On-Demand Bridge Monitoring using EcoSense

(1)

TVE-F 17 026 juni

Examensarbete 15 hp Juni 2017

On-Demand Bridge Monitoring using EcoSense

Scavenging transient vibrations induced by passing train to activate sensor node Lucas Wassénius

Daniel Stenbacka

(2)

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0

Postadress:

Box 536 751 21 Uppsala

Telefon:

018 – 471 30 03

Telefax:

018 – 471 30 00

Hemsida:

http://www.teknat.uu.se/student

Abstract

On-Demand Bridge Monitoring using EcoSense

Lucas Wassénius and Daniel Stenbacka

The project aim is to evaluate if vibrations, in a train bridge, can be used in an on-demand bridgemonitoring system, for vibrational stress, using a technique called EcoSense. The EcoSense

architectureis used to reduce the power consumption of sensors by only allowing them to use power when a desiredmeasurement is wanted (e.g. a train passes). The main task is to analyse how such a system can bebuilt to perform as efficiently as possible. The system should be tested at Lidingö bridge in Stockholm.The resulting circuit was successful to be used as an on-demand switch for a sensor on Lidingö bridge.Testing and evaluation of different specifications of the circuit was done. The final conclusion is thatEcoSense is applicable for bridge vibrational sensing and designs can be tailored to fit the specificationsof a bridge.

Ämnesgranskare: Gopi Tummala Handledare: Ye Liu

(3)

Popular Science Summary - Energy Efficient Bridge Monitoring

Bridges are known to have stability problems because of vibrations. Therefore the need to monitor and correlate how vibrations and tension in bridges are connected to the damage in the structure. The vibrational stress sensors are often in remote locations and need to be energy efficient. Therefore there is a

need that the sensors only consumes energy (i.e. the sensor is turned on) when a train passes by. A possible solution would be to harvest energy from vibrations of a passing train and use that energy to turn

on the sensor. By using an on-demand wake-up design, the power consumption of sensors could be reduced. A design for the switching unit was constructed and adjusted for a field test on Lidingö bridge.

The field test succeeded and the sensor turned on by just using the vibrations from the train. A great result for remote sensing.

(4)

1 Introduction

With today’s technology, internet of things is growing by the minute. The ability to be connected to internet is an attractive feature for many producers today. More and more laptops, phones, clocks and all kind of sensors are constructed to be mobile and be able to connect to the internet everywhere. One of the biggest problems of today, with staying mobile and connected, is the power management. There is no technology yet for long distance wireless charging. Therefore when the unit is discharged it needs to be connected via cable which not always can be found. This is a rather large problem for sensors which are deployed in remote places. One of these situations is when monitoring vibrations and stress in bridges.

The aim of this project is to evaluate the EcoSense [1] architecture for a train bridge monitoring system. The goal for the whole project is to see the possibility of an EcoSense switch on the Lidingö-bridge in Stockholm.

Since the sensor network will be in a remote and dangerous location, it is required to be as self-sustaining as possible. To achieve this there is a need for power management. EcoSense is a great solution to the power management problem if it can be applied to this network.

EcoSense is a hardware architecture that imposes low-energy consumption techniques to sensors. Instead of using traditional means, like duty-cycles and adaptive sampling, to lower energy consumption EcoSense uses on-demand wake up. The system will only collect and transmit data during desired events and this is done by having the power supply turned on only when a harvesting unit gets enough energy. The source by which the harvesting unit collects energy can be chosen to be the same event that is being measured. In this project EcoSense will be used to measure vibration in a railway bridge and therefore the harvesting unit can use the railway track’s vibrations.

The EcoSense architecture consists of five units, shown in figure 1. There is a power supply unit, an on- demand connection unit, a sensing, a processing, and a communication unit. The on-demand connection unit uses a switch to connect the sensor units (sensing, processing, communication unit) to the power supply unit. To close the switch a capacitor will be charged from the harvesting unit. When the charging capacitor reaches a threshold the switch closes. In order to keep the switch closed, while the measurements are taken and transmitted, there might be a need for a storage capacitor so the switch does not oscillate. In order for the sensor units to work properly there is a requirement on the reaction time and work duration. The reaction time should be short enough that the sensor unit can measure the vibrations and transmit the data from the passing train and the work duration should fit the train passing by so that no energy is wasted and enough measurements are taken.

Figure 1: A simplified architecture of EcoSense. [1]

(6)

Some questions to be answered: will the energy harvester get enough energy from the vibrations to get the wanted working duration and reaction time? How are the components chosen to achieve the best possible properties?

2 Theory

This section will explain all the necessary theory to fully understand the project. The different component blocks that have either been under consideration or used in the circuits are explained in this section.

2.1 Rectifier

Most commonly a signal is using alternating current (AC) because this has lower energy loss in power lines.

