Proceedings of the 2023 International Conference on Mechatronics and Smart Systems

DOI: 10.54254/2755-2721/28/20230269

ECG circuit design and analysis algorithm

Yixin Yang
Carnegie Institute of Technology, Carnegie Mellon University, Pittsburgh, PA 15213,
United States of America

[email protected]

Abstract. Nowadays, many people have a sub-health lifestyle: poor diet, lack of exercise, high
pressure, inadequate sleep, long periods of sitting, etc. All the unhealthy behaviours will lead to
chronic disease, especially heart disease. People then use wearable devices like Apple Watch to
monitor their heartbeat and try to find out if anything goes wrong in advance. This paper is then
going to introduce how wearable devices detect the heartbeat from two perspectives. The
hardware part gives a detailed implementation of a complete design of circuits within the
wearable device used to detect heartbreak. The circuit is designed to be power-efficient, resistant
to noise, and capable of amplifying the input voltage by about 100 times. The circuit design
would contain three main parts: the human body circuit, the three-amplifier system, and the filter
system. For each part, detailed information about how to choose the value of components is
provided, and how the chosen value of components meets the requirement is also illustrated. On
the other hand, it will introduce three algorithms used to analyze the ECG signals output from
the circuit.

Keywords: ECG circuit design, pseudo resistors, right leg driven circuit, three-amplifier system.

1. Introduction
Even though the technology associated with medical treatment has much remarkable progress, heart
disease has still been the leading cause of death for men and women in the United States [1]. According
to the data, one person would die due to cardiovascular disease about every 34 seconds in the United
States [2]. Unlike other chronic diseases like cancer, where patients could receive proper therapy after
they are diagnosed, patients with heart diseases may suffer from heart attacks and strokes, which will
take the life of patients within several minutes, or even several seconds. In addition, although older
people are still the main part of the dead population, the number of young people who die from heart
disease is still considerable; in the current competitive society, young people would face challenges
from both work and study, which makes staying up late a common behavior of the young generation.
Moreover, covid has just passed less than one year. People who recover from covid may still suffer from
sequela such as inflammatory cardiomyopathy.
While once the patients face the outbreak of heart disease, other people and themselves should be
prepared to conduct first-aid, ensuring they can then be sent to the hospital for further treatment.
Otherwise, they may pass away before they arrive at the hospital. To prepare before the onset of a heart
attack, patients should first know their own conditions well. As some heart diseases like cardiac-
arrhythmias cannot be diagnosed within a short time, people need to monitor their heart condition

© 2023 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(https://creativecommons.org/licenses/by/4.0/).

150
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

continuously. Meanwhile, monitoring heart conditions could also help people avoid a heart attack; when
they find strange behaviors of the heart, they could go to the doctors or stop their unhealthy behaviors
immediately before a heart attack occurs.
Wearable devices like a watch could be optimal types of equipment that meet the expectation. For
example, you can take a quick 30-second ECG test on your Apple Watch and get a brief summary about
if your heart is working well. This paper will then introduce how this technique is realized from both
hardware and software aspects.

2. ECG circuit design

Since each beat is initiated by the electrical stimulus generated by the sinoatrial node (SA node), the
electrical signal could be detected and measured to assess the performance of the heart. The method that
monitors the heartbeat by recording the electrical signal is called Electrocardiogram (ECG or EKG). To
evaluate the performance of an ECG circuit, aspects including the differential gain, the ability to deal
with the noise, and the total power consumption. ECG circuit with high quality should be able to amplify
the signal it received to the expected magnitude (typically, the gain should be about 40dB, which means
the output voltage is about 100 times the input voltage) and cancel out the internal and exterior noise
with low power [3]. This part is going to discuss the implementation of ECG. The implementation could
be divided into three main parts: the human body circuit, the three-amplifier system, and the filter system.

2.1. Human body circuit

Since the human body is conductive, the potential created by the heartbeat could then be transmitted to
the body’s surface. To measure the electrical signals from the body surface, electrodes and leads are
commonly used components. The electrode is a small conductive pad attached to the skin to receive the
potential, and the lead refers to a pair of electrodes. Leads could be simply classified into two types:
unipolar leads and bipolar leads. While bipolar leads measure the differential potential across a pair of
electrodes, unipolar leads measure a common potential. According to Wilson’s Central Terminal, the
common potential will be the average of the potentials from the right arm, left arm, and left leg.
R2

R1 1000k
VRA
2k
C1

50n

R4

R8 R3 1000k
Vecg- Vecg+ VLA
200 2k
C2

R9 R7 50n

200 200

R6

R5 1000k
VRL
2k
C3

I1 C4 50n
200p
SINE(0 0.1u 60)

Figure 1. Human body circuit design.

