Abstract— A novel time domain measurement technique is desired bandwidth (BW), and other essential requirements [2].
proposed to facilitate the simultaneous measurement of electri- The microwave and RF material characterization methods can
cal properties (complex relative permittivity) and geometrical be broadly classified as resonant and nonresonant types. The
parameters (thickness) of the material under test (MUT). The
overall process is noninvasive and noncontacting, which uses the resonant methods are found to have better sensitivity as com-
measured scattering data of the MUT in the equivalent time pared with the nonresonant methods, but are usually suitable
domain or spatial domain. The effective time domain scattering for low loss materials providing measurement at some specific
data are employed to detect the primary and secondary peaks of frequency point [3]–[6]. On the other hand, the nonresonant
the overall reflection and transmission coefficients. To this end, methods can usually provide wideband RF characterization
a novel algorithm is proposed to obtain the complex permittivity
and thickness of the MUT in terms of extracted reflection and with good amount of accuracy, and are valid for materials
transmission power peaks. From the practical point of view, the with moderate losses [6]. The nonresonant methods, which
main advantage of the proposed scheme is that one avoids the mostly require measurements of reflection and transmission
complicated calibration procedure normally required to define coefficient data, can be broadly classified into the guided trans-
the reference plane. For increasing the accuracy of the overall mission line [7]–[10], and the free-space method [11]–[24].
reconstruction process, an automated optimization procedure
based on parameter sensitivity analysis is proposed, which uses It may be noted that the free-space methods are quite con-
standard time gating procedure to implement the corresponding venient from the user point of view especially in the higher
direct problem. The proposed technique is validated by extracting frequency region, as practically no sample preparation is
the relative permittivity, the dielectric loss (effective conduc- required. The free-space methods can, in principle, be further
tivity), and the thickness of various standard materials, such classified under the frequency (spectral) domain and time
as polyethylene, Plexiglas, PVC, mortar, nylon, and so on, and
comparing the extracted data with their values available in the (spatial) domains depending upon the methodology adopted
literature. for the reconstruction of dielectric properties in terms of
measured data. The frequency domain free-space methods
Index Terms— Dielectric material characterization, effective
conductivity determination, Fourier transform, free-space usually require the complicated calibration procedure, which
measurement, material measurement, relative permittivity sometimes becomes difficult due to nonavailability of accurate
measurement, remote material characterization, time domain standards (conventional standard, i.e., open short through) for
reflectometery, wave propagation, wideband. nonguided media.
The traditional time domain material testing methods are
I. I NTRODUCTION
usually based on analytical formulations, and appear to provide
expressed as
where
PL1 (d1 ) two way path loss between the
interface I1 and I f ;
R1 (εr0 , εrt , σ0 , σt ) local reflection from the interface I f
of MUT;
τ1 (εr0 , σ0 , d1 ) time corresponding to primary
reflection from interface I f ; u(t) unit
Fig. 2. Bounce diagram for the geometry shown in Fig. 1. A transient wave step function.
undergoes an infinite number of transmission and reflection at the interfaces.
The second reflection, i.e., the reflection from the back
interface of MUT (Ib ), observed by TR1 , as shown in Fig. 2,
opposite side of the interface at a distance d2 from the MUT. can also be expressed as
The incident, reflected, and transmitted power components
are represented as Pinc , Pr , and Pt , respectively. It is to be Pr2 (εr0 , εrt , σ0 , σt , d1 , dt , Pinc ) = PL2 R2 A2 u(t − τ2 )Pinc
noted here that the background medium is considered to be (4)
lossless and dispersion-less throughout the formulation, which
can represent the free space in most of the cases. In this paper, where
various primary and secondary reflection and transmission
PL2 (d1 , dt ) two way path loss between the
power quantities are directly taken from the vector network
interface I1 and interface Ib ;
analyzer (VNA), which basically makes use of the complex
R2 (εr0 , εrt , σ0 , σt ) two transmission (i.e., forward
scattering data (both magnitude and phase) to convert them
and reverse way) at the
into the equivalent time domain using the standard inverse
interface I f and a reflection
fast Fourier transform (IFFT) routines. Hence, although the
from interface Ib ;
VNA makes use of both magnitude and phase information of
A2 (εrt , σt , dt ) material absorption;
the scattering data, the present formulation makes use of the
τ2 (εr0 , εrt , σ0 , σt , d1 , dt ) time corresponding to the
power quantity only for determining the material properties
second reflection.
and the thickness of the test specimen.