Most machines however use direct current (DC) and to covert AC to DC one could use a rectifier. There are different kinds of rectifiers, a commonly used rectifier is the diode bridge rectifier, see figure 2 for a schematic of a full-wave diode bridge rectifier. All diodes have a certain threshold voltage for when they become forward biased. A normal value for the forward voltage is 0.7 V, [2] and for Schottky diodes a normal value is 0.25 V. [3] For low-power applications there might be a need to be able to start up the circuits at a lower voltage. Then one could consider using a transistor rectifier, these generally have a lower threshold to become conducting. In most cases there is a capacitor between the output and ground in order to get a more stable voltage.

Figure 2: A schematic of a full-wave diode bridge rectifier with a smoothing capacitor.

2.2 Buck Converter

A buck converter or a step down converter is used in electronics to lower the voltage and increase the current in a circuit [5]. The buck converter is a switching power supply which means that the output is regulated by a switch which can be turned on in either a synchronous or asynchronous mode. This output will be regulated using semiconductors and a storage element. The semiconductors can be either diodes or transistors and different set ups with these can be used. The storage element can be a capacitor or an inductor or the two together. The buck converter is an efficient power supply and can reach an efficiency of 90 %, this is much greater than linear regulators.

By controlling when the switch is closed the unregulated DC input voltage can produce a regulated DC output. When the switch is closed the storage elements starts to charge until the switch opens. The charged storage elements will then keep the output voltage at the regulated level. If the switching frequency is high

(7)

enough the voltage drop from the discharging of the storage elements will be negligible. The output voltage is controlled by adjusting the on and off time for the switch.

2.3 Voltage Multiplier

A voltage multiplier is used in circuits to boost the amplitude of a signal. The signal can be either AC or DC, which there are different voltage multiplier solutions respectively. One solution for AC to DC multiplier is the Cockcroft-Walton charge pump. The Cockcroft-Walton consists of two capacitors and two diodes per stage, see figure 3 for a two stage circuit design. Using more stages results in a higher output voltage. The output voltage is given by Vout = 2N · Vin, where N is the number of stages and Vinis the amplitude of the incoming AC signal.

Figure 3: A schematic of a two stage Cockcroft-Walton charge pump.

Another type of voltage multiplier is the Dickson charge pump. This is a modification of the Cockcroft- Walton multiplier. There are still two capacitors and two diodes but arranged differently, see figure 4 for a two stage Dickson charge pump. In most Dickson charge pumps there are two synchronous clocks that allows charging in different capacitors at different times. It is possible to use AC and DC input signals for this design. Using the design in figure 4 type removes the need for clock signals and limits the input to AC. The Dickson charge pump has the same voltage multiplication as the Cockcroft-Walton multiplier, Vout = 2N · Vin. If Schottky diodes is used in the configuration, the charge pump will work for low input voltages.

Figure 4: A schematic of a two stage Dickson charge pump.

(8)

2.4 Transistor as Switch

A transistor has properties that make them ideal as a switch. [4] A transistor allows a small current to control a much larger current. There are two different types of transistors, one is the bipolar junction transistor (BJT) and the other is the field-effect transistor (FET). Both can be used as a switch but the FET has a lower threshold voltage to enter the saturation region (on-state). For low power applications this often the preferred behaviour. The signal, Vsig, at the gate, when implementation of a FET, can control if there will be any potential drop over the load resistor. The transistor switch circuit is illustrated in figure 5.

Figure 5: A schematic of a NMOS transistor as a switch.

2.5 Energy Harvesting Unit

An energy harvester is a sensor that act as a voltage source when the sensor can extrapolate energy from its environment. An example is a solar panel, this will harvest energy from the sunlight and convert it into electrical energy. This will only be active during the sunlight hours. The energy harvester that is needed in this project is a vibrational sensor, which converts kinetic energy to electrical energy using electromagnetic induction.

2.6 Simulation

Electronic circuits can be simulated to test how they function in reality. Simulations are effective to find what properties the components need to have in order to get the wanted behaviour. Simulations can also be used to find the limiting factors of circuits. One simulation program is called PSpice from the company Cadence. Cadence is known for creating many programs that facilitate simulations. PSpice allows graphical circuit building and has a large default component library.

There are multiple options when running simulations, the two used in this project are the DC-sweep and transient analysis. The transient analysis is a time dependent simulation that calculates all currents, voltages and power in every node for every time step taken. The start and stop time can be configured. The DC-sweep is a time invariant simulation allows DC-voltages to be varied. If a component model is not included in the default library it can be imported by downloading a model file that is supplied by most manufacturers of components.

PSpice allows voltage sources to be controlled by external text (.txt) files by using a component in the Sources library called V P W L_F ILE. The component can be linked to a text file on the computer.

(9)

3 Component List

In the component list section the most specific components, that are used in the project, are listed and explained to give a further understanding of why they are used.

3.1 Evaluation module for BQ25570

An evaluation module (EVM) [7] for the BQ25570 chip [6] from Texas Instruments, shown in figure 6, can be used to control the output voltage. The BQ25570 chip contain several important features that can be used for energy harvesting. Two of the main features is the buck stage and voltage multiplier stage which are described in the theory. Another feature of the chip is the storage which gives the chip the possibility of a quick start up mode. If energy is stored from earlier signals then the energy needed for the next start up is less. The chip also has a battery which is used to regulate the output signal to always have a constant voltage.

