Readout Circuits For Capacitive Sensors

micromachines
Review
Readout Circuits for Capacitive Sensors
Yongsang Yoo and Byong-Deok Choi *
Department of Electronic Engineering, Hanyang University, Seoul 04763, Korea; yongsangyoo@hanyang.ac.kr

* Correspondence: bdchoi@hanyang.ac.kr
Abstract: The development of microelectromechanical system (MEMS) processes enables the in-
tegration of capacitive sensors into silicon integrated circuits. These sensors have been gaining
considerable attention as a solution for mobile and internet of things (IoT) devices because of their
low power consumption. In this study, we introduce the operating principle of representative capaci-
tive sensors and discuss the major technical challenges, solutions, and future tasks for a capacitive
readout system. The signal-to-noise ratio (SNR) is the most important performance parameter for
a sensor system that measures changes in physical quantities; in addition, power consumption is
another important factor because of the characteristics of mobile and IoT devices. Signal power
degradation and noise, which degrade the SNR in the sensor readout system, are analyzed; circuit
design approaches for degradation prevention are discussed. Further, we discuss the previous efforts
and existing studies that focus on low power consumption. We present detailed circuit techniques
and illustrate their effectiveness in suppressing signal power degradation and achieving lower noise
levels via application to a design example of an actual MEMS microphone readout system.
Keywords: capacitive sensors; capacitive readout system; self-capacitance readout circuit; mutual ca-
pacitance readout circuit; signal power degradation; noise reduction; charge-balancing compensation;
parasitic capacitance; chopper stabilization; (microelectromechanical systems) MEMS microphone
readout circuit

Citation: Yoo, Y.; Choi, B.-D.

Readout Circuits for Capacitive
Sensors. Micromachines 2021, 12, 960. 1. Introduction
https://doi.org/10.3390/mi12080960 Recent growth in mobile devices and the internet of things (IoT) market has led to an
increased demand for various types of sensors [1]. Mobile and IoT devices have a small
Academic Editor: Jae-Sung An size and are equipped with batteries; further, because of their size limitations, a small
form factor is required for the sensor. Low power consumption is important for battery-
Received: 6 July 2021
powered devices that cannot operate for longer periods with their smaller batteries [2].
Accepted: 10 August 2021
The electrical characteristics of the sensor change according to the physical quantity to
Published: 13 August 2021
be measured. Thus, depending on the type of change in the electrical property being
measured, sensors can be classified as resistive, capacitive, inductive, and magnetic [3].
Publisher’s Note: MDPI stays neutral
Among these, capacitive sensors have received considerable research attention because
with regard to jurisdictional claims in
of the recent development of the microelectromechanical system (MEMS) processes [4–6].
published maps and institutional affil-
Further, capacitive sensors have become a popular research topic as they are advantageous
iations.
because of their small form factor and low power consumption [5–7].
Resistive, inductive, and magnetic sensors have static current paths, which inevitably
consume power when operating sensors. The direct current(DC) does not flow because
there is no DC current path in the capacitive sensor; this is advantageous in terms of power
Copyright: © 2021 by the authors.
consumption. Some capacitive sensors require a high-voltage DC bias; however, even in
Licensee MDPI, Basel, Switzerland.
these cases, there is no DC current flow. It is possible to decrease the power consumption
This article is an open access article
of the sensor using various driving methods such as charge–discharge cycle adjustment
distributed under the terms and
or charge-sharing techniques because the power consumption of a capacitive sensor is
conditions of the Creative Commons
Attribution (CC BY) license (https://
determined by the charge–discharge charge, formulated as Q = C · V. Sensors that require
creativecommons.org/licenses/by/
mechanical movement among conventional capacitors (gyroscope, accelerator, pressure
4.0/). sensor, etc.) are difficult to implement in small sizes because of problems such as the
Micromachines 2021, 12, 960. https://doi.org/10.3390/mi12080960 https://www.mdpi.com/journal/micromachines

Micromachines 2021, 12, x 2 of 32
Micromachines 2021, 12, 960 2 of 30
pressure sensor, etc.) are difficult to implement in small sizes because of problems such
as the production, assembly, and durability of mechanical parts [7]. The recent develop-
production,
ment of MEMS assembly,
processes andhasdurability
facilitated of mechanical
the development parts [7]. The recentsensors
of capacitive development
with me-of
MEMS processes has facilitated the development of capacitive sensors with mechanical
chanical movement to be implemented on a very small scale on a semiconductor chip,
movement to be implemented on a very small scale on a semiconductor chip, which meets
which meets the needs of the small form factor.
the needs of the small form factor.
A capacitive sensor system comprises a capacitive sensor and readout system. The
A capacitive sensor system comprises a capacitive sensor and readout system. The
readout system and sensor are very important in terms of the overall system performance
readout system and sensor are very important in terms of the overall system performance
and power consumption. Although the capacitive sensor has few noise sources, most of
and power consumption. Although the capacitive sensor has few noise sources, most of
the noise is introduced from the outside or from the other major circuits of the readout
the noise is introduced from the outside or from the other major circuits of the readout
system. Further, because the power consumption of the capacitive sensor is low, the
system. Further, because the power consumption of the capacitive sensor is low, the power
power consumptions of other circuits (analog front-end circuit, analog-to-digital con-
consumptions of other circuits (analog front-end circuit, analog-to-digital converter, post-
verter, post-processor, etc.) of the readout system are dominant. Therefore, most major
processor, etc.) of the readout system are dominant. Therefore, most major specifications,
specifications, such as the signal-to-noise ratio(SNR) and power consumption required for
such as the signal-to-noise ratio(SNR) and power consumption required for a capacitive
asensor
capacitive
system,sensor
are system, are more
more affected by affected
the readoutby the readout
system system
than than the
the sensor sensor
itself. itself.
Thus, the
Thus, the capacitive readout system is very important for achieving
capacitive readout system is very important for achieving specific goals such as meeting specific goals such as
meeting SNR or power consumption requirements. Among
SNR or power consumption requirements. Among the capacitive readout systems, the the capacitive readout sys-
tems,
analog the analog front-end
front-end circuitry, circuitry,
which is which
the firstis of
thethe
first of the systems
readout readout to systems to be con-
be connected to
nected to the sensors, uses different structures and driving methods
the sensors, uses different structures and driving methods based on the characteristics of based on the charac-
teristics
the sensor;of the sensor;
further, further,
it has it has a significant
a significant influence on influence on the
the overall overall performance
performance of
of the entire
the entire readout system. That is, the effective performances of
readout system. That is, the effective performances of other circuits such as analog-to-other circuits such as an-
alog-to-digital
digital converters converters
(ADCs) (ADCs) and post-processing
and post-processing logic that
logic circuits circuits thatsignals
receive receivefrom
signals
the
from the analog front-end circuit and the design target specifications
analog front-end circuit and the design target specifications considerably depend on the considerably depend
on the performance
performance of the analog
of the analog front-endfront-end
circuit.circuit.
The
The rest
rest ofofthis
thispaper
paperisisorganized
organized as asfollows.
follows. TheThedifferent
different types
types of
of capacitive
capacitive sensors
sensors
and the structure of the basic readout system are discussed in Section
and the structure of the basic readout system are discussed in Section 2. The main technical 2. The main tech-
nical issues that degrade the performance of the capacitive readout
issues that degrade the performance of the capacitive readout system are presented in system are presented
in Section
Section 3, addition
3, in in addition to the
to the methods
methods employed
employed to resolve
to resolve thesethese issues.
issues. Finally,
Finally, in Sec-
in Section 4,
tion 4, we introduce
we introduce the readout
the readout circuitcircuit
designdesign
method method
through through
an actualan actual
designdesign
example example
of the
of the analog
analog front-endfront-end
circuitcircuit
for theforMEMSthe MEMS microphone.
microphone.
2. Capacitive
2. Capacitive Sensor
Sensor and
and Readout
Readout System
System
In this
In this section,
section, the
the basic
basic configuration
configuration and and features
features of
of the capacitive
capacitive sensor
sensor and its
readout system
readout systemwillwillbebe
introduced.
introduced. TheThe
typical configuration
typical of a capacitive
configuration sensor readout
of a capacitive sensor
system is
readout shownis in
system Figure
shown in1.Figure
The system comprises
1. The system an analog
comprises front-end
an analog circuit,circuit,
front-end which
interfaces with the sensor and converts the capacitance change caused
which interfaces with the sensor and converts the capacitance change caused by the by the change in
the physical quantity to be measured into voltage or current; an ADC,
change in the physical quantity to be measured into voltage or current; an ADC, which which converts
analog signals
converts analogto digitaltosignals;
signals and a post-processor
digital signals; for signal
and a post-processor for post-processing. Based
signal post-processing.
on the application, the sensor and ADC can be directly connected except
Based on the application, the sensor and ADC can be directly connected except for the for the analog
front-end
analog circuit;circuit;
front-end if the output is analog,
if the output it may only
is analog, it maycomprise the sensor
only comprise the and theand
sensor analog
the
front-end circuit.
analog front-end circuit.
Figure
Figure 1.
1. Block
Blockdiagram
diagram of
of capacitive
capacitive sensor
sensor and
and readout
readout system.
system.
2.1. Capacitive
2.1. Capacitive Sensor
Sensor
There are
There are various
various types
typesofofcapacitive
capacitivesensors;
sensors;however,
however, for forthe
theconvenience
convenienceof ofunder-
under-
standing, if we model
standing, model the
thecapacitance
capacitancebetween
betweentwo
twoflat
flatelectrodes
electrodes = εA/d);
(C(C thethe
= εA/d); capacitive
capaci-
sensors
tive can can
sensors measure the permittivity,
measure the distance
the permittivity, between
the distance the two
between the electrodes, or the
two electrodes, orarea
the
change
area caused
change by thebychange
caused in the in
the change physical quantity
the physical to be measured;
quantity the capacitive
to be measured; sensors
the capacitive
can be classified
sensors based on
can be classified theseon
based changes.
these changes.

An accelerometer, gyroscope, pressure sensor, and microphone are representative
An accelerometer,
capacitive sensors based gyroscope, pressure
on the change sensor, between
in distance and microphone are representative
the two electrodes. These sen-
capacitive
sors are sensors based on
manufactured the change
through in distance
the MEMS between
process, whichtheallows
two electrodes. These sen-
the mechanical move-
An accelerometer, gyroscope, pressure sensor, and microphone are representative ca-
sors are manufactured through the MEMS process, which allows
ment of the electrode via the addition of the membranes to the IC fabrication process. A the mechanical move-
pacitive sensors based on the change in distance between the two electrodes. These sensors
ment of electrode
fixed the electrode via the addition
andthrough
membrane, which ofisthe membranes
a movable to the IC fabrication
electrode, process.
in the A
are manufactured the MEMS process, which allows forms a capacitor
the mechanical movement sen-
fixed
sor.electrode
The and membrane,
distance between thewhich
two is a movable
electrodes is electrode,
changed byforms
a a capacitor
physical force in the sen-
(accelerating
of the electrode via the addition of the membranes to the IC fabrication process. A fixed
sor. The pressure,
force, distance between
etc.), andthe two electrodes
when is changed by a physical force (accelerating
electrode and membrane, whichthe is aphysical
movableforce is reduced,
electrode, formsthe distance
a capacitor returns
in to the
the sensor.
force, pressure,
original value etc.), and
because when
oftwo the physical
the electrodes
elastic force force is reduced,
of the membrane. the distance
As examples returns to
of sensors the
that
The distance between the is changed by a physical force (accelerating force,
original
follow value
this because
principle, of
thethe elastic
accelerometerforce of
and the
themembrane.
pressure As
sensor examples
are shownof sensors
in Figurethat
2a,b.
pressure, etc.), and when the physical force is reduced, the distance returns to the original
follow this principle,
In Figure 2, theofgraytheand
accelerometer
green and the pressure sensor are
andshown in Figure respec-
2a,b.
value because the elastic forceparts
of therepresent
membrane. the electrode
As examples the membrane,
of sensors that follow
In tively.
FigureFor 2, the grayamplitude
and greenand parts represent thesensors
electrode and the membrane, respec-
this principle, the accelerometer and the pressure sensor are shown in Figure 2a,b. bias
signal linearity, some need the application of a DC In
tively.
Figure For
between 2, signal amplitude
the gray
the electrodes ofand
and green the linearity,
sensor
parts to some
represent thesensors
generate need thethe
an electrostatic
electrode and application ofrespectively.
force, as the
membrane, a membrane
DC bias
between
needs
For theexhibit
to
signal electrodes of the
appropriate
amplitude and sensor
elastic
linearity, to generate
force.
some sensors an need electrostatic force,ofas
the application the bias
a DC membrane
between
needs to exhibit appropriate elastic force.
the electrodes of the sensor to generate an electrostatic force, as the membrane needs to
exhibit appropriate elastic force.
Acceleration Acceleration
Acceleration Acceleration Pressure
Pressure
CS↑
CS↑ Proof CS↓ Pressure
Proof
mass CS↓ Pressure
mass Proof CS↓
CS↓ Proof
mass CS↓
CS↓ CS↑ mass CS↑
CS↑ CS↑
(a) (b)
(a) (b)
Figure 2. Distance-based capacitive sensors: (a) accelerometer; (b) pressure sensor.
Figure 2. 2.
Figure Distance-based capacitive
Distance-based sensors:
capacitive (a)(a)
sensors: accelerometer; (b)(b)
accelerometer; pressure sensor.
pressure sensor.
Capacitive sensors based on changes in permittivity between two electrodes include
Capacitive
Capacitive
a humidity sensors
sensors
sensor, based
based
particle ononchanges
changes
concentration inin
permittivity
permittivity
sensor, between
between
and chemical two
and twoelectrodes
electrodesinclude
biomedical include
substance
a humidity sensor,
aconcentration
humidity sensor, particle Asconcentration
particle
sensors. concentration
shown in Figuresensor,
3a, aand
sensor, andchemical
chemical
capacitive andandbiomedical
sensor a fixedsubstance
biomedical
has substance
sensor elec-
concentration
concentration sensors.
trode; the electrodesensors.As shown
can As shown
be arrangedin Figure 3a, a3a,
inhorizontally,
Figure capacitive sensor
aascapacitive
shown has a fixed
in sensor
Figure has asensor
3b. Because elec- is
fixed there
sensor
trode; the electrode
electrode;
no physical themovement,can be
electrode canarranged
no be horizontally,
arranged
special suchasasshown
horizontally,
process as shown
the in Figure
MEMS 3b. Because
inprocess
Figure 3b.required,
is there
Because is it
there
and
noiscan
physical
nobe movement,
physical movement, no special
no process
special process such as
such the
as MEMS
the MEMS process
process
manufactured relatively easily via the IC process without the need for bias voltage. is required,
is required, and
andit it
can
can be manufactured
be manufactured relatively
relatively easily
easily via
viathe IC
the process
IC process without
without the
In addition, a fast sensing speed is not required because the concentration of the material need
the need for bias
for voltage.
bias voltage.
InIn addition,
addition,
to a afast
be measured fastsensing
bysensing
thesespeedspeedisisnot
sensors notrequired
does required because the
because
not demonstrate concentration
concentration
abrupt of
changes easily.of the
thematerial
material
In to
addition
tobebemeasured
to measured
the bybythese
these
above-mentioned sensors
sensors does
method doesnot demonstrate
fornot demonstrate
detecting abrupt
changes changes
abrupt
in changes
the easily. In
easily.
distance andaddition to the
Inpermittivity
addition
toabove-mentioned
the above-mentioned
between method
two electrodes, a for
method detecting
displacement changes
for detecting in thesensor
changes
and position distance and permittivity
in the isdistance
used to and
measure between
permittivity
the capac-
two
between
itanceelectrodes,
two
change a displacement
electrodes,
caused abydisplacementand position
the change and
in sensor
theposition is used
sensor
overlapping tobetween
is used
area measure
to measure theelectrodes
the capacitance
the capac-de-
change
itance caused
change by
caused thebychange
the in the
change overlapping
in the area
overlapping between
pending on the horizontal movement between the electrodes of the two objects. area the
between electrodes
the depending
electrodes de-on
the horizontal movement between the electrodes of the
pending on the horizontal movement between the electrodes of the two objects. two objects.
ε
ε↓ ε↑ ε
CS↓
ε↓ CS↑
ε↑ CS
CS
CS↓ CS↑
(a) (b)
(a) (b)
Figure3.3.Permittivity-based
Figure Permittivity-basedcapacitive
capacitivesensors:
sensors:(a)
(a)vertically
verticallyplaced
placedconcentration
concentrationsensor;
sensor;(b)
(b)horizontally
horizontally placed
placed concen-
concen-
Figure 3. sensor.
tration
tration Permittivity-based
sensor. capacitive sensors: (a) vertically placed concentration sensor; (b) horizontally placed concen-
tration sensor.
Most capacitive sensors are configured either alone or in pairs for the differential.
Because the touch sensor is used in the form of an array on a display panel with a large
Most capacitive sensors are configured either alone or in pairs for the differential.
Because the touch sensor is used in the form of an array on a display panel with a large
physical size, there is is aa considerable
considerable difference
difference in in the
the environment
environment from from other
other capacitive
capacitive
sensors. In other sensors, the noise of the readout readout circuit
circuit is
is the
the dominant
dominant noise noise source.
source. InIn
the touch sensor,
sensor, the
the display
display noisenoise becomes
becomes thethe most
most dominant
dominant noisenoise source
source because
because thethe
touch and display panels are very close to each other, other, which
which results
results inin aa large
large parasitic
parasitic
capacitance between
between eacheachother.
other.Further,
Further,forforthis
this same reason, the
same reason, the noise fromnoise from outside
outside the
the sensor
sensor (such (such as charger
as charger noisenoise and human
and human noise) noise) is a significant
is a significant noise The
noise source. source. The
required
required
report ratereport
in arate in aapplication
touch touch application
is loweris lower
than thatthanofthat
theofother
the other sensors;
sensors; however,
however, the
the sensing
sensing timetime
per per sensor
sensor is shorter
is shorter thanthan
thatthat of other
of other sensors
sensors because
because the the sensor
sensor is inisthe
in
the form of an array. The time available for sensing per sensor is
form of an array. The time available for sensing per sensor is very limited because touch very limited because
touch
sensingsensing
is onlyisperformed
only performedduring during
the idlethe idle of
period period of the driving
the display display to driving to cope
cope with the
with the display noise. Further, the touch sensor can be divided into
display noise. Further, the touch sensor can be divided into self-capacitive and mutual- self-capacitive and
mutual-capacitive
capacitive types, astypes,shownasinshown Figurein4a,b,
Figure 4a,b, respectively,
respectively, accordingaccording
to the sensorto the sensor
structure
structure and the principle
and the principle of sensingofcapacitance.
sensing capacitance.
CS↓
CS↑
(a) (b)
Figure 4.
Figure 4. Capacitive
Capacitive touch
touch sensors:
sensors: (a)
(a) self-capacitive
self-capacitive type;
type; (b)
(b) mutual-capacitive
mutual-capacitive type.
type.
For the self-capacitive

