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

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

A Low-Power On-chip ECG Monitoring System

Based on MWCNT/PDMS Dry Electrodes
Nishat T. Tasneem, Student Member, IEEE, Salvatore A. Pullano, Member, IEEE, Costantino D. Critello,
Antonino S. Fiorillo, Senior Member, IEEE, Ifana Mahbub, Member, IEEE

Abstract—Wearable technologies have established novel strate- impedance is higher for the dry electrodes than the wet
gies for continuous electrocardiogram (ECG) monitoring, ex- ones, recording ECG signals may require additional applied
ploiting recent developments in sensors and integrated circuit pressure. It may also cause discomfort to the patient when
technology. Advances in dry electrodes fabrication improved the
growth of wearable devices. Herein, we present an on-chip inte- high pressure is applied to rigid dry electrodes. Different
grated ECG signal acquisition system along with dry electrodes types of dry electrodes such as microneedle arrays (MNA)
for wearable and long-term monitoring. Custom electrodes are [15], textile electrodes on woven fabrics [16], non-woven
fabricated using multi-walled carbon nanotube (MWCNT) and fabrics [17], and soft-polymer based electrodes [18]–[22] have
Polydimethylsiloxane (PDMS) composite on a flexible substrate. A been recently proposed. These electrodes typically have a
1.5 µW fully-differential operational transconductance amplifier
(OTA) with a capacitive-resistive feedback network is designed high skin-electrode impedance because of the weak contact
in 130 nm CMOS process achieving a mid-band gain of 43.09 dB with the skin. In order to improve the long-term monitoring
and an input-referred noise of 2.81 µVrms in the range from 0.175 performances, soft conductive polymer such as polyvinylidene
Hz to 1.636 kHz. Reconfigurability of the lower cut-off frequency fluoride (PVDF) and polydimethylsiloxane (PDMS) based
is achieved by controlling the bias voltage of the triple-well nMOS flexible dry electrodes are preferred [23]–[25]. As a potential
transistors acting as the feedback pseudoresistor. Electrodes
with different surface area and MWCNT concentrations were polymer-based dry electrode, carbon nanotube (CNT) and
investigated in a real-time ECG measurement scenario. The PDMS based composite was investigated for its high elec-
recorded signals achieve a high signal-to-noise ratio which spans trical conductivity [8], [26]–[28]. This work proposes multi-
from 35.7 dB to 38.6 dB with the increasing MWCNT con- walled carbon nanotube (MWCNT) and PDMS composite
centration. Experimental results demonstrate that the proposed as the sensor layer for the ECG recording system. Kapton
system allows for high performance and low-power ECG signal
recording for the long-term and wearable applications. (polyimide film) is employed as the flexible substrate for its
broad temperature stability for the proposed dry electrode [29].
Index Terms—Flexible and wearable dry electrodes, Different geometries and MWCNT/PDMS compositions are
MWCNT/PDMS, Kapton substrate, ECG signal recording,
two-stage differential amplifier. investigated to improve the signal-to-noise ratio (SNR) and
thus the accuracy of the CV assessment.
The front-end amplifier for ECG signal acquisition is one
I. I NTRODUCTION of the substantial elements in the ECG recording system [30],
[31]. It requires a very low power consumption (< 4µW )
E LECTROCARDIOGRAM (ECG) has been used as a
diagnostic tool since its discovery, becoming an ordinary
procedure now-a-days in clinical settings [1]–[3]. Long-term
for long-term monitoring and wearability. As the amplitude
of the ECG signal can be up to 5 mV with a bandwidth of
ECG monitoring allows for a more accurate and manage- 0.1 Hz to 150 Hz [32], the signal is prone to low-frequency
able cardiovascular risk (CV) assessment, the detection and Flicker (1/f) noise and artifacts, imposing design constraints in
analysis of cardiac events such as arrhythmia, sudden cardiac terms of the input-referred noise and power consumption [31].
arrest (SCA), atrial fibrillation, etc. [4]–[6]. In this context, Previous works reported ECG amplifiers with a low gain and
wearable devices have been rapidly developed and moved high power consumption, that may not be suitable for portable
from the prototype stage to a more established consumer ECG monitoring systems [33], [34]. In this work, a two-
and clinical trial device stage [7]. Electrodes are one of stage operational transconductance amplifier (OTA) topology
the core components in the acquisition of biopotentials [8]. is adopted for its simple architecture which includes DC-
Standard wet electrodes are extensively used for ECG signal blocking capacitors (to reject large DC offset voltage at the
acquisition [9]–[11]. However, problems arising from the long- electrode-skin interface) and two diode-connected triple-well
term instability and skin irritations along with the prohibitive nMOS-bipolar transistors as the pseudo-resistors for obtaining
movement artifacts have led to the development of alternative sub-Hz lower cut-off frequency [35]. The performance of
technologies such as dry electrodes, allowing better stability the proposed electrodes and the recording system was also
over days or weeks without signal degradation and patient compared with the commercially available ECG electrodes.
discomfort [7], [8], [12]–[14]. Since the skin-electrode contact The novelty of the proposed work is the integration of a low-
power, and low-noise application specific integrated circuit
Nishat T. Tasneem, and Ifana Mahbub are with the Department of Electrical (ASIC) based ECG signal acquisition amplifier with a flexible,
Engineering, University of North Texas, Denton, TX 76207 USA (email: thin-film, highly conductive dry electrode that make the system
nishattarannumtasneem@my.unt.edu, ifana.mahbub@unt.edu). fully wearable and miniaturized achieving a high signal to
Salvatore A. Pullano, Costantino D. Critello, and Antonino S. Fiorillo
are with the Department of Health Sciences, University Magna Graecia of noise ratio (SNR). Though active CMOS based biosensor
Catanzaro, Catanzaro, Italy. array-based research have been carried out previously [36], to

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

Fig. 2: Fabricated MWCNT/PDMS dry electrodes having a width

ranging from 3 to 5 mm and a length ranging from 40 to 60 mm.