Figure 6: A picture of an evaluation module of the BQ25570 nano power boost charger and buck converter for energy harvester powered applications from Texas

Instruments.

The whole evaluation module is constructed to specify the values for the chip to match a harvesting unit.

The lowest cold start voltage, with the EVM used, is 0.33 V and if the storage is charged the lowest start up voltage is 0.13 V. The output voltage of the EVM gives a DC-signal of 1.8 V when the battery is charged.

3.2 Schottky Diode

A Schottky diode is a semiconducting diode that has a low forward voltage drop. This is needed when using energy harvesting units as the voltages in the circuit are low. The two different kinds of Schottky diodes used in this project are the SMS7630-079LF [8] from Skyworks Solutions and the HSMS-2860-TR1G [9] from

(10)

Farnell. The SMS7630 diode has a forward voltage between 135 mV to 240 mV at 1 mA. The HSMS-diode has a forward voltage between 250 mV to 350 mV at 1 mA.

3.3 MOSFET

The MOSFET used in the project is a transistor with a low threshold voltage from Nexperia called BSH105 transistor. [10] The low threshold voltage has a typical value of 0.57 V when the drain current is 1 mA. It also has quick switching which is needed for its purpose.

3.4 Energy Harvester

A company called ReVibe Energy construct and sell vibrational energy harvesters. RevibeEnergy’s model D [11]

is shown in figure 7. The harvester use electromagnetic induction to generate an electrical signal by an accel- erating magnet inside a coil. The magnitude of the signal is dependent on the acceleration, both magnitude and frequency.

Figure 7: The model D vibrational sensor from ReVibe Energy.

4 Method

There was several parts to the project and in this section the methods and reason for them will be explained.

The first part was to simulate all components. After that construction and field testing was done. Lastly some complimentary simulations helped to finalize the circuit. Some methods depend on the results from earlier methods.

4.1 Simulation

Multiple simulations have been done during this project and this section will cover what circuits were analysed and the settings used.

To get the most realistic results from the simulations, data from the ReVibe harvester was sampled using an oscilloscope. The oscilloscope test probe was connected to the plus output of the harvester and the oscilloscope probe’s ground to the minus output of the harvester. The harvester was placed on a robust table and the table was forced into vibration by hitting the table. This data was sampled and exported to MATLAB. In MATLAB the data can be modified freely (e.g. changing the amplitude) repeating the data to get

(11)

more pulses, scaling the time axis to get higher frequency. A graph of the sampled pulse repeated with a frequency of 427 ms is shown in figure 8.

0 200 400 600 800 1000 1200 1400 1600 1800 2000

Time(ms) -1.5

-1 -0.5 0 0.5 1 1.5

Voltage(V)

Figure 8: Five periods of a signal gathered with the energy harvester repeated with a frequency of 427 ms.

4.1.1 Simulation of Voltage Multipliers

The simulation is a comparison between the Dickson charge pump and the Cockcroft-Walton charge pump.

All diodes used are Schottky diodes of the type HSMS-2860-TR1G. The capacitors used are from the default analog library. The 2-stage versions of the Cockcroft-Walton and Dickson are shown in figure 3 and 4. In the simulations the capacitor values and the amount of stages were varied. The voltage source is always configured as a sinus-wave with an amplitude of 0.2 V but with varied frequency. The frequency is set to 10 Hz or 20 Hz. The output signal is measured after the Nth stage if the number of stages in the simulation is N. The simulation was set to transient and the end time set to 10 seconds.

Another simulation was made using the one stage Dickson charge pump with the same type of Schottky diodes. The input voltage source was of the type V P W L_F ILE and was linked to the sampled vibration pulse data file. The simulation was run with varied capacitor values, 10 nF, 0.1 µF, 4.7 µF and 50 µF. The simulation was set to transient and end time set to 10 seconds. This was done to compare the output signal’s magnitude depending on the capacitances of the charge pump.

(12)

4.1.2 Simulation of EVM

The BQ25570 evaluation module (EVM) uses a chip called BQ25570 and is made by Texas Instruments.

There is a PSpice model available for the chip which is used in the simulation, the rest of the EVM circuit is built as the documentation shows. All resistors, inductors and capacitors are picked from the default library.

On VIN a voltage source of the type VDC is connected with the amplitude of 1.2 V. The simulation was set to transient and the end time was set to 20 seconds. The simulation was done to understand the EVM further.

4.2 Circuit construction

When the initial simulations were done the hardware evaluation started. The simulations do not take all factors into consideration. When briefly understanding the components function and properties, the construction of the circuit was initialized.

The work progression during the construction was to build all different solutions and try and measure the output both with a DC-source as input and with the harvester as input. When using the harvester as input, it was placed on a table that was forced into vibration by hitting it lightly. To get measurement from the designs, with the harvester as input, an oscilloscope was connected. The oscilloscope observed the input and output from the charge pump, and the potential difference, between gate and drain, in the transistor. To make sure that the switch was working a test sensor and two 1.5 V batteries in series were connected between the drain and source of the transistor. The sensor was configured to turn on a LED when the voltage input was higher than 2.8 V. For this stage of the circuit construction the SMS-diodes were used.

