Power Processing Circuits For Piezoelect PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

4170 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO.

12, DECEMBER 2010

Power Processing Circuits for Piezoelectric


Vibration-Based Energy Harvesters
Reinhilde D’hulst, Tom Sterken, Member, IEEE, Robert Puers, Senior Member, IEEE,
Geert Deconinck, Senior Member, IEEE, and Johan Driesen, Member, IEEE

Abstract—The behavior of a piezoelectric vibration-driven en- The output voltage of an energy harvester, generally, is not
ergy harvester with different power processing circuits is evalu- directly compatible with what is needed to power the load elec-
ated. Two load types are considered: a resistive load and an ac–dc tronics; moreover, the power transfer from an energy harvester,
rectifier load. An optimal resistive and optimal dc-voltage load for
the harvester is analytically calculated. The difference between the generally, can be maximized by optimizing the load impedance
optimal output power flow from the harvester to both load circuits connected to the harvester. Thus, a power processing circuit
depends on the coupling coefficient of the harvester. Two power needs to be connected between harvester and load. To date in
processing circuits are designed and built, the first emulating a literature, the work on power processing circuits for vibration-
resistive input impedance and the second with a constant input based energy harvesters can be roughly classified into two
voltage. It is shown that, in order to design an optimal harvesting
system, the combination of both the ability of the circuit to harvest different approaches. In a first approach, the efficiency of the
the optimal harvester power and the processing circuit efficiency power processing circuit itself is the major point of attention,
needs to be considered and optimized. Simulations and exper- including the design of efficient control circuitry for the inter-
imental validation using a custom-made piezoelectric harvester face circuit (see, among others, [5]–[7]). In a second approach,
show that the efficiency of the overall system is 64% with a buck maximizing the output power transfer from the energy harvester
converter as a power processing circuit, whereas an efficiency of
only 40% is reached using a resistor-emulating approach. is the main focus (see, among others, [8]–[10]). In this paper,
it will be shown that, in order to design an optimal harvesting
Index Terms—Energy efficiency, energy harvester, piezoelectric system, the combination of both the efficiency and the ability of
devices, power conditioning.
the processing circuit to harvest the maximum available output
power needs to be optimal. This optimal combination depends
I. I NTRODUCTION on the harvester design.
Two different load types are considered in this paper: a
T HE current advances in performance and functionality of
micro- and nanosystems have stimulated the development
of intelligent networks of autonomous systems. The demand
resistive load and an ac–dc rectifier load. In literature, much
work has been done concerning the optimal power flow of
for a small, mobile, and reliable energy supply for each au- the harvester in case the load is a linear resistor (see, among
tonomous network node has led to the development of a new others, [11]). Since the vibration-based harvester provides a
type of generators, as the use of conventional electrochemical varying ac power and because electronic loads typically need
batteries is not always an option because of the need for a stable dc power supply, it is useful to analyze the harvester
replacement and the volume dependence on the amount of behavior when connected to an ac–dc rectifier. Other load types,
stored energy. Motion energy or vibrations are an attractive as described in [6], [12], and [13], are not considered in this
source for powering miniature energy-harvesting generators paper.
[1]. Vibration energy can be converted into electrical energy The remainder of this paper is organized as follows.
through piezoelectric [2], electromagnetic [3], and electrostatic Section II discusses the modeling of piezoelectric energy har-
[4] devices. This paper focuses on piezoelectric devices. The vesters. Section III describes the behavior of the model with
output power of such devices, made using micromachining the two different load types. In Section IV, the design and,
techniques, is limited, ranging from milliwatts down to only a in particular, the efficiency of the power processing circuits
few microwatts. are discussed. In Section V, all findings are illustrated by
simulations and measurements on a custom-made piezoelectric
harvester.

Manuscript received May 8, 2009; revised July 29, 2009; accepted


November 8, 2009. Date of publication March 1, 2010; date of current version II. H ARVESTER M ODEL
November 10, 2010.
R. D’hulst was with the Department of Electrical Engineering, Katholieke Energy harvesters of the inertial type are considered, i.e., the
Universiteit Leuven, 3000 Leuven, Belgium. She is now with VITO. motion of the vibration source is coupled to the generator by
T. Sterken was with the Interuniversitary Center for Microelectronics, 3001
Leuven, Belgium. He is now with the University of Ghent, 9000 Ghent, means of the inertia of a seismic mass. This mass m is modeled
Belgium. as being suspended by a spring with spring constant k, while its
R. Puers, G. Deconinck, and J. Driesen are with the Department of Electrical motion is damped by a parasitic damping d due to friction and
Engineering, Katholieke Universiteit Leuven, 3000 Leuven, Belgium (e-mail:
johan.driesen@esat.kuleuven.be). air. The mass is also damped by the generator, the piezoelectric
Digital Object Identifier 10.1109/TIE.2010.2044126 transducer, exerting a force Fg . The displacement of the mass

0278-0046/$26.00 © 2010 IEEE


D’HULST et al.: POWER PROCESSING CIRCUITS FOR PIEZOELECTRIC VIBRATION-BASED ENERGY HARVESTERS 4171

Fig. 1. Electrical circuit equivalent of vibration-based harvester model.

