Accepted Manuscript

Analysis of the energy extracted by a harvester based on a piezoelectric tile

Furio Cascetta, Alessandro Lo Schiavo, Aldo Minardo, Marilena Musto, Giuseppe

Rotondo, Alessio Calcagni

PII: S1567-1739(18)30111-1
DOI: 10.1016/j.cap.2018.04.015
Reference: CAP 4734

To appear in: Current Applied Physics

Received Date: 15 December 2017

Revised Date: 21 March 2018
Accepted Date: 20 April 2018

Please cite this article as: F. Cascetta, A.L. Schiavo, A. Minardo, M. Musto, G. Rotondo, A. Calcagni,
Analysis of the energy extracted by a harvester based on a piezoelectric tile, Current Applied Physics
(2018), doi: 10.1016/j.cap.2018.04.015.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

1 Analysis of the energy extracted

2 by a harvester based on a piezoelectric tile
3
4 Furio Cascetta1, Alessandro Lo Schiavo1, Aldo Minardo1*, Marilena Musto2, Giuseppe Rotondo2, Alessio Calcagni3
5

PT
6
7 In this paper, we analyze the maximum energy that can be extracted from a piezoelectric harvester
8 subject to pulsed excitation, with an interface circuit composed by a standard bridge rectifier. We show

RI
9 that the optimal voltage of the DC load of the bridge rectifier is a fraction, comprised between 1/3 and
10 ½, of the open-circuit voltage, depending on the piezoelectric losses and excitation time. A simple
11 analytical model is provided, whose accuracy has been assessed against SPICE simulations. Furthermore,

SC
12 preliminary experimental tests carried out over a commercial piezoelectric tile confirm the validity of
13 the proposed model.
14
15 Keywords: Energy harvesting, Piezoelectricity, Vibration

U
16
17 1. Introduction
AN
18 Among the various energy harvesting technologies, piezoelectric vibration has emerged as a method for
19 harvesting from to macro-to-micro scale [1-4]. Piezoelectric materials can be designed to handle a wide
20 range of input frequencies and forces allowing for energy harvesting to occur. Although studies on
M

21 piezoelectric energy harvesting systems have been extensively conducted over the past years [5-7], this
22 application is still in development and therefore, its potential has not yet been fully exploited. Studies
23 based on piezoelectric energy harvesting from human force include the energy harvested from the
D

24 bending of elbow or finger joints [8], implants in the knee joints [9], piezoelectric modules inserted
25 under the soles of shoes [10-12], or motion of the human limbs [13]. However, these cases cannot be
TE

26 considered as macro-sources because of their limited installation area, but independent units such as
27 piezoelectric tiles can be planted over a wider area; thus, they can be used as macro-power sources [14].
28 Key aspects of the use of piezoelectric tiles on a large scale are cost reduction and quality of harvesting
EP

29 systems. The first aspect mainly depends on the efficient use of the system, the integration of various
30 piezoelectric technologies, the reduced device break times by searching for new materials and designing
31 appropriate energy transfer facilities. In this regard, Adnan M. Elhalwagy et al. [15] provided a guide to
C

32 facilitate embedding piezoelectric tile technology in the designs as a part of the demanded low energy
33 consumption in the buildings, while Xiaofeng Li et al [16] carried out an optimization of the piezoelectric
AC

34 tile deployment in according to the frequency of pedestrian mobility. Their study confirms that selecting
35 high traffic areas is critical for the optimization of the energy harvesting efficiency; furthermore, the
36 orientation of the tile pavement significantly affects the total amount of the harvested energy. The
37 second aspect depends on the efficiency in transferring mechanical energy from the environment to the

1
Università degli Studi della Campania “Luigi Vanvitelli”, Dipartimento di Ingegneria Industriale e
dell’Informazione, via Roma 29, 81031 Aversa (CE)
2
Università degli Studi di Napoli, Dipartimento di Ingegneria Industriale, P. le Tecchio 80, 80125 - Napoli
3
Luche S.r.l., via Cagliari 215, 09012, Capoterra (CA)
* Email address: aldo.minardo@unicampania.it
 ACCEPTED MANUSCRIPT