4.2.1 Circuit design 1

The first construction consisted of a few components. The harvester was the first component of the design.

The harvester was connected to the input of a one stage Dickson charge pump which rectified the signal. The charge pump had capacitors with 4.7 µF capacitance and the diodes used were the SMS-diodes. The output of the charge pump was connected to node Vstor of the EVM power module. This was used to store the signal, to get a longer working duration. The output from the EVM was the node Vbat. The last component was a BSH105 FET-transistor which acted as a switch. The output of the EVM, Vbat, was connected to the gate of the transistor. The schematic is shown in figure 9.

(13)

Figure 9: A schematic of circuit design 1 with a load resistor connected to the drain of the transistor.

4.2.2 Circuit design 2

The second circuit design was similar to the first design. The energy harvester was connected to a one stage Dickson charge pump which rectified the signal. The charge pump had capacitors with 4.7 µF capacitance and the diodes used was the SMS-diodes. The output of the charge pump was connected to the capacitor parallel to the gate and source of the BSH105 transistor switch. The capacitor acted as a storage and had a value of 200 µF. The schematic is shown in figure 10.

Figure 10: A schematic of circuit design 2 with a load resistor connected to the drain of the transistor.

4.3 Field Testing

One of the circuit designs was to be used in a field test on Lidingö bridge. Circuit design 1 was chosen as the EVM has a higher overload tolerance. The test gave the opportunity to measure and record the harvester’s output signal from the train vibrations and to test how the circuit responded to the input signal.

(14)

To test if the gathered energy from the harvester was sufficient to close the switch; a test sensor and power supply was connected to the circuit in between the transistor’s drain and source. The test sensor had a LED that would flash when the transistor was in the saturation region. The purpose of the field test was also to see which positions of the energy harvester were viable. Three train passes were recorded with an oscilloscope. Three different position of the energy harvester was tested, one for each recording. One test had the harvester positioned on wooden platform close to one of the main cross beams of the bridge, referenced as harvester position one. Another position for the harvester was on the foundation of a metal fence, referenced as harvester position two. The last test the harvester was positioned on the wood platform close to the rails, referenced as harvester position three. The three different positions are photographed and shown in appendix B.

4.4 Complimentary Simulation

Complimentary simulations are done in order to improve the constructed model using the field test data. The simulations tested the Dickson charge pump similar to the earlier voltage multiplier simulations (changing capacitor values and amount of stages) but using a text file, containing the field test data from the harvester, as input. The simulation end time was set to 14 seconds.

5 Result and Discussion

The results from the method is explained and discussed in this section. The discussion is to help understand all the following results.

5.1 Simulation

There were several results from the primary simulations. The result from the simulation between the different voltage multiplier solutions is shown in table 1. The table shows the different charge times for the voltage multipliers and also the voltage at 10 seconds. This simulation used a 0.2 V sine-wave input with frequency of 20 Hz and capacitor values used for Cockcroft was 10 µF and for Dickson 0.1 µF.

(15)

Table 1: Results from simulation comparison between the Dickson charge pump and the Cockcroft-Walton charge pump and different amount of stages.

Type Charge Time

0.2 V [s]

Charge Time

0.4 V [s] Voltage at 10 s [V]

One stage Dickson 0.10 - 0.29

Cockcroft-Walton 2.74 - 0.26

Two stage Dickson 0.14 0.40 0.58

Cockcroft-Walton 0.40 1.54 0.49

Three stage Dickson 0.14 0.40 0.88

Four stage Dickson 0.14 0.44 1.15

Five stage Dickson 0.14 0.44 1.43

The Dickson charge pump is considerably quicker to charge than the Cockcroft-Walton charge pump, es- pecially the one and two stage to reach 0.2 V. For all stages the Dickson charge pump reaches a higher voltage at 10 s. A simulation using the same setup except the sine signal being 10 Hz is shown in table 2 (see appendix A). It also shows that the Dickson charge pump both has a lower charge time and higher voltage at 10 s for all stages in a similar way to the 20 Hz simulation. The Dickson charge pump is more suited for this project than the Cockcroft-Walton charge pump.

The results from the one stage Dickson charge pump with varied capacitance values can be seen in figure 11.

(16)

0 50 100 150 200 250 300 350 400 450 500 Time[ms]

-2 -1 0 1 2 3 4

Voltage[V]

Input Signal 10nF 0.1 F 4.7 F 50 F

Figure 11: The graph shows the simulation results of the output voltages using the one stage Dickson charge pump. The simulation was done with different capacitor values and the same input signal which is also shown in the graph.

The x-axis is set to 0 ms to 500 ms which includes two periods of the input signal. The 10 nF discharges quickly and will therefore never be able to boost the input signal unless the signal has higher frequency.

