application of piezo electronic sensor in jump counting
TRANSCRIPT
Thinh Ha
Application of Piezo Electronic Sensor in Jump Counting
Metropolia University of Applied Sciences
Bachelor of Engineering
Electronics
Bachelor’s Thesis
1 March 2021
Abstract
Author: Thinh Ha
Title: Application of Piezo Electronic Sensor in Jump Counting
Number of Pages: 14 pages + 1 appendices
Date: 1 March 2021
Degree: Bachelor of Engineering
Degree Programme: Electronics
Professional Major: Senior Lecturer
Supervisors: Anssi Ikonen
The aim of the project was to design and test a prototype of a system that can be
attached to a trampoline to sense impacts caused by jump movement. The system will
detect the movement and store the number of jumps to a counter, which value is sent
to a mobile application.
The system is comprised of the following main components: piezo electronic sensor, a
filter circuit which is used to modify the raw signal from the piezo sensor, a
microcontroller which will detect the jump movement from the sensor signal and send
the number of jumps to a mobile application.
The main goal of the project was to extract the correct number of jumps when attached
to the trampoline. Several prototypes was designed and tested in order to evaluate the
performance of the difference types of filter circuits and position of the sensor. The final
system designed was able to successfully detect jump movement and number of jumps
in a specific test scenario. However, in the future the performance of the system could
be improved smart motion detection algorithm, efficient power management and
versatile internet connectivity.
Keywords: Piezo electronic sensor
Contents
List of Abbreviations 5
1 Introduction 1
2 Piezo Electronic Sensor 2
2.1 Historical Background 2
2.2 Dielectric Polarization And Piezoelectric Effect 2
2.3 Piezo-Electronic Sensors 4
2.4 Introduction To Poly-Vinylidene Fluoride 5
2.5 Modelling Theory 8
2.5.1 Charge Mode Configuration 8
2.5.2 Voltage Mode Configuration 9
3 Hardware and Software Tools 10
3.1 Piezo Electronic Sensor Minisense 100 10
3.2 Particle Photon Microcontroller 14
3.3 Operational Amplifier LM358 14
3.4 Pads logic and Pads layout 15
3.5 Particle Web IDE 15
3.6 Blynk 16
4 Physical implementation of the device 17
4.1 Jump Counter Device 17
4.2 Piezoelectric Sensor 17
4.3 Design Architecture And Measurement Method 19
4.4 Circuit designing 20
4.4.1 Raw Piezo Electronic Signal 20
4.4.2 Pre-Amplifier Module 21
4.4.3 Sallen-Key low pass filter 23
4.4.4 Offset control and final gain 25
4.5 Software Implementation 27
4.6 Pulse Peak Detection 28
5 Testing 30
6 Limitation and future possibilities 32
7 Conclusion 33
References 34
List of Figure 1
Main code 3
Schematic of the circuit 6
PCB layout of the circuit 7
List of Abbreviations
OP AMP: Operational amplifier, high impedance electronic component, mainly
use to amplify weak signals
PVDF: Poly-Vinylidene Fluoride
SMDs: Surface Mounted Devices
IoT: Internet of Thing
PCB: Printed Circuit Board
1
1 Introduction
Shock and vibration take part in an important role in electronic and
electromechanical system. They can shorten the life of electronic system since
delicate leads and bond tend to be applied by strain and stress during the
frequent use, in which lack of could result in a system failure. Therefore, the
technology of measuring motion, vibration and shock is always priority
considered in designing and manufacturing process of almost all electronic
system. This technology has been mandatorily required in multiple technology
sectors such as automation, aerospace, consumer electronics.
The aim of this thesis work was to learn more about one of the most popular
shock and vibration sensing technology these days, the piezoelectric sensor. A
deep literature study was made about the general principle of the sensor, I have
compared and verified of learning thing with a commercial sensor by designing
and prototyping a simple application of the sensor. I conclude the thesis with a
set of measurement and result of the prototype as well as my personal discuss
about this technology and their use in electronic aspect.
2
2 Piezo Electronic Sensor
2.1 Historical Background
Piezo electronic is the property of a material that can gain a certain electrical
charge due to the deformation. The prefix ‘piezo’ come from the Greek, that mean
“to press” or “to squeeze”. This phenomenon was first discovered by the brothers
Pierre and Jacques Curie. They found that in certain crystal such as tourmaline,
topaz, quartz, electrical charges are created on the surface when applied with a
mechanical force to the crystal. In the following year, Gabriel Lipmann discovered
the inverse piezo electronic effect, this was soon to be confirmed by the Curie
brother by the same year. However, until the 1950s, by the discovered of new
material with better properties, included Barium Titanate, Zirconate Titanate,
manufacturers began to apply the new material in sensing applications [1] [2].
2.2 Dielectric Polarization And Piezoelectric Effect
The polarization of a material is the state when negative charges and positive
charges of the material are displaced from another. This usually achieved by
induce the material with an electric field. However, some material with a
permanent polar structurer can also gain the same polarization states from a
mechanical deform. That is the piezoelectric effect of a material, which can be
explained in the figure 1 (a).
3
Figure 1. Piezo electronic effect explained (a) At rest turbed molecule (b) Material under external force (c) Polarization effects under external force (d) Neutralized current induce by the external force [3].