TABLE I
Dimensions of The Designed Electrodes

Concentration (wt%) Dimensions (mm2 )

Fig. 1: Schematic of the fabricated dry electrode design and the layers 4, 8 and 12 60×3 50×4 40×5
composition.

the best of authors’ knowledge, this work is the first prototype

that demonstrates the integration of dry electrodes with a and 12%) and the dimensions of the fabricated electrodes are
custom-designed IC for ECG signal recording. The rest of the reported in Table I. The fabricated electrodes are shown in Fig.
paper is organized as follows: Section II presents the detailed 2, on the flexible Kapton substrate, with the MWCNT/PDMS
fabrication process of the dry electrodes, followed by the composite and the exposed Ag contact. The Ag is used as the
amplifier design architecture in section III. The measurement contact for the external wire-bonding of the electrode with the
results of the electrode and the front-end amplifier are pre- chip.
sented in section IV. Concluding remarks are then presented
in section V. III. A MPLIFIER D ESIGN A RCHITECTURE
A fully differential amplifier is designed in 130 nm standard
II. D RY E LECTRODE FABRICATION P ROCEDURE CMOS process with 7 metal layers and 1 poly layer. The
The fabrication of the proposed dry electrode realizes the schematic of the proposed OTA is shown in Fig. 3. A two-
MWCNT/PDMS composite as the sensing layer. Kapton is stage topology with the common-mode feedback (CMFB) is
used as the substrate due to its hydrophilicity, chemical implemented in this work [37]. A two-stage OTA maximizes
stability and the wide temperature range, which ensure its the output swing through the second stage as well as increases
stability during the fabrication process (Fig. 1) [29]. In this the open-loop gain [38]. A capacitive-resistive (C2 − R2 )
work, a thin film of Kapton (25 µm in thickness) is used feedback network is chosen as the feedback in the closed-
as the flexible substrate. Through a 3D printed stencil mask, loop system. The mid-band gain of the amplifier is set by the
Silver (Ag) electrode is evaporated on the Kapton substrate by C1 − C2 capacitors as follows:
e-beam physical vapor deposition (Varian technologies, USA) C1
at a pressure of 5 × 10−6 P a with a deposition rate of 1.0 A0 = (1)
C2
nm/s, resulting in a Ag layer of 250 nm thickness. MWCNTs
The transistor biasing current, Iin is set to be 1.25 µA
with a purity of > 95% (US Researchers Nanomaterial, Inc.,
keeping the total power consumption to be less than 1.5 µW
Houston, USA) is used as a nano-filler material with an inner
with 1.2 V supply voltage. Any mismatch in the skin-electrode
and outer diameter of 5-10 nm and 10-20 nm respectively.
impedance, which results in a large DC offset voltage is
The MWCNT also achieves an electrical conductivity of 100
reduced by the two DC-blocking capacitors (C1 ) [39]. The
S.cm−1 , a tap density of 0.22 gm.cm−3 , and a true density
lower and higher cut-off frequencies of the amplifier are set
of 2.1 gm.cm−3 . Due to its tunable mechanical properties
by the feedback network and the load capacitor, CL using the
and biocompatibility, PDMS (SylgardTM 184, Dowsil, Mid-
following equations [40].
land, MI, USA) is preferred as the polymeric matrix in the
composite fabrication [29]. Firstly, MWCNT is dispersed in 1
fL = (2)
Isopropyl Alcohol (IPA) solution with 10:1 weight ratio. Then 2πR2 C2
the mixture is stirred for 15 minutes and sonicated for 20 min- gm
fH = (3)
utes (40 kHz, 100 W) for temporarily reducing the nanotube A0 CL
aggregation. Then, the elastomer is added, mixed, and again The CL is chosen to be 2 pF to have a higher cut-off fre-
sonicated for 20 minutes. The Alcohol content in the mixture quency >1 kHz in this work. Here, gm is the transconductance
is then reduced by evaporation for 6 hours at 70 ◦ C. After that, of the input pair transistors M1 and M2 . R2 represents the
the curing agent is added to the composite at 10:1 ratio, and very large pseudo-resistor to keep the lower cut-off frequency
the mixture is vigorously stirred for 15 minutes. Finally, the <1 Hz, to record the sub-Hz frequency ECG signal [32]. R2
solution is spin-coated on the Kapton substrate (2000 rpm/min, is implemented by the two diode-connected triple-well nMOS
60 s) and cured at 80 ◦ C for 2 hours. A total of 9 electrodes are transistors connected in series as shown in Fig. 3. By using
fabricated with three different MWCNT concentrations (4, 8, the triple-well configuration, a reduction in the input-referred

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

Fig. 3: Schematic of the proposed OTA and inset with the designed circuit.

noise is achieved. The body of the nMOS is kept isolated from

the p-substrate by having a deep n-well [41].
The equivalent resistor can be designed to achieve the
GΩ range in the diode-connected configuration. The nMOS
transistors operate in the sub-threshold region as the drain
to source voltage Vds of the transistor is very low. The
relationship between the sub-threshold drain current (ID ) and
vDS can be presented as below [42]:
dID I0 VGS − Vth − Vof f
= exp ( ) (4)
dvDS VT n.VT Fig. 4: (a) Incremental resistance for different bias voltages, (b)
incremental resistance at ∆V = 0.6 V.
W
I0 ∝ (5) input transistors are designed to have an aspect ratio of 100/0.5
L
µm/µm since the noise varies inversely with the gate area of
where Vth is the threshold voltage of the nMOS, Vof f is the input pairs. The nMOS current mirrors M11 , M12 , M13 ,
the offset voltage when VGS = 0, and n is the subthreshold and M14 , and the pMOS current mirrors M5 , M6 , and M7 are
slope factor. As can be seen from eqn. (4), as the gate biased to operate in the strong inversion region.
voltage of the nMOS (Vbias in Fig. 3) increases, the change A resistor, Rcm of 250 kΩ and a capacitor, Ccm of 500
in the drain current increases and eventually the value of the fF are connected to the output nodes of the amplifier from a
pseudoresistor decreases. From eqn. (5), it is evident that the CMFB circuit to maintain the common-mode voltage, Vcm as
resistance of the pseudoresistor also depends on the W/L 0.6 V. This loop rejects the common-mode signal components
ratio of the transistors. The W/L ratio of the triple-well and increases the CMRR (common-mode rejection ratio) of
transistors are chosen to be 4 µm/0.6 µm to achieve the the OTA [43]. A compensation resistor, RC of 25 kΩ and a
maximum impedance by biasing the transistor in the weak capacitor, CC of 1 pF are implemented to compensate for the
inversion region. The simulated incremental resistance value oscillation caused by the positive feedback.
is calculated as ∆V /∆I and is shown in Fig. 4 (a) for different
gate voltages: 0.6, 0.9 and 1.2 V. The resistance decreases as
IV. E XPERIMENTAL R ESULTS
the current through the nMOS transistors increases as a result
of increase in Vbias (4 (b)). Thus the incremental resistance, A. Dry Electrode Characterization
and eventually the lower cut-off frequency of the amplifier Dry electrode’s electrical characterization is investigated
can be controlled by changing the Vbias . The re-configurable using the Impedance analyzer VersaSTAT 4 (Ametek Scientific
sub-Hz lower cut-off frequency allows the amplifier to record Instrument, TN, USA). The impedance analyzer is used in
ECG signals in different low-frequency bandwidths. the potentiostatic mode (i.e., water-based solution) with 10
In order to reduce the 1/f noise component in the acquired mV sinusoidal signal with the frequency range of 0.1 Hz - 1
ECG signals, two pMOS transistors operating in the weak kHz applied between the working and the Ag/AgCl standard
inversion region are used as the input pair of the amplifier reference electrode. In analogy with the wet electrodes, the
[4]. The objective for doing this is to achieve a high gain by equivalent electrical model of the skin-electrode interface can
maximizing the transconductance of the transistors. Also, the be modeled as a series resistance, which is usually not crucial

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

TABLE II
Evaluation of Equivalent Circuit Parameters the MWCNT concentration increases, as shown in Table II.
The representative conductivity data are collected from the 50
Concentration Rp (kΩ) Cp (nF) Conductivity (mS/cm) × 4 mm2 electrodes.
12 wt% 117 4 685 Raman analysis is widely used qualitatively and quantita-
8 wt% 187 5.8 515 tively for the characterization of rubbers in order to evaluate
4 wt% 304 6.2 481 the transition from the pristine form (characterized by low
electrical conductivity) to the compounded form with mineral
fillers, characterized by higher electrical conductivity [47],
in terms of the skin impedance and depends on the type of [48]. Raman spectra of MWCNT/PDMS for various filler con-
the electrode material. In the absence of the conductive gel, centration depicts the disorder band (D-peak at 1447 cm−1 ),
dry electrode coupling can be considered as capacitive, while the graphite band (G-peak at 1510 cm−1 ) and the second order
the skin humidity and sweat cause resistance between the D band (2D-peak at 1550 cm−1 ), as shown in Fig. 7. The
electrode and the skin. Moreover, the electrolyte-like function quantity of structural defects can be evaluated from the ID /IG
of the skin humidity/sweat can be modeled as a half-cell ratio, while the I2D /ID ratio takes into account the overall
potential [41], [44]. Fig. 5(a) and (b) show the measured crystalline quality of the material [49]. The I2D /IG ratio,
parallel resistance, Rp and, capacitance, Cp for the fabricated which accounts for electrons-exchange effect of the molecules
electrodes with the three different MWCNT concentrations (4, in the composite, ranges from 0.56 (4 wt%) to as high as
8, and 12 wt%) and three different dimensions (40 × 5, 50 × 0.89 (12 wt%) [50]. The ID /IG ratio decreases from 1.08 (4
4, and 60 × 3 mm2 ). Two samples are measured for each wt%) down to 1.03 (12 wt%) evidencing less defects in the
category, and there is not any significant variations in the electrodes.
Rp and Cp values in the measurement. From the measured
impedance results of the three different lengths of electrodes, B. Front-end Amplifier Measurements
the average equivalent resistances and capacitances of two This section presents the performance of the amplifier in
sample representatives of each concentration of 60 × 3 mm2 terms of the gain, power consumption, noise, and bandwidth
at 100 Hz frequency are summarized in Table II. The measured characteristics. The amplifier is designed with a gain of 40
values almost follow a linear relationship with respect to the dB by simply choosing C1 as 1 nF and C2 as 1 pF. Since
MWCNT concentration change. From Fig. 5 it is evident that, the measured equivalent dry electrode impedance is very high
as the MWCNT concentration increases, the resistance and (∼90-140 kΩ at 100 Hz for the three concentrations), the
the capacitance values decrease for all the electrodes. Also, amplifier is designed to have a high input impedance (∼350
the capacitance values are higher for larger electrode size (40 kΩ at 100 Hz) to match with it [51]. The bias voltage for
× 5 mm2 and 50 × 4 mm2 ) than the smaller (60 × 3 mm2 ) controlling the pseudoresistor value is chosen to be as 0.6 V.
one. The die microphotograph of the chip is shown in Fig. 8(a),
The morphology of the electrode is investigated using the while the packaged chip along with the Printed Circuit Board
scanning electron microscopy, as reported in Fig. 6 for the (PCB) is shown in Fig. 8(b). The fabricated amplifier occupies
8 wt% electrode. It shows an excellent uniformity of the an active die area of 0.14 mm2 while the total chip area is 2.25
composite layer with an average roughness of the layer that mm2 . The fabricated bare die is packaged using a 7 × 7 mm2
is evaluated with the help of the profilometer (VEECO - Quad flat no lead (QFN) package with 44 pins manufactured
Dektak 6M) and is measured to be 15.6 ± 5.2 µm. Main by SEMPEC Inc. The experimental mid-band gain of the
morphological differences among the electrodes are due to the amplifier is evaluated to be 43.09 dB. The measured lower
variations in thickness of the MWCNT/PDMS layer, which cut-off frequency is 0.175 Hz that is suitable for recording
ranges from 122.4 µm to 152.3 µm. The electrical conduc- low-frequency ECG signals. The higher cut-off frequency is
tivity of the Ag and the composite layers are also evaluated. measured as 1.636 kHz by setting the load capacitor as 2 pF
It follows the percolation theory that models the conductive
particle dispersed inside an insulating matrix [45], [46]. The
conductivity increase from 481 to 685 mS/cm is observed as

Fig. 5: Measured (a) resistances and (b) capacitances of the fabricated Fig. 6: Cross-sectional SEM image (400X magnification) of the
electrodes for each MWCNT concentration at 100 Hz. representative fabricated dry electrode.

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

Fig. 10: (a) Recorded ECG signals from electrode with 4 wt%
MWCNT concentration, (b) analysis of the average peak to peak
voltage as a function of MWCNT concentrations.

Fig. 7: Raman spectra of the representative electrode for each

MWCNT concentration.

Fig. 11: (a) Recorded ECG signals from commercial electrode, (b)
test setup for ECG acquisition with the proposed dry and commercial
electrodes.

precordial lead (modified lead V2) to assess the heart rate

[54]. Thus, the use of a minimum set of dry electrodes can
Fig. 8: (a) Chip microphotograph of the amplifier, (b) packaged chip
with the test bench. be employed for clinical assessment even though not in a
clinical setting. To investigate the dry electrode performance,
(Fig. 9(a)). The total power consumption is 1.5 µW with a typical ECG signals are acquired from the dry electrodes
voltage supply of 1.2 V. Measured input-referred noise of the along with the designed amplifier. To detect the ECG signal,
amplifier is calculated as 2.81 µVrms over the bandwidth of we placed the electrodes in 3-leads configuration (RA-right
0.175 Hz to 1.636 kHz (Fig. 9(b)). arm, LA-left arm, and LL-left leg). ECG pattern is recorded
(Fig. 10(a)) with the three electrodes having the concentration
of 4, 8, and 12 wt%. The representative signal is achieved
C. ECG Signal Acquisition
using the 60 × 3 mm2 electrode of 4 wt% concentration.
The 12-Lead standard ECG is clinically accepted for the The peak-to-peak amplitude is found to be dependent on the
detection of cardiac events even though the literature has MWCNT concentrations as shown in Fig. 10(b). With the
reported how a higher number of recording sites may allow for 12 wt% electrode, the recorded ECG signal’s peak-to-peak
improved diagnostic potential [52]. However, the reconstruc- amplitude is 0.27 V while that with the 4 wt% electrode
tion of the 12-leads ECG signal is possible by a reduced set is 0.24 V. The ECG signals are also acquired from the
of precordial leads [53]. In cardiovascular related pathologies commercial Ag/AgCl electrode from Karlsson Robotics using
such as apnea, ECG activity is monitored using a single the proposed amplifier to compare with the performance of
the dry electrodes (Fig. 11(a)). The peak-to-peak amplitude
is found as 0.06 V (with a baseline wandering of 45 mV).
Compared to the signal acquired by the commercial electrode,
the proposed dry electrode (0.24 Vp−p with 15 mV baseline
wandering) prove to be less prone to noise. From the analysis
of the recorded data, it is evident that the SNR is better
for the proposed dry electrodes compared to the commercial
ones. The SNR values are calculated from the power spectral
density, and it is observed that as the MWCNT concentration
of the dry electrodes increases, the SNR increases from 35.76
dB to 38.67 dB. The lowest concentrated proposed electrode
achieves a ∼11.3 dB improvement in SNR compared to the
Fig. 9: (a) Experimental closed-loop gain, and (b) measured input- commercial one. The dry electrodes perform better because
referred noise of the proposed amplifier. of the fact that the skin-electrode impedance is lower for

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

TABLE III
Comparison of Performances of ECG Electrodes V. C ONCLUSION
[16] [17] [26] This work An ECG signal acquisition system is designed with a CMOS
Material PEDOT:PSS Ag/AgCl MWCNT/PDMS MWCNT/PDMS
Substrate polyester polyurethane PET Kapton fully differential amplifier along with the proposed dry and
Concentration(wt%) - - 2-8 4-12
Thickness (mm) 4 - 0.3 0.5
flexible electrodes. MWCNT/PDMS composite polymer is
Dimension (mm2 )
SNR
30×30 (S)
nr
10-30 (C)
16.30-28.72
20×20 (S)
nr
See Table I
35.8-38.7
integrated with the Kapton substrate to fabricate the flexible
Conductivity (mS/cm) 424-575 nr ∼500 481-685 and biocompatible electrodes that are suitable for ECG signal
C: Circular, S: Square, nr: Not Reported.
measurements. As the conductivity of the proposed electrode
TABLE IV ranges between 481-685 mS/cm, this results in an improved
Comparison of Performances of Amplifier Unit skin-electrode impedance compared to the other works re-
ported in the literature as flexible and wearable ECG sensors.
[33] [34] This work
Technology (nm) 130 180 130 The on-chip amplifier shows a high gain in the reconfigurable
Supply voltage (V) 1 1.5 1.2 ECG frequency bandwidth with a low-power consumption.
Power consumption (µW ) 12.1 3.23 1.5
Mid-band gain (dB) 40 39.8 43.09 The amplifier is designed in such a way to reduce the low-
Bandwidth (Hz) 0.05-10.5k 10-10k 0.17-1.6k
Input-referred noise (µVrms ) 2.2 3 2.81 frequency flicker noise. Thus, these thin and flexible elec-
trodes with the CMOS read-out circuitry exhibiting a high
SNR builds a novel ECG acquisition system. The proposed
ECG recording system has the potential to be wearable and
the proposed MWCNT-PDMS electrodes compared to that of
miniaturized to be implemented in the ubiquitous monitoring
the Ag/AgCl electrodes. There are no variabilities in terms
system.
of the measurement set-up (e.g. contact pressure, area size,
moisture of the skin, or other factors) between the proposed
electrodes and the commercial electrodes for obtaining the R EFERENCES
results in Fig. 10(a) and 11(a). Also, another reason for [1] H. Khamis, R. Weiss, Y. Xie, C.-W. Chang, N. H. Lovell, and S. J.
better performance in terms of flexibility and elasticity is that Redmond, “QRS detection algorithm for telehealth electrocardiogram
the proposed electrode uses polyimide film (Kapton) as the recordings,” IEEE Transactions on Biomedical Engineering, vol. 63,
no. 7, pp. 1377–1388, 2016.
substrate, whereas the commercial electrodes use Polyethylene [2] M. Oehler, V. Ling, K. Melhorn, and M. Schilling, “A multichannel
(PE) foam. The Young’s modulus for polyimide is 2.07 GPa, portable ECG system with capacitive sensors,” Physiological Measure-
and for the PE film it is 210 GPa [55]. Since the lower the ment, vol. 29, no. 7, p. 783, 2008.
Young’s modulus of the material, the more is it’s elasticity. [3] T.-W. Shen, W. Tompkins, and Y. Hu, “One-lead ECG for identity
verification,” in Proceedings of the Second Joint 24th Annual Con-
Kapton film allows more flexibility than the commercial ECG ference and the Annual Fall Meeting of the Biomedical Engineering
electrodes. The test setup for obtaining the ECG data with the Society][Engineering in Medicine and Biology, vol. 1. IEEE, 2002,
proposed MWCNT/PDMS based dry electrodes (commercial pp. 62–63.
[4] E. Nemati, M. J. Deen, and T. Mondal, “A wireless wearable ECG sensor
electrodes in the inset) is presented in Fig. 11(b). Harmless for long-term applications,” IEEE Communications Magazine, vol. 50,
adhesive tapes are used to attach the electrodes to the skin. no. 1, pp. 36–43, 2012.
[5] K.-P. Hoffmann and R. Ruff, “Flexible dry surface-electrodes for ECG
A comparison analysis of the flexible dry electrode and long-term monitoring,” in 2007 29th Annual International Conference
the front-end amplifier among the other previously published of the IEEE Engineering in Medicine and Biology Society. IEEE, 2007,
works is presented in Table III and Table IV, respectively. pp. 5739–5742.
[6] S. A. Pullano, C. Critello, and A. S. Fiorillo, “Triboelectric-induced
Results evidence that the proposed dry electrodes, apart from pseudo-icg for cardiovascular risk assessment on flexible electronics,”
their simple fabrication and cost-effective design, provide Nano Energy, p. 104278, 2019.
better ECG signal quality compared to the traditional Ag/AgCl [7] J. Heikenfeld, A. Jajack, J. Rogers, P. Gutruf, L. Tian, T. Pan, R. Li,
M. Khine, J. Kim, and J. Wang, “Wearable sensors: modalities, chal-
and other dry electrode designs in terms of size, SNR and lenges, and prospects,” Lab on a Chip, vol. 18, no. 2, pp. 217–248,
conductivity. No significant variations have been observed in 2018.
the ECG signal amplitude by varying the electrode dimensions. [8] H.-C. Jung, J.-H. Moon, D.-H. Baek, J.-H. Lee, Y.-Y. Choi, J.-S. Hong,
and S.-H. Lee, “CNT/PDMS composite flexible dry electrodes for long-
The MWCNT concentration variation significantly impacts term ECG monitoring,” IEEE Transactions on Biomedical Engineering,
the output voltage showing an increase of 12.5% with the vol. 59, no. 5, pp. 1472–1479, 2012.
12 wt% electrode compared to the 4 wt% electrode. As far [9] B.-C. Kang and T.-J. Ha, “Wearable carbon nanotube based dry-
as the duration of the ECG signal recording and parameters electrodes for electrophysiological sensors,” Japanese Journal of Applied
Physics, vol. 57, no. 5S, p. 05GD02, 2018.
are concerned, 30 seconds of recorded ECG signal has been [10] M. Fernandez and R. Pallas-Areny, “Ag-Agcl electrode noise in high-
analyzed that showed an average P-R interval of 0.16 s, Q- resolution ECG measurements.” Biomedical Instrumentation & Technol-
T interval of 0.4 s, P wave of 0.08 s and QRS complex of ogy, vol. 34, no. 2, pp. 125–130, 2000.
[11] A. Searle and L. Kirkup, “A direct comparison of wet, dry and insulating
0.11 s. The above-reported results are in good agreement with bioelectric recording electrodes,” Physiological Measurement, vol. 21,
the typical ECG parameters range of healthy subjects. The no. 2, p. 271, 2000.
electronic interface designed on-chip has shown a low-power [12] S. Majumder, L. Chen, O. Marinov, C.-H. Chen, T. Mondal, and
M. J. Deen, “Noncontact wearable wireless ECG systems for long-term
performance of 1.5 µW with a gain of 43 dB, which exhibits monitoring,” IEEE Reviews in Biomedical Engineering, vol. 11, pp. 306–
to be a good match for the ECG acquisition system. It can 321, 2018.
also be seen from the Table IV, the proposed ECG amplifier [13] E. Özkaya and P. Kavlak Bozkurt, “Allergic contact dermatitis caused by
self-adhesive electrocardiography electrodes: a rare case with concomi-
is superior than the other amplifiers in-terms of total power tant roles of nickel and acrylates,” Contact Dermatitis, vol. 70, no. 2,
consumption, mid-band gain and bandwidth reconfigurability. pp. 121–123, 2014.

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.3001209, IEEE Sensors
Journal

[14] T. Niederhauser, A. Haeberlin, T. Marisa, M. Jungo, J. Goette, M. Ja- [35] S. Shamsir, I. Mahbub, S. A. Pullano, S. Islam, and A. Fiorillo, “Design
comet, R. Abächerli, and R. Vogel, “Electrodes for long-term esophageal of a charge amplifier for a low-power respiration monitoring system,”
electrocardiography,” IEEE Transactions on Biomedical Engineering, IET Circuits, Devices & Systems, 2019.
vol. 60, no. 9, pp. 2576–2584, 2013. [36] T. S. Pui, Y. Chen, C. C. Wong, R. Nadipalli, R. Weerasekera, S. K.
[15] H. Zhang, W. Pei, Y. Chen, X. Guo, X. Wu, X. Yang, and H. Chen, “A Arya, H. Yu, and A. R. Rahman, “High density cmos electrode array for
motion interference-insensitive flexible dry electrode,” IEEE Transac- high-throughput and automated cell counting,” Sensors and Actuators B:
tions on Biomedical Engineering, vol. 63, no. 6, pp. 1136–1144, 2016. Chemical, vol. 181, pp. 842–849, 2013.
[16] D. Pani, A. Dessı̀, J. F. Saenz-Cogollo, G. Barabino, B. Fraboni, [37] N. T. Tasneem and I. Mahbub, “A low-power low-noise reconfigurable
and A. Bonfiglio, “Fully textile, PEDOT: PSS based electrodes for bandwidth BiCMOS neural amplifier,” in 2018 IEEE 13th Dallas
wearable ECG monitoring systems,” IEEE Transactions on Biomedical Circuits and Systems Conference (DCAS). IEEE, 2018, pp. 1–5.
Engineering, vol. 63, no. 3, pp. 540–549, 2016. [38] M. H. Hamzah, A. B. Jambek, and U. Hashim, “Design and analysis of
[17] M. A. Yokus and J. S. Jur, “Fabric-based wearable dry electrodes for a two-stage cmos op-amp using silterra’s 0.13 µm technology,” in 2014
body surface biopotential recording,” IEEE Transactions on Biomedical IEEE Symposium on Computer Applications and Industrial Electronics
Engineering, vol. 63, no. 2, pp. 423–430, 2016. (ISCAIE). IEEE, 2014, pp. 55–59.
[18] J.-Y. Baek, J.-H. An, J.-M. Choi, K.-S. Park, and S.-H. Lee, “Flexible [39] T. Degen and H. Jäckel, “Continuous monitoring of electrode–skin
polymeric dry electrodes for the long-term monitoring of ECG,” Sensors impedance mismatch during bioelectric recordings,” IEEE Transactions
and Actuators A: Physical, vol. 143, no. 2, pp. 423–429, 2008. on Biomedical Engineering, vol. 55, no. 6, pp. 1711–1715, 2008.
[19] M. Fernandes, K. Lee, R. Ram, J. Correia, and P. Mendes, “Flexible [40] I. Mahbub, S. A. Pullano, H. Wang, S. K. Islam, A. S. Fiorillo, G. To,
PDMS-based dry electrodes for electro-optic acquisition of ECG signals and M. Mahfouz, “A low-power wireless piezoelectric sensor-based
in wearable devices,” in 2010 Annual International Conference of the respiration monitoring system realized in CMOS process,” IEEE Sensors
IEEE Engineering in Medicine and Biology. IEEE, 2010, pp. 3503– Journal, vol. 17, no. 6, pp. 1858–1864, 2017.
3506. [41] X. Zou, X. Xu, L. Yao, and Y. Lian, “A 1-v 450-nw fully integrated
[20] P. F. Shahandashti, H. Pourkheyrollah, A. Jahanshahi, and H. Ghafoori- programmable biomedical sensor interface chip,” IEEE Journal of Solid-
fard, “Highly conformable stretchable dry electrodes based on inex- state Circuits, vol. 44, no. 4, pp. 1067–1077, 2009.
pensive flex substrate for long-term biopotential (emg/ecg) monitoring,” [42] R. H. Olsson, D. L. Buhl, A. M. Sirota, G. Buzsáki, and K. D.
Sensors and Actuators A: Physical, 2019. Wise, “Band-tunable and multiplexed integrated circuits for simulta-
[21] D. Pani, A. Achilli, A. Spanu, A. Bonfiglio, M. Gazzoni, and A. Bot- neous recording and stimulation with microelectrode arrays,” IEEE
ter, “Validation of polymer-based screen-printed textile electrodes for transactions on biomedical engineering, vol. 52, no. 7, pp. 1303–1311,
surface emg detection,” IEEE Transactions on Neural Systems and 2005.
Rehabilitation Engineering, 2019. [43] T. V. Cao, D. T. Wisland, T. S. Lande, and F. Moradi, “Rail-to-rail
[22] F. J. Romero, E. Castillo, A. Rivadeneyra, A. Toral-Lopez, M. Becherer, low-power fully differential ota utilizing adaptive biasing and partial
F. G. Ruiz, N. Rodriguez, and D. P. Morales, “Inexpensive and flexible feedback,” in Proceedings of 2010 IEEE International Symposium on
nanographene-based electrodes for ubiquitous electrocardiogram moni- Circuits and Systems. IEEE, 2010, pp. 2820–2823.
toring,” npj Flexible Electronics, vol. 3, no. 1, p. 12, 2019. [44] Y. M. Chi, T.-P. Jung, and G. Cauwenberghs, “Dry-contact and noncon-
tact biopotential electrodes: Methodological review,” IEEE Reviews in
[23] A. A. Chlaihawi, B. B. Narakathu, S. Emamian, A. Eshkeiti, S. G. A.
Biomedical Engineering, vol. 3, pp. 106–119, 2010.
Reddy, B. J. Bazuin, and M. Z. Atashbar, “Development of flexible dry
[45] A. Fiorillo, C. Critello, and A. Pullano, “Theory, technology and
ECG electrodes based on MWCNT/PDMS composite,” in 2015 IEEE
applications of piezoresistive sensors: A review,” Sensors and Actuators
