sensors
Article
Passive Wireless Pressure Gradient Measurement System for
Fluid Flow Analysis
Partha P. Dutta 1 , Alexander C. Benken 1, *, Tao Li 2 , John Richard Ordonez-Varela 3
and Yogesh B. Gianchandani 1
Citation: Dutta, P.P.; Benken, A.C.; Li,
T.; Ordonez-Varela, J.R.; Gianchandani,
Y.B. Passive Wireless Pressure
Gradient Measurement System for
Fluid Flow Analysis. Sensors 2023, 23,
2525. https://doi.org/10.3390/
s23052525
Academic Editors: Edmond Cretu
and Chang Ge
Received: 29 January 2023
Revised: 21 February 2023
Accepted: 22 February 2023
Published: 24 February 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Sensors 2023, 23, 2525. https://doi.org/10.3390/s23052525
1 Center for Wireless Integrated MicroSensing and Systems (WIMS2 ), ECE Division, EECS Department,
University of Michigan, Ann Arbor, MI 48109, USA
2 Department of Electrical and Computer Engineering, University of Cincinnati, Cincinnati, OH 45219, USA
3 TotalEnergies, Centre Scientifique et Technique Jean Féger (CSTJF), Av. Larribau, CEDEX, 64018 Pau, France
* Correspondence: acbenken@umich.edu
Abstract: Using distributed MEMS pressure sensors to measure small flow rates in high resistance
fluidic channels is fraught with challenges far beyond the performance of the pressure sensing
element. In a typical core-flood experiment, which may last several months, flow-induced pressure
gradients are generated in porous rock core samples wrapped in a polymer sheath. Measuring
these pressure gradients along the flow path requires high resolution pressure measurement while
contending with difficult test conditions such as large bias pressures (up to 20 bar) and temperatures
(up to 125 ◦ C), as well as the presence of corrosive fluids. This work is directed at a system for using
passive wireless inductive-capacitive (LC) pressure sensors that are distributed along the flow path to
measure the pressure gradient. The sensors are wirelessly interrogated with readout electronics placed
exterior to the polymer sheath for continuous monitoring of experiments. Using microfabricated
pressure sensors that are smaller than ø15 × 3.0 mm3 , an LC sensor design model for minimizing
pressure resolution, accounting for sensor packaging and environmental artifacts is investigated and
experimentally validated. A test setup, built to provide fluid-flow pressure differentials to LC sensors
with conditions that mimic placement of the sensors within the wall of the sheath, is used to test
the system. Experimental results show the microsystem operating over full-scale pressure range of
20,700 mbar and temperatures up to 125 ◦ C, while achieving pressure resolution of <1 mbar, and
resolving gradients of 10–30 mL/min, which are typical in core-flood experiments.
Keywords: physical sensors; energy industry; high resolution; differential; harsh environment
1. Introduction
As the need for wireless, compact, and cost-effective sensing solutions rises in de-
mand [1–4], microelectromechanical systems (MEMS) become a necessary replacement
for conventional macro-scale sensors [5,6]. Considered one of the greatest successes of
the MEMS industry, pressure sensors have been widely utilized in industrial, automotive,
and healthcare sectors [7–9]. Prior work has primarily focused on exploration of pressure
sensing methodologies [10–13] such as piezoresistive [14], capacitive [15], micro-plasma
discharge [16,17], and optical fiber Fabry–Perot [18,19], along with specific advancements in
sensor properties, such as reduction in size and temperature coefficients or improvements
in sensitivity over a large full-scale range [20]. While both piezoresistive and capacitive
sensing technologies are widely manufacturable, capacitive pressure sensors allow scalable
precision over a wide operating temperature range [21,22]. Capacitive pressure sensors
with multiple sensing diaphragms also improve sensitivity and reduce noise [23,24]. Over
the years, several research investigations have been directed at fluid flow using microfabri-
cated pressure sensors; these range from directing flow through an orifice in a piezoresistive
pressure sensing diaphragm [25] to using a differential pressure sensor to measure pres-
sure drop across two sensing nodes along a flow channel [26]. While these efforts have
https://www.mdpi.com/journal/sensors
Sensors 2023, 23, 2525 2 of 21

significantly advanced our capabilities to study pipe flow rates and flow-induced pressure
gradients, understanding these gradients across multiple nodes along a long flow channel,
with varying fluidic resistance, and at small flow rates remains challenging. Gradient
determination requires a full system solution which extracts miniscule pressure differen-
tials across multiple sensing nodes that are precisely located. In addition, environmental
constraints such as a large baseline pressure head, a broad operating temperature range,
and corrosive test fluids influence the pressure resolution through the overall system de-
sign strategy, integration, and packaging, far beyond the sole capabilities of the pressure
sensing element.
Whereas a variety of fluidic systems may benefit from the type of distributed pressure
gradient measurement capability described above, core-flood experiments present an inter-
esting test case. Core-flood experiments refer to the study of the porosity and permeability
of rock core samples by subjecting them to pressurized fluid flow; they are used in the
energy industry to plan and assess oil and natural gas recovery strategies [27,28]. The rock
core sample is encased in a polymer sheath and then placed inside a high pressure (HP)
chamber test cell, with lateral pressure up to 250 bar and temperature up to 125 ◦ C [29].
Fluids are pumped into the rock core longitudinally at pressure levels up to 20 bar. Al-
though the resulting pressure gradients provide valuable information, the measurement
locations are limited to the ends of the core. Pressure measurements along the length of the
rock core could improve insight as well as permit early detection of failure in these tests,
which sometimes extend for months [30]. In past work, X-ray CT scans [31] were used
for understanding spatial distribution of fluids and minerals along the rock core during
a core-flood experiment. The recorded images helped explain the formation of vugs and
wormholes inside the core, thus providing information on the changing permeability of the
rock core. Contrary to this high-cost approach which uses custom test setups and image
processing algorithms, the physical measurement of forming pressure gradients along the
rock core can provide a low cost, continuous monitoring alternate to extract changing rock
core permeability information, derived from Darcy’s law [32], as simply represented for a
continuous porous medium by Equation (1):

Φµ f ∆L
K= (1)
A∆P
where K denotes permeability of rock core, Φ denotes flow rate of test fluid, A denotes
cross-sectional area of rock core, µf denotes viscosity of test fluid, and ∆P denotes measured
pressure gradient across a ∆L length of the rock core.
In order to realize a pressure gradient measurement system that addresses core-flood
experiments, several challenges must be addressed: Any interference with the fluid flow
that is caused by the presence of the sensor must be minimized. This necessitates that only
miniature sensing elements be placed along the flow path (at interface between the rock core
and polymer sheath), whereas all readout electronics must be placed exterior to the flow
chamber. The electrical lead transfer from sensor to interrogation electronics through flow
path boundary across the polymer sheath must allow a hermetic seal. The sensing elements
must be packaged in a manner that allows pressure to be transmitted through the package,
while providing the necessary robustness to withstand large mechanical forces during rock
core insertion into polymer sheath, a large baseline pressure head during the core-flood
experiment, and the presence of corrosive test fluids such as brine and organic solvents.
In the presence of high lateral pressure and temperature inside the flow chamber, the
sensors must measure pressure gradients on the order of few mbar induced by the small
flow rates, which are typically 10–30 mL/min. One method for overcoming these problems
is to employ an inductive-capacitive (LC) transduction method. This approach utilizes
a fully passive sensing element, permitting great reduction in sensor size by eliminating
the largest components (e.g., battery and electronics) from the sensed environment [33].
Furthermore, it removes the requirement of a wired connection between sensor and readout
electronics [34]. However, its success depends on the integration strategy for interface
 coil, and corresponding readout nodes to remotely interrogate individual LC sensors.
Each readout node consists of an inductor coil, standing wave ratio bridge circuit, and
sinusoidal excitation circuitry; a microcontroller unit (MCU) controls each node and dig-
itizes the collected data. An external Raspberry PiTM (R-Pi) microcomputer (Raspberry
Sensors 2023, 23, 2525 PiTM Foundation, UK) interfaces with the MCU, collecting stored data and running 3 ofpro-
21
cessing algorithms to extract pressure dependent resonant frequency. A laptop connected
to R-Pi permits user control through graphical user interface (GUI) and can upload data
to the cloud
electronics, for further
which processing and
must interrogate the LCremote retrieval.
sensor to determine the pressure-dependent
Although this approach does not
resonant peak [35], and perform the necessary signal limit the number of LC sensors
conditioning andthat can be incorpo-
post-processing,
rated within a PGM system, the manifestation described in this paper
and a robust sensor packaging approach which preserves the necessary pressure sensitivity uses four LC sensors
incorporated
and resolution. into a flow test setup, wirelessly interrogated by a readout printed circuit
boardThis(PCB),
paper and an external
describes a passive control
wirelessunitpressure
locatedgradient
outside measurement
the flow chamber. (PGM)The PGM
system,
system
shown inwas successfully
Figure 1, that uses used to measure
wireless flow-induced
LC pressure sensing pressure
elementsgradients
(denoted with
as LCasensors,
pressure
LCresolution of <1 mbar,
n ), each comprised ofmeasuring
a capacitive ≈ 10 mbar pressure
pressure transducer dropandbetween
planar adjacent
inductivesensors
coil, andat a
flow rate of 14 mL/min representing values for a typical core-flood
corresponding readout nodes to remotely interrogate individual LC sensors. Each readout experiment. If de-
ployed in a core-flood setup, the four LC sensors would be arranged
node consists of an inductor coil, standing wave ratio bridge circuit, and sinusoidal excita- along the rock core
within
tion the polymer
circuitry; sheath in the
a microcontroller unitflow
(MCU)chamber, readout
controls PCB with
each node interrogation
and digitizes electron-
the collected
ics may
data. An be placedRaspberry
external on the outsidePiTM of the polymer
(R-Pi) sheath within
microcomputer the HP
(Raspberry TM Foundation,
Pichamber, and ex-
ternal
UK) controlwith
interfaces unitthemayMCU,be placed outside
collecting thedata
stored testand
cell running
to permit real-time algorithms
processing user interface.
to
The system
extract pressure design is described
dependent in Section
resonant 2, while
frequency. fabrication
A laptop and packaging
connected are described
to R-Pi permits user
in Section
control through3. Experimental
graphical user test results(GUI)
interface are described in Section
and can upload data4,tofollowed
the cloud byfor
discussion
further
processing
in Sectionand remote
5, with retrieval. and summary in Section 6.
a conclusion

Figure
Figure 1. 1.Cross-sectional
Cross-sectionalview
viewofofa apassive
passivewireless
wirelesspressure
pressuregradient
gradientmeasurement
measurement(PGM)
(PGM)system.
system.
LC sensors and readout nodes can be scaled to n elements, n = 4 in this
LC sensors and readout nodes can be scaled to n elements, n = 4 in this work. work.

Although this approach does not limit the number of LC sensors that can be incorpo-
rated within a PGM system, the manifestation described in this paper uses four LC sensors
incorporated into a flow test setup, wirelessly interrogated by a readout printed circuit
board (PCB), and an external control unit located outside the flow chamber. The PGM
system was successfully used to measure flow-induced pressure gradients with a pressure
resolution of <1 mbar, measuring ≈ 10 mbar pressure drop between adjacent sensors at a
flow rate of 14 mL/min representing values for a typical core-flood experiment. If deployed
in a core-flood setup, the four LC sensors would be arranged along the rock core within the
polymer sheath in the flow chamber, readout PCB with interrogation electronics may be
placed on the outside of the polymer sheath within the HP chamber, and external control
unit may be placed outside the test cell to permit real-time user interface. The system
design is described in Section 2, while fabrication and packaging are described in Section 3.
Experimental test results are described in Section 4, followed by discussion in Section 5,
with a conclusion and summary in Section 6.
Sensors 2023, 23, x FOR PEER REVIEW 4 of 21
Sensors 2023, 23, 2525 4 of 21

2. System Design
2. System Design
This section describes
This section the main
describes thecomponents and design
main components and methodology used in realiz-
design methodology used in real-
ing theizing
PGM the PGM system. The first subsection provides the circuit modelmathematical
system. The first subsection provides the circuit model and and mathematical
background of the of
background inductor-capacitor (LC) sensor
the inductor-capacitor and the
(LC) sensor andreadout circuit.circuit.
the readout Based Based
on thison this
mathematical model and analysis, the LC sensor design strategy is used
mathematical model and analysis, the LC sensor design strategy is used in the secondin the second
subsection to obtain
subsection a figure
to obtain of merit
a figure that isthat
of merit related to the to
is related pressure resolution
the pressure of the of
resolution LCthe LC
sensor;sensor;
this figure
this of meritofaccounts
figure for impact
merit accounts for of LC sensor
impact of LCpackaging materials materials
sensor packaging and de- and
ployment environment.
deployment The final subsection
environment. addresses addresses
The final subsection the readout circuit
the design
readout as real-
circuit design as
ized on a flexible printed circuit board.
realized on a flexible printed circuit board.