The human body circuit design is shown in Figure 1. In this design, three augmented limb leads, a
specific type of unipolar lead, are implemented. Even though both leads will be common for medical

151
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

usage, unipolar tends to be more pervasive for its easier introduction and better position of the lead [4].
The graph above is our design for the human body circuit. The ecg+ and ecg- tags are the input potential
that is read from the ECG. To mimic the noise (always 60 Hz) in real life, a current source I1 is attached
to the input. VRA corresponds to the potential of the right arm, VLA corresponds to the potential of the
left arm, and VRL corresponds to the potential of the right leg.

2.2. Three-amplifier system

The average voltage generated by one heartbeat is about 75 mini volts. The amount is so tiny that it is
not suitable for further analysis. To ensure the heartbeat signal could be used for analysis, the op-amp
system should amplify the signal by 40dB (the output will be about 100 times the input voltage). Since
a 40dB gain is expected for the circuit, one op-amp is not an appropriate choice (typically, a power
amplifier would have a 20 dB gain). Therefore, a two-step amplification is needed. In addition, this
system would have two inputs: voltage from the right arm and voltage from the left arm. This system
then needs to convert the differential input signals to a single-ended output. A three-op-amp system is
then chosen. The common three-amplifier system is shown in Figure 2.
For a traditional three-op-amp model, the differential gain of the first stage 𝐴𝑑1 and second stage
𝐴𝑑2 will be:
𝑅1+𝑅2
𝐴𝑑1 = 1 + (1)
𝑅𝑔𝑎𝑖𝑛
𝑅4
𝐴𝑑2 = (2)
𝑅3
X1

V1 Vin+ VDD R3 R4
Vout
Vin- R R

R1
R

X3

Rgain Vin+ VDD

Vout Vout
R
Vin-

R2
X2 R

V2 Vin+ VDD R3 R4
Vout
Vin- R R

Figure 2. Common three-amplifier system.

Instead of using resistors to control the amplification amount, capacitors are used to control the
amount of amplification.
Pseudo-resistors are implemented to replace resistors. With the same resistance, pseudo resistors tend
to occupy less space and would create less noise than a typical resistor [5]. Since the ECG circuit would
be implemented in wearable devices and wearable devices should be small for people to carry along
with them daily, pseudo resistors then become a more desirable choice. The construction of pseudo
resistor is shown in Figure 3.

152
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

V_in V_out

M1

M2

M3

M4
PMOS

PMOS

PMOS

PMOS
Figure 3. Construction of pseudo resistor.
Therefore, the new gain will be:
𝑅𝐶1 +𝑅𝐶2
𝐴𝑑1 = 1 + 𝑅𝐶𝑔𝑎𝑖𝑛
(3)
1 1
+
𝑗𝑤𝐶1 𝑗𝑤𝐶2
𝐴𝑑1 = 1 + 1 (4)
𝑗𝑤𝐶𝑔𝑎𝑖𝑛

𝐶𝑔𝑎𝑖𝑛(𝐶1+𝐶2)
𝐴𝑑1 = 1 + (5)
𝐶1𝐶2
𝑅
𝐴𝑑2 = 𝑅𝐶4 (6)
𝐶3
1
𝑗𝑤𝐶4
𝐴𝑑2 = 1 (7)
𝑗𝑤𝐶3

𝐶3
𝐴𝑑2 = 𝐶4 (8)
The revised three-amplifier system is shown in Figure 4. According to the data from the following
graph, the first-stage differential gain will be 20, and the second-stage differential gain will be 5.
Therefore, the overall gain will be 100, which equals 40 dB. The requirement for the amplifier system
is met.
VDD

X1

VRA Vin+ VDD C3 C4

Vout
Vin- 50p 5p
V_in

C1 V_in V_out
PR1

X4
10p X3
V_out

Vin+ VDD
C5 Vout Vout
1p Vin-
V_in

Cgain
PR2
V_out

100p
VDD

C6

1p
V_in
VDD

C2 V_out V_in
PR3

10p X5
V_out

X2

VLA Vin+ VDD C3_2 C4_2

Vout Vref
Vin- 50p 5p

Figure 4. Revised three-amplifier system.