38 piezoelectric material, and on the coupling between the piezoelectric material and the electronic circuits
39 for rectifying, regulating electricity and storing the energy. From the design point of view, the biggest
40 challenge in transferring the mechanical energy to the piezoelectric harvester, is the low input
41 frequency and acceleration of typical mechanical energy sources. Most vibrational energy harvesters are
42 designed to operate in resonance mode and the half-power bandwidth is usually small. As a result,
43 matching a low frequency and acceleration input to a piezoelectric harvester architecture becomes
44 challenging.

PT
45 In this paper, we focus our analysis to a piezoelectric tile employing direct conversion energy from a
46 human footstep. In particular, we report an analytical model that predicts the optimal DC voltage load in
47 a piezoelectric tile subjected to a pulsed excitation. We show that the optimal voltage load is different

RI
48 from the optimal value found in case of sinusoidal excitation. In particular, the model shows that the
49 optimal voltage is comprised between 1/3 and ½ of the open-circuit voltage, depending on the
50 characteristics of the mechanical solicitation, and the piezoelectric capacitance discharge time.

SC
51 Numerical SPICE simulations, performed for different durations of the mechanical solicitation, confirm
52 the validity of the proposed model. Furthermore, we have characterized the energy extracted from a
53 commercial piezoelectric tile, under typical load conditions. The analysis of the energy provided by the

U
54 tile, as a function of output voltage, confirm the predictions of the proposed model.
55
AN
56 2. Analytical model

57 Various load circuits have been proposed and demonstrated to maximize the energy (and the power)
58 extracted from a piezoelectric harvester. All essentially function by properly regulating the force that the
M

59 transducer presents to the mechanical system [17]. The force presented by the piezoelectric transducer
60 to the mechanical system is influenced by the impedance and operation of the electrical circuit that is
61 connected to it. The transducer interface circuit can, therefore, be designed to allow modification of the
D

62 damping force, and different types of circuit are capable of modifying the electrical damping by different
63 amounts. Because of the need to transform the voltage from time-varying to DC, the most widely used
TE

64 circuit connected to a piezoelectric harvester is a standard bridge rectifier with a DC smoothing

65 capacitor. The voltage level of the DC capacitor significantly influences the force presented by the
66 transducer to the mechanical system and, hence, the power extracted by the harvester. This justifies the
EP

67 significant research efforts both to model the dependence of the extracted power from the DC voltage
68 level [17-18], and to actively regulate this voltage level through MPPT (Maximum Power Point Tracking)
69 algorithms implemented in DC/DC switching converters [19]. Although most of the studies and results
70 are focused on purely sinusoidal mechanical vibrations, there exist no less significant cases where
C

71 mechanical vibrations are not sinusoidal, as in the case of piezoelectric harvesters placed in floor tiles.
72 Therefore, the principal aim of our analysis is determining the optimal dc voltage ensuring the maximum
AC

73 energy (and power) extracted from the tile, for any mechanical stress applied on it.
74 We start our analysis from the model expressed in Ref. [18], which governs the piezoelectric vibrator
75 under the spring mass damper approximation:
76

77
  +  
 + Θ  = , (1)

78 −Θ
 +  
 = −  (2)

79
 ACCEPTED MANUSCRIPT

80 In Eqs. (1-2),  is the displacement of the mass
;  is the voltage across the piezoelectric element;
81  is the forcing function;  is the damper coefficient; Θ amd  are the effective piezoelectric
82 coefficient and capacitance of the piezoelectric element, respectively;   is the current provided by
83 the piezoelectric vibrator. Under the common assumption of weak electro-mechanical displacement
84 [17], any level of damping from the electrical side makes negligible difference to the velocity of the
85 proof mass. In such a case, the piezoelectric harvester can be simply modeled by a current generator
  = Θ
 and a shunt capacitance  . As earlier discussed, the transducer interface circuit is

PT
86
87 composed by a full-bridge rectifier followed by a DC smoothing capacitor. Taking into account that the
88 large value of the capacitance makes the voltage variation negligible during a single step, the DC
89 capacitor is here modeled by a constant voltage source, imposing a voltage  . Finally, we add the

RI
90 parallel resistor  in order to take into account the leakage loss of the piezoelectric capacitance. The
91 overall circuit is represented in Fig. 1.

SC
92
93

Piezoelectric element Bridge rectifier

iC iDC

U
AN
ip Cp Rp Vp VDC
M

94
95 Fig.1 Schematic of the simplified piezoelectric circuit.
D

96
97 As Eq. (2) suggests, the electrical current  provided by the piezoelectric element is proportional to the
TE

98 displacement velocity 
. Under the action of a footstep, we assume that the relevant quantities
99 follow the temporal behavior depicted in Fig. 2. In particular, the displacement is supposed to follow a
100 trapezoidal waveform, with a rise time equal to ∆ , a fall time equal to ∆ , and a uniform forcing interval
equal to . As indicated in Fig. 2, the time interval ∆ can be further split in two intervals: in the first one,
EP

101
102 of duration  , the current  charges the capacitance  up to  . The charge placed on  during this
103 interval can be calculated applying the charge conservation law:
104
C

$
&  − ! "# =  
%
105 (3)
AC

106
107 From Eq. (3) we easily derive:
108
()*
 τ − τ =  
 !+ 
109 (4)
110
111 And therefore:
+ ()*
112 τ = (5)
,+ -()* /!+ 
113
 ACCEPTED MANUSCRIPT

114
115

PT
RI
U SC
AN
116
117
M

Fig. 2 Voltage and current waveforms under impulsive forcing.

118
119 During the second part of the interval, i.e. for  /  / 0, the voltage  keeps stable to  as the
120 diodes of rectifiers are on (we neglect the voltage drop on diodes). During this time, the output current
D

121 is nonzero and the energy transferred to the DC load, i.e. to  , is:
122
TE

3% 3%
1 = 2  − ! " # = 2  − ! " # =
& $%
()* + ()* ()*
123 = Δ − τ  56 − 7  = 5Δ − 8 7 56 − 7  (6)
EP

!+ + -()* /!+  !+
124
125 It is useful to introduce the quantity  _: , which is the open-circuit voltage of the piezoelectric element.
C

126 This voltage, measured when the bridge rectifier is not connected to the piezoelectric element, reaches
127 a peak value, which can be calculated following these passages:
AC

128
∆% (+_;< 8+ ∆%
129 &  − ! "# =   _:
6 ∆ − ∆ =   _:
 _: = (7)
 !+ + =∆/!+ 
130
131 Note that the circuit in Fig. 2 only produces power if the voltage rise due to the current source is high
132 enough to cause current to flow to the DC load, that is  _: >  .
133 From  = 0 to  = 0 + , the displacement is constant so that both piezoelectric and DC currents are
134 zero. During this time, the voltage on the piezoelectric capacitance discharges by a loss factor
135 @ = A -B/!+ + .
136 During the second charging phase, the charge placed on  is:
 ACCEPTED MANUSCRIPT

137
∆% =B=CD
138 ∆  − ! "# =  1 + @ (8)
% =B
139
140 where τ is the time after which the piezoelectric voltage reaches − . From Eq. (8) we easily derive:
141
(
142 6 τ −  !
)*
τ ≈  1 + @ (9)

PT
+
+ ()* =G
143 τ = (10)
8+ -()* /!+ 
144
After this time, the voltage  keeps stable to − as the diodes of rectifiers are on, while the DC

RI
145
146 current is nonzero. In particular, the energy transferred to the load during this time is:
147

SC
3 =B=3D ()* + ()* =G ()*
148 1 = 3 %=B=C  − ! " # = Δ − τ  56 − !+
7  = 5Δ − 8 7 56 − 7 
% D + -()* /!+  !+
149 (11)
150

U
151 Assuming for simplicity Δ = Δ = ∆, the total energy transferred to the load is:
152
AN
+ ()* =G
153 1 = 1 + 1 =  6  H2∆ − J (12)
8+D
154
()* ()*
where 6  ≡ 6 − and 6  ≡ 6 −
M