SENSORS. IEEE, 2015, pp. 1–4.
A: Physical, 2018.
[24] T. I. Oh, S. Yoon, T. E. Kim, H. Wi, K. J. Kim, E. J. Woo, and R. J.
[46] N. Meziane, J. Webster, M. Attari, and A. Nimunkar, “Dry electrodes
Sadleir, “Nanofiber web textile dry electrodes for long-term biopotential
for electrocardiography,” Physiological Measurement, vol. 34, no. 9, p.
recording,” IEEE Transactions on Biomedical Circuits and Systems,
R47, 2013.
vol. 7, no. 2, pp. 204–211, 2013.
[47] L. Bokobza, J.-L. Bruneel, and M. Couzi, “Raman spectra of carbon-
[25] Y.-H. Chen, M. de Beeck, L. Vanderheyden, E. Carrette, V. Mihajlović,
based materials (from graphite to carbon black) and of some silicone
K. Vanstreels, B. Grundlehner, S. Gadeyne, P. Boon, and C. Van Hoof,
composites,” C, vol. 1, no. 1, pp. 77–94, 2015.
“Soft, comfortable polymer dry electrodes for high quality ECG and
[48] J. H. Kim, J.-Y. Hwang, H. R. Hwang, H. S. Kim, J. H. Lee, J.-W. Seo,
EEG recording,” Sensors, vol. 14, no. 12, pp. 23 758–23 780, 2014.
U. S. Shin, and S.-H. Lee, “Simple and cost-effective method of highly
[26] H.-L. Peng, J.-Q. Liu, H.-C. Tian, B. Xu, Y.-Z. Dong, B. Yang, conductive and elastic carbon nanotube/polydimethylsiloxane composite
X. Chen, and C.-S. Yang, “Flexible dry electrode based on Carbon nan- for wearable electronics,” Scientific reports, vol. 8, no. 1, p. 1375, 2018.
otube/polymer hybrid micropillars for biopotential recording,” Sensors [49] S. Santangelo, G. Messina, G. Faggio, M. Lanza, and C. Milone,
and Actuators A: Physical, vol. 235, pp. 48–56, 2015. “Evaluation of crystalline perfection degree of multi-walled carbon
[27] J. H. Lee, Y. W. Nam, H.-C. Jung, D.-H. Baek, S.-H. Lee, and J. S. nanotubes: correlations between thermal kinetic analysis and micro-
Hong, “Shear induced CNT/PDMS conducting thin film for electrode raman spectroscopy,” Journal of Raman Spectroscopy, vol. 42, no. 4,
cardiogram (ECG) electrode,” BioChip Journal, vol. 6, no. 1, pp. 91– pp. 593–602, 2011.
98, 2012. [50] C. Lou, R. Li, Z. Li, T. Liang, Z. Wei, M. Run, X. Yan, and X. Liu,
[28] A. A. Chlaihawi, B. B. Narakathu, S. Emamian, B. J. Bazuin, and M. Z. “Flexible graphene electrodes for prolonged dynamic ecg monitoring,”
Atashbar, “Development of printed and flexible dry ECG electrodes,” Sensors, vol. 16, no. 11, p. 1833, 2016.
Sensing and Bio-sensing Research, vol. 20, pp. 9–15, 2018. [51] G. Gargiulo, P. Bifulco, M. Cesarelli, M. Ruffo, M. Romano, R. A.
[29] P. Taptimthong, J. Rittinger, M. C. Wurz, and L. Rissing, “Flexible Calvo, C. Jin, and A. van Schaik, “An ultra-high input impedance
magnetic writing/reading system: Polyimide film as flexible substrate,” ecg amplifier for long-term monitoring of athletes,” Medical devices
Procedia Technology, vol. 15, pp. 230–237, 2014. (Auckland, NZ), vol. 3, p. 1, 2010.
[30] J. Zhang, H. Zhang, Q. Sun, and R. Zhang, “A low-noise, low-power [52] F. Kornreich, R. S. MacLeod, and R. L. Lux, “Supplemented standard
amplifier with current-reused ota for ecg recordings,” IEEE Transactions 12-lead electrocardiogram for optimal diagnosis and reconstruction of
on Biomedical Circuits and Systems, vol. 12, no. 3, pp. 700–708, 2018. significant body surface map patterns,” Journal of Electrocardiology,
[31] R. R. Harrison and C. Charles, “A low-power low-noise CMOS amplifier vol. 41, no. 3, pp. 251–256, 2008.
for neural recording applications,” IEEE Journal of Solid-state Circuits, [53] S. P. Nelwan, J. A. Kors, and S. H. Meij, “Minimal lead sets for recon-
vol. 38, no. 6, pp. 958–965, 2003. struction of 12-lead electrocardiograms,” Journal of Electrocardiology,
[32] K. A. Ng and P. K. Chan, “A CMOS analog front-end IC for portable vol. 33, pp. 163–166, 2000.
eeg/ecg monitoring applications,” IEEE Transactions on Circuits and [54] S. A. Pullano, I. Mahbub, M. G. Bianco, S. Shamsir, S. K. Islam, M. S.
Systems I: Regular Papers, vol. 52, no. 11, pp. 2335–2347, 2005. Gaylord, V. Lorch, and A. S. Fiorillo, “Medical devices for pediatric
[33] F. Zhang, J. Holleman, and B. P. Otis, “Design of ultra-low power apnea monitoring and therapy: past and new trends,” IEEE Reviews in
biopotential amplifiers for biosignal acquisition applications,” IEEE Biomedical Engineering, vol. 10, pp. 199–212, 2017.
Transactions on Biomedical Circuits and Systems, vol. 6, no. 4, pp. [55] Y. Xu, C. Luo, F.-G. Zeng, J. C. Middlebrooks, H. W. Lin, and Z. You,
344–355, 2012. “Study of the carrier-aided thin film electrode array design for cochlear
[34] G. Deepika and K. Rao, “A low power low noise amplifier for recording insertion,” Micromachines, vol. 9, no. 5, p. 206, 2018.
neural signals,” ARPN Journal of Engineering and Applied Sciences,
vol. 11, no. 5, 2016.

1558-1748 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Edinburgh. Downloaded on June 15,2020 at 03:19:39 UTC from IEEE Xplore. Restrictions apply.