153
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

2.3. Filter system

The frequency of ECG will vary depending on its usage. For example, for clinical ECG, the frequency
would range from 0.05Hz to 100Hz; the monitoring mode will have a frequency from 0.05Hz to 35 Hz;
and the surgical mode will have a frequency from 1HZ to 25 Hz [6]. For this design, the frequency range
would be 0 to 250 Hz.
The bandpass filter consists of a high pass filter which only allows signals with a frequency larger
than 0 Hz to pass, and a low pass filter which only allows signals with a frequency less than 250Hz to
pass.
The frequency of the bandpass filter is calculated by:
1
𝑓𝐿 = 2𝜋𝑅 (9)
𝐿 𝐶𝐿
1
𝑓𝐻 = (10)
2𝜋𝑅𝐻 𝐶𝐻

The filter system design is shown in Figure 5. For the circuit below, the 𝑅𝐿 is 𝑅1 , the 𝐶𝐿 is 𝐶9 , the
𝑅𝐻 is 𝑅4 , and the 𝐶𝐻 is 𝐶10. The 𝑅2 and 𝐶11 are connected to the reference voltage to cancel out the
noise.
C4

5p

V_out V_in R1
Vdd

Vdd
PR4 7.96e9 C10
X3
X6 200p
Vin+ VDD C8 C9
X7
Vout Vin+ VDD R3 R4
Vin- 200p 200p Vout Vin+ VDD
Vin- 3.18e6 3.18e6 Vout Vout
Vin-
R2 C11
7.96e9
200p
Vdd

Vref

Vref

Figure 5. Filter system design.

2.4. Overall design and summary

The overall design is shown in Figure 6.
X1
Vdd
Vin+ VDD C3 C4
Vout
Vin- 50p 5p
V_out V_in
V_in

V_out V_in R1
C1
PR5
PR1

Vdd

Vdd

PR4 7.96e9 C10

V_out

10p X3
X6 200p
C12 Vin+ VDD C8 C9
X7
Vout Vin+ VDD R3 R4
C6 1p Vin- 200p 200p Vout Vin+ VDD
3.18e6 3.18e6
Vref Vin-
Vin-
Vout Vout

5p X4
R2
7.96e9
C11

200p
VLA
Vecg+

Vecg+ Vecg+
Vdd

VRA
Vecg- Vecg-
Vref
V_out
Vdd

VRL
Cgain C13
Vref
PR2

100p 1p
Vref
V_in

V1 V2 V3

0.5 0.25 SINE(0 1m 50)

AC 1m
Vdd

C2
V_out

10p
PR3

V_out V_in
V_in

PR5
Vecg-

X2

Vin+ VDD C5 C6
Vout Vref
f Vin- 50p 5p
Vecg+
Vref
Vdd

V_out V_in
V1 V2 V3
PR6

0.5 0.25 SINE(0 1m 50)

C7 AC 1m

100p
Vecg-

Vdd X9

VDD Vin+ Vref

Vout
Vin-

Figure 6. Overall ECG circuit design.

154
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

Aside from the three main parts mentioned above, a right-leg-driven circuit is connected to the design.
After the first amplification stage, this part would sense the common-mode voltage, invert the voltage,
magnify it, and send it back to the right leg circuit. This feedback circuit would help cancel out the
common mode interference created during the amplification process, and it is incredibly efficient in
undermining the influence of the 60Hz noise in real life [7].
As mentioned above, the amplifying amount should be around 40 dB. According to the simulation
of Figure 7, between 1-250 Hz, the amplifying amount is stable at around 43 dB, which meets the
requirement.
While both the right leg feedback circuit, pseudo resistors, and the filter system contribute to noise
reduction, the noise of the system is optimal, which is 170.86 microvolts.
V(vout)
50dB 160°

40dB
80°
30dB

20dB 0°

10dB
-80°
0dB

-10dB -160°

-20dB
-240°
-30dB

-40dB -320°

-50dB
-400°
-60dB

-70dB -480°

-80dB
-560°
-90dB

-100dB -640°
1mHz 10mHz 100mHz 1Hz 10Hz 100Hz 1KHz 10KHz 100KHz 1MHz 10MHz 100MHz

Figure 7. DB gain vs. frequency.

3. ECG analysis
Before walking into the algorithm that analyzes the ECG, the characteristic of the output signal should
first be introduced.

Figure 8. ECG wave [8].

The ECG wave can be classified into three waves – P, QRS, and T wave. Figure 8 is the ECG wave.
The P wave is the result of atrial activation, and the PR interval is the distance between the start of the
P wave and the start of the QRS complex. The PR Segment measures the distance between the end of
the P wave and the start of the QRS complex. This line can also be considered as the baseline, where
any amplitude will be the difference between the point and this line. The QRS complex is the result of
ventricular depolarization. It will be optimal to have a shorter time interval between the beginning of
the QRS complex and its end because this means the speed of ventricular depolarization is high. The T
wave results from ventricular repolarization. The QT interval will be the distance between the start of
the QRS complex and the end of the T wave, which is also the total time for ventricular depolarization

155
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

and repolarization. There is an inverse connection between the length of the QT interval and the heart
rate. In other words, a longer QT interval means a slower heart rate, and vice versa.

3.1. Peak detection

Peak detection is one of the most critical analyzing algorithms for ECG signals. Like its name, this
algorithm will track the R wave in the QRS complex, because the R wave always reaches the highest
amplitude of one heartbeat. Every time the heartbeat signals fill the window, the maximum value will
be recorded (the R wave with the largest amplitude will be recorded). If the peak is found in the central
part of the window and the new amplitude is higher than the past threshold, the threshold will be updated
to the new amplitude [9].

3.2. SQRS algorithm

Unlike peak detection, which uses the ECG signal as direct input, before the ECG signals go into the
SQRS algorithm, they will first pass a Finite Impulse Response filter which will create an approximated
slope for the signal. If the filtered signal is larger than the original threshold, it will have two possible
outcomes. First, if there are two to four signals detected within 160ms, it turns out that a normal heartbeat
is recorded. Second, if more than four detections happen within 200ms, it indicates a problem. The
algorithm will then reset if it faces either outcome [10].

3.3. WQRS algorithm

This algorithm will highly depend on the length transform of ECG. The algorithm will detect whether
there is a rising slope of the length transform. If a length transform is detected, it means that the QRS
complex begins. It could also find the end of the QRS complex through length transform. This algorithm
is also able to offer extra information about the end of the QRS complex. Compared to the peak detection
algorithm and the SQRS algorithm, the measurement of the WQRS algorithm would be more accurate
(highest sensitivity and high positive predictivity). However, this accuracy is in exchange for processing
time [10].

4. Conclusion
In conclusion, the hardware part is an ECG circuit that could be implemented into the wearable device.
It could amplify the electrical signal to about 43dB, and the noise is only about 170 microvolts.
Additionally, to save more space, the total capacitor is minimized, and resistors with large impedance
are changed into pseudo resistors. While our circuit design could fit the basic requirements – high
amplifying amount and low noise – for the ECG circuit, the real-life condition is much more complicated.
Each component would be cautiously evaluated and then can be chosen as the built-up parts. For
example, producers should consider which amplifier model would occupy a smaller space, become more
energy-efficient, and cost less. While all the results are based on theories and online simulation, more
data about hands-on experiments should be recorded to provide further information. Although many
different algorithms solve the problem, there is no best algorithm for the software. People should
evaluate their needs and find the algorithms that best fit their requirements.

References
[1] CDC 2000 Facts about heart disease and stroke 2020 M2 Presswire
[2] CDC 2022 Heart Disease Facts
[3] Wu B, Li M, Yang Y, and Zhang W 2012 ECG acquisition circuit design based on C8051F330.
Proc. 2012 IEEE-EMBS Int. Conf. Biomed. Health Inf. pp 93-96 doi: 10.1109/BHI.2012.621
1515.
[4] Kersschot I 1994 Unipolar versus bipolar leads Cardiac Pacing and Electrophysiology ed Aubert
A E, Ector H, and Stroobandt R (Springer, Dordrecht) doi: 10.1007/978-94-011-0872-0_20.
[5] Emanuele G, Fabio T, Francesco Z, Giuseppe S, Alireza M, Marco S and Giorgio F 2020 High-
Value Tunable Pseudo-Resistors Design IEEE J. Solid-state Circuits 55(8) pp 2094-2105 doi:

156
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/28/20230269

10.1109/JSSC.2020.2973639.
[6] Gifari M W, Zakaria H, and Mengko R 2015 Design of ECG Homecare:12-lead ECG acquisition
using single channel ECG device developed on AD8232 analog front end. 2015 Int. Conf.
Electri. Eng. Inf. pp 371-376 doi: 10.1109/ICEEI.2015.7352529.
[7] Li Y 2021 Design of ECG Signal Detection Circuit Based on Pacing Pulse Suppression. 2021
IEEE Int. Conf. Electron. Tech. pp 664-669 doi: 10.1109/ICETCI53161.2021.9563387.
[8] Aziz S, Ahmed S and Alouini 2021 M S ECG-based machine-learning algorithms for heartbeat
classification. Sci. Rep. 11 pp 18738 doi: 10.1038/s41598-021-97118-5
[9] Xiang J, Dong Y, Xue X, and Xiong H 2019 Electronics of a Wearable ECG With Level Crossing
Sampling and Human Body Communication. IEEE Trans. Biomed. Circuits Sys. 13(1) pp 68-
79, doi: 10.1109/TBCAS.2018.2879818
[10] Pino E, Ohno-Machado L, Wiechmann E, and Curtis D 2005 Real-time ECG algorithms for
ambulatory patient monitoring AMIA Annu Symp. Proc. 604(8)

157