8 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 17, NO.

1, JANUARY 2002

Fundamental Limits on Energy Transfer and Circuit

Considerations for Piezoelectric Transformers
Anita M. Flynn and Seth R. Sanders, Member, IEEE

Abstract—This work investigates fundamental limits on 400–500 V to run at a 1–5 W power level and have an input
electromechanical energy conversion capacity of piezoelectric impedance on the order of hundreds of kilohms, a good match
transformers by considering a work cycle analysis. Fundamental for the piezoelectric transformer.
limitations in a lossless piezoelectric transformer are imposed by
maximum electric field strength, maximum surface charge density, While piezoelectric transformers were originally developed
maximum stress, and maximum strain. For the lossless case, our in the United States in the 1950s [1], [2], they were not com-
analysis indicates that the mechanical stress limit is the effective mercially pursued to a large extent due to poor materials relia-
constraint in typical PZT materials. For a specific PZT-5H sample bility and competition from magnetic flyback transformers for
considered, a mechanical stress-limited work cycle indicates that cathode ray tubes. Recently, a number of researchers, mainly
this material can handle 330 W/cm3 at 100 kHz.
A second direction this work has taken has been an investiga- in Japan, have begun new efforts at producing compact, effi-
tion into a soft-switching drive and control circuit, that does not cient, bulk-ceramic piezoelectric transformers for portable ap-
require any magnetic components. The theory of operation of soft- plications, most notably for laptop flat panel displays. These ef-
switching resonant drive circuitry is discussed, and experimental forts have yielded reported efficiencies from 82% [3] to as high
results on a soft-switching inverter incorporating no magnetic com- as 92% [4].
ponents are reported.
In this paper, we discuss fundamental limits of energy density
Index Terms—Piezoelectric, soft-switching, transformer, work and power throughput of an ideally lossless piezoelectric ele-
cycle. ment based on materials considerations and a work cycle anal-
ysis. The work cycle considered is shown to be consistent with a
I. INTRODUCTION square-wave voltage drive. However, such a hard-switched drive
scheme in an inverter circuit leads to considerable switching

P OWER electronic circuits have conventionally been based

on magnetic technology and until recently have not been
part of the tide of miniaturization and integration advances from
losses as the static input capacitance of a piezoelectric trans-
former can be large. We subsequently describe a strategy for
achieving soft-switching, and report on an inverter circuit that
which signal-processing integrated circuits have benefited. In
we have built which demonstrates soft-switching and high effi-
many power electronic systems today, the magnetic components
ciency without the use of magnetic components.
remain as the bulkiest and costliest components. In an effort to
achieve miniaturization, we are investigating alternative tech-
II. ELECTROMECHANICAL POWER TRANSFORMATION—
nologies to magnetics for provision of reactance, galvanic isola-
FUNDAMENTAL LIMITS
tion and voltage/current scaling (transformer function). Specif-
ically, we are investigating acoustic coupling in place of induc- A piezoelectric transformer works by using the direct and
tive coupling to create electro-mechanical transformers. converse piezoelectric effects to acoustically transform power
Acoustic coupling can be achieved through piezoelectric from one voltage and current level to another. That is, power
or magnetostrictive means. A piezoelectric transformer offers is converted electromechanically through a vibrating piezoelec-
the potential advantages of a thinner profile than a magnetic tric structure rather than electromagnetically as in an inductive
transformer, and a simpler manufacturing process due to the transformer. Electromechanical coupling, in one dimension, is
lack of windings. Application areas range from battery-pow- described by the piezoelectric constitutive relations [6]
ered consumer electronics and cellular telephones to wearable
computers and head-mounted displays. The driving application (1)
at present for a piezoelectric transformer is the step-up trans- (2)
formation required to power a cold cathode fluorescent lamp
in a flat panel display from a laptop computer battery. These The convention here is that the poling direction is taken as
cold cathode fluorescent lamps require 1000 V to start and the 3-direction in a 1-2-3 labeled orthogonal coordinate system.
Variable is the electric field strength applied in the direc-
tion the material was poled, while is the resulting surface
Manuscript received January 17, 2000; revised May 2, 2001. Recommended charge density. Variables and are the reduced matrix no-
by Associate Editor K. Ngo.
A. M. Flynn is with the MicroPropulsion Corporation, Berkeley, CA 94704 tation representations for the stress tensor component, , and
USA (e-mail: aflynn@micropropulsion.com). strain tensor component, , respectively. Parameter is the
S. R. Sanders is with the Electrical Engineering and Computer Science compliance taken at constant electric field, while is the per-
Department, University of California, Berkeley, CA 94720 USA (e-mail:
sanders@eecs.berkeley.edu). mittivity taken at constant stress, and is the piezoelectric
Publisher Item Identifier S 0885-8993(02)02167-1. coupling coefficient.
0885–8993/02$17.00 © 2002 IEEE
FLYNN AND SANDERS: FUNDAMENTAL LIMITS ON ENERGY TRANSFER AND CIRCUIT CONSIDERATIONS 9