At rest, the dipoles are perfectly balanced since the positive charge cancels out
the negative one due to the symmetrical arrangement of the crystal structure as
Figure 1 a. As seen in Figure 1 b, if some external force, squeeze, or stretch, for
example, is applied that causes the deform of the structure, the symmetry
structure is lost. Some atoms are clumped together, while others are scattered
which will then create net electrical charges. This phenomenon forms in the whole
structure of the material, create a temporary polarization that can be seen in
Figure 1 c. The effects can be seen from transverse, longitudinal, or shear forces
depend on the structure of the material. In figure 1 d, two metal electro nodes are
attached to two positive side of the polarization directions, and then short
circuiting each other with conductor. The deform of the material will deposit
opposite charges on the surface of each electro nodes. The free charge will flow
through the conductor in order to neutralized the polarization of the material.
4
2.3 Piezo-Electronic Sensors
Piezoelectric sensors are usually classified by the kinds of material they are made
from. There are three kind of materials as ceramic, crystal or polymeric. Some
commercially available piezo-electronic sensor is shown in Figure 2.
Figure 2. Common commercial piezo electronic sensors (a) Buzzer and transducer ceramic type (b) Pressure sensor using piezo electronic crystal [4] (c) Polymeric type [5].
Single crystal piezoelectronic is often seen in pressure transducers application.
The most common single crystal material sensor is lead magnesium niobate-lead
titanate crystal family. They have an extremelly high sensitivity due to its high
piezoelectric coefficient, very broad bandwidth. Ceramic Piezo Electric has good
sensitivity, less stable compares to the single-crystal one but they are very cheap.
The most common material of ceramic sensors are Lead Zirconate Titanate,
Barium Titanate and Lead Titanate. They are popular due to its low cost and
acceptable sensitivy. However, they are generally brittle, not durable and have a
narrow working temperature. Polymer Piezo Electronic offer even higher
sensitivy, also offer a excellent physcial properties such as high strength, high
impact resistance, durable, low density and flexiable [6].
5
2.4 Introduction To Poly-Vinylidene Fluoride
In 1969, Heiji Kawai found that the PVDF material could gain and remain a high
Piezo-, Pyro- and Ferro-electric properties after a polarization process with high
temperature and pressure [7]. After this breakthrough, more and more
researches have been conducted to discover more about PVDF material and its
application.
Poly-Vinylidene fluoride is a linear polymer chain made of multiple identical
monomers. The material is mainly produced by the free radical polymerization
process of gaseous Vinylidene Fluoride. There are approximately 2000
monomers tied together in a chain, the length is 0.5 um [8] and the weight is about
100000 g/Mol [9]. Each monomer has the chemical formula is CH2CF2, and
different phase structure formulas can be seen in Figure 3.
Figure 3. Structure formula of CH2CF2.
PVDF has three main conformations: α-, β- and γ-phase [10]. In each monomer
structure, the side with two fluorine atoms is highly negatively charged while the
side with two hydrogen atoms is oppositely positively charged. However, the
structure direction arrangement of Fluoride and Hydrogen atoms made the polar
properties of each phase different. In the α-phase, each charge has a different
atoms group that is pointed in the opposing directions, the charges cancel out
result in a non-polar structure. In the β- phase, each charge only has one
orientation, the negative and positive charges point opposite directions, resulting
in a dipole molecular. The direction of the dipole is perpendicular to the molecular
axis. Therefore, PVDF in β- phase has a permanent dipole structure and this is
6
the only structure that can be used as a piece of electronic material. The other
polymorphs, γ-phase, and δ-phase are slightly polar and show a much weaker
piece electronic effect than the β- phase.
Figure 4. Structure conformation of PVDF (a) α-phase (b) β- phase (c) γ-phase [2].
PVDF has a high piezoelectric property, that is the ability to produce an
electrical charge in response to structures deform caused by an external
mechanical force. The sensor used in this paper is made of a thin film PVDF
and the piezoelectric equation [11] can be express below:
[𝐷31
𝐷32
𝐷33
] = [0 0 00 0 0
𝑑31 𝑑32 𝑑33
0 𝑑15 0
𝑑24 0 00 0 0
] ∗
[ 𝑇1
𝑇2
𝑇3
𝑇4
𝑇5
𝑇6]
(1)
7
In the equation, denote the symbol Xij that ‘I’ refers to measurement direction and
‘j’ is the external mechanical direction. The directions of the piezo film can be
shown in Figure 5.
Figure 5. Direction of PVDF film sensor [12].
D31, D32 and D33 are electrical displacement in directions 1, 2 and 3; d31, d32
and d33 are the piezoelectric coefficient in directions 1, 2, and 3; d24 and d15
are piezoelectric coefficients in shearing strain directions 1 and 2. However, the
shearing coefficients are much smaller than the strain coefficients and thus can
be neglected. T1, T2 and T3 are tensile stress in directions 1, 2, and 3; T4, T5
and T6 are shear stress in directions 1,2 and 3 [13]. The output charge can be
calculated based on the electrode area A as equation (2)
𝑄 = (𝑑31 ∗ 𝑇1 + 𝑑32 ∗ 𝑇2 + 𝑑33 ∗ 𝑇3) ∗ 𝐴 (2)
𝑄 = 𝑑31 ∗ 𝑇1 ∗ 𝐴 (3)
The sensor used in this paper is a thin film, it is nearly impossible to deform the
sensor in directions 3 (thickness directions), which means T3 is 0. One end of
the film is horizontally connected to a hard PCB board. This board strengthens
the film in directions 2, thus made the deform in this directions can be
neglected, T2 is 0. Therefore, the equation () can be approximately calculated
by the equation (3). The piezoelectric coefficient is now approximately equal
8
d31, the higher the constant the stronger the piezo electronic effect of the
material.