Fig. 2. Electronic equivalent of a velocity-damped inertial energy harvester.

is z(t), and the displacement of the package is y(t). I and U Fig. 3. Optimal output power of an energy harvester with resistive load, nor-
are the outgoing current and voltage, respectively. This system malized to Plim , versus operating frequency and squared coupling coefficient
(ζ = 0.016).
is governed by the following differential equation:
B. Resistive Load
mÿ = mz̈ + dż + kz + Fg . (1)
The harvested power dissipated in a resistive load is
Different configurations and implementations of the piezoelec- analytically calculated using the equivalent circuit in Fig. 1. For
tric generator lead to the same general electromechanical model every operating frequency ω, an optimal load resistance Ropt
of a piezo tranducer [14]–[16] exists. The expression for the optimal load resistance written
 in terms of the natural resonance frequency ωn = k/m,
Fg = Kz + ΓU
(2) the normalized parasitic damping factor ζ = d/2mωn , the
I = Γż − C0 U̇ .
normalized frequency Ω = ω/ωn , and usingthe definition of
Parameter Γ is a measure for the electromechanical coupling the electromechanical coupling factor κ = Γ2 /(kC0 + Γ2 )
of the piezo element, and C0 is the clamped capacitance of the gives [16], [17]
piezo element. K is a measure for the stiffness of the piezo 
1 (1 − Ω2 )2 + 4ζ 2 Ω2
element. Fig. 1 shows an electrical circuit equivalent of the Ropt =    . (4)
ωC0 2 2 + 4ζ 2 Ω2
1−κ2 − Ω
1
harvester model, including the piezoelectric element. In this
electrical circuit, the voltages represent forces, and charges
represent displacements. The stiffness of the piezo element K The corresponding output power dissipated in the optimal load
is neglected in Fig. 1 and is assumed to be incorporated into the resistance is given by the following equation, with Y0 being the
mechanical spring constant k. amplitude of the input vibration:
2
Ω κ
mω 3 Y02 8ζ 1−κ2
Pres =  .
III. O PTIMAL P OWER O UTPUT 
 1 2
 −Ω 2
+ 2ζΩ + (1−Ω
2 2
κ2 ) 2
A. Power Output Limit 1−κ2 2ζΩ 2ζΩ + 1−κ2ζΩ

The output power of both load cases considered, resistive (5)


load and ac–dc rectifier load, is compared to the theoretical
absolute maximum output power of an inertial energy harvester, The optimal harvested power depends on the electrome-
i.e., the output power of a harvester with the damping force of chanical coupling factor κ. Fig. 3 shows a plot of the optimal
the generator proportional to the velocity of the seismic mass. harvester output power, normalized to Plim , versus the
An electronic equivalent of this theoretical harvester is shown in normalized operating frequency Ω and the squared coupling
Fig. 2. The output power dissipated in the load dampingRload coefficient κ2 with the damping factor ζ arbitrarily set to 0.016.
reaches a maximum at the resonance frequency ωn = k/m At both open- and short-circuit resonance frequencies, the
if the load resistor has an impedance that is identical to the output power tends to Plim for larger coupling factors.
parasitic damping d. The maximal output power is given by the
following equation, with a being the acceleration imposed on C. AC–DC Rectifier Load
the system:
The output power of an energy harvester connected to an
a2 m2 ac–dc rectifier, with a fixed voltage Ucc at the dc side, is
Plim = . (3)
8d analytically calculated as well. The relation between the mass
4172 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 12, DECEMBER 2010

Fig. 5. Maximal harvester power for both load cases versus squared coupling
factor (ζ = 0.016).

Fig. 4. Harvester output power, normalized to Plim , transferred to the optimal


dc load via an ac–dc rectifier versus operating frequency and squared coupling
factor (ζ = 0.016).

displacement amplitude zM and the dc voltage Ucc is given


by (6), with a being the (constant) acceleration of the mass
and ϕ being the phase difference between the applied force
Fig. 6. Schematic overview of the piezo bimorph harvesting structure.
to the mass (i.e., the displacement of the package) and the
displacement of the mass [16], [18]. The phase difference ϕ TABLE I
is not negligible if the electromechanical coupling of the device M EASURED M ODEL PARAMETERS OF B IMORPH H ARVESTING
D EVICE AT AN I NPUT ACCELERATION OF 5 m/s2
is not too weak.

−ma cos(φ) = −mzM ω 2 + kzM + Ucc Γ
(6)
−mazM sin(φ) π2 = −dzM 2
ω π2 + 2Ucc (ΓzM − Ucc C0 ).

The forward voltage loss over the needed rectifier is neglected


in (6) in order to have a fair comparison with the resistive load
case where no rectifier was taken into account either.
The expression for the average output power, dissipated in
the fixed-voltage sink Ucc is
ω
Prect = 2Ucc (ΓzM − Ucc C0 ). (7)
π
Prect varies with Ucc . Similar to the resistive load case, an
optimal voltage Ucc can be found for every operating frequency
[16], [18]. It is however not possible to calculate this optimum
analytically. Fig. 4 shows the plot of a numerically determined
optimal Prect , normalized to Plim , versus normalized frequency
and squared coupling coefficient. For large coupling factors, the
optimal power at both resonance frequencies also tends to Plim .
There is no difference in maximal output power between both
Fig. 7. Predicted and measured output powers of bimorph device versus
discussed load cases for a harvester device with large coupling (a) resistive load and (b) ac–dc rectifier load.
coefficient. However, for weakly coupled harvester devices, a
resistive load tends to capture more power from the harvester upon a substrate to create a cantilever beam configuration, and
device, as can be seen in Fig. 5. a small mass of 1 g is attached to its tip. The harvester is
excited by a shaker connected to a function generator via a
power amplifier. The open-circuit resonance frequency of the
D. Harvester Example
device lies around 301 Hz. The model parameters of the device
To illustrate the previous analysis, a commercially available are measured according to the procedure given in [19] and are
piezo bimorph element is chosen as the energy-harvesting given in Table I. The parameters are measured while imposing
device. Fig. 6 shows a schematic diagram of the harvesting an acceleration of 5 m/s2 to the harvesting device. Fig. 7
structure with its dimensions. The bimorph element is glued shows the output power of the bimorph device as predicted
D’HULST et al.: POWER PROCESSING CIRCUITS FOR PIEZOELECTRIC VIBRATION-BASED ENERGY HARVESTERS 4173

Fig. 8. Electrical circuit schematic of a buck–boost converter without an input


filter capacitor.

by the aforementioned modeling, together with the measured Fig. 9. Input current of a buck–boost converter without an input filter
capacitor.
output power, for both a varying resistive load and a varying
output dc voltage after being rectified by a full-bridge diode
rectifier. The output power measurements are accurate to 1 μW. input vibration frequency of the harvester. The output voltage
The measured forward voltage drop over the rectifier diodes is Uout of the converter is assumed to be held constant.
0.45 V. In order to have a clear point of comparison for both the
modeled and the measured output power, the output power is 2Lfs
Rin = . (8)
plotted against the amplitude of the voltage on the piezo output δ2
capacitor C0 in Fig. 7(b). The output power is measured at the
dc side of the rectifier and does not incorporate the losses in the For the efficiency analysis, the losses in the switch (switching
rectifier and, hence, the difference between the modeled and losses as well as conduction losses due to its series resistance
measured powers. The squared mechanical coupling factor κ2 Rsw ), the losses in diode D, in the full-bridge rectifier, and the
of the device is 0.03. As was mentioned before (see Fig. 5), losses in the inductor L due to its equivalent series resistance
the ac–dc resistive load performs better at lower coupling ResrL are taken into account. The harvester is modeled as a
factors. current source parallel to the output capacitor C0 to simplify the
determination of the voltages and currents in the converter [21]
(see Fig. 8). The current amplitude of the source Ih is defined
IV. D ESIGN OF P OWER P ROCESSING C IRCUITS by the optimal harvester output Pres power and the optimal
As already stated, a power management circuit is needed to resistive load Ropt
transform the harvester output voltage to a dc voltage of an  
appropriate level, and moreover, the power management circuit 2Pres Ropt 1 + ω 2 C02 Ropt
2

should also provide the optimal load to the harvesting device. Ih = . (9)
Ropt
The choice can be made to provide the optimal resistive load
to the harvester, but according to previous analysis, a fixed dc The current through the inductor is calculated for the two
voltage may allow the harvester to perform nearly optimal as phases of the switching period (switch closed and switch open)
well, depending on the electromechanical coupling factor of the by evaluating the differential equations describing the circuit.
harvester. The question is now which power processing circuit The losses in this circuit are as follows.
will lead to the optimal harvesting system. For that reason,
efficiency analysis is made for both circuits. An analytic loss 1) MOSFET loss: The MOSFET loss Pmos is the sum of
model is set up for both power processing circuits in order to the conduction loss Pon , the switching loss Psw , and
be able to find the most efficient circuit configuration in a quick the gatedrive loss Pdrv [22]. The conduction loss of the
yet accurate way. switch is calculated as the loss in Rsw , the on-resistance
of the switch. As the converter operates in discontinuous
conduction mode, the current is zero at the beginning of
A. Processing Circuit Providing Resistive Load each switching period, so that the switching loss during
A buck–boost dc–dc converter without an input filter capac- on-switching of the switch is negligible. The gate-drive
itor (see Fig. 8), operating in discontinuous conduction mode, loss of the switch is given by [23], [24]
has a “resistive” input impedance and may thus be well suited
to operate as a power processing circuit for harvesting devices Pdrv = Ugs Qg fs (10)
[10], [20]. Fig. 9 shows the input current of the converter.
The input resistance of the circuit Rin is controlled through where Ugs and Qg are the gate–source voltage at the ON
the duty cycle of the switching element, generally a MOSFET, state and the gate charge, respectively.
according to (8), with L being the used inductance, fs being 2) Diode loss: The diode loss Pdiode is given by (11),
the switching frequency, and δ = Ton fs being the duty cycle of with UD being the diode forward voltage and iavg_out
the switching element. Note that (8) only holds if fs  fh , the being the average output current of the converter. Reverse
4174 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 12, DECEMBER 2010

TABLE II
PARAMETERS OF D EVICES U SED IN P OWER P ROCESSING C IRCUITS

Fig. 11. Electrical circuit schematic of a buck converter with an input filter
capacitor.

B. Processing Circuit Providing Fixed Voltage


A fixed dc voltage can be provided through, e.g., a buck
converter with a sufficiently large input filter capacitor (see
Fig. 11). The input voltage of the converter is controlled
through the duty cycle. Because of the very low power
processed by the circuit, the converter is most likely to operate
in discontinuous conduction mode. Analogous to the previous
section, the harvester is modeled as a current source to simplify
the determination of the voltages and currents in the converter
(see Fig. 11). The input current of the converter Iin is defined by
the optimal harvester output power Prect and the optimal fixed
output voltage Uopt [see (14)]. In the analytic calculations, the
Fig. 10. Efficiency of a buck–boost converter with a resistive input impedance forward voltage drop over the rectifier diodes UD was not taken
of 580 kΩ and an input power of 49.4 μW. into account. In the analytic loss model, however, these voltage
drops are not neglected, and thus, the converter has to provide a
recovery of the diode is neglected in this analytic loss voltage at its input equal to Uopt minus twice the diode voltage
model drop UD .
Prect
Pdiode = UD iavg_out . (11) Iin = . (14)
Uopt + 2UD
3) Inductor loss: The loss PL in the inductor is calculated as The current through the inductor is again calculated for the two
the loss in its equivalent series parasitic resistance ResrL . phases in the switching period. By using the expressions for
4) Bridge rectifier loss: The power loss in the bridge rectifier the inductor current, the input voltage of the converter can be
Prect is given by the following equation, with iavg_in calculated. The duty cycle δneeded to obtain the optimal Uopt
being the average input current of the converter: is then found through iteration.
The losses in this circuit are as follows.
Prect = 2iavg_in UD . (12) 1) MOSFET loss: The switch conduction loss Pon can be
obtained from the rms current through the switch and
The efficiency of the converter is then expressed as the switch on-resistance Rsw . As only discontinuous con-
duction is considered, no switching loss occurs during
PRopt
η= . (13) on-switching of the switch. The gate-drive loss is given
PRopt + Pmos + Pdiode + PL + Prect
by (10).
2) Diode loss: The diode loss can be obtained using (11),
The efficiency of a converter used as a power processing with iavg_out being the average output current of the
circuit for the previously mentioned bimorph harvester device converter.
is calculated for varying switching frequency and inductance. 3) Inductor loss: The loss in the inductor due to its series
The optimal resistive load conditions of the bimorph harvesting resistance is calculated using the rms inductor current.
device (see Table I) are taken as input for the efficiency calcu- 4) Bridge rectifier loss: The loss in the bridge rectifier is
lations. The device parameters used are listed in Table II. An
output voltage of 3 V is chosen. Fig. 10 shows a plot of the cal- Prect = 2Iin UD . (15)
culated efficiency versus switching frequency and inductance.
The highest efficiency, i.e., 51%, is reached using a switching 5) Input filter loss: The loss in the input filter capacitor is
frequency of 3 kHz and an inductance of 150 μH. The duty calculated as the loss in the series and parallel parasitic
cycle at this point is 0.125%. Switching frequencies beyond resistors (ResrC and RparC , respectively).
3 kHz are not suitable anymore, as the switching frequency is The efficiency of the converter is then calculated using (13).
not sufficiently large anymore compared to the vibration source The efficiency of a buck converter is calculated for varying
frequency. switching frequency and inductance, using the optimal dc-load
D’HULST et al.: POWER PROCESSING CIRCUITS FOR PIEZOELECTRIC VIBRATION-BASED ENERGY HARVESTERS 4175

TABLE III
C OMPARISON OF PS PICE S IMULATION R ESULTS AND T HEORETICAL
E FFICIENCY C ALCULATIONS

Fig. 12. Efficiency of a buck converter with an input voltage of 4.9 V and an V. S IMULATIONS AND M EASUREMENTS
input power of 39.8 μW.
Both power management circuits have been implemented in
conditions of the bimorph harvesting device as input (see PSpice and have been simulated together with the electronic
Table I). An output voltage of 3 V is chosen. Fig. 12 shows circuit equivalent of the bimorph piezo energy harvester. The
a plot of the calculated efficiency versus switching frequency used MOSFET model is a 2N7002, and the diode model is a
and inductance. The highest efficiency, i.e., 60.3%, is reached BAS16W. The energy harvester is driven at its short-circuit res-
using a switching frequency of 3 kHz and an inductance of onance frequency. The switching frequency and the inductance
150 μH. The duty cycle in the optimal point is 0.18%. value of both power management circuits are taken from the
The conclusion drawn from Figs. 10 and 12 is that the opti- optimal efficiency point in previous calculations.
mal efficiency of a buck converter with a fixed input voltage is A comparison of the simulation results with the theoreti-
considerably higher than the efficiency of a buckboost converter cal calculations is given in Table III. Circuit1 represents the
emulating a resistive input impedance. It must be noted that the buck–boost converter emulating a resistive load, Circuit2 is the
theoretical optimal efficiencies are rather low because of the buck converter providing a fixed input voltage to the harvester.
very low given power budget. First, the simulations show that both circuits are very well able
to capture the available power out of the harvester. Second,
the simulated losses in the circuits are consistent with the
C. Influence of Control Circuitry
analytic calculated ones, except that the gate-drive losses of
The analytically calculated efficiencies mentioned are the the MOSFET are slightly lower in simulations because the gate
open-loop efficiencies. Adding adequate control circuitry to the charge Qg is lower than given in the datasheet of the MOSFET
power processing circuits might lead to an increased efficiency due to the very low drain–source currents. The simulations also
of the design. In the rectifier load case, the input voltage of the show that the modeling of the losses in the diode bridge rectifier
converter should always be regulated toward the optimal value. is not so accurate.
By using hysteretic (thermostatlike) control, this is a control The simulation results also indicate that, although the output
strategy where the converter is switched on only when the power of the harvester when connected to the resistive load
voltage on the filter capacitor rises above a previously set limit, is considerably higher than the output power of the harvester
the overall efficiency of the circuit is dramatically increased, with rectifier load, the overall output power of both processing
since the circuit dissipates only a portion of the time, as shown circuits is comparable.
in [5]. Both dc–dc converters have been built on a printed cir-
Such type of control, switching on and off of the circuit, is cuit board using commercially available discrete devices. The
not suitable for power processing circuits providing resistive used MOSFET is a 2N7002, and the diode is a BAS16W.
load, as the input of these converters is not buffered. This The inductors are ferrite drum core inductors in an SMD
indicates that the difference in efficiency between both types 1210-package. A PIC18F4550 microcontroller is used to gen-
of converters can be increased even more by adding control erate the pulsewidth-modulated (PWM) gate-drive signals, and
circuitry. an HCPL3180 is the gate driver. The microcontroller and the
The analytic calculations show that the overall output power gate-drive circuitry are powered from an external source, as
of the example piezoelectric harvesting system (i.e., the har- the optimization toward low power consumption of the control
vester with its power conditioning circuit) is comparable for components is beyond the scope of this paper. The minimal
both load cases, although the output power of the harvester PWM frequency generated by the microcontroller is 3 kHz,
itself is much higher when the load is resistive, compared to the with a minimal duty cycle of 0.2%. An inductor with an induc-
rectifier load. By adding a control circuitry, the overall output tance value of 470 μH leads to a duty cycle that is larger than
power of the harvesting system using the buck converter as the minimal 0.2% for both circuits, and hence, this inductance
interface circuit will be higher compared to the output power of value has been used in the practical realization. As can be seen
the system with a buck–boost converter providing the optimal in Fig. 10, the theoretical efficiency of the buck–boost con-
resistive load. verter providing the resistive load with an inductance value of
4176 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 12, DECEMBER 2010

TABLE IV designed: the first one emulating a resistive input impedance


M EASUREMENTS ON B OTH P OWER P ROCESSING C IRCUITS
and the second one with a constant input voltage. A buck–boost
dc–dc converter without input filter capacitor, operating in
discontinuous conduction mode, is shown to have a resistive
input impedance. A buck converter with input filter capacitor is
used to evaluate the rectifier load case. An analytic loss model
of both converters is set up to determine the optimal operating
point of the converters. If the model parameters of a certain
energy harvester are known, the output power of the energy
470 μH is 51%. The efficiency of the buck converter with the harvester after power processing can be calculated using the
same inductor is 60%, as shown in Fig. 12. analytic loss models of the converters.
The output voltage of both circuits is set to 3 V, and the Generally, it can be concluded that, in order to design an
piezo bimorph harvester is driven at its open circuit resonance optimal harvesting system, it is not sufficient to employ the
frequency with an acceleration of 5 m/s2 . A summary of the power processing strategy that is able to harvest the maximal
main measurements on both circuits is given in Table IV. amount of energy out of the harvester. The efficiency of the
The output power of the harvester, connected to Circuit1, is power processing circuit is equally important. This has been
measured to be about 48 μW. This value is consistent with the illustrated by means of a piezo bimorph taken as harvester
theoretical optimal output of the harvester as shown in Fig. 7. device. The harvester generates more power if connected to
From this measurement, it can be concluded that the input the optimal resistive load than if connected to the optimal dc
resistance emulated by the buck–boost converter is correctly load. However, the loss calculations show that the efficiency
seen by the energy harvester. The efficiency of the converter of the converter emulating a resistive input impedance is much
is rather low, about 40%, without taking the gate-drive losses lower (51%) than the efficiency of the buck converter (60%)
into account. at their optimal operating points. The efficiency of the latter
Using Circuit2, the voltage at the output of the harvester, processing circuit can be increased even more by adding control
after rectification, is set to 4.8 V, the measured output voltage circuitry. Hence, for this example, it can be concluded that the
is about 39 μW, and this value is also quite consistent with efficiency of the overall system of a harvesting device with
the theoretical value as shown in Fig. 7. The efficiency of the power processing circuit will be better if the processing circuit
converter is about 64%, without taking the power loss in the has a fixed dc voltage as input. This conclusion is validated
gate driver into account. through simulations and experimental measurements. An out-
The measured and calculated power losses are also shown put power of 19 μW was measured after power processing
in Table IV. The differences between the measured and the with the resistor-emulation approach, whereas a 25-μW output
modeled power loss are about 15% for Circuit2 and 25% for power was measured with the fixed-voltage approach.
Circuit1, without taking the gate-drive losses into account.
The higher measurement losses are probably due to parasitic
effects that have not been taken into account in the model, e.g., R EFERENCES
parasitic capacitances that become relatively important because [1] S. Roundy, P. K. Wright, and J. Rabaey, “A study of low level vibrations
of the high frequencies present. as a power source for wireless sensor nodes,” Comput. Commun., vol. 26,
no. 11, pp. 1131–1144, Jul. 2003.
From the measurements, it can be concluded that, for this [2] S. Roundy and P. K. Wright, “A piezoelectric vibration based generator for
particular harvester configuration, the open-loop efficiency of wireless electronics,” Smart Mater. Struct., vol. 13, no. 5, pp. 1131–1142,
the standard buck-converter processing circuit is considerably Oct. 2004.
[3] S. P. Beeby, R. N. Torah, M. J. Tudor, P. Glynne-Jones, T. O’Donnell,
higher than the efficiency of the buck–boost converter, as was C. R. Saha, and S. Roy, “A micro electromagnetic generator for vibration
predicted by theoretical calculations and simulations. More- energy harvesting,” J. Micromech. Microeng., vol. 17, no. 7, pp. 1257–
over, the measurements show that the overall power output 1265, Jul. 2007.
[4] P. D. Mitcheson, T. Sterken, C. He, M. Kiziroglou, E. M. Yeatman, and
of the piezoelectric harvesting system with the standard buck R. Puers, “Electrostatic microgenerators,” Meas. Control, vol. 41, no. 4,
converter as power processing circuit is higher, even though the pp. 114–119, May 2008.
harvester power output with the converter providing the optimal [5] M. D. Seeman, S. R. Sanders, and J. M. Rabaey, “An ultra-low-power
power management IC for energy-scavenged wireless sensor nodes,” in
resistive load is higher. Proc. IEEE Power Electron. Spec. Conf., Rhodes, Turkey, Jun. 2008,
pp. 925–931.
[6] S. Xu, K. D. T. Ngo, T. Nishida, G. B. Chung, and A. Sharma, “Low
VI. C ONCLUSION frequency pulsed resonant converter for energy harvesting,” IEEE Trans.
Power Electron., vol. 22, no. 1, pp. 63–68, Jan. 2007.
In this paper, the behavior of a piezoelectric vibration-driven [7] M. Marzencki, Y. Ammar, and S. Basrour, “Design, fabrication and char-
energy harvester has been assessed with two different power acterization of a piezoelectric microgenerator including a power man-
processing circuits. An optimal linear resistive load, as well agement circuit,” in Proc. Symp. DTIP MEMS/MOEMS, Stresa, Italy,
Apr. 25–27, 2007.
as an optimal dc voltage load of the energy harvester, can [8] G. K. Ottman, H. F. Hofmann, and G. A. Lesieutre, “Optimized piezoelec-
be analytically calculated for every operating frequency. To tric energy harvesting circuit using step-down converter in discontinuous
determine which load leads to the optimal overall efficiency conduction mode,” IEEE Trans. Power Electron., vol. 18, no. 2, pp. 696–
703, Mar. 2003.
of a system consisting of an energy harvester connected to [9] D. Guyomar, A. Badel, E. Lefeuvre, and C. Richard, “Towards energy har-
a power processing circuit, two power processing circuits are vesting using active materials and conversion improvement by nonlinear
D’HULST et al.: POWER PROCESSING CIRCUITS FOR PIEZOELECTRIC VIBRATION-BASED ENERGY HARVESTERS 4177

processing,” IEEE Trans. Ultrason., Ferroelectr., Freq. Control, vol. 52, Tom Sterken (M’02) received the Diploma degree in
no. 4, pp. 584–595, Apr. 2005. electrical engineering from the University of Ghent,
[10] E. Lefeuvre, D. Audigier, C. Richard, and D. Guyomar, “Buck-boost Ghent, Belgium, in 2001 and the Ph.D. degree
converter for sensorless power optimization of piezoelectric energy har- from the Katholieke Universiteit Leuven, Leuven,
vester,” IEEE Trans. Power Electron., vol. 22, no. 5, pp. 2018–2025, Belgium, in 2009.
Sep. 2007. From 2002 to 2008, he was with the Interuniver-
[11] P. Mitcheson, T. Green, E. Yeatman, and A. Holmes, “Architectures sitary Center for Microelectronics, Leuven, working
for vibration-driven micropower generators,” J. Microelectromech. Syst., on the design, modeling, and fabrication of miniature
vol. 13, no. 3, pp. 429–440, Jun. 2004. power generators based on MEMS technology. This
[12] Y. Hu, H. Xue, T. Hu, and H. Hu, “Nonlinear interface between the research resulted in his Ph.D. degree. He is currently
piezoelectric harvesting structure and the modulating circuit of an energy with the Centre for Microsystems Technology group,
harvester with a real storage battery,” IEEE Trans. Ultrason., Ferroelectr., University of Ghent, where he is working on ultrathin chip packaging and
Freq. Control, vol. 55, no. 1, pp. 148–160, Jan. 2008. stretchable electronics.
[13] L. Garbuio, M. Lallart, D. Guyomar, C. Richard, and D. Audigier, “Me-
chanical energy harvester with ultralow threshold rectification based on
SSHI nonlinear technique,” IEEE Trans. Ind. Electron., vol. 56, no. 4, Robert Puers (SM’95) was born in Antwerp,
pp. 1048–1056, Apr. 2009. Belgium, in 1953. He received the B.S. degree in
[14] A. Tabesh and L. Frechette, “An accurate analytical model for electro- electrical engineering from University of Ghent,
mechanical energy conversion in a piezoelectric cantilever beam for Ghent, Belgium, in 1974 and the M.S. and Ph.D.
energy harvesting,” in Proc. 7th Int. Workshop Micro Nanotechnol. degrees from the Katholieke Universiteit Leuven
Power Gener. Energy Convers. Appl. PowerMEMS, Freiburg, Germany, (K.U.Leuven), Leuven, Belgium, in 1977 and 1986,
Nov. 2007, pp. 61–64. respectively.
[15] A. Badel, A. Benayad, E. Lefeuvre, L. Lebrun, C. Richard, and In 1980, he joined ESAT, K.U.Leuven, as a Re-
D. Guyomar, “Single crystals and nonlinear process for outstanding search Assistant, where he became the Director
vibration-powered electrical generators,” IEEE Trans. Ultrason., Ferro- (NFWO) of the clean room facilities for silicon
electr., Freq. Control, vol. 53, no. 4, pp. 673–684, Apr. 2006. and hybrid circuit technology at the ESAT–MICAS
[16] R. D’hulst, T. Sterken, P. Fiorini, R. Puers, and J. Driesen, “Energy laboratories in 1986. He was a pioneer in the European research efforts
scavengers : Modeling and behavior with different load circuits,” in Proc. in silicon micromachined sensors, microelectromechanical systems (MEMS),
33rd IEEE IECON, Taipei, Taiwan, Nov. 2007, pp. 2169–2174. and packaging techniques, for biomedical implantable systems as well as for
[17] N. duToit, B. Wardle, and S. Kim, “Design considerations for MEMS- industrial devices. He is currently a full Professor with K.U.Leuven, teaching
scale piezoelectric mechanical vibration energy harvesters,” Integr. Fer- courses in “microsystems and sensors” and “biomedical instrumentation and
roelectr., vol. 71, no. 1, pp. 121–160, Jul. 2005. stimulation” and also a basic course in “lectronics, system control, and infor-
[18] Y. C. Shu and I. C. Lien, “Analysis of power output for piezoelectric mation technology.” He is the author or coauthor of more than 350 papers on
energy harvesting systems,” Smart Mater. Struct., vol. 15, no. 6, pp. 1499– biotelemetry, sensors, MEMS, and packaging in reviewed journals or interna-
1512, Dec. 2006. tional conferences. He is the Editor-in-Chief of the Journal of Micromechanics
[19] M. Renaud, T. Sterken, A. Schmitz, P. Fiorini, C. Van Hoof, and R. Puers, and Microengineering of the Institute of Physics (IOP).
“Piezoelectric harvesters and MEMS technology: Fabrication, model- Dr. Puers is a Fellow of the IOP (U.K.) and a council member of the Interna-
ing and measurements,” in Proc. 14th TRANSDUCERS, Lyon, France, tional Microelectronics and Packaging Society. He is the General Chairman of
Jun. 2007, pp. 891–894. the Eurosensors conferences.
[20] R. D’hulst, P. Mitcheson, and J. Driesen, “CMOS buck-boost power
processing circuitry for PowerMEMS harvesters,” in Proc. 6th Int. Work- Geert Deconinck (SM’01) received the M.Sc. de-
shop Micro Nanotechnol. Power Gener. Energy Convers. Appl. Power- gree in electrical engineering and the Ph.D. degree in
MEMS, Berkeley, CA, Nov. 29–Dec. 1, 2006. engineering from the Katholieke Universiteit Leuven
[21] G. K. Ottman, H. F. Hofmann, A. C. Bhatt, and G. A. Lesieutre, “Adaptive (K.U.Leuven), Leuven, Belgium, in 1991 and 1996,
piezoelectric energy harvesting circuit for wireless remote power supply,” respectively.
IEEE Trans. Power Electron., vol. 17, no. 5, pp. 669–676, Sep. 2002. He was a Postdoctoral Fellow of the Fund
[22] R. Sodhi, S. Brown, Sr., and D. Kinzer, “Integrated design environment for Scientific Research—Flanders from 1997 to
for DC/DC converter FET optimization,” in Proc. 11th ISPSD, Toronto, 2003. He is currently a full Professor (hoogleraar)
ON, Canada, May 1999, pp. 241–244. with K.U.Leuven. He is also a Staff Member of
[23] S. Musunuri and P. L. Chapman, “Optimization of CMOS transistors for ESAT/ELECTA (Electrical Energy and Computing
low power DC-DC converters,” in Proc. IEEE 36th Power Electron. Spec. Architectures), K.U.Leuven, where he performs re-
Conf., Recife, Brazil, Jun. 2005, pp. 2151–2157. search on designing dependable system architectures for industrial automa-
[24] Y. Katayama, M. Edo, T. Denta, T. Kawashima, and T. Ninomiya, “Op- tion and control, assessing their dependability attributes and characterizing
timum design method of CMOS IC for DC-DC converter that integrates infrastructure interdependences.
power stage MOSFETs,” in Proc. IEEE 35th PESC, Jun. 2004, vol. 6, Dr. Deconinck is a member of the IEEE SMC Technical Committee on
pp. 4486–4491. Infrastructure Systems and Services, the Royal Flemish Engineering Society,
and the Institute of Engineering and Technology. He is a senior member of the
IEEE Reliability Society, IEEE Computer Society, and IEEE Power and Energy
Society. He is the Chairman of the TI society BIRA on industrial automation.

Johan Driesen (M’97) was born in Belgium in


1973. He received the M.Sc. degree and the Ph.D.
degree in electrical engineering from the Katholieke
Universiteit Leuven (K.U.Leuven), Leuven,
Belgium, in 1996 and 2000, respectively. His Ph.D.
Reinhilde D’hulst received the M.Sc. degree in focused on the finite-element solution of coupled
electrical engineering and the Ph.D. degree from thermal–electromagnetic problems and related appli-
the Katholieke Universiteit Leuven (K.U.Leuven), cations in electrical machines and drives, micro-
Leuven, Belgium, in 2004 and 2009, respectively. systems, and power quality issues.
As a member of the Electrical Energy research From 2000 to 2001, he was a Visiting Researcher
group (ELECTA), Department of Electrical Engi- with the Imperial College of Science, Technology
neering, K.U.Leuven, she worked on power process- and Medicine, London, U.K. In 2002, he was with the University of California,
ing circuits for energy harvesters. She is currently Berkeley. He is currently an Associate Professor with K.U.Leuven, where he
with VITO, the Flemisch Institute for Technological teaches power electronics and drives. Currently, he conducts research on distrib-
Research, where she conducts research on power uted energy resources, including renewable energy systems, power electronics,
electronics in a smart grid environment. and its applications, for instance, in renewable energy and electric vehicles.

You might also like