Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Dynamic Analysis of a New Piezoelectric Flextensional

Actuator Using the J1- J4 Optical Interferometric Method


Luiz A. P. Marçala, José V. F. Leãob, Gilder Naderc, Emílio C. N. Silvad, Ricardo
T. Higutie, and Cláudio Kitanof
a,b,e,f
Department of Electrical Engineering, São Paulo State University - UNESP
P.O. BOX 31 - 15385-000, Ilha Solteira, S.P., Brazil. Phone: +55 (18) 3743-1150
c,d
Department of Mechatronics and Mechanical Systems Engineering - Escola Politécnica
da Universidade de São Paulo, P.O. BOX 31 - 05508-900 – São Paulo, S.P., Brazil.
a,b,e,f
kitano@dee.feis.unesp.br; c,decnsilva@usp.br

Abstract Piezoelectric actuators are widely used in positioning systems which demand
high resolution such as scanning microscopy, fast mirror scanners, vibration cancellation,
cell manipulation, etc. In this work a piezoelectric flextensional actuator (PFA), designed
with the topology optimization method, is experimentally characterized by the
measurement of its nanometric displacements using a Michelson interferometer. Because
this detection process is non-linear, adequate techniques must be applied to obtain a linear
relationship between an output electrical signal and the induced optical phase shift.
Ideally, the bias phase shift in the interferometer should remain constant, but in practice it
suffers from fading. The J1-J4 spectral analysis method provides a linear and direct
measurement of dynamic phase shift in a no-feedback and no-phase bias optical
homodyne interferometer. PFA application such as micromanipulation in biotechnology
demands fast and precise movements. So, in order to operate with arbitrary control signals
the PFA must have frequency bandwidth of several kHz. However as the natural
frequencies of the PFA are low, unwanted dynamics of the structure are often a problem,
especially for scanning motion, but also if trajectories have to be followed with high
velocities, because of the tracking error phenomenon. So the PFA must be designed in
such a manner that the first mechanical resonance occurs far beyond this band. Thus it is
important to know all the PFA resonance frequencies. In this work the linearity and
frequency response of the PFA are evaluated up to 50 kHz using optical interferometry
and the J1-J4 method.

1. INTRODUCTION

Piezoelectric transducers are widely used in positioning systems that demand nanometer
resolution. In addition they can present large force generation, sub-millisecond response, no
magnetic fields, extremely low steady state power consumption, etc [1]. One kind of this
transducer is the piezoelectric flextensional actuator (PFA), which consists of three parts: the
metal caps or shell, the active piezoelectric element, and an epoxy adhesive. The basic
schematic of the flextensional transducer is shown in Fig.1. The operational principle of the
PFA is that the caps convert and “amplify” the small radial displacement and vibration
velocity of the piezoceramic into a much larger axial displacement and vibration velocity
normal to the surface of the caps. Two examples of these devices are given by the moonie
and cymbal piezoactuators [2]. Novel models of optimised PFA’s have been designed by
using topology optimisation technique [3], [4].

Figure 1: Schematic of the piezoelectric flextensional actuator.

The problem of designing a PFA is posed as the design of a flexible structure coupled to the
piezoceramic that maximizes the output displacement and generative force in some specified
direction. Recently, successful prototypes (similar to that in Fig.1) designed by topology
optimisation of the PFA’s were manufactured and analysed to characterize their behaviour
[5], [6]. However in those cases the aluminum flexible structures have a closed shape endcap.
In the present work the endcap edges are opened (like shown in Fig.1) and the piezoceramics
is directly bonded to the flexible structure.

Applications in bio-manipulation (such as pro-nuclear injection, intra-cytoplasmic sperm


injection, and nuclear transfer) demand fast and smooth pipette movements to penetrate cell
membranes, with no perceptible lag or overshoot between signal control and pipette, the
same performance being expected in ultra-fast solution switching positioning systems [7]. So
the fast response time is an important criterion for the PFA’s and it can be defined as the time
to achieve the quick and precise response of the actuator without overshoot and ringing.
Currently, commercial positioning systems provide a 100 µm displacement in 1 ms for
working with cell manipulation. However the mechanical resonance of the system limits the
practical actuating range. The actuator should be used in the linear range of their spectrum
and thus it is important to establish the PFA frequency response in order to determine the
resonance frequencies, for which the device experiences a maximum in displacement.