2.5 Modelling Theory
In the simplest form, a piezoelectric sensor can be modelled as a charge source
parallel with a shunt capacitor and resistor [14]. However, the piezoelectric
operation is unique and depends on multiple conditions. Therefore, in actual
design and modelling, sensors are modelled in either two modes, charge mode
configuration or voltage mode configuration.
2.5.1 Charge Mode Configuration
Charge mode configuration is typically used when the electronics are connected
far from the sensor, the configuration is shown in Figure 6. In this mode the
amplifier must have a low bias input current since it does not charge or discharge
the gain capacitor. A typical CMOS operational amplifier is recommended since
it has a very low input current. Since the long distance between sensor and
amplifier, the capacitance effect of long cable is considered as CC, any charge
created by the sensor will charge the cable capacitor and thus create a voltage
between amplifier input. Since the amplifier has a very high gain, this voltage is
nulled by the sourcing and creates the same amount of charge to the feedback
capacitor and resistor. Measuring this will result in a high accuracy charge change
from the sensor.
9
Figure 6. Charge mode configuration [15].
2.5.2 Voltage Mode Configuration
Voltage mode is generally used when the sensor is placed close to the
amplifier, input charge is presented directed to high impedance input and then
amplified by the op-amp. The basic configuration is described in Figure 7. The
gain, in this case, can be adjusted by input and feedback resistor. The output
voltage can be calculated by equation (4) [15]. The capacitance of the cable
CC affects the actual gain of the op amplifier, and it must be as small as
possible to minimize its effect. Therefore, in practice the sensor needs to be
placed as close as possible to the operational amplifier.
𝑉𝑜𝑢𝑡 =𝑄𝑠
𝐶𝑠 + 𝐶𝑐∗ (1 +
𝑅𝑓
𝑅𝑔) (4)
One advantage of this configuration is that it provides a natural passband
frequency response. The low-pass and high pass cutoff frequency can be
defined by this equation (5) and (6) [15]. Resistor R is used to correct biases in
the Op-amp. Since this sensor application only measures real human jumps on
a trampoline, which are in the low-frequency range, we need to get the lowest
10
natural high pass threshold. One solution is to increase the value of resistor R
to be as high as possible, typically in the mega ohm range.
𝑓𝐻𝑃 =1
2𝜋 ∗𝑅𝑆∗𝑅
𝑅𝑆+𝑅∗ (𝐶𝑠 + 𝐶𝑐)
(5)
𝑓𝐿𝑃 =1
2𝜋 ∗ 𝑅𝑓 ∗ 𝐶𝑓 (6)
Figure 7. Voltage mode configuration [15].
3 Hardware and Software Tools
3.1 Piezo Electronic Sensor Minisense 100
The Minisense 100 is a piezo electronic vibration sensor manufactured by
Measurement Specialties. The sensor is made from PVDF material, into a thin
film that is flexible and durable. One end of the sensor is attached to a mass; the
purpose was to increase the sensitivity when operates at low frequencies. The
active sensing film is coated with a radio frequency interference and
electromagnetic interference rejection layer. The sensor has a dynamic range
and excellent linearity response property. The sensor is small, lightweight with
only 0.5 gram, and can be operated between -20C to 60C (Specialties, 2019),
thus can be used in various applications e.g. load balancing, motion sensor,
11
tamper detection, impact sensing. Figure 5. also indicates the Minisense 100
sensor used in this module.
In general, the sensor performs as a cantilever-beam accelerometer as shown in
Figure 8. The sensor is usually mounted in a way that the area of the thin film is
perpendicular to the measurement axis. When an acceleration is applied to the
system, the vertical plane of the beam is bending due to the inertia of the mass
of the end of the film. This strain creates a piezo electronic effect on the beam,
which results in a voltage charge that can be measured across the output pin.
Figure 8. Minisense100 bending due to the vibration [12].
In the situation of the sensor is mounted off-axis of the desired above, that is
horizontally rotated through the length of the film in the x-axis or vertically around
its centre in the y-axis. The sensitivity of the sensor follows a cosine law with the
same relation in either rotation axis. A 90 degrees’ rotation in either plane will
theoretically result in a zero sensitivity, at 30 degrees the sensitivity change is -
1.25dB (87% of original), and at 60 degrees, the sensitivity is only half of the
maximum (-6 dB or 50%) [12].
12
Figure 9. Off axis response of the sensor (a) Off axis direction (b) cosine low relation of the off-axis rotation and sensitivity [12].
The sensors are designed to withstand high shock overload and can be operated
in continuous or high impact. The sensor purposely design to work in low-
frequency range, which an upper limit frequency of 42 Hz [12]. The voltage
sensitivity of the open circuit sensor is 1.1 V/g when operating below the resonant
frequency of 75Hz. The voltage sensitivity at resonance frequency is significantly
higher, which can go up to 6V/g [12]. Therefore, in practical application, high-
frequency impacts that is potentially induced by resonance could distort the
measurement and an external filter is required to remove this response. One of
the main advantages of the Minisense 100 is its excellent linearity performance.
Figure 10 shows the linearity relationship between the charge output and the
applied acceleration input, the linearity is less than 1% provided by the
manufacturer.
13
Figure 10. Frequency response and linearity of Minisense 100 [12].