Both the 0.1 µF and the 4.7 µF charges quickly and has higher magnitude than the input signal, while the 50 µF is a bit slower to reach higher magnitudes. However the higher the capacitance value, the slower it will discharge. Which is helpful if the frequency of the input signal is low. Using the HSMS-diode a value between 0.1 µF to 4.7 µF is suitable choice of capacitor.

A simulation for the Dickson charge pump with SMS-diodes and capacitors with 0.1 µF is done to show how the capacitors discharge in between periods. The simulation is shown in figure 12.

(17)

0 50 100 150 200 250 300 350 400 450 500 Time[ms]

-1.5 -1 -0.5 0 0.5 1 1.5

Voltage][V]

Output Voltage Input Voltage

Figure 12: The graph shows the simulation results of the output voltages using the one stage Dickson charge pump. The capacitors had 0.1 µF and SMS-diodes

was used in the simulation.

When the charge pump has capacitors with too low values the capacitors will completely discharge before the next signal comes. Therefore the capacitor should be chosen so the discharge, between the peaks of the signal, is slow. To manage this, the frequencies of the peaks need to be considered.

Comparing figure 11 and figure 12, the difference between the HSMS-diodes and SMS-diodes can be seen.

When the Dickson charge pump uses HSMS-diodes in figure 11 the discharging of capacitors are slower than the discharging with the SMS-diodes, shown in figure 12. This is due to that the HSMS-diodes having a higher backwards voltage. Higher backwards voltage make the current away from the capacitors less. The HSMS-diode also have a higher forward voltage which make the charging of the capacitors slower. Therefore both can be applicable to use but need to be managed with different capacitors.

The result from the simulation of the EVM module, with a constant DC voltage as input, is shown in figure 13. In the simulation the input was connected to Vin and several nodes are observed.

(18)

0 50 100 150 Time[ms]

0 1 2 3 4 5 6 7

Voltage[V]

V_in V_out V_bat V_stor

Figure 13: The graph shows the simulation results of the EVM with 1.2 V input voltage using Vinon the EVM module and the output is taken on several nodes.

The x-axis is set to 0 ms to 150 ms because circuit reached a steady state after 60 ms. From the graph in figure 13 it was noticed that the EVM module charges the Vstor and Vbat nodes before Vout gives a signal.

Therefore the option of using Vstor or Vbat as output became a viable option. The reason why a different setup might be needed was because the input is restricted in power and the charge time would increase when the input is changed from a DC-generator to the vibrational harvester. Another discussion was if the input node even could be switched to be Vstorto decrease the charge time even further.

5.2 Construction

When circuit design 1 was used, with an input voltage of 1.2 V at Vstor, an oscilloscope was used to measure several nodes on the EVM. The nodes measured were, Vin, Vbat, Vstor and Vout and can be seen in figure 14.

The oscilloscope data was exported as a comma separated file (.csv) and imported to MATLAB to make the graph.

(19)

0 5 10 15 20 25 Time[ms]

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Voltage[V]

Vin Vout Vbat Vstor

Figure 14: Oscilloscope data using the EVM-module with an input signal of 1.2 V at Vstor.

The two oscilloscope probes connected at Vinand Voutshowed 0 V during the whole measurement. The node Vout did not give any voltage as Vbat did not reach its threshold voltage. The DC power supply was turned on at 9 ms and Vbatis charged close to 90 % of Vstorat 15 ms. If this result is compared to the rejected EVM design, see figure 15, the charge time for Vbat as output is shorter than using Vout as output. One problem with circuit design 1 was that it did not use the boost-stage of the EVM. Though, this was not an issue because the signal would be boosted using the Dickson charge pump. Another problem is that the design does not have a regulated output. Which means that the components could be overloaded and damaged if the output voltage from the Dickson charge pump is too high.

Two circuit designs were chosen and the difference between them is the storage. The storage in design 1 was the EVM and for design 2 it was a capacitor. The difference between these is that the EVM uses larger capacitors and some diodes to hold the signal for a longer time while only the capacitor can have a smaller value though cannot hold the signal as long. The problem with the EVM’s large capacitance is that the output load from the harvester becomes larger so that the output signal decreases in amplitude. The problem with just the capacitor with lower value is that it keeps the signal for about 3 minutes in comparison to 5 minutes with the EVM.

5.3 Rejected designs

There were many different designs that were tested during the construction phase. This section describes which solutions were tested and why they were rejected.

The first circuit solution that was tested was one using the EVM power module. The same setup as in circuit design 1 was used except for the input and output from the EVM. In this design the input node to the EVM was Vinand the output node was Vout. The favourable thing about this design was that the output was constant. Specifically, the EVM was designed to give out a 1.8 V signal when the storage and battery reached a threshold voltage. However, a FET normally reaches the saturation region at a lower voltage than

(20)

1.8 V. Therefore the output is unnecessary slow, see figure 15. Another problem with this design is that the node Vinhas a high input resistance, and the harvester signal is reduced to less than half the magnitude.

