See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/318740474

Interdigitated Sensors: A Design Principle for Accurately Measuring the

Permittivity of Industrial Oils

Article  in  IEEE Sensors Journal · July 2017

DOI: 10.1109/JSEN.2017.2732942

CITATIONS READS

3 249

4 authors, including:

Alex Risos Arvid Hunze

University of Auckland Victoria University of Wellington
9 PUBLICATIONS   15 CITATIONS    32 PUBLICATIONS   239 CITATIONS   

SEE PROFILE SEE PROFILE

Gideon Gouws
Victoria University of Wellington
40 PUBLICATIONS   105 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Alex Risos on 17 February 2018.

The user has requested enhancement of the downloaded file.

6232 IEEE SENSORS JOURNAL, VOL. 17, NO. 19, OCTOBER 1, 2017

Interdigitated Sensors: A Design Principle for

Accurately Measuring the Permittivity
of Industrial Oils
Alex Risos, Nick Long, Arvid Hunze, and Gideon Gouws

Abstract— An accurate measurement of the relative permit- TABLE I

tivity, ε r , of industrial oils using interdigitated dielectrometry T YPICAL VALUES OF THE R ELATIVE S TATIC P ERMITTIVITIES
sensors (IDS) is challenging due to the parasitic capacitance FOR S ELECTED O ILS AT R OOM T EMPERATURE
caused by the IDS spatial structure. In this paper, the investigated
IDS structure consists of a substrate, a ground plane, an insulat-
ing layer below the electrodes and a solder mask above the elec-
trodes having a periodicity of λ. This structure was simulated to
determine the influence on the expected values of capacitance C,
sensitivity S, and deviation D (accuracy) for typical values of ε r
for industrial oils. It is shown that for particular structure ratios
and values of ε r , the parasitic capacitance can be compensated
in order to provide an accurate measurement of ε r . A standard
rapid and cost efficient printed circuit board (PCB) technology
is used to manufacture an IDS based on the simulations to verify
the expected high accuracy in practice. Experiments conducted
with this IDS structure on both new and degraded oils showed a
high accuracy ≈99.5%, verifying the simulations. These results
have the potential to reduce the number of IDS design iterations
in order to carry out highly accurate and precise measurements.
This may go beyond the in-situ monitoring of oil degradation,
such as a number of related sensing applications employing
interdigitated sensor platforms.
Index Terms— Interdigitated dielectrometry sensor (IDS),
permittivity, tan delta, dissipation factor, acidity, oxidation,
aging/degradation, transformer/lubricant/hydraulic oil, in-situ/
online measurement.

I. I NTRODUCTION

O ILS play many roles in industry. They are used as

lubricants, hydraulic fluids, or as insulators and cooling
fluids in transformers. Particularly in the case of trans-
former oils, the oil inevitably degrades with service time,
losing the desired characteristics. This deterioration is linked
to the increase in oxidation, acidity, moisture and sludge
Fig. 1. (a) A cross section view showing the general structure of the IDS.
(b) FEM simulation geometry of the IDS and boundary conditions. Depicted
are the relative permittivities and distances for each layer with the
IDS surrounded at a distance by a grounded boundary. Electric field
lines (dotted) which contribute to the sensing are shown schematically.

Manuscript received June 30, 2017; accepted July 20, 2017. Date of publi-
cation July 27, 2017; date of current version September 8, 2017. This work
was supported in part by Ministry of Business, Innovation and Employment
New Zealand acquired by the Robinson Research Institute as part of the
Victoria University of Wellington under Grant RTVU 1304 and in part by the
ETEL Limited, Auckland. The associate editor coordinating the review of this
paper and approving it for publication was Prof. Tarikul Islam. (Corresponding
author: Alex Risos.)
A. Risos is with the School of Chemical and Physical Sciences, Victoria
University of Wellington, Wellington 6012, New Zealand, and also
with the Robinson Research Institute, Victoria University of Wellington,
Lower Hutt 5010, New Zealand (e-mail: alex.risos@protonmail.com).
N. Long and A. Hunze are with the Robinson Research Institute,
Victoria University of Wellington, Lower Hutt 5010, New Zealand (e-mail:
nick.long@vuw.ac.nz; arvid.hunze@vuw.ac.nz).
G. Gouws is with the School of Engineering and Computer Sciences,
Victoria University of Wellington, Wellington 6012, New Zealand (e-mail:
gideon.gouws@vuw.ac.nz).
Digital Object Identifier 10.1109/JSEN.2017.2732942
content, affecting its lubricating or electrical breakdown
properties [1]–[3]. Thus oil quality monitoring can be a key
element for predicting transformer failures and is therefore
highly desirable in the electrical distribution industry. A mea-
sure of the relative permittivity, εr , has been shown to be
a good indicator of the oil condition [4]. The standard for
insulating liquid measurements [5] specifies the procedure to
measure εr for monitoring the oil condition. It is known that
the εr value of oils increase with service time [1], [3], [5]–[10],
so that a measure of εr in theory enables the real time health
monitoring of oil. Table 1 provides an overview of industrial
oils, their application and nominal εr values.
The standard process in industry determining the condi-
tion of an oil via εr is a time consuming and thus costly