The Minisense 100 can be modeled as a voltage source serially connected with
an active capacitor. The voltage output is linearly proportional with the applied
acceleration when operated at a low-frequency range. According to the
manufacturer, Minisen100 has an internal capacitance C0 of approximately 244
pF. Any external resistance connected with the sensor will form a high-pass filter.
The cut off frequency of this high pass filter can be simply calculated by the
equation
𝑓𝑐 =1
2𝜋𝑅𝐶𝑜 (7)
Figure 11. PVDF sensor model in series with load resistor R.
Figure 11 shows the relation between the lower limiting frequency or cut-off
frequency with different load resistance connections. In order to work with low
14
frequency, a direct connection with a high impedance component is
recommended. In our specific design of measuring an actual human jump, the
sensor needs to work in the frequency range between 0.25 Hz to 4 Hz, therefore
the proposed solution is to connect the sensor direct with an operational amplifier,
which has a high input impedance.
Figure 12. Relation between load impedance and the lower limit of working frequency [12].
3.2 Particle Photon Microcontroller
The photon is a low-cost IoT microcontroller provided by Particle. The board is
powered with a combination of Broadcom BCM43362 Wi-Fi chip, provides
standard 802.11b/g/n Wi-Fi connection, and STM32 120Mhz ARM Cortex M3
microcontroller [16]. The photon offers all fundamental features of a
microcontroller in addition to free access to the Particle cloud. This enables the
application to extend it physical limit to connect over-the-air with other projects,
software app. There are three main functions of the photon in this application:
converting the filter analog signal to digital, implementing the peak detection
algorithm to detect jump motion and sending the jump count on air to Blynk
software on phone.
3.3 Operational Amplifier LM358
In order to minimize the high-pass filter effect of the internal capacitance, the
sensor needs to first connect with a high impedance component, ideally an
operational amplifier. In this paper, our module uses LM358 for different circuit
purposes. The chosen Op-amp has a low input bias current (45 nA), high input
15
impedance (1 GΩ) and a rail to rail power supply feature [17]. This electrical
characteristic allows the sensor to work in our desired frequency range In our
module, the Op amp is served in the signal amplifier circuit, offset controller
circuit, and active low pass filter circuit. The LM358n package comes with two
independent Op-amp thus we need to have at least two devices embedded in the
module [18].
3.4 Pads logic and Pads layout
These are the two main PCB design software used in this projects. PADS Logics
is an easy to use software that allow user to build schematics of a circuit. PADS
Layout Is a linked software, is used to create the ready to print PCB layout from
the PADS Logics schematics.
3.5 Particle Web IDE
Particle Device cloud is cloud service provided by the Particle, this cloud enables
Particle device to communicate on-air securely and effortlessly. Each device has
a specific RSA public-private key pairs to ensure that the cloud is communicated
with the right device. The Particle cloud connection uses constrained application
protocol; sessions are securely encrypted using DTLS over UDP or AES over
TCP depend on the device.
The Particle Web IDE is a software development application running in web
browser. It communicates with the device through Particle Device cloud and thus
no direct port connection is needed.
In order to use Web IDE feature, initial setups are required. Firstly, user have to
register a free account of the Particle cloud through mobile app or the main web
IDE. The device then has to be initialized through that Particle App account in
order to set up a default Wi-Fi connection and registered to the user account.
After the setup, the device has an ownership connection with the specific account.
16
User can create, compile and embedded application directly to the specific
device, as long as it is connected to an internet available Wi-Fi. Particle Web IDE
also has a strong supported community library hosted on GitHub that can simply
include in any application.
3.6 Blynk
One of the feature of the system developed in this project is to transmit the jump
count into a mobile application. After some research of available solutions, Blynk
is chosen due to its efficiency, simplicity and low cost. Blynk is an Internet of
Things platform that can establish communication between smartphone and
hardware devices through its own server. There are three majors’ components in
this platform:
Blynk App is a mobile phone application that can help users to create their own user interfaces from the provided widgets.
Blynk Server is a private, local server operated by Blynk. Hardware devices and smartphone apps are communicated through this server.
Blynk Libraries are available in multiple device platforms, included Particle Web IDE. This library enables the hardware part to connect to the cloud and communicate with the Blynk App. The library required a key that is specific to each user, which helps the cloud to distinguish each device.
17
Figure 13. Three major components of Blynk flatform [19].
4 Physical implementation of the device
4.1 Jump Counter Device
The purpose of the project was to create a device that can be attached to a
trampoline, and count the quantity of jump as well as indicate that amount to a
smartphone when someone is jumping on the trampoline. The circuit is based on
the application of the piezo electronic sensor. Even though the piezo electronic
sensor generates a very low energy signal, it could potentially result in a high
voltage swing. Therefore, a high impedance amplifier is a must, in order to draw
significant current into it and convert the initial signal into a stable voltage signal.
Since the target of the device was to count the number of the human jumps, which
is generally low frequency, a suitable low pass filter needs to be implemented.
And then offset control is needed to broaden the signal range and thus increase
the resolution of the measurement.
4.2 Piezoelectric Sensor
First, let’s take a look at the frequency response of a typical piezo electronic
sensor. Depend on the type of the piezoelectric sensor, the resonant frequency
18
can be varied. In most piezo electronic applications, the electronic device has to
remove the resonance frequency, which can call spark and do harm to the device.
This can be done by implementing a suitable low pass filter. The measurement
value of application falls into the low-frequency range, typically equal to nature
human jump frequency, from observation, that is around 0.25Hz to 4 Hz.
