An experimental study of ultra-low power wireless sensor-based autonomous energy harvesting system

The recent progress in wireless technology makes portable devices and wireless sensors popular, which in turn increases the demand of self-powering technology. Wireless sensors are now commonly used in structural health monitoring (SHM) systems.¹ They usually require their own power supply, which uses the conventional electro-chemical battery technology in most cases. It is almost impossible for sensor networks with battery-powered nodes to meet the ultimate lifetime design goals. The battery must be replaced once it is depleted. The task of replacing the battery for wireless sensors is very costly, which can be alleviated to a large extent through the application of a self-powering device.^2,3 In recent years, a series of self-managing, low power wireless network systems that are built from wireless nodes have been developed, which can easily communicate through the wireless network once the powering issue is solved.^4–6 

Energy-harvesting, which can convert various kinds of ambient energy into electrical energy, has emerged as a self-powering option.⁷ Solar-based energy harvesting, radio frequency-based energy harvesting, thermal-based energy harvesting, flow-based energy harvesting, human-based energy harvesting, and motion energy have been studied for sensor network power solutions.^8–10 However, every kind of energy harvesting has its own limitation. For example, silicon cells for solar energy require enough sunshine, and the wind energy demands for a wind turbine system which carries a high cost.

In the recent decade, increasing efforts have been focused on vibration energy harvesting based on piezoelectric effects.^11–15 Mechanical vibrations exist in many systems, from rotating mechanical machines, bridges where numerous cars run through, or even shoes when people walk. In these situations, piezoceramic materials are well suited for the mechanical-to-electrical energy conversion due to their solid state nature, facilitated by their low cost and various shapes for different applications.¹⁶ For the monitoring systems embedded underground, vibration (including stress waves) energy may be the only feasible source for energy harvesting.

Many factors can affect the results of harvested power energy. For the mechanical system, the power extraction usually depends on the input vibrational characteristics such as acceleration, the mass of the generator, the electrical load, the natural frequency, the mechanical damping ratio, and the electromechanical coupling coefficient of the system.^17,18 For the electrical system, a vibrating piezoelectric element generates an AC voltage, while the wireless sensor requires a stabilized DC voltage, and so, the energy harvesting circuit usually consists of an AC-DC rectifier, a voltage regulator, a DC-DC step-down converter, and an energy storage device (capacitor/supercapacitor).^19–22 In the actual application, there are some additional challenges. As the transfer efficiency of a piezoceramic energy harvester is very low, the amount of charge generated from mechanical vibration is typically quite small.²³ Therefore, charging a capacitor with piezoceramic output takes a long time. The piezoceramic output may come with poor characteristics of high voltage, low current, and high impedance,^24,25 which makes it hard for regulating the circuit and storage device to store the electrical energy without great loss. Meanwhile, the system must ensure impedance matching with the wireless load and electrical compatibility.²⁶ In this way, the optimization scheme for power extraction depends on both the mechanical structure and the electrical system.

The mechanical energy exists in the ambient environment and usually unchangeable. To improve the power extraction capability from piezoelectric harvesters, many research focused on the circuit technology. In the literature,²⁷ a bias-flip rectifier circuit and a voltage doubler by greater than 4X are implemented in a 0.35 μm CMOS process. This efficient control circuit is proposed to regulate the output voltage of the rectifier in order to charge a storage capacitor. In the literature,^28–32 during a certain interval in each cycle of piezoelectric energy harvesting, energy returns from the electrical part to the mechanical part is inevitable. With the use of a technique named synchronized switch harvesting on inductor (SSHI), this energy return problem is overcome and the harvesting efficiency can be greatly enhanced. However, the approach being used to evaluate the performance of the SSHI circuit is based on the assumption that the periodic excitation and the speed of mass are in-phase. Based on the synchronized switch harvesting technique, a new enhanced synchronized switch harvesting technique with double synchronized switch harvesting (DSSH) for optimized energy harvesting allows extracting energy which is independent of the load. An increase in harvested power by almost 300% is achieved in the same vibration condition.^33–35 

After the energy is harvested, the capacitor is needed, by being charged, to store the energy and later work as a power source for wireless devices. The capacitor is suitable for the piezoceramic energy harvesting system because it is small and can be charged with small current. Although a supercapacitor has a large storage capacity, it can only power a wireless sensor with the required high current in a short time. Once the energy is fully discharged to a wireless sensor, the supercapacitor needs to recharge repeatedly. While a wireless sensor is working, it consumes a large instantaneous current to transmit or receive data packets. The sensor node will run out of power before completing a task because of the leakage current that drains its storage capacitor at a rate much faster than the energy harvester can compensate for. Compared to the small charging current of the capacitor in the energy harvesting, the discharging current of the capacitor while a wireless sensor is active is much larger. To solve this problem, many ways were discussed to split the charging and discharging process into two separate steps completely. In the literature,³⁶ a wake-up radio receiver was applied to reduce the wireless sensor node activity and the communication energy so that the power consumption was brought to the nano-Watt level. The wake-up receiver which is powered by a designed MEMS piezoelectric harvester is always on and can receive wake-up messages which have to be sent from a transmitter when it intends to wake up the node. In the literature,³⁷ a Lead Zirconate Titanate (PZT) film prepared by single-step deposition and postannealing was successfully utilized to fabricate a high-performance flexible piezoelectric harvester via a fast, simple, and cost-effective method. The generated electricity was used to charge a capacitor up to 4.3 V within about 45 min and operate a wireless sensor node 18 times. However, the consumption of leakage current is not considered, which drains its storage capacitor in practical application.

Four vibration-powered generators designed to power standalone systems are compared in the literature,³⁸ and their respective effectiveness is discussed. The literature³⁹ presented an overview of current available commercial low power electronics for energy harvesting wireless sensors. It addressed several circuit requirements, including the low-voltage power conversion and the reset circuits which are needed to start-up from a depleted state. Also, it discussed that Micro Control Unit controlled the power conversion stage and applied Maximum Power Point Tracking algorithms. In the literature,⁴⁰ it surveys hybrid energy harvesting systems with multiple harvesters on the same platform. Considering various energy harvesting sources that can be used by Wireless Sensor Networks (WSNs), the literature⁴¹ thoroughly reviewed the energy harvesting mechanism, the harvester hardware design, the amount of energy harvested, and the efficiency of the harvester recently. However, each device must have a suitable interface circuit and energy-awareness to put an additional demand on the embedded device.

Our work demonstrates the low frequency vibration energy harvesting based WSNs, which aim for structural health monitoring (SHM) application. Our work provides a novel and feasible solution for the interface circuit and energy-awareness design particularly based on vibration energy harvesting. The experiment was conducted at single frequency (resonance frequency 2.42 Hz) to get the largest vibration amplitude and therefore the largest PZT output. Usually, the vibration of machines, bridges, or vehicles is of multifrequency, which means that the excitation of PZT is a complex signal. However, the energy can be harvested based on the same principle. It is worth noting that continuous steel highway bridge vibration modes generated by the vibration of moving heavy trucks are at the frequencies of about 3 Hz,⁴² which is very close to our applied frequency of 2.42 Hz. In practical applications, we can enlarge the output amplitude of PZT by applying multiple or large-sized PZT patches or mount PZT patch somewhere of high vibration intensity. Based on the potential application of our developed system in structural health monitoring, we can imagine that future smart bridges could perform self-diagnosis, powered by the energy of vibration or stress induced by vehicles and pedestrians.

In this paper, for a wireless communication system with a sensor-manager pair or wireless sensor nodes communicating to each other in the network, a PZT based energy harvester was designed and tested and the harvested electrical energy was stored in a supercapacitor and used to power a wireless sensor. Due to the very small charging current, the accumulation of electric charge to a desired level was a long process. To avoid the wireless sensor energy consumption in the charging period, an isolation interface circuit between the energy harvester and the wireless sensor was applied. Experimentally, an autonomous energy harvesting based wireless sensor system was successfully verified.

The general purpose of this paper is to design a standalone vibration energy based energy harvesting wireless sensor network and to verify its feasibility experimentally. Specifically, we utilize voltage indicator signals on the energy harvester as the wake up signal to “wake up” the wireless node automatically. We design a new interface circuit that connects the energy harvester and the wireless sensor, and it is used as a physical switch to minimize the energy consumption of the wireless sensor during the charging process and to provide the power to the wireless sensor only when the capacitor is sufficiently charged.

The specific objectives of this paper are (1) to design an energy-efficient, piezoceramic based energy transfer circuit, (2) to design a new switching interface to accomplish the energy harvesting function which can effectively power wireless sensor nodes for data transmission in a wireless network, and (3) then to construct and verify a self-powered autonomous wireless system.

The entire experimental system using a piezoelectric patch to power a wireless sensor is illustrated in Fig. 1. There are four major components in the system, namely, the mechanical structure that vibrates, the energy harvesting module that includes the supercapacitor, and the wireless sensor node that works with the Manager in a wireless sensor network and the interface circuit.

Vibrations exist in many civil and mechanical structures. Vibration energy can be converted from mechanical energy into electrical energy by the PZT patch for energy harvesting.^43–45 In this research, a mechanical structure was built as a vibration platform to provide the mechanical energy. The vibration platform, which was mainly composed of a cantilever beam, a shaker motor, a function generator, a power amplifier, and a weight, was used to produce mechanical energy. The PZT patch was mounted on the fixed end, and the weight was mounted on the free end of the beam. A sinusoidal signal from the function generator was amplified by the power amplifier and then used to excite the shaker motor. The shaker motor then drove the cantilever beam to vibrate at the same frequency with the signal from the function generator.

The main function of the energy harvesting module is to rectify the AC signal generated from the PZT to a DC signal and store the energy in a supercapacitor.^46,47 A circuit based on the LTC3588 chip was designed and developed. Although LTC3588 has been widely used in energy harvesting studies, our work emphasizes designing a practical and completely standalone energy harvesting based wireless system, which is different from many researchers' work. Specifically, we researched different scenarios to fully understand the system properties and overcame several issues by changing the typical design. Figure 2 shows the block diagram of the energy harvesting module based on LTC3588.

The circuit mainly includes the LTC3588 chip, an input capacitor, and an output capacitor which is usually a supercapacitor.⁴⁸ A full-wave bridge rectifier in LTC3588 is connected to the PZT AC inputs (PZT1 and PZT2) in a differential way. The rectified output signal is stored in the input capacitor as a charge reservoir. An ultralow quiescent current in the voltage lockout mode with a wide hysteresis window allows charge to be accumulated on the input capacitor until the switch-mode DC-DC middle converter can efficiently transfer the stored charge to the output capacitor. During the regulation, the converter turns on when the input capacitor is fully charged. Then, the charge transfers from the input capacitor to the output capacitor. After that, the input capacitor becomes fully discharged and the middle converter turns off as needed to maintain the regulation. The voltage indicator signal “PGOOD” changes to the logic high when the voltage “Vout” of the output capacitor is above a threshold, which is 92% of the desired regulation voltage. The “PGOOD” signal remains at the high level so long as the output voltage “Vout” is above the threshold voltage with an output current of 100 mA. This adaptive control technique for the DC-DC converter is used to continuously implement the optimal power transfer and maximize the power stored in the output capacitor.⁴⁹ Then, the energy stored in the output capacitor can be used to power a wireless sensor.

Normally, a wireless network consists of several wireless sensor nodes and a network manager. The DC9003 which integrates the LTC5800-IPM Smart-Mesh IP Mote-on-Chip operates in the 2.4 GHz to 2.4835 GHz international license-free frequency band.⁵⁰ With a powerful 32-bit ARM Cortex-M3, the radio, flash, RAM, and peripherals, the LTC5800 is one of the world's most energy efficient IEEE 802.15 4 compliant platforms with ultra-low power as a wireless sensor node. It requires 4.5 mA to receive a packet, 5.4 mA to transmit at 0 dBm, and 9.7 mA to transmit at 8 dBm. The chip build-in temperature sensor is used to acquire the local temperature data for wireless transmission. The Network Manager DC2274 provides the core networking functionality, enabling the network to achieve unsurpassed levels of resilience, reliability, and scalability. As long as the wireless sensor is fully powered, it would turn to the auto-join state and transmit data packets to the Network Manager through a wireless network as programmed. Otherwise, it remains in the idle state with the ultra-low current of 1.2 μA.

The interface circuit connects the energy harvesting module and the wireless sensor and is used to efficiently utilize the harvested energy. Specifically, an analog switch circuit using ISL43L210, which is controlled by the voltage indicator signal “PGOOD” from the energy harvesting module, is used to optimize the power charging efficiency,^51,52 as shown in Fig. 3. The “PGOOD” is used to control the on/off logic switch in ISL43L210, it ensures that the wireless node and the energy harvesting module output are dis-connected until the harvested voltage “Vout” reaches a desired regulation value, and then, the “PGOOD” signal becomes logic high, which indicates that the output voltage is high enough to power the wireless node. Within a period of consumption, the energy in the energy harvesting module decreases gradually until the voltage “Vout” is lower than the threshold, and “PGOOD” becomes logic low, which indicates that the output voltage cannot meet the power requirement of the wireless sensor. Only when the “PGOOD” turned logically high, did the switch circuit turn to the “on” state, which allowed the harvested energy to power the wireless sensor. Otherwise, the wireless sensor was isolated from the energy harvesting circuit. In other words, the output capacitor kept charging in a free-load mode until the “PGOOD” turned high. With the switching circuit, the charging process of the energy harvester and the powering process of the wireless sensor are seamlessly controlled so that both processes are functional in an alternating fashion.

The energy harvesting experiments are designed as follows:

1. Set up a vibration system and test the output of PZT when the beam vibrates at the resonant frequency to ensure that it has enough power (voltage) to drive a load;

2. Develop an effective circuit according to the characteristics of the PZT output, for energy transfer and storage to accomplish piezoelectric energy harvesting applications;

3. Drive a resistor or a wireless sensor directly with this harvesting circuit to observe its load characteristic;

4. Develop an interface circuit between the energy harvesting module and the wireless sensor, test the isolation function of it before the capacitor is fully charged, and record the charging time;

5. Switch on the interface automatically after the capacitor is well charged, drive the wireless sensor with the harvested energy to transmit data packets, and time the powering period.

Figure 4(a) shows a photo of the vibration platform. A cantilever beam system was set up with a PZT patch at its fixed end, and the fixed end was connected to a shaker motor which was used to excite the beam to vibrate at its first modal frequency of 2.42 Hz. Table I shows the parameters of the vibration platform. Figure 4(b) shows the periodical output signal generated from the PZT patch when the driving signal from the function generator was Vpp = 3 V. When the PZT patch was not connected with the energy harvesting module, the output voltage amplitude approached 40 V.

The energy harvesting module helps to accomplish the energy transfer and storage in high efficiency. The harvested energy is stored in an output capacitor to power a wireless sensor. In the following part, we will discuss the output performance of the energy harvesting module, while it is connected to the wireless sensor in different ways, and observe how the voltage indicator signal “PGOOD” works.

Figure 5(a) shows the connection between the PZT patch and the energy harvesting module. Without connecting any wireless sensor, there is no load to consume the energy from the energy harvesting module. In this case, the output voltage is usually called open circuit voltage.

Figures 5(b)–5(d) show the experimental results when the driving signal from the function generator was Vpp = 3 V, the amplitude of the PZT input signal was 20 V, and the output capacitance was $C$ = 100 μF. After 20-s charging, the output voltage “Vout” reached the target voltage of 2.5 V. Meantime, the voltage indicator “PGOOD” kept logic high. To protect the energy harvesting circuit, the input voltage was limited to 20 V.

When the energy harvesting module is used as a power source, it should provide enough energy at a certain high voltage level with a lasting current for long enough time. To obtain the load effect of the power, we tested its output voltage by simply connecting a resistor with different resistance values but approximate to the impedance of a wireless sensor, which is shown in Fig. 6(a).

Keeping the driving signal from the function generator at Vpp = 3 V, we connected a resistor with different resistance values, $R1$ = 15 kΩ, $R2$ = 72 kΩ, and $R3$ = 180 kΩ, as the load of the energy harvesting module, respectively. In this case, the output capacitor of the energy harvesting module has been discharging to the resistor continuously. The capacitor $C$ and the load $R$ form a $RC$ loop, and the time constant $τ=RC$ changes with the load resistance value. Based on the output capacitor $C$=100 μF, the time constant is $τ1=R1C=$ 1.5 s, $τ2=R2C=$ 7.2 s, and $τ3=R3C=$ 18 s. The larger the $R$ value is, the larger the τ value is and the more slowly the capacitor discharges. Therefore with a different load $R$⁠, the discharging curve is different. For a load with a smaller resistance, the output voltage was very small and vice versa.

Figures 6(b)–6(d) show that the output voltage, “Vout,” which is not a constant, and the maximum voltage amplitude change with the load resistance value. For the resistance $R1$=15 kΩ, the maximum value of “Vout” is 1.2 V. For the resistance $R2$ = 72 kΩ, the maximum value of “Vout” is 1.4 V and the maximum voltage of “Vout” is 1.8 V for resistance $R3$ = 180 kΩ. When the load resistance is changed to a larger value, the load voltage changes to a higher value.

The vibration of a cantilever beam was controlled by a function generator, and the amplitude of the PZT output changed by that of the driving signal. The frequency of the driving signal was kept at 2.42 Hz. The output capacitor in the energy harvesting module was $C$ = 100 μF, and a resistor of $R$=500 Ω was randomly chosen as a low load. When the driving signal was kept at Vpp = 4 V, Vpp = 5 V, and Vpp = 6 V, the charge/discharge rate of the output capacitor increased with the amplitude of the driving signal. However, the maximum value of “Vout” remained the same as the load resistance was constant, which is shown in Fig. 7.

In this section, a wireless communication system with a wireless sensor-manager pair was studied as a load. Figure 8(a) shows the connection between the energy harvesting module and the wireless sensor. “Vsupply” (power supply pin) was connected to the energy harvesting power output pin “Vout” directly. The open circuit output voltage of the energy harvesting module is “Vout” = 2.5 V, and the operation voltage “Vsupply” for the wireless sensor LTC5800 contained in DC9003 is between 2.2 V and 3.76 V. The manager DC2274, which managed the wireless network and received the data from the wireless sensor, was connected to a hosting computer.

The driving signal from the function generator was Vpp = 4 V, and the output capacitor in the energy harvesting module was $C$ = 100 μF. “Vout”/“Vsupply” changed periodically due to charging and discharging of the output capacitor and kept at a low voltage, as shown in Fig. 8(b). By the discharging curve, the equivalent impedance of the wireless sensor was computed to be $R$ = 50 kΩ and the maximum discharging current to be 30 μA. Due to a 30 μA leakage current flowing through the wireless node, the output of the capacitor could never reach the node's operating voltage. So, it was unable to provide enough energy for node's operation.

From the above experimental results, we found that the equivalent impedance of a load would affect the output performance of the energy harvesting module. The equivalent impedance of the wireless sensor was 50 kΩ. To match the high output impedance requirement of the energy harvesting module, we connected a large resistor (⁠$R$ = 160 kΩ) in series with the wireless sensor as shown in Fig. 9. Due to the increase in the load, the output of the energy harvesting module got obvious, as a 2.5 V “Vout” signal lasted for about 5 s periodically. “Vout” changed with the charging and discharging of the output capacitor. Correspondingly, “PGOOD” changed with the “Vout” from logic high to logic low periodically. The wireless sensor was powered by a source of about 0.6 V, which is too small to turn it on.

Comparing the above results, we obtain the following observations:

1. As a load of the energy harvesting module, the wireless sensor or resistor kept consuming energy before the output capacitor is fully charged.

2. When the “Vout” pin was connected to the “Vsupply” pin directly, the output voltage and current of the energy harvesting module were too small to power the wireless sensor.

3. To obtain a desired voltage, the load impedance must match with the impedance of the energy harvesting module.

We concluded that it is necessary to separate the energy harvesting module and the wireless sensor during the capacitor charging period to avoid the continuous energy consumption.

To avoid the continuous consumption of the harvested energy, an on/off switch circuit ISL43L210 contained in DC9003 was applied. Figure 10 shows the experimental setup. The energy harvesting module output was connected with the wireless sensor through the switch circuit, and the performance of the entire energy harvesting system was observed by measuring the voltage and current.

Through the applied switching interface between the energy harvesting module and the wireless sensor, the harvested energy stored in a capacitor, once exceeding a certain level, will power a wireless sensor. In this section, the performance of the energy harvesting module with the output capacitor using different capacitance values (100 μF, 0.33 F,…, and 1.5 F) will be tested. Through comparison, the optimal capacitance value can be selected.

The output capacitor of the energy harvesting module was a tantalum capacitor $C$ = 100 μF. Figure 11 shows the results when the drive signal from the function generator was Vpp = 4 V. While “PGOOD” was turned to logic high, “Vsupply” demonstrated a 2.5 V periodical pulse signal that has the same amplitude as “Vout.” This indicated that the “PGOOD” signal could effectively realize the circuit isolation between the energy harvesting module and the wireless sensor to guarantee that no power was consumed by the wireless sensor while the output capacitor was charging. After charging the output capacitor to 92% of the target value, “Vsupply” was connected to “Vout,” causing the sudden increase in signal “Vsupply.” Due to the wireless node's significant energy consumption, “Vout” suddenly decreased to a value that was less than the threshold, then, “PGOOD” was turned to logic low, and the switch was turned off. Meanwhile, the output capacitor kept charging and its voltage kept increasing. As we can see, the “Vsupply” signal formed a sequence of pulses.

When the capacitor of 100uF was used as the output capacitor, the “Vsupply” signal was transient pulses. It was not enough for the wireless sensor to transmit packets because of the short duration time of the pulses. To obtain a constant “Vsupply” voltage, a supercapacitor with a larger capacitance (C = 0.33 F) was used. Keeping the drive signal from the function generator at Vpp = 4 V, the experimental results during the charging process for the 0.33 F supercapacitor are shown in Figs. 12–14.

At the beginning, when the voltage “Vout” exceeded the threshold voltage (2.3 V) that was 92% of the target voltage (2.5 V), the “PGOOD” signal turned logic high and “Vsupply” had the same voltage amplitude as “Vout.” The wireless sensor was able to consume the harvested energy although the “Vsupply” was still a series of short pulses. The consumption resulted in the decrease in “Vout” until “Vout” was lower than 2.3 V and “PGOOD” became logic low, which caused the disconnection between the energy harvesting module and the wireless sensor. With continuous charging, “Vout” kept increasing. “Vsupply” gradually became square waves and finally reached a steady power of 2.5 V. This indicated that by using a 0.33 F supercapacitor, the harvested energy was enough to provide a continuous operational voltage for the wireless sensor under the control of the “PGOOD” signal.

The driving signal from the function generator was kept at Vpp = 4 V. A supercapacitor with capacitance values of C = 0.66 F, 0.99 F, 1.32 F, and 1.5 F was connected to the output of the energy harvesting module. The charging/discharging time for these capacitors with different capacitance values was different. “Vsupply” would be kept at the desired value for a longer time if a larger supercapacitor is used.

By using the designed mechanical vibration system, through the applied interface circuit, a 100 μF tantalum capacitor could be fully charged in couple of seconds and provide a burst of current that lasted for about 100 ms. To test the powering capability of the energy harvesting module with different output capacitors, we programmed the wireless sensor to transmit data packets continuously. For the same driving signal from the function generator Vpp = 4 V, the charging time and powering time are compared in Table II. The larger the capacitance value is, the longer the charging time and powering time will be.

The main task of the wireless sensor is to collect and transmit data to the network manager. Once the wireless sensor starts to work, it consumes energy. A wireless sensor requires different working current levels to work in different states. By monitoring the current of the wireless sensor in different working states, we can evaluate the performance of the entire energy harvesting system.

A wireless node should join a network before it communicates with the network manager. To join a network, the node listens to advertisement messages sent by the network manager. Once it hears such an advertisement, it synchronizes to it and starts the network joining process with the manager. As soon as the joining process is finished, the node becomes operational and sends a data packet to the manager over the wireless network. It is assigned the “base bandwidth” from the manager, allowing it to send infrequent traffic to the manager. When the wireless node was powered by the energy harvesting module which was connected to a 0.33 F output capacitor, the current in different working states was measured and is shown in Fig. 15.

As demonstrated in Fig. 15(a), the joining network process consumed a significant amount of energy. The current was around 5.5 mA when the node was listening at 100% duty cycle. It spent about 10 s listening before it could synchronize and dropped to a lower power cost mode, typically less than 50 μA.

From the above results, we know that “Vsupply” of a wireless node is available only when the “Vout” is at least above 2.3 V. When the energy harvesting module power supply “Vout” changes from 2.3 V to 2.5 V, the total energy stored in the output capacitor is obtained

Then, the charge in the capacitor which can be used to power the wireless sensor is

where $V2$ = 2.5 V and $V1$ = 2.3 V are applied. The current for the wireless sensor in the listening state is around $I$ = 0.0055 A, and hence, the total time $ΔT$ that the wireless sensor can listen is

for the supercapacitor of $C$ = 0.33 F and $ΔT$ = 12 s.

When the wireless sensor transmits data packets to the manager, it consumes a short pulse of current. Figure 15(b) shows the current pulses (1.6 mA–1.8 mA) while each data packet was transmitted. Theoretically and experimentally, a 0.33 F supercapacitor as an output capacitor for the energy harvesting module was enough to power a wireless sensor joining the network and transmitting data packets. If we choose a larger supercapacitor, it can provide more energy to transmit more data packets.

In the wireless network, the wireless sensors which are powered by the energy harvesting module can communicate with each other. The data packets can be transmitted at different transmitting power levels and different numbers of packets by programming Application Program Interface parameters. We customized the software program in the wireless sensor based on our application so that the wireless sensor listens, joins the network, and transmits data packets as long as the wireless sensor is powered on.

Figure 16 shows the current curves when a wireless sensor, which was powered by the energy harvesting module, sent 50 data packets each time to another node with transmitting powers of 8 dB and 0 dB, respectively. The pulse current was 2.5 mA for the 0 dB transmitting power and 4 mA for the 8 dB transmitting power. Every time the wireless sensor sent a packet, there was a pulse current consumption with a difference of 1.5 mA.

Figure 17 shows the current consumption curves when the wireless sensor which was powered by the energy harvesting module received data packets from a transmission sensor. When a wireless sensor received data packets, the current consumption was around 5 mA until it finished the receiving process.

In this paper, an autonomous piezoceramic-based energy harvesting system was implemented and verified experimentally by powering a wireless sensor node. Specifically, an LTC3588-based energy harvesting circuit was designed to realize the energy transfer and storage. A comprehensive study of LTC3588 properties with different loads is presented. Based on experimental results and analysis, we selected the appropriate value for the output capacitor in our designed LTC3588 based energy harvesting module. To the best of our knowledge, this is the first work that integrated a micropower energy harvesting power supply LTC3588 and an ultra-low power wireless mote LTC5800 in the application and experimentally demonstrated the feasible solution. The integration was enabled by an applied switching interface which can control the power on/off of a wireless sensor so that it guarantees the supercapacitor to be charged under no-load conditions.

We systematically and experimentally studied the system's performance under different scenarios. Experimental results showed that with 12 h charging, the stored energy in the 0.33 F supercapacitor was abundant for the wireless sensors to join the network and transmit data packets to each other. By using new ultra-low power wireless module LTC5800, our experimental study demonstrates that harvested vibration energy can fully power a wireless sensor node. With the increasing urgency and importance of structural health monitoring,^53,54 the developed method has the potential to power wireless sensors in structural health monitoring systems, and this prototype provides the design solution for other researchers.

The reported work was partially supported by the U.S. National Science Foundation (NSF) (Grant No. ECCS-1101547) and Shandong Natural Science Foundation (Grant No. ZR2013EEL004).