2.1. LC2.1.
Sensor Model Model
LC Sensor
The pressure
The pressure sensingsensing
elementelement
of the of
PGMthe system
PGM system
is the is
LCthe LC sensor,
sensor, comprised
comprised of a of a
capacitive
capacitive pressure
pressure transducer, PXDCRP
transducer, , with
XDCR , with variable
variable capacitance,
capacitance, C XDCR C
, and
XDCR , and inductive
inductive coil coil
with inductance, L ; a
with inductance, LS; a simplified
S simplified circuit model is shown in Figure 2a. Here, C
circuit model is shown in Figure 2a. Here, CPar is parasiticPar is parasitic
capacitance of an LC sensor; R parasitic series resistance of an LC sensor coil. The The
capacitance of an LC sensor; RS,Coil isS,Coil is parasitic series resistance of an LC sensor coil.
equivalent
equivalent series series resistance,
resistance, ESR, results
ESR, results from resistance
from resistance of thinoffilm
thinmetal
film metal electrodes
electrodes of of
PXDCR. PRelevant
XDCR . Relevant
design design
equationsequations are summarized
are summarized in Tablein1.Table 1.

(a) (b)
Figure 2. (a) Wireless inductive Pressure Gradient Measurement (PGM) System circuit model illus-
Figure 2. (a) Wireless inductive Pressure Gradient Measurement (PGM) System circuit model
trating the LC sensor and readout coil and circuitry. (b) Two Re{Zin} datasets illustrating relevant
illustrating the LC sensor and readout coil and circuitry. (b) Two Re{Zin } datasets illustrating
parameters.
relevant parameters.
The resonant frequency of the LC sensor, f0 (2) is sensitive to pressure through the
The resonant frequency of the LC sensor, f 0 (2) is sensitive to pressure through the
changes in sensor
changes capacitance,
in sensor CXDCR. When
capacitance, CXDCRthe . Whenreadout coil withcoil
the readout inductance, LRO, and LC
with inductance, LRO , and
sensorLCcoilsensor
with inductance, L S, are in close proximity, they become coupled, resulting in
coil with inductance, LS , are in close proximity, they become coupled, resulting
mutual ininductance, M (4). When
mutual inductance, M (4).theWhen
readout thecoil is excited
readout coil iswith input
excited voltage,
with inputVvoltage,
RO and
VRO
input current, I RO, with a resulting complex input impedance, Zin (3), the pressure-depend-
and input current, IRO , with a resulting complex input impedance, Zin (3), the pressure-
ent resonant
dependentfrequency,
resonant f0, frequency,
of the mutually coupled
f 0 , of the mutuallyLC sensor,
coupled manifests
LC sensor,as manifests
a peak in the
as a peak
real part of the input impedance, Re{Z in}, of the readout coil. The value of Re{Zin} is moni-
in the real part of the input impedance, Re{Zin }, of the readout coil. The value of Re{Zin } is
tored by the readout
monitored by the circuitry,
readoutdescribed
circuitry,below.
described below.
To extract the pressure dependent
To extract the pressure dependent f0 of thef 0 LC sensor,
of the a set ofa discrete
LC sensor, data points
set of discrete are are
data points
collected, where each data point is the magnitude of Re{Z } at the
collected, where each data point is the magnitude of Re{Zin } at the specific interrogated
in specific interrogated
frequency. Two typical
frequency. captured
Two typical datasets
captured of Re{Z
datasets ofin}Re{Z
are inshown in Figure
} are shown 2b; a Gaussian
in Figure 2b; a Gaussian
curve,curve,
referred to as fitted
referred frequency
to as fitted response
frequency waveform,
response FZin (8)FZin
waveform, is computed for each
(8) is computed for each
dataset. From this, f can be determined through interpolation.
dataset. From this, f 0 can be determined through interpolation. This method allows
0 This method allows fre-
quency resolutionresolution
frequency to be achieved to be without
achievedrequiring
withoutcollection
requiringof an intractably
collection of an large num- large
intractably
ber of number
data points.
of data points.
Bandwidth, σ, is defined as f0/Q, and mean (μ) is defined as f0. In this work, σ is de-
fined as the full width at half maximum, in the frequency span of the Re{Zin} dataset. Rin,
defined in Equation (9), is determined by first identifying the maximum value of Re{Zin},
Sensors 2023, 23, 2525 5 of 21

Table 1. Inductor-Capacitor (LC) Sensor Design Equations.

Parameter Equation
Resonant 1
f0 = √ (2)
Frequency LS (CXDCR +CPar )

Complex Input Impedance Zin = VIRO

RO (3)
√
Mutual Ind. M = k LS L RO (4)
2
Eff. Series Res. Of CXDCR ESR (CXDCR ) (5)
R ESR =
(CXDCR +CPar )2
Total Eff. LC Sensor Res. RS = R ESR + RS,Coil (6)
Total Eff. LC Sensor Cap. CS = CXDCR + CPar (7)
  !
−( f −µ)2 −( f − f 0 )2
Fitted Gaussian Curve FZin ≈ Rin exp σ2
= Rin exp  2
f
(8)
0
Q
n  o  2 
f
Mag. Of Input Impedance Rin = Re{ Zin ( f 0 )} − Re Zin 100 ≈ LMRO Q
LS
(9)

Quality f0 1
q
LS 1
q
LS
Q= ≈ = (10)
Factor σ RS (CXDCR +CPar ) RS CS
√ √  −1
Absolute ( LS CS )
−1
− LS (CS +∆C )
∆ f0 (11)
Response AR = ∆PFS ≈ ∆PFS
Relative 
AR
 
AR· Q

RR = = (12)
Response σ f0
h i
Figure of Merit, FOM AR· Q · Rin (13)
FOM = RR· Rin = f0

Simulated 
f0

ResolSim = C (14)
Pres. Resol. AR· Q· Rin

Bandwidth, σ, is defined as f 0 /Q, and mean (µ) is defined as f 0 . In this work, σ is

defined as the full width at half maximum, in the frequency span of the Re{Zin } dataset. <in ,
defined in Equation (9), is determined by first identifying the maximum value of Re{Zin },
found at approximately f 0 , and minimum value of Re{Zin }, found far away from f 0 (at
approximately f 0 /10); it can also be estimated using M, Q, LRO , and LS . The quality factor,
Q, defined in (10), is f 0 divided by σ; it can also be estimated using total effective resistance,
RS (6), LS , and total capacitance, CS (7). The shift in resonant frequency, ∆f 0 , with applied
pressure, ∆P, (caused by the change in capacitance, ∆C, of the transducer over the full-scale
pressure range, ∆PFS ) is defined as Absolute Response, AR, (11). AR when normalized to
the bandwidth, σ is defined as Relative Response, RR, (12).

2.2. Design Methodology

In an initial effort to minimize the resolvable pressure of the PGM system, a balance
must be found between <in , Q, and AR; furthermore, both the readout coil and the LC
sensor must be designed in tandem in order to ensure M is large enough such that <in is
above the minimum detectable signal for the utilized readout electronics. In designing the
LC sensor, there are two main factors that must be considered: (1) f 0 is extracted through
curve fitting of discrete data points; and (2) <in , Q and AR are often in direct conflict—as
one is increased, the other is reduced. (For example, Q can be increased by adding parallel
capacitance, additional CPar , to reduce RS contributed by transducer ESR, but this addition
will result in a reduction in AR). This inverse relationship between AR and Q balances
RR which depends on the product of these terms. The final design parameters must
remain within the physical boundary conditions of the specific application; the boundary
conditions applicable to the core-flood experiments targeted in this work are shown in
Table 2.
Sensors 2023, 23, 2525 6 of 21

Table 2. Design Constraints for LC Sensor and Readout Coil.

Design Parameter Dimension Comments

To maintain packaged LC sensor
LC sensor, max. diameter <13 mm
diameter ≤15 mm
LRO , max. diameter <25 mm LC sensor pitch is 50 mm
Min. L trace width,
0.125 mm Min. manufacturable dimension
spacing
Nominal Interrogation Polymer sheath thickness in
4 mm
Gap (IG) of LS and LRO core-flood experiment
Full-scale pressure range of
Full-scale pressure, ∆PFS 20 bar
core-flood experiments
CXDCR offset cap, C0 4.0 pF CXDCR cap. at 0 app. pres.
CXDCR cap. change Total capacitance change of CXDCR
3.2 pF
over ∆PFS , ∆CFS over ∆PFS
Equivalent Series Resistance (ESR)
CXDCR ESR 200 Ω
of CXDCR
Vertical spacing between LC sensor and
Interrogation Gap, IG 4 mm
readout coil
AM w.r.t. the centers of paired LC sensor
Axial Misalignment, AM 0–2 mm
and readout coil

The frequency range over which Re{Zin } is measured from an LC sensor (i.e., the
interrogation frequency range, IFR) must be wide enough so that curve fitting to extract
f 0 can be completed. To capture >99% of the sensor readout across the full-scale pressure,
a frequency range of approximately ±3σ plus the span of ∆f 0 is required. In addition to
IFR, the step size between data points (frequency step size, fss ) must be chosen. With fewer
data points, data collection bandwidth can be improved but curve fitting performance (and
resolution) may decrease. Empirical testing with the implemented curve fitting algorithm
within the R-Pi microcomputer revealed that performance saturated beyond a data density
of approximately 10 points within the LC sensor frequency response bandwidth, σ (f 0 /Q).
Therefore, a fss of σ/10 (f 0 /(10 Q)) is used.
The bandwidth normalization in RR facilitates the comparison of different LC sensor
AR, to select the most suitable design. With the discrete nature of the captured dataset of
Re{Zin }, it is important to note that, regardless of absolute values of Q and σ, the Re{Zin }
datasets for all LC sensors look identical over their IFR, when bandwidth normalization is
utilized. Therefore, the frequency shift due to applied pressure for a given LC sensor design
may be then easily expressed in terms of their respective signal bandwidth, as shown in
Figure 3.
As previously discussed, in addition to AR and Q (now captured in RR), <in also
impacts resolution. Assuming a constant white noise across the frequency spectrum,
<in is directly proportional to SNR. In other words, if <in is low, such as due to low
mutual coupling between sensor and readout coil (low M), the ability of the readout circuit
to accurately decipher f 0 through curve fitting shall be low. Thus, <in must exceed a
minimum threshold governed by the capabilities of the readout circuit and curve fitting
algorithm to allow accurate f 0 capture, thus being proportional to SNR. Increasing <in can
be accomplished by increasing M and/or Q of LC sensor. M is primarily dependent on
physical parameters (such as turns and diameter) and spacing and orientation between
sensor and readout inductors. Whereas Q can be increased by increasing LS or decreasing
CS , it is primarily impacted by reducing RS . <in and minimum resolvable pressure were
approximately determined to bear an empirically linear relationship (i.e., a 2× increase in
<in will improve resolution by ≈2×).
 Sensors 2023, 23, x FOR PEER REVIEW 7 of 21

Sensors 2023, 23, 2525 in Table 3. Re{Zin} is plotted over each designs normalized bandwidth, σ. Figure 3 shows
7 of 21
the shift in Re{Zin} for an applied pressure of 20 bar; the designs individually maximize
Absolute Response, AR (red), and Relative Response, RR (orange).

Figure3.3.Comparison
Comparisonof of example
example Re{Z
Re{Z in} datasets at 0 and 20 bar applied pressure for three LC
Figure in } datasets at 0 and 20 bar applied pressure for three LC sensor
sensor designs (given in Table 3). All datasets plotted over their bandwidth (i.e., bandwidth nor-
designs (given in Table 3). All datasets plotted over their bandwidth (i.e., bandwidth normalized).
malized).

Table 3. Design Parameters used for Figure 3 Simulations.

Table 3. Design Parameters used for Figure 3 Simulations.
Max.
Max.ARAR Max.RR
Max. RR Max.FOM
Max. FOM
Equ.
Equ. (Red) (Orange) (Green)
(Red) (Orange) (Green)
LS Diameter
LS Diameter — --- ø13.0
ø13.0mm
mm
TraceTrace
WidthWidth — --- 0.925 mm
0.925 mm 0.125
0.125mm
mm 0.125
0.125mm
mm
Inductor LayersLayers
Inductor — --- 22 44 22
Inductance
Inductance [µH] [μH] — --- 0.60.6 36.4
36.4 10.5
10.5
f0 [MHz]
f 0 [MHz] (2) (2) 52.8
52.8 6.7
6.7 12.5
12.5
σ [kHz]
σ [kHz] — --- 7178
7178 167
167 494
494
Q (10) 7.3 40.1 25.3
Q (10) 7.3 40.1 25.3
Rinnorm --- 1.0 0.35 0.88
Rinnorm — 1.0 0.35 0.88
AR [Hz/mbar] (11) 0.189 0.024 0.045
AR [Hz/mbar]
RR [ppm/mbar] (11) (12) 0.189
827 0.024
4163 0.045
2662
RR [ppm/mbar]
FOMnorm (12) (13) 827
0.33 4163
0.63 2662
1.00
FOMnorm (13) 0.33 0.63 1.00
The design maximizing AR has the smallest RR (as its σ is >14x larger than the other
designs).
To furtherThe design maximizing
illustrate how Q, RR, RRAR,
results
andin theare
<in smallest Rin (due to
interconnected, high 3Rshows
Figure S of its LS
a
coil). However, the design maximizing FOM, which is the product of
SPICE simulation of three Re{Zin } datasets for different designs using parameters as listed RR and R in, Equation

in(13),
Tableresults
3. Re{Zin middling RR and Rin values, shown in green. This parameter is suitable as
in } is plotted over each designs normalized bandwidth, σ. Figure 3 shows
a figure
the shift in ofRe{Z
merit }because it accounts for the combined impact of Q and AR through RR,
in for an applied pressure of 20 bar; the designs individually maximize
Equation (12), as
Absolute Response, AR (red),well as theand
impact of RResponse,
Relative in on the SNR. The design with the maximum
RR (orange).
value
The design maximizing AR has the smallestdesign
returned by the FOM is the LC sensor RR (aswhich
its σ iswill
>14theoretically
× larger thanpermit
the other the
minimum resolvable pressure. The inverse of FOM is proportional to pressure
designs). The design maximizing RR results in the smallest <in (due to high RS of its LS coil). resolution;
when a fitted
However, proportionality
the design maximizing factor,
FOM, C, which
is included,
is the absolute
product of values for <
RR and predicted pressure
in, Equation (13),
resolution,
results ResolSim, RR
in middling canand
be found, Equation
<in values, shown (14).in green. This parameter is suitable as a
figureAn of LC sensor
merit design
because optimization
it accounts program
for the combined was implemented
impact of Q and in MATLAB
AR through to sweep
RR,
physical LC sensor and readout coil design parameters within
Equation (12), as well as the impact of <in on the SNR. The design with the maximum their boundary conditions
(Tablereturned
value 2) and automatically
by the FOM iscalculate the FOM
the LC sensor for each
design which design. Approximations
will theoretically permit forthe
in-
ductance [36] and coil resistance [37] were utilized while k
minimum resolvable pressure. The inverse of FOM is proportional to pressure resolution; and M were determined
through
when finite
a fitted element analysis
proportionality (FEA)
factor, C, ismodeling
included, in the absence
absolute valuesofforaccurate
predictedclosed form
pressure
equations.ResolSim , can be found, Equation (14).
resolution,
An LC sensor design optimization program was implemented in MATLAB to sweep
physical LC sensor and readout coil design parameters within their boundary conditions
(Table 2) and automatically calculate the FOM for each design. Approximations for induc-
Sensors 2023, 23, 2525 8 of 21

Sensors 2023, 23, x FOR PEER REVIEW 8 of 21

tance [36] and coil resistance [37] were utilized while k and M were determined through
finite element analysis (FEA) modeling in the absence of accurate closed form equations.
The simplified
The simplified closed
closed form
form equations
equations that
that may
may bebe used
used toto estimate
estimate M M do
do not
not account
account
for the
for themagnetic
magneticcoupling
couplingand andleakage
leakagefluxflux [38];
[38]; additionally,
additionally, thethe magnetic
magnetic behavior
behavior of
of the
the packaging
packaging materials
materials and other
and other nonidealities
nonidealities have have an impact
an impact on theon the value.
value. Conse-
Consequently,
quently,
FEA was FEA wasinutilized
utilized COMSOL in COMSOL
MultiphysicsMultiphysics
® to extract® to extract M between LS and LRO.
M between LS and LRO . Figure 4
shows the 2D axisymmetric model geometry, including the packagethe
Figure 4 shows the 2D axisymmetric model geometry, including packagematerials
structure; structure; of
the LC sensor are noted detailed in Section 3 and listed in Table 4. The magnetic The
materials of the LC sensor are noted detailed in Section 3 and listed in Table 4. fieldmag-
lines
netic field
between lines between
readout coil and readout
LC sensor coil
atand LC sensorfrequency
an excitation at an excitation frequency
of 13 MHz ofshown
are also 13 MHz in
are also
Figure 4. shown
The field in lines
Figure 4. The
show field lines
the impact show
of the top the impact
metal coverofandthebottom
top metal
metal cover and
stiffener:
bottom
the metalofstiffener:
bending field linestheinbending of field
the presence oflines
theseinelements
the presence of these
reduces elements
coupling reduces
between the
coupling
LC sensorbetween
and center theof
LCthesensor andcoil,
readout center of the readout
moving the maximumcoil, moving theconcentration
field line maximum field to
lineouter
the concentration
edge of the to LC
the sensor
outer edge
coil. of the LC sensor coil.

Figure 4. Model geometry of LC sensor and readout coils and Magnetic field distribution at 13 MHz
Figure 4. Model
excitation. geometry
Assumed of LC
material sensor and
properties for readout coils
FEA given inand Magnetic
Table 4. field distribution at 13 MHz
excitation. Assumed material properties for FEA given in Table 4.
Table 4. Assumed material properties for FEA.
Table 4. Assumed material properties for FEA.
Density Young’s Poisson’s Elect. Cond.
Material Density Young’s Poisson’s Elect. Cond.
Material [kg/m 3] Modulus [GPa] Ratio [S/m]
[kg/m3 ] Modulus [GPa] Ratio [S/m] -2
Air - - - 1.00 × 10
Air
Aluminum 2700- 70.0- -
0.33 1.00 ×
3.77 107−2
× 10
Copper
Aluminum 8700
2700 130
70.0 0.34
0.33 3.77 ××10
5.99 1077
Mineral
Copper oil 850
8700 -
130 -
0.34 1.75
5.99××10
-1
107
Polyimide
Mineral oil 1300
850 3.10- 0.34
- 6.66
1.75 ××10
-16
10−1
VitonTM 1840 0.01 0.47 2.1 × 10-9−16
Polyimide 1300 3.10 0.34 6.66 × 10
VitonTM 2.1 × 10 as the −9
In the context of this1840
work, measured 0.01
pressure resolution,0.47ResolMeas, is defined
RMS error (e.g., ±standard deviation) of at least 50 readings (n ≥ 50) of the LC sensor taken
by the PGM
In the system.
context Thework,
of this measured pressure
measured resolution,
pressure ResolMeas
resolution, was
ResolMeascalculated
, is definedbyas the
the
RMSerror
RMS error(e.g.,
of f0 by AR converting
±standard to units
deviation) of atof pressure,
least (15): (n ≥ 50) of the LC sensor taken
50 readings
by the PGM system. The measured pressure𝑅𝑀𝑆 𝐸𝑟𝑟𝑜𝑟 𝑜𝑓Resol
resolution, 𝑓 Meas was calculated by the
RMS error of f 0 by AR converting 𝑅𝑒𝑠𝑜𝑙 = pressure, (15):
to units of (15)
𝐴𝑅
In order to verify FOM, four types of LC RMS sensors
Error were
o f f 0 fabricated on planar inductor
Resol Meas = (15)
coils using varying PXDCR protection materials, andARceramic capacitors were electrically
connected in parallel with CXDCR to emulate varying CPar values, resulting in ten distinct
In order to verify FOM, four types of LC sensors were fabricated on planar inductor
designs. A total of 50 discrete datasets for all ten designs were taken and used to both
coils using varying PXDCR protection materials, and ceramic capacitors were electrically
calculate ResolMeas and fit CPar and RS,Coil values to the LC sensor circuit in Figure 2a (using
connected in parallel with CXDCR to emulate varying CPar values, resulting in ten distinct
previously measured values for CXDCR, ESR, LS, and LRO) via SPICE modeling. These com-
designs. A total of 50 discrete datasets for all ten designs were taken and used to both
ponent values were coded into the MATLAB model and the additional CPar value was
calculate ResolMeas and fit CPar and RS,Coil values to the LC sensor circuit in Figure 2a (using
swept from 0 to 50 pF while returning ResolSim (C = 1). The ResolSim and ResolMeas for these
designs are plotted in Figure 5, showing an agreement within 11%, confirming the ability
Sensors 2023, 23, 2525 9 of 21

previously measured values for CXDCR , ESR, LS , and LRO ) via SPICE modeling. These
Sensors 2023, 23, x FOR PEER REVIEW
component values were coded into the MATLAB model and the additional CPar value9 was of 21
swept from 0 to 50 pF while returning ResolSim (C = 1). The ResolSim and ResolMeas for these
designs are plotted in Figure 5, showing an agreement within 11%, confirming the ability
ofofthe
theproposed
proposed FOM
FOM to to
reliably
reliablypredict pressure
predict resolution.
pressure Whereas
resolution. Whereas the resolution of theof
the resolution
unprotected LC sensor varied from ≈ 1 mbar (with 0 pF
the unprotected LC sensor varied from ≈1 mbar (with 0 pF additionaladditional C Par ) to ≈
CPar) to ≈1.5(with
1.5 mbar mbar
50 pF additional
(with CPar ), the
50 pF additional CParresolution of an LC
), the resolution sensor
of an protected
LC sensor with 15–5
protected withstainless steel
15-5 stainless
had
steel a larger variation
had a larger from ≈from
variation 2 mbar (with (with
≈2 mbar 0 pF additional
0 pF additional CPar≈)12
CPar ) to to mbar (with(with
≈12 mbar 50 pF 50
additional C Par ).
pF additional CPar).

Figure5.5.Verification
Figure Verificationof of
FOMFOM comparing
comparing measured
measured andand predicted
predicted pressure
pressure resolution
resolution for varying
for varying LC
LC sensor designs.
sensor designs.

Thestudy
The study yielded
yielded thethe preferred
preferred LC sensor
LC sensor design—corresponding
design—corresponding to the to the highest
highest value
value of FOM, which was 0.515—a double layer inductor with a coil outer diameter
of FOM, which was 0.515—a double layer inductor with a coil outer diameter of 13 mm of 13
mm and 17 turns, with a trace width and spacing of 0.125 mm. The associated readout
and 17 turns, with a trace width and spacing of 0.125 mm. The associated readout inductor
inductor
design haddesign had
an outer andiameter
coil outer coilofdiameter
17.5 mm of 17.5
and 21mm and
turns, 21 turns,
with with trace
trace width width and
and spacing of
spacing
0.15 mm. of 0.15 mm.

2.3.
2.3.Readout
ReadoutCircuit
Circuit
As
Asnoted
notedearlier
earlierand
andillustrated
illustratedininFigure
Figure1,1,individual
individualLC LCsensors
sensors(LC (LCnn))were
werewire-
wire-
lessly
lesslycoupled
coupled to specific readout
to specific inductor
readout coils integrated
inductor into each
coils integrated intocorresponding readout
each corresponding
circuit
readout node (Readout
circuit node Node n). Each
(Readout NodeReadout
n). EachNode, illustrated
Readout Node, in Figure 6,in
illustrated was controlled
Figure 6, was
by
controlled by an individual MCU (C8051F990, Silicon Laboratories, Inc., USA) withan
an individual MCU (C8051F990, Silicon Laboratories, Inc., Austin, TX, USA) with an
embedded
embedded12-bit 12-bitanalog
analogtotodigital
digitalconverter
converter(ADC).(ADC).AAdirectdirectdigital
digitalsynthesizer
synthesizer(DDS) (DDS)
(AD9850,
(AD9850,Analog AnalogDevices,
Devices,Inc.,
Inc.,Wilmington,
Wilmington,MA, MA,USA)USA)generated
generatedaasinusoidal sinusoidalsignalsignaltoto
excite
exciteaastanding
standingwave waveratio
ratiobridge
bridgecircuit.
circuit.The
Thereadout
readoutcoil,
coil,with
withits itsimpedance
impedancevarying varying
because
becauseofofthe thecoupled
coupledLC LCsensor,
sensor,formed
formedone onebranch
branchofofthe thestanding
standingwave waveratio ratiobridge
bridge
circuit; the remaining branches were implemented with fixed value
circuit; the remaining branches were implemented with fixed value resistors. Four volt- resistors. Four voltages
ofages
eachofbranch of the standing
each branch wave ratio
of the standing wave bridge bridge V
ratio circuit, z , Vs , V
circuit, Vs, VVr,i ,and
Vrz,, and were Vi,amplified
were am-
by differential amplifiers, realized with ADA4891 (Analog Devices,
plified by differential amplifiers, realized with ADA4891 (Analog Devices, Inc., Wilming- Inc., Wilmington, MA,
USA). Amplified signals were digitized using ADCs embedded
ton, MA, USA). Amplified signals were digitized using ADCs embedded within the MCU. within the MCU. The
Re{Z } was calculated within the MCU using (16) and (17), where
TheinRe{Zin} was calculated within the MCU using (16) and (17), wherez |Vz|, s|Vs|, |V |V |, |V |, |V r r|,and
|, and
|V i | represented peak values of the voltages
|Vi| represented peak values of the voltages [39]. [39].
 !2 
𝑉⃗
 50 50→ / V→s 𝑉⃗ + + 50 2 𝑆𝑊𝑅
50  SWR (16)
𝑅𝑒{𝑍 } = VZ
50 (𝑆𝑊𝑅 + 1)
Re{ Zin } = (16)
50 (SWR2 + 1)
1 1
𝑆𝑊𝑅 = 𝑉⃗ + 𝑉⃗ / 𝑉⃗ − 𝑉⃗ (17)
2 2
Sensors 2023, 23, 2525 10 of 21

Sensors 2023, 23, x FOR PEER REVIEW 1→ → 1→ → 10 of 21

   
SWR = Vi + Vr / V − Vr (17)
2 2 i

Sensors 2023, 23, x FOR PEER REVIEW 10 of 21

Figure
Figure 6.6.Circuit
Circuitschematic
schematicofof readout
readout node
node onon Readout
Readout PCB.
PCB.

3.3.System
SystemFabrication
Fabricationand andPackaging
Packaging
3.1. LC Sensor
3.1. Figure
LC Sensor Fabrication
Fabrication
6. Circuit schematic of readout node on Readout PCB.
TheTheLC LCsensor
sensorinductor
inductorcoil coilwas
wasfabricated
fabricatedonona aflexible flexiblepolyimide
polyimideprinted printedcircuit
circuit
3. System Fabrication and Packaging
board (PCB) substrate, manufactured by FlexPCB (Santa
board (PCB) substrate, manufactured by FlexPCB (Santa Ana, CA, USA). It contained Ana, CA, USA). It contained
elec-
3.1. LC Sensor Fabrication
electrical contact
trical contact padspads at the
at the center
center for for mounting
mounting the capacitive
the capacitive pressure
pressure transducer.
transducer. The CThe
XDCR
CXDCR Theelectrically
was LC sensor inductor coil was
connected fabricated
using Duralco on a flexible a polyimide
120, temperature printed circuit
high temperature silverfrom
epoxy
wasboard
electrically connected using Duralco 120, a high
(PCB) substrate, manufactured by FlexPCB (Santa Ana, CA, USA). It contained elec- silver epoxy Co-
from Cotronics
tronics (Brooklyn,(Brooklyn,
NY, NY,
USA). USA).
Capacitive Capacitive
pressure pressure
transducerstransducers
used
trical contact pads at the center for mounting the capacitive pressure transducer. The CXDCR in used
this in
workthis work
were de-
were developed
was at
electrically the University
connected using of Michigan
Duralco 120, a [23].
high These
temperature capacitive
veloped at the University of Michigan [23]. These capacitive pressure transducers provide silver pressure
epoxy from Co-transducers
provide
special special
tronics
propertiesproperties
(Brooklyn, NY,found
not notCapacitive
USA). found
in inpressure
commercially
commercially transducers
available available
used transducers
in this
transducers work such
high as
wereasde-
such high
sensitiv-
velopedlow
sensitivity, at the University
value of Michigan
baseline and [23]. These
parasitic capacitive pressure
capacitance, low transducers provide
temperature coefficients, and
ity, special
low value baseline
properties
and parasitic
not found in
capacitance,
commercially available
low temperature
transducers
coefficients,
such as high
and small
sensitiv- capacitive
small
volume.volume. The pressure range is extended through electrically insulated
ity, lowThevaluepressure
baseline andrange is extended
parasitic capacitance, through electrically
low temperature insulated
coefficients, capacitive elec-
and small
electrodes, which can operate in contact mode. Furthermore, the substrate under the
trodes,
volume.whichThe can operate
pressure range in contact mode.
is extended through Furthermore,
electrically insulatedthe capacitive
substrateelec- under the dia-
diaphragm
trodes, offers
which natural
can operateover-pressure
in contact protection.
mode. Furthermore, Figurethe 7b showsunder
substrate the assembled
the dia- flexible
phragm offers natural over-pressure protection. Figure 7b shows the assembled flexible
PCB inductor coil with one capacitive pressure transducer; an inset image of the pressure
PCBphragm
inductor offers natural over-pressure protection. Figure 7b shows the assembled flexible
coil with one capacitive pressure transducer; an inset image of the pressure
PCB inductor
transducer element coilis
with one
also capacitive
shown. Thepressure
responsetransducer;
of the an inset imagedue
transducer of thetopressure
applied pressure
transducer
transducer element
element is
is alsoshown.
also shown. TheThe response
response of the of the transducer
transducer due to due pressure
applied to applied pressure
is shown in Figure 7c. The transducer has very low baseline capacitance of ≈4 pF, along
is shown
is shown ininFigure
Figure 7c. Thetransducer
7c. The transducer has has
very very low baseline
low baseline capacitance capacitance of ≈4 pF, along
of ≈4 pF, along
with a ≈4.5 pF linear full-scale capacitive response over a 0–25 bar applied pressure range.
withwith
a ≈4.5 pFpFlinear
a ≈4.5 linear full-scale capacitive
full-scale capacitive response
response over a over
0–25 bara 0–25
appliedbarpressure
applied pressure range.
range.

Figure 7. (a) Cross-section of packaged LC sensor (b). Unpackaged LC sensor coil and inset image
(a) Cross-section
Figureof7.capacitive of packaged
pressure transducer. LC sensor.
(c) Response (b). Unpackaged
of capacitive LC sensor
pressure transducer. coil
(d). 3D and inset image
printed
bottom metal
of capacitive substrate
pressure stiffener (left)
transducer. (c)and top metalofcover
Response (right). I pressure
capacitive Fully packaged LC sensor.
transducer. (f) 3D printed
(d).
Readout PCB showing two readout nodes.
bottom metal substrate stiffener (left) and top metal cover (right). (e) Fully packaged LC sensor.
(f)Figure
Readout7. (a)
PCBCross-section
showing two ofreadout
packaged LC sensor (b). Unpackaged LC sensor coil and inset image
nodes.
of capacitive pressure transducer. (c) Response of capacitive pressure transducer. (d). 3D printed
bottom metal substrate stiffener (left) and top metal cover (right). I Fully packaged LC sensor. (f)
Readout PCB showing two readout nodes.
Sensors 2023, 23, x FOR PEER REVIEW 11 of 21

Sensors 2023, 23, 2525 11 of 21

3.2. LC Sensor Packaging

The LC sensor packaging must be sufficiently robust to provide protection against
3.2. LC Sensor Packaging
mechanical forces and withstand environments inside core-flood experiment equipment
The LC
(containing sensor chemicals
corrosive packagingand must be sufficiently
temperatures up robust
to 125to provide
°C) while protection against
still permitting
mechanical forces and withstand environments inside
pressure transmission to the transducer. To limit any disturbance in core-flood experiment equipment
the fluid flow during
(containing
a core-flood corrosive chemicals
experiment, the packaged andsensor
temperatures up to 125 ◦within
may be integrated C) whilethe still permitting
thickness of
the polymer sheath encasing the rock core. For a typical polymer sheath thicknessduring
pressure transmission to the transducer. To limit any disturbance in the fluid flow of 5 a
core-flood experiment, the packaged sensor may be integrated within the thickness of the
mm, the packaged sensor should be less than 3 mm in thickness to enable integration
polymer sheath encasing the rock core. For a typical polymer sheath thickness of 5 mm,
without compromising the integrity of the polymer sheath. The package is composed of a
the packaged sensor should be less than 3 mm in thickness to enable integration without
top metal cover and bottom metal substrate stiffener, VitonTM rubber encapsulation, and
compromising the integrity of the polymer sheath. The package is composed of a top metal
thin film polyimide coating; a cross section is shownTM in Figure 7a.
cover and bottom metal substrate stiffener, Viton rubber encapsulation, and thin film
The top metal cover and bottom metal substrate stiffener (Figure 7d) were used to
polyimide coating; a cross section is shown in Figure 7a.
protect the capacitive pressure transducer, providing mechanical support to prevent any
The top metal cover and bottom metal substrate stiffener (Figure 7d) were used to
shear forces or static high pressure from causing delamination. They were fabricated by
protect the capacitive pressure transducer, providing mechanical support to prevent any
Fathom Advanced Manufacturing (Oakland, CA, USA) using the direct metal laser sin-
shear forces or static high pressure from causing delamination. They were fabricated
tering (DMLS) 3D printing process. An AlSi10Mg aluminum alloy [40] was used to both
by Fathom Advanced Manufacturing (Oakland, CA, USA) using the direct metal laser
minimize size and reduce impact on the LC sensor coupling to the readout coil while still
sintering (DMLS) 3D printing process. An AlSi10 Mg aluminum alloy [40] was used to both
providing
minimize maximum
size and support and protection.
reduce impact on the LCThe topcoupling
sensor of the toptometal cover and
the readout coil bottom
while still
of the bottommaximum
providing metal substrate
support stiffener were curved
and protection. The topto match
of the the
top curvature
metal cover radius of the of
and bottom
rock core. To secure them to the LC sensor, a high temperature flexible
the bottom metal substrate stiffener were curved to match the curvature radius of the rock epoxy (Duralco
4538, Brooklyn,
core. To secure NY, USA)
them to was
the LCemployed.
sensor, aThis
highflexible epoxy also
temperature surrounded
flexible the trans-
epoxy (Duralco 4538,
ducer and transducer diaphragms, providing additional protection
Brooklyn, NY, USA) was employed. This flexible epoxy also surrounded the transducer while still permitting
pressure to be transmitted
and transducer diaphragms,to the diaphragms.
providing additional protection while still permitting pressure
toTobe determine
transmitted the
tonecessary thicknesses of the top metal cover and bottom metal sub-
the diaphragms.
strate stiffener, an FEAthe
To determine model was developed
necessary in COMSOL
thicknesses of the top Multiphysics
metal cover. and
® This model was
bottom metal
used to evaluate
substrate the an
stiffener, stress
FEAdistribution
model was and deflection
developed of the AlSi
in COMSOL 10Mg aluminum
Multiphysics ® . Thisalloy
model
under high pressure. The simulations showed that the maximum von
was used to evaluate the stress distribution and deflection of the AlSi10 Mg aluminum Mises stress in a 1.2
alloy
mm thickhigh
under AlSi10pressure.
Mg aluminum alloy top metal
The simulations cover
showed was
that the90maximum
MPa at 230von barMises
applied pres-in a
stress
sure
1.2(Figure
mm thick 8a),AlSi
permitting a safety factor of >3x for the 300 MPa yield strength of the
10 Mg aluminum alloy top metal cover was 90 MPa at 230 bar applied
AlSi 10Mg aluminum
pressure (Figure 8a),alloy. The maximum
permitting deflection
a safety factor of >3of × the top300
for the metal
MPa cover
yieldwas ≈3.6 μm
strength of the
at 230
AlSibar
10 Mgapplied
aluminum pressure,
alloy. indicating
The maximum that the designed
deflection of clearance
the top of
metal 200
cover μm was was
≈ suffi-
3.6 µm at
cient.
230 bar applied pressure, indicating that the designed clearance of 200 µm was sufficient.

(a) (b)
Figure 8. (a)
Figure FEA
8. (a) modelling
FEA of von
modelling Mises
of von stress
Mises distribution
stress in LC
distribution sensor
in LC packaging
sensor under
packaging applied
under applied
pressure of 230 bar. (b) Experimental Re{Zin} datasets for LC sensors in unpackaged and packaged
pressure of 230 bar. (b) Experimental Re{Zin } datasets for LC sensors in unpackaged and packaged
configurations. Assumed material properties for FEA given in Table 4.
configurations. Assumed material properties for FEA given in Table 4.

To To
confirm thethe
confirm strength of the
strength metal
of the cover,
metal a sample
cover, a sample of of
thethe
AlSi 10Mg top metal cover
AlSi10 Mg top metal cover
andand
bottom metal stiffener were mounted to the L S coil using Duralco 4538; the dummy
bottom metal stiffener were mounted to the L coil using Duralco 4538; the dummy LC
S
Sensors 2023, 23, 2525 12 of 21

sensor was pressurized to 250 bar and heated to 125 ◦ C for 120 min. At the conclusion of this
test, no visible deformation of the top metal cover and bottom metal stiffener was observed.
For this work, it is important that the top metal cover and bottom metal substrate
stiffener are fabricated from materials with relative magnetic permeability (µr ) as low as
possible. Materials with high µr placed near the inductor will act as a ferrite core; at high
frequencies, this will greatly increase the effective value of RS,Coil , reducing Q and <in ,
worsening pressure resolution. For comparison two additional sets of results are presented
here: (i) the results of an LC sensor without top metal cover and bottom metal substrate
stiffener installed (unprotected), and (ii) an LC sensor protected with top metal covers
and bottom metal substrate stiffeners fabricated from 15-5 stainless steel (µr ≈ 95), and
316 L stainless steel (µr ≈ 1.4). These are compared to the results from the metal cover of
AlSi10 Mg aluminum alloy (µr ≈ 1.0) are shown in Figure 8b. The 15-5 stainless steel results
in a Q and <in reduction of >50% compared to an unprotected LC sensor; however, the
AlSi10 Mg aluminum alloy results in a reduction of <10%. These reductions in both Q and
<in combine to reduce pressure resolution significantly. Given that pressure resolution is
indirectly proportional to both Q and <in , (14), a reduction in these values by >50% worsens
pressure resolution by >4×; however, a reduction of <10% worsens resolution by <1.2×.
After the top metal cover and bottom metal substrate stiffener were secured, the LC
sensors were encapsulated in VitonTM rubber, which is known to tolerate elevated tem-
perature and pressure [41]. This was performed by using a custom mold and Fluorodyn
VitonTM Caulk from Thermodyn Global Sealing (Houston, TX, USA). As VitonTM is incom-
patible with certain solvents used in core-flood experiments, such as toluene, a 100 µm thin
film polyimide coating (PI 2610, HD MicroSystems, Wilmington, DE, USA) was applied to
the VitonTM surface. A fully packaged LC sensor is shown in Figure 7e.

3.3. Readout Circuit Fabrication

The readout PCB, containing four readout coils and associated readout nodes, was
fabricated on a flexible polyimide PCB substrate manufactured by FlexPCB (Santa Ana,
CA, USA). The readout PCB also included a 5-pin connector for power supply and commu-
nication with the external microcomputer. A segment of the fabricated PCB showing the
connector and two readout nodes is shown in Figure 7f.

4. Test Results
A number of experimental tests were performed to assess the wireless readout and
flow gradient measurement capability of the implemented PGM system. A test setup was
constructed to enable in-house PGM introducing controlled flow-induced pressure differen-
tials to LC sensors; PGM results were recorded from when the system was integrated into
the setup. Operation of the PGM system over the full-scale temperature (25–125 ◦ C) and
pressure (0 bar–20 bar) range was verified. Key factors that determine system performance
were investigated; these include pressure and temperature calibration of LC sensors, digital
improvements through oversampling, and accounting for variation in system integration
for use in real world applications.

4.1. LC Sensor Readout System

As described in Section 2, LC sensors were interrogated with a coupled readout coil
and associated readout node. Ideally, each LC sensor is axially aligned to a readout coil
with an IG of 4 mm. On receiving a command from the user through the graphical user
interface (GUI), the R-Pi external microcomputer serially communicates over an I2 C bus
with the MCU in each readout node to trigger a measurement cycle. During a measurement
cycle, the MCU commands the DDS to supply a sinusoidal excitation signal to the standing
wave ratio bridge circuit between the IFR of 11–15 MHz with an fss of 35 kHz. The dwell
time at each discrete frequency step is 100 ms, during which the ADC samples the four
voltages of the standing wave ratio bridge circuit, collecting and averaging 250 samples to
reduce uncorrelated white noise. These four voltages are then used to determine Re{Zin } at
Sensors 2023, 23, 2525 13 of 21

each discrete interrogation frequency. After completion of each measurement cycle, Re{Zin }
values stored in the MCU Flash memory are transferred to the R-Pi, where a Python-based
program extracts f 0 by applying a Gaussian fit, FZin , to the Re{Zin } dataset. The total time
required by the PGM system to trigger four readout nodes into a measurement cycle,
interrogate each LC sensor over the IFR, transfer the Re{Zin } dataset from the MCU at each
node to the R-Pi, and extract f 0 by fitting FZin , was 25 s. The ability of the PGM readout
system to accurately extract LC sensor f 0 was characterized using twelve LC sensor loads
with f 0 values between 3 and 12 MHz as measured using a benchtop network analyzer
(Keysight E5061B). These loads were constructed from planar inductor coils fabricated
on standard PCB manufacturing methods and surface mounted ceramic capacitors. The
f 0 values of these loads as measured by the PGM readout system were within 6% of the
benchtop network analyzer measurements.

4.2. Dynamic Pressure Response and Flow Resolution

To apply a pressure gradient across multiple LC sensors through fluid flow and record
PGM, a custom test setup was constructed (Figure 9a). The setup consisted of a peristaltic
pump to generate constant fluid (mineral oil) flow, individual pressure chambers to house
LC sensors, LCn , (IG = 4 mm, AM = 0 mm), and check valves, CVn , to create known pressure
differential, ∆PCV , between each chamber; commercial pressure gauges, PGn , [42]
Sensors 2023, 23, x FOR PEER REVIEW 14 were
of 21
used to monitor pressure in each LC sensor chamber. Each PGn acts as a high-resolution
pressure reference at each sensing node with a limited operating pressure range of 500 mbar.
The analog output of each wired PGn was digitized using a commercial data acquisition
board (USB-6363-OEM DAQ, National Instruments128𝜇 𝐿𝛷 Corp., USA) controlled in LabVIEW™
∆𝑃 = (19)
(National Instruments Corp., Austin, TX, USA). 𝜋𝐷All fluidic components in the test setup
were interconnected using fluidic connectors and polyurethane tubing with an 1/8” I.D.
and 3/16” O.D., manufactured by ∆𝑃𝑆
ATP = ∆𝑃 + ∆𝑃(Milford
Pneumatics + ∆𝑃 Center, OH, USA). (20)

(a) (b)
Figure 9. (a) Flow test setup illustrating LC sensors (LCn) and wired pressure gauges (PG) in flow
Figure 9. (a) Flow test setup illustrating LC sensors (LCn ) and wired pressure gauges (PG) in
path with wirelessly coupled Readout nodes (RNn). Check valves (CVn) create pressure drop. (b)
flow path with wirelessly coupled Readout nodes (RN ). Check valves (CVn ) create pressure drop.
Measured pressure of LC sensors (LCn) and flow rate nduring flow experiment at 25 °C (left), and
(b) Measured pressure of LC gauges
sensors (PG
(LCnn))during
and flow rateexperiment
during flow experiment ◦ C (left), and
at 25 for
measured pressure of wired flow (right). Error bars measure-
measured
ment pressure
resolution notofvisible
wiredatgauges (PGn )confidence
scale (95% during flowinterval:
experiment
<1.5(right).
mbar). Error bars for measurement
resolution not visible at scale (95% confidence interval: <1.5 mbar).
In a typical PGM experiment, LC sensors were placed into each pressure chamber
Duringaligned
and axially a typical pressure
(axial drop calibration
misalignment, of the
AM = 0 mm) flow testcoils
to readout setup,
on when a flowPCB
the Readout rate
of ≈ 32 mL/min was applied, CV and CV provided a ∆P
with an interrogation gap, IG of 41 mm; AM3 and IG are further of 60 ± 5 mbar, while CV
CV discussed in Section 4.5.2
provided a ∆PCV of 35 ± 5 mbar. The pressure drop across each check valve (∆PCV ) was
The test setup was then saturated with mineral oil by initiating fluid flow until steady
estimated using Equation (18),
state flow was established; fluid flow was then ceased and measurements were taken (i.e.,
pressure readings during a flow ∆Prate of 0 mL/min). The fluid flow was restarted and(18) in-
CV = ( PG1 − PG2 ) − ∆PC
creased in three discrete steps of 14 mL/min, 24 mL/min, and 32 mL/min; the flow rate
was held constant at each flow step for 225 min. Figure 9b left plot shows a time-series of
pressure measurements from four LC sensors (shown with colored diamond markers);
440 samples were averaged per measurement. Fluid flow rate during experiment is plot-
ted with red line. In the presence of flow, LC1, being nearest to the pump, recorded the
highest pressure head (inlet pressure); as the flow encounters resistance at each check
Sensors 2023, 23, 2525 14 of 21

where PG1 and PG2 are the commercial pressure gauge readouts at the two ends of the
check valve and ∆PC is the pressure drop in the associated fluidic connectors, estimated
to be ≈3 mbar at 25 mL/min fluid infusion rate and ≈1 mbar at 5 mL/min infusion rate.
Uncertainty in ∆PCV arises from resolution of commercial gauges, varying pressure head
due to consumption of oil in the upstream reservoir, and pressure gauge placement on the
opposite side of the pressure chamber relative to the LC sensor. Furthermore, CV’s have a
variation in pressure drop of up to 10% [43].
The pressure drop in the polyurethane tube of the test setup, for white mineral oil flow,
was estimated using the simplified Hagen-Poiseuille’s equation [44] shown in Equation
(19), where ∆PT is the pressure drop in the tube, µoil is the dynamic viscosity of white
mineral oil, L is the tube length, Φ is the flow rate, and D is the tube inner diameter. For a
tube length of 2.54 cm, tube inner diameter of 3.175 mm, mineral oil density of 838 Kg/m3 ,
dynamic viscosity of 0.0103 N-s/m2 and flow rate of 25 mL/min, the calculated pressure
drop is 0.437 mbar. Using the same assumptions but substituting in a 5 mL/min flow rate,
the pressure drop is 0.0873 mbar. Hence, the pressure gradient between two adjacent LC
sensor channels x and y (∆PSxy ) shown in Equation (20), is the summation of ∆PCV , ∆PC ,
and ∆PT .
128µoil LΦ
∆PT = (19)
πD4
∆PSxy = ∆PCV + ∆PC + ∆PT (20)
In a typical PGM experiment, LC sensors were placed into each pressure chamber and
axially aligned (axial misalignment, AM = 0 mm) to readout coils on the Readout PCB with
an interrogation gap, IG of 4 mm; AM and IG are further discussed in Section 4.5. The test
setup was then saturated with mineral oil by initiating fluid flow until steady state flow
was established; fluid flow was then ceased and measurements were taken (i.e., pressure
readings during a flow rate of 0 mL/min). The fluid flow was restarted and increased in
three discrete steps of 14 mL/min, 24 mL/min, and 32 mL/min; the flow rate was held
constant at each flow step for 225 min. Figure 9b left plot shows a time-series of pressure
measurements from four LC sensors (shown with colored diamond markers); 440 samples
were averaged per measurement. Fluid flow rate during experiment is plotted with red
line. In the presence of flow, LC1 , being nearest to the pump, recorded the highest pressure
head (inlet pressure); as the flow encounters resistance at each check valve, the pressure
head drops sequentially from LC2 to LC3 to LC4 . Recorded pressure measurements from
the commercial pressure gauges (PGn ) are shown to the right in Figure 9b. The LC sensor
measurements are in close agreement with the PG measurements, differing by <5% on
average. This difference in pressure between LC sensor and commercial PG measurements
can be accounted for since the LC sensor and corresponding PG are not exactly collocated;
fluctuations in ambient pressure in the room and flow resistance provided by the tube and
connectors between the PG and LC sensor can then result in minor differences between the
two pressure levels at the LC sensor and PG.
Although the demonstrated PGM system uses a passive wireless interrogation ap-
proach, low power computing, and compact readout electronics on a flexible PCB substrate,
the system can successfully decipher very low PGM (e.g., ≈10 mbar between LC2 and
LC3 at 14 mL/min flow rate). These results demonstrate the ability of the PGM system
to measure static pressure with a resolution of up to 1.0 mbar when oversampling and
averaging is utilized and detect pressure gradients due to fluid flow in a dynamic fluid
flow environment. The following subsections illustrate the methodology used to achieve
the high pressure measurement resolution which enables PGM using passive LC sensors.

4.3. System Static Pressure and Thermal Response

To alleviate the impact of variation in parasitic capacitance, planar coil inductance, and
capacitance response of each pressure transducer, along with variation originating from
LC sensor integration and packaging steps, each packaged LC sensor was individually
pressure and temperature calibrated to ensure a high pressure measurement resolution. To
 ally pressure and temperature calibrated to ensure a high pressure measurement resolu-
tion. To perform this calibration operation over the full-scale temperature (25 °C–125 °C)
and pressure (0 bar–20 bar) range, LC sensors were axially aligned to each readout coil on
the PCB with an IG of 4 mm. The PGM system with four LC sensors was placed inside a
chamber and pressurized using nitrogen gas. Pressure was raised to 20,700 mbar (300 psi)
Sensors 2023, 23, 2525 15 of 21
at 25 °C in steps of 3447 mbar (50 psi). Readings from the four LC sensors were taken at
each pressure step and are shown in Figure 10; LC sensors showed an average response
of approximately 38 Hz/mbar from 0 bar to 20.7 bar.
perform this
The calibration
PGM systemoperation
was thenover the into
placed full-scale temperature
an oven (25–125 ◦pressure)
(at atmospheric C) and pressure
and the
(0temperature
bar–20 bar) range, LC sensors were axially aligned to each readout
ramped from 25 °C to 125 °C while readings were continually coil on the PCB with
taken. The
IG of 4remained
ansystem mm. Thefully
PGMfunctional
system with four LC sensors was placed inside a chamber
with no change in performance. f0 shift due to tempera- and
pressurized using nitrogen gas. Pressure was raised to 20,700 mbar (300 psi) at 25 ◦ C in
ture (where a shift of “0” refers to f0 at 0 applied pressure and ambient temperature of 25
steps of 3447 mbar (50 psi). Readings from the four LC sensors were taken at each pressure
°C) for four LC sensors (LC1–LC4) over the full-scale temperature range is shown in Figure
step and are shown in Figure 10; LC sensors showed an average response of approximately
10. All LC sensors showed a linear shift in f0 of ≈2.7 kHz/ °C from 25 °C–100 °C, reducing
38 Hz/mbar from 0 bar to 20.7 bar.
to a shift of approximately 0.0 kHz/ °C from 100 °C to 125 °C.

Figure 10. This figure shows the shift in the resonant frequency, f0, resulting from applied pressure
Figure 10. This figure shows the shift in the resonant frequency, f 0 , resulting from applied pressure
(left) and the applied temperature (right). Error bars not visible at scale (95% confidence interval:
(left) and the applied temperature (right). Error bars not visible at scale (95% confidence interval:
±0.041 kHz).
±0.041 kHz).
4.4. Conversion of Resonant Frequency to Pressure
The PGM system was then placed into an oven (at atmospheric pressure) and the
In order
temperature ramped fromf025to◦ applied
to convert C to 125 pressure
◦ C whilewhile alsowere
readings accounting
continually for changes in ambi-
taken. The sys-
entremained
tem temperature fullywhich also affect
functional with no f0, multi-dimensional
change in performance. regression
f 0 shift(i.e.,
duehigher order pol-
to temperature
ynomial
(where curve
a shift of fitting)
“0” refers[45]towas
f 0 atutilized.
0 applied This approach
pressure andinvolves
ambientfitting
temperature of 25 ◦ C)
a 3-D polynomial
surface to the LC sensor response, where one dimension is
for four LC sensors (LC1 –LC4 ) over the full-scale temperature range is shown in Figure applied pressure, one dimen-
10.
sion is ambient temperature, and the third dimension ◦
is extracted
All LC sensors showed a linear shift in f 0 of ≈2.7 kHz/ C from 25–100 C, reducingmulti- f0 . A ◦ third order to a
dimensional
shift polynomial
of approximately was fit
0.0 kHz/ ◦ Ctofrom 100 ◦ Cpressure
the static to 125 ◦ C.
response as measured over the full-
scale pressure range of 0–20.7 bar and temperature range of 25 °C to 125 °C. The calibra-
4.4. Conversion
tion equation of forResonant Frequency
conversion to Pressure
of extracted f0 and temperature, T, to applied pressure, PApp,
is shown in Equation (21), where a
In order to convert f 0 to applied pressure 1 –a 6 are fittedwhile
coefficients using leastfor
also accounting squares
changesregression
in am-
to static
bient full-scale which
temperature pressurealsoand temperature
affect test data.
f 0 , multi-dimensional regression (i.e., higher order
polynomial curve fitting) [45] was utilized. This approach involves fitting a 3-D polyno-
𝑃 = 𝑎 (𝑓 ) + 𝑎 (𝑓 ) + 𝑎 (𝑓 ) + 𝑎 + 𝑎 (𝑓 ∙ 𝑇) + 𝑎 (𝑇) (21)
mial surface to the LC sensor response, where one dimension is applied pressure, one
dimensionThe calibration
is ambient equation
temperature, was andfound thetothird
convert f0 and measured
dimension is extracted temperature
f 0 . A thirdwith
orderan
multi-dimensional
accuracy of <±0.25% polynomial
of full-scalewas(<±50
fit tombar)
the static
overpressure
the designedresponse as measured
full-scale pressureover
and
the full-scale pressure range of 0–20.7 bar and temperature range of 25 ◦ C to 125 ◦ C. The
temperature range (0–21,000 mbar and 25 °C–125 °C, respectively). However, it is im-
calibration
portant toequation
note thatforthe conversion
accuracy of with which fthe
extracted 0 and
testtemperature,
apparatus could T, to applied pressure,to
apply pressure
PApp
the, PGM
is shownsystem in was
Equation (<±50 ambar
(21), where
comparable 1 –a6 are fitted coefficients
or <0.25% of full-scale).using least squares
regression to static full-scale pressure and temperature test data.

PApp = a1 ( f 0 )3 + a2 ( f 0 )2 + a3 ( f 0 ) + a4 + a5 ( f 0 · T ) + a6 ( T ) (21)

The calibration equation was found to convert f 0 and measured temperature with an
accuracy of <±0.25% of full-scale (<±50 mbar) over the designed full-scale pressure and
temperature range (0–21,000 mbar and 25–125 ◦ C, respectively). However, it is important
to note that the accuracy with which the test apparatus could apply pressure to the PGM
system was comparable (<±50 mbar or <0.25% of full-scale).
Sensors 2023,
Sensors 23,23,
2023, 2525
x FOR PEER REVIEW 16
16 ofof2121

4.5.
4.5.Pressure
PressureResolution
ResolutionEnhancement
Enhancement
With
Withthetheuse
useofof
additional
additional digital
digitalsignal processing
signal processing through
through oversampling
oversampling andand
averag-
aver-
ing, in-band
aging, noisenoise
in-band power can be
power canlowered, thereby
be lowered, improving
thereby effective
improving pressure
effective resolution
pressure of
resolu-
LC sensors.
tion In the target
of LC sensors. application,
In the the signalthe
target application, tests can tests
signal be performed for months,
can be performed forand the
months,
signal
and thechanges
signalcan occurcan
changes in geological time scales,
occur in geological presenting
time ample opportunity
scales, presenting for this
ample opportunity
approach. Each quadrupling of the number of averaged samples
for this approach. Each quadrupling of the number of averaged samples lowers in-band lowers in-band noise by
6 noise
dB, improving pressure resolution by a factor of two [46]. Figure 11
by 6 dB, improving pressure resolution by a factor of two [46]. Figure 11 shows theshows the resulting
instantaneous ResolMeas ofResol
resulting instantaneous 46 mbar permitted by the PGM system along with improvement
Meas of 46 mbar permitted by the PGM system along with im-
inprovement
pressure resolution
in pressure when oversampling
resolution and averagingand
when oversampling were used; when
averaging were320 samples
used; when
were averaged, resolution improved to 1.0 mbar. As seen in the plot,
320 samples were averaged, resolution improved to 1.0 mbar. As seen in the plot, the im- the improvement in
pressure resolution with number of averaged samples is rapid to begin
provement in pressure resolution with number of averaged samples is rapid to begin with with but eventually
saturates to a point
but eventually of diminishing
saturates to a pointreturns; the estimated
of diminishing returns;pressure resolution
the estimated also being
pressure resolu-
dependent on the chosen confidence interval of measurement. In
tion also being dependent on the chosen confidence interval of measurement. In core- core-flood experiments,
which
floodcan be several months
experiments, which can in duration,
be severalhigh bandwidth
months measurements
in duration, are generally
high bandwidth not
measure-
necessary, permitting a large number of samples to be averaged for resolution
ments are generally not necessary, permitting a large number of samples to be averaged improvement
without impacting system efficacy.
for resolution improvement without impacting system efficacy.

Figure11.
Figure 11.Measured
Measuredpressure
pressureresolution
resolution(Resol
(ResolMeas))of
ofPGM
PGMsystem.
system.
Meas

4.6.
4.6.System
SystemDeployment
DeploymentVariations
Variations
InIndeploying
deployingthe thePGMPGMsystem
systeminina acore-flood
core-floodexperiment
experimentsetup, setup,ororininother
otherreal
realworld
world
PGM
PGMapplications,
applications,ititmaymaybebemore morechallenging
challengingtotoachieveachieveperfect
perfectaxialaxialalignment
alignmentbetween
between
the
theLC LCsensor
sensor and
andits its
corresponding
corresponding flexible readout
flexible readoutcoil.coil.
In addition,
In addition,different applications
different applica-
may
tionsnecessitate different
may necessitate interrogation
different gap, IGgap,
interrogation requirement between
IG requirement each LC
between eachsensor and
LC sensor
readout coil pair.
and readout coil To
pair.understand
To understand such variations
such variations in deployment
in deployment requirement,
requirement, the impact
the im-
ofpact
axialofmisalignment,
axial misalignment, AM, (Figure
AM, (Figure 12a) of upof
12a) toup 4 mm
to 4 was
mm also
was alsostudied as a as
studied function of
a function
IG.
of In
IG.particular, IG values
In particular, of 3 mm,
IG values of 3 mm, 4 mm, and 5and
4 mm, mm5were
mm wereexperimentally
experimentallyevaluated and
evaluated
are
andplotted in Figure
are plotted 12b using
in Figure 12b AMusing of AM
0 mm of and
0 mm IG and
of 4 IG
mmofas4 amm reference. It is important
as a reference. It is im-
toportant
note that the plot depicts absolute values of pressure resolution
to note that the plot depicts absolute values of pressure resolution relative relative to the 0 to mm the
AM, 4 mm IG measurement, i.e., a lower value in the plot translates
0 mm AM, 4 mm IG measurement, i.e., a lower value in the plot translates to a superior to a superior pressure
resolution. The resultsThe
pressure resolution. show that the
results showpresence
that theof both IG and
presence of AM
bothhave a significant
IG and AM havenegative
a signifi-
impact
cant negative impact on pressure resolution. In the absence of misalignment (withthe
on pressure resolution. In the absence of misalignment (with AM = 0 mm), as AM IG =
was changed 4 mm to 3 mm, the improvement in pressure resolution
0 mm), as the IG was changed 4 mm to 3 mm, the improvement in pressure resolution was was ≈ 25%, and as
IG≈25%,
was changed 4 mm to 5 mm, the loss in pressure resolution
and as IG was changed 4 mm to 5 mm, the loss in pressure resolution was ≈145%.was ≈ 145%. For a constant
interrogation
For a constant gap (with IG = gap
interrogation 4 mm), (withasIGthe= AM
4 mm), wasaschanged
the AM was 0 mm to 2 mm,
changed 0 mm thetoloss in
2 mm,
pressure resolution was ≈ 10%, and as IG was changed 0 mm to
the loss in pressure resolution was ≈10%, and as IG was changed 0 mm to 4 mm, the loss 4 mm, the loss in pressure
resolution
in pressure was ≈46%. These
resolution was ≈46%.results highlight
These resultsthe significance
highlight of considering
the significance AM and
of considering
IG during LC sensor design process to improve performance with changing deployment
AM and IG during LC sensor design process to improve performance with changing de-
requirements. For instance, the optimal LC sensor and readout coil design (i.e., the design
ployment requirements. For instance, the optimal LC sensor and readout coil design (i.e.,
which maximizes FOM) for an IG of 4 mm is not necessarily the same for an IG of 5 mm.
the design which maximizes FOM) for an IG of 4 mm is not necessarily the same for an IG
The results also show across various values of IG, the loss in pressure resolution for an
of 5 mm. The results also show across various values of IG, the loss in pressure resolution
 Sensors
Sensors 2023,
2023, 23,23, x FOR PEER REVIEW
2525 17ofof2121
17

AMforup
anto
AM up to
2 mm is 2low
mm is low thus
(<10%), (<10%), thus allowing
allowing some flexibility
some flexibility duringassembly
during system system assem-
in a
bly in a core-flood experiment
core-flood experiment setup. setup.

(a) (b)
Figure 12. (a) Illustration of axial misalignment (AM) and interrogation gap (IG). (b) Measured
Figure 12. (a) Illustration of axial misalignment (AM) and interrogation gap (IG). (b) Measured
pressure resolution with non-ideal AM and IG.
pressure resolution with non-ideal AM and IG.

5.5.Discussion
Discussion
InInthis
thiswork,
work,the the combined
combined interrogation time time for forallallfour
fournodes
nodes was was limited
limited by bytwo
two primary
primary factors:
factors: (1) the
(1) the settling
settling timetime
of theof the
outputoutput response,
response, and and (2) available
(2) the the available MCU
MCUFlashFlash
storagestorage
(<2 kB)(<2ofkB) theofdigitized
the digitized |Vx | data.
|Vx| value valueThe data. The settling
settling time was timethewas theof
result
result of time necessary at each interrogation frequency for the
time necessary at each interrogation frequency for the DDS frequency output and stand- DDS frequency output and
standing
ing wave waveratioratio circuit
circuit response
response to beto stable
be stable before
before digitization
digitization (≈50(≈ms).
50 ms).ThisThis settling
settling time
time caused
caused a delay
a delay at each
at each interrogated
interrogated frequency;
frequency; therefore,
therefore, the total
the total timetime required
required to
to inter-
interrogate
rogate all 112 all 112 frequencies
frequencies acrossacross the IFR
the full IFR ≈6.5
full was was sec ≈6.5 s (including
(including an additional
an additional ≈5 ms
≈to5 ms to sample
sample the actual
the actual |Vx | values).
|Vx| values). The limited
The limited flash memory
flash memory storagestorage
of the MCU of therequired
MCU
required that the digitized |V | values be transferred from
that the digitized |Vx| values be transferred from the MCU to the external microcomputer
x the MCU to the external
microcomputer
after each frequency,after each frequency,
taking ≈4.5 s.taking ≈4.5 s. Furthermore,
Furthermore, the data transfer the data transfer
between between
the readout
the readout PCB and microcomputer was limited to a
PCB and microcomputer was limited to a single serial communication bus. Because single serial communication bus. a
Because a modular design approach was utilized for the readout
modular design approach was utilized for the readout circuit (i.e., each readout node con- circuit (i.e., each readout
node
tainedcontained its own complete
its own complete circuit), thecircuit), the data
data transfer transfer
speed speed was
was further further
limited, limited,
as each of the
asfoureachindividual
of the four individual MCUs was required to wait for
MCUs was required to wait for the preceding MCU to complete its data the preceding MCU to
complete its data transfer to the microcomputer before sending
transfer to the microcomputer before sending its own data, resulting in a total data trans- its own data, resulting
infera total
perioddata oftransfer
≈18.5 s period of ≈18.5
for all four MCUs. s forInallanfour MCUs.application
alternate In an alternate scenarioapplication
requiring
scenario requiring higher operational bandwidth (and where
higher operational bandwidth (and where the number of readout nodes is limited), the number of readout nodes fur-
is ther
limited), further refinement of the standing wave ratio bridge
refinement of the standing wave ratio bridge circuit could be performed to reduce circuit could be performed
tosettling
reduce settling time; a common
time; a common excitationexcitation and readout
and readout electronicselectronics
platform platform
with awith a single,
single, higher
higher performance MCU and DDS to control all readout nodes simultaneously may be
performance MCU and DDS to control all readout nodes simultaneously may be imple-
implemented, which could significantly reduce interrogation time. Use of a single, more
mented, which could significantly reduce interrogation time. Use of a single, more com-
computationally capable MCU would enable fitting algorithms to run real-time on the
putationally capable MCU would enable fitting algorithms to run real-time on the readout
readout PCB itself, significantly reducing the requirement on external processing and data
PCB itself, significantly reducing the requirement on external processing and data trans-
transfer. In terms of the present circuit, an external Flash memory at each readout node
fer. In terms of the present circuit, an external Flash memory at each readout node could
could also eliminate the need for repeated communication with the external unit at each
also eliminate the need for repeated communication with the external unit at each fre-
frequency step through local data storage, thus reducing overall interrogation time.
quency step through local data storage, thus reducing overall interrogation time.
The approach used to design the PGM system has notable features that can extend
The approach used to design the PGM system has notable features that can extend
its use to other applications. This work is scalable to a large number of sensors because
its use to other applications. This work is scalable to a large number of sensors because of
of the use of a modular readout PCB design. The system hardware and software were
the use of a modular readout PCB design. The system hardware and software were co-
co-designed to maximize pressure resolution. For instance, the LC sensor design which
designed to
maximizes maximize
pressure pressurewas
resolution resolution.
selectedFor instance,
while operatingthe LC sensorthe
within design which max-
capabilities of
imizes pressure resolution was selected while operating
the deployed curve fitting algorithm. The FOM presented in this work encompasses within the capabilities of the de-
allployed
major curve
designfitting algorithm.ofThe
dependencies the FOM presented
LC sensor and in this work
readout system encompasses
which may allbemajor
in
design dependencies of the LC sensor and readout system which
direct conflict with one another (e.g., interrogation gap, inductor coil diameter, capacitive may be in direct conflict
with one
pressure anotherresponse,
transducer (e.g., interrogation
etc.), in contextgap,ofinductor coil diameter,
unique challenges posedcapacitive pressure
by the application
transducer response, etc.), in context of unique challenges
(robust packaging, flexible PCB substrate, custom pressure transducers with unique ∆C/C posed by the application
0
Sensors 2023, 23, 2525 18 of 21

response, misalignment tolerance, interrogation gap, etc.). The same design refinement
model for LC sensors can be extended across different PGM applications.

6. Conclusions and Summary

This work has focused on the investigation and realization of a passive wireless
pressure sensing system using inductive-capacitive (LC) transduction to measure ultra-
small flow-induced pressure gradients (≈few mbar) at small flow rates (10–32 mL/min) in
high resistance fluidic channels such as a rock core, with a pressure measurement resolution
of <1 mbar, while being subjected to a large lateral pressure up to 250 bar and temperature
up to 125 ◦ C. The design of the LC sensor and packaging approach, sensor interrogation
and readout electronics, data analysis and software enhancement methods, and external
interface for user control were presented with application in core-flood experiments. The
sensing system demonstrated PGM capability during fluid flow with a ≈10 mbar successful
PGM between adjacent LC sensors caused by a 14 mL/min flow rate.
While application driven requirements of a high pressure measurement resolution,
which improves PGM and flow analysis, is typically addressed through specific improve-
ments in sensor element and/or interface electronics, this work addressed the same problem
through understanding relationships between all system components (including sensor
design, sensor packaging, interface electronics, processing capability, firmware, etc.) and
their relative impact on overall system performance. The resulting co-design of hardware
and software elements lead to a more cost-effective solution. An LC sensor design re-
finement model for analytically predicting and minimizing pressure resolution, which
accounted for both sensor non-idealities and secondary packaging effects, was investigated
and experimentally validated. The LC sensor was comprised of a package using 3D printed
aluminum caps to protect the capacitive transducer, encased in VitonTM and thin film
polyimide to provide chemical resistance to the environment and still permit pressure
transduction to the transducer, and measured <ø15 × 3 mm3 .
The LC sensor interrogation system was comprised of a readout PCB and external unit
containing a microcomputer and remote laptop with custom GUI for user control. Inter-
rogation electronics extracted and digitized the real part of the input impedance, Re{Zin },
of an inductor coil inductively coupled to the LC sensor over a set of discrete frequencies.
This discrete dataset was then transferred to the external microcomputer which extracted
the pressure-dependent resonant frequency, f 0 , via interpolation by fitting a continuous
Gaussian curve. A custom GUI running on a remotely connected laptop permitted real-time
user control of the system. A modular system design approach allows easy customization
of number of LC sensors used to fit varying application needs while use of commercial
manufacturing practices promote volume scalability of the proposed solution.
The system was characterized over a full-scale pressure range of 0–21,000 mbar and
temperature range of 25–125 ◦ C. Measurements of flow-induced pressure gradients were
successfully taken by integrating the full system into a custom test setup. A pressure
resolution of <1 mbar was demonstrated when oversampling and averaging software
enhancement techniques were utilized.
Applications in which measurement of flow-induced pressure gradients were pre-
viously thought impossible or impractical are now within the realm of possibility. Per-
formance may be improved by utilizing a higher performance circuitry, such as an MCU
with additional memory or an ADC with a higher bit resolution and faster data capture
rate. While this work focused on core-flood experiments, the LC sensor design refinement
techniques, readout circuit approach, f 0 extraction procedure, and software enhancement
methods can easily be used to create passive wireless flow gradient sensing systems to
meet the requirements of many different applications such as in high performance liquid
chromatography and unmanned navigation of underwater vehicles.
Sensors 2023, 23, 2525 19 of 21

7. Patents
Y. Gianchandani, T. Li, P. Dutta, A. Benken, J.-R. Ordonez-Varela, “Distributed Pressure
Measurement System for Core Flood Experiments”, disclosure filed 2 June 2019, UM file no.
2019-452, provisional patent 62/910,828 filed 4 October 2019, PCT/US2020/053970 filed
2 October 2020.