Figure 14. Ideal frequency response of testing piezo electronic sensor before and after signal processing [20].
Based on preliminary observation of a trampoline when being used, it can be
estimated that jumps on a trampoline have a high impact and low frequency.
Next, the device has to be small enough to not affect the functioning of the
trampoline. To satisfy those conditions, a suitable piezo electronic sensor with a
resonance frequency of 75Hz and size of 3.0 mm x 15.0 mm was chosen. One
end of the film is attached to a metal ring that enables a free swing when there is
an external impact. Another end of the film is attached firmly to a printed wired
mini circuit board, which can be soldered to the application digital board. The
resonance frequency is provided by the manufacture is approximately 75 Hz and
thus in our purpose of measure the human jump. Figure 15 shows the picture of
the sensor we use. The length of the film and mass of the swing ring is determined
to generate a desired resonance frequency of 75Hz.
19
Figure 15. Chosen piezo electronic sensor for prototyping.
4.3 Design Architecture And Measurement Method
The device is comprised of two part: the physical module that process the raw
signal into a smoother and easy to detect peak and the software is in charge of
translating the analog signal into digital data, process that data to detect peak
and then send the peak data to smartphone application. The overall architecture
of the module is indicated in Figure 16.
Figure 16. Signal processing flow
20
Particle photon offer five 12-bit resolution ADC pins, and also the serial
communication. In order to testify the working of each module, the output of each
step is connected to pin A0 and then measurement is made. Photon will first
convert the voltage signal to the 12-bit value correspond to the scale of 0-5V.
Then a code in implement to reverse that data back into voltage value. The
convert equation is express as equation below.
𝑉𝑜𝑙𝑡𝑎𝑔𝑒 =𝑑𝑎𝑡𝑎
4096∗ 5 [𝑉]
The voltage data will be sent to Arduino Serial Plotter app through a serial
connection. The result is a real time graph of the output voltage. The graph can
be analysed to assess the performance of each module.
4.4 Circuit designing
4.4.1 Raw Piezo Electronic Signal
Minisense 100 is directly connected to pin A0 of Photon. Then the signal is
generated by finger tapping to the surface of the breadboard. This set up will give
us a general observation of the working sensor. The breadboard setup and the
result graph generated by tapping the surface is shown in Figure 17.
21
(a)
(b)
Figure 17. Directly connected setup (a) setup on breadboard (b) raw signal graph
4.4.2 Pre-Amplifier Module
The sensor signal will go through a non-inverting amplifier. Connecting a load
directly to the sensor will form a high pass filter with the internal capacitance of
the sensor. Since the MiniSense 100 behaves electrically as an active capacitor,
As showing in Figure 12, the sensor has a high impedance of 650M at 1Hz.
Therefore, in order to collect the most of the charge output from the sensor, a
high gain operational amplifier is ideal since its input is a virtual ground to the
22
sensor. The piezoelectric sensor is placed close to the high impedance amplifier,
the possible capacitance effect of interface cable (Cc) between them is generally
really low, typically less than 0.5pF for 5 mm mounted coper path. Because of
that, the voltage mode configuration was chosen due to its simplicity. Resistor
RBIAS provides a bias path for the input amplifier, this resistor is required to be
as high as possible. From the datasheet of the piezoelectric, a measurement of
QS is around 0.35 pC/G The set up, equivalent circuit and the graph of the
processed signal can be seen from Figure 18.
(a)
(b)
23
(c)
Figure 18. Setup with pre-amplifier and equivalent circuit (a) Breadboard setup (b) equivalent circuit (c) processed signal
4.4.3 Sallen-Key low pass filter
Since the use of LM358 as high impedance amplifier, a recommended value of
bias resistor of 10M Ohm is used [18]. The amplifier uses 5V rail to rail power,
thus the output will swing between 0V to 5V. After going through a high
impedance amplifier, the signal needs to filter out all of the unnecessary high
frequency. Second-order Sallen-Key low pass filter was chosen because of its
stable and high slew rate. The value of the capacitor needs to be small but not
too much to avoid build-up high voltage differences that can disturb the signal.
Therefore, an 8.2 nF capacitance was chosen and then we can calculate resistor
value through equation (4), with fc=15Hz.
𝑓𝑐 =1
2𝜋√𝑅1𝑅2𝐶1𝐶2
(4)
24
Figure 19. Sallen-Key low pass filter
The combination circuit with two modules: the pre-amplifier and the Sallen-key
low pass filter is set up on breadboard and the output signal is collected and
graph. The set up and the graph of the processed signal can be found in Figure
20.
25
(a)
(b)
Figure 20. Pre amplifier with low pass filter combination (a) the set up (b) the graph of the processed signal
4.4.4 Offset control and final gain
The filtered signal will go through an offset control circuit; the purpose was to
adjust the minimum value of the signal to the desired stage of 0 [V]. This will
enable the output signal to swing in the range of 0V – 5V. After that, a final gain
step is implemented to amplify the signal to a suitable and readable range. After
a few tests of the impact, a gain of 20 dB was chosen. The final circuit and
output signal result are shown in Figure 21.
26
(a)
(b)
(c)
Figure 21. Offset control and final gain circuit (a) the bread board set up (b) equivalent circuit of the offset and final gain circuit only (c) the signal after processed through three module.