self-capacitive type sensor,sensor, the the finger
finger is is sensed
sensed using
using capacitance
capacitance change
change
caused
caused bybythethedistance
distancebetween
between thethe
finger andand
finger the the
touch electrode.
touch Because
electrode. all capacitances
Because all capaci-
generated from afrom
tances generated single sensing
a single electrode
sensing (the point
electrode at which
(the point touchtouch
at which is to is
betosensed) are
be sensed)
measured,
are measured, the the
parasitic
parasitic capacitance
capacitance is very
is very large
largeininthe
thecase
caseofofaalarge-area
large-areatouchtouch panel,
which isis disadvantageous
disadvantageous in that the signal power caused by the touch is not not sufficiently
sufficiently
large. The mutual-capacitive type sensor senses changes in in capacitance between the
capacitance between the two
two
electrodes
electrodes caused by fringe field interference. The mutual-capacitive type does not suffer
caused by fringe field interference. The mutual-capacitive type does not suffer
from
from signal
signalpower
powerdegradation
degradation caused
caused by byparasitic
parasitic capacitance
capacitance as inas
theinself-capacitive type;
the self-capacitive
however, because
type; however, the fringe
because the field
fringebetween the two the
field between electrodes is used, one
two electrodes disadvantage
is used, one disad- is
that the amount
vantage of change
is that the amountinofcapacitance is rapidly reduced
change in capacitance when
is rapidly the distance
reduced when between the
the distance
finger
between and electrodes
the finger and in electrodes
the touch sensor increases
in the touch as the
sensor covering
increases as glass becomesglass
the covering thicker.
be-
This sensor does not suffer from a lack of signal power such as that
comes thicker. This sensor does not suffer from a lack of signal power such as that in the in the self-capacitive
type sensor; however,
self-capacitive type sensor;the use of the the
however, fringe
use field
of thebetween
fringe fieldthebetween
two electrodes
the twocauses the
electrodes
amount of change in capacitance to rapidly become insignificant
causes the amount of change in capacitance to rapidly become insignificant owing to the owing to the distance
between the fingerthe
distance between and the electrodes.
finger and the electrodes.
2.2. Analog Front-End Circuit
2.2. Analog Front-End Circuit
An analog front-end circuit can be classified into a circuit that reads the self-capacitance
An analog front-end circuit can be classified into a circuit that reads the self-capaci-
and another that reads the mutual capacitance. The former measures all capacitances occur-
tance and another that reads the mutual capacitance. The former measures all capacitances
ring in one node connected to the sensor, and the latter measures the capacitance between
occurring in one node connected to the sensor, and the latter measures the capacitance
the two separated connections with a connection to drive the sensor and a connection for
between the
measuring two
the separated connections with a connection to drive the sensor and a con-
sensor.
nection for measuring the sensor.
2.2.1. Self-Capacitance Readout Circuit
The most common method of sensing self-capacitance is measuring the amount of
charge required to charge the sensor. The circuit shown in Figure 5a is an example of a
self-capacitance readout circuit that uses a charge amplifier. The node where the readout
ΔVIN = −ΔVS
ΔCS
= VBIAS
CS
ΔCS
circuit and sensor are connected (NS ) is virtually shorted with = the referenceVBIAS
node (NREF )
CS 0 + ΔCS
by the charge amplifier. The charge amplifier charges and discharges the sensor such that
the voltage across the sensor In (V
theS )case
is equal to the voltage
of a MEMS applied
microphone, anto the NREF (V
appropriate REF ). force
elastic The must b
charge required to charge and discharge the sensor is provided via a feedback capacitor
the membrane to exploit the performance of the sensor; to this end, a relati
(CFB ) by the change in the output
voltage mustsignal (∆VSIG ). ∆V
be constantly REF is added
applied to the together with the
sensor. Figure 6b voltage
shows an exam
change required for sensor
widelycharging
used inand discharging
microphone to ∆VSIG . ∆V
applications. ASIG is obtaineddiode
back-to-back as is used to
impedance DC bias, and a p-channel transistor source follower is used to
∆VSIG signal
= ∆Vwith + (CS0 +∆CSDC
)(∆VS =∆VREF )
voltage REF a ground CFB voltage. This increases the DC level and l
(1)
+∆CS

+ CS0CF
pedance of=the1 analog ∆VREF
front-end
B
circuit output so that it can match the input
of the ADC or the output buffer that receives the signal from the source fol
CFB
VREF ΔVREF
Sensor
t
NS
VSIG VSIG
CS0 ΔCS VS VREF
ΔVSIG
t
(a)
ФRST
Ф1
Ф2
ФRST
VP VS VP
CFB ΔVS
VREF
Sensor Ф1
Ф2
t
NS
VSIG VSIG
CS0 ΔCS VS VREF VREF
ΔVSIG
t
(b)
Figure
Figure 5. 5.Readout
Readoutcircuits
circuits for
for self-capacitance
self-capacitance sensing:
sensing:(a)(a)
using charge
using amplifier;
charge (b) using
amplifier; charge integr
(b) using
charge integrator.
VBIAS VBIAS
Sensor Sensor
As shown in Equation (2), the change in the self-capacitance (∆CS ) can be calculated VDD
from ∆VSIG as
∆VSIG

ΔCS IREF
CS0 ΔCS VS ∆CS = − 1 CFB −CCS0S0 VS (2)
∆VREF
V
In contrast, as shown in N the VSIG5b that uses the switch-capacitorNIN
IN circuit in Figure charge
M1
integrator, the pre-charge and measurement phases are separated through the discrete-
time operation. Further, thereRHZ is an advantage that signals can be easily accumulatedRHZ by
repeating the pre-charge and measurement phases without resetting the stored charge in
the CFB . The sensor capacitor(a) charged to the pre-charge voltage (VP ) during (b) phase 1 (Φ1 )
is discharged to VREF during phase 2 (Φ2 ). The amount of charge required to discharge
Figure 6. Directcapacitor
the sensor current (DC) biased self-capacitance
is provided by the CFB . readout circuit:
After that, if (a) using1resistor
phases and 2 and
are buffer;
repeated(b) using ba
diode and source follower.
without resetting the stored charge in the CFB , the amount of charge sensed in the CFB is
accumulated, and a large signal is obtained. If the phase 1–2 operation is repeated N times,
∆VSIG and ∆CS are obtained as
CS0 + ∆CS
∆VSIG = − N (VP − VREF ) (3)
CFB
∆VSIG
∆CS = − CFB − CS0 (4)
N (VP − VREF )
CFB
VREF ΔVREF
Sensor
t
NS
VSIG VSIG
CS0 ΔCS VS VREF
ΔVSIG
t
(a)
The above-mentioned circuits ФRST measure the capacitance of the sensor based on the
principle of detecting the amount Ф1 of charge required to charge or discharge the sensor
Ф2
capacitor. The ФRST
voltage difference between the charging and discharging of the sensor
VP
capacitor should be large to increase
VS the output signal;
VP therefore, the voltage applied to the
CFB
sensor (VS ) changes significantly during measurement. ΔV However, some sensors require a
VREF S
Sensor Ф1
Ф
fixed bias voltage, and in this case, a readout circuit, as shown in Figure 6a, can be used
2
t
NSinstead of the circuits shown in Figure 5. In this circuit, a high-impedance path is formed at
VSIG V SIG
CS0 ΔCS VSthe node
VREFwhere the sensor and readout circuit are connected
VREF (NIN ) to apply the DC bias of
the node; however, the movement of the charge is prevented. Thus, the bias voltage of the
capacitor is maintained, and the amount of charge in the ΔVSIG
capacitor does not change. If the
capacitance changes because of the physical quantity change to be measured, the voltage
changes as t
(b)∆Q = ∆(CS )·VS + CS ∆VS
(5)
= 0charge amplifier; (b) using charge integrator.
Figure 5. Readout circuits for self-capacitance sensing: (a) using
VBIAS VBIAS
Sensor Sensor VDD
ΔCS IREF
CS0 ΔCS VS CS0 VS
VSIG
NIN VSIG NIN
M1
RHZ RHZ
(a) (b)
Figure6.6.Direct
Figure Directcurrent
current(DC)
(DC)biased
biasedself-capacitance
self-capacitancereadout
readoutcircuit:
circuit:(a)
(a)using
usingresistor
resistorand
andbuffer;
buffer;(b)
(b)using
usingback-to-back
back-to-back
Further, the voltage at node NIN changes as
∆VI N = −∆VS
= ∆C S
CS VBI AS (6)
∆CS
= CS0 +∆CS VBI AS
In the case of a MEMS microphone, an appropriate elastic force must be generated

in the membrane to exploit the performance of the sensor; to this end, a relatively high
bias voltage must be constantly applied to the sensor. Figure 6b shows an example of a
circuit widely used in microphone applications. A back-to-back diode is used to realize
a high impedance DC bias, and a p-channel transistor source follower is used to receive
an AC voltage signal with a ground DC voltage. This increases the DC level and lowers
the impedance of the analog front-end circuit output so that it can match the input voltage
range of the ADC or the output buffer that receives the signal from the source follower.
2.2.2. Mutual-Capacitance Readout Circuit

A mutual capacitance readout circuit measures the capacitance between the electrodes
of the sensor capacitor. It is different from the self-capacitance readout circuit in that
the nodes that apply the charge–discharge voltage (NTX ) and measure the change in the
amount of charge generated (NRX ) are separated. The simplest method to sense mutual
capacitance is using a charge amplifier, as shown in Figure 7a. When a voltage change is
applied to node NTX (∆VTX ), VS changes by −∆VTX ; a change in the amount of charge is
required to charge and discharge the sensor capacitor. This charge is provided by the CFB
via ∆VSIG . The generated ∆VSIG is given by
CS0 + ∆CS
∆VSIG = − ∆VTX (7)
CFB
Sensor CFB VTX ΔVTX

CS0
t
NRX
VTX ΔCS
VSIG VSIG
VREF ΔVSIG
VS t
(a)
ФRST
Ф1
Ф2
ФRST
VP VS VREF
Sensor
CFB
CS0 ΔVS
Ф1 -VP
Ф2
t
NRX
TX ΔCS
VSIG VSIG
Ф2 Ф1 VREF
VS ΔVSIG
VREF
t
(b)
Figure
Figure 7.
7. Readout
Readout circuits
circuits for
for mutual-capacitance
mutual-capacitance sensing:
sensing: (a)
(a) using
using charge
charge amplifier;
amplifier; (b)
(b) using
using charge integrator.
charge integrator.
2.3.Analog-to-Digital
The capacitanceConverters (ADC)
of the sensor can be calculated using
The current or voltage signal from the analog front-end circuit is an analog signal
∆VSIG
with various levels. The current∆C = − signal
or Svoltage CFB − Cfrom
read S0 the sensor requires post-pro- (8)
∆VTX
cessing to improve the overall sensor readout system performance; most of this post-pro-
cessingIf ∆Vis performed in the digital domain. Therefore, the sensor readout system often
SIG is not sufficiently large, the output signal can be accumulated using a
includes an ADC that
charge integrator composed convertsof the analogcapacitor,
a switch output ofasthe analog
shown in front-end
Figure 7b.circuit
Unlike intothea
digital signal. Since ADCs are widely used in many different fields,
self-capacitance readout circuit shown in Figure 5b, the mutual capacitance readout circuit their types are very
diverse,
can increase and their characteristics
the pre-charged and performance
voltage by changingranges are alsoofvery
the voltage nodediverse.
NRX (VTherefore,
RX ) in Φ1
it is necessary
because nodestoNselect
TX and the most
NRX aresuitable
separated.ADC considering
During Φ1 , VS the characteristics
is charged to −VPof the during
, and sensor
and itsisapplication
Φ2 , it dischargedfield. An. ADC
to VREF Whilefor VSa changes
sensor that −(Vmeasures
P + V REF the
) change
during Φ 1 in
andphysical
Φ 2 , thechar-
self-
acteristics
capacitancedoes not circuit
readout requireonly
high sample-per-second
changes VP − VREF in the (SPS)
sameperformance
process. When because of the
the number
characteristics
of repetitions of of Φ
the sensor.
1 and Φ2 Most ∆VSIG and
is N, sensors require a samplingcapacitance
the measured rate of severalare,tens of SPS to
respectively
several
given askSPS, and a slightly higher (200 kSPS) is required for a microphone that measures
∆Carrays,
acoustic signals. For touch∆V sensors
SIG = − N CS0C+FB
arranged as S
(−VP 100 to 10,000
− VREF ) sensors are placed on
one panel based on the size of the display ∆CS used. The report rate, which is the meas-
panel
CS0 + (9)
= N CFB (VP + VREF )
urement rate for the entire panel, requires 60–180 Hz; therefore, the SPS required for the
ADC in touch applications is∆C several kSPS∆VtoSIG several MSPS.
S = CFB − CS0 (10)
N ( V + V )
The resolution of the ADC is expressed by the number of bits of the ADC. Because
P REF
the Signal-to-quantization-noise ratio(SQNR) that can be achieved with a specific number
of bits has the value
SQNR = 6.02 ⋅ bits + 1.76 dB (11)
The number of bits should be determined such that they have a higher SQNR perfor-
mance than the SNR performance of the analog front-end circuit.
With the above method, noise reduction can be performed by arranging a filter to
remove noise after changing the capacitance of the sensor into the form of voltage, current,
etc. Subsequently, the signal is applied to the ADC, or if the analog signal is directly output
to the outside; the output buffer is used as required.
2.3. Analog-to-Digital Converters (ADC)

The current or voltage signal from the analog front-end circuit is an analog signal with
various levels. The current or voltage signal read from the sensor requires post-processing
to improve the overall sensor readout system performance; most of this post-processing is
performed in the digital domain. Therefore, the sensor readout system often includes an
ADC that converts the analog output of the analog front-end circuit into a digital signal.
Since ADCs are widely used in many different fields, their types are very diverse, and their
characteristics and performance ranges are also very diverse. Therefore, it is necessary to
select the most suitable ADC considering the characteristics of the sensor and its application
field. An ADC for a sensor that measures the change in physical characteristics does not
require high sample-per-second (SPS) performance because of the characteristics of the
sensor. Most sensors require a sampling rate of several tens of SPS to several kSPS, and
a slightly higher (200 kSPS) is required for a microphone that measures acoustic signals.
For touch sensors arranged as arrays, 100 to 10,000 sensors are placed on one panel based
on the size of the display panel used. The report rate, which is the measurement rate for
the entire panel, requires 60–180 Hz; therefore, the SPS required for the ADC in touch
applications is several kSPS to several MSPS.
The resolution of the ADC is expressed by the number of bits of the ADC. Because the
Signal-to-quantization-noise ratio(SQNR) that can be achieved with a specific number of
bits has the value
SQNR = 6.02·bits + 1.76 dB (11)
The number of bits should be determined such that they have a higher SQNR perfor-
mance than the SNR performance of the analog front-end circuit.
In general, the sensor and analog front-end circuit have an SNR of 60–120 dB. In the
case of a touch sensor, there is a considerable amount of noise that comes from the outside,
and therefore, it has an SNR of 30–80 dB. Equation (12) must be satisfied for the ADC to
not become the bottleneck of the overall readout system performance.
SNR − 1.76 dB
Bits > (12)
6.02
A resolution of 10–20 bits is required to satisfy the above SNR range; 5–13 bits are
required for a touch sensor.
Figure 8 shows the approximate sampling rate and resolution characteristics of three
types of ADCs: pipeline, SAR, and delta-sigma, which are commonly used as ADC struc-
tures. A delta-sigma ADC is used when tens of SPS to several kSPS sampling rates with
16 bits or more resolution are required; when several tens of kSPS sampling rates with
16 bits or less resolution are required or when power consumption is important, SAR ADC
is used.
In the case of the delta-sigma ADC, a high SNR is achieved through a noise-shaping
effect and oversampling. SQNR is the theoretical SNR only considering quantization noise,
and it is affected by the number of bits of the comparator and DAC(N), the delta-sigma
loop order(L), and the oversampling ratio(OSR). SQNR is calculated as
π 2L