(a)
(a) (b)

(b)
Fig. 1. Both transformers shown here are long thin bars of piezoelectric (c)
material driven in a longitudinal mode resonance (displacements in the
3-direction) [1]. (a) Two segments are poled as shown and electroded faces
perpendicular to the poling directions provide contacts for the input and output
voltages. (b) The input segment is poled transversely to the longitudinal mode
displacements, utilizing the d effect (d = d ). The generator segment
uses d coupling, a larger effect than d , which assists in creating larger
voltage amplification.

Rosen first described piezoelectric transformers in [1], [2], (d)

and two of the simplest structures are shown in Fig. 1. In the Fig. 2. Work cycle viewed in both the electrical and mechanical domains. As
transformers shown in the figure, the converse piezoelectric ef- the work cycle is traversed, the area inside each parallelogram is the energy
fect, in which an applied electric field produces a resulting strain transformed from electrical to mechanical energy or vice versa. (a) Electrical
domain. (b) Mechanical domain. (c) E versus time. Points 1, 2, 3, and 4
in a body, is used to first convert an oscillating electric field ap- represent the square-wave drive corresponding to the respective points in the
plied to the left half of each bar into an extensional vibrational work cycle of (a) and (b) when the piezoelectric structure is cycled at its
mode of the entire bar. If driven at resonance, standing-wave resonant frequency. (d) S versus time. The strain is at its maximum positive
value during the instantaneous time that the electric field moves from its
distributions of large amplitudes of stress and strain result. The maximum positive value at position 1 to its most negative value at position 2.
resonantly amplified strain in the right half of each bar is then Conversely, the strain is at its most negative value when E transitions from its
converted to a voltage across the output terminals by the direct most negative value at state 3 to its maximum positive value at state 4.
piezoelectric effect. Depending upon the geometry and mate-
rials parameters, voltage amplifications of various magnitudes then to some negative value (point 2). At this point, the field
can be obtained, with associated step-downs in current levels. is held constant and the mechanical block is removed, where-
upon the structure moves from a state of positive strain to a state
A. Power Density Calculation of negative strain (point 3), and the system has performed me-
In designing piezoelectric transformers, it is useful to first chanical work in moving from point 2 to point 3. The element
investigate the fundamental limits of power density, i.e., the is then mechanically blocked again in this new strain state and
amount of power able to be transferred through the device per the electric field is increased (point 4). The electric field is then
unit volume. During each cycle of electrical input, some amount held constant and the mechanical block is removed, moving the
of work is transformed in the left half of the bar from elec- system back to its original state of positive strain (point 1). The
trical energy to mechanical energy, and then similarly, in the area enclosed by the work cycle represents the transformed en-
right half of the bar, that same mechanical energy is transformed ergy density. Specifically, the area enclosed by the parallelo-
back to electrical energy. The maximum power density could gram in Fig. 2(a) represents the supplied electrical energy (den-
be limited by a number of mechanisms such as electric break- sity), while the area enclosed by the parallelogram in Fig. 2(b)
down strength, maximum surface charge density (i.e., satura- represents the mechanical work (density). Note that these two
tion), maximum stress, maximum strain, or depoling due to elec- areas are equal for this ideal lossless work cycle. The cycle is
tric fields. To address the power density issue, we consider an repeated periodically and the power throughput density is equal
infinitesimal cellular element situated at a point in one half of a to the energy density multiplied by the frequency of operation.
piezoelectric transformer, such as a point in the left half of the We note that the trajectory of the work cycle of Fig. 2(a) and
device in Fig. 1(a). (b) is actually consistent with a square-wave voltage drive, as
We determine the power density by examining a work cycle. readily achieved with a square-wave inverter, or some variant
Let us start by assuming a lossless system that is in state 1 of thereof. Fig. 2(c) illustrates such a square wave drive. In this
its work cycle (point 1 of Fig. 2). The structure is then blocked case, operation could be near resonance, where the voltage drive
mechanically as the electric field is reduced first to and is switched from maximum (point 1 of Fig. 2) to minimum
10 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 17, NO. 1, JANUARY 2002