In a similar manner, the primary and secondary transmitted
B. Algorithm Description power peaks can be expressed as
The typical reflection and the transmission bounce diagram
Pt 1 (εr0 , εrt , σ0 , σt , d1 , dt , d2 , Pinc ) = PL3 R3 A3 u(t − τ3 )Pinc
in the time domain representing the reflected and transmitted
powers from the test media are shown in Fig. 2, where (5)
S1 and S2 represent the antenna/sensor elements being respon- Pt 2 (εr0 , εrt , σ0 , σt , d1 , dt , d2 , Pinc ) = PL4 R4 A4 u(t − τ4 )Pinc
sible for transmission and reception of the RF power. (6)
The reflection from or transmission through the test
media (i ) can be expressed in the following generalized
where
form:
PL3 (d1 , dt , d2 ) and
i = PLi Ri Ai u(t − τi )Pinc . (2) PL4 (d1 , dt , d2 ) path losses;
The term PLi in the above expression describes the propa- R3 (εr0 , εrt , σ0 , σt ) transmission factor at the interface
gation condition, i.e., the path loss, which models the actual I f and Ib successively;
propagation condition in different wave propagation situations R4 (εr0 , εrt , σ0 , σt ) transmission factor at the interface
viz., the planar, the cylindrical, and the spherical [19]. Here, I f and Ib and reflection factor at the
for the sake of simplicity, a plane EM wave incident nor- interface Ib and I f , respectively;
mally over the test media is assumed. The term Ri considers A3 (εrt , σt , dt ) and
various reflection and transmission factors when the wave A4 (εrt , σt , dt ) material absorption;
propagates through the test media whereas Ai models the τ3 (εr0 , εrt , σ0 , σt ,
effective absorption through the test media. The term u(t − τi ) d1 , dt , d2 ) and
is the unit step function, which signifies the received waveform τ4 (εr0 , εrt , σ0 , σt ,
with τi representing the arrival time of various reflection and d1 , dt , d2 ) time corresponding to the first and
transmission signals at sensor/antenna. the second transmission, respectively.
The first reflection sensed by the receiver, i.e., reflection The magnitude of reflection and transmission peaks
from the front interface (I f ), observed by S1 in Fig. 2 can be described earlier in terms of incident power assuming a plane
AKHTER AND AKHTAR: FREE-SPACE TIME DOMAIN POSITION INSENSITIVE TECHNIQUE 2397
wave propagation can also be written in the following form: from the equivalent time domain representation of
S-parameter data obtained in Step 1.
Pr1 ∼
= |t | Pinc
2
(7)
Step 3: Finally, with the help of (10)–(14), the values of
1
Pr2 ∼
= |t |2 (1 − |t |2 )2 exp −2σt η0 √ dt Pinc (8) the relative permittivity (εrt ), the effective electrical
εrt conductivity (σt ), and the thickness (dt ) of the test
1
Pt 1 ∼
specimen are determined.
= (1 − |t |2 )2 exp −σt η0 √ dt Pinc (9)
εrt The proposed approach makes use of time corresponding to
1 occurrence of the first transmission peak in order to identify
Pt 2 ∼
= (1 − |t |2 )2 |t |4 exp −3σt η0 √ dt Pinc (10) various power terms required to reconstruct the dielectric
εrt
properties and thickness of the test specimen. The problem
where η0 = 377
(free-space intrinsic impedance) and Pinc is of background reflections, if any, can be resolved with the
the reference incident power, as shown in Fig. 1. Pr1 and Pr2 help of absorbing boundaries by facilitating the measurements
represent the powers reflected from front and back interface in the anechoic environment as discussed earlier. The time of
of the test slab whereas Pt 1 and Pt 2 represent the direct path occurrence of various power peaks can be verified by placing
transmission and the consecutive transmission in time. a metal plate at the material interface. It should be, however,
Equations (7)–(10) are taken as approximate in the sense noted that the metal plate is required here only for verification
that the term PLi in all these equations is taken as 1 indicating and not for any reference measurement.
a perfect plane wave or a TEM wave. Now, it is well known
that the perfect plane wave concept is always an approxima- C. Limitations on Bandwidth and Number
tion, since under practical situation, the phase front over the of Frequency Samples
full cross section of the test specimen might not represent
The BW (pulse rise time) and the number of frequency
an ideal plane. Equations (7)–(10) would be exact when the
sample are very crucial parameters in order to get accurate
measurements are carried out in a guided medium representing
reconstruction of the properties of test specimen using the
a TEM wave.
proposed approach. The proposed technique makes use of
From (7)–(10)
various primary and secondary reflection/transmission peaks
Pr1 Pt 2 |t |2 to determine the properties of MUT, and the time separation
= =k (11) between these peaks directly depend on BW of the measured
Pt 1 Pr2 (1 − |t |2 )
scattering data. Therefore, a compromise has to be made here
where k is a constant, which can be determined using various between the thickness of MUT and BW of the scattering data
reflection and transmission power peaks. Equation (11) can such that various reflection and transmission peaks can be
then be solved to determine the magnitude of the reflection detected in the equivalent time domain. In other words, a
coefficient with the constraint that |t | < 1. minimum BW is required to resolve two adjacent interfaces
Once the magnitude of the local reflection coefficient |t | (i.e., axial resolution) of the test specimen. The axial resolution
is determined, the relative permittivity and the thickness of δz in the range direction is determined approximately by the
test material can determined with the help of the following signal BW of the measured RF signal [27]
expression:
c 10.71
1 + |t | 2 dmin = δz (mm) ≈ = √ . (15)
εrt = (12) 2 × BW εrt
1 − |t |
Accordingly, a large signal BW results in an equivalent short
c ∗ τ
dt = √ (13) pulse duration and hence provides a high range resolution.
2 εrt In most of the network analyzer, the number of points for the
where τ represents the time delay between Pr1 and Pr2 , and time domain representation is directly linked to the number
c is the velocity of EM wave in free space. After determining of sample/points in frequency sweep (equal in most cases).
the permittivity and the thickness, the conductivity of the slab Therefore, reducing the number of points in the frequency
can be computed using domain also results into less number of points in the time
domain representation [28], which ultimately limits the ability
1 ε0 εr Pr1
σt ∼
= ln (1 − |t |2 )2 or of the procedure to detect and resolve various peaks in the
2dt μ0 Pr2
delay domain. For example, if the spectral domain measure-
1 ε0 εr Pt 1
= ln |t |4 . (14) ments are carried out over a BW, then the procedure cannot
2dt μ0 Pt 2 properly characterize a material having relative permittivity εrt
The derivation of (14) is facilitated with the help of (7)–(10). if its thickness is below the minimum value given in (15).
The details of the proposed inversion algorithm can be The intermediate frequency (IF) BW is another important
explained using the following steps. parameter in the VNA measurements, which basically cor-
Step 1 (Optional): The measured scattering data (Fig. 1) responds to the BW of the IF filter. The smaller IF BW
are converted into equivalent time domain using the implies less broadband noise, which increases the effective
IFFT routines of the VNA. dynamic range available for measurement. For lower permit-
Step 2: The second step involves the extraction of various tivity samples, the magnitude of the reflection coefficient is
parameters, such as Pr1 , Pr2 , Pt 1 , Pt 2 , and τ , low, and hence, the detection of various higher orders peaks,
2398 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 65, NO. 10, OCTOBER 2016
TABLE I
R ECONSTRUCTION OF MUT’s PARAMETERS
such as Pr2 and Pt 2 , becomes difficult. In order to detect these The following observation can be made from Table I.
peaks, the reference noise label should be considerably low, 1) The proposed technique can estimate the relative per-
which can be achieved by lowering the value of IF BW during mittivity of various standard samples having different
measurement. It should, however, be mentioned that lowering dielectric contrasts with respect to the background media
the IF BW also slows the overall measurement process. having a typical accuracy of 98% or more.