155 .
!+ !+
156
157 It should be observed that, according to (12), the energy 1 transferred to the load reduces in the typical
D

158 case of small piezoelectric losses, that is when @ is large (@ ≈ 1). Actually, in this case,  is much smaller
159 than   and the piezoelectric capacitance does not have enough time to discharge from + to 0
TE

160 before the second current impulse arrives. Thus, the current generator,  , must reverse the charge on
161 the capacitance before it can supply the dc load. As shown in Fig. 2, the second current impulse in the dc
162 load, M , is shorter than the first impulse, even if the two generated impulses,  , have equal duration
(asymmetric operation). On the other hand, if  is sufficiently large (@ ≈ 0), the capacitance discharges
EP

163
164 to zero and the second current impulse in the dc load, M , is equal to the first one (symmetric
165 operation).
C

166 Moreover, if we compare the situation here analyzed with that of sinusoidal forcing, the currently
167 analyzed case is more advantageous because, when the first positive impulse arrives, the capacitance
AC

168 has no charge, while in sinusoidal operation, the capacitance has always a charge opposite to the
169 incoming half-wave.
170 In order to determine the optimal output voltage, we calculate the first derivative of energy and put it to
171 zero. For this calculation, we assume for simplicity that 6  e 6  are only weakly dependent on  , i.e.
()*
172 6 ≫ !+
. In this approximation, we can write:
173
MO 8+%
174 ≈ 26  ∆ − 2  2 + @ (13)
M()* 8+D
175
 ACCEPTED MANUSCRIPT

176 Therefore, we achieve:

177
8+D ∆ 8+ ∆
178 _: P = ≈ (14)
+ =G + =G
179
180 Neglecting loss on resistor  , the optimal output voltage can be expressed as:
181

PT
(+_;<
182 _: P ≈ (15)
=G
183
184 Equation (15) is the main result of our work: it shows that, in case of low loss (@ ≈ 1, i.e.  ≪   ),

RI
185 the optimal voltage is close to one third of the open-circuit voltage, in contrast with the value of one-
186 half of the open-circuit voltage found in Ref. [18], and referring to the case of sinusoidal forcing. Vice
versa, in conditions of high loss (@ ≈ 0, . A.  ≫   , the optimal voltage is half the open-circuit

SC
187
188 voltage, as in the sinusoidal case. Note that in case of @ ≈ 0, the absolute voltage variation on the
189 capacitance  is the same for the positive and the negative pulses of  , that is | |. A symmetric
190 operation happens also in case of sinusoidal forcing and justifies the same result. On the other hand, as

U
191 described above, the operation in the typical case of low piezoelectric losses, is asymmetric and leads to
192 a result different from the one holding in the sinusoidal case.
AN
193 Using the optimal voltage _: P expressed by Eq. (15), we derive the maximum energy transferred to
194 the load:
195
8+% 8+D ∆D
M

196 1 ST = (16)
+ =G
197
In case of U piezoelectric elements connected in parallel and interfaced to separate rectifiers, the total
D

198
199 energy is:
TE

200
+ ()* =G
201 1BVB =  U  6  H2∆ − J (17)
8+D
202
where  is an efficiency factor, which takes into account possible unbalances between forcing of various
EP

203
204 transducers. Finally, we note that the time waveform of current is less important that its area. In fact,
205 the extracted energy can be rewritten as:
C

206
+ ()* =G
207 1BVB ≈  U W  H2 − X+
J (18)
AC

208
209 where W = 6 ∆ is the total charge generated by the transducer if resistance loss are neglected.
210
211
212 3. Numerical validation

213 In order to validate the model presented in the previous section, we have used SPICE to simulate a
214 system composed by three piezoelectric elements with separate rectifiers, and connected in parallel to
215 the same DC load. Each piezoelectric vibrator was modeled with a shunt capacitance  = 20 Y and
 ACCEPTED MANUSCRIPT