This could be done by using an electrical impedance meter to measure the magnitude and
phase of the piezoactuator impedance. However the authors have noticed that peaks in the
lower range of impedance (or admittance) spectra (below 20 kHz) exhibit small magnitudes
and could not reveal with precision all the resonance occurrences [8]. In this portion of the
spectra the optical interferometry is more adequate to determine these resonances, by direct
measurement of the microscopic movement with the necessary accuracy.

In optical interferometry the displacement/vibration information modulates the phase of an


optical carrier. Owing to the high value of the light velocity and the low value of the
corresponding wavelength, the interferometer sensitivity to optical phase modulation is
extremely high, enabling measurements in the nanometer scale. Next, by using quadratic law
photo-detectors this phase modulation can be linearly converted from the optical domain to a
low frequency phase modulation in the electrical domain, where some adequate type of
analog/digital demodulation processing can be used [9].

Nevertheless, as the interferometric technique is sensitive to very weak stimuli, in practice it


suffers from fading: small changes in ambient temperature and pressure induce random phase
drifts that add unpredictably to the true signal’s phase shift. So the detected signal may
fluctuate randomly in a wide range and during brief periods of time due to the perturbation
agents. The random drifts may be tracked and compensated for to maintain the interferometer
operating in phase quadrature regime. This is commonly done by electronic feedback of the
photo-detector output to some kind of auxiliary optical phase modulator in one arm of the
interferometer. However, in several applications there is a need for a simple and reliable
method of phase demodulation. The J1-J4 method provides a linear, self-consistent and direct
readout of the dynamic phase shift in a homodyne interferometer, irrespective of random
phase drifts due to ambient temperature and pressure fluctuations, source instabilities, and
changes in visibility [10]. According to this method, the photo-detector output frequency
spectrum is obtained (determined from a fast Fourier transform) and the voltage amplitudes
at the fundamental frequency and its next three harmonics are used to calculate the dynamic
phase modulation index. This phase measurement technique is applicable to optical
interferometry in general for the measurement of dynamic sinusoidal displacements.

In the present paper the new PFA designed by the topology optimization method in previous
works [3], [4] is experimentally characterized by the measurement of nanometric
displacements using a homodyne Michelson interferometer based on the J1-J4 spectral
analysis method. In the Section 2 the PFA assembly is described; in Section 3 the J1-J4
method is introduced and in Section 4 the experimental results are listed, emphasizing the
tracking error phenomenon. Linearity and frequency response of the PFA displacement are
evaluated. Conclusions are presented in Section 5.

2. FLEXTENSIONAL PIEZOELECTRIC ACTUATOR ASSEMBLY

The PFA used in this work was modelled by using finite element method (FEM) through
software ANSYS [3]-[6]. A two-dimensional FEM model was built based on plane stress
assumption due to the rectangular shape of the piezoceramic and extrusion-like geometry of
this actuator prototype. The topology optimisation method applied to design this actuator is
based on the homogenisation design method. Essentially, the method consists of finding the
optimal material distribution in a perforated structure with infinite micro-scale voids. The
piezoactuator prototype is shown in Fig. 2.

The piezoceramic was poled in the z-direction (direction 3) and electrode surfaces are normal
to this direction. The designed actuator is composed of an aluminum flexible structure
bonded to a PZT5A (American Piezoceramics, 14 mm x 30 mm x 23 mm in directions 1, 2
and 3, respectively) piezoceramic by using a thin film of epoxy resin. The aluminum flexible
structure was manufactured by using wire EDM (Electrical Discharge Machining). The PFA
is fixed to a holder (Fig. 2 (c)) by three points, perpendicular to the displacement to be
measured. Thus the actuator is free to vibrate in directions 1 and 3. Both the d31 and d33
coefficients of the piezoelectric ceramic contribute to the axial displacement of the PFA.