2) The thickness of the various test samples can also be
III. N UMERICAL VALIDATION accurately determined with a typical accuracy of 98%
A. Simulation Procedure or more.
Before testing the validity of the proposed technique against 3) The proposed technique can also be used to evaluate the
the experimental data, it is tested against the independent effective conductivity of various lossy dielectrics, and
simulation data to closely observe the accuracy and range of the typical accuracy can be quantified as follows.
validity of the proposed technique. For this purpose, the CST a) The lossy samples with σt 10−3 or more can be
microwave studio is used to simulate the actual configuration, characterized with a typical accuracy ranging from
as shown in Fig. 1. The S-parameter is calculated for the 99% to 97%.
frequency range of 26–40 GHz. In order to satisfy the plane b) The dielectric samples with σt 10−4 or less can
wave condition, the boundary conditions are chosen, such be characterized with a typical accuracy ranging
that the structure supports the TEM mode of propagation as from 90% to 85%.
its dominant mode. In the present case, the parallel plate The sensitivity analysis has been done here for the MUT’s
waveguide boundary condition (i.e., electric and magnetic parameter in order to find out the applicability of the proposed
boundaries along the mutually perpendicular faces) is used to technique, and to obtain range of parameters values for which
facilitate the TEM mode of propagation. The waveguide ports the optimum accuracy can be achieved.
are kept at a distance of 20 × λ f c from the MUT surface in
both the directions, where λ f c is the free-space wavelength B. Sensitivity and Error Analysis
at a center frequency of 33 GHz of band of interest. The In order to obtain the sensitivity of the proposed method, a
thickness dt of the MUT is ranging from 25–30 mm while the detailed parametric analysis is carried out in this paper with
cross-sectional dimensions, i.e., length and width, are taken respect to the electrical properties, i.e., the relative permittivity
as 15 × λ f c . and the effective electrical conductivity. To this end, the test
The itemized inversion algorithm presented in Section II has sample’s electrical properties parametric variations against the
been applied on the simulated S-parameters. It is to be noted measurable quantities, i.e., Pr1 , Pr2 , Pt 1 , Pt 2 , and τ , are
here that the number of frequency samples plays an important closely analyzed. The variations in the primary/secondary
role as described earlier, and it is desirable to have more reflected and transmitted powers, i.e., Pr1 , Pt 1 /Pr2 , Pt 2 , with
number of frequency samples for accurate peak detection. The respect to the relative permittivity εrt of the MUT are pre-
number of frequency samples and the mesh cells considered sented in Fig. 3.
in the current simulation are 10 000 and 2 862 250 (after In Fig. 3, the associated scales of primary and secondary
symmetry plane reduction), respectively. For the validation of reflected and transmitted powers are represented along the left
the proposed concept, a number of standard samples (with and the right sides of the y-axis, respectively. The following
low-to-high dielectric contrasts with respect to free-space observations can be made from the detailed analysis.
background) were used from the CST material library. The 1) The proposed algorithm is well suited for low as well as
actual and the reconstructed relative permittivity, effective high dielectric contrast objects. However, the sensitivity
conductivity, and thickness of various standard materials are is more for samples with low dielectric contrast as
tabulated in Table I. The reconstructed material properties compared with those having higher dielectric contrasts.
shown in Table I are obtained directly using the proposed set 2) The powers variation graph can play an impor-
of closed form equations without any optimization. tant role during the optimization process (discussed
AKHTER AND AKHTAR: FREE-SPACE TIME DOMAIN POSITION INSENSITIVE TECHNIQUE 2399
Fig. 5. Assessment of the proposed scheme for different sample thicknesses. (a) Comparison of actual and reconstructed values of relative permittivity and
thickness. (b) Comparison of actual and reconstructed effective conductivity (test sample: Arlon 600, εrt = 6, and σt = 0.01168).
TABLE II
M EASURED A SSESSMENT OF VARIOUS MUT’s PARAMETERS
input to the corresponding direct problem of the optimization that while utilizing the optimization procedure under practical
scheme. The formulation of direct problem is done here using scenario, the time gating function of the VNA has been used
the standard T-matrix approach. The direct problem here is an during the measurement procedure to collect the scattering
independent MATLAB program that calculates the scattering data termed as [S11 , S21 ]m in Fig. 7. The bandpass time gating
parameters of the three layer problem, i.e., air–dielectric–air. is then performed, which considers the primary and secondary
It is to be noted here that the accuracy of the proposed reflection/transmission peaks, i.e., Pr1 , Pt 1 /Pr2 , Pt 2 . Finally,
optimization scheme strongly depends on the accurateness with the deployment of time gating, the existing measurement
of direct problem formulation. It should be mentioned here state is converted into three layered media, i.e., air–MUT–air,
2402 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 65, NO. 10, OCTOBER 2016
Fig. 8. Junction and transmission line representation of time gated domain of discontinuous media.
as shown in Fig. 8. Now, in order to implement the correspond- or Pt 1 and Pt 2 , and τ1 = τ1 − τg_start and τ3 = τg_stop,
ing direct problem, the electrical and geometrical properties of where τg_start and τg_stop are the starting and stop time
each layer should be known in advance. corresponding the bandpass gate function, respectively.
The initial values of these properties are obtained using 5) The S-parameter of the corresponding direct problem
the earlier proposed analytical time domain procedure. The termed here as [S11 , S21 ]s is then calculated with the
procedure regarding the increment/decrement of the electrical help of T -matrix approach [35]. The number of fre-
and geometrical properties is given in Fig. 9. The important quency sample should be kept equal to the measured
points regarding the proposed algorithm can be itemized as data samples so that they can be compared.
follows. 6) The calculated S-parameters are then compared with the
1) In order to apply the direct problem formulation over measured time gated S-parameters and accordingly the
the reduced spatial domain, the parameters, such as d1 , objective function, i.e., the goal of the optimization,
d2 , εr0 , σ0 , should first be determined using the earlier expressed as the mean square error between the mea-
proposed analytical time domain procedure. sured (superscript m) and the simulated (superscript s)
2) The decision regarding the increment/decrement of the S-parameters, is defined using the following expression:
electrical and geometrical properties is taken using a ⎧
⎫
methodology given in Fig. 8. ⎪
⎪ re S m (i ) − S s (i ) 2 ⎪
⎪
Nf ⎪⎪ 11
m 11 ⎪
2
⎪
3) The background medium in the present situation is ⎨
+ img S11 (i ) − S11 (i )
s ⎬
considered to be air. ∃=
⎪ + re S m (i ) − S s (i ) 2 ⎪ ⎪
i=1 ⎪⎪
4) The thicknesses of background front and back ⎪
⎩ 21m 21
2
⎪
⎪
⎭
+ img S21 (i ) − S21 (i )
s
layer d1 and d2 shown in Fig. 8 are calculated with the
help of the following expressions:
where N f represents the number of frequency samples.
c × τ1 7) If the mean square error is in acceptable limit, the
d1 = (16) optimization routine will automatically stop. Otherwise,
2
τ it will update the variable optimization variable, i.e., εrt
u,
d2 = τ3 − × c − d1 (17) σt , and dt . The smart variable update, a subroutine-
u u
2
based process, which is responsible for updating
where c is the velocity of light in free space, τvalues the optimization variables after each step, is shown
represent the time separation between Pr1 and Pr2 , in Fig. 8.
AKHTER AND AKHTAR: FREE-SPACE TIME DOMAIN POSITION INSENSITIVE TECHNIQUE 2403
TABLE III
M EASURED A SSESSMENT OF VARIOUS O PTIMIZED MUT’s PARAMETERS
[33] A. K. Jha and M. J. Akhtar, “A generalized rectangular cavity approach Mohammad Jaleel Akhtar (S’99–M’03–
for determination of complex permittivity of materials,” IEEE Trans. SM’09) received the Ph.D. and Dr. Ing.
Instrum. Meas., vol. 63, no. 11, pp. 2632–2641, Nov. 2014. degrees in electrical engineering from
[34] C. Yeh and F. I. Shimabukuro, “Circular dielectric waveguides,” in the Otto-von-Guericke University of
The Essence of Dielectric Waveguides. Boston, MA, USA: Springer, Magdeburg, Magdeburg, Germany, in 2003.
2008. He was a Scientist with the Central Electronics
[35] M. J. Akhtar, L. E. Feher, and M. Thumm, “Noninvasive procedure Engineering Research Institute, Pilani, India, from
for measuring the complex permittivity of resins, catalysts, and 1994 to 1997, where he was involved in the
other liquids using a partially filled rectangular waveguide structure,” design and development of high power microwave
IEEE Trans. Microw. Theory Techn., vol. 57, no. 2, pp. 458–470, tubes. From 2003 to 2009, he was a Post-Doctoral
Feb. 2009. Research Scientist and a Project Leader with the
Institute for Pulsed Power and Microwave Technology, Karlsruhe Institute
of Technology, Karlsruhe, Germany, where he was involved in a number of
projects in the field of microwave material processing. In 2009, he joined
the Department of Electrical Engineering, IIT Kanpur, Kanpur, India, where
he is currently an Associate Professor. He has authored two books, two
Zubair Akhter (S’13) received the bachelor’s
book chapters, and has authored or co-authored over 100 papers in various
degree in electronics and instrumentation engi-
peer-reviewed international journals and conference proceedings. He holds
neering from Anand Engineering College Agra,
one patent on coplanar based RF sensors. His current research interests
Agra, India, in 2008, and the M.Tech. degree in
include RF, microwave and THz imaging, microwave nondestructive testing,
RF and microwave engineering from IIT Roorkee,
RF sensors, functional materials, wideband electromagnetic absorbers, UWB
Roorkee, India, in 2011. He is currently pur-
antennas for imaging, and design of RF filters and components using the
suing the Ph.D. degree in electrical engineering
electromagnetic inverse scattering.
with IIT Kanpur (IITK), Kanpur, India, with a
Dr. Akhtar is a fellow of the Institution of Electronics and
focus on microwave imaging, material characteri-
Telecommunication Engineers, New Delhi, India, and a Life Member
zation, non-destructive testing of materials, through
of the Indian Physics Association and the Indo-French Technical Association.
wall imaging and ultra-wide antennas for imaging
He is a recipient of the CST University Publication Award in 2009 from the
applications.
CST AG, Darmstadt, Germany. He served as a Chair of the IEEE Microwave
He is currently involved in cost effective solutions for microwave imaging
Theory and Techniques Society-S Uttar Pradesh Chapter from 2013 to 2015,
of concealed object, especially for homeland security applications. He has
and the Vice-Chair of the IEEE Uttar Pradesh Section in 2015.
authored or co-authored over 20 scientific contributions published in peer-
reviewed journals and various admired international conferences. His current
research interests include developing a cost-effective microwave millimeter
wave imaging setup for real-time imaging of concealed object where objects
can be classified on the basis of their dielectric signatures rather that its
physical parameters, such as shape, and size.
Mr. Akhter is the Founder and a Chair of the IEEE Antennas and Propaga-
tion Society Student Branch Chapter IITK, Uttar Pradesh Section, India.