Figure 15: Oscilloscope data using the EVM-module with an input signal of 1.2 V at Vin. The y-axis shows voltage and is set to 1 V/DIV. The x-axis shows time and is set to 20 ms/DIV. The signals are the following: 1 is Vstor, 2 is Vin, 3

is Vbat, 4 is Vout.

Another rejected design was using a diode bridge instead of a voltage multiplier to rectify the AC-signal from the harvester. The problem using the diode bridge was that if the vibrations were not strong enough, the transistor switch would never turn on. The reason is because the diode bridge does not amplify the signal and uses four diodes with a forward voltage of about 0.3 V.

During the construction part we tested using different capacitors on the Dickson charge pump. Using the one stage and two stage Dickson charge pump with the SMS-diodes, the capacitor values were varied between 0.1 µF, 1 µF, 4.7 µF, 10 µF, and 50 µF. The capacitors with lower values, 0.1 µF and 1 µF, had the problem that the input needed higher frequency for the charge pump to boost the signal. This was because the capacitors discharged between the periods of the harvester. The two larger capacitors, 10 µF and 50 µF, had the problem that they reduced the output voltage because of the high load. Therefore they needed a higher voltage input and was rejected to the design.

5.4 Field Testing

Figure 16 shows the measurements collected at harvester position one, when a train passed by. It was recorded with an oscilloscope for 14 seconds and the probes were connected to the harvester output, the Dickson charge pump output and the EVM output.

(21)

0 2 4 6 8 10 12 14 Time[s]

-1.5 -1 -0.5 0 0.5 1 1.5 2

Voltage[V]

Harvester Output Charge Pump Output EVM Output

Figure 16: The figure shows the recorded nodes from the field test where the harvester was placed at position one. The three nodes which were observed are the output from the energy harvester, the output from the Dickson charge pump

and the output from the EVM power module.

In figure 16, it is shown that the harvester starts producing a higher voltage at around 4 s which indicates that the train is close to the harvester. When the train is close, the Dickson charge pump starts charging and the EVM output follows quickly as well. Within a second from when the train got close to the sensor the EVM output is large enough to turn on the transistor switch which the test sensor LED also indicated.

When the train had driven past the sensor, the charged capacitor held the switch closed for several minutes.

which is enough time for a sensor to transmit all values.

Similar to figure 16, the result for harvester position two was recorded. The same measurements were taken and is shown in figure 17.

(22)

0 2 4 6 8 10 12 14 Time[s]

-1.5 -1 -0.5 0 0.5 1 1.5 2

Voltage[V]

Figure 17: The figure shows the recorded nodes from the field test where the harvester was placed at position two. The three nodes which were observed are the output from the energy harvester, the output from the Dickson charge pump

At harvester position two, the signal from the energy harvester was lower than harvester position one. The reason could be that the fence was further away from the train and more stable. Even though the voltage was lower than position one, it was enough to charge the gate of the transistor to close the switch which the test sensor LED indicated. It was a bit slower than harvester position one, though still successful.

The last harvester position was measured using the same method as the other positions. The result for harvester position three is shown in figure 18.

(23)

0 2 4 6 8 10 12 14 Time[s]

-1.5 -1 -0.5 0 0.5 1 1.5 2

Voltage[V]

Figure 18: The figure shows the recorded nodes from the field test where the harvester was placed at position three. The three nodes which were observed are the output from the energy harvester, the output from the Dickson charge pump

The result from the harvester position three was not as successful as the other two positions. The wooden deck close to the railway did not give a high enough potential for the transistor switch to turn on. This was because the wooden deck did not have as high acceleration peaks as the other positions. In position three the harvester gave a signal with more constant magnitude then the other two positions.

The three different field tests results show similar results they all charge the capacitor to the maximum voltage and then keep the signal for several minutes. The field test was successful as two of the tests managed to fulfil the task. The improvement that could be done is to find a way for the switch to activate earlier. The transistor switch reached its threshold voltage at approximately 0.8 V because of the high current through test sensor. If the current through the sensor is lowered the threshold voltage for the switch will also decrease. A solution to the problem is to have more stages in the Dickson charge pump to increase the voltage gain. Though increasing the number of stages will increase the time to charge the capacitor and the capacitance load of the circuit. Therefore there is a trade-off between time for the storage to charge and higher voltage for the switch.

5.5 Complimentary Simulations

A simulation using a one stage Dickson charge pump and capacitor values of 10 nF, 0.1 µF, 4.7 µF and 50 µF.

The result using harvester data from position one is shown in figure 19.

(24)

0 2 4 6 8 10 12 14 Time[s]

-1 -0.5 0 0.5 1 1.5 2 2.5 3

Voltage[V]

Figure 19: Simulation of the one stage Dickson charge pump (open circuit) using capacitors with different values. The input signal is field data from the energy

harvester at harvester position one.

Just as in the simulation using input data of the table forced to vibrate it is clear that the 10 nF capacitor is discharging too quickly. The 0.1 µF capacitor is charging to a higher voltage and also has a longer discharge time than the 10 nF capacitor. The 4.7 µF capacitor has lower max voltage but a much longer discharge time and the 50 µF capacitor has the same behaviour but even lower max voltage and longer discharge time. A suitable capacitor value using diodes of the type HSMS for harvester position one is in the range 0.1 µF to 4.7 µF. In figure 20 the results from the same simulation is shown using field test data at harvester position two.

(25)

0 2 4 6 8 10 12 14 Time[s]

-1 -0.5 0 0.5 1 1.5 2 2.5 3

Voltage[V]

Figure 20: Simulation of the Dickson charge pump (open circuit) using capacitors with different values. The input signal is field data from the energy

harvester at harvester position two.

Figure 20 shows similar results as figure 19, the 10 nF capacitor is discharging too quickly and the higher value chosen the longer the discharge time. But choosing too high value means that the maximum amplitude decreases. Even though the 4.7 µF capacitor almost reaches the same maximum amplitude as the 0.1 µF capacitor it is a bit slower to charge. The most important property of the Dickson charge pump is to reach a high peak voltage early when the train comes. The reason for this is because the switch needs to turned on as soon as the train comes, otherwise important measurements are lost. The 50 µF capacitor has the same behaviour as at harvester position one; low maximum peaks. The results from harvester position two supports that a suitable capacitor is in the range 0.1 µF to 4.7 µF.

A simulation of the increment of stages was done. The circuit consisted of the Dickson charge pump with one, two or three stages. The sampled harvester input for harvester position one and two was used in the simulation. The charge pump used capacitors with 4.7 µF and the HSMS-diodes. Both of the results is shown in figure 21 where (a) is harvester position one and (b) is harvester position two.

(26)

0 2 4 6 8 10 12 14 Time[s]

-2 0 2 4 6

Voltage[V]

(a) Harvester Position One Input Signal

One Stage Two Stage Three Stage

0 2 4 6 8 10 12 14

Time[s]

-2 0 2 4 6

Voltage[V]

(b) Harvester Position Two Input Signal

One Stage Two Stage Three Stage

Figure 21: Simulation of the Dickson charge pump (open circuit) with different number of stages. The input signal is field data from the energy harvester at different positions. In graph (a) the harvester is at position one and in (b) at

harvester position two.

The result in figure 21 shows how the number of stages increase the output from the charge pump by approximately N-times where N is the number of stages. It also indicates how the boost is dependent on the input voltage from the harvester. The reason why the simulation does not correspond directly to the real circuit is because the input signal is equal independent on the number of stages. If the test is done with the energy harvester, the output from the harvester would decrease with the number of stages. This is because if the number of stages increase, the input impedance of the Dickson charge pump is increased. Therefore, to boost the signal more stages of the charge pump can be used, but the load of the stages is needed to be considered. If more stages are implemented the capacitor will need to have lower capacitance for the load to be constant.

(27)

5.6 Evaluation and Further Development

The design tested on Lidingö bridge succeeded to close the transistor switch. Therefore this is an applicable design. The design can still be improved to turn on the switch earlier and also a way of opening the switch when the measurements are completed and transmitted. Another improvement to the circuit is to implement a voltage regulator in order to protect the circuit from overloading.

An example of a working design for Lidingö bridge is shown in figure 22. The circuit is constructed of a one stage Dickson charge pump with 4.7 µF capacitors and HSMS-diodes and a storage capacitor of 10 µF.

The figure illustrates the measurements from the oscilloscope. The blue curve in the graph show the output from the energy harvester and the red curve show the voltage over the storage capacitor, which is the same as VGS for the transistor switch. This design is fast to charge and can keep the transistor closed for about two minutes.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Time[s]

-2 -1 0 1 2 3 4

Voltage[V]

Harvester Signal Storage Voltage

Figure 22: The oscilloscope result from one example of the final construction.

The construction contained a one stage Dickson charge pump (using HSMS-diodes and capacitors with 4.7 µF capacitance) and a storage capacitor

with 10 µF capacitance and a MOSFET with an average threshold voltage is 0.57 V.

A solution for faster reaction time has been discussed and one solution could be to lower the capacitor values in the Dickson charge pump and storage to reduce the load. However, decreasing the capacitor values will

(28)

shorten the work duration. Another solution to the problem is to position the harvester further in front of the sensor to activate the switch before the train reaches the sensor. The downside of this solution is that the measurements can only be taken when the train is passing in one direction. This solution can be further developed by having two harvesters, one for each direction.

In order to save energy using the design shown in 22 the working duration will have to be adjusted to how long the collection and transmission of data takes. The discharge time of the storage capacitor is dependent on what voltage it reaches. The time it takes to discharge can be longer than 5 minutes which probably is too long depending on how busy the railway is. Therefore, a way to adjust the working duration is needed.

A solution to open the switch at desired time could be to let the sensor’s integrated circuit send a signal, which short circuit the storage capacitor, when the transmission of data is finished. A design for this solution is shown in figure 23. In the figure a signal, Vshort, is sent to a transistor which shorten the storage. The signal Vincould be a Dickson charge pump output.

Figure 23: A schematic showing the implementation of transistor to shorten the storage using a signal, Vshort, from the sensor’s integrated circuit.

All designs, except the ones using the EVM, does not have a voltage regulator. The voltage regulator makes the circuit resistant against overvoltage. A regulator is needed if the output from the harvester reaches higher voltages. A typical regulator consists of a series of diodes which will shorten the circuit if the voltage in a node is too high. Though, implementing a regulator might increase the load for the circuit and decrease the working duration.

6 Conclusion

The final design for an EcoSense on-demand switch for vibrational sensing in bridges are dependent on various factors. Some examples are: the type of harvester, diodes, capacitors and number of stages and can be modified for the purpose. The harvester will give a signal dependent on both the acceleration of the vibrations and the input impedance of circuit connected to it. To rectify and boost the signal a Dickson charge pump is a reliable option. The charge pump output is increased with the number of stages and the capacitor values. If the input signal to the charge pump increases in frequency, the capacitor values can be decreased. A storage capacitor is needed to maintain the switch signal for the required amount of time. A

(29)

transistor switch can be used with a threshold voltage which is dependent on the transistor and the current through the Drain-Source of the transistor.

For the Lidingö Bridge a good version of the design is a one stage Dickson charge pump with 4.7 µF capacitor and HSMS-diodes and a storage capacitor of 10 µF. In figure 22 the solution is shown with vibrational input from a table forced into vibration by hitting.

EcoSense is applicable for on-demand vibrational sensing in bridges. The design can be modified to the specifications which is needed for different bridges. EcoSense is therefore an energy efficient way of power management during the time in between passing trains.

(30)

7 References

[1] EcoSense: A Hardware Approach to On-Demand Sensing in the Internet of Things, Ye Liu, Qi Chen, Guangchi Liu, Hao Liu, Qing Yang, IEEE Communications Magazine, December 2016.

[2] Neil Storey, ELECTRONICS, A SYSTEMS APPROACH, Pearson Education limited, Edinburgh Gate, United Kingdom, 5th ed, pp. 317, 2013.

[3] Neil Storey, ELECTRONICS, A SYSTEMS APPROACH, Pearson Education limited, Edinburgh Gate, United Kingdom, 5th ed, pp. 327, 2013.

[4] Neil Storey, ELECTRONICS, A SYSTEMS APPROACH, Pearson Education limited, Edinburgh Gate, United Kingdom, 5th ed, pp. 634–635, 2013.

[5] Learn about Electronics, Buck Converter.

http://www.learnabout-electronics.org/PSU/psu31.php accessed May 19th 2017.

[6] Texas Instruments, BQ25570 Nano Power Boost Charger and Buck Converter for Energy Harvester Powered Applications.

http://www.ti.com/lit/ds/symlink/bq25570.pdf accessed May 19th 2017.

[7] Texas Instruments, User’s Guide for BQ25570 Battery Charger Evaluation Module for Energy Harvesting.

http://www.ti.com/lit/ug/sluuaa7a/sluuaa7a.pdf accessed May 19th 2017.

[8] Skyworks, Surface Mount Mixer and Detector Schottky Diodes.

http://www.skyworksinc.com/uploads/documents/Surface_Mount_Schottky_Diodes_200041AC.pdf accessed May 19th 2017.

[9] Avago Technologies, HSMS-286x Series, Surface Mount Microwave Schottky Detector Diodes.

http://www.farnell.com/datasheets/631088.pdf accessed May 19th 2017.

[10] Philips Semiconductors, N-channel enhancement mode BSH105 MOS transistor.

http://www.farnell.com/datasheets/454160.pdf accessed May 19th 2017.

[11] ReVibe Energy, The modelD harvester: enabling wireless power supply.

http://revibeenergy.com/modeld/ accessed May 19th 2017.

(31)

A Dickson charge pump test with 10 Hz input signal

Using 10 Hz input signal.

Table 2: Results from the voltage multiplier simulations.

Type Charge Time

0.2V [s]

Charge Time

0.4V [s] Voltage at 10s [V ]

One stage Dickson 0.21 - 0.28

Cockcroft-Walton 2.78 - 0.26

Two stage Dickson 0.28 0.83 0.55

Three stage Dickson 0.32 0.88 0.82

Four stage Dickson 0.30 0.90 1.06

Five stage Dickson 0.30 0.90 1.25

(32)

B Harvester Positions

Pictures which shows the different harvester positions.

Figure 24: A picture showing harvester position one.

Figure 25: A picture showing harvester position two.

(33)

Figure 26: A picture showing harvester position three.

On-Demand Bridge Monitoring using EcoSense

Examensarbete 15 hp Juni 2017

On-Demand Bridge Monitoring using EcoSense

Scavenging transient vibrations induced by passing train to activate sensor node Lucas Wassénius

Daniel Stenbacka

Abstract

On-Demand Bridge Monitoring using EcoSense

Popular Science Summary - Energy Efficient Bridge Monitoring

Contents

1 Introduction

2 Theory

3 Component List

4 Method

5 Result and Discussion

6 Conclusion

7 References

A Dickson charge pump test with 10 Hz input signal

B Harvester Positions