Author Contributions: Conceptualization, Y.B.G., A.C.B., P.P.D., T.L. and J.R.O.-V.; software, P.P.D.;
validation, P.P.D., A.C.B. and J.R.O.-V.; investigation, A.C.B., P.P.D. and T.L.; resources, Y.B.G. and
J.R.O.-V.; data curation, A.C.B.; writing—original draft preparation, P.P.D. and A.B; writing—review
and editing, Y.B.G. and A.C.B.; supervision, Y.B.G., T.L. and J.R.O.-V.; project administration, Y.B.G.
and J.R.O.-V.; funding acquisition, Y.B.G. and J.R.O.-V.; All authors have read and agreed to the
published version of the manuscript.
Funding: This work was supported in part by a research contract from Total Energies, Pau, France,
to the University of Michigan, Ann Arbor, MI, USA.
Institutional Review Board Statement: Not Applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author. The data are not publicly available.
Acknowledgments: The authors thank T. Ben-Ghzaiel and A. Chenieux, at Total Energies, for aid in
facilitating project development. The authors thank V.K. Gondrala and the research team at University
of Cincinnati for assistance in development of the PGM system readout nodes. Facilities used for this
work include the Lurie Nanofabrication Facility (LNF) at the University of Michigan, Ann Arbor.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Park, J.S.; Gianchandani, Y. A low cost batch-sealed capacitive pressure sensor. In Proceedings of the IEEE/ASME 12th
International Conference on Micro Electro Mechanical Systems (MEMS 99), Orlando, FL, USA, 17–21 January 1999; pp. 82–87.
2. Grimes, C.A.; Jain, M.K.; Singh, R.S.; Cai, Q.; Mason, A.; Takahata, K.; Gianchandani, Y. Magnetoelastic microsensors for
environmental monitoring. In Proceedings of the IEEE/ASME 14th International Conference on Micro Electro Mechanical
Systems (MEMS 01), Interlaken, Switzerland, 21–25 January 2001; pp. 278–281.
3. Takahata, K.; DeHennis, A.; Wise, K.; Gianchandani, Y. A wireless microsensor for monitoring flow and pressure in a blood
vessel utilizing a dual-inductor antenna stent and two pressure sensors. In Proceedings of the IEEE/ASME 17th International
Conference on Micro Electro Mechanical Systems (MEMS 04), Maastricht, The Netherlands, 25–29 January 2004; pp. 216–219.
4. Chapman, D.; Trybula, W. Meeting the Challenges of Oilfield Exploration using Intelligent Micro and Nano-Scale Sensors. In
Proceedings of the IEEE 12th International Conference on Nanotechnology, Birmingham, UK, 20–23 August 2012; pp. 20–23.
5. Fleming, W. Overview of automotive sensors. IEEE Sens. J. 2001, 1, 296–308. [CrossRef]
6. Wu, C.H.; Zorman, C.; Mehregany, M. Fabrication and testing of bulk micromachined silicon carbide piezoresistive pressure
sensors for high temperature applications. IEEE Sens. J. 2006, 6, 316–324.
7. Chavan, A.V.; Wise, K.D. Batch-processed vacuum-sealed capacitive pressure sensors. J. Microelectromech. Syst. 2001, 10, 580–588.
[CrossRef]
8. Gianchandani, Y.; Wilson, C.; Park, J.-S. Micromachined pressure sensors: Devices, interface circuits, and performance limits. In
The MEMS Handbook; Gad-el-Hak, M., Ed.; CRC Press: Boca Raton, FL, USA, 2006; pp. 1–44.
9. Benken, A.; Gianchandani, Y. Passive Wireless Pressure Sensing for Gastric Manometry. Micromachines 2019, 10, 868. [CrossRef]
10. O’Neal, C.B.; Malshe, A.P.; Singh, S.B.; Brown, W.D.; Eaton, W.P. Challenges in the packaging of MEMS. In Proceedings of the
International Symposium on Advanced Packaging Materials, Processes, Properties and Interfaces, Braselton, GA, USA, 14–17
March 1999; pp. 41–47.
11. Savazzi, S.; Guardiano, S.; Spagnolini, U. Wireless sensor network modeling and deployment challenges in oil and gas refinery
plants. Int. J. Distrib. Sens. Netw. 2013, 9. [CrossRef]
12. Khan, W.Z.; Aalsalem, M.Y.; Gharibi, W.; Arshad, Q. Oil and Gas monitoring using Wireless Sensor Networks: Requirements,
issues and challenges. In Proceedings of the IEEE International Conference on Radar, Antenna, Microwave, Electronics, and
Telecommunications (ICRAMET), Jakarta, Indonesia, 3–5 October 2016; pp. 31–35.
13. Hester, J.; Sorber, J. The future of sensing is batteryless, intermittent, and awesome. In Proceedings of the 15th ACM Conference
on Embedded Network Sensor Systems, New York, NY, USA, 6–8 November 2017; pp. 1–6.
14. Samaun; Wise, K.D.; Angell, J.B. An IC piezoresistive pressure sensor for biomedical instrumentation. IEEE Trans. Biomed. Eng.
1973, 20, 101–109. [CrossRef]
Sensors 2023, 23, 2525 20 of 21

15. Sander, S.; Knutti, J.W.; Meindl, J.D. A monolithic capacitive pressure sensor with pulse-period output. IEEE Trans. Electron.
Devices 1980, 27, 927–930. [CrossRef]
16. Bui, T.; Dinh, T.; Terebessy, T. Pressure sensor based on bipolar discharge corona configuration. Sens. Actuators A Phys. 2016, 237,
81–90.
17. Luo, X.; Gianchandani, Y.B. A Microdischarge-Based Pressure Sensor Fabricated Using Through-Wafer Isolated Bulk-Silicon Lead
Transfer. J. Microelectromech. Syst. 2018, 27, 365–373. [CrossRef]
18. Ma, W.; Jiang, Y.; Hu, J.; Jiang, L.; Zhang, T. Microelectromechanical system-based, high-finesse, optical fiber Fabry–Perot
interferometric pressure sensors. Sens. Actuators A Phys. 2020, 302, 111795. [CrossRef]
19. Feng, F.; Jia, P.; Qian, J.; Hu, Z.; An, G.; Qin, L. High-Consistency Optical Fiber Fabry–Perot Pressure Sensor Based on Silicon
MEMS Technology for High Temperature Environment. Micromachines 2021, 12, 623. [CrossRef]
20. Song, P.; Ma, Z.; Ma, J.; Yang, L.; Wei, J.; Zhao, Y.; Zhang, M.; Yang, F.; Wang, X. Recent progress of miniature MEMS pressure
sensors. Micromachines 2020, 11, 56. [CrossRef]
21. Giachino, J.M.; Haeberle, R.J.; Crow, J.W. Method for Manufacturing Variable Capacitance Pressure Transducers. U.S. Patent
4,386,453, 7 June 1983.
22. Ji, J.; Cho, S.T.; Zhang, Y.; Najafi, K. An Ultraminiature CMOS Pressure Sensor for a Multiplexed Cardiovascular Catheter. IEEE
Trans. Electron. Devices 1992, 39, 2260–2267. [CrossRef]
23. Benken, A.; Gianchandani, Y.B. A High-Yield Microfabrication Process for Sapphire Substrate Capacitive Pressure Sensors
Providing 70 MPa Range and 0.5 kPa Resolution. IEEE Sens. J. 2021, 21, 27315–27324. [CrossRef]
24. Liao, P.; Hong, H.; Chiu, Y. A CMOS-MEMS Capacitive Pressure Sensor with Differential Sensing Electrodes and On-Chip
Frequency Output Circuits. In Proceedings of the IEEE 21th International Conference on Solid-State Sensors, Actuators and
Microsystems (Transducers), Orlando, FL, USA, 20–24 June 2021; pp. 90–93.
25. Richter, M.; Wackerle, M.; Woias, P.; Hillerich, B. A novel flow sensor with high time resolution based on differential pressure
principle. In Proceedings of the IEEE 12th International Conference on Micro Electro Mechanical Systems, Cat. No. 99CH36291.
Orlando, FL, USA, 21 January 1999; pp. 118–123.
26. Deng, Y.; Jin, N.; Yang, Q.; Wang, D. A differential pressure sensor coupled with conductance sensors to evaluate pressure drop
prediction models of gas-water two-phase flow in a vertical small pipe. Sensors 2019, 19, 2723. [CrossRef]
27. Kamali, F.; Hussain, F.; Cinar, Y. A laboratory and numerical-simulation study of co-optimizing CO2 storage and CO2 enhanced
oil recovery. SPE J. 2015, 20, 1227–1237. [CrossRef]
28. Gilliland, R. Apparatus and Method for Measuring Relative Permeability and Capillary Pressure of Porous Rock. U.S. Patent No.
5,297,420, 29 March, 1994.
29. Dutta, P.P.; Li, T.; Gianchandani, Y.B. A Sensing Microsystem for High Resolution Pressure Gradient Measurement in Core-Flood
Experiments. In Proceedings of the IEEE SENSORS, Montreal, QC, Canada, 27–30 October 2019; pp. 1–4.
30. Sun, Y.; Li, Q.; Yang, D.; Liu, X. Laboratory core flooding experimental systems for CO2 geosequestration: An updated review
over the past decade. J. Rock Mech. Geotech. Eng. 2016, 8, 113–126. [CrossRef]
31. Ott, H.; de Kloe, K.; van Bakel, M.; van Pelt, A.; Legerstee, P.; Bauer, A.; Eide, K.; van der Linden, A.; Berg, S.; Makurat, A.
Core-flood experiment for transport of reactive fluids in rocks. Rev. Sci. Instrum. 2012, 83, 084501. [CrossRef]
32. Prada, A.; Civan, F. Modification of Darcy’s law for the threshold pressure gradient. J. Pet. Sci. Eng. 1999, 22, 237–240. [CrossRef]
33. Fonseca, M.A.; English, J.; Arx, M.; Allen, M.G. Wireless micromachined ceramic pressure sensor for high-temperature applica-
tions. J. Microelectromech. Syst. 2002, 11, 337–343. [CrossRef]
34. Henao-Sepulveda, J.A.; Toledo-Quinones, M.; Jia, Y. Contactless monitoring of ball bearing temperature. In Proceedings of the
IEEE Instrument Measurement Technology Conference, Ottawa, ON, Canada, 17–19 May 2005; pp. 1571–1573.
35. Collins, C. Miniature passive pressure transensor for implanting in the eye. IEEE Trans. Biomed. Eng. 1967, 14, 74–83. [CrossRef]
[PubMed]
36. Mohan, S.; Hershenson, M.; Boyd, S.; Lee, T. Simple accurate expressions for planar spiral inductances. IEEE J. Solid-State Circuits
1999, 34, 1419–1424. [CrossRef]
37. Ferreira, J. Improved analytical modeling of conductive losses in magnetic components. IEEE Trans. Power Electron. 1994, 9,
127–131. [CrossRef]
38. Pichorim, S.; Abatti, P. Design of coils for millimeter-and submillimeter-sized biotelemetry. IEEE Trans. Biomed. Eng. 2004, 51,
1487–1489. [CrossRef]
39. Zhang, C.; Wang, L.; Huang, J.; Huang, Q. An LC-type passive wireless humidity sensor system with portable telemetry unit. J.
Microelectromech. Syst. 2014, 24, 575–581. [CrossRef]
40. EOS GmbH, Germany. Aluminum AlSi10Mg Material Data Sheet. Available online: https://www.eos.info/en/additive-
manufacturing/3d-printing-metal/dmls-metal-materials/aluminium-al (accessed on 15 July 2022).
41. DuPont. Viton Fluoroelastomer Brochure. Available online: https://www.ber-pa.it/UserFiles/Download/Viton_Brochure%2020
10.pdf (accessed on 4 March 2022).
42. NXP Semiconductors Inc. 50 kPa On-Chip Temperature Compensated and Calibrated Silicon Pressure Sensors Datasheet.
Available online: https://www.nxp.com/docs/en/data-sheet/MPX2053.pdf (accessed on 10 October 2022).
43. Smart Products USA, Inc. Check/Pressure Relief Valves Product Catalog, Series 200, Model #214. Available online: https:
//www.smartproducts.com/docs/Smart_Products_Product_Catalog.pdf (accessed on 12 January 2022).
Sensors 2023, 23, 2525 21 of 21

44. Loudon, C.; McCulloh, K. Application of the Hagen—Poiseuille Equation to Fluid Feeding through Short Tubes. Ann. Entomol.
Soc. Am. 1999, 92, 153–158. [CrossRef]
45. Crary, S.B.; Baer, W.G.; Cowles, J.C.; Wise, K.D. Digital compensation of high-performance silicon pressure transducers. Sens.
Actuators A Phys. 1990, 21, 70–72. [CrossRef]
46. Oppenheim, A.; Buck, J.; Schafer, R. Discrete-Time Signal Processing; Prentice Hall: Hoboken, NJ, USA, 1999.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.