1558-1748 © 2017 Crown Copyright

RISOS et al.: IDS: A DESIGN PRINCIPLE FOR ACCURATELY MEASURING THE PERMITTIVITY OF INDUSTRIAL OILS
procedure [5]. Co-axial probes are commonly used for mea-
suring the permittivity. However, they are used at microwave
frequencies to determine the reflection coefficient for lossy
mediums such as water [16]. This method requires a prior
calibration due to the two wire measurement principle causing
parasitic effects and has been reported to exhibit a maximum
accuracy of 94% at 1 GHz and 80% at 2.45 MHz [17].
However, it has been shown that εr can potentially be measured
with interdigitated dielectrometry sensors (IDS) [18]–[22] in
order to facilitate an on-line measurement system. The use
of IDS benefits from a capacitance scalability (i.e. using
more electrodes to suit the measurement system) and four
wire impedance measurement principle. This is especially
true when measuring impedances of insulating liquids at line
frequencies (∼50 Hz) as per the defined standard [5]; larger
capacitances due to the material under test help to improve
the signal to noise ratio in the measurement. Furthermore,
a four wire measurement system is designed to be free of
lead impedances through e.g. BNC measurement connectors.
Here, we particularly show how the IDS technology can be
improved by reducing parasitic capacitances using a special
design.
The insight provided by this work may be useful beyond
the application of IDS structures for the measurement
of oil degradation, as these sensor platforms has been
shown by many workers to be suitable for measurements
in a range of physical, chemical and biological sensing
applications [23]–[27].
The value of εr can be readily determined by the ratio of
the IDS capacitance with oil vs. the IDS capacitance with air
where εr = Coil /Cair as εr,air ∼
= 1.
However, the measured value of Coil can be a result of
contributions from the dielectric liquid, the sensor structure
as well as the sensor environment. The value of permittivity
calculated for the oil will thus be influenced by parasitic
capacitances of each measurement for oil Cparasitic, and for