SQNR [dB] = 6.02· N + 1.76 + (20L + 10) log10 OSR − 10 log10 (13)
2L + 1
The resolution, loop order, and OSR of the delta-sigma ADC were determined based
on the target SNR. The delta-sigma ADC can achieve a very high SQNR depending on its
configuration; however, there are disadvantages in that the power consumption increases
A resolution of 10–20 bits is required to satisfy the above SNR range; 5–13 bits are
required for a touch sensor.
Figure 8 shows the approximate sampling rate and resolution characteristics of three
Micromachines 2021, 12, 960 types of ADCs: pipeline, SAR, and delta-sigma, which are commonly used as ADC9struc-of 30
tures. A delta-sigma ADC is used when tens of SPS to several kSPS sampling rates with
16 bits or more resolution are required; when several tens of kSPS sampling rates with 16
bits or less resolution are required or when power consumption is important, SAR ADC
because
is used.the sampling frequency increases in proportion to the oversampling ratio, and an
amplifier with a high bandwidth is required.
24
22
Resolution (bits) 20
Σ-Δ
18
16
SAR
14
12 PIPELINE
10
8
10 100 1k 10k 100k 1M 10M 100M 1G

Sampling rate (Hz)
Figure8.8.Analog-to-digital
Figure Analog-to-digitalconverter(ADC)
converter(ADC)architectures,
architectures,resolution,
resolution,and
andsampling
samplingrates
rates[8].
[8].
The SARcase
In the ADC of has
the an advantageADC,
delta-sigma in thata ithigh
is easy
SNR tois
implement
achieved low power
through compared to
a noise-shaping
other
effectADCs
and by using a capacitor
oversampling. SQNR DAC. Thetheoretical
is the SAR ADC SNR performsonlya considering
binary searchquantization
from most
significant
noise, andbit it (MSB) to least
is affected significant
by the numberbit of(LSB)
bits ofwith
the the charge-sharing
comparator operation
and DAC(N), theofdelta-
the
capacitor DAC, and it then performs the analog-to-digital
sigma loop order(L), and the oversampling ratio(OSR). SQNR is calculated as conversion. Because the capacitor
DAC does not require an amplifier, it is advantageous in terms of power consumption.
 π 2 L  ADC, and
However, the SAR SQNRADC [dB]does not⋅ Napply
= 6.02 + ( 20shaping
+ 1.76noise L + 10 ) logsuch as the delta-sigma
10 OSR − 10 log10   (13)
therefore, the SQNR increase caused by oversampling is small comparedto + 1delta-sigma
2 Lthe 
ADC, as shown by
The resolution, loop order, and OSR of the delta-sigma ADC were determined based
on the target SNR. TheSQNR delta-sigma ADC
[dB] = 6.02 · Ncan achieve
+ 1.76 + 10a· log
very high SQNR depending on
10 OSR (14)its
configuration; however, there are disadvantages in that the power consumption increases
That the
because is, when an ADC
sampling of the same
frequency resolution
increases is used, to
in proportion thethe
SQNR is lower than
oversampling that
ratio, andof an
a
delta-sigma
amplifier with ADC. a high bandwidth is required.
The SAR ADC has an advantage in that it is easy to implement low power compared
2.4. Post-Processing
to other ADCs by using a capacitor DAC. The SAR ADC performs a binary search from
most significantreadout
The sensor bit (MSB) system uses
to least a post-processor
significant bit (LSB)towith achieve higher performance
the charge-sharing or
operation
support various additional functions by removing noise and errors
of the capacitor DAC, and it then performs the analog-to-digital conversion. Because the difficult to remove
using only DAC
capacitor an analog
does not front-end
requirecircuit throughita isdigital-domain
an amplifier, advantageous operation.
in terms ofErrors
powerthat con-
are difficult to remove using the analog front-end circuit alone
sumption. However, the SAR ADC does not apply noise shaping such as the delta-sigma include external noise in
the frequency
ADC, within the
and therefore, the SQNR
sensingincrease
signal band,
causederrors caused by sensor
by oversampling is smallmismatches,
compared to and the
errors of the ADC. It
delta-sigma ADC, as shown byis not only difficult to design a circuit to cope with these errors, but
it is also disadvantageous in terms of power consumption and area as the analog filter is
vulnerable to nonlinearity. However, digital-domain filters are relatively less affected by
these problems [9]. Further, post-processing has the advantage of improving performance
even after chip manufacturing by enabling programmable changes to the filter according
to the environment or user needs through a firmware update.
Post-processing allows compensating sensor characteristics such as the temperature
variation of the sensor and process variation between sensors and circuit limitations such
as offset and gain error in analog front-end circuits or ADCs. Further, it compensates
for external noise. In the case of touch sensors, capacitance changes may occur because
of water droplets or palms, and problems that cannot be solved with capacitor sensors
or analog front-end circuits alone are eliminated in post-processing by comprehensively
evaluating the capacitance change pattern, intensity, and so on. Post-processing also allows
calculating the touched position coordinates based on the capacitance measured by the
touch sensor.
3. Technical Issues of the Capacitive Sensor Readout System

This section introduces the technical issues of the capacitive sensor readout system
and also the solutions that have been presented for the major issues. Table 1 summarizes
the technical issues for each application covered in capacitive sensor readout system studies
published in major journals and conferences in the field of sensors and circuits for the
past 10 years (2011–2021). The gyroscope was included in the accelerometer because of
the similarity of its operation principle to that of the accelerometer; the microphone was
included in the pressure sensor because it senses sound signals through pressure. Sensors
that have common features for sensing the concentration of substances, such as particle
meters and chemical sensors, are classified as concentration sensors.
Table 1. Paper counts on technical issues of capacitive sensor readout system published in major
sensor conferences and journals from 2011 to 2021.
Sensor Application Total

Issues Touch Accelerometer Pressure Concentration Etc.
SNR 12 15 2 1 3 33
Power consumption 5 1 5 1 1 13
Sensing speed 4 - 2 - - 6
Sensitivity 1 - - 2 2 5
Etc. 2 1 - - 2 5
Readout method 1 1 - 1 - 3
Total 25 18 9 5 8 65
Table 1 indicates that the SNR is a very important performance indicator for the
accurate measurement of the physical quantity because the core function of the sensor
readout system is to measure the physical quantity. Further, power consumption is an
important consideration because capacitive sensors are used extensively in small devices.
Sensing speed is not a major design issue because the rate of change of the physical quantity
to be measured is not fast in most sensor systems. However, touch applications operate as
a matrix based on the time-division method. The number of channels and parasitic R and
C increase as the physical size of the sensor increases; further, the report rate requirement
for smooth touch increases. Thus, unlike in other sensor readout systems, sensing speed
issues should be considered. However, the sensitivity is related to the performance of the
sensor itself rather than the readout system.
3.1. Signal-to-Noise Ratio (SNR): Signal Power

It is necessary to increase the signal power or reduce the noise power to achieve a
higher SNR because SNR is the ratio of signal power to noise power. Conversely, SNR
is lowered when the signal power is degraded or noise is increased. Figure 9 shows
the capacitive sensor and basic self-capacitance readout circuit previously introduced in
Figure 5. Figure 9 shows the base capacitance (CS0 ) and capacitance change (∆CS ) caused
by a physical quantity, and the parasitic capacitance (CP ) between the sensor and input of
the readout circuit, including the input capacitance of the readout circuit, for a capacitive
sensor. When VREF is changed by ∆VS , the amount of charge on the sensor capacitor
changes; the resulting ∆QSIG is given by
∆QSIG = (CP + CS0 + ∆CS )∆VS (15)
Further, the output voltage change (∆VSIG ) caused by ∆QSIG can be obtained as:
∆QSIG
∆VSIG = + ∆VS
CFB
CP +CS0 +∆CS

= C FB
+ 1 ∆VS (16)
+ 1 ∆VS + ∆C

CP +CS0
CFB ∆VS
S
= CFB
Equation (16) indicates that ΔVSIG includes not only the terms caused by sensor ca-
pacitance change (ΔCS) but also the term caused by the parasitic and base capacitances. If
CP and CS0 are large, a large feedback capacitor (CFB) must be used to maintain ΔVSIG within
Micromachines 2021, 12, 960
the power supply range without clipping. Accordingly, there is a problem in that the
11 of 30
amount of voltage change ( 𝛥𝑉 ) in response to the change in the capacitance of the
sensor decreases, which results in a decrease in signal power.
CFB
Sensor ΔVREF
NS
VSIG
CS0 ΔCS VS CP VREF
ΔVREF
Figure 9. Base and parasitic capacitances in a self-capacitance readout circuit.

Figure 9. Base and parasitic capacitances in a self-capacitance readout circuit.
A DC-biased
Equation (16) self-capacitance
indicates that ∆V readout circuit is used for sensors that require the DC
SIG includes not only the terms caused by sensor
bias (V BIAS), as shown in Figure 6. Figure 10 shows the sensor, CP, and readout circuit. The
capacitance change (∆CS ) but also the term caused by the parasitic and base capacitances.
Ifoutput
C and
P
ofCthisare
S0
circuit is aalso
large, affected
large by Ccapacitor
feedback P.
(C ) must be used to maintain ∆V
FB SIG
within the power supply range without clipping. Accordingly, there is a problem in that
the amount of voltage change ( ∆C CFB ∆VS ) in response to the change in the capacitance of the
S
sensor decreases, which results in a decrease in signal power.

Micromachines 2021, 12, x
A DC-biased self-capacitance readout circuit is used for sensors that require the12DC of 32
bias (VBIAS ), as shown in Figure 6. Figure 10 shows the sensor, CP , and readout circuit. The
output of this circuit is also affected by CP .
∆C ΔC S
ΔV
∆VSIGN = = C S+ C VVBIBIAS
SIGNAL
AL (17)
(17)
CS0 S+0 CP P AS
VBIAS
Sensor
CS0 ΔCS VS
NIN VSIG
CP RHZ
Figure10.
Figure 10.Base
Baseand
andparasitic
parasiticcapacitances
capacitancesinina aDC-biased
DC-biasedself-capacitance
self-capacitancereadout
readoutcircuit.
circuit.
Equation(17)
Equation (17)indicates
indicatesthat
thatthethesignal
signalpower
power decreases
decreases with
with anan increase
increase in in
CPCand
P and CS0. .
CS0
ItItisisnot
notpossible
possibletotoreduce
reduceCC S0S0inin the
the analog
analog front-end
front-end circuit
circuit because
because itit
isis determined
determined by
by
thesensor
the sensor design.
design. However,
However, CP C is affected
isPaffected byphysical
by the the physical
design,design,
such assuch as the routing
the routing metal
metal of
layout layout of the connection
the connection between between the and
the sensor sensor
theand the circuit,
circuit, becausebecause
CP occurs CP occurs
between be-
tween
the the output
sensor sensor output
and theand inputtheofinput of the readout
the readout circuit circuit alongthe
along with with the input
input capaci-
capacitance
oftance of the readout
the readout circuit.circuit.
When When designing
designing the the layout
layout connecting
connecting thethe sensor
sensor andandreadout
readout
circuit,efforts
circuit, effortsshould
shouldbebemade
madetotominimize
minimizethe theparasitic
parasiticcapacitance.
capacitance.Despite
Despitethesetheseefforts,
efforts,
thegate
the gatecapacitance
capacitanceofofinput
inputtransistor,
transistor,thetheparasitic
parasiticcapacitance
capacitanceofofthethepad,
pad,andandrouting
routing
metalremain.
metal remain.
3.1.1.
3.1.1.Mutual-Capacitance
Mutual-CapacitanceSensingSensing
AAmutual-capacitance
mutual-capacitancesensing
sensingmethod
methodisiswidely
widelyused
usedforformany
manysensors
sensorstotosolve
solvethe
the
problem of signal power reduction caused by the parasitic capacitance of self-capacitance
problem of signal power reduction caused by the parasitic capacitance of self-capacitance
sensing.
sensing.InInthe
themutual-capacitance
mutual-capacitancesensing
sensingmethod,
method, the TX
the TXnode
node (N(N ) )that
TXTX thatapplies
appliesvoltage
voltage
totothe sensor capacitor and the RX
the sensor capacitor and the RX node (Nnode (N RX ) that measures the change in charge are
RX) that measures the change in charge are
separated.
separated. AsAs shown
shown inin Figure
Figure 11,
11,thethevoltage
voltageatatnode
nodeNRXNRX to VtoREFVby
is fixed
is fixed byOp
REFthe theAmp.
Op
Amp.
WhenWhen a driving
a driving voltage
voltage (ΔVTX(∆V ) is applied
) isTXapplied to sensor,
to the the sensor,
the the voltage
voltage across
across thethe sensor
sensor ca-
capacitor
pacitor and the amount of charge changes; however, the voltage does not change at at
and the amount of charge changes; however, the voltage does not change the
parasitic capacitor in NRX, and the parasitic capacitor does not affect the change in the
amount of charge or the output signal (ΔVSIG), which is expressed as Equation (7). How-
ever, the effect of parasitic capacitance on mutual capacitance sensing is not fully elimi-
nated. VRX is not exactly equal to VREF and varies depending on the output voltage of Op
the parasitic capacitor in NRX , and the parasitic capacitor does not affect the change in
the amount of charge or the output signal (∆VSIG ), which is expressed as Equation (7).
However, the effect of parasitic capacitance on mutual capacitance sensing is not fully
eliminated. VRX is not exactly equal to VREF and varies depending on the output voltage
of Op Amp (VSIG ) because the real Op Amp does not have an infinite open-loop gain (A).
A (CS0 + ∆CS )
Micromachines 2021, 12, x ∆VSIG = − ∆VTX 13 (18)
of 32
A + 1 CS0 +CP + CFB
A +1
Sensor CFB
CS0
NRX
VTX ΔCS
VSIG
ΔVTX CP VREF
Figure 11.
Figure 11. Base
Base and
and parasitic
parasitic capacitances
capacitances in
in aa mutual-capacitance
mutual-capacitance readout
readout circuit.
circuit.
For sensors that require

∆VSIG includes the DC bias used in some applications, it is difficult to equa-
use a
The the open-loop CS 0 + gain
CS of Op CAmp, shown in Equation (18). The
mutual-capacitance
tion indicates that the influence  V
readout SIG = −
circuit because
of CP is almost  V a −
largeCOMP
 V
signal
TX reduced by COMP the open-loop gain; however,TX
cannot be applied to the it
node. Further, the effect of C was CFB
eliminated, but Cthe
FB effect of CS0 was not.
does not completely disappear. P
For sensors that require=the CS 0 V TX +used