216 a parallel resistance Z = 20
Ω. These values were chosen based on measurements done on the
217 piezoelectric diaphragms, composing the tile used for the experimental tests presented later. The other
218 parameters have been chosen as follows: 6 = 25 ]^; Δ = 85 `a;  = 200 `a. Finally, we assume
219 an efficiency  = 1. The chosen parameters correspond to a loss factor α ≈ 0.61.
220

PT
u
1
u
2
u
3

RI
Voltages [V]
0 0.2 0.4 0.6 0.8 1 1.2
Time [s]

25

SC
i p1
i p2
0
i
p3

-25

U
0 0.2 0.4 0.6 0.8 1 1.2
Time [s]
221
AN
222 (a) (b)
223 Fig. 3 SPICE simulation of a system composed by three piezoelectric harvesters connected in parallel, in absence of
224 load: (a) displacements and corresponding input currents; (b) piezoelectric voltages.
M

225
226 Figure 3 shows the results of the simulation performed when no voltage generator is connected to the
227 output of the bridge rectifiers, and with the three piezoelectric vibrators forced by trapezoidal
228 displacements shifted in time by 100 ms . The displacement applied to each piezoelectric element cause
D

229 a positive and a negative current pulse for each piezoelectric current generator. The latter first charges
230 and then discharges, together with the resistance Rp, its piezoelectric capacitance, as shown by the
TE

231 piezoelectric voltage waveforms. Due to the parallel connection of the three bridge rectifiers, the output
232 DC voltage follows the greater absolute value among the three piezoelectric voltages, producing a multi-
233 peak waveform that will be observed also in the experimental results presented in the next Section.
234 As a next step, we consider the case of piezoelectric harvesters connected to a voltage generator 
EP

235 imposing a fixed output voltage. As an example, we show in Fig. 4 the simulated waveforms in case of an
236 output voltage fixed to 30 V. In particular, we report in the same graph the energy cumulated on the
237 load, 1BVB , (blue solid line), together with the output current  (red dashed line). We see that, after
C

238 the completion of the sixth current pulse (i.e. at t = 670 ms), the transferred energy has reached a value
239 of ≈ 220 µJ.
AC

240
 ACCEPTED MANUSCRIPT

Output current [ A]
Energy [ J]

PT
RI
SC
241
242 Fig. 4 SPICE simulation of a system composed by three piezoelectric harvesters connected in parallel, with an

U
243 output voltage of 30V. The solid line is the extracted energy, while the dashed line is output current.

244
AN
245 Finally, we compare in Fig. 5 the conveyed energy as a function of the output voltage, as obtained by
246 SPICE simulations or computed using our model (Eq. (17)). We see that a good agreement exists
247 between numerical and analytical results, at least in the region of interest (maximum energy). In
M

248 particular, the more relevant parameters derived from SPICE simulations are: open-circuit voltage
249  _: = 96 ; optimal output voltage _: P = 37; maximum extracted energy 1 ST = 227 f. The
corresponding analytical values, obtained by use of Eqs. (7), (15) and (16), are  _: = 96 , _: P =
D

250
251 36.8  and 1 ST = 231 f. It is clear that SPICE simulations confirm the validity of the proposed
TE

252 model.
240
SPICE simulation
220 Analtyical model
EP

200

180

160
Energy [ J]
C

140
AC

120

100

80

60

40
0 10 20 30 40 50 60 70
V [V]
253 DC
254 Fig. 5 Extracted energy as a function of output voltage, as obtained by SPICE simulations (blue solid line), or using
255 our model (red circles).
 ACCEPTED MANUSCRIPT

256 In order to assess the accuracy of our model for different pulse durations, we have performed new
257 SPICE simulations with  varying from 100 ms to 2s. All other parameters were kept constant. We report
258 in Fig. 6 the optimal output voltage (left vertical axis) and maximum energy (right vertical axis) as a
259 function of pulse duration, as retrieved using SPICE (solid lines), or our analytical model (circles). Note
260 that, the loss factors α corresponding to the chosen pulse durations  are reported on top of the graph.
261 We see that analytical and numerical results are in good agreement, especially in case of low loss
262 (@ ≈ 1. For increasing loss (@ → 0, our model provides larger values of optimal output voltage and

PT
263 maximum energy. In any case, the deviation between SPICE-derived optimal voltage and analytical
264 results is less that ≈ 1% over the considered pulse duration interval, while the deviation relative to
265 maximum energy is less than ≈ 2%.

RI
0.78 0.61 0.47 0.37 0.29 0.22 0.17 0.14 0.11 0.08
48 300

SC
290
46

280
44
270

U
42 260
AN
40 250

240
38
230
M

36
220

34 210
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
D

266 T [s]

267 Fig. 6 Optimal output voltage (left vertical axis) and maximum energy (right vertical axis), as a function of pulse
TE

268 duration T, as obtained by SPICE simulations (solid lines), or using our model (circles).

269 Using the same simulation data, we have also computed the ratio between the optimal output voltage
270 and the open-circuit voltage, as a function of the loss factor α. The simulation results are shown in Fig. 7,
EP

271 together with the values provided by our model. Still, SPICE simulations and data from our analytical
272 model are in overall agreement, confirming the general trend expressed by the approximate formula
273 (15).
C
AC
 ACCEPTED MANUSCRIPT

0.46
SPICE simulation
Analtyical model
0.44

0.42

PT
0.4

0.38

RI
0.36

0.34

SC
0.32
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
274
275 Fig. 7 Ratio between optimal output voltage and open-circuit voltage, as a function of the loss factor α, as

U
276 computed using SPICE (solid line), or our model (red circle).

277
AN
278 4. Experimental results

279 We have carried out a number of experimental tests over a piezoelectric tile produced by Luche S.r.l.
M

280 [20]. A picture of the sample used for the tests is shown in Fig. 8(a). The tile uses Lead Zirconate Titanate
281 as piezoelectric material. The input mechanical energy is transformed into electrical energy through nine
282 stacks, placed under the upper surface of the tile and each composed by five piezoelectric diaphragms
D

283 (buzzers) spaced by rings to allow the vibration, as shown in Fig. 8(b) and Fig.8(c). Each buzzer, a
284 Piezotite 7BB-35-3L0 by Murata [21], is connected to a full-bridge rectifier, and the outputs of all bridges
TE

285 are connected in parallel to the output terminals of the tile (see the scheme of Fig. 9).
286
C EP
AC
 ACCEPTED MANUSCRIPT

PT
287 (a)

RI
U SC
AN
288 (b) (c)
289 Fig. 8 Piezoelectric tile used for experimental tests. (a) picture (b) internal structure with nine housings for the
290 stacks of piezoelectric diaphragms (c) Each stack composed by piezoelectric diaphragms spaced by rings.
291
M
D
TE
C EP

292
293 Fig. 9 Scheme of the electrical connections among piezoelectric diaphragms in the piezoelectric tile used for
AC

294 experiments (N=45).

295
296 The electrical characterization was performed using a digital sampling oscilloscope with a 500MHz
297 bandwidth. The oscilloscope was used to capture voltage and current signals, while two single-output
298 stabilized power supplies, capable of generating voltages up to 32 V, were used in order to impose the
299 desired voltage on the rectifier bridges.
300 The first experiment was aimed to determine the open-circuit voltage provided by the piezoelectric tile
301 in the presence of a typical mechanical stress. The latter was produced by the footstep of a man of
302 about 85 kg, walking over it, and resting on a single foot. The typical recorded voltage waveform is
 ACCEPTED MANUSCRIPT

303 shown in Fig. 10. This test was repeated several times under the same operating conditions, in order to
304 evaluate the repeatability of the measured data. The tests have shown that the open-circuit voltage was
305 comprised between 100 V and 110 V. The variability of this voltage is attributed to the sensitivity of the
306 tile to the exact distribution of the applied load over its surface. After application of the mechanical
307 stress, the open-circuit voltage at the electrical terminals returned to zero after about 3 s, showing an
308 exponentially decaying response. The open-circuit voltage waveform also reveals the presence of
309 several peaks. This behavior is compatible both with the rectification of the AC voltage generated by

PT
310 each piezoelectric transducer, as well as with the non-synchronized mechanical forcing of the various
311 piezoelectric transducers located within our tile.
312

RI
120

SC
100

80

U
Voltage [V]

60
AN
40

20
M

-20
0 0.5 1 1.5 2 2.5 3 3.5
D

313 Time [s]

314 Fig. 10. Typical waveform of the open-circuit voltage generated by the piezoelectric tile under a footstep of a man
TE

315 of about 85Kg.

316
317 Successive tests were aimed to determine the energy extracted by the tile, when subjected to
318 mechanical stress, while varying the voltage imposed by the generator connected to the tile’s output
EP

319 terminals. For this test, the current supplied by the tile at the output terminals was measured though a
320 test resistance of 1 kΩ, connected in series to the applied DC voltage. The test resistance was chosen to
321 be small enough, in order to cause a negligible voltage drop compared to the DC voltage. The latter has
C

322 been varied from 0 V to 64 V, which was lower than the maximum open-circuit voltage, but larger than
323 the value providing maximum energy extraction. The results of three test series, together with a fitting
AC

324 curve obtained by using the average data, are reported in Fig. 11.
 ACCEPTED MANUSCRIPT

PT
RI
U SC
325
AN
326 Fig. 11 Extracted energy as a function of load voltage. Experimental and analytical data.

327 In order to compare the experimental results with analytical results from our model, we have used it
328 with parameters identical to those used in the previous paragraph, except the number of piezoelectric
elements increased up to U = 5 × 9 = 45, and the efficiency  set to 0.7. The computed energies are
M

329
330 compared to experimental data in Fig. 11. A reasonable agreement between experimental and model
331 data is observed, especially in terms of optimal output voltage. In fact, the experimental optimal
D

332 voltage, extracted from the fitting curve, is 35 V, while our model predicts an optimal voltage of 37 V. In
333 any cases the ratio between the optimal voltage and experimentally determined open-circuit voltage
value ( ;< ≈ 105 ), is ≈ 0.35, in decent agreement with the factor provided by Eq. (15) for α = 0.61,
TE

334
335 i.e. 1⁄2 + @ = 0.38. Finally, it is important to observe that, while the efficiency factor used for the
336 data shown in Fig. 11 was chosen arbitrarily, in order to quantitatively match the experimentally
337 measured energies, this choice does not influence the ratio between optimal output voltage and open-
EP

338 circuit voltage (see Eq. (15)), which is the real parameter of interest in our analysis.
339
340
C

341 5. Conclusions
AC

342 A simple analytical expression for the optimal output voltage in piezoelectric harvesters subjected to
343 pulsed excitation has been derived, in the typical context of a piezoelectric tile. The model reveals that
344 the optimal output voltage is comprised between 1/3 and ½ of the open-circuit voltage, depending on
345 the ratio between the discharge time of the piezo capacitance and the duration of the pulsed excitation.
346 The developed model has been assessed against SPICE simulations, as well as experimental tests
347 performed over a commercial piezoelectric tile.

348

349 Acknowledgements
 ACCEPTED MANUSCRIPT

350 This work has been partly funded by Luche srl (www.veranu.eu).

351

352

353 References

PT
354 [1] S Roundy and P K Wright, “A piezoelectric vibration based generator for wireless electronics,” Smart
355 Mater. Struct. 13 1131 (2004).

356 [2] Steven R. Anton and Henry A. Sodano, “A review of power harvesting using piezoelectric materials

RI
357 (2003–2006),” Smart Mater. Struct. 16 R1 (2007).