Fig. 3. Maximum energy density for a stress-limited material can be calculated

by maximizing over S the area of the parallelogram enclosed by the work cycle.
Fig. 4. Work cycle for stress-limited PZT-5H yields an energy density of
(point 2), when the strain, and corresponding displacement, is 3920 J/m .
at its positive extreme. Conceptually, the voltage transition oc-
curs instantaneously, and no change in the strain state occurs Maximizing the area with respect to yields the optimum
during this transition. Then the voltage drive is held constant at strain,
its minimum while the inertia of the resonant motion carries the
(4)
transition from point 2 to point 3, thus changing from a positive
strain state to a negative strain state, as is illustrated in Fig. 2(d). This value of must be checked to ensure that it is less than
The respective transitions from point 3 to point 4 and from point . If so, this strain yields a maximum energy density of
4 to point 1 of Fig. 2 occur analogously to the transitions from
point 1 to point 2 and from point 2 to point 3, consistent with Joules/m (5)
the symmetry of the trajectory.
Maximum power throughput depends on the maximum Maximum power density corresponds to maximum energy den-
values of electric field, charge density, stress and strain which sity per unit time, yielding
can be achieved with a given material. We consider the lossless
case here and thus the computed power densities do not reflect Watts/m (6)
thermally limited power transmission. In practice, steady state
power density may also be limited by temperature rise due to where is the frequency at which the work cycle is traversed.
internal losses [10]. For the lossless case, the maximum-area Fig. 4 illustrates the maximum area stress-limited work
work cycle is bounded by one or more of the material’s limits, cycle for the example PZT-5H material with the characteristics
or . To explore the largest feasible shown in Table I. Note that maximum values for electric field,
work cycle, we pick a corner corresponding to maximum values charge density, and strain are not exceeded. The calculated
in either domain, such as and map this state energy density from (5) is 3920 J/m . The resulting maximum
into the - domain with the one-dimensional piezoelectric power density from (6), in the lossless case, at 100 kHz is thus
constitutive relations (1), (2). 330 W/cm . For a ceramic piece 0.5 mm thick, this translates
For an example material such as the PZT-5H ceramic [5], to 16.5 W/cm . As previously mentioned, this calculation is
specified in Table I, the point maps outside the not based on a thermal limit, and hence may not provide a
bounding box. The bounding box is the region in steady state maximum power density. Work addressing thermal
the stress-strain plane that will not cause fracture. The mate- limitations of piezoelectric transformers can be found in [10].
rial is thus mechanical-field-strength limited, i.e., it is either
maximum stress limited or maximum strain limited. Assume III. MAGNETICS-FREE SOFT-SWITCHED DRIVE
the material is stress limited. From known values of and
slope , a value of strain, , can be found which maximizes The preceding analysis was based on a square-wave drive,
the enclosed area, where the enclosed area is the energy trans- applied exactly at resonance. As reported in the literature
formed per unit volume. Fig. 3 shows the construction of the [7]–[9], the impedance at the input terminals of the piezoelec-
maximum-area work cycle for a stress-limited material. For a tric transformer is actually inductive at a frequency just above
maximum stress of , and compliance at constant electric resonance, as is the case with a conventional series resonant
field, , the area of the parallelogram (energy density) is - tank circuit. Fig. 5(a) shows a simplified equivalent circuit
model of a piezoelectric transformer, corresponding to a model
Joules/m (3) that accounts only for the principle resonant mode. Here,
FLYNN AND SANDERS: FUNDAMENTAL LIMITS ON ENERGY TRANSFER AND CIRCUIT CONSIDERATIONS 11