CCOMPinV C
− DC bias some
COMP applications,
− S VTX it is difficult to use(22) a
3.1.2. Charge-Balancing Compensation: Using
mutual-capacitance readout circuit becauseCaFBlarge signal cannot a Compensation CFB be Capacitor
applied to the TX node.
Further, the effect of CP was
A charge-balancing compensationCS
eliminated, but the effect
technique using of aCS0 was
CS 0 not.
compensation capacitor (CCOMP)
can be used to eliminate the effect = −  V , when  V = − V
CFBof CS0Using
. Figure 12 shows the charge-balancing compen-
TX COMP TX
3.1.2. CCOMP
sationCharge-Balancing
circuit using CCOMP Compensation:
in the self-capacitance a Compensation
readout circuit. Capacitor
In this circuit, CCOMP is
added AWhen
charge-balancing
to theusing
nodethe where compensation
methodthe parasitic technique using
with thecapacitance
additional C COMPa
occurs compensation
, the
(N compensation
S), capacitor
and the voltage charge (Camount
on the other)
COMP
can
andbe
side of used
the CCOMP to(Veliminate
charge amount
COMP the
by Ceffect
) is driven S0inand ofC
the CPS0must
same . Figure
phase 12 shows
exactly
and match
polarity the
toas charge-balancing
be removed.
the voltage Therefore, compen-
of node NS.forIf
sation
accurate
the circuit
change inusing
compensation,
charge CCOMP
amount in the
either Cself-capacitance
COMP isbecause
changes implementedofreadout as acircuit.
CS0 and Cvariable
P (ΔQInOFFthis circuit,
capacitor
) and theor CCOMP
VCOMPis
change is
in
added
charge to
implemented the node
amountascaused where
a DAC the parasitic
by [10].
ΔVCOMP capacitance
Further,
and CCOMP SAR(ΔQ occurs
logicCOMPis (N
) areused
S ), and the voltage
to determine
determined on
to bethe the other
the correct
same,
side
ΔQ of is
OFF CCOMP
compensation (V
supplied by C)[11,12].
charge
COMP is driven
COMP andEven in the
not same
in the
the case
feedbackphase and polarity
ofcapacitor
variable (CFB).as
capacitors theand
This isvoltage
DACs, of node
formulated as N
there isSa.
Iflimit
the change in charge amount changes
in that the perfect compensation is difficult becausebecause of C S0 and C
of the (∆Q
P limitation
OFF ) and the change
of the resolution. in
charge amount caused
In addition, when using signals by ∆V COMP
Δ Q and
OFF
of other
= (
C C P + C
waveforms
COMP (∆Q
S0 ) Δ V S
such
COMP ) are determined to be
as sine waves instead of accurate the same,
∆Q OFF issuch
pulses supplied
as ΔV byREFCCOMP
in the and not the = Δfeedback
self-capacitanceQCOMP readoutcapacitor (CFB ).orThis
circuit ΔVTX is formulated
in the mutual- as
(19)
+ C(S0Δ)V∆V − ΔVifS )the accurate negative polarity

capacitance readout circuit, there may=be an errorCOMP
source
∆QOFF = (CC COMP
of compensation voltage driving is not Prealized. ToS solve this problem, a structure that
The voltage ∆Q
=for COMP (19)
compensates for of ΔVCOMP required
a negative capacitance this canusing
circuit be obtained as
an additional Op amplifier without
=C (∆VCOMP − ∆VS )
applying a negative polarity signal hasCOMP
been proposed [13,14,15].
 C + CS 0 
ΔVCOMP =  P + 1 ΔVS (20)
 CCOMP
CCOMP 
VCOMP
Further, the influence of CS0 and CP on the change in output voltage (ΔVSIG) is re-
moved by the charge-balancing compensation, and only the change
CFB ΔVCOMP
in sensor capacitance
is reflected.
Sensor
( C + CP ) ΔVS − CCOMP ( ΔVCOMP − ΔVS ) ΔCS
N
ΔVSIG = SS 0 + ΔVS + ΔVS
CFB C
VFB
SIG
CS0 ΔCS VS CP VREF (21)
 ΔCS 
= + 1 ΔΔV
VS REF
 CFB 
Figure 12. Charge-balanced
A12.
Figure Charge-balancedcompensation
charge-balancing compensation using
compensation using aa compensation
method compensation
can be appliedcapacitor inaaself-capacitance
to eliminate
capacitor in self-capacitance
the effect of
readout circuit.
the basecircuit.
readout capacitance in both self-capacitance sensing and mutual-capacitance sensing (Fig-
ure 13). When applied to a mutual capacitance readout circuit, the effect of the base ca-
Sensor if the sum of the charge due to the base capacitance (CS0ΔVTX)
pacitance can be removed
CS0
and the charge due to the compensation capacitor (CCOMPΔVCOMP) is 0, as
CS 0 + ΔCS C C
VTX ΔCΔSVSIG = − ΔVTX − COMP ΔFB
VCOMP
CFB CFB
FB FB
C C
= − S VTX , when VCOMP = − S 0 VTX
CFB CCOMP
When using the method with the additional CCOMP, the compensation charge amount
and the charge amount by CS0 and CP must exactly match to be removed. Therefore, for
accurate compensation, either CCOMP is implemented as a variable capacitor or VCOMP is
implemented as a DAC [10]. Further, SAR logic is used to determine the correct
compensation
The voltagecharge
of ∆V[11,12]. Even in the case of variable capacitors and DACs, there is a
COMP required for this can be obtained as
limit in that the perfect compensation is difficult
becauseof the limitation of the resolution.
C +
In addition, when using signals of other waveforms
P C S0 such as sine waves instead of accurate
∆VCOMP = + 1 ∆VS (20)
pulses such as ΔVREF in the self-capacitance CCOMP readout circuit or ΔVTX in the mutual-
capacitance readout circuit, there may be an error source if the accurate negative polarity
Further, the voltage
of compensation influence of CS0isand
driving not C P on theTo
realized. change in output
solve this voltage
problem, (∆VSIGthat
a structure ) is
removed by the charge-balancing compensation, and only the change in sensor capacitance
compensates for a negative capacitance circuit using an additional Op amplifier without
is reflected.
applying a negative polarity signal has been proposed [13,14,15].
(CS0 +CP )∆VS −CCOMP (∆VCOMP −∆VS ) ∆CS
∆VSIG = CFB + CFB ∆VS + ∆VS

∆CS
CCOMP (21)
= CFB + 1 ∆VS
VCOMP
A charge-balancing compensation method can CFBbe applied toΔVeliminate
COMP the effect of
the base capacitance in both self-capacitance sensing and mutual-capacitance sensing
Sensor
(Figure 13). When applied to a mutual capacitance readout circuit, the effect of the base
NS if the sum of the charge due to the base capacitance (C ∆V )
capacitance can be removed S0 TX
and the charge due to the compensation capacitor (CCOMP ∆VCOMP VSIG ) is 0, as
CS0 ΔCS VS CP VREF
∆VSIG = − CS0C+FB∆CΔV∆VTX − CCOMP
S REF
CFB ∆VCOMP
CS0 ∆VTX +CCOMP ∆VCOMP
=− CFB − ∆C
CFB ∆VTX
S
(22)
∆C
Figure 12. Charge-balanced=compensation using a compensation capacitor
CS0 in a self-capacitance
− CFB ∆VTX , when ∆VCOMP = − CCOMP ∆VTX
S
readout circuit.
Sensor
CS0
CFB
VTX ΔCS
ΔVTX NRX
VSIG
ΔVCOMP CCOMP CP VREF
VCOMP
Figure 13.
Figure 13. Charge-balanced
Charge-balanced compensation
compensation using
using aa compensation
compensation capacitor
capacitor in
in aa mutual-capacitance
mutual-capacitance
readout circuit.
readout circuit.
3.1.3.When
Charge-Balancing Compensation:
using the method Using a Differential
with the additional Sensor
CCOMP , the compensation charge amount
and the chargemethod
Another amount ofby CS0 and CP must
compensation exactly
uses an match
another to be
sensor removed.
instead of a Therefore,
compensation for
accurate
capacitor.compensation,
Figure 14 shows CCOMP
eithersuch is implemented as
a charge-balancing a variable
circuit. Therecapacitor VCOMP to
exists aormethod is
implemented as a DAC [10]. Further, SAR logic is used to determine the correct
configure two identical sensors such that the base capacitance is the same as that shown compensa-
tion charge14a
in Figure [11,12].
or a Even
singleinsensor
the casecomposed
of variableofcapacitors and DACs,
two differential there is aas
capacitors, limit in that
shown in
the perfect compensation is difficult because of the limitation of the resolution.
Figure 14b. Instead of balancing the charge from the added compensation capacitor and In addition,
when using signals of other waveforms such as sine waves instead of accurate pulses such
as ∆VREF in the self-capacitance readout circuit or ∆VTX in the mutual-capacitance readout
circuit, there may be an error source if the accurate negative polarity of compensation
voltage driving is not realized. To solve this problem, a structure that compensates for a
negative capacitance circuit using an additional Op amplifier without applying a negative
polarity signal has been proposed [13–15].
3.1.3. Charge-Balancing Compensation: Using a Differential Sensor

Another method of compensation uses an another sensor instead of a compensation
capacitor. Figure 14 shows such a charge-balancing circuit. There exists a method to
configure two identical sensors such that the base capacitance is the same as that shown
in Figure 14a or a single sensor composed of two differential capacitors, as shown in
sensor are used are, respectively, given by
CS 0 + CS + C + CS −
VSIG1 = − VTX − S 0 ( −VTX )
CFB CFB
(23)
CS + − CS −
Micromachines 2021, 12, 960 =− VTX 14 of 30
CFB
CS 0 + CS C − ( −CS )
VSIG 2 = − VTX − S 0 ( −VTX )
Figure 14b. Instead of balancing the CFBcharge from theCadded
FB compensation capacitor and
(24)
2CS in the sensor, the base capacitance of the two sensors
the charge from the base capacitance
= − V
are very similar even if they are not completely
TX matched because both methods in Figure 14
CFB
use the same two sensors; thus, an improved compensation is possible. The output signal
(∆VAs ) when in
SIG1shown twoEquation
identical(23),
sensors
the and output in
difference signal
the (∆V SIG2 ) when
capacitance a differential
change (ΔCS+ − sensor
ΔCS-)
are usedthe
between are,two
respectively,
sensors isgiven by and therefore, it is difficult to measure if a similar
measured,
change in the physical quantity is applied
C +∆CS+
to the twoCsensors. Thus, there is a limitation in
S0 + ∆CS−
that one sensor must∆V beSIG1 − aS0place
= in
placed CFB ∆V
that is
TX −
not affected
CFB (−by∆V
the )
TXphysical quantity(23)
or
∆CS+ −∆CS−
−
placed far away so that similar=physical ∆VTX
CFBquantity changes do not occur in the two sensors
or require post-processing. When a differential sensor is used, the difference between the
CS0 +∆C S0 −(− ∆CS )
∆VSIG2rather
two capacitances becomes = −largeCFB
because
S
− Ccapacitance
∆VTX the CFB ∆V
(−of the sensor for the same
TX )
physical quantity forms a differential 2∆CSand reacts with opposite polarity. Therefore, in the (24)
= − CFB ∆VTX
case of the differential, it is possible to obtain the effect of increasing the signal power and
removing the effect of the base capacitance.
VTX Sensor
ΔVTX CFB
CS0 ΔCS+
NRX
Sensor
VSIG1
CP VREF
CS0 ΔCS-
ΔVTXb
VTXb
(a)
VTX Sensor
CFB
ΔVTX
CS0 ΔCS
NRX
VSIG2
VREF
CS0 -ΔCS CP
ΔVTXb
VTXb
(b)
Figure
Figure14.
14.Charge-balanced
Charge-balancedcompensation
compensationusing
usingaacompensation
compensationcapacitor
capacitorininaamutual-capacitance
mutual-capacitance
readout circuit. (a) configured identical sensors such that the base capacitance; (b) a single sensor
readout circuit. (a) configured identical sensors such that the base capacitance; (b) a single sensor
composed of two differential capacitors.
composed of two differential capacitors.
The compensation method (23),using the sensorindescribed above change

assumes(∆C
that the base
As shown in Equation the difference the capacitance S+ − ∆CS− )
capacitance of the two sensors is the same; however, the effect of the base capacitance
between the two sensors is measured, and therefore, it is difficult to measure if a similar is
change in the physical quantity is applied to the two sensors. Thus, there is a limitation in
that one sensor must be placed in a place that is not affected by the physical quantity or
placed far away so that similar physical quantity changes do not occur in the two sensors
or require post-processing. When a differential sensor is used, the difference between the
two capacitances becomes rather large because the capacitance of the sensor for the same
physical quantity forms a differential and reacts with opposite polarity. Therefore, in the
case of the differential, it is possible to obtain the effect of increasing the signal power and
removing the effect of the base capacitance.
The compensation method using the sensor described above assumes that the base
capacitance of the two sensors is the same; however, the effect of the base capacitance is
not fully eliminated because a mismatch occurs because of process variation. Sensors and
= VOS +  
A
CFB
Further, ΔVERROR can be calculated as the capacitance measurement error of the

sensor as
V
CS , ERROR = − ERROR CFB
VTX
VOS CFB (26)
compensation capacitors are sometimes used = − together
VTX CPfor
+ 2charge-balancing
CS 0 compensation
+
to eliminate these residual errors [16–18]. Further, 1non-ideality such as the offset and open-
AC
loop gain error of Op Amp and sensor mismatch cause theFBparasitic and base capacitances
of theTherefore,
sensor to not
it isbenecessary
fully eliminated. To explain
to increase this, the offset
the open-loop voltage
gain (A) (VOS
of the Op) and
Ampopen-
and
loop gain (A) were assumed for Op Amp, as shown in Figure 15. Even if it
design to have a small offset to minimize the error caused by the non-ideality of the Op is assumed
that
Amp.theHowever,
parasitic and
thisbase capacitances
circuit of the two
design increases thesensors
powermatch, the output
consumption andvoltage error
area, which
(∆V
should
ERROR caused by V
be)considered OS and AOversampling
together. are given by successive approximation circuit structures
and methods have been proposed to remove gain error using methods besides the design
of a high-performance Op ∆VAmp = VOS
ERROR[19,20]. + ∆Qmethods
These
ERROR
reduce the gain error gradually
CFB
∆VSIG (25)
by accumulating and repeating the phase for −storing
A P +2Ccompensating
(Cand S0 ) the gain error
= VOS + CFB
several times to obtain one output result.
VTX Sensor
CFB
CS0 ΔCS
NRX
A VSIG
VREF
CS0 -ΔCS CP
VOS
VTXb
Figure15.
Figure 15. Non-idealities
Non-idealities of
ofOp
OpAmp
Ampin
inaacharge-balanced
charge-balancedcompensation
compensationcircuit.
circuit.
Further, ∆VERROR
3.2. Signal-to-Noise can (SNR):
Ratio be calculated
Internalas the capacitance measurement error of the
Noise
sensor as internal noise of the readout system can be divided into the sensor, analog front-
The
∆CS,ERROR = − ∆V∆V ERROR
CFB
end circuit, and ADC noise. In the case of the analog
TX front-end circuit, the Op Amp is the
VOS
= − ∆VTX CFB (26)
main noise source, and flicker noise is dominant because
1+ PAC the
C +2CS0 signal frequency band is low.
FB
Chopper stabilization is widely used as a circuit technique to remove such flicker noise.
Therefore,
Figure 16 shows it isannecessary
exampleto ofincrease
a circuitthe open-loop
that gain (A) of
applies chopper the Op AmpAs
stabilization. and design
shown in
to have a small offset to minimize the error caused by the non-ideality
Figure 16b, when low-frequency noise such as flicker is combined with the signal and is of the Op Amp.
However,
difficult tothis circuit design
distinguish, increases
as shown the power
in Figure 16c, theconsumption and area,
signal frequency which should
is modulated by the
be considered together. Oversampling successive approximation circuit
chopping frequency to distinguish the signal from noise. Further, as shown in Figure structures and
methods have been proposed to remove gain error using methods besides
16d,e, the signal that has passed the demodulation is moved to the low-frequency band, the design of
a high-performance Op Amp [19,20]. These methods reduce the gain error gradually by
accumulating and repeating the phase for storing and compensating the gain error several
times to obtain one output result.
3.2. Signal-to-Noise Ratio (SNR): Internal Noise

The internal noise of the readout system can be divided into the sensor, analog front-
end circuit, and ADC noise. In the case of the analog front-end circuit, the Op Amp is the
main noise source, and flicker noise is dominant because the signal frequency band is low.
Chopper stabilization is widely used as a circuit technique to remove such flicker noise.
Figure 16 shows an example of a circuit that applies chopper stabilization. As shown in
Figure 16b, when low-frequency noise such as flicker is combined with the signal and is
difficult to distinguish, as shown in Figure 16c, the signal frequency is modulated by the
chopping frequency to distinguish the signal from noise. Further, as shown in Figure 16d,e,
the signal that has passed the demodulation is moved to the low-frequency band, and
the noise in the low-frequency band is moved to the chopping frequency. Subsequently, a
low-pass filter (LPF) is applied, and only noise is removed.
CFB
Sensor
ФCLK ФCLK
CS0 ΔCS
VOP
VTX VN
LPF
VON
CS0 -ΔCS
Point Point Point