air Cparasitic so that

εr = (Coil + Cparasitic )/(Cair + Cparasitic ), (1)
where Coil/air = (1/V )ε0 ∫s1 Ed A with s1 is the upper
surface of the V0 electrode and Cparasitic = (1/V )ε0 ∫s2 Ed A
where s2 is the lower surface of the V0 electrode [28]. The
electrostatic potential between the electrodes was V = 1 in
the simulations in section II, E is the electric field and ε0 is
the permittivity of free space.
Indications for such parasitic capacitances of IDS were
observed in previous work [20] and [29]. The expected
increase of εr during the oil’s lifetime is generally about +5%
to +10% [7]–[10], [30], which can easily be swamped by the
parasitic capacitance in the system. A more accurate IDS is
thus needed to allow the measurement of a true value of εr
and provide information on the state of degradation in an oil
over time.
This paper identifies the parasitic capacitances of an
IDS in order to design an optimised IDS using printed circuit
board (PCB) technology measuring εr accurately. A cross
sectional view of the IDS structure is shown in Fig. 1a.
This model is transferred to the FEM simulation using the
6233
simulated geometry depicted in Fig. 1b having sharp electrode
boundaries. The electric field lines (dotted lines) represent
schematically the sensing mechanism; these electric field lines
penetrate the oil as well as dielectric layers, causing parasitic
capacitances [20]. The IDS structure employs a common
ground plane as this is believed to reduce major parasitic
capacitances (GND = 0 V ) [20], [31].
The electrodes “VAC ” and “V0 ” with a periodicity of λ are
separated from this ground layer by an insulator layer with
distance dins. A soldermask (an epoxy resin) is a standard
protection coating for PCBs, as for IDS in a liquid, certain
dielectrics can dissociate due to electrolysis or react with
the electrode’s surface in other ways. In order to investi-
gate its effects of the IDS figures of merits the soldermask
was considered with thickness dSM in the simulations. This
structure was simulated with the aid of the finite element
method (FEM) to determine the impact of the spatial structure
on the capacitance C by varying the ratios dins/λ, and dSM /λ
while changing εr .
Simulations were performed for a structure of
20 electrodes (10 λ). The first figure of merit considered
in the simulation is the capacitance C, which is important
for maximizing the signal to noise ratio (S/N). From this
capacitance, the figure of merit sensitivity S is defined as:
S = CIDS /εr, IDS . (2)
The deviation D from a reference value (accuracy) is
determined by the ratio of the IDS permittivity, εr,IDS , over
the reference value εr,ref :
D = (εr,ref − εr,IDS )/εr,ref , (3)
with the measured permittivity by the IDS given by:
εr,IDS = Coil,IDS /Cair,IDS (4)
and the reference value, measured as:
εr,ref = Coil,ref /Cair,ref . (5)
The value εr,ref is used in section III and was obtained via
a plane parallel test cell exhibiting a standard deviation and
accuracy of λ ≤ 0.01%, as shown in Fig. 9. It represents
a typical value of new, unused paraffinic oils as commonly
used as coolants in an electrical environment [2], see Table 1.
Thus it was used by the authors for testing and verifying the
IDS simulations.
II. FEM S IMULATIONS WITH C OMSOL M ULTIPHYSICS
The simulations in Comsol Multiphysics version 4.2 used
electrostatics for a simplification as the conductivity for oil
at 50 Hz can be neglected [18]. The electrostatics module
computes the potential field based on the Laplace equation by
applying Dirichlet boundary conditions. The resulting surface
charges on the electrodes VAC and V0 are used to compute
the capacitance. The geometry in Fig. 1 depicts the Dirichlet
FEM boundary held at zero potential (GND). This surrounds
the IDS structure with spacing to the grounded structure
of 12.5λ. This is mimicking a grounded plane to left and
6234
to the right at infinity. Here, the deviation D in the sim-
ulations was obtained similar to Equation 3 through D =
(εr − εr,IDS )/εr , where εr,IDS is the permittivity calculated
from Coil,IDS /Cair,IDS as a result of the FEM simulation with
the IDS structure. Therefore a comparison of the defined
permittivity as entered in the simulation as material under test
above the IDS electrode structure (see Fig. 1) εr was possible.
The width of the electrode and the gap between each
pair of electrodes were equal as adopted from previous
designs [13], [18]–[21], [28], [31], [32]. The electrodes were
set to a potential of VAC = 1 V and V0 = 0 V . The
oil was located above the IDS geometry. The thickness of
the soldermask was modeled for values of dSM /λ = 0
to 0.1 and the insulator distance was varied from dins/λ = 0
to 0.2. The oil layer, doil = h − (dSM + dins) ∼ = 25, was
effectively infinite relative to the soldermask and insulator
distances. These values are likely to be encountered in a typical
IDS design. In a standard PCB structure, the dielectrics of the
soldermask and insulator are a resin based type coating and
FR4 laminate which is a standard material and thus widely
available. The permittivity of FR4 is known to be ≈ 4.5 in kHz
regime [33], [34].
No information regarding the permittivity of the used solder-
mask was available from the supplier (Pcbway.com, Shenzhen
JDB Technology Co., Ltd, Shenzhen, China), therefore the per-
mittivity value for the soldermask was approximated to 4.5 as
εr = 4 to 5 is a common value for epoxies in literature [35].
Using this value has shown to give accurate results with
experimental observations in preceding research [29]. While
the relative permittivity of the soldermask εr,SM and insulator
εr,ins was held constant at 4.5, the permittivity of the oil,
εr , was varied from 2 to 5, covering a wide range of oil
permittivities. Fig. 2 shows the results of our simulations
presented as contour plots.
The first set of simulations, Fig. 2a to Fig. 2c, show the
impact that changes in εr and dins/λ have on C, S and D.
The value of dSM /λ was held constant at 0.007 and the values
of C, S and D were calculated as shown in Fig. 2a. The
results show an increase in the capacitance C as both εr and
dins/λ increase. The sensitivity S is predominantly affected
by an increase in dins/λ, and is less sensitive to εr , as is seen
in Fig. 2b. The accuracy D, shown in Fig. 2c, is also primarily
affected by dins/λ. At low values of dins/λ, D is rising for
dins/λ ∼= 0.03 up to ≈ 15%, then follows a linear decrease
to −30 % at dins /λ ∼ = 0.2. This results in a zero cross-
ing at dins /λ ∼ = 0.075 resulting in high accuracy at this
ratio.
The second set of simulations, shown in Fig. 2d to Fig. 2f,
illustrates the change of C, S and D when εr and dSM /λ are
varied. Here the value of dins/λ was held constant at 0.03.
In Fig. 2d it can be seen that the capacitance C is pre-
dominantly affected by a change in εr and less by dSM /λ,
contrary to the previous observation varying dins/λ. The
sensitivity S in Fig. 2e varies with εr and dSM /λ. This occurs
in a way that low values for dSM /λ increase the sensitivity
with increasing εr , whereas this pattern is the opposite at
high values of dSM /λ. The sensitivity decreases as the para-
sitic capacitance through this soldermask gradually increases.
IEEE SENSORS JOURNAL, VOL. 17, NO. 19, OCTOBER 1, 2017
This does not appear to significantly affect the accuracy D as
shown in Fig. 2f; it exhibits a similar behaviour as observed
in Fig. 2c, exhibiting a steady decrease with increasing dSM /λ.
More specifically, D decreases throughout dSM /λ from approx-
imately +10% down to -30%. This in turn also exhibits a zero
crossing for the deviation at around dSM /λ = 0.025.
The third set of figures, Fig. 2g to Fig. 2i, shows the
impact of varying dSM /λ and dins /λ on C, S and D when
εr = 3.5 is held constant. Fig. 2g plots an increase in
capacitance C with increasing dins/λ while dSM /λ has a less
significant effect on C. However, Fig. 2h demonstrates an
opposite behaviour as can be seen via the contour lines. The
sensitivity S generally increases with dins/λ but at high dSM /λ
this is a weak effect. The accuracy D was affected by both,
dSM /λ and dins/λ, Fig. 2i. In general, high values of dSM /λ
and dins /λ produce a large D of more than −25% in the
negative space. However, small values of dSM /λ increase D
with increasing dins /λ similar to Fig. 2c. It is possible to find
a zero deviation for a certain combination of dSM and dins,
as well as whendSM and dins is equal to zero which would
represent an IDS having no substrate dielectrics – as this is
commonly assumed in approximations [32].
The simulations show that the spatial structure strongly
affects the figures of merit of an IDS. The soldermask
decreases the sensitivity throughout as it provides a direct
parasitic capacitance in between the sensed oil and electrodes.
Increasing the insulator distance in turn increases C and S, as
a larger separation of the electrodes to the ground plane results
in fewer field lines being pulled downwards to the ground. The
insulator thickness can control the accuracy while maintaining
a high sensitivity.
In order to facilitate a measurement of εr with high accuracy,
one aim of the work was to find the design for an IDS with
the best possible accuracy, ideally D = 0.
For an experimental verification of these simulations,
the practical design for a possible IDS was informed with
the aid of the data shown in Fig. 2i. It can be seen clearly
that a particular ratio of dins /λ is needed in order to achieve
a high accuracy. With aid of this knowledge, a suitable
IDS measuring accurately a specific value of εr in practice can
be created. A new insulating oil having εr,ref = 2.1065 was
used to simulate this specific case, Fig. 5. This Fig. 5 shows
a similar pattern as Fig. 2i. With aid of the FEM simulations,
it could be identified that the presence of a soldermask,
dSM , is detrimental to the sensitivity. Therefore, dSM was
removed in the manufactured IDS. The thickness dins is fixed
by the manufacturer at ∼ = 120 μm. From Fig. 5, a D = 0%
can be created for dins /λ ∼ = 0.096 resulting in an electrode
periodicity of λ = 1.247 mm.
This simulated D = 0% is achieved by a sensing mech-
anism which is dependent on the dimensions and dielectric
properties of the IDS and the εr of the oil under test. The
electric field distribution changes for different values of εr ,
as shown in Fig. 6. This affects the sensed capacitance and
thus εr in a convoluted manner. With reference to Eq. 1,
an accurate εr can be measured when no parasitic capacitance

is present or when a particular set of Cparasitic, Cparasitic ,
Coil and Cair compensates the remaining parasitic capacitance.
RISOS et al.: IDS: A DESIGN PRINCIPLE FOR ACCURATELY MEASURING THE PERMITTIVITY OF INDUSTRIAL OILS 6235

Fig. 2. FEM Simulations for the capacitance C, sensitivity S and accuracy D of εr,IDS . The first row (a, b, c) shows the figures of merit while varying εr
and dins of the IDS structure. The second row (d, e, f) shows the figures of merit when altering εr and dSM . The third row (g, h, i) shows the figures of merit
dependent on dSM and dins having εr,oil = 3.5.

The latter is case of the IDS; choosing λ = 1.247 for
εr = 2.1065 has such an effect on εr,IDS .
The physics in this case is predominantly based on two
effects, illustrated with Fig. 6. For the capacitance with εr = 1
(Fig. 6a), the field lines emanated from VAC are strongly
pulled downwards to the ground plane (GND) due to the
vicinity of the FR4 dielectric (εr,FR4 = 4.5 > εr,air ). This is
causing a lowered air capacitance, Cair , and a strong parasitic

capacitance, Cparasitic , through field lines connecting to the
underside of V0 (see circle in Fig. 6a). When measuring
the oil (εr > 1) the field distribution changes as illustrated
in Fig. 6b. This is yielding an increased E-field density through
the oil than air. This produces an increased capacitance, Coil ,
through εr and weak parasitic capacitance, Cparasitic. This
parasitic capacitance vanishes for εr = εr,FR4 and becomes
negative for εr > εr,FR4 as more field lines progressively
connect to V0 than to GND, Fig. 6c, causing a capacitance
over-proportional to εr . This special interaction is dependent
on the insulator thickness dins allowing an optimal value of λ
in the sensor design.
III. D EVELOPING AND TESTING ACCURATE IDS
A PCB-IDS based on λ = 1.247 for a target deviation
D = 0% was designed and fabricated without soldermask
as the soldermask is detrimental to the sensing proper-
ties. A gold finish onto the electrodes provided sufficient
electrochemical inertness in the oil environment. Measure-
ments on this IDS were done at 50 Hz using a standard
impedance analyser (Hioki IM3590, Hioki, Nagano, Japan)
with guarded four wire BNC connectors. Fig. 3 shows the
6236
Fig. 3. Designed IDS in standard PCB technology with four point measure-
ment principle through shielded BNC connectors. Sensor surface is Ni as a
diffusion barrier for copper and subsequently Au plated as corrosion reduction.
Active IDS sensing area highlighted and electrically grounded areas are
marked.
Fig. 4. A SEM cross sectional image of the manufactured IDS as a cross
sectional view showing the electrode width, thickness and separation from the
common ground layer as dins .
Fig. 5. Simulation results for the deviation D dependent on dSM and dIns
as a specific case for the MUT having εr = 2.1065.
manufactured PCB IDS. The four BNC connectors provided
shielded measurement leads. These four leads merge onto the
PCB board surface in a shielded manner at the electrodes.
The SEM image in Fig. 4 shows a cross sectional view of the
electrodes and indicates the actual dimensions. The ground
layer is indicated with “GND”. This confirms the designed
parameter of width = gap = 312μm for dins ∼ = 120 μm.
The SEM image is accurate within ±1 μm. These dimensions
were consistent within ±1 μm for three examined IDS cross
sections.
In order to verify the functionality of the IDS, showing the
capability to measure εr of the new and degraded oil accu-
rately, six litres of a new insulating transformer oil (Savita Oil
Technologies Ltd. Grade BS 148 Class-I) were degraded
in an accelerated manner. This oil is a purified paraffinic
oil without additives which has properties equivalent to a
IEEE SENSORS JOURNAL, VOL. 17, NO. 19, OCTOBER 1, 2017
Fig. 6. Electric field lines (1 to 8) depicted for one pair of the simulated
electrodes in Log(E/E0 ) where E0 is 1 V/m. The field lines curl above
the electrodes and reconnect to the opposite electrode dependent on the
permittivity under test (red circle). A grey arrow supports the visual field
line tracking representatively between 1 and 5. (a) Two field lines connected
to the upper side of V0 while one reconnects to the underside. (b) Three field
lies connect to the upper side when measuring oil. (c) For higher εr values
of oil, the increasing relative field intensity above the sense electrodes is
responsible for a non-constant parasitic capacitance.
standard industrial oil base. This oil was heated to 140 °C
with 500 g of transformer insulating paper ‘Diamond-Paper’
(Weidmann Diamond Dotted 380 μm). The high temperature
treatment and presence of a thin copper sheet with a sur-
face area of 40 cm2 accelerated the aging. The aging time
was 600 h, which results in aged oil equivalent to its end of
life condition [36]. The measured moisture value of the opaque
brown coloured oil was 103 ±5 ppm, measured via Karl Fisher
titration by eNZoil Ltd (Lower Hutt, New Zealand).
Two vials, one containing the new oil and the other the
degraded oil, were prepared. In Fig. 7, we measured air, new
oil and degraded oil manually with our optimized IDS. The
procedure was as follows: after measuring the air capacitance
with the IDS, the IDS was placed in a vial containing new oil
for about 20 seconds. The IDS was removed and exposed to
air for 10 seconds. Subsequently, the IDS was inserted into
RISOS et al.: IDS: A DESIGN PRINCIPLE FOR ACCURATELY MEASURING THE PERMITTIVITY OF INDUSTRIAL OILS
Fig. 7. Repeatability and response of the sensor showing air and oil
capacitance without hysteresis effect. The IDS was exposed to air and
manually inserted into vials containing new and old oil (each two times).
Fig. 8. (a) Capacitance C measured vs. simulated. (b) Sensitivity S measured
vs. simulated. FEM simulation showing similar values to the experimental
data.
the vial containing degraded oil for about 20 seconds. This
process was repeated again and no cleaning of the IDS was
done. This sensor exhibited an instantaneous response without
lag or hysteresis effect to the air and different oils. This
shows that this IDS is capable of reflecting changes in the
oil instantaneously. This is possible when the capacitance is
directly and linearly proportional to εr . In this experiment,
the capacitance of the IDS followed εr accurately within the
measurement uncertainty of 0.5%.
For a comparison of the measured vs. simulated figures of
merit C, S and D, the capacitance was measured over 51 data
points at 50 Hz in accordance with the standard [5]. Both
results of C and S are shown in Fig. 8. The capacitance C
(Fig. 8a) and the sensitivity S (Fig. 8b) are slightly higher
than simulated. In Fig 9, the measured permittivity of oil
via the manufactured IDS (Fig. 3) is compared to reference
values (εr,IDS vs. εr,ref ) for the new (εr,ref = 2.1065) and
aged oil (εr,ref = 2.2453). The mean value yields a slight
consistent positive deviation D of about +0.45% due to the
manufacturing tolerances. We have conducted the experiment
6237
Fig. 9. Permittivity via IDS measured εr,IDS compared with reference
values εr,ref showing a good agreement. The exact value for the new oil was
εr,ref = 2.1065 and for the aged oil εr,ref = 2.2453 from which the
εr,IDS differs of about 0.5 %.
using three sensors from which we observed a consistent
deviation of 0.4 - 0.5% emphasising a manufacturing issue.
Examining Fig. 5, a lower value of dins/λ causes a trend
towards a positive deviation. The reduced thickness of the
insulator dins = 119 μm vs. designed 120 μm, explains this
shift. Still, this result is within the measurement error of the
experiment which was ± 0.5%. We did not observe any capac-
itance fluctuations dependent on the environment or sensor’s
temperature. An explanation for this lays in the thermal expan-
sion of dins. The thermal expansion of the FR4 insulator layer
may result in a dimensional change of not more than 100 nm
per 10 °C which is well below the 1 μm manufacturing
tolerance causing the 0.5% deviation.
This is a valuable outcome and verifies that standard
PCB production with FR4 laminate is capable of manufac-
turing the proposed design.
IV. SUMMARY AND C ONCLUSION
A procedure has been demonstrated for the development of
an accurate interdigitated sensor for measuring the degradation
of industrial oils. This design procedure simulated figures of
merit of the sensor for particular combinations of dimensions
and sensor structure. The examined figures of merit (capac-
itance, sensitivity and accuracy) are strongly affected by the
spatial structure of the IDS making it important to know the
dimensions and permittivities of the physical layout. It was
found that the soldermask throughout has a negative impact
on the sensitivity but if this layer thickness is well con-
trolled, it can improve accuracy. A similar result regarding
accuracy was obtained from a variation of the insulator dis-
tance dins between the IDS electrodes and a common ground
plane (GND) without a soldermask. This also exhibited a
positive impact on the sensitivity.
The high accuracy emerges from parasitic capacitances
which are auto-compensated choosing a right combination of
the IDS insulator dins and wavelength λ for a given εr of the
oil. The accuracy is achieved at the ratio of dins /λ ∼
= 0.95 with-
out the soldermask. Standard PCB technology was then used
to manufacture an IDS with this given ratio.
The efficacy of this design in discriminating changes in
the properties of industrial oil during degradation was shown
by a measurement of the permittivity on unused and aged
insulating oil and by comparing measured values to the values
6238
from a calibrated instrument. Subsequent measurements were
consistent with the FEM predictions in this research, verifying
the possibility to manufacture an IDS with PCB technology
measuring εr accurately using dins/λ ∼ = 0.096. This particular
ratio will change when a different material is used for the
insulator (such as Rogers™or alumina). The changed insulator
permittivity yields a new dins/λ for the D = 0 condition.
ACKNOWLEDGMENT
The authors thank Joseph Bailey, Mike Davis and
James Storey for helpful discussions. Also, thank Bhaba Das
(ETEL, Auckland, New Zealand) for valuable discussions.
R EFERENCES
[1] T. O. Rouse, “Mineral insulating oil in transformers,” IEEE Elect. Insul.
Mag., vol. 14, no. 3, pp. 6–16, May 1998.
[2] I. Fofana, “50 years in the development of insulating liquids,” IEEE
Elect. Insul. Mag., vol. 29, no. 5, pp. 13–25, Sep./Oct. 2013.
[3] J. W. Jung, Y. B. Lim, J. S. Jung, G. H. Yi, and H. Y. Park, “The
analysis of dielectric characteristics of transformer oil suffering ageing
treatment,” presented at the IEEE 8th Int. Conf. Properties Appl.
Dielectr. Mater., Bali, Indonesia, 2006.
[4] A. Risos, N. Long, A. Hunze, and G. Gouws, “Mid-IR absorbance
and its relation to static permittivity as a robust in-field tool tracking
oil deterioration,” Proc. SPIE, vol. 10215, pp. 102150O-1–102150O-10,
May 2017.
[5] Insulating Liquids. Measurement of Relative Permittivity, Dielectric Dis-
sipation Factor (Tan δ) and D.C. Resistivity, Standard IEC 60247:2004,
British Standards Institution, United Kingdom, 2004, p. 27.
[6] C. F. Ten, M. A. R. M. Fernando, and Z. D. Wang, “Dielectric properties
measurements of transformer oil, paper and pressboard with the effect of
moisture and ageing,” in Proc. Annu. Rep.-Conf. Elect. Insul. Dielectr.
Phenomena, Oct. 2007, pp. 727–730.
[7] Y. Du, T. Wu, and J. Cheng, “Age detection of lubricating oil with
on-line sensors,” presented at the IEEE Sensors, Busan, South Korea,
2015.
[8] Suwarno and J. Marbun, “Effect of thermal aging on the dielectric
properties of mixture between mineral oil and natural ester,” presented
at the IEEE Region 10 Conf. TENCON, Macao, China, Nov. 2015.
[9] D. Yang, X. Zhang, Z. Hu, and Y. Yang, “Oil contamination monitoring
based on dielectric constant measurement,” in Proc. Int. Conf. Measuring
Technol. Mechatronics Autom., Apr. 2009, pp. 249–252.
[10] A. Agoston, N. Dorr, and B. Jakoby, “Online application of sensors mon-
itoring lubricating oil properties in biogas engines,” in Proc. 5th IEEE
Conf. Sensors, Oct. 2006, pp. 1099–1102.
[11] M. Rafiq et al., “Use of vegetable oils as transformer oils—A review,”
Renew. Sustain. Energy Rev., vol. 52, pp. 308–324, Dec. 2015.
[12] C. T. Dervos, C. D. Paraskevas, P. Skafidas, and P. Vassiliou, “Dielectric
characterization of power transformer oils as a diagnostic life pre-
diction method,” IEEE Elect. Insul. Mag., vol. 21, no. 1, pp. 11–19,
Jan. 2005.
[13] A. E. Seaver, A. V. Mamishev, Y. Du, B. C. Lesieutre, and M. Zahn,
“Development and applications of fringing electric field dielectrometry
sensors and parameter estimation algorithms,” J. Electrostatics, vol. 46,
nos. 2–3, pp. 109–123, Apr. 1999.
[14] J. Ulrych and V. Mentlík, “Dielectric properties of sunflower, rapeseed
and commonly used mineral oil,” in Proc. 17th Int. Sci. Conf. Electr.
Power Eng. (EPE), May 2016, pp. 1–4.
[15] T. A. Prevost, “Dielectric properties of natural esters and their influence
on transformer insulation system design and performance,” in Proc.
IEEE Power Energy Soc. Gen. Meet., May 2009, pp. 1–7.
[16] D. Berube, F. M. Ghannouchi, and P. Savard, “A comparative study of
four open-ended coaxial probe models for permittivity measurements
of lossy dielectric/biological materials at microwave frequencies,” IEEE
Trans. Microw. Theory Techn., vol. 44, no. 10, pp. 1928–1934, Oct. 1996.
[17] S. Seewattanapon and P. Akkaraekthalin, “A broadband complex per-
mittivity probe using stepped coaxial line,” J. Electromagn. Anal. Appl.,
vol. 3, no. 8, pp. 312–318, 2011.
[18] A. V. Mamishev, K. Sundara-Rajan, F. Yang, Y. Du, and M. Zahn,
“Interdigital sensors and transducers,” Proc. IEEE, vol. 92, no. 5,
pp. 808–845, May 2004.
IEEE SENSORS JOURNAL, VOL. 17, NO. 19, OCTOBER 1, 2017
[19] A. Mamishev, Y. Du, and M. Zahn, “Measurement of dielectric property
distributions using interdigital dielectrometry sensors,” in Proc. Conf.
Elect. Insul. Dielectr. Phenom., Virginia Beach, VA, USA, Oct. 1995,
pp. 309–312.
[20] A. V. Mamishev, B. C. Lesieutre, and M. Zahn, “Optimization of multi-
wavelength interdigital dielectrometry instrumentation and algorithms,”
IEEE Trans. Dielectr. Electr. Insul., vol. 5, no. 3, pp. 408–420, Jun. 1998.
[21] A. V. Mamishev, S. R. Cantrell, Y. Du, B. C. Lesieutre, and
M. Zahn, “Uncertainty in multiple penetration depth fringing electric
field sensor measurements,” IEEE Trans. Instrum. Meas., vol. 51, no. 6,
pp. 1192–1199, Dec. 2002.
[22] Y. Sheiretov and M. Zahn, “Dielectrometry measurements of moisture
dynamics in oil-impregnated pressboard,” in Proc. Conf. Rec. IEEE Int.
Symp. Elect. Insul., Jun. 1994, pp. 33–36.
[23] L. Yang, Y. Li, C. L. Griffis, and M. G. Johnson, “Interdigitated
microelectrode (IME) impedance sensor for the detection of viable
Salmonella typhimurium,” Biosensors Bioelectron., vol. 19, no. 10,
pp. 1139–1147, May 2004.
[24] M. Varshney and Y. Li, “Interdigitated array microelectrodes based
impedance biosensors for detection of bacterial cells,” Biosensors
Bioelectron., vol. 24, no. 10, pp. 2951–2960, 2009.
[25] T. Yang, Y. Z. Yu, L. S. Zhu, X. Wu, X. H. Wang, and J. Zhang,
“Fabrication of silver interdigitated electrodes on polyimide films via
surface modification and ion-exchange technique and its flexible humid-
ity sensor application,” Sens. Actuators B, Chem., vol. 208, pp. 327–333,
Mar. 2015.
[26] K. Kim, X. Liu, Y. Zhang, and Y. Sun, “Nanonewton force-controlled
manipulation of biological cells using a monolithic MEMS microgripper
with two-axis force feedback,” J. Micromech. Microeng., vol. 18, no. 5,
p. 055013, Apr. 2008.
[27] Y. Zhang, X. Liu, C. Ru, Y. L. Zhang, L. Dong, and Y. Sun,
“Piezoresistivity characterization of synthetic silicon nanowires using a
MEMS device,” J. Microelectromech. Syst., vol. 20, no. 4, pp. 959–967,
Aug. 2011.
[28] M. W. den Otter, “Approximate expressions for the capacitance and
electrostatic potential of interdigitated electrodes,” Sens. Actuators A,
Phys., vol. 96, nos. 2–3, pp. 140–144, Feb. 2002.
[29] A. Risos, N. Long, A. Hunze, and G. Gouws, “A 3D Faraday shield
for interdigitated dielectrometry sensors and its effect on capacitance,”
Sensors, vol. 17, no. 1, p. 77, 2017.
[30] C. T. Dervos, C. D. Paraskevas, P. D. Skafidas, and P. Vassiliou,
“A complex permittivity based sensor for the electrical characterization
of high-voltage transformer oils,” Sensors, vol. 5, no. 4, pp. 302–316,
2005.
[31] X. B. Li, S. D. Larson, A. S. Zyuzin, and A. V. Mamishev, “Design prin-
ciples for multichannel fringing electric field sensors,” IEEE Sensors J.,
vol. 6, no. 2, pp. 434–440, Apr. 2006.
[32] Y. C. Lim and R. A. Moore, “Properties of alternately charged coplanar
parallel strips by conformal mappings,” IEEE Trans. Electron Devices,
vol. 15, no. 3, pp. 173–180, Mar. 1968.
[33] E. L. Holzman, “Wideband measurement of the dielectric constant
of an FR4 substrate using a parallel-coupled microstrip resonator,”
IEEE Trans. Microw. Theory Techn., vol. 54, no. 7, pp. 3127–3130,
Jul. 2006.
[34] A. Chirap and V. Popa, “Insertion loss measurement of a lowpass
microwave filter manufactured on FR4 laminate,” in Proc. Int. Conf.
Develop. Appl. Syst. (DAS), May 2016, pp. 231–234.
[35] M. S. Naidu, High Voltage Engineering. New York, NY, USA:
McGraw-Hill, 1995.
[36] Power Transformers—Part 7: Loading Guide for Mineral Oil-Immersed
Power Transformers, Standard IEC 60076-7, International Electrotech-
nical Commission, United Kingdom, 2015, p. 27.
Alex Risos received the B.Sc. degree in physical
technologies in 2011, and the M.Sc. degree in optical
engineering and photonics in Gottingen, Germany,
in 2013. He is currently pursuing the Ph.D. degree
in physics with the Victoria University of Welling-
ton, New Zealand. He is experience includes AMO
physics, finite-element method, and analytical analy-
sis and electronics.
RISOS et al.: IDS: A DESIGN PRINCIPLE FOR ACCURATELY MEASURING THE PERMITTIVITY OF INDUSTRIAL OILS 6239

Nick Long received the master’s degree in physics
from Victoria University and the Ph.D. degree in
physics from the University of Southern California.
He is currently Director of the Robinson Research
Institute, Victoria University of Wellington, New
Zealand. From 1994 to 2013, he was with Industrial
Research Ltd. His primary research has been in high
temperature superconductivity. He was a Founder
of the program with IRL to develop HTS Roebel
cables including development of manufacturing and
characterisation processes. His other HTS work has
focused on enhancing wire performance and understanding the phenom-
enology of critical currents. He is also involved in many projects with
NZ companies in the application of electromagnetic sensing. His previous
research experience includes quantum transport phenomenon in low temper-
ature physics and materials research in oxygen ion conductors.
Gideon Gouws received the Ph.D. degree in physics
from the University of Port Elizabeth, South Africa.
He then joined the CSIR and was involved in the
development of thin film II–VI semiconductors by
organometallic vapor phase epitaxy. He also spent
several years with Rakon Ltd, New Zealand, where
he developed commercial quartz crystal resonators.
He is currently on the academic staff in engineering
and computer science with the Victoria University
of Wellington, New Zealand. His research interest is
in materials and sensor instrumentation, with focus
on impedance-based sensing for environmental and industrial applications and
the fabrication of microstructured materials for sensing and energy transfer
applications.

Arvid Hunze received the M.Sc. degree in physics

from the University of Saarbruecken in 1999 and
the Ph.D. degree in materials science from the Uni-
versity Erlangen-Nuernberg in 2003. From 2003 to
2011, he was a Senior Scientist and Scientific Con-
sultant in the area of organic LEDs and photovoltaic
in the research and renewable energy business,
Department of Siemens, Germany. From 2011, he
was a Senior Scientist with IRL, Callaghan Inno-
vation, and Victoria University of Wellington, New
Zealand, developing optical and magnetic sensors
and systems for non-destructive testing and monitoring assets for energy
transmission.

View publication stats