27
4.5 Software Implementation
The system developed in this project uses Photon as a micro-controller unit to
serve as an analog to digital converter as well as a wireless communication
module. The filtered signal is connected to one analog pin A0 of the Photon. The
ADC module of Photon receives the filtered signal in a range of 0V - 5V and
converts it to a 12-bit digital signal with a sampling frequency of 100Hz. A
program embedded in Photon also implements a Pulse peak detector algorithm
to detect jumps in real-time. The software is built from Particle Web IDE.
And then the jump count data will be transmitted through wireless connection to
a Blynk app on a mobile phone and indicate the jump in real-time. In order to use
the Blynk app, the library of Blynk need to be included in the project. After the
registration, each device is assigned with an authentication keys. By implement
that key into the project source code, the Blynk cloud could be able to let the
software to communicate with the right device. The implement code can be seen
below:
#include <blynk.h> #include library
char auth[] = "Aw7kXSv1glByw-AROxoQwcWnOII9i-rC"; #authentication key
Blynk.begin(auth); #first time set up
Blynk.run(); #inside loop run
The jump count data is send to the Blynk app through a virtual port V1, that can
be done directly inside loop. A reset button and also sensitive adjust is add to the
blynk application, the user can use that to reset the jump count and adjust the
sensitivity of the pulse peak detection algorithm. The code of sending and
receiving of the Blynk can be seen below:
28
BLYNK_WRITE(V0) // This function gets called when sensity value is adjust in
the app
sensity = param.asInt();
Blynk.virtualWrite(V1,jumpcount); #assign the jump count to virtual port V1
4.6 Pulse Peak Detection
Each jump on the trampoline always results in an abnormal high impact force on
the surface. In real-time measuring, each jump is represented by a peak. The
purpose of the pulse peak detection algorithm is to detect those peaks which a
specific delay that matches the human jump frequency. The algorithm considers
a signal value as a valid peak if and only if it could satisfy three conditions:
That signal value is higher than a number of standard deviations away from some moving average value.
That signal value is higher than both the previous value and the next value.
That signal is not within 300ms after and before another valid peak.
In real-time measurement, the algorithm takes 10 latest consecutive signals to
compute the standard deviation and average value. The threshold is updated as
the sum of average value and a factor multiply of standard deviation. That factor
is a fixed value, indicates the signal power difference between human-made
impact and when human is in air. After multiple testing of the device with multiple
jump scenarios, a factor of 3.5 was chosen.
When the new signal vale is inserted into the set, the algorithm will recalculate
the mean, standard deviation, and threshold. However, after a valid peak is
detected, if that peak signal value is also inserted into the set, the new standard
deviation and mean can be significantly higher which will result in a much higher
threshold. This could potentially cause a situation when the new valid peak is
smaller than the abnormal high threshold. Therefore, whenever a valid peak is
detected, the signal value is multiplied by an influencing factor, to reduce the
effect of that peak to the mean value. In testing with the real jumps, a peak signal
29
has a value that is roughly three times higher than the mean, thus a suitable
influence of 0.33 was chosen.
When a new signal is read, the previous signal will be tested with the algorithm
process. After every valid peak is detected, a simple timer is implemented to
eliminate all other peaks which occur within 300ms. The pseudo code for the
algorithm is shown from the appendix 5.
Figure 22. Pulse peak algorithm diagram
30
5 Testing
After building the working prototype, a premier test is carried out to evaluate the
actual work. A small gym trampoline with a diameter of roughly 1 meter is used
to test. The device is attached to the reel of the trampoline and real-time data is
collected through a serial connection to the PC. The jump is arranged as naturally
as possible and in different poses for a non-stop jump of 10 seconds. The test
carries out three times with different people of different weights. The number of
jump and timing is simply counted visually since the speed is slow and quite
obvious to observe. Figure 23 shown the implemented prototype and testing
setting with a one-person trampoline.
Figure 23. Prototype with the size of 3cm x 6cm
A sample result is shown in Figure 24, data is collected at 100Hz, the window
size is 500 samples, resulting in 5 seconds of data. The algorithm detected for
peaks at the given time result a correct number of jumps in testing.
31
Figure 24. Testing result sample
The purple line indicates the real time signal collected from the hardware part.
The red line shown the detected jump.
32
6 Limitation and future possibilities
The algorithm of peak detecting has some limitations. The use of a simple timer
delay mechanism is to avoid detecting multiple peaks in a short duration.
However, this delay lowers the operation frequency of the device. The device is
unable to correctly detect jump in some circumstances e.g. small jump,
resonance of the trampoline. The jump detection would be improved by using a
neural network algorithm.
The low-frequency operation of the current device is suitable for multiple
applications. The new signal processing algorithm can be embedded into the
software in order to detect the graph's different features in different uses. For
example, the device can be used to detected washing machine load imbalance;
tracking vehicle motion.
Modification of the filter circuit would enable the working in different scenarios. A
biomedical device could potentially be developed by improving the circuit size
and filter properties. For example, increasing the final gain would enable the
device to sense tiny signals of the blood pulse or heart rate. Implementation of
SMDs would significantly reduce the size, would also help develop a comfortable
in-ear or wrist band device.
33
7 Conclusion
The goal of this project was to design a piezo electronic sensor device that can
count the number of jumps on a trampoline. In order to achieve the goal, deep
research of the related sensor has been conducted and a proposed working
circuit is implemented. To conclude the goal, the device needs to correctly detect
the jump based on the signal received from the hardware circuit.
There are various difficulties during the project, mainly in designing a working
circuit and calibrating the suitable components' value. As a result of this project,
a prototype of the system was developed and tested to function as specified.
However, there is still a variety of aspects of the device that needs an
improvement. The detection is not perfect and thus requires a better algorithm,
the device has a limited working frequency. In the end, the testing on the gym
trampoline provided very promising results, thus the prime goal was partially
achieved.
34
References
1 Aiguo Song; Yezhen Han; Haihua Hu; Jianqing Li, 2014. IEEE
Transactions on Instrumentation and Measurement. A Novel Texture
Sensor for Fabric Texture Measurement and Classification, 63(7), pp.
1739 - 1747.
2 Azoderimo, 2019. Piezo Film Sensor, Vibration, LDT Series. [Online]
Available at:
https://www.azoderimo.com/index.php?main_page=product_info&produc
ts_id=156240
[Accessed 9 8 2021].
3 Bartolome, E., 2010. Signal conditioning for piezoelectric sensors, Texas:
Texas Instruments Incorporated.
4 Blynk, 2020. How Blynk Works. [Online]
Available at: https://docs.blynk.cc/
[Accessed 9 8 2021].
5 Blynk, n.d. Intro How Blynk Works. [Online]
Available at: https://docs.blynk.cc/
[Accessed 28 5 2021].
6 Electrokit, 2017. Vibration sensor piezo right-angle. [Online]
Available at: https://www.electrokit.com/en/product/vibration-sensor-
piezo-right-angle/
7 GH, H., 1991. Buchanan RC Piezoelectric and electro-optic ceramics in
ceramic materials for electronics 2nd edition. New York: Marcel Dekker.
35
8 Instruments, T., 2014. Datasheet for LMx58-N Low-Power, Dual-
Operational Amplifiers. Texas: Texas Instruments.
9 Instruments, T., 2021. LM358 Dual standard operational amplifier.
[Online]
Available at: https://www.ti.com/product/LM358
10 Jaffe B, Jaffe H, Cook WR, 1971. Piezoelectric ceramics 1st Ed..
London: Academic Press.
11 Karki, J., 2000. Voltage Mode Amplifier. In: Signal Conditioning
Piezoelectric Sensors. s.l.:s.n., p. 2.
12 Kawai, H., 1969. The Piezoelectricity of Poly (vinylidene Fluoride).
Japanese Journal of Applied Physics, Volume 8, p. 975.
13 Liu, Huicong; Zhong, Junwen; Lee, Chengkuo; Lee, Seung-Wuk; Lin,
Liwei, 2018. A comprehensive review on piezoelectric energy harvesting
technology: Materials, mechanisms, and applications. s.l.:Applied
Physics Reviews.
14 M. G. Broadhurst, G. T. Davis, 1980. Electrets. Berlin: Springer-Verlag.
15 Mohammadi B., Yousefi A.A., Bellah S.M, 2007. Effect of tensile strain
rate and elongation on crystalline structure and piezoelectric properties
of PVDF thin films. In: Polymer testing. s.l.:s.n., pp. 42-50.
16 One, S., 2016. DAC series piezoelectric pressure sensors. [Online]
Available at: https://www.sensorsone.com/dac-series-piezoelectric-
pressure-sensors/
[Accessed 9 8 2021].
36
17 PARTICLE, 2019. PARTICLE PHOTON: WI-FI. [Online]
Available at: https://docs.particle.io/photon/
[Accessed 28 5 2021].
18 Ping Yu , Weiting Liu , Chunxin Gu , Xiaoying Cheng , Xin Fu , 2016.
Flexible Piezoelectric Tactile Sensor Array for Dynamic Three-Axis Force
Measurement. Sensors , 16(6), p. 819.
19 R.S. Dahiya, M. Valle, 2013. Fundamentals of Piezoelectricity. In:
Robotic Tactile Sensing. s.l.:s.n., pp. 199-200.
20 Robert Littrell, Ronald Gagnon, n.d. Frequency response of a
piezoelectric MEMS. In: PIEZOELECTRIC MEMS MICROPHONE
NOISE SOURCES. s.l.:s.n., p. 260.
21 Salimi, A., & Yousefi, A. A., 2003. Analysis method: FTIR studies of β-
phase crystal formation in stretched PVDF films. In: Polymer
international. s.l.:s.n., pp. 699-704.
22 Specialties, M., 2019. Datasheet for MiniSense 100 Vibration Sensor.
s.l.:Measurement Specialties.