CFB
A B C
(a)
N N LPF LPF
S S S N S
f f N
DC fCH DC fCH f f
Point A DC fCH DC fCH
Point B Point C
(b) (c) (d) (e)
Figure 16. Chopper-stabilized fully differential amplifier with a differential sensor: (a) Circuit diagram; (b) signal and
Figure 16. Chopper-stabilized fully differential amplifier with a differential sensor: (a) Circuit diagram; (b) signal and noise
noise frequency without chopper stabilization; (c) with chopper stabilization at point A; (d) point B; (e) point C.
frequency without chopper stabilization; (c) with chopper stabilization at point A; (d) point B; (e) point C.
Correlated double sampling (CDS) is a widely used low-frequency noise-reduction
Correlated double sampling (CDS) is a widely used low-frequency noise-reduction
method. The CDS takes only the readout signal, which is the difference after sensing the
method. The CDS takes only the readout signal, which is the difference after sensing the
output
output by
by thethe low-frequency
low-frequencynoise noiseand andoutput,
output, including
including bothboth
thethe low-frequency
low-frequency noisenoise
and
and readout signal. This method is performed in post-processing
readout signal. This method is performed in post-processing after sampling and conversion after sampling and con-
version
twice intwice
ADC,in orADC, or it is implemented
it is implemented to subtract tonoise
subtract noise
in the analog in the analog
circuit when circuit when
measuring
measuring the readout signal after sampling the noise value into
the readout signal after sampling the noise value into the capacitor. Alternatively, CDS can the capacitor. Alterna-
tively, CDS can beasimplemented
be implemented a time-domain as adifferential
time-domain thatdifferential
measures each that measures
output with each output
opposite
with opposite sensor-driving polarity and then takes the difference
sensor-driving polarity and then takes the difference between the two [21]. The CDS between the two [21].
The CDS method is only applicable to low-frequency noise because
method is only applicable to low-frequency noise because the time for sampling only noise the time for sampling
only noise
and the timeand forthe time forasampling
sampling a signal containing
signal containing noise are not noise
thearesame.not the same.
The
The process-voltage-temperature(PVT)
process-voltage-temperature(PVT) variation variation of of the
the sensor
sensor and and analog
analog circuit
circuit has
has
aa significant
significant influence
influence on on the
the accuracy
accuracy and and reliability
reliability ofof the
the sensor readout system.
sensor readout system. The
The
charge-balancing compensationmethod
charge-balancing compensation methodcan canbe besaid
saidtotobebe a countermeasure
a countermeasure against
against pro-
process
cess variation through circuit design. In addition to the method
variation through circuit design. In addition to the method of circuit design, the method of circuit design, the
method of storing the output value caused by the process
of storing the output value caused by the process variation of the sensor in advancevariation of the sensor in ad-
vance and reflecting
and reflecting it in post-processing
it in post-processing is oftenis often
usedused[22]. [22]. Temperature
Temperature variation
variation affects
affects the
the capacitance
capacitance of the
of the capacitive
capacitive sensor
sensor andand affects
affects thethe characteristics
characteristics of of
thethe devices
devices used
used in
in
thethe readout
readout circuit,which
circuit, whichresults
resultsininchanges
changesininthe theoffset
offsetandand gain
gain ofof the
the Op Amp and
ADC.
ADC. The most common method method used
used to to eliminate
eliminatethe theeffect
effectofoftemperature
temperatureisistotoinclude
includea
atemperature
temperature sensor
sensor in the
in the readout
readout circuit
circuit to compensate
to compensate for thefor the of
effect effect of temperature
temperature change
change
based on based
the on the measured
measured temperature.
temperature. For compensation,
For compensation, it is necessary
it is necessary to analyze
to analyze how
how the capacitive
the capacitive sensor sensor and readout
and readout systemsystem are affected
are affected by temperature
by temperature [23,24]. [23,24]. The
The charge-
balancing compensation
charge-balancing compensationmethodmethod
discussed in Section
discussed in 3.1.2
Sectioncan3.1.2
be used
can asbe aused
compensation
as a com-
method. The
pensation effectThe
method. of temperature variation variation
effect of temperature on the output
on thevoltage
outputcan be eliminated
voltage can be elim-by
using abycompensation
inated using a compensation capacitor capacitor
as a temperature sensor [25].
as a temperature sensorAn [25].
orthogonal pseudo-
An orthogonal
random signal issignal
pseudo-random placed is in the signal
placed in thepathsignal while
pathoperating, and only
while operating, andthe output
only by the
the output
pseudo-random signal is separated using the orthogonal characteristic
by the pseudo-random signal is separated using the orthogonal characteristic from the from the combined
output to compensate
combined for the gainfor
output to compensate variation
the gain ofvariation
the Op Amp caused
of the Op Amp by temperature without
caused by temper-
usingwithout
ature an additional
using temperature
an additionalsensor. Then, the
temperature gain Then,
sensor. of Op theAmp canofbeOp
gain calculated
Amp canand be
compensated
calculated andfrom the separated
compensated fromsignal [26].
the separated signal [26].
3.3. Signal-to-Noise Ratio (SNR): External Noise

The effect of external noise on the capacitive sensor readout system is a topic addressed
in touch sensors because, unlike other sensors, the noise originating from the outside is
larger than the noise of the readout circuit itself in touch applications. Further, in the case
of touch sensors, the area constituting the mutual capacitance is large, and the parasitic
capacitance that occurs between them is very large (up to hundreds of pF) because the
distance between the sensor electrode and display panel is very close and the physical size
of the display/touch panel is large. Further, the pixels of TFT-LCD or AMOLED are driven
with a relatively large data voltage, and therefore, the noise flowing into the touch sensor
electrode is large. In addition, noise inflow is caused by the reference potential difference,
such as charger noise and human noise, because the size of the sensor itself is larger than
that of other sensors; touch sensors are subject to external inflow signals as noise such as
an antenna. Therefore, in touch applications, the focus is on removing these external noises
rather than the noise of the circuit.
3.3.1. Fully Differential Structure and Chopper Stabilization

The fully differential structure and chopper stabilization were used to remove the noise
of the readout circuit, as indicated in Section 3.2. This method is also used to eliminate
external noise in touch applications. Among external noises, the display noise, which
has a major influence on the SNR of the touch readout system, is common noise (locally
independent), and therefore, it is removed by the common-mode rejection performance
of the fully differential Op Amp. The display noise entering the touch sensor can be
suppressed by using two adjacent sensors in a differential structure. Although the two
sensors are adjacent to each other, there is a difference in the location because of the
difference in the connection of the signal lines, and the incoming noise is slightly different.
Chopper stabilization used to remove external noise can be implemented using the circuit
shown in Figure 17a. As shown in the operation example in Figure 17, the signal from the
sensor is demodulated to DC using the chopping frequency (fCH ), which is the same as the
frequency of the sensor-driving signal (fTX ); only the sensor signal in the low-frequency
band is left using an LPF. The modulated noise with the high-frequency band is removed by
the LPF. Thus, not only the external noise but also the noise of the readout circuit, such as
offset error, are removed. Recent studies used bandpass filters (BPFs) before demodulation
and LPF to achieve higher noise-reduction performances [27,28]. This method can remove
noises at different frequencies from signals; however, there is a limit in that it is difficult
to remove noises that exist over the signal frequency. To solve this problem, instead of a
fixed chopping frequency, a method to find and change the optimal chopping frequency
to remove noise, and a method for applying pseudo-noise to the chopping frequency to
provide a dithering effect to the chopping frequency, were proposed [29,30].
3.3.2. Averaging, Integration, and Parallel Driving Method

Random noise has white noise characteristics evenly distributed over the entire fre-
quency range. It is not easy to remove noise through a frequency filter because noise is
mixed at the same frequency as the signal. In this case, the effect of noise with the same
frequency as the signal can be reduced through integration and averaging. If a sensor is
sensed multiple times instead of only once, a constant value is measured for the signal
because of the change in the amount of the sensor; however, the random noise can ideally
be eliminated by averaging the repeatedly measured output values. Accumulating the
results of repeated sensing through the configuration shown in Figure 18 also has the same
average effect on random noise. Further, there is a method to remove the periodic display
noise with an average effect by using periodicity inversely [31]. Both the method of the
repeated measurement for averaging the output value and the method of integration have
a disadvantage in that the sensing time is increased by the number of repetitions compared
to the one-time sensing method. Further, the random noise removal performance shows
higher performance as the number of repetitions increases; however, there is a problem in

that it is difficult to increase the number of repetitions because the touch sensor is composed
of an array, and there are many sensors that need to be measured.
CFB
Sensor
ФCLK
CS0 ΔCS+ CN
VOP
VTX VN
LPF
VON
CN
CS0 ΔCS-
Sensor
Point Point Point
CFB
A B C
(a)
ФCH
VTX
ΔCS+>ΔCS- N S N
VON
Point A
VOP
t f
DC fCH = f TX
VON LPF
Point B
VOP
t f
DC fCH
VON
LPF
Point C
VOP
t f
DC fCH
(b)
Micromachines 2021, 12, x Figure 20 of 32
Figure17.17.
Chopper-stabilized fully
Chopper-stabilized differential
fully amplifier
differential for afor
amplifier touch sensor:
a touch (a) circuit
sensor: diagram;
(a) circuit (b)
diagram;
time-domain and frequency-domain signals.
(b) time-domain and frequency-domain signals.
3.3.2. Averaging, Integration, and Parallel Driving Method

CFB
Sensor Random noise has white noise characteristics ФCH
evenly distributed over the entire fre-
quency range. ФCLKIt is not easy to remove noise through VTX a frequency filter because noise is
CS0 ΔCS+ mixed at the same frequency as the signal. In this case, the effect of noise with the same
frequency as the signal can be reduced through VOP integration and ΔCS+ >ΔCS-
averaging. If a sensor
VON is
VTX sensed multiple times instead of only once, a constant value is measured for the signal
because of the change in the amount of the VON sensor; however, the random noise can ideally
CS0 ΔCS- be eliminated by averaging the repeatedly measured output values. Accumulating the
results of repeated sensing through the configuration shown in Figure 18 also has the VOPsame
Sensor average effect on random noise. Further, there is a method to remove the periodict display
noise with an average effect CFB by using periodicity inversely [31]. Both the method of the
repeated
Figure 18. Fully measurement
differential amplifierfor averaging
with the
integration inoutput value and thereadout
a mutual-capacitance methodcircuit.
of integration have
Figure 18. Fully differential amplifier with integration in a mutual-capacitance readout circuit.
a disadvantage in that the sensing time is increased by the number of repetitions com-
pared A toAparallel
the one-time
parallel sensing
driving
driving method
method method. Further,
(PDM)that
(PDM) the randomsenses
thatsimultaneously
simultaneously noise multiple
sensesremoval performance
multiplesensors
sensorsinineach
each
shows
channel higher
channelhas performance
hasbeen
been developed
developed as the
to to number
overcome
overcome of repetitions
thisthis
timetime increases;
constraint
constraint however,
problem.
problem. there
Sensing
Sensing is sen-
the
the sensorsa
problem
ofsors
each in that
ofchannel it is
each channeldifficult to increase
withdivision
with time the
time division number of repetitions
is an existing
is an existing generalgeneral because
method method the touch
that is that sensor
is similar
similar to timeto
isdivision
composed
time division of an array, and there are many
(TDMA) sensors
used in that need to be
communication. measured.
multiple access (TDMA) used in communication. Unlike the general method,general
multiple access Unlike the PDM
method, PDM is similar to code division multiple access (CDMA) and frequency division
multiple access (FDMA) used in communication. For multiple-frequency driving similar
to FDMA, as shown in Figure 19a, signals with different frequencies are applied to several
TX channels simultaneously in parallel, and the combined signals of various frequencies
are measured from the readout circuit of each channel; further, each sensor is measured
is similar to code division multiple access (CDMA) and frequency division multiple access
(FDMA) used in communication. For multiple-frequency driving similar to FDMA, as
shown in Figure 19a, signals with different frequencies are applied to several TX channels
simultaneously in parallel, and the combined signals of various frequencies are measured
from the readout circuit of each channel; further, each sensor is measured simultaneously
by separating it by frequency through post-processing such as fast Fourier transform
(FFT) [32]. Orthogonal frequency division multiplexing (OFDM) can be applied because
the frequencies used for driving may interfere with each other [33]. For code-division
multiple sensing similar to CDMA, orthogonal signals are applied to each TX line, and
the output value for each sensor is separated through post-processing calculations [34,35].
Figure 19b shows an example using the 4 × 4 matrix Walsh–Hadamard matrix. In Figure 19,
the values from all four sensors are superimposed on the output; however, because the
applied code has an orthogonal characteristic, it can be separated by calculation. In the
example, the code is expressed as a pulse; a continuous-time AC signal can be encoded as
a code and used instead [36]. Because all PDM methods measure multiple sensors in the
array simultaneously, the average readout time does not increase. However, depending
on the code, the output signal may be saturated to output range. To prevent this from
occurring, efficient orthogonal code was proposed [27,37–41]. As an additional problem, if
the same code is repeatedly used for PDM, energy is concentrated at a specific frequency,
and therefore, it may not be able to remove random noise along with the EMI problem
properly. A method for creating and using a pseudo-random orthogonal code generator
was proposed to solve this problem [42]. The disadvantages of these PDM methods are
that the area and power consumption of a sensor-driver circuit required for creating and
applying multiple frequencies and a digital circuit for encoding and decoding orthogonal
codes increase. Depending on the code, the output signal may become saturated beyond
the output range as the signals for several sensors overlap. A more efficient orthogonal
code was proposed [27,37–41]. PDM method may not be able to properly remove random
noise along with the EMI problem if the same code is repeatedly used and concentrated
at a specific frequency. To solve this problem, a method of creating and using a pseudo-
random orthogonal code generator has been proposed [42]. The disadvantages of these
PDM methods are that the area and power consumption of a sensor-driver circuit for
creating and applying multiple-frequency and a digital circuit for encoding and decoding
orthogonal codes increase.
3.3.3. Frequency Adaption

Although various driving methods and circuit techniques have been introduced to
cope with external noise, there is an underlying limitation that noise cannot be properly
removed when the signal and noise frequencies overlap. It is difficult to expect a noise
reduction effect by averaging and integrating non-random noise. It may be possible to
determine a noise-free frequency band and use it as the sensor-driving frequency if the
noise characteristics of the display panel and environment where the touch sensor will be
used can be identified in advance. In touch applications, a large amount of external noise is
introduced because of the large size of the display panel, and in the case of mobile devices,
the size and frequency of the incoming noise vary greatly depending on the location where
it is used. Therefore, using a specific sensor-driving frequency and reducing noise is not a
realistic solution. As shown in the example spectrum of Figure 20, frequency adaptation is a
method wherein the readout system identifies and uses a frequency that is good for driving
and sensing the sensor because it has the least noise among various frequency bands. As
shown in the block diagram of Figure 20, the readout system measures the noise and uses
the FFT processor and MCU to determine the frequency band with the least noise, which
decides to use the frequency band as the sensor-driving frequency [30,32]. Through this
operation, optimal SNR performances can be achieved even in various noise environments.
VTXA Sensor
CFB
CS0 ΔCSA
NRX
VSIG
ADC FFT
CP VREF
CS0 ΔCSB
VTXB Sensor
Signal
power
VTXA VSIG
VTXB
t t f
DC fTXA fTXB
(a)
Sensor
VTXA
CS0 ΔCSA
VTXB CFB
CS0 ΔCSB
NRX
VTXC VSIG
ADC Processor
CP VREF
CS0 ΔCSC
VTXD
CS0 ΔCSD
VRST
VTXA VRST
1 1 1 1 VSIG[2]
VTXB VSIG[1]
1 -1 -1 1 VSIG
VTXC VSIG[3]
1 -1 1 -1
VSIG[0]
VTXD
1 1 -1 -1 t
(b)
Figure 19. Parallel driving method: (a) multiple-frequency driving; (b) code-division multiple sensing.
Figure 19. Parallel driving method: (a) multiple-frequency driving; (b) code-division multiple sensing.
3.3.3. Frequency Adaption
3.3.4. Post-Processing
Although various driving methods and circuit techniques have been introduced to
The touch application not only senses the capacitance change but also determines
cope with external noise, there is an underlying limitation that noise cannot be properly
whether a touch is made by the user’s finger. Besides human fingers, various physical noise
removed when the signal and noise frequencies overlap. It is difficult to expect a noise
sources, such as water or palms, can come into contact with the touch panel. These sources
reduction effect by averaging and integrating non-random noise. It may be possible to
have a certain amount of electric charge similar to that for a finger, and this can affect the
determine a of
capacitance noise-free
the touch frequency
sensor. Itband anddifficult
is very use it as
tothe sensor-driving
detect and removefrequency if the
those physical
noise sources from the analog front-end circuit, and therefore, the role of post-processing be
noise characteristics of the display panel and environment where the touch sensor will is
used can beinidentified
important in advance.
touch applications In touch
because applications,
it can determinea whether
large amount
it is aoffinger
external noise
touch or
is introduced because of the large size of the display panel, and in the case of mobile de-
vices, the size and frequency of the incoming noise vary greatly depending on the location
where it is used. Therefore, using a specific sensor-driving frequency and reducing noise
Micromachines 2021, 12, 960 is not a realistic solution. As shown in the example spectrum of Figure 20, frequency 21 ofad-
30
aptation is a method wherein the readout system identifies and uses a frequency that is
good for driving and sensing the sensor because it has the least noise among various fre-
quency bands. As shown in the block diagram of Figure 20, the readout system measures
not by comprehensively
the noise and uses the FFT evaluating
processortheandamount
MCU toofdetermine
capacitance
thechange andband
frequency the changed
with the
distribution pattern in the array [43,44]. Post-processing is an operation performed
least noise, which decides to use the frequency band as the sensor-driving frequency in the
digital
[30,32]. Through this operation, optimal SNR performances can be achieved even initvari-
domain, and therefore, if an improved noise removal algorithm is developed, has
the advantage in that
ous noise environments. it can be used in an existing system via a firmware update.
Touch Readout
ADC
sensor circuit
Sensor FFT
MCU
driver processor
Noise
power As driving frequency
Figure 20.
Figure 20. Concept
Concept of
of frequency
frequency adaption
adaption technique
technique for
for capacitive
capacitive sensor
sensor readout
readoutsystem.
system.
3.3.4.
3.4. Post-Processing
Power Consumption
Power consumption
The touch application is an important
not only senses issuethein addition
capacitance to the SNR in
change butthealso
capacitive sen-
determines
sor readout
whether circuit.isThe
a touch madeimportance of mobile
by the user’s finger.and IoT devices
Besides human cannot
fingers,be overemphasized.
various physical
In a readout
noise sources,system
such for capacitive
as water sensors,
or palms, canpower
come into consumption
contact withcan the
be divided into sensor
touch panel. These
drivers,
sources analog
have a front-end circuits,
certain amount of ADC,
electricand logic similar
charge circuits to
forthat
post-processing.
for a finger, and this can
affectThe
thepower of theof
capacitance sensor driver
the touch is determined
sensor. It is veryby the capacitance
difficult to detect and of the sensor,
remove ex-
those
citation voltage, and driving frequency. A low excitation voltage
physical noise sources from the analog front-end circuit, and therefore, the role of post- can be used to lower
the power consumption;
processing is important inhowever, there is a because
touch applications problemitincan that the signal
determine power itcan
whether is aalso
fin-
be
gerreduced.
touch orThe notmethod of reducing power
by comprehensively consumption
evaluating the amountby lowering the sensor-driving
of capacitance change and
frequency
the changed can be used as pattern
distribution a method to significantly
in the array [43,44].reduce power consumption
Post-processing in touch
is an operation per-
applications
formed in the where thedomain,
digital parasiticand
capacitance
therefore, is very
if an large because
improved it isremoval
noise composed of a large-
algorithm is
area array [45]; further, a method of reducing driving energy through
developed, it has the advantage in that it can be used in an existing system via a firmware stepwise driving
can also be used [46]. For analog front-end circuits, there is a limit to lowering power con-
update.
sumption because of the trade-off between power consumption and noise. If the method of
increasing the signal power discussed in Section 3.1 is applied, there is room for reducing
3.4. Power Consumption
power consumption because the noise performance required to achieve the same SNR
Power consumption is an important issue in addition to the SNR in the capacitive
is alleviated [47]. Alternatively, a structure that removes the power of the pre-amplifier
sensor readout circuit. The importance of mobile and IoT devices cannot be overempha-
with a capacitance-to-digital (CDC) that connects the ADC directly to the sensor without
sized. In a readout system for capacitive sensors, power consumption can be divided into
using the pre-amplifier or a structure that eliminates overlapping circuits by combining the
sensor drivers, analog front-end circuits, ADC, and logic circuits for post-processing.
analog front-end pre-amplifier and ADC is suggested to reduce the power consumption of
The power of the sensor driver is determined by the capacitance of the sensor, exci-
analog front-end circuits [48–52].
tation voltage, and driving frequency. A low excitation voltage can be used to lower the
When adopting ADC for capacitive sensor readout systems, SAR ADCs with basic
power
low-powerconsumption;
characteristicshowever,
are the there is a problem in
most advantageous in terms
that the signalconsumption.
of power power can also Afterbe
sampling the input, the SAR ADC performs a binary search and compares the number
of binary bits required for each step by charge sharing through the capacitor DAC. For
this, the sensor undergoes the charge–discharge cycle for every sample, and charge sharing
of the capacitor DAC occurs at every conversion step. Therefore, a method to reduce the
conversion energy required per sample by reducing the conversion step or the number of
samplings is used [53,54].
The delta-sigma ADC used when a higher SNR is required employs a higher over-
sampling ratio for a higher noise shaping effect; therefore, it has a disadvantage in that it
consumes significant power as the sampling frequency is increased. Therefore, it is very
important to reduce the power consumption of the delta-sigma ADC in the readout circuit.
The bandwidth of the Op Amp must be increased to drive at a high driving frequency,
which leads to an increase in the power consumption of the Op Amp. Therefore, a structure
that does not need to increase the bandwidth of the Op Amp is adopted to lower the power
consumption of the delta-sigma ADC [55], or an inverter-based amplifier is used to lower
the power consumption and maintain a high bandwidth [56].
Recently, considerable research has been conducted on the method for lowering the
supply voltage to lower the power consumption in other fields. In the field of capacitive
sensor readout, analog circuits that represent information in the time domain instead
of expressing information as a voltage amplitude and operate at a low supply voltage
have also been proposed [57–60]. As one of these methods, an oscillator-based voltage-
controlled oscillator (VCO), as shown in Figure 21, is sometimes used as an ADC instead of
a traditional voltage-based ADC. The VCO can operate at a low supply voltage because it
is fabricated as a current hungry inverter; further, it consumes little power and can achieve
high resolution because the conversion resolution is in the time domain. However, it is
difficult to apply the method of increasing the signal power introduced in Section 3.1, the
method of reducing flicker noise and thermal noise, and the method of compensating
for external noise introduced in Sections 3.2 and 3.3. In addition, the linearity is poor
Micromachines 2021, 12, x because of the characteristics of the VCO. To overcome these shortcomings, methods24such of 32
as using the PLL structure [58], applying chopper stabilization [59,60], and implementing
the delta-sigma modulator with VCO [60] have been proposed.
TSIG
VSIG
RFB
VIH
VS
VSIG VIL
VREF
CS VS t
VS VT
VX
VSIG
t
CS VS
VX
VSIG
TSIG
Figure 21.
Figure 21. Operation
Operation of
of voltage-controlled
voltage-controlled oscillator.
oscillator.
Recently,efforts
Recently, effortstotoreduce
reducepower
powerconsumption
consumption through
through transistor-level
transistor-level design
design such
such as
analog front-end
as analog front-end circuits andand
circuits ADCsADCsandand
methods for lowering
methods power
for lowering consumption
power consumptionat theat
algorithm levellevel
the algorithm considering actual
considering applications
actual are studied.
applications In theIncase
are studied. theofcase
a microphone,
of a micro-
the performance
phone, and power
the performance andconsumption are lowered
power consumption when there
are lowered is no
when sound
there input
is no to the
sound in-
readout
put to thesystem,
readout andsystem,
only the input
and onlyofthe
theinput
soundofsignal is detected
the sound signalwith a low SNR.
is detected withWhen
a low
the
SNR.sound
When inputtheissound
detected, theisoperation
input detected,algorithm is programmed
the operation algorithm toisquickly return to
programmed to
the normal
quickly statetoand
return thesense
normalthe state
inputand
sound signal
sense the with
inputa sound
normalsignal
SNR. Therefore,
with a normal a method
SNR.
that achieves
Therefore, very lowthat
a method power consumption
achieves very lowinpower
an actual use environment
consumption is suggested
in an actual [61].
use environ-
ment is suggested [61].
4. Design Example
A design example of a capacitive sensor readout circuit for a MEMS microphone sys-
tem will be presented in this section, where it can be found how the design techniques are
4. Design Example
A design example of a capacitive sensor readout circuit for a MEMS microphone
system will be presented in this section, where it can be found how the design techniques
are applied to a real sensor readout system. There has been significant growth in the
demand for microphone sensors owing to the growth of voice recognition services for
various home appliances, including smartphones, and voice recognition AI speakers.
Recently, cordless earphones that employ two or more microphones for voice recognition
and calls have been developed, and the use of four to six microphones for anti-noise
cancellation (ANC) functions is gaining popularity. Hearing aids are similar to cordless
earphones, and the demand is expected to increase rapidly in the future because of the
aging population. The MEMS microphone sensor, which is smaller than other microphone
sensors, is used because the size of these two products is very important. Further, a low
noise level and low power consumption are important requirements for voice recognition,
ANC function, and battery life.
The capacitor of the capacitive MEMS microphone sensor comprises a flexible di-
aphragm made of a membrane and a fixed back plate. A high-voltage bias must be applied
between the diaphragm and the back plate to operate the sensor by generating an appro-
priate electrostatic force. A bias voltage of 5–20 V is required in general, and the sensor is
directly affected by the bias voltage, as shown in Equation (5). The bias voltage fluctuations
change in the gain of the sound being converted to voltage, and therefore, the linearity may
deteriorate. Thus, an analog front-end circuit with a high-impedance DC bias (Figure 6b)
issensor
used capacitance. However,
to apply a steady DC in this
bias tosensing method,
the sensor whilesignal degradation
sensing the change is caused by par-
in the sensor
asitic capacitance,
capacitance. However,as indicated in Equation
in this sensing method,(17). Therefore,
signal degradationsignal power by
is caused degradation
parasitic
caused by the
capacitance, parasitic capacitance
as indicated in Equation must be considered
(17). Therefore, signalinpower
the design to achieve
degradation a high
caused by
SNR.
the parasitic capacitance must be considered in the design to achieve a high SNR.
Figure
Figure22 22shows
showsaablock
blockdiagram
diagramof ofthe
thedesigned
designedoverall
overallreadout
readoutsystem.
system.The Thesensor
sensor
isisconnected
connectedtotoananESD ESDpadpadwithwitha built-in back-to-back
a built-in back-to-backdiode for high-impedance
diode for high-impedance DC bias;
DC
abias;
source follower is used as a pre-amplifier to drive the PGA, and it is
a source follower is used as a pre-amplifier to drive the PGA, and it is used as anused as an output
buffer
outputfor proper
buffer for gain
proper and low
gain andoutput impedance.
low output A charge
impedance. pumppump
A charge is used to generate
is used to gen-
the
eratebias
thevoltage required
bias voltage for membrane
required capacitive
for membrane sensorsensor
capacitive operation. It has It
operation. anhas
additional
an addi-
regulator (LDO) (LDO)
tional regulator and is placed away toaway
and is placed avoid tomutual interference
avoid mutual with the
interference readout
with block.
the readout
Nonvolatile memory
block. Nonvolatile (NVM)(NVM)
memory is usedistoused
program the DCthe
to program bias
DC and signal
bias gain, and
and signal a logic
gain, and a
circuit for NVM operation and one-wire data communication
logic circuit for NVM operation and one-wire data communication is used. is used.
VBIAS LPF Charge pump
R-DAC Oscillator
Regulator
Logic
POR BGR NVM VPROG
circuit
Regulator
Pre-amplifier Output buffer

VIN PAD Active shield
(source follower) VOUT
(PGA)
Figure 22. Block diagram of analog frontend for MEMS microphone.

Figure 22. Block diagram of analog frontend for MEMS microphone.
4.1. Design for Minimizing Signal Power Degradation
Input capacitance includes the input capacitance of the pre-amplifier and parasitic
capacitances caused by the bonding wire, input pad, and routing. It is important to mini-
mize the input capacitance of the readout system to minimize signal power degradation.
However, the flicker noise of the input transistor in the pre-amplifier is the dominant noise
4.1. Design for Minimizing Signal Power Degradation

Input capacitance includes the input capacitance of the pre-amplifier and parasitic
capacitances caused by the bonding wire, input pad, and routing. It is important to mini-
mize the input capacitance of the readout system to minimize signal power degradation.
However, the flicker noise of the input transistor in the pre-amplifier is the dominant noise
in the analog front-end circuit because the acoustic signal has a relatively low-frequency
band (10–100 kHz). It is necessary to design the area of the input transistor in the pre-
amplifier as far as possible to reduce flicker noise; therefore, minimizing the other input
capacitances (parasitic capacitance of the bonding wire, input pad, and routing) becomes
more important.
The wire should maintain its distance from other wires to minimize the parasitic
capacitance of the bonding wire. The input pad that connects the sensor and analog
front-end circuit is placed away from other pads to minimize the parasitic capacitance of
the bonding wire. The input pad needs to have lower parasitic capacitance and should
protect the ESD from the outside. Instead of the pad IP provided by the manufacturer with
the high ESD protection and a large parasitic capacitance, the following techniques were
applied to the custom-designed input pad. (1) As shown in Figure 23, the back-to-back
diode required for high impedance bias in the analog front-end circuit is arranged in the
Micromachines 2021, 12, x pad to serve the ESD diode. The size of the diode and resistor are determined to satisfy 26 of 32
the human body model (HBM) of 500 V through simulation. (2) An active shield metal is
added to minimize the parasitic capacitance in the customized input pad. (3) The routing
between the pad and the pre-amplifier is made as short as possible by placing the input
close
pad to the
close pre-amplifier;
to the further,
pre-amplifier; an an
further, active shield
active metal
shield is placed
metal around
is placed thethe
around routing. (4)
routing.
Dummy metal to resolve the design rules for density is not placed around the input
(4) Dummy metal to resolve the design rules for density is not placed around the input pad pad
andits
and itsrouting.
routing.
PAD CORE
Figure 23. Input pad circuit using back-to-back diode.

Figure 23. Input pad circuit using back-to-back diode.
Table2 2summarizes
Table summarizesthethepost-layout
post-layoutsimulation
simulationresults
resultson
onthe
theeffect
effectofofparasitic
parasiticca-
ca-
pacitance and signal degradation based on whether an active shield is applied among
pacitance and signal degradation based on whether an active shield is applied among the the
customizedpad
customized padand
andthe
thepad
padIPIPprovided
providedbybythe
themanufacturer.
manufacturer.
Table2.2.Post-layout
Table Post-layoutsimulated
simulatedcomparison
comparisonbetween
betweenPAD
PADIPIP and
and Customized
Customized PAD.
PAD.
Customized PAD
PAD IP Customized PAD
PAD IP W/O Shield WithShield
Shield
W/O Shield With
Parasitic capacitance
Parasitic capacitance
2.705 pF
2.705 pF
0.334 pF
0.334 pF
0.201 pF
0.201 pF
Signal
Signal degradation
degradation 1 1 −12.053
−12.053 dB dB −2.738
−2.738 dB dB −1.734 dBdB
−1.734
1 Signal degradation is calculated by Equation (17) with 0.9 pF of sensor capacitance.
1 Signal degradation is calculated by Equation (17) with 0.9 pF of sensor capacitance.
4.2.Design
4.2. Designfor
forNoise
NoiseReduction
Reduction
Asshown
As shownininthe
theblock
blockdiagram
diagramofofFigure
Figure22,
22,the
thepre-amplifier
pre-amplifierand
andPGAPGAareareaffected
affected
by the noise of the bandgap reference (BGR) through the reference voltage
by the noise of the bandgap reference (BGR) through the reference voltage or reference or reference
current. AA lower
current. lower noise
noiseofofthe
theBGR
BGRisisrequired
requiredtotolower
lowerthethenoise
noiseofofthe
theentire
entirereadout
readout
circuit. Further, an LPF was added to the output of the BGR to lower the noise,
circuit. Further, an LPF was added to the output of the BGR to lower the noise, as shown as shown
in
in Figure
Figure 24. The
24. The cut-off
cut-off frequency
frequency of theofLPF
themust
LPF must
be verybelow
verybecause
low because the frequency
the frequency band
band of the sound signal is low; further, a very low cut-off frequency makes the start-up
settling of the BGR very slow. The power-on reset (POR) and a switch were used to settle
when power was applied for overcoming the slow start-up settling, as shown in Figure
25. Using the noise simulation function of Synopsys’ hspice simulator, the noise level of
the BGR for the frequency range of 10–100 kHz in the post-layout was −91.38 dBV without
4.2. Design for Noise Reduction
As shown in the block diagram of Figure 22, the pre-amplifier and PGA are affected
by the noise of the bandgap reference (BGR) through the reference voltage or reference
Micromachines 2021, 12, 960 current. A lower noise of the BGR is required to lower the noise of the entire readout 25 of 30
circuit. Further, an LPF was added to the output of the BGR to lower the noise, as shown
in Figure 24. The cut-off frequency of the LPF must be very low because the frequency
band of the sound signal is low; further, a very low cut-off frequency makes the start-up
of the sound signal is low; further, a very low cut-off frequency makes the start-up settling
settling of the BGR very slow. The power-on reset (POR) and a switch were used to settle
of the BGR very slow. The power-on reset (POR) and a switch were used to settle when
when was
power power was applied
applied for overcoming
for overcoming the slow the slowsettling,
start-up start-upassettling,
shown inas Figure
shown25.in Using
Figure
25. Using the noise simulation function of Synopsys’ hspice simulator, the noise
the noise simulation function of Synopsys’ hspice simulator, the noise level of the BGR for level of
the BGR for the frequency range of 10–100 kHz in the post-layout was −91.38
the frequency range of 10–100 kHz in the post-layout was −91.38 dBV without the LPF dBV without
the LPF
and and dBV
−101.92 −101.92 dBV
with thewith
LPF. the
TheLPF. The reduction
reduction was confirmed
was confirmed to dBV.
to be 10.54 be 10.54 dBV.
VDD
Start Fast Start-Up LPF
-up
Flicker noise is necessary because the frequency range RLPFof the acoustic signal is very
low (10–100
VBGR kHz). The flicker noise of a transistor is inversely proportional to the gate
VBGR_LPF
capacitance. TheFast
size of the input transistorVBGRof the pre-amplifier
POR was designed
VBGR_LPFto be as
Start-Up
large as possible to minimize flicker noise. Further, LPF was added to the structure of the
LPF the noise transmitted from the current reference,CasLPF
pre-amplifier to reduce shown in Fig-
ure 25a. The noise level of the designed pre-amplifier and the effect of the added LPF were
confirmed by a post-layout noise simulation. The result of the noise simulation is shown
in Figure 25b, where the noise for the 10–100 kHz frequency range is significantly reduced
if LPF is present. The total noise level within the frequency range of 10 Hz–100 kHz was
−95.42 dBV without LPF and −113.00 dBV with LPF, which confirmed an 18.58 noise-re-
Figure 24. BGR circuit and fast start-up LPF.
duction effect.
Figure 24. BGR circuit and fast start-up LPF.
VNOISE [dBV]
VDD 110
CLPF 120
MP1 MP2
130
RLPF
W/O LPF
VOUT LPF 140
150
VIN MP0 IREF
160 With LPF
f
10Hz 100Hz 1kHz 10kHz 100kHz
(a) (b)
Figure
Figure25.
25.Pre-amplifier
Pre-amplifierwith
withLPF:
LPF:(a)
(a)schematic;
schematic;(b)
(b)noise
noisesimulation
simulationresult.
result.
The output
Flicker signal
noise of the pre-amplifier
is necessary because theisfrequency
connectedrange
to the of
PGA,
the which
acousticacts as anisout-
signal very
put
low (10–100 kHz). The flicker noise of a transistor is inversely proportional as
buffer. PGA uses a capacitive closed-loop multiplier with a negative gain, shown
to the gate
incapacitance.
Figure 26a. AThe multiplier composed
size of the of an input
input transistor ofcapacitor and feedback
the pre-amplifier was capacitor
designed is toused
be as
tolarge
separate the output
as possible operatingflicker
to minimize point of the pre-amplifier
noise. Further, LPFandwasthe inputtooperating
added pointof
the structure
ofthe
thepre-amplifier
PGA. As shown in Figure
to reduce the 26a,
noisethe general negative
transmitted from thegain multiplier
current follows
reference, the ref-in
as shown
erence
Figureinput;
25a. The the noise
noise level
from of the preamplifier
the and the noise
designed pre-amplifier of the
and theeffect
reference
of thevoltage
added are
LPF
included in the output. As shown in Figure 26b, the positive gain multiplier
were confirmed by a post-layout noise simulation. The result of the noise simulation is does not re-
ceive
showntheinreference
Figure 25b, voltage
where input, and therefore,
the noise the noise
for the 10–100 of the reference
kHz frequency range voltage is not
is significantly
reflected
reducedin if the
LPFoutput.
is present.TheThe
post-layout noise
total noise levelsimulation
within the of the negative
frequency rangegain
of 10and positive
Hz–100 kHz
−95.42 dBVusing
wasmultipliers
gain without
the LPF
sameand Op−Amp
113.00and
dBVsame
withBGR
LPF, reference
which confirmed
are shownan 18.58 noise-
in Figure
reduction
26c. The total effect.
noise level for the 10 Hz–100 kHz frequency range was −85.96 dBV for the
negativeThegain
output signal of
multiplier the−103.28
and pre-amplifier
dBV foris connected to gain
the positive the PGA, which it
multiplier; acts
wasas 17.32
an output
dB
buffer. PGA uses a capacitive closed-loop
lower than that with the positive gain multiplier. multiplier with a negative gain, as shown in
Figure 26a. A multiplier composed of an input capacitor and feedback capacitor is used to
separate the output operating point of the pre-amplifier and the input operating point of
the PGA. As shown in Figure 26a, the general negative gain multiplier follows the reference
input; the noise from the preamplifier and the noise of the reference voltage are included
in the output. As shown in Figure 26b, the positive gain multiplier does not receive the
reference voltage input, and therefore, the noise of the reference voltage is not reflected
in the output. The post-layout noise simulation of the negative gain and positive gain
multipliers using the same Op Amp and same BGR reference are shown in Figure 26c. The
Micromachines 2021, 12, x
total noise level for the 10 Hz–100 kHz frequency range was −85.96 dBV for the negative
28 of 32
gain multiplier and −103.28 dBV for the positive gain multiplier; it was 17.32 dB lower
than that with the positive gain multiplier.
CFB CFB
CIN CIN
VN,IN
VIN VN,AMP VN,AMP
VSIG VSIG
VREF VIN
VN,REF VN,REF
(a) (b)
VNOISE [dBV]
100
110
120
Negative gain
130
140
150
Positive gain
f
10Hz 100Hz 1kHz 10kHz 100kHz
(c)
Figure
Figure 26.
26. Noise
Noisein
inPGAs:
PGAs:(a)
(a)negative
negativegain;
gain;(b)
(b)positive
positivegain;
gain;(c)
(c)post-layout
post-layout noise
noise simulation
simulation result.
result.
4.3.
4.3.Measurement
Measurement
The
Thefabricated
fabricatedMEMSMEMSmicrophone
microphonereadout readoutcircuit
circuithadhad aa size
size of
of 1000
1000 μm µm ××10001000μm,
µm,
as
asshown
showninin Figure
Figure27.27.
Figure 28 shows
Figure the environment
28 shows the environment for measuring the readout
for measuring circuit.
the readout
An Agilent-U8903A
circuit. audio analyzer
An Agilent-U8903A was used
audio analyzer as used
was the measurement
as the measurementequipment, and theand
equipment, A-
weighting unit reflecting
the A-weighting the characteristics
unit reflecting the characteristicsof the audible
of the audiblefrequency
frequencyused usedinin sound
sound
measurement
measurement was wasapplied.
applied. Signal
Signal degradation
degradation cannot cannot be checked
be checked when thewhen the low-
low-impedance
impedance
generator ofgenerator of thedevice
the measuring measuring device
is directly is directly
connected connected
to the input oftothe the inputbecause
circuit of the
signal degradation
circuit because signal occurs in the readout
degradation occurscircuit
in theowing
readout to the parasitic
circuit owing capacitance of the
to the parasitic
input node. of
capacitance Therefore,
the inputusing
node.a discrete
Therefore, capacitor
using athat corresponds
discrete capacitorto the
thatbase capacitance
corresponds to
of 0.9
the base pFcapacitance
of the microphone
of 0.9 pF ofsensor, an environment
the microphone sensor,for anAC coupling the
environment signal
for AC of the
coupling
generator
the signal ofwastheconstructed.
generator was When measuring
constructed. the sensitivity
When measuringand theSNR of the microphone,
sensitivity and SNR of
the microphone,
the standard for measuring
the standard theforoutput of an acoustic
measuring signal
the output ofwith a size ofsignal
an acoustic 94 dB with
SPL or 1 Pa
a size
and a frequency of 1 kHz was followed [62]. Sensitivity refers to the
of 94 dB SPL or 1 Pa and a frequency of 1 kHz was followed [62]. Sensitivity refers to the size of the microphone
signal
size of for
the the sound of 94
microphone dB SPL
signal for or
the1 sound
Pa pressure;
of 94 dBtheSPLratioorof1sensitivity
Pa pressure; andthe noise level
ratio of
becomes the SNR of the microphone. From the acoustic performance
sensitivity and noise level becomes the SNR of the microphone. From the acoustic measurement result
of the microphone
performance provided byresult
measurement the sensor
of themanufacturer,
microphone the magnitude
provided ofby the the
microphone
sensor
sensor output for
manufacturer, the the 1 kHz frequency
magnitude and 94 dBSPL
of the microphone audio
sensor inputfor
output was −38
the dBV;frequency
1 kHz therefore,
the signal
and 94 dBSPLinput frominput
audio the generator
was −38 was dBV;determined
therefore, the to be −38 input
signal dBV. Measurements
from the generator were
carried out using a battery in a shield box to prevent external noise
was determined to be −38 dBV. Measurements were carried out using a battery in a shield because the noise level
to be measured is very low. Table 3 summarizes the post-layout
box to prevent external noise because the noise level to be measured is very low. Table simulation results and3
measurement results when the PGA gain values were set to
summarizes the post-layout simulation results and measurement results when the PGA0 dB and +5 dB.
gain values were set to 0 dB and +5 dB.
OUT
IN Pre
-amp.
Output VSS OUT
IN Pre buffer
Output VSS
-amp. (PGA)
buffer BGR
Regulator BGR
(PGA)
(ROIC)
Regulator
(ROIC)
POR VDD
POR VDD
μmμm
Logic
Regulator
1000
Logic
(Charge pump)
Regulator
1000
DAC
(Charge pump) DAC
NVM
Charge pump NVM
Charge pump
BIAS PROG
BIAS PROG
1000 μm
1000 μm
Figure27.
Figure 27.Micrograph
Micrographofofthe
thefabricated
fabricatedMEMS
MEMSmicrophone
microphonereadout
readoutchip.
chip.
Figure 27. Micrograph of the fabricated MEMS microphone readout chip.
Shield can
Shield can
4.2 V battery
4.2 V battery
VDD VSS
VDD VSS Analyzer
Readout Analyzer
0.9 pF Readout
circuit Audio
0.9 pF circuit Audio
IN OUT analyzer
IN OUT analyzer
Generator
Generator
Figure 28. Measurement setup of the MEMS microphone readout chip.
Table 3. Post-layout simulated and measured results of PGA.
Table 3. Post-layout simulated and measured results of PGA.
Table 3. Post-layout
Post-Layoutsimulated and measured results of PGA.
Simulation Measurement
Program Code Post-Layout Simulation
0 dB Gain 5 dB Gain 0 dB Gain Measurement 5 dB Gain
Post-Layout Simulation Measurement
Program Code
Sensitivity 0 Program
dB Gain
−38.60 dBVCode 50 dB
−34.13 Gain
dBV
dB Gain 0 dB
−39.11
5 dB Gain Gain
dBV 0 dB Gain 5 dB
−33.76 Gain
5 dBdBV
Gain
Sensitivity
Measured gain −38.60
−0.6 dBdBV
Sensitivity −34.13
+3.87
− dBV
38.60 dB
dBV −39.11 dBV
−1.11 dB −39.11 dBV
−34.13 dBV −33.76
+4.24 dBV
dB
−33.76 dBV
TotalMeasured gain
noise level (A) 1 −0.6 dBV
dB gain
Measured
−103.84 +3.87
−0.6 dB
−102.11 dB
dBV dB−1.11dBV
+3.87−106.64 dB −1.11 dB +4.24 dB
+4.24
−101.86 dBVdB
TotalSNR
noise(A)
level (A) 1 Total
−103.84 level (A) 1
noisedBV −103.84 dBV
−102.11 dBV −102.11 −106.64
dBV −106.64 dBV
dBV −101.86
−101.86 dBV
dBV
1 65.24SNRdBA(A) 1 67.97
65.24 dBA
dBA 66.08 dBA
67.97 dBA 66.08 dBA 68.23 dBA
68.23 dBA
SNR (A) 1 1
65.24 dBA 67.97
1 (A) denotes dBA
A-weighted. 66.08 dBA 68.23 dBA
(A) denotes A-weighted.
1 (A) denotes A-weighted.
5.5.Conclusions
Conclusions
5. Conclusions
Wereviewed
reviewedvarious
variousstate-of-the-art
state-of-the-arttechnologies
technologiesandandchallenges
challengestotobebeaddressed
addressedinin
We
the We reviewed
capacitive sensorvarious state-of-the-art
readout system. The technologies
types and and challenges
characteristics of to be addressed
capacitive sensorsin
the capacitive sensor readout system. The types and characteristics of capacitive sensors
the the
and capacitive sensor
structure of thereadout
basic system.circuit
readout The types
were and characteristics
discussed; the main of capacitive
issues of the sensors
capac-
and the structure of the basic readout circuit were discussed; the main issues of the
and the structure
itive sensorsensor
readoutof the basic
system, readout
SNR, and circuit
power were discussed; the main issues of the capac-
capacitive readout system, SNR, andconsumption were analyzed;
power consumption and their limits
were analyzed; and
itivelimits
and
their sensor
solutionsreadout
and were system,
wereSNR,
presented.
solutions and power
Through
presented. severalconsumption
Throughperformance were analyzed;
indicators,
several performance SNRandand
their
indicators, limits
power
SNR
andandpower
solutions
consumption werebecome
have
consumptionpresented.
more
have Through
important
become several
more than performance
others
important indicators,
in recent
than mobile
others SNR
and
in recent and
IoT power
applica-
mobile and
consumption have become more important than others in recent mobile and IoT applica-
IoT applications. The readout circuit has a very strong influence on the SNR, and the
power consumption of the entire capacitive sensor system was discussed. A circuit design
technique that can increase the signal power and/or reduce noise is important to improve
the SNR. Further, it is important to consider the signal power and noise in all processes
from circuit architecture, transistor-level design, chip floor planning, and physical layout.
It is very challenging to achieve low power while achieving a high SNR because power
consumption has a trade-off relationship with noise performance. Further, research and
development of low-power-consumption ADCs are very important because the ADC
consumes the largest amount of power in the readout system. In addition, post-processing
in digital domains has been actively conducted to contribute to high SNR, low power
consumption, and various additional functions. In conclusion, to implement a high-SNR
and low-consumption capacitive sensor readout system, innovation in each part from the
analog frontend (AFE), ADC, post-processing, and sensors is required.
Funding: This research received no external funding.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Bogue, R. Towards the trillion sensors market. Sens. Rev. 2014, 34, 137–142. [CrossRef]
2. Jayakumar, H.; Lee, K.; Lee, W.S.; Raha, A.; Kim, Y.; Raghunathan, V. Powering the internet of things. In Proceedings of the 2014
International Symposium on Low Power Electronics and Design, La Jolla, CA, USA, 11–13 August 2014; ACM: New York, NY,
USA, 2014; Volume 2, pp. 375–380.
3. Winncy, Y.D. Resistive, Capacitive, Inductive, and Magnetic Sensor Technologies; CRC Press: Boca Raton, FL, USA, 2014;
ISBN 9780367864651.
4. Tsuchiya, T. Technologies, applications, and reliabilities of microelectromechanical systems (MEMS). In Proceedings of the 9th
SEGJ International Symposium, Sapporo, Japan, 12–14 October 2009.
5. Rao, K.; Liu, H.; Wei, X.; Wu, W.; Hu, C.; Fan, J.; Liu, J.; Tu, L. A high-resolution area-change-based capacitive MEMS tilt sensor.
Sens. Actuators A Phys. 2020, 313, 112191. [CrossRef]
6. Elko, G.W.; Pardo, F.; López, D.; Bishop, D.; Gammel, P. Capacitive MEMS microphones. Bell Syst. Tech. J. 2005, 10, 187–198.
[CrossRef]
7. Mishra, M.K.; Dubey, V.; Mishra, P.M.; Khan, I. MEMS Technology: A Review. J. Eng. Res. Rep. 2019, 1–24. [CrossRef]
8. Analog Devices Which ADC Architecture Is Right for Your Application? Available online: https://www.analog.com/en/analog-
dialogue/articles/the-right-adc-architecture.html# (accessed on 16 June 2021).
9. Ko, S.H.; Yang, B.D. An Ultra-Compact Low Power Self-Capacitive Touch Screen Readout IC Embedding Reconfigurable Noise
Immunity and Current-Driven Capacitance Compensation. IEEE Trans. Circuits Syst. II Express Briefs 2018, 66, 1321–1325.
[CrossRef]
10. Ouh, H.K.; Choi, J.; Lee, J.; Han, S.; Kim, S.; Seo, J.; Lim, K.; Ko, H. Sub-fF trimmable readout circuit for tri-axes capacitive
microaccelerometers. In Proceedings of the 2012 IEEE Sensors, Taipei, Taiwan, 28–31 October 2012; pp. 1–4.
11. Shin, D.-Y.; Lee, H.; Kim, S. A Delta–Sigma Interface Circuit for Capacitive Sensors with an Automatically Calibrated Zero Point.
IEEE Trans. Circuits Syst. II Express Briefs 2011, 58, 90–94. [CrossRef]
12. Song, H.; Park, Y.; Kim, H.; Cho, D.-I.D.; Ko, H. Fully Integrated Low-Noise Readout Circuit with Automatic Offset Cancellation
Loop for Capacitive Microsensors. Sensors 2015, 15, 26009–26017. [CrossRef]
13. Lee, C.-J.; Park, J.K.; Piao, C.; Seo, H.-E.; Choi, J.; Chun, J.-H. Mutual Capacitive Sensing Touch Screen Controller for Ultrathin
Display with Extended Signal Passband Using Negative Capacitance. Sensors 2018, 18, 3637. [CrossRef]
14. Wang, S.; Koickal, T.J.; Hamilton, A.; Mastropaolo, E.; Cheung, R.; Abel, A.; Smith, L.S.; Wang, L. A Power-Efficient Capacitive
Read-Out Circuit With Parasitic-Cancellation for MEMS Cochlea Sensors. IEEE Trans. Biomed. Circuits Syst. 2016, 10, 25–37.
[CrossRef] [PubMed]
15. Sisinni, E.; Depari, A.; Flammini, A.; Ferri, G.; Stornelli, V.; Barile, G. Full-Analog Parasitic Capacitance Compensation for
AC-Excited Differential Sensors. IEEE Trans. Instrum. Meas. 2020, 69, 5890–5899. [CrossRef]
16. Ciccarella, P.; Carminati, M.; Sampietro, M.; Ferrari, G. Multichannel 65 zF rms Resolution CMOS Monolithic Capacitive Sensor
for Counting Single Micrometer-Sized Airborne Particles on Chip. IEEE J. Solid-State Circuits 2016, 51, 2545–2553. [CrossRef]
17. Sanchez-Chiva, J.M.; Michalik, P.; Fernandez, D.; Madrenas, J. A CMOS BEOL accelerometer low-noise readout amplifier with 4.2
√
zF/ Hz total noise floor. In Proceedings of the 2015 IEEE Sensors, Busan, Korea, 1–4 November 2015; IEEE: Busan, Korea, 2015;
pp. 1–4.
18. Huh, Y.; Hong, S.-W.; Park, S.-H.; Shin, C.; Bang, J.-S.; Park, C.; Park, S.; Cho, G.-H. A 10.1” 183- µW/electrode, 0.73-mm2 /sensor
High-SNR 3-D Hover Sensor Based on Enhanced Signal Refining and Fine Error Calibrating Techniques. IEEE J. Solid State
Circuits 2017, 53, 1079–1088. [CrossRef]
19. Zhong, L.; Lai, X.; Xu, D. Oversampling Successive Approximation Technique for MEMS Differential Capacitive Sensor. IEEE J.
Solid State Circuits 2018, 53, 2240–2251. [CrossRef]
20. Zhong, L.; Lai, X.; Song, H.; Xu, D. Differential Capacitive Readout Circuit Using Oversampling Successive Approximation
Technique. IEEE Trans. Circuits Syst. I Regul. Pap. 2018, 65, 4072–4085. [CrossRef]
21. Wang, Y.M.; Chan, P.K.; Li, H.; Ong, S.-E. A Low-Power Highly Sensitive Capacitive Accelerometer IC Using Auto-Zero
Time-Multiplexed Differential Technique. IEEE Sens. J. 2015, 15, 6179–6191. [CrossRef]
22. Yeo, D.-H.; Kim, S.-H.; Noh, H.-K.; Kim, B.; Sim, J.-Y.; Park, H.-J. A SNR-Enhanced Mutual-Capacitive Touch-Sensor ROIC using
an Averaging with Three Specific TX Frequencies, a Noise Memory, and a Compact Delay Compensation Circuit. IEEE Sens. J.
2016, 16, 6931–69381. [CrossRef]
23. Nicollini, G.; Devecchi, D. MEMS Capacitive Microphones: Acoustical, Electrical, and Hidden Thermal-Related Issues. IEEE Sens.
J. 2018, 18, 5386–5394. [CrossRef]
24. Zhao, Y.; Zhao, J.; Wang, X.; Xia, G.M.; Shi, Q.; Qiu, A.P.; Xu, Y.P. A Sub-0.1◦ /h Bias-Instability Split-Mode MEMS Gyroscope
With CMOS Readout Circuit. IEEE J. Solid-State Circuits 2018, 53, 2636–2650. [CrossRef]
25. Pérez Sanjurjo, J.; Prefasi, E.; Buffa, C.; Gaggl, R. A Capacitance-To-Digital Converter for MEMS Sensors for Smart Applications.
Sensors 2017, 17, 1312. [CrossRef] [PubMed]
26. Heinssen, S.; Hellwege, N.; Heidmann, N.; Paul, S.; Peters-Drolshagen, D. Robust digital calibration engine for MEMS inertial
sensor systems. In Proceedings of the 2015 IEEE Sensors, Busan, Korea, 1–4 November 2015; IEEE: Busan, Korea, 2015; pp. 1–4.
27. Park, J.-E.; Park, J.; Hwang, Y.-H.; Oh, J.; Jeong, D.-K. A Noise-Immunity-Enhanced Analog Front-End for 36 × 64 Touch-Screen
Controllers With 20-VPP Noise Tolerance at 100 kHz. IEEE J. Solid-State Circuits 2019, 54, 1497–1510. [CrossRef]
28. Park, J.-E.; Lim, D.-H.; Jeong, D.-K. A Reconfigurable 40-to-67 dB SNR, 50-to-6400 Hz Frame-Rate, Column-Parallel Readout IC
for Capacitive Touch-Screen Panels. IEEE J. Solid-State Circuits 2014, 49, 2305–2318. [CrossRef]
29. Petkov, V.P.; Balachandran, G.K.; Beintner, J. A Fully Differential Charge-Balanced Accelerometer for Electronic Stability Control.
IEEE J. Solid-State Circuits 2014, 49, 262–270. [CrossRef]
30. An, J.-S.; Jung, S.-J.; Hong, S.-K.; Kwon, O.-K. A Highly Noise-Immune Capacitive Touch Sensing System Using an Adaptive
Chopper Stabilization Method. IEEE Sens. J. 2017, 17, 803–811. [CrossRef]
31. Lee, J.-S.; Yeo, D.-H.; Kwon, H.-J.; Kim, B.; Sim, J.-Y.; Park, H.-J. An LCD-VCOM-Noise Resilient Mutual-Capacitive Touch-Sensor
IC Chip With a Low-Voltage Driving Signal. IEEE Sens. J. 2015, 15, 4595–4602. [CrossRef]
32. An, J.-S.; Han, S.-H.; Kim, J.E.; Yoon, D.-H.; Kim, Y.-H.; Hong, H.-H.; Ye, J.-H.; Jung, S.-J.; Lee, S.-H.; Jeong, J.-Y.; et al. A
3.9-kHz Frame Rate and 61.0-dB SNR Analog Front-End IC with 6-bit Pressure and Tilt Angle Expressions of Active Stylus Using
Multiple-Frequency Driving Method for Capacitive Touch Screen Panels. IEEE J. Solid-State Circuits 2018, 53, 187–203. [CrossRef]
33. Kim, S.; Rikan, B.S.; Pu, Y.; Yoo, S.-S.; Lee, M.; Hwang, K.C.; Yang, Y.; Lee, K.-Y. A High Noise Immunity, 28 × 16-Channel Finger
Touch Sensing IC Using OFDM and Frequency Translation Technique. Sensors 2018, 18, 1652. [CrossRef]
34. Luo, Y.; Heng, C. An 8.2 µw 0.14 mm 2 16-channel CDMA-like period modulation capacitance-to-digital converter with reduced
data throughput. In Proceedings of the 2018 IEEE Symposium on VLSI Circuits, Honolulu, HI, USA, 18–22 June 2018; IEEE:
Honolulu, HI, USA, 2018; pp. 165–166.
35. Jang, H.; Shin, H.; Ko, S.; Yun, I.; Lee, K. 2D Coded-aperture-based ultra-compact capacitive touch-screen controller with 40
reconfigurable channels. In Proceedings of the 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers
(ISSCC), San Francisco, CA, USA, 9–13 February 2014; IEEE: San Francisco, CA, USA, 2014; Volume 57, pp. 218–219.
36. Heo, S.; Ma, H.; Song, J.; Park, K.; Choi, E.-H.; Kim, J.J.; Bien, F. 72 dB SNR, 240 Hz Frame Rate Readout IC With Differential
Continuous-Mode Parallel Architecture for Larger Touch-Screen Panel Applications. IEEE Trans. Circuits Syst. I Regul. Pap. 2016,
63, 960–971. [CrossRef]
37. Park, C.; Park, S.S.; Kim, K.D.; Park, S.S.; Park, J.; Huh, Y.; Kang, B.; Cho, G.H. A pen-pressure-sensitive capacitive touch system
using electrically coupled resonance pen. Dig. Tech. Pap. IEEE Int. Solid-State Circuits Conf. 2015, 58, 124–125. [CrossRef]
38. Heo, S.; Park, K.; Choi, E.-H.; Bien, F. Differential Coded Multiple Signaling Method With Fully Differential Receiver for Mutual
Capacitive Fingerprint TSP. IEEE Trans. Circuits Syst. I Regul. Pap. 2020, 67, 74–85. [CrossRef]
39. Ko, S.H. A -24 dB in-Band Noise-Immunity Mutual Capacitance Readout System for Variable Refresh Rate of Active-Matrix
OLED Display. IEEE Trans. Circuits Syst. I Regul. Pap. 2020, 67, 1126–1135. [CrossRef]
40. Kim, P.; Han, S.; Jung, Y.; Lee, S. A PAPR Reduction Technique for Fast Touch Sensors Adopting a Multiple Frequency Driving
Method on Large Display Panels. Sensors 2021, 21, 429. [CrossRef] [PubMed]
41. Ko, S.H. A Mutual-Capacitance 3-D Gesture Sensing System Based on 2-D Orthogonal Encoding and an Un-Folding Coordinate
Interpolation Schemes. IEEE Sens. J. 2020, 20, 4778–4786. [CrossRef]
42. Miyamoto, M.; Hamaguchi, M.; Nagao, A. A 143 × 81 Mutual-Capacitance Touch-Sensing Analog Front-End With Parallel Drive
and Differential Sensing Architecture. IEEE J. Solid-State Circuits 2015, 50, 335–343. [CrossRef]
43. Zhong, L.; Lai, X.; Xu, D.; Liao, X.; Yang, C.; Fang, Z.; Zheng, Y. Capacitive Touch Panel With Low Sensitivity to Water Drop
Employing Mutual-Coupling Electrical Field Shaping Technique. IEEE Trans. Circuits Syst. I Regul. Pap. 2019, 66, 1393–1404.
[CrossRef]
44. Huang, H.-M.; YU, H.C.; Chih-Hung, K.; Huang, S.-H. Adsorption of Cationic Dyes onto Mesoporous Silica. US9280243B2,
1 May 2011.
45. Ha, H.; Sylvester, D.; Blaauw, D.; Sim, J.Y. A 160nW 63.9fJ/conversion-step capacitance-to-digital converter for ultra-low-power
wireless sensor nodes. Dig. Tech. Pap. IEEE Int. Solid-State Circuits Conf. 2014, 57, 220–221. [CrossRef]
46. Park, J.; Hwang, Y.-H.; Oh, J.; Song, Y.; Park, J.-E.; Jeong, D.-K. A Mutual Capacitance Touch Readout IC With 64% Reduced-Power
Adiabatic Driving Over Heavily Coupled Touch Screen. IEEE J. Solid-State Circuits 2019, 54, 1694–1704. [CrossRef]
47. Oh, S.; Lee, Y.; Wang, J.; Foo, Z.; Kim, Y.; Jung, W.; Li, Z.; Blaauw, D.; Sylvester, D. A Dual-Slope Capacitance-to-Digital Converter
Integrated in an Implantable Pressure-Sensing System. IEEE J. Solid-State Circuits 2015, 50, 1581–1591. [CrossRef]
48. Kim, H.; Lee, B.; Mun, Y.; Kim, J.; Han, K.; Roh, Y.; Song, D.; Huh, S.; Ko, H. Reconfigurable Sensor Analog Front-End Using
Low-Noise Chopper-Stabilized Delta-Sigma Capacitance-to-Digital Converter. Micromachines 2018, 9, 347. [CrossRef]
49. Amini, S.; Johns, D.A. A Flexible Charge-Balanced Ratiometric Open-Loop Readout System for Capacitive Inertial Sensors. IEEE
Trans. Circuits Syst. II Express Briefs 2015, 62, 317–321. [CrossRef]
50. Omran, H.; Alhoshany, A.; Alahmadi, H.; Salama, K.N. A 35fJ/Step differential successive approximation capacitive sensor
readout circuit with quasi-dynamic operation. IEEE Symp. VLSI Circuits Dig. Tech. Pap. 2016, 2016, 1–2. [CrossRef]
51. Omran, H.; Alhoshany, A.; Alahmadi, H.; Salama, K.N. A 33fJ/Step SAR Capacitance-to-Digital Converter Using a Chain of
Inverter-Based Amplifiers. IEEE Trans. Circuits Syst. I Regul. Pap. 2017, 64, 310–321. [CrossRef]
52. Liu, B.; Hoseini, Z.; Lee, K.-S.; Lee, Y.-M. On-Chip Touch Sensor Readout Circuit Using Passive Sigma-Delta Modulator
Capacitance-to-Digital Converter. IEEE Sens. J. 2015, 15, 3893–3902. [CrossRef]
53. Xin, H.; Pelzers, K.; Baltus, P.; Cantatore, E.; Harpe, P. A 4.3 fJ/conversion-step 6440 µm 2 all-dynamic capacitance-to-digital
converter with energy-efficient charge reuse. In Proceedings of the 2020 IEEE Symposium on VLSI Circuits, Honolulu, HI, USA,
14–19 June 2020; IEEE: Honolulu, HI, USA, 2020; Volume 2020, pp. 1–2.
54. Hussaini, S.; Jiang, H.; Walsh, P.; MacSweeney, D.; Makinwa, K.A.A. A 15-nW per Sensor Interference-Immune Readout IC for
Capacitive Touch Sensors. IEEE J. Solid-State Circuits 2019, 54, 1874–1882. [CrossRef]
55. Zhong, L.; Yang, J.; Xu, D.; Lai, X. Bandwidth-Enhanced Oversampling Successive Approximation Readout Technique for
Low-Noise Power-Efficient MEMS Capacitive Accelerometer. IEEE J. Solid-State Circuits 2020, 55, 2529–2538. [CrossRef]
56. Duan, Q.; Wang, Z.; Roh, J. A 7.5-µW 0.08-mm 2 single-ended SC delta-sigma ADC for acoustic sensor applications. Int. J. Circuit
Theory Appl. 2015, 44, 1173–1185. [CrossRef]
57. Quintero, A.; Cardes, F.; Perez, C.; Buffa, C.; Wiesbauer, A.; Hernandez, L. A VCO-Based CMOS Readout Circuit for Capacitive
MEMS Microphones. Sensors 2019, 19, 4126. [CrossRef] [PubMed]
58. Li, S.; Zhao, W.; Xu, B.; Yang, X.; Tang, X.; Shen, L.; Lu, N.; Pan, D.Z.; Sun, N. A 0.025-mm 2 0.8-V 78.5dB-SNDR VCO-based
sensor readout circuit in a hybrid PLL-∆ΣM structure. In Proceedings of the 2019 IEEE Custom Integrated Circuits Conference
(CICC), Austin, TX, USA, 14–17 April 2019; IEEE: Austin, TX, USA, 2019; Volume 2019, pp. 1–3.
59. Quintero, A.; Cardes, F.; Hernandez, L.; Buffa, C.; Wiesbauer, A. A capacitance-to-digital converter based on a ring oscillator with
flicker noise reduction. In Proceedings of the 2016 Austrochip Workshop on Microelectronics (Austrochip), Villach, Austria, 19
October 2016; IEEE: Villach, Austria, 2016; pp. 40–44.
60. Tu, C.C.; Wang, Y.K.; Lin, T.H. A 0.06 mm2 ± 50 mV range-82 dB THD chopper VCO-based sensor readout circuit in 40 nm
CMOS. In Proceedings of the IEEE Symposium on VLSI Circuits, Digest of Technical Papers, Kyoto, Japan, 5–8 June 2017; IEEE:
Kyoto, Japan, 2017; pp. C84–C85.
61. Yang, Y.; Lee, B.; Cho, J.S.; Kim, S.; Lee, H. A Digital Capacitive MEMS Microphone for Speech Recognition With Fast Wake-Up
Feature Using a Sound Activity Detector. IEEE Trans. Circuits Syst. II Express Briefs 2020, 67, 1509–1513. [CrossRef]
62. Brüel&Kjær. Microphone Handbook Vol. 1: Theory; Brüel&Kjær: Naerum, Denmark, 2019.