358 [3] Huidong Li, Chuan Tian, and Z. Daniel Denga, “Energy harvesting from low frequency applications

SC
359 using piezoelectric materials,” Applied Physics Reviews 1, 041301 (2014)

360 [4] Alperen Toprak and Onur Tigli, “Piezoelectric energy harvesting: State-of-the-art and challenges,”
361 Applied Physics Reviews 1, 031104 (2014)

U
362 [5] Y.B. Jeon, R. Sood, J.H. Jeong, S.G. Kim, “MEMS power generator with transverse mode thin film
363 PZT,” Sensors and Actuators, A: Physical, 122 (2005).
AN
364 [6] H.W. Kim, S. Priya, K. Uchino, R.E. Newnham, “Piezoelectric energy harvesting under high prestressed
365 cyclic vibrations,” Journal of Electroceramics, 15 (2005) 27.
M

366 [7] H. Kim, S. Priya, K. Uchino, “Modeling of piezoelectric energy harvesting using cymbal transducers,”
367 Jpn. J. Appl. Phys., Part 1, 45 (2006).
D

368 [8] B. Yang, K.S. Yun, “Piezoelectric shell structures as wearable energy harvesters for effective power
369 generation at low-frequency movement,” Sensors and Actuators, A: Physical, 188 (2012)
TE

370 [9] S.R. Platt, S. Farritor, K. Garvin, H. Haider, “The use of piezoelectric ceramics for electric power
371 generation within orthopedic implants, IEEE/ASME Trans. Mechatron., 10 (2005).

372 [10] L. Moro, D. Benasciutti, Harvested power and sensitivity analysis of vibrating shoe-mounted
EP

373 piezoelectric cantilevers, Smart Materials and Structures, 19 (2010).

374 [11] N.S. Shenck, J.A. Paradiso, “Energy scavenging with shoe-mounted piezoelectrics,” IEEE Micro, 21
375 (2001).
C

376 [12] T. Ogawa, R. Sugisawa, Y. Sakurada, H. Aoshima, M. Hikida, H. Akaishi, “Energy harvesting devices
AC

377 utilizing resonance vibration of piezoelectric buzzer,” Jpn. J. Appl. Phys., 52 (2013).

378 [13] M. Renaud, P. Fiorini, R. Van Schaijk, C. Van Hoof, “Harvesting energy from the motion of human
379 limbs: The design and analysis of an impact-based piezoelectric generator,” Smart Mater. Struct., 18
380 (2009).

381 [14] S. J. Hwang, H. J. Jung, J. H. Kim, J. H. Ahn, D. Song, Y. Song, H. L. Lee, S. P.Moon, H. Park, T. H. Sung,
382 “Designing and Manufacturing a Piezoelectric Tile for Harvesting Energy from Footsteps”, Current
383 Applied Physics, vol. 15, no. 15, vol. 6, pp. 669-674, 2015.
 ACCEPTED MANUSCRIPT

384 [15] Adnan M. Elhalwagy, Mahmoud Yousef M. Ghoneem, Mohamed Elhadidi. “Feasibility Study for
385 Using Piezoelectric Energy Harvesting Floor in Buildings’ Interior Spaces”, International Conference –
386 Alternative and Renewable Energy Quest, AREQ 2017, 1-3 February 2017, Spain.

387 [16] Xiaofeng Li, Vladimir Strezov. “Modelling piezoelectric energy harvesting potential in an educational
388 building”, Energy Conversion and Management 85 (2014) 435–442.

PT
389 [17] J. Dicken, P. D. Mitcheson, I. Stoianov and E. M. Yeatman, "Power-Extraction Circuits for
390 Piezoelectric Energy Harvesters in Miniature and Low-Power Applications," IEEE Transactions on Power
391 Electronics, vol. 27, no. 11, pp. 4514-4529, Nov. 2012.

RI
392 [18] Y C Shu, I C Lien, “Analysis of power output for piezoelectric energy harvesting systems” Smart
393 Mater. Struct. 15 (2006) 1499–1512.

SC
394 [19] G. D. Szarka, B. H. Stark and S. G. Burrow, "Review of power conditioning for kinetic energy
395 harvesting systems," IEEE Transactions on Power Electronics, vol. 27, no. 2, pp. 803-815, Feb. 2012.

396 [20] http://www.veranu.eu/it/

U
397 [21] https://www.murata.com/
AN
398
M
D
TE
C EP
AC