(a) (b) (c)


Figure 2: Prototype of manufactured flextensional piezoactuator. a) Lateral view b) Top view. c) Holder.

3. MICHELSON INTERFEROMETER AND J1-J4 METHOD

The Michelson interferometer setup, used to measure the PFA vibration amplitudes, is shown
in Fig. 3 (a). A 50/50 % beam splitter divides a 5 mW Helium-Neon laser beam (λ = 0.6328
µm) in two equal components: a reference beam and a sensor beam, which are directed to
two different mirrors. The reflected sensor beam re-enters the beam splitter where it is
directed to a photo-detector (a PIN photodiode). The reflected reference beam re-enters the
beam splitter and is also focused onto the photo-detector. The basic principle in the two-
beams Michelson interferometer is that an applied stimulus in the sensor arm induces a phase
shift relative to the reference arm. In this work the stimulus is the displacement/vibration of
the actuator and causes a φ(t) phase modulation. At the photo-detector, both beams are
recombined, and the light intensity is time averaged, resulting in [9]:

I0
I (t ) = {1 + V . cos[φ 0 + φ (t )]} (1)
2

where I0 is the laser optical intensity, V is the visibility, φ0 is the global static phase difference
between the arms and φ(t) is related to the displacement to be measured. In Fig.3 (b) it is
illustrated the normalized curve I/I0 as a function of total phase shift Φ = φ0 + φ ( t ) , when
the visibility is equal to unity.

As can be seen from (1) the interferometer output is a non-linear, sinusoidal function of the
optical phase shift induced by the time-dependent signal. Hence, a linear measurement of the
signal’s phase from the photo-detector output is not possible directly, unless for the case of
small phase modulation index, as sketched in Fig. 3 (b) for a triangular drive signal operating
in quadrature regime (φ0 = π/2 rad). In this case there is a direct conversion of the optical
phase modulation to an intensity amplitude modulation. For any other forms of operation the
signal output is a distorted version of the input drive signal. Besides, the phase φ0 ideally
should remain constant, but in practice it does not happen due to fading. A direct and self-
consistent readout of the dynamic phase shift φ(t) in a homodyne interferometer, without
feedback, source stabilization or visibility control, may be obtained by using the J1-J4
technique [10], [11].
(a)
(b)
Figure 3: Homodyne interferometry used to characterize the flextensional piezoactuator. (a) Michelson
interferometer setup. (b) Transfer function and small modulation index operation.

When the PFA is excited with a sinusoidal signal with angular frequency ωs, a phase shift in
the form φ ( t ) = x sin( ωst ) is produced, where x [rad] is the phase modulation index. The
instantaneous output voltage in the photo-detector v(t), which is proportional to the optical
intensity (1), can also be written as:

I0 ⎧ ⎡ ∞
⎤ ⎡ ∞ ⎤⎫
v(t ) = R ⎨1 + V . cos φ J
0⎢ 0

( x ) + 2∑ J 2n ( x ) cos( 2 nω t
s ⎥)

− V .sinφ0 ⎢2∑ J 2 n −1 ( x) sin[(2n − 1)ωs t ]⎥ ⎬
⎣ n =1 ⎦⎭
2 ⎩ n =1

(2)

where the terms Jn's are Bessel functions of the first kind and order n, and R is the voltage
photo-detector responsivity. The J1-J4 technique employs a Bessel recurrence relation to
measure x directly [10]:
V2V3
x 2 = 24 (3)
(V1 + V3 ). (V2 + V4 )

where Vi is associated to the photo voltage magnitude, expressed by (2), at the ith frequency.
One can see from (3) that factors cosφ0 (or sinφ0), R, I0 and V that multiply each Jn in (2) are
cancelled. This justifies why phase measurement is unaffected by random phase drifts,
fluctuation in source intensity, and variations in visibility. The piezoactuator vibration
amplitude or surface displacement at the measurement point, ∆L, is given by:


∆L = λ (4)
x

where λ is the laser wavelength in vacuum. At low values of modulation index (x), the Bessel
terms J3 and J4 become small compared with J1 and J2, and so, the values of V3 and V4
components would be inaccurate when they are close to noise floor. So the use of the J1-J4
technique requires the generation of phase modulation indexes (x) greater than 0.1 rad for a
typical interferometric setup [11].
4. EXPERIMENTAL RESULTS

A low cost version of the interferometric setup sketched in Fig. 3 was built and tested. Only
the flexural vibration mode of the metal shell could be detected. Due to the difficulties to
polish the irregular piezoactuator surface to an optical degree, a 200 µm thickness mirror,
obtained by aluminum vaporization over a glass plate, was bonded to the actuator surface
with epoxy resin (see Fig. 2). The PIN photodiode output signal is amplified, digitized by an
oscilloscope (Tektronix TDS2022) and transferred to a microcomputer via RS232 port. The
data is then processed in a Matlab program.

4.1 Dynamic Performance of the Piezoelectric Flextensional Actuator


In positioning systems applications it is desirable that infinitesimally small changes in
operating drive voltage of the piezoactuator is converted to smooth movements. In the
following example, the flextensional piezoactuator is driven by the triangular voltage signal
(4 V amplitude and 100 Hz frequency) shown in Fig. 4 (a). The a.c. output signal measured
with the interferometer operating under small phase modulation index and quadrature
conditions is shown in Fig.4 (b). This output signal presents some noise level, which could be
electronically suppressed. The displacements involved are in nanometer scale.

(a) (b)
Figure 4: Response of the flextensional actuator to a ± 4 V, 100 Hz triangular drive signal.

Fast response is one of the desirable features of piezoactuators. A rapid drive voltage change
should results in a rapid position change. The flextensional actuator is interferometrically
tested at 815, 1800, and 2800 Hz frequencies of triangular drive voltages. In Fig.5 (a-c) it is
illustrated their respective photo-detected voltages, which are proportional to the actuator
surface displacements. Apparently, it seems that the signals are noisy, however, a detailed
analysis reveals that the behavior is due to the tracking error phenomenon. As the resonance
frequencies of the piezoactuator are not sufficiently high, unwanted dynamics of the stage are
often a problem, especially for scanning motion with high frequency, but also if trajectories
have to be followed with high velocities. In Fig.5 (d) it is illustrated a simulated signal
corresponding to a 2 kHz triangular waveform superposed to a 10% amplitude and 23 kHz
sinusoidal waveform. The result is very similar to Fig. 5(c). Under this mode of operation
there is no more direct relationship between the control voltage and the PFA displacement.

As it is well known, to a τ =2 ms seesaw movement associated to an electrical triangular


waveform corresponds a 2/τ = 1 kHz spectral bandwidth, which is the range where its more
significant spectral components are present. So, in principle, if the resonance frequencies of
the piezoactuator occur far beyond 1 kHz there would be no problems with tracking errors in
this kind of operation. However, this is not exactly true as revealed by the results obtained
with the present interferometer. In Fig. 6 (a) it is illustrated a 5 V and 110 Hz square
waveform that drives the PFA, and in Fig. 6 (b) the corresponding detected photo-voltage for
the interferometer operation under small phase modulation index. This frequency was
selected because the oscillations in the displacements vanish at the end of the half-cycle.
Carefully inspecting Fig.6 (b) it is observed that there are 211 cycles of oscillation per period
(9.2 ms) of the square waveform signal. Thus the oscillation frequency is close to 23 kHz,
which is a PFA resonance frequency.

(a) (b)

(d)
(c)
Figure 5: Occurrence of tracking error phenomenon for triangular drive voltage. (a) 815 Hz. (b) 1800 Hz. (c)
2800 Hz. (d) Simulated output at 2000 Hz.