Readout Circuits For Capacitive Sensors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Readout Circuits For Capacitive Sensors

Uploaded by

Copyright:

Available Formats

micromachines

Department of Electronic Engineering, Hanyang University, Seoul 04763, Korea; yongsangyoo@hanyang.ac.kr

Citation: Yoo, Y.; Choi, B.-D.

Micromachines 2021, 12, 960. https://doi.org/10.3390/mi12080960 https://www.mdpi.com/journal/micromachines

Micromachines 2021, 12, 960 2 of 30

Micromachines 2021, 12, 960 3 of 30

Micromachines 2021, 12, 960 4 of 30

For the self-capacitive

Further, the voltage at node NIN changes as

In the case of a MEMS microphone, an appropriate elastic force must be generated

2.2.2. Mutual-Capacitance Readout Circuit

Sensor CFB VTX ΔVTX

2.3. Analog-to-Digital Converters (ADC)

10 100 1k 10k 100k 1M 10M 100M 1G

3. Technical Issues of the Capacitive Sensor Readout System

Sensor Application Total

3.1. Signal-to-Noise Ratio (SNR): Signal Power

∆QSIG = (CP + CS0 + ∆CS )∆VS (15)

Figure 9. Base and parasitic capacitances in a self-capacitance readout circuit.

sensor decreases, which results in a decrease in signal power.

For sensors that require

For sensors that require=the CS 0 V TX +used

+ C(S0Δ)V∆V − ΔVifS )the accurate negative polarity

3.1.3. Charge-Balancing Compensation: Using a Differential Sensor

The compensation method (23),using the sensorindescribed above change

Further, ΔVERROR can be calculated as the capacitance measurement error of the

3.2. Signal-to-Noise Ratio (SNR): Internal Noise

Point Point Point

3.3. Signal-to-Noise Ratio (SNR): External Noise

3.3.1. Fully Differential Structure and Chopper Stabilization

3.3.2. Averaging, Integration, and Parallel Driving Method

Micromachines 2021, 12, x 19 of 32

3.3.2. Averaging, Integration, and Parallel Driving Method

3.3.3. Frequency Adaption

VBIAS LPF Charge pump

Pre-amplifier Output buffer

Figure 22. Block diagram of analog frontend for MEMS microphone.

4.1. Design for Minimizing Signal Power Degradation

Figure 23. Input pad circuit using back-to-back diode.

Funding: This research received no external funding.

You might also like