This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

A Nonintrusive Magnetically Coupled Sensor for

Measuring Liquid Level
Sandra K. R., Student Member, IEEE, Boby George, Senior Member, IEEE, and Jagadeesh Kumar V., Senior
Member, IEEE

Abstract— A new, linear, nonintrusive, and cost-effective type but need installations at the lid of the liquid container.
inductive liquid level sensor is designed and developed. The sensor Moreover, these types of sensors are expensive and need
consists of a simple coil that floats on the liquid surface. A couple constant maintenance. The accuracy of this method is limited
of printed planar coils partially surround the liquid column and due to the parasitic reflections caused by barriers on the liquid
sense the position of the floating coil and hence the liquid level. The surface, including foams, residues, and surface ripples, and the
planar coils are arranged with overlap such that the floating coil is readings may be affected by temperature, pressure, and dust [3],
always in proximity to both the coils. The coils are excited, and [4]. Dust deposited over a period can even shut down the sensor.
their response processed, such that the final output is proportional Optical systems are lightweight, immune from electromagnetic
to the liquid level. The input-output characteristics of the proposed interference, but in general, are expensive and also need
liquid level measurement are linear, with a worst-case percentage
constant maintenance [8], [9].
nonlinearity < 1%. A prototype sensor built for a range of 180 mm,
had a resolution of 0.73 mm. The output of the prototype has very Inductive sensors are widely used in the industry. They
low sensitivity to the position of the floating coil in the x-y plane, provide reliable output even when exposed to a harsh
indicating that the movement of the floating coil on the surface of environment [10]. They use the principle of eddy currents or
the liquid column does not affect the liquid level measurement. change in the magnetic field for level measurement [11]-[15].
Since the sensing part is just made of passive elements, the lifetime Water level measurement from the phase values of inductive
of the sensor can be made very high. measurements has been proposed [11]. A fixed planar coil
Index Terms— Liquid level sensor, inductive sensor, planar coil inductive sensor is placed near the water surface and excited. In
sensor, nonintrusive sensor, ratio-metric output.
this method, the change in impedance of the coil due to the
induction of eddy currents in water is measured, and the
I. INTRODUCTION correlation between its phase and the water level is presented.
This method requires free space above the water level. Another
L iquid level sensing is an essential requirement in the
process, chemical, petroleum, pharmaceutical and life
science, food, and energy industries [1], [2]. Depending on
eddy current based technique with two coils was proposed for
liquid level measurement [12]. In this method, the liquid level is
the operating environment and the type of the liquid whose level measured by calculating the phase angle of the mutual
is to be measured, different methods are being used. The inductance between the excitation and induction coils. Liquid
commonly used sensors in industry include potentiometric, level measurement using an electromagnetic resonance circuit,
capacitive, hydrostatic, radar, ultrasonic, laser, photoelectric, presented earlier, consists of a coil that surrounds the liquid
vessel-weighing type load cells and float based sensors [3]. The column. A capacitor is connected in parallel with this coil to
method is chosen such that it should perform reliably well in form a tank circuit. It is shown that the resonance frequency is a
those application conditions and suit its economic structure, at function of the liquid level [13]. An inductive type liquid level
the same time. sensor proposed earlier uses two fixed coils of different sizes to
infer conductivity and the liquid level simultaneously [14]. A
In general, liquid level sensors can be classified as contact technique has been proposed in [15] to measure the liquid level
type and non-contact type. In the contact type, the liquid is in of a float type sensor, using a magnetic rod attached to the float.
electrical contact with the measurement probe, and hence its The position of the float is detected using a differential
applications are limited [3], [4]. Of the non-contact type level inductance transducer. All the sensors presented above require
sensors, the hydrostatic method is dependent on the density of a wired connection to the measurement unit [11]-[15]. Hence,
the liquid whose level is to be measured [3]. Capacitive type they are not suitable for level measurement in closed airtight
liquid level measurement systems are good as they have: no containers.
moving parts, low power consumption, and linearity. However,
they are best suited for low conductivity liquids possessing a We now present the design and development of a simple,
constant dielectric property. The accuracy of capacitance type cost-effective, linear, inductive liquid level sensor suited for
liquid level measurement is adversely affected by parasitic sealed containers. The sensor uses a wire-wound coil made to
capacitances and conductivity variations, which is inevitable in float on the surface of the liquid column. A set of planar coils
field operations [5] - [7]. surround, but not encapsulate the liquid column. The
arrangement utilizes the mutual coupling between the planar
Non-contact sensors have longer life period; they are more coils and the floating coil. The proposed sensor has the
reliable, stable, and robust and can be used for liquids in airtight following advantages: (a) It is nonintrusive; no electrical
containers, or for detecting toxic substances. Radar and connection is required to the floating part. (b) The installation
ultrasonic type measurement methods are of the non-contact

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

the top of the coil that has uniformly increasing length as a

function of h, then the mutual inductance as given in (1) varies
d as shown in Fig. 2(b). For plotting the characteristics in Fig.
2(b), W was set as 114 mm, and dw was set as 2 mm. In
obtaining the plot in Fig. 2(b), few valid assumptions were
l made; namely, the effective length l at any height h will be that
of the triangular-shaped coil of Fig. 2(a) (end effects are
neglected) and only the wire in the triangular-shaped coil that is
Fig. 1. Two straight parallel equal wires. in close proximity alone contributes to M with the wire W
(mutual inductance is calculated considering one wire from each
of the sensor is very easy, and it requires very little maintenance. coil at a time, neglecting the contributions from the adjacent
(c) As the moving part does not require electrical contacts, wires of the triangular-shaped coil). The structure proposed in
higher reliability compared to the contact type sensors is Fig. 2 was only for illustration, as in practice when we wind a
ensured. (d) The sensing part in the float is made of passive coil, the ends of the adjacent wires need to be connected to make
elements that guarantee a long operational life. (e) The planar a complete coil and that results in the magnetic field (and
coils are relatively easy to manufacture. (f) The sensor range can coupling) generated by one set of the wires to be suppressed by
be easily increased by increasing the number of planar coils, the field generated by the set of ‘return’ wires (since they carry
outside the container. The method works well to sense the level the current in opposite directions) that connect the ends of the
of liquids with low conductivity and suited for nonconducting wires shown in Fig. 2(a). Thus, the triangular coil design is
liquids stored in nonconducting containers. impractical.

II. THE PROPOSED NONINTRUSIVE LIQUID LEVEL SENSOR A solution to this problem is to use a simple, square-shaped
planar coil shown in Fig. 3(a). It is easily seen that four such
If two coils or even two wires share a common flux, then a triangular coil patterns marked 1, 2, 3, and 4 in Fig. 3(a) are
mutual magnetic coupling exists between them. The mutual obtained from a single square-shaped planar coil. If we now
inductance M depends on many parameters and it is possible to keep the floating coil, as shown in Fig. 3(b), then only portions
design such that M becomes independent of all parameters, 1 and 3 will contribute to the mutual inductance as the currents
except one chosen parameter. This principle is exploited here to in the floating coil are parallel to each other. The currents in
design a liquid lever sensor consisting of two fixed planar coils portions 2 and 4 will not contribute to the mutual inductance as
and a floating coil. the current vectors are displaced by 90. Fig. 3(c) shows the plot
A. Principle of design of the proposed technique. of mutual inductance M as the position of the floating coil is
varied from 0 to h in the vertical axis. In order to maintain
The mutual inductance, M, between two parallel equal
compatibility, here too Wc was chosen as 114 mm, and d is set
straight wires of length, l and distance of separation, d, as shown
as 2 mm. When the sensor coil was moved from 0 to Wc along
in Fig. 1 is given by [16]:
h, the mutual inductance decreased in the region from 0 to Wc 2
    and then increased in the region from Wc 2 to Wc . The
l l2  d2 d
M = 0.002 l loge  + 1 + 2  − 1 + 2 +  . (1)
 d d  l l assumptions made in plotting Fig. 2(b) are applicable in
    obtaining the plot in Fig. 3(c) as well.
Equation (1) provides an insight that if we make the length of However, a value of M (on the y-axis in Fig. 3(c) corresponds
interaction to linearly vary with level h, then we get the mutual to two different positions h of the floating coil. One method to
inductance, M to vary almost linearly with h. Such a solution is resolve this issue is to use only one half of the planar coil and
shown in Fig. 2(a), where the coil length keeps increasing as we fold the other half. With this solution, the nonlinearity will
increase the height h. If the second wire is of length W (only one persist. Instead, if we place two planar coils A and B such that
wire) and we move this wire, in the h axis, from the bottom to 50% of each coil’s area overlaps with the other’s as illustrated
in Fig. 3(d), the mutual inductances MA and MB of coils A and B
i with the floating coil vary in a push-pull manner in the region
between Wc 2 to Wc as seen in Fig. 3(e). If we adopt a
dw ratiometric operation, we will obtain an output that is linear to
the position h of the floating coil. If required, we can extend this
concept, as shown in Fig. 4, to enhance the range of sensing the
W height beyond Wc 2 .
h
It should be noted here that the floating coil must be made
lightweight and added with additional buoyancy so that it floats
on the surface of the liquid. The well-known Archimedes’
W principle is utilized to design the float. According to this, the
(a) (b) material of the float, its size, and shape are selected such that the
up-thrust force of the fluid on the float is much greater than the
Fig. 2. (a) Schematic of the triangular coil (front view) (b) Plot of h vs. M. total weight W of the object. The up-thrust force or buoyant

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

i
0
Floating coil
1

h h
4 2

0
Readout coil

(a) (b) (c)

0
coil B

Floating
coil
coil A
0
(d) (e)

Fig. 3. (a) Square shaped planar spiral coil (b) Spiral coil bent around the curvature of the floating coil (c) Plot of mutual inductance vs. height for a single
readout coil (d) Two readout coils aligned one on top of the other with 50% overlap and bent around the curvature of the floating coil (e) Plot of mutual
inductance vs. height for a pair of readout coils.

force FB is given by FB = Vg , where ρ is the density of the liquid with different density will differ depending on the
liquid, V is the volume of the object, and g is the acceleration difference in the densities of the liquids. This will introduce an
due to gravity. Once a float is designed such that its weight, offset in the output that can be corrected. It is also essential that
W << Vg (over-designed), the change in the vertical position the coil is as close to the fixed planar coils on the container. With
of the float due to variation in the density of the liquid will be these conditions, we may have to have the diameter of the
negligible. Hollow floats filled with air is one of the methods to floating coil to be almost equal to the diameter of the container
realize the above condition. If we use a float designed for a and at the same time ensure that the floating coil is able to move
liquid of a certain density, the amount that it will sink for another up and down inside the container with the liquid level, without
any friction and lateral movements. Now, as the liquid level
changes, the mutual inductance, between the coils vary, only as
a function of h.
Tank Coil D
Floating coil III. FINITE ELEMENT ANALYSIS
Coil C
Coil B To check the validity of the proposed coil sensor system, the
.. .. Readout coil Coil A Finite Element Analysis (FEA) of the proposed liquid level
sensor was carried out using the COMSOL multiphysics FEA
solver. For the simulation, the floating coil, along with one
Measurement planar coil, was considered. A square planar spiral coil of side
Liquid Output
Unit 100 mm with 12 turns was modeled in COMSOL. The curvature
of the planar coil was not introduced in the model in order to
Fig. 4. Cross sectional view of the measurement set up with floating coil reduce the complexity of the model, without affecting the
and readout coils in place. accuracy of the output. Therefore, the floating coil also was

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

Floating coil S2 RA
×10-5 S1 RF
0
8 E 0 1 MA
O' O E'
6 1
4
2 LA
0
VR CE LF CF
h iA
-2
-4
-6
-8

Readout coil Fig. 7. Equivalent circuit of the proposed readout scheme when
(a) (b) coil A alone is introduced.
Fig. 5. (a) FEA slice plot of magnetic flux density of readout coil. The
magnetic flux that links with the floating coil alone is shown. (b) Plot of are explained here. The explanations and derivations can be
height vs. mutual inductance for a single readout coil, obtained from FEA . extended to any number of stacked planar coils. In order to
measure the mutual inductance (MA or MB) between the floating
coil and the planar coil (A or B), a measurement scheme, as
modeled as a simple square coil of side 120 mm and 8 turns,
shown in Fig. 6 is adopted. As can be seen from the figure, a
instead of a circular coil. This was intentionally done to have an
tank circuit is formed using the floating coil by connecting a
overhang of 10 mm on each side. A finite element study has
capacitor CF across the terminals of it. One terminal each (A1
been conducted to obtain mutual inductance versus the vertical
and B1) of the planar coils A and B are tied together and the
position of the floating coil. In this study, an alternating current
current in the node converted to a voltage using an i-to-v
of magnitude 1 A and frequency 100 kHz was injected into the
converter, as shown in Fig. 6. The other terminals, namely A2
planar coil and the open-circuit voltage induced in the floating
and B2, are tied to the terminals of a single pole double throw
coil for different positions (vertical) of the floating coil in the
(SPDT) switch S2. The pole of S2 is, in turn, connected to one of
range 0 mm to 100 mm in steps of 2 mm was recorded. From
the output terminals of a second SPDT, S1. While the other
the recorded values, the flux-linkage and the mutual inductance
terminal of S1 is connected to the reference voltage VR, its pole
are computed and shown in Fig. 5.
is connected to one terminal of a capacitor CE, whose other
It is seen from Fig. 5 that in the first half section of the coil, terminal is grounded. Voltage vv across the capacitor CE and the
the output voltage increases initially and then decreases as the voltage vi proportional to the current i are acquired and
length of coil filaments decreases. At both the ends of the coil processed to obtain the height h of the liquid.
(origin O and end E), the impact from the neighboring
Initially switches S1 and S2 are kept in position 0 till steady-
conducting filaments is less due to which there is a change in
state occurs, and CE is charged to VR. Then switch S1 alone is
the trend of the plot (in the ranges OO' and EE'). As obvious, the
changed to position 1. Capacitor CE is now connected to coil A
induced voltage in the second half of the coil is the mirror image
and discharges. The current and voltage during discharge are
of the first half as the coil is symmetric about its center. From
acquired and stored until the charge in CE becomes zero. Next,
these results, it can be easily deduced that the region between O
switch S1 is set to position 0, and S2 set to position 1. After
and E is optimal for setting up the range of the height of the
steady-state (CE is charged to VR), switch S1 alone is changed to
liquid column to be measured.
position 1. Now coil B is connected across CE and CE
discharges. Once again, the current and voltage during discharge
IV. MEASUREMENT SCHEME are acquired and stored until the charge in CE becomes zero.
Though any number of coils can be stacked to obtain liquid From the stored data, the values of MA or MB are computed as
level measurement as indicated in Fig. 4, for ease of explained next.
understanding, only stacking of two coils, namely coil A and B
A. Computation of MA and MB.
S2 When CE is charged to VR and then made to discharge through
S1 0
DEMUX

0 1 coil A, the equivalent circuit will be as indicated in Fig. 7. Here

1 LA and RA are the self-inductance and winding resistance of
VR B2 A2 M
CE vS1 vS2 coil A. Similarly, the self-inductance and winding resistance of
Coil B

the floating coil are represented by LF and RF, respectively. The

Coil A

CF
mutual inductance, MA, between the floating and readout coil A
B1 LF
A1 is given by M A = K L A L F , where K is the coupling
vv Ri
Control i coefficient. To analyse the circuit, we need to transform the
Data impedance Zr(s) of the floating coil to the coil A side as [17]:
Algorithm _
&
Sealed Container − M A2CF s 3
Z r (s ) =
Computation vi +
(2)
LF CF s 2 + RF CF s + 1

Since the impedance is made of inductances and capacitances,

Fig. 6. Block diagram of the proposed readout scheme.
exponentially decreasing oscillatory current will flow. These

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

oscillations will correspond to the natural frequencies of the Virtual instrument

system, and the system here has two series resonance
frequencies and one parallel resonance frequency [18]. The Container
parallel resonance frequency, fF of the floating coil is:
1 1 Readout coils
fF = → ωF = . (3)
2π LF C F LF C F Data
Acquisition
Writing Kirchhoff’s Voltage Law in the s-domain, we get: System
Measurement

I A (s ) VR
circuitry

− + LAsI A (s ) + RAI A (s ) + Zr (s )I A (s ) = 0 (4)

CE s s
Floating coil
Rearranging (4), we obtain the impedance Z(s) of the circuit as:

Z (s ) =
(L C
A Es
2
)
+ RACE s + 1 − M A2CE CF s 4 (a)

(
CE LF CF s 2 + RF CF s + 1 ) (5)
Capacitor Floating coil
The imaginary part of Z(s) in the frequency domain can be
obtained as:

ω 5 M A 2 C F 2 RF
Im(Z ( jω) ) = R Aω +
(1 − L C ω ) + (ωR C
(6)
F F
2 2
F F )2
Substituting ωF as in (3) in (6), we get:
130 mm
 3M 2
Im(Z ( jωF )) = RAF + F A (7) mmmm
160 mm
RF
(b) (c)
It should be noted here that if the floating coil is absent, then
Im(Z(jωF)) will be just RAωF. Thus, the measurement of ωF
Fig. 8 (a). Measurement setup (b) photograph of the floating part (c) the 3D
without the floating coil will give us RAωF, and from (7), we can model of the floating coil in a casing to minimize friction between the
deduce MA. Similarly, from the second set of charge and floating part and the container wall.
discharge through, coil B, we can obtain:

F 3M B 2 glass beaker used has a capacity of 5 liters, with an inner

Im(Z ( jωF ) ) = RBF + , (8) diameter of 162 mm and thickness, 2.5 mm. A set of three
RF
identical planar square coils fabricated on a thin flexible printed
and hence MB. In order to get the imaginary part of the circuit board (PCB) of thickness 0.2 mm, side of 114 mm, with
impedance, a multiplier type phase sensitive detector (PSD) 55 turns, were used as the readout coils. The copper tracks on
logic is used to extract the in-phase and quadrature components, the PCB coil possessed a thickness of 35 µm and a width of 0.4
at the resonance frequency ωF, of the voltage output vi, first mm, with a separation of 0.6 mm between the conductors. Three
when coil A is connected, and then coil B is connected. To such PCB coils were stacked one with an overlap of about 60 %
facilitate this, a reference sine wave at ωF was generated and between two adjacent coils. This will ensure that measurement
used. Thus, the values of MA and MB are determined. During the range is within O-E, avoiding the ranges OO’ and EE’, where
measurement, the planar readout coils are excited one at a time. nonlinearity is high as indicated in Fig. 4. Three planar coils are
When one coil is excited, the other coil is kept in an open circuit. used for demonstration; however, this can be extended to any
Therefore, the measured mutual inductance is solely between number according to the required range of measurement. The
the coil that is excited and the floating coil. From the measured coils were bent to fit the curvature of the cylindrical container
values of MA and MB, the height of the liquid is determined as: and stuck to it in a way that the printed side faces the container.
This whole arrangement serves as the readout coil set. The
M A − MB floating coil is a wire wound coil with 6 turns and a diameter of
h= . (9)
M A+ M B 160 mm. The glass container used in the prototype has a smooth
inner surface. Hence the friction between the float and the
V. EXPERIMENTAL RESULTS container is very low. Additionally, the radius of the floating
coil is kept 1 mm smaller than that of the container such that it
To demonstrate the feasibility of the proposed level sensor, moves in the container without touching the walls and hence
a prototype sensor was developed and tested. The sensor was without any friction. The floating coil is wound over a structure
tested in a glass beaker to measure the level of water contained printed using PLA (Polylactide) to make it float in the water. A
in it. A photograph of the measurement setup, with the ceramic capacitor, CF, was connected in parallel with the
developed sensor incorporated in the glass beaker, along with floating coil. The weight of the floating coil was 49 g, and that
the readout coils and the floating coil, is shown in Fig. 8 (a). The of the capacitor was 0.12 g. Due to the smaller size and weight

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

TABLE I. PARAMETERS OF THE PROTOTYPE SENSOR

Without floating coil FFT of vi

Values vS1
Parameters
A B C
Dimensions 114 mm × 114 mm
Readout Number of turns 55
Coil Coil Inductance (µH) 151.8 152.2 149.6 vi
Coil Resistance (Ω) 19.7 20.7 20.3
Diameter of the coil 160 mm
Floating Number of turns 6
Coil Coil Inductance (µH) 15
(a)
Coil Resistance (Ω) 0.46

of the capacitor in comparison with the floating coil, there were With floating coil FFT of vi

no balancing issues for the float. vS1

A photograph of the floating part of the prototype sensor, fF

along with the coil and the capacitor is shown in Fig. 8(b). If the
container wall is not polished and has ridges, a casing, as shown vi
in Fig. 8(c) can be used to minimize friction. The value of
capacitor CE was chosen as 100 µF, the value of CF was chosen
as 10 nF and the value for VR was set as 15 V. The demultiplexer
IC MAX4709 served as the switches and the op amp OPA551
(b)
is used for realizing the i-to-v converter. Digilent Analog
Discovery 2 was used for acquiring of the capacitor voltage and Fig. 9. Relevant waveforms with and without the floating coil. (a) Plot of vi and
the circuit current. The control of the switches, acquisition of vS1 without the floating coil. FFT of vi shown inset. (b) Plot of vi and vS1 with the
voltage and current, and implementation of the measurement floating coil in place. FFT of vi shown inset.
algorithm were implemented using a specially developed virtual
instrument under the LabVIEW environment. The important frequency [18]. From the expression of the equivalent
parameters of the prototype sensor are listed in Table I. In order impedance seen from the readout coil side, given in [18], the
to test the prototype sensor, water is added to the container in local minimum in between the two peaks corresponds to the
steps of 50 ml, which corresponds to an increment of 2.42 mm parallel resonance frequency, fF. At the parallel resonance, the
along h. For each step, the position of the floating coil was impedance is maximum, and the corresponding voltage vi will
determined and recorded. be minimum. Fig. 9(b) shows the output of the i-to-v converter
with the floating coil in place. The two series resonance
Fig. 9 shows the relevant waveforms obtained from the frequencies and the parallel resonance frequency, fF are visible.
circuit, vi, along with the control signal vS1, with and without the From the figure, fF was found to be 432 kHz. Calculations with
floating coil in place. Fig. 9(a) shows the voltage output of the the LF and CF values indicate that fF to be 411 kHz. The
i-to-v converter without the floating coil in place. When the difference between theoretical and practical value for the
floating coil (LC tank circuit) is introduced, the system will have frequencies is due to the presence of distributed and parasitic
two series resonance frequencies and one parallel resonance capacitances. MA, MB and MC were computed using (7) and
plotted as shown in Fig. 10(a). Fig. 10(b) shows the plot of the
sum of the mutual inductances with the height.
ALGORITHM 1: PSEUDO-CODE
As per (9), a ratio-metric calculation of the two piece-wise
1: If ((MA+MB+MC) < x) && (MC < MA) && (MC < MB); linear sections of the mutual inductances that change in a push-
then pull manner, within the liquid surface is required to resolve the
2: compute output = G1MA + k1; position of the floating coil h. To resolve which two out of the
three mutual inductances to be employed for computation of h,
3: else if (MC < MA) && ((MA+MB+MC) > x); then a simple algorithm is used. The algorithm given in terms of a C-
4: compute output = G2MAB* + k2; style pseudocode is presented below.

5: else if (MA < MC) && ((MA+MB+MC) > x); then In Algorithm 1, k1, k2, k3, and k4 are the constants used to
adjust the offset and G1, G2 and G3 is used to adjust the gain. For
6: compute output = G2MBC** + k3; the prototype, x = 6.2, k1 = 2.38, k2 = 0.67, k3 = 1.1, k4 = ˗0.62,
G1 = ˗1/µH, G2 = 2.25, and G3 = 1/µH, the final output h obtained
7: else if ((MA+MB+MC) < x) && (MA < MB) && (MA < is as shown in Fig. 11. Since the output is always obtained using
MC); then the ratio-metric computation as given in (9), the input-output
8: compute output = G3MC + k4; characteristics obtained is not only linear, it is also insensitive
to lateral misalignments. From Fig. 10 and Algorithm 1, it is
9: end if seen that a ratio-metric output is obtained in the regions h1 to h2
*MAB=(MA-MB)/(MA+MB)
and h2 to h3. This would result in a linear output proportional to
**MBC=(MB-MC)/(MB+MC) h, and at the same time, less sensitive to any lateral

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

(a)

h1 h2 h3 h4
(b)

Fig. 10. Experimental output (a) Plot of the mutual inductances MA, MB, MC
with height of the liquid column (b) Plot of MA+MB+MC with height. Fig. 12. Change in output with intentionally introduced misalignment in the
horizontal direction. The variation in the final ratiometric output with
misalignment is shown inset.

The output signal vi is found to decrease with the increase in

lateral misalignment as expected. As the same amount of
decrease occurs in the output signals from both the planar coils,
the ratio-metric output remains very less affected.
2) Effect of Temperature and Change in Density of the Liquid
When the temperature of the liquid changes, the resistance R
of the floating coil will change as in (10). In (10), Rref is the
resistance of the conductor at reference temperature Tref, T is the
current temperature of the conductor, and α is the temperature
coefficient of resistance of the conductor material [19].
 (
R = Rref 1 +  T − Tref , ) (10)

The liquid used for testing the prototype was water, hence a
temperature range from room temperature (20° C) to 100° C was
Fig. 11. Final ratiometric output along with percentage nonlinearity.
studied. The resistance Rref of the floating coil at Tref = 20° C is
460 mΩ. From (10), the value of R at 100° C becomes 607 mΩ,
misalignments or factors that would affect both the mutual where α = 0.0038 for Copper. To emulate this, a variable
inductances. However, this feature is not achievable at the ends
of the readout coil, 0 to h1 and h3 to h4, but a linear output
proportional to h is obtained here as well. The error in the FFT of vi
measured height h is also plotted in Fig. 11. From Fig. 11 we
see that the percentage nonlinearity is less than 1% of full scale
(FS).

VI. SENSOR ANALYSIS AND DISCUSSION

fF
The sensor was experimentally analyzed to quantify the
effects of lateral misalignment, variation in the temperature of
the liquid, and density. The details of the study are presented
below.
1) Effect of Lateral Misalignment of the Floating Coil
The proposed sensor is designed keeping a small gap between
the floating coil and the container to ensure friction-free travel
of the floating part in the vertical direction. Due to this gap, there
is a possibility of lateral movement of the floating part. To
evaluate the effect, the horizontal misalignment was
intentionally introduced in the prototype, in the range of 2 mm
to 6 mm in steps of 2 mm. The output recorded are plotted and Fig. 13. Plot of the output signals with and without Rext and their difference.
Plot of the FFTs of the outputs with and without Rext is shown inset.
given in Fig. 12. The final ratio-metric output is shown inset.

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

resistor, Rext, of 150 mΩ was introduced in series with the [26] are intrusive type. The installation, and maintenance of the
floating coil, and the output vi was recorded. The recorded proposed sensor are very easy as it is nonintrusive and requires
output waveforms, with and without Rext, along with their no electrical connection to the floating part. The prototype
difference, is plotted in Fig. 13. The FFTs of the output vi, with sensor developed has a resolution of 0.7 mm, while the level
and without Rext is shown inset. Even though the quality factor sensors in [23] and [25] have resolution of 2.54 mm and 1 mm,
of the system decreased with an increase in the resistance of the respectively. The proposed sensor is easy to manufacture as it
floating coil, the oscillations were still present, and the parallel uses planar coil structures. The sensor has a measuring error of
resonance was correctly observed at fF. The final ratio-metric < 1%, which is comparable with that of the other intrusive
output was calculated with Rext, and the output had a decrease of
sensors in the market [23]-[25].
1.3 %. Also, at 100° C, the density of water decreases from 998
kg/m3 to 958 kg/m3 [20]. Since the float is overdesigned, 4%
variation in the density will not introduce a noticeable error in VIII. CONCLUSION
the output. A linear nonintrusive inductive type liquid level
measurement system was designed and developed. The
The sensor was prototyped using a glass beaker with a
proposed system is suitable for sealed containers as well. The
uniform circular cross-section. Once the outer diameter D of the
sensor consists of a lightweight, wire wound coil on a flexible
container is known, one can select the height (= width) of the
floating structure which can be easily pushed into the container
planar coil. As the planar coil employed is of square-shaped, the
before it is sealed. A set of planar coils with square-shaped spiral
maximum height of the coil can be D. In such a case, the coil
patterns etched on it using PCB fabrication technology are
will completely surround the container. In practice, the user may
stacked one above the other with about 60% overlap, and they
prefer to use, say, only 1/4th of the periphery of the container
are bent and wrapped around the walls of the container. This
(like in the prototype developed). In such a case, the coil height
serves as the readout coils and outputs the change in mutual
will be D/4. For tall containers, multiple planar readout coils inductance due to the changing position of the floating coil,
will be required to get the required range, i.e., the number of which is proportional to the height of the liquid column. The
planar coils in the readout can be increased in the vertical sensor designed and developed has a range of 180 mm, and a
direction to increase the vertical sensing range [21]. For any resolution of 0.7 mm, and the worst-case percentage
shaped container with a uniform cross-section, the float and the nonlinearity was less than 1%. The performance analysis of the
floating coil can be designed according to the shape and size of sensor indicates that the output is insensitive to any
the container. The sensitivity and resolution of the proposed misalignment in the x-y plane. No electrical connection to the
sensor are limited by the inter-turn distance in the readout coil. moving part of the sensor makes it more reliable and improves
Hence, once the height of the readout coil is decided, it is its operational life. It is easy to manufacture at a low cost and
recommended to include as many number of turns as possible, will be useful for many industrial level sensing applications.
keeping the minimum gap between the turns limited by the
fabrication facility. The number of turns of the floating coil is
REFERENCES
designed in such a way that it gives sufficient mutual inductance
[1] ElProCus, “Different types of level sensors and their workings,” Elprocus,
that is measurable by the measurement system. In addition, the June 23, 2019. [Online]. Available: https://www.elprocus.com/different-
coil resistance should be kept minimum to achieve high quality types-level-sensors-applications/
factor for the LC tank circuit. The weight of the floating coil [2] Senix distance and level sensors, “Liquid Level Sensor Applications,”
should be minimum to ensure that the float will not sink in the Senix distance and level sensors, June 22, 2019. [Online]. Available:
https://senix.com/liquid-level/
liquid. Another factor to be considered is the effect of vibrations [3] Alan S. Morris, “Level Measurement,” in Measurement and
on the sensor output. The sensor output will be sensitive to the Instrumentation Principles, 3rd ed. Oxford, U.K.: Butterworth-Heinemann,
waves on the surface of the liquid if it is exposed to vibrations. 2001, pp. 340-351.
Computing the average of the output for a specified time [4] William C. Dunn, “Level,” in Introduction to Instrumentation, Sensors
and Process Control, Norwood, MA: Artech House, Inc., 2005,
window will remove the effect from the output. pp. 115-128.
In the prototype sensor, the floating part of the floating coil is [5] S. C. Bera, H. Mandal, S. Saha and A. Dutta, "Study of a Modified
printed using Polylactide (PLA). Other than this, lightweight Capacitance-Type Level Transducer for Any Type of Liquid," IEEE
and heat resistant plastic materials typically used in 3D printing Trans. Instrum. Meas., vol. 63, no. 3, pp. 641-649, March 2014.
such as Alumide or the commonly used float materials in [6] S. Chakraborty, S. K. Bera, N. Mandal and S. C. Bera, "Study on Further
commercially available float switches like Polypropelene (PP) Modification of Non-Contact Capacitance Type-Level Transducer for a
Conducting Liquid," IEEE Sensors J., vol. 15, no. 11, pp. 6678-6688, Nov.
or Polyphenylene Sulfide (PPS) can be used as the floating part 2015.
as it can withstand relatively high thermal conditions. PP can [7] H. Canbolat, "A Novel Level Measurement Technique Using Three
withstand up to 130° C, and PPS can withstand up to 220° C Capacitive Sensors for Liquids," IEEE Trans. Instrum. Meas., vol. 58, no.
without deformation [22]. 10, pp. 3762-3768, Oct. 2009.
[8] D. Wang, Y. Zhang, B. Jin, Y. Wang and M. Zhang, "Quasi-Distributed
Optical Fiber Sensor for Liquid-Level Measurement," IEEE Photonics J.,
VII. COMPARISON STUDY vol. 9, no. 6, pp. 1-7, Dec. 2017.
The proposed sensor has numerous applications in industry, [9] G. Betta, A. Pietrosanto and A. Scaglione, "A Gray-code-based fiber optic
liquid level transducer," IEEE Trans. Instrum. Meas., vol. 47, no. 1, pp.
including chemical, pharmaceutical, and food, where 174-178, Feb. 1998.
nonintrusive sensing of liquid level in sealed containers is a
requirement. Most of the commercially available sensors, [23] -

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2020.2982842, IEEE
Transactions on Instrumentation and Measurement

[10] M. R. Nabavi and S. Nihtianov, "Eddy-Current Sensor Interface for [19] E. O. Doebelin, “Temperature and Heat-flux Measurements,” in
Advanced Industrial Applications," in IEEE Transactions on Industrial Measurement Systems Application and Design, 4th ed, USA: McGraw-Hill
Electronics, vol. 58, no. 9, pp. 4414-4423, Sept. 2011 Publishing Company, 1990, ch. 8, pp. 643-648.
[11] Y. M. Tan, W. Yin and A. J. Peyton, "Non-contact measurement of water [20] M. Pauken, “Laying the foundation of Thermodynamics,” in
surface level from phase values of inductive measurements," Proc. IEEE Thermodynamics for Dummies, NJ.: John Wiley & Sons, Inc., 2011, ch. 2,
Int. Instrum. Meas. Technol. Conf. (I2MTC) Proc, Graz, 2012, pp. 1621- pp. 21-22.
1624. [21] N. Anandan, A. Varma Muppala and B. George, "A Flexible, Planar-Coil-
[12] Q. Zhao, K. Xu, Y. Xie and W. Yin, "Measurement of liquid level with a Based Sensor for Through-Shaft Angle Sensing," IEEE Sens. J., vol. 18,
small surface area using high frequency electromagnetic sensing no. 24, pp. 10217-10224, 15 Dec.15, 2018
technique," Proc. IEEE Int. Instrum. Meas. Technol. Conf. (I2MTC), Pisa, [22] Polymer Properties, Omnexus, The material Selection, accessed on Feb.
2015, pp. 1414-1419. 24, 2020 [Online]. Available: https://omnexus.specialchem.com/polymer-
[13] B L Hauge and L E Helseth, “An electromagnetic resonance circuit for properties/properties/max-continuous-service-temperature
liquid level detection,” European Journal of Physics, vol.33, no. 3, pp. [23] Modbus Ultrasonic Level Sensor, Automation Products Group, Inc.,
525-530, March 2012. accessed on Nov. 21, 2019 [Online]. Available:
[14] W. Yin, A. J. Peyton, G. Zysko and R. Denno, "Simultaneous Noncontact https://store.apgsensors.com/default.aspx?appname=Store&tabname=Pro
Measurement of Water Level and Conductivity," IEEE Trans. Instrum. ducts&PartID=124124-8001
Meas., vol. 57, no. 11, pp. 2665-2669, Nov. 2008. [24] Capacitive probe for level measurement in liquids, Endress+Hauser
[15] S. Marick, S. K. Bera, S. C. Bera, “A Float Type Liquid Level Measuring (India) Pvt. Ltd., accessed on Nov. 21, 2019 [Online]. Available:
System Using a Modified Inductive Transducer,” Sensors & Transducers, https://www.e-direct.endress.com/in/en/two-rod-capacitive-probe-
vol. 182, no. 11, Nov. 2014, pp. 111-118. liquicap-t-fmi21-for-level-measurement-in-liquids
[16] F. W. Grover, “Parallel elements of equal length,” in Inductance [25] Low Cost Submersible Liquid Level Transmitter Series: PT-510/510W
Calculations, N. Y.: Dover Publications, Inc., 1946, ch. 5, pp. 31-35. Datasheet, Automation Products Group, Inc., accessed on Nov. 21, 2019
[17] A. Babu and B. George, "An Efficient Readout Scheme for Simultaneous [Online]. Available: https://www.apgsensors.com/sites/default/files/datas
Measurement From Multiple Wireless Passive LC Sensors," IEEE Trans. heets/PT-510-510W.pdf
Instrum. Meas., vol. 67, no. 5, pp. 1161-1168, May 2018. [26] Free space radar sensor for non-contact level measurement,
[18] A. Babu and B. George, “A linear and high sensitive interfacing scheme Endress+Hauser (India) Pvt. Ltd., accessed on Nov. 21, 2019 [Online].
for wireless passive LC sensors,” IEEE Sensors J., vol. 16, no. 23, pp. Available: https://www.e-direct.endress.com/in/en/radar-sensor-for-non-
8608–8616, Dec. 2016. contact-level-measurement-micropilot-fmr10

0018-9456 (c) 2020 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.
Authorized licensed use limited to: Macquarie University. Downloaded on April 11,2020 at 15:08:15 UTC from IEEE Xplore. Restrictions apply.