improper contact between the obstacle
can be described as shown in Table III. and microstrip conductor, the impact of transmission line loss,
The correct step-length value for the simulations is 5 mm and and the error in geometry of the microstrip line are the other
therefore the correct line length is 10 mm. For error estimations, sources of measurement errors. The influence of overall trans-
a 2% error in the step length was implemented by moving the mission line loss on the real part of the measured permittivity of
obstacle positions by 0.1 mm. Fig. 11 shows the simulated errors FR4 is found as 0.00085 and 0.0039 at 5 and 11 GHz, which is
in relative permittivity (in percentage) corresponding to various about 0.018% and 0.086% of the substrate permittivity, respec-
errors in step lengths. In Case 1, both obstacles 1 and 3 moved tively. Therefore, the influence of the transmission line loss, par-
0.1 mm towards the center, which introduced an error of 2% ticularly the conductor loss at high frequency, on the real part
2790 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 11, NOVEMBER 2014

of the measured permittivity is negligible. The overall loss of [3] L. P. Ligthart, “A fast computational technique for accurate permit-
the remaining high-frequency substrate materials is smaller than tivity determination using transmission line methods,” IEEE Trans. Mi-
crow. Theory Techn., vol. MTT-31, no. 3, pp. 249–254, Mar. 1983.
the loss of FR4. Therefore, the impact of loss on the measured [4] J. B. Jarvis, E. J. Vanzura, and W. A. Kissick, “Improved technique
permittivity can be considered as negligible for the remaining for determining complex permittivity with the transmission/reflection
substrates. However, the loss tangent of the materials has large method,” IEEE Trans. Microwave Theory Techn., vol. 38, no. 8, pp.
1096–1103, Aug. 1990.
error. The method is not preferable for the measurement of the [5] K. H Baek, H. Y. Sung, and W. S Park, “A 3-position transmission/
loss tangent. The error in geometry of the microstrip line due reflection method for measuring the permittivity of low loss materials,”
to under etching can cause error in the measured permittivity IEEE Microw. Guided Wave Lett., vol. 5, no. 1, pp. 3–5, Jan. 1995.
[6] P. Queffelec, P. Gelin, J. Gieraltowski, and J. Loaec, “A microstrip
as the formula used to calculate relative permittivity from ef- device for the broad band simultaneous measurement of complex per-
fective permittivity uses the dimensions of the microstrip line. meability and permittivity,” IEEE Trans. Magn., vol. 30, no. 2, pp.
The errors due to under etching can be overcome by using ac- 224–231, Mar. 1994.
[7] N. K. Das, S. M. Voda, and D. M. Pozar, “Two methods for the mea-
curate manufacturing methods such as a PCB milling machine. surement of substrate dielectric constant,” IEEE Trans. Microw. Theory
The effect of the error in the dimensions of the microstrip line Techn., vol. MTT-35, no. 7, pp. 636–642, Jul. 1987.
and the effect of conductor loss on the relative permittivity will [8] J. Hinojosa, “ -parameter broadband measurements on-microstrip
and fast extraction of the substrate intrinsic properties,” IEEE Microw.
increase for the microstrip lines with smaller trace width and Wireless Compon. Lett., vol. 11, no. 7, pp. 305–397, Jul. 2001.
higher frequencies. [9] Z. Zhou and K. L. Melde, “A comprehensive technique to determine the
broadband physically consistent material characteristics of microstrip
lines,” IEEE Trans. Microw. Theory Techn., vol. 58, no. 1, pp. 185–194,
VI. CONCLUSION Jan. 2010.
[10] I. Rolfes and B. Schiek, “A novel method for the determination of the
A novel method for broadband permittivity measurement of dielectric properties of liquids at microwave frequencies,” in 36th Eur.
planar dielectric substrates has been presented. A microstrip line Microw. Conf., Manchester, U.K., 2006, pp. 399–402.
has been used as the measurement cell, which eliminates the [11] I. Rolfes, “Obstacle-based self-calibration techniques for the determi-
nation of the permittivity of liquids,” Adv. Radio Sci., vol. 5, pp. 29–35,
error due to the air gap between the trace of the microstrip and 2007.
the sample. Additionally, the method can minimize the effect [12] J. Baker-Jarvis, M. D. Janezic, J. H. Grosvenor, and R. G. Geyer,
of connector nonreproducibility and impedance mismatch prob- “Transmission/reflection and short-circuit line methods for measuring
permittivity and permeability,” NIST, Boulder, CO, USA, Tech. Note
lems, which are prominent in transmission line methods. How- 1355-R, Dec. 1993.
ever, the method is highly sensitive to the error in positioning [13] W. Barry, “A broad-band, automated, stripline technique for the simul-
of the obstacle. For example, simulations show an error of more taneous measurement of complex permittivity and permeability,” IEEE
Trans. Microw. Theory Techn., vol. MTT-34, no. 1, pp. 80–84, Jan.
than 10% of the permittivity value for frequencies below 4 GHz 1986.
corresponding to an error of 2% in step lengths. Therefore, the [14] H. Heuermann and B. Schiek, “Line network network (LNN): An alter-
method is not preferable for measurements at lower frequencies, native in-fixture calibration procedure,” IEEE Trans. Microw. Theory
Techn., vol. 45, no. 3, pp. 408–413, Mar. 1997.
below 4 GHz. The effect of positioning error can be minimized [15] H. Heuermann and B. Schiek, “The in-fixture calibration procedure
by increasing the line length. In addition, accurate obstacle po- line–network–network-LNN,” in 23rd Eur. Microw. Conf., Madrid,
sitioning methods such as motor positioning can minimize the Spain, 1993, pp. 500–503.
[16] G. Zou, H. Grnqvist, J. Piotr Starski, and J. Liu, “Characterization of
positioning error. Simulation and measurement results of rel- liquid crystal polymer for high frequency system-in-a-package appli-
ative permittivity for four different substrate materials, TLX8, cations,” IEEE Trans. Adv. Packag., vol. 25, no. 4, pp. 503–508, Nov.
FR4, RF60A, and CER10, have been presented. The worst case 2002.
[17] B. Schiek, “Principles of network analyzer calibration,” Bochum, pp.
simulation and measurement errors have been found to be less 38–39, Jul. 1996.
than 16% for CER 10 and 10% for all other substrates, within [18] M.-Q. Lee and S. Nam, “An accurate broadband measurement of sub-
the measured frequency range. The effects of design param- strate dielectric constant,” IEEE Microw. Guided Wave Lett., vol. 6, no.
4, pp. 168–170, Apr. 1996.
eters, such as the line length and obstacle size, on the band- [19] E. McGibney, J. Barton, L. Floyd, P. Tassie, and J. Barrett, “The high
width and accuracy of the method have been discussed. The pro- frequency electrical properties of interconnects on a flexible polymide
posed method finds applications in permittivity measurements substrate including the effects of humidity,” IEEE Trans. Compon.,
Packag., Manuf. Technol., vol. 1, no. 1, pp. 4–15, Jan. 2011.
of planar substrate materials with a wide range of thicknesses [20] A. M. Nicolson and G. F. Ross, “Measurement of the intrinsic prop-
and permittivity. erties of materials by time-domain techniques,” IEEE Trans. Instrum.
Meas., vol. IM-19, no. 4, pp. 377–382, Nov. 1970.
[21] R. Sandoval, R. A. MacLachlan, M. Y. Oh, and C. N. Riviere, “Po-
REFERENCES sitioning accuracy of neurosurgeons,” in Proc. 29th Annu. Int. IEEE
[1] P. M. Narayanan, P. Rulikowski, and J. Barrett, “Miniaturization of EMBS Conf., Lyon, France, 2007, pp. 206–209.
non-uniform transmission lines for ultra-wideband pulse shaping,” in [22] C. N. Riviere and P. K. Khosla, “Accuracy in positioning of handheld
IEEE Int. Microw., Commun., Antennas, Electron. Syst. Conf., Tel- instruments,” in 18th Int. IEEE Eng. Med. Biol. Soc. Conf., Amsterdam,
Aviv, Israel, , 2011, pp. 1–6. The Netherlands, 1996, pp. 212–213.
[2] K. K. Li, G. Arjavalingam, A. Dienes, and J. R. Whinnery, “Propaga-
tion of picosecond pulses on microwave striplines,” IEEE Trans. Mi- Prasanth Moolakuzhy Narayanan, photograph and biography not available at
crow. Theory Techn., vol. MTT-30, no. 8, pp. 249–254, Aug. 1982. time of publication.

You might also like