A piezoactuator can reach its nominal displacement in approximately one period of the
resonance frequency, albeit with significant overshoot [4]. So an actuator with a 23 kHz
resonant frequency should reach its nominal displacement within 43 µs, although, as reveals
Fig.6 (b), the displacement stabilization time is 100 times larger (approximately 4.6 ms)
because of the transient response.

(a) (b)
Figure 6: PFA dynamics. (a) Square waveform-driving voltage. (b) PFA’s photo-detected response.

Next the PFA is driven by 700 and 1000 Hz square waveforms (whose periods are smaller
than 4.6 ms), as illustrated in Fig.7 (a-b) respectively. From the step response of the system
the natural frequencies could be obtained, however, the square waveform is more illustrative
permitting the comparison between oscillating response directly to its period. As can be
observed, in all cases the spectral components of the signals located far beyond their –3 dB-
spectral bandwidth coupled energy to the 23 kHz PFA resonance frequency. In fact, it was
counted 33 and 23 oscillation cycles per period of the respective square waveform in Fig 7
(a) and (b), respectively, which result in 23 kHz frequency oscillations. The spectral plots
corresponding to the instantaneous output signals certify the occurrence of a resonance in 23
kHz. Nevertheless, Fig. 7 (a) reveals that smaller resonances near 15, 32 and 48 kHz are also
present.

To eliminate vibration, methods have been proposed which are commonly called “input
shaping” or “command shaping”, which consist in “breaking” the input step into a staircase-
shaped signals as described in [12]. However, this problem will not be treated in this work.

(a)

(b)
Figure 7: Interferometric response (small phase modulation index) to square waveform voltage: instantaneous
waveform, input and output spectra. (a) At 700 Hz. (b) At 1000 Hz.

4.2 Linearity and Frequency Response Curves


The graphical results presented in the previous sections were obtained under small phase
modulation index conditions, which demand careful adjust of φ0 to π/2 rad. Measurements of
linearity and frequency response require higher accuracy and will be accomplished by using
the J1-J4 method. As an example, Fig. 8 (a) shows a typical output interferometric waveform
generated in such case, corresponding to the light intensity detected when an exciting
sinusoidal voltage (amplitude of 9.2 V and frequency of 32 kHz) is applied to the
piezoactuator. The corresponding Fourier spectrum (in dB) is show in Fig. 8 (b), where V1,
V2, V3 and V4 are the voltage magnitudes of the fundamental and the next three harmonics.

Using (3) and (4) the modulation index and the displacement amplitude can be obtained,
respectively. At the frequencies of 15 kHz and 23 kHz with varying amplitudes applied to the
piezoactuator, the input-output relations are shown in Fig. 9 (a). These frequencies
correspond to two PFA resonances. The curves show a linear response in the displacement
range considered. The slope of the curves at 15 kHz and 23 kHz are equal to 5.49 nm/V and
86.35 nm/V, respectively. To provide accurate measurements (above noise floor), the
minimum detected displacement was limited down to 80 nm. The spectrum analysis in this
work was able to display only the magnitudes of the spectral components V1 … V4, which
have positive values. Thereby the maximum displacement was limited up to the value where
x reaches 3.83 rad, when J1 becomes negative, even though modifications in the J1-J4 method
have been published, improving the dynamic range up to 6.3 rad [13].

(a) (b)
Figure 8: Typical output interferometric signal observed when the PFA is driven with a sinusoidal voltage. a)
Voltage waveform corresponding to the intensity of detected light. b) The corresponding spectrum.

The frequency response of the actuator, obtained between 1 kHz and 50 kHz, is shown in Fig.
9 (b), in terms of the amplitude displacement normalized by the input voltage. Resonance
frequencies are found at 4.6, 15, 23, 32, and 48.6 kHz, which confirms the previous results.
From d.c. to 4.5 kHz the PFA frequency response is approximately flat.

(a) (b)
Figure 9: PFA interferometric characterization. (a) Displacement amplitudes as a function of input voltage. The
least-squares fit to these plots is also shown. (b) Frequency response.

5. CONCLUSIONS

A new PFA, designed with the topology optimization method, was implemented and
characterized. Its dynamic performance was investigated emphasizing the tracking error
problem. Linearity and frequency response of the displacement amplitudes were evaluated
using a homodyne Michelson-type interferometer by applying the J1-J4 spectral method,
which is not influenced by random drifts. All the resonance frequencies up to 50 kHz were
determined. The gain was fairly smooth from d.c. up to 4.5 kHz approximately.

ACKNOWLEDGEMENTS

Support for this project from the sponsor agency CAPES is gratefully acknowledged.

REFERENCES
[1] K. Uchino, “Recent Trends of Piezoelectric Actuator Developments,” In: Proceedings of the 1999
International Symposium on Micromechatronics and Human Science,” pp. 3-9, 1999.
[2] A. Dogan, K. Uchino, R.E. Newnham, “Composite Piezoelectric Transducer with Truncated Conical
Endcaps “Cymbals”,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol.44,
no.3, pp.597-605, May 1997.
[3] E.C.N. Silva, N. Kikuchi, “Design of Piezoelectric Transducers Using Topology Optimization,” Smart
Mater. Struct., vol. 8, pp.350-364, 1999.
[4] E.C.N. Silva, N. Kikuchi, “Topology Optimization Design of Flextensional Actuators,” IEEE
Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol.47, no.3, pp.597-605, May 2000.
[5] E. C. N. Silva, G. Nader, A.B. Shirahige, J.C. Adamowski, “Characterization of Novel Flextensional
Actuators Designed by Using Topology Optimization Method,” Journal of Intelligent Materials and
Structures, vol. 14, pp.297-308, 2003.
[6] G. Nader, E.C.N. Silva, J.C. Adamowski, “Effective Damping Value of Piezoelectric Transducers
Determined by Experimental Techniques and Numerical Analysis,” In: Proceedings of the 17th
International Congress of Mechanical Engineering – COBEM, November 10-14, São Paulo, SP, Brazil,
2003.
[7] Burleigh Instruments Inc., “PZT actuator/PZS-100HS and PZS-200 microstage,” Application notes,
http://www. burleighlifescience. com Web Site, 2003.
[8] L.A.P. Marçal, J.V.F. Leão, G. Nader, E.C.N. Silva, R.T. Higuti, C. Kitano, “Analysis of Linearity and
Frequency Response of a New Piezoelectric Actuator Using a Homodyne Interferometer and the J1-J4
Method,” to be presented at the 2005 IEEE Instrumentation and Measurement Technology Conference,
Ottawa, Ontario, Canada, 2005.
[9] C.B. Scruby, L.E. Drain, “Laser Ultrasonics, Techniques and Applications,” Adam Hilger, New York,
USA, 1990.
[10] V.S. Sudarshanam, K. Srinivasan, “Linear Readout of Dynamic Phase Change in a Fiber-Optic
Homodyne Interferometer,” Optics Letters, vol.14, no.2, pp.140-142, Jan. 1989.
[11] V.S. Sudarshanam, B. Desu, “Measurement of Dynamic Polarization Modulation Depth Utilizing the J1-
J4 Method of Spectrum Analysis,” Rev. Sci. Instrum. 65 (7), pp. 2337-2343, Jul. 1994.
[12] A. Bergender, W. Driesen, T. Varidel, and J.M. Breguet, “Development of Miniature Manipulators for
Applications in Biology and Nanotechnologies,” In: Proceedings of Workshop "Microrobotics for
Biomanipulation", IEEE IROS 2003, Las Vegas, USA, pp. 11-35, October 27-31, 2003.
[13] V.S. Sudarshanam, R.O. Claus, “Generic J1...J4 Method of Optical Phase Detection-Accuracy and Range
Enhancement,” Journal of Modern Optics, vol.40, no. 3, pp.483-492, 1993.

You might also like