TABLE I
PZT-5H MATERIAL CONSTANTS

the principle resonant mode, and the mass corresponds to a

lumped parameter value for the same resonant mode. Ideal
transformers in Fig. 5(a) indicate coupling between electrical
and mechanical domains.
The resonant characteristic of the piezoelectric transformer
can be used to overcome switching losses associated with
(a) switching or commutating the static input capacitance if the
transformer is driven above resonance. That is, by operating the
inverter at a frequency slightly above resonance, and by pro-
viding for an appropriate deadtime on each inverter switching
transition, it is possible to obtain resonant soft-switching. This
constitutes a variation of the work cycle introduced in the
preceding section and is illustrated in Fig. 5(b), and discussed
below.
If a deadtime DT is provided after turning off and be-
fore turning on , the transformer input capacitance, in com-
bination with the mosfet drain-source capacitances, is free to be
charged up through the resonant mechanical motion of the trans-
former device. Specifically, during this deadtime, the variable
in Fig. 5(a), representing velocity of the lumped parameter
mass, is negative. The continuing mechanical motion causes a
corresponding change in strain state and displacement , and a
subsequent change in the charge-voltage state at the electrical
terminals. The change in strain and displacement can easily
be large enough to supply all the charge needed to drive the
complete voltage transition from negative state to positive state
during the deadtime, even considering the charge required by the
transformer input capacitance and also the drain-source capac-
itances of the power mosfets. This resonant transition is analo-
(b) gous to that which is used to achieve zero-voltage switching in
a half-bridge series resonant converter [13]. A complementary
Fig. 5. (a) Inverter circuit for driving the primary of a piezoelectric trans- zero-voltage switching transition is effected during the deadtime
former. The equivalent circuit model for the piezoelectric transformer is
enclosed in the dashed lines. (b) Drive waveforms consistent with zero-voltage DT .
switching. There is a subtle difference from the work cycle of Fig. 2 with
the proposed soft-switching cycle. Specifically, in Fig. 2, the
transition from points 1 to 2 occurs with the piezoelectric slab
the principle resonant mode is the extensional mode of the mechanically blocked, and this transition is effectively driven
piezoelectric device that coincides with the drive frequency. at the electrical port. In the proposed soft-switching cycle, the
Transformers of this type have been used in the fundamental change in state of electric field occurs while the piezoelectric
and second extensional resonant modes [12]. In the model of slab is in motion, and with the electrical terminals open-cir-
Fig. 5(a), the input capacitance corresponds to static electrical cuited. Thus, the soft-switching transition occurs with a con-
capacitance with the transformer mechanically blocked. The stant value of charge density . An analogous complementary
compliance corresponds to a lumped parameter value for transition occurs from points 3 to 4.
12 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 17, NO. 1, JANUARY 2002

(a)

Fig. 6. Primary of the piezoelectric transformer is driven by a MOSFET half-

bridge and the cold cathode fluorescent lamp is the load on the secondary. A
feedback signal is tapped off the transformer secondary via a capacitive voltage
divider and fed back to the gate drivers to create a self-oscillating system. The
feedback signal is phase shifted such that a zero-crossing comparator produces
the appropriately phased drive signal for the gate drivers.

IV. EXPERIMENTAL RESULTS

An experimental setup was built to verify the proposed soft-
switching inverter scheme. The purpose was not to verify op-
eration at maximum power density. Fig. 6 is a block diagram
of a circuit realization of this inverter scheme that is self-oscil- (b)
lating. The power train is a half-bridge, always driven at 50% Fig. 7. Equivalent model of the piezoelectric transformer consists of two ideal
duty cycle. The gate drive block provides gate drive for the transformers which model the mechanical-electrical and electrical-mechanical
energy conversion processes respectively. L and C represent
power train, along with appropriately adjusted dead times. The the resonant characteristic of the transformer. During the deadtimes when both
remainder of the circuit is a very simple feedback scheme to MOSFET’s are off, the transformer impedance appears inductive, creating a
cause the inