You are on page 1of 7

2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

Free-space Measurement Using Explicit, Reference-plane and


Thickness-invariant Method for Permittivity Determination of
Planar Materials
Kok Yeow You1 , Man Seng Sim1, 2 , Hafizah Mutadza1 , Fahmiruddin Esa2 , and Yi Lin Chan2
1
Faculty of Electrical Engineering, Universiti Teknologi Malaysia
Skudai, Johor 81310, Malaysia
2
Faculty of Science, Technology and Human Development
Universiti Tun Hussein Onn Malaysia, Batu Pahat, Johor 86400, Malaysia

Abstract— In this paper, transmission phase-shift method was applied on free-space measure-
ments for thin planar material characterization. The free-space measurement system consists
of two pyramidal horn antennas, coaxial cables, network analyser, sliding guide, two antenna
holders/stands, sample holder and two precision positioners. The free-space measurement sys-
tem was operated in frequency range from 10 GHz to 15 GHz. The planar acrylic samples with
different thicknesses were used for measurement validation. The relative permittivity of those
samples were predicted from the measured reflection/transmission coefficients based on trans-
mission phase shift concept. Before the actual measurement is done, systematic errors due to
multiple reflections along the measurement distance between the horn antenna aperture and
the planar sample surface were calibrated by using Thru-Reflect-Line (TRL) technique. In the
free-space measurements, performance of the horn antenna and the distance between the two
horn antennas played an important role in the measurement set-up. However, the mentioned
measurement set-up factors were usually less discussed in previous literatures. Hence, in this
study, these issues will be discussed in order to improve accuracy and reduce the multi-reflection
in the measurement. Besides, the effect of thickness for the sample in permittivity prediction
was also discussed and analysed in detail. Formerly, the thickness of the planar sample was not
recommended to be more than a quarter of the wavelength in order to avoid phase-shift am-
biguity problems and effects on the accuracy of the broadband relative permittivity prediction.
Here, the phase-shift ambiguity issue due to the sample thickness was solved by the MATLAB
command “unwrap” during the mathematical calculation of the permittivity value. In addition,
this proposed method is sample position-invariant. Hence, the measurement errors due to the
uncertainty of the sample position can be avoided.

1. INTRODUCTION
Microwave free-space measurement method is increasingly popular for the use in large-scale indus-
trial control systems, especially for drying process control systems. To be specific, this method is
contactless, sensitive, nondestructive, large sensing area and suitable for product under test with
different temperatures. The free-space measurements in laboratory are usually used to determine
the constitutive parameters (permittivity, εr , permeability, μr and conductivity, σ) for thin planar
form sample at high operating frequency (broadband) and temperature.
Recently, the transmission phase shift (TPS) method has been applied to calibration-independent
permittivity (εr = εr − jεr ) measurements using rectangular waveguides [1, 2]. In fact, the TPS
method was first used in free-space measurement by Kraszewski and Nelson [3]. However, in [3],
only closed-form equations of εr extraction were proposed, which are not highly accurate to be used
in loss factor, εr measurement for low-loss sample. In this study, the exact εr extraction equations
are used for εr prediction for low-loss samples. The advantages of this method are simple, sam-
ple position-independent, sample thickness-invariant, and non-iterative (explicit), except that this
method is only for non-magnetic samples (μr = 1). Although, commonly used Nicolson Ross Weir
(NRW) method [4, 5] is explicit and capable of determining εr and μr simultaneously, this method
is position- and thickness-dependent. Formerly, the thickness, d of the planar sample should not be
more than a quarter of the wavelength (λ/4) in order to avoid phase-shift ambiguity problems. In
this study, the phase-shift ambiguity issue was solved by the MATLAB command “unwrap” during
the mathematical calculation of the εr value.
In addition, by using NRW, the phase of the measured S-parameters (S11 and S21 ) is required to
be on the position of the calibration reference plane and in consistent position of the sample at the
center between the two horns Fortunately the TPS method is a calibration-independent and sample

222
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

position-invariant technique, which can reduce the complexity of the de-embedding procedures and
also the uncertainty of measurement due to the effect of sample position precision This method
only requires two measurement topologies, namely measurement of S-parameters for air (without
placing sample) and the measurement which repeated by placing a sample as shown in Figure 1. In
this study, the uncertainty in the sample thickness, Δd and phase measurements, Δφ contributing
to the error in the εr determination have been quantitatively evaluated. Besides, issues of distance
selection between transmitting and receiving horns were analysed in term of measurement, which
lacks discussion in previous literatures.

Sample

S11A Air S21A S11S S21S


θ¸
Horn φ21A Horn Horn φ21S Horn

l
d
(a) (b)

Figure 1. Free-space measurements (a) without sample and (b) with thin planar sample.

2. THEORY
2.1. Explicit εr Extraction Formulations
Based on the transmission phase shift and wave propagation theories, the index refraction, n of a
sample under test can be written as:
 
√ 1 φ21A − φ21S
n = εr μr = γo − jα + (1)
βo d

From (1), for non-magnetic samples (μr = 1), the real and imaginary parts of εr can be separated
as [1, 2]:
  
1 φ21A − φ21S 2
εr = 2 γo + − α2 (2a)
βo d
 
2α φ21A − φ21S
εr = 2 γo + (2b)
βo d

where βo = 2πf /c and d are the propagation constant of free space (c = 2.99792458 ms−1 ) and
thickness of the planar sample. The parameters φ21A and φ21S (in radian) are the measured
transmission phase shift of the air (without sample) and sample, respectively. Symbol α is the
dielectric attenuation constant for the sample in nepers/meter.

1.15129254     
α≈− log10 |S11S |2 + |S21S |2 − log10 |S11A |2 + |S21A |2 (3)
d

where |S11S | and |S21S | are the measured linear magnitudes of the reflection coefficient and the
transmission coefficient for the sample, respectively. While, |S11A | and |S21A | are the measured
linear magnitudes of the reflection coefficient and the transmission coefficient for the air (without
sample), respectively. The γ o is the z-component of the propagation constant for air as:

γo = βo 1 − sin2 θ (4)

For far-field measurements, the incident wave from the transmitting horn is assumed to be in
plane-wave form (θ = 0◦ , thus, (4) can be simplified as γo = βo .

223
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

2.2. Uncertainties Analysis for εr


The uncertainties for extracted εr obtained from (2a) and (2b) can be analytically calculated as:

∂εr ∂ε ∂ε


Δεr = Δφ+ r Δd+ r Δα
∂ (φ21A −φ21S ) ∂d ∂α
   
βo +(φ21A −φ21S )/d βo +(φ21A −φ21S )/d 2α
= 2 Δφ−2 (φ21A −φ21S ) Δd− 2 Δα (5a)
βo2 d βo2 d2 βo
and
∂εr ∂ε ∂ε
Δεr = Δφ+ r Δd+ r Δα
∂ (φ21A −φ21S ) ∂d ∂α
   
βo +(φ21A −φ21S )/d 2α (φ21A −φ21S ) 2 φ21A −φ21S
= 2α Δφ− Δd+ 2 βo + Δα (5b)
βo2 βo2 d2 βo d

where Δφ, Δd, and Δα are the uncertainties on the phase shift (in radian), sample thickness
(in meter), and attenuation constant (nepers/meter), respectively. In this study, acrylic sample
(typical: εr = 2.6, εr = 0.049 [2]) is used in uncertainty analysis in order to predict the effects
of errors in measured (φ21A − φ21S ), d, and α to the accuracy of εr and εr . The Δφ and Δα
uncertainties are mainly given by the vector network analyzer (VNA) and the horns specifications.
Whereas, uncertainty of Δd is mainly dependent on the specification of the micrometer/vernier
calipers used and uniformity of the sample thickness. In Figures 2(a) and (b), a deviation of Δεr
respected to the errors of Δφ and Δd are plotted.
From Figure 2, some conclusions can be summarized as:

0.4 Δφ=0.01 rad 0.2 0.13 0.1


Δφ=0.015 rad
Δφ=0.02 rad 0.18 d = 2 mm 0.12 d = 3 mm 0.09 d = 4 mm
0.35 Δφ=0.025 rad
0.11
0.16 0.08
0.1
0.3
0.14 0.07
0.09
Δεr'

0.25 0.12 0.08 0.06

0.07
0.1 0.05
0.2
0.06
0.08 0.04
0.05
0.15 d = 1 mm
0.06 0.03
0.04

0.1 0.04 0.03 0.02


10 12.5 15 10 12.5 15 10 12.5 15 10 12.5 15
f (GHz) f (GHz) f (GHz) f (GHz)
(a)
0.8 0.4 0.2
0.26
Δd=0.1 mm
0.7 Δd=0.2 mm 0.35 0.24 0.18
Δd=0.3 mm
Δd=0.4 mm 0.22
0.16
0.6 0.3
0.2
0.14
0.5 0.25 0.18
d = 1 mm d = 2 mm d = 3 mm d = 4 mm
Δεr'

0.16 0.12
0.4 0.2
0.14 0.1
0.3 0.15 0.12
0.08
0.1
0.2 0.1 0.06
0.08

0.1 0.05 0.06 0.04


10 12.5 15 10 12.5 15 10 12.5 15 10 12.5 15
f (GHz) f (GHz) f (GHz) f (GHz)
(b)

Figure 2. (a) Phase shift noise, Δφ (assume: Δd = 0 and Δα = 0) and (b) thickness tolerance, Δd (assume:
Δφ = 0 and Δα = 0) effects in uncertainty of predicted εr for various thickness of acrylic samples.

224
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

a) The effect of Δφ on Δεr is variable with operating frequency. The Δεr will decrease when
the frequency, f increases.
b) The effect of Δd on Δεr is always constant against the operating frequency, f .
From Equation (5a), typically, the uncertainty Δα only contributes to Δεr less than 10−3 .
In Figure 3, the influence of the Δφ, Δd, and Δα in case of Δεr are observed. Factor Δd does
not have a significant effect on Δεr . On the other hand, Δεr is strongly influenced by the effect of
Δα. Overall, the samples that have thicker sizes, will be able to reduce the uncertainty effects of
Δφ and Δd on Δεr and Δεr .
-4 -3
x 10 x 10
13 13
Δφ=0.01 rad Δα=0.2 Np
12
12 Δφ=0.015 rad Δα=0.4 Np
Δφ=0.02 rad Δα=0.6 Np
11
Δφ=0.025 rad 10 Δα=0.8 Np
10

9 8

Δεr"
Δεr"

8
6
7

6
4
5

4 2
10 11 12 13 14 15 10 11 12 13 14 15
f (GHz) f (GHz)
(a) (b)
-3 -3 -3 -3
x 10 x 10 x 10 x 10
8 4 2.6 2

Δd=0.1 2.4
mm 3.5 1.8
7
Δd=0.2 mm 2.2
Δd=0.3 mm 1.6
6 3 2
Δd=0.4 mm
1.4
1.8
2.5
5
Δεr"

1.6 1.2
2
4 1.4
1
1.5 1.2
3 d = 1 mm 0.8
d = 2 mm 1 d = 3 mm d = 4 mm
1 0.6
2 0.8
0.5 0.6 0.4
10 12.5 15 10 12.5 15 10 12.5 15 10 12.5 15
f (GHz) f (GHz) f (GHz) f (GHz)
(c)

Figure 3. (a) Phase shift noise, Δφ (assume: Δd = 0 and Δα = 0), (b) resolution noise, Δα (assume:
Δd = 0 and Δφ = 0) and (c) thickness tolerance, Δd (assume: Δφ = 0 and Δα = 0) effects in uncertainty
of predicted εr for various thickness of acrylic samples.

3. MEASUREMENT SET-UP
3.1. Experimental Set-up
The free-space measurement set-up consists of two LB-75-10-A model pyramidal horn antennas
manufactured by A-Info Inc., measurement fixtures, and Keysight E5071C vector network analyzer
(VNA) as shown in Figure 4. The technical specification of the LB-75-10-A is tabulated in Ta-
ble 1 [6]. The distance, l between the transmitting and receiving horn apertures is fixed to be 14 cm
and the thin planar sample (with 20 cm width × 20 cm length) is placed between the center of the
horn antennas.
3.2. Horns Distance, l Estimation
The distance, l between the transmitting and receiving horns must be estimated correctly, in
order to optimize the quality and accuracy of measurement. If the distance, l is too far, the
transmitted signal received by receiving horn is small and severe noise will occur in the S-parameter
measurements. On the contrary, if the distance, l is too close, the signal that is transmitted to

225
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

Table 1. LB-75-10-A model pyramidal horn specification [6].

f 3 dB Beamwidth (◦ ) Gain
(GHz) H-plane E-plane (dB)
10 48.94 53.36 10.15
11 50.74 53.36 10.98
12 42.02 46.36 11.65
13 41.02 42.21 12.35
14 39.56 35.99 13.00
15 35.63 35.66 13.58
Planar sample

Receiving horn
Transmitting horn

Micrometer
Micrometer

Figure 4. Measurement set-up.

planar surface sample is not in plane wave form (θ = 0◦ , thus causing less precise determination of
εr for the sample using (2a) and (2b).
To overcome such issues, the minimum distance, l between the transmitting and receiving horns
is based on the size, D of the aperture horn as l = 2D2 /λ [7]. Previously, a spot focusing horn
lens antenna (high gain and narrow beamwidth) was used to minimize diffraction effect, so that
the sample surface area can be smaller and the distance between the horns could be placed far [8].
Alternatively, the rectangular dielectric waveguide (RDWG) method has been used to focus the
field in transmitted direction [9]. Recently, collinear lenses have been used to generate Gaussian
plane wave at the sample location [7].
In this work, the l was determined using monostatic reflection measurement in which the metal
plate was gradually moved far from the horn aperture as shown in Figure 5(a). Figure 5(b) shows the
measured linear magnitude of reflection coefficient, |S11 | versus distance, l between horn aperture
and metal plate. Level |S11 | = 0.2 (reflected signal < 10%) is used as a reference to determine the
minimum distance between the horn aperture and the surface sample. Value of |S11 | has decreased

1
10.5 GHz
|S11| y 12.5 GHz
x 0.8 15.0 GHz

z 0.6
|S11|

Metal plate
0.4

l' 0.2

0
0 5 7 10 15
Distance, l ' (cm)
(a) (b)

Figure 5. (a) Monostatic |S11 | measurement. (b) Linear magnitude of |S11 | versus distance l .

226
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

and reached ∼ 0.2, when l equals 7 cm. After l > 7 cm, the |S11 | seems almost constant with
distance, l . Thus, the l(= 2l ) was determined at 14 cm for our free-space measurement system.
3.3. Calibrations
Two calibrations are done before measurement. To eliminate systematic errors of VNA and coaxial
cables, full two-port open-short-load-thru (OSLT) calibration is applied at the end of the coaxial
cables for port-1 and port-2 using Keysight 85052D calibration kit from 10 GHz to 15 GHz. The
coaxial cables are respectively connected to the transmitting and receiving horn antennas. Then,
two-port thru-reflect-line (TRL) calibration [8] is done using built-in TRL model of E5071C VNA
in order to eliminate multiple reflections between the transmitting and receiving antennas. The
thru standard is configured by keeping the 14 cm of distance between the two antennas. The reflect
standard is obtained by placing a 3 mm thickness of aluminum plate (with 25 cm width ×25 cm
length) at the center of the two horn antennas and the receiving horn is moved backward by
3 mm using precision micrometer. The line standard is achieved by increasing the distance of the
transmitting and receiving antennas by 6 mm (quarter of the wavelength) using micrometer. Thus,
the total distance between the two horns for the line standard is equal to 14.6 cm.
3.4. S-Parameters Measurement
After calibration, the planar sample (with 20 cm width × 20 cm length) is placed at the center
between two horns. The linear magnitude, |S11 |, |S21 |, and the phase shift, φ21 , for air and the
planar sample were measured using VNA from 10 GHz to 15 GHz. Once the measured |S11 |, |S21 |,
and φ21 for air and sample are obtained, then replace into (2a), (2b), and (3) in order to extract
the values of εr and εr for the sample. The acrylic samples with various thickness, respectively,
were measured for validation.
4. RESULTS AND ANALYSIS
Figures 6(a) and (b) shows the extracted εr of acrylic samples using conventional NWR and pro-
posed TPS methods respectively. The least squares method was applied to the εr points in Fig-
ure 6(b), in order to smooth the noisy curve as shown in Figure 6(c). Both methods show significant
errors in extraction εr for thin acrylics. For acrylics thicknesses that are too thin (d ≤ 2 mm), low

d =1 mm d =1.7 mm d =2.6 mm d =3.8 mm d =5.0 mm d =5.8 mm d=1 mm d=1.7 mm d=2.6 mm d=3.8 mm d=5 mm d=5.8 mm d=7.8 mm

3.5
3.5
3
3
2.5
ε'
r

r
ε'

2
2.5

1.5 2
Acrylic (NRW method)
1
Acrylic (TPS method)
1.5
10 11 12 13 14 15 15.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 15.5
f (GHz) f (GHz)
(a) (b)
d=1 mm d=1.7 mm d=2.6 mm d=3.8 mm d=5 mm d=5.8 mm d=7.8 mm 0.2
d=3.8 mm d=5 mm d=5.8 mm d=7.8 mm
3.5
0.1

0
εr"
r
ε'

2.5

-0.1
2
Acrylic (TPS method + Least squares method) Acrylic (TPS method)
1.5 -0.2
10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 15.5
f (GHz) f (GHz)
(c) (d)

Figure 6. (a), (b), (c) Extracted εr and (d) εr versus operating frequency, f for various thicknesses of acrylic
samples.

227
2017 Progress In Electromagnetics Research Symposium — Fall (PIERS — FALL), Singapore, 19–22 November

frequency waves are less sensitive when propagating through the acrylic samples. The coupling
fringing field between the front and back of the surface sample is likely to occur, and the wave
propagates in the thin sample are disturbed by this effect [2]. In addition, the uncertainty of the
small value of d is high.
However, when the thickness of the acrylic sample is thick (d > 4 mm), NRW method fails
to give precise εr values, which is mainly due to the phase ambiguity issue. On the other hand,
TPS method is not influenced by the acrylic that has thick size. Besides, from Table 1 [6], it is
known that at low frequencies, 3 dB beamwidth of horns used are large, thus the diffraction effect
of transmitted wave is higher and may causes the measurements to be scattered. Figure 6(d) shows
the extracted εr versus frequency, f . Obviously, the scatter for the points of εr will decrease, when
the sample thickness increases.
5. CONCLUSIONS
A simple, non-iterative, sample position- and thickness-independent free-space measurement has
been presented. The proposed TPS method is able to determine the εr using two measurements: one
measurement is for the air (without sample) using two horn antennas, and the other for one planar
sample placed at arbitrary position between the two horns. The selection of the distance, l between
the horns, horn specifications, sample thickness, d and VNA performance are the major factors in
free-space measurement system. Likewise other two-port free-space techniques, the TPS method
also presents high uncertainties in measuring small value of εr . Among two-port S-parameter-based
methods, the coaxial waveguides is more precise in the measurement of εr which has a small value.
ACKNOWLEDGMENT
This study was supported by a Research University Grant (GUP) from Universiti Teknologi Malaysia
under project number Q.J130000.2523.15H30 and the Ministry of Higher Education of Malaysia
(MOHE).
REFERENCES
1. You, K. Y., Y. S. Lee, L. Zahih, M. F. A. Malek, K. Y. Lee, E. M. Cheng, and N. H. Khamis,
“Dielectric measurements for low-loss materials using transmission phase-shift method,” Jurnal
Teknologi, Vol. 77, No. 10, 69–77, 2015.
2. You, K. Y., “Effects of sample thickness for dielectric measurements using transmission phase-
shift method,” International Journal of Advances in Microwave Technology. Vol. 1, No. 3,
64–67, 2016.
3. Kraszewski, A. and S. Nelson, “Study on grain permittivity measurements in free space,” J.
Microwave Power & EE., Vol. 25, No. 4, 202–210, 1990.
4. Nicolson, A. M. and G. F. Ross, “Measurement of the intrinsic properties of materials by
time-domain techniques,” IEEE Trans. Instrum. and Meas., Vol. 19, No. 4, 377–382, 1970.
5. Weir, W. B., “Automatic measurement of complex dielectric constant and permeability at
microwave frequencies,” Proc. IEEE., Vol. 62, No. 1, 33–36, 1974.
6. A-Info, LB-75-10 10.0–15.0 GHz Standard Gain Horn Antenna: Technical Specification. Avail-
able: http://www.ainfoinc.com/en/pro pdf/new products/antenna/Standard%20Gain%20Ho-
rn%20Antenna/tr LB-75-10.pdf.
7. Sabielny, M., “An investigation into the feasibility of S-parameter measurements at X-band
frequencies with Gaussian beam techniques,” Proceedings of EADS Conference, 2003.
8. Ghodgaonkar, D. K., V. V. Varadan, and V. K. Varadan, “Free-space measurement of complex
permittivity and complex permeability of magnetic materials at microwave frequencies,” IEEE
Trans. Instrum. Meas., Vol. 39, No. 2, 387–394, 1990.
9. Abbas, Z., R. D. Pollard, and R. W. Kelsall, “Complex permittivity measurements at Ka-
band using rectangular dielectric waveguide,” IEEE Trans. Instrum. Meas., Vol. 50, No. 5,
1334–1342, 2001.

228

You might also like