Appendix
1 (7)
List of Figure
Figure 1. Piezo electronic effect explained (a) At rest turbed molecule (b) Material
under external force (c) Polarization effects under external force (d) Neutralized
current induce by the external force (R.S. Dahiya, M. Valle, 2013) ..................... 3
Figure 2. Common commercial piezo electronic sensors (a) Buzzer and
transducer ceramic type (b) Pressure sensor using piezo electronic crystal (One,
2016) (c) Polymeric type (Azoderimo, 2019). ...................................................... 4
Figure 3. Structure formula of CH2CF2 .............................................................. 5
Figure 4. Structure conformation of PVDF (a) α-phase (b) β- phase (c) γ-phase
(GH, 1991) .......................................................................................................... 6
Figure 5. Direction of PVDF film sensor (Specialties, 2019) ............................... 7
Figure 6. Charge mode configuration (Karki, 2000) ............................................ 9
Figure 7. Voltage mode configuration (Karki, 2000) .......................................... 10
Figure 8. Minisense100 bending due to the vibration (Specialties, 2019) ......... 11
Figure 9. Off axis response of the sensor (a) Off axis direction (b) cosine low
relation of the off-axis rotation and sensitivity (Specialties, 2019). .................... 12
Figure 10. Frequency response and linearity of Minisense 100 (Specialties, 2019)
.......................................................................................................................... 13
Figure 11. PVDF sensor model in series with load resistor R ........................... 13
Figure 12. Relation between load impedance and the lower limit of working
frequency (Specialties, 2019) ........................................................................... 14
Figure 13. Three major components of Blynk flatform (Blynk, 2020). ............... 17
Figure 14. Ideal frequency response of testing piezo electronic sensor before and
after signal processing (Robert Littrell, Ronald Gagnon, n.d.) .......................... 18
Figure 15. Chosen piezo electronic sensor for prototyping. .............................. 19
Figure 16. Signal processing flow ..................................................................... 19
Figure 17. Directly connected setup (a) setup on breadboard (b) raw signal graph
.......................................................................................................................... 21
Figure 18. Setup with pre-amplifier and equivalent circuit (a) Breadboard setup
(b) equivalent circuit (c) processed signal ......................................................... 23
Figure 19. Sallen-Key low pass filter ................................................................. 24
Appendix
2 (7)
Figure 20. Pre amplifier with low pass filter combination (a) the set up (b) the
graph of the processed signal ........................................................................... 25
Figure 21. Offset control and final gain circuit (a) the bread board set up (b)
equivalent circuit of the offset and final gain circuit only (c) the signal after
processed through three module. ..................................................................... 26
Figure 22. Pulse peak algorithm diagram ......................................................... 29
Figure 23. Prototype with the size of 3cm x 6cm ............................................... 30
Figure 24. Testing result sample ....................................................................... 31
Appendix
3 (7)
Main code
#include <blynk.h>
#define BLYNK_PRINT Serial
#include <blynk.h>
#include <math.h>
char auth[] = "Aw7kXSv1glByw-AROxoQwcWnOII9i-rC";
int led = D7;
int timer = 0;
int analogPin = A0;
float past_value=0;
float curr_value = 0;
float next_value = 0;
int data_length = 10;
float data_list[10];
float curr_mean = 0;
float curr_std = 0;
float threshold = 0;
float influence = 0.75;
int sensity = 100;
int delayms = 200; //appriximate 600ms
int jumpcount = 0;
float mean(float data[], int len); //calculate the average value of most
recent 10 samples data
float mean(float data[], int len)
float sum = 0.0, mean = 0.0;
int i;
for(i=0; i<len; ++i)
sum += data[i];
mean = sum/len;
return mean;
float stdd(float data[],float mean, int len); //calculate the standard
deviation value of working data series
float stdd(float data[],float mean, int len)
float standardDeviation = 0.0;
int i;
for(i=0; i<len; ++i)
standardDeviation += pow(data[i] - mean, 2);
return sqrt(standardDeviation/len);
void insert(float data[], float value, int len); //insert new data to the
working data series
void insert(float data[], float value, int len)
int i;
for(i=0;i<len-1;i++)
data[i]=data[i+1];
data[len-1]=value;
int ispeak(float past_point, float current_point, float next_point, float
threshold); //check if this is peak
Appendix
4 (7)
int ispeak(float past_point, float current_point, float next_point, float
threshold)
if
((current_point>=past_point)&&(current_point>=next_point)&&(current_point>thre
shold))
return 1;
else
return 0;
BLYNK_WRITE(V0) // This function gets called when sensity value is adjust in
the app
sensity = param.asInt();
void setup()
Serial.begin(9600);
Blynk.begin(auth);
//initialize the first 10 sample signal
int i;
for(i=0;i<data_length;i++)
data_list[i] = analogRead(A0); //read from pin A0
past_value = data_list[data_length-2];
curr_value = data_list[data_length-1];
curr_mean = mean(data_list,data_length);
curr_std = stdd(data_list,curr_mean,data_length);
//led signal indicates finish setup
pinMode(led, OUTPUT);
digitalWrite(led, HIGH);
delay(1000);
digitalWrite(led, LOW);
delay(1000);
void loop()
next_value = analogRead(A0); //read from pin A0
//printout serial data
Serial.print(next_value/4095*5);
Serial.print(",");
Serial.print(threshold/4095*5);
Serial.print(",");
Serial.print(delayms/4095*5);
Serial.print(",");
threshold = curr_mean + curr_std+sensity;
if (ispeak(past_value,curr_value,next_value,threshold)==1)
if(timer == 0)
jumpcount++;
Blynk.virtualWrite(V1,jumpcount);
timer =1;
curr_value = curr_value * influence;
if (timer == 1)
Appendix
5 (7)
Serial.println(1);
timer++;
else if ((timer>1)&&(timer<delayms))
Serial.println(-1);
timer ++;
else if (timer == delayms)
Serial.println(-1);
timer = 0;
else
Serial.println(-1);
insert(data_list,next_value,data_length);
curr_mean = mean(data_list,data_length);
curr_std = stdd(data_list,curr_mean,data_length);
past_value = curr_value;
curr_value = next_value;
delay(1);
if(digitalRead(D7)==HIGH)
jumpcount = 0;
Blynk.virtualWrite(V1,jumpcount);
Blynk.run();
Appendix
6 (7)
Schematic of the circuit
Appendix
7 (7)
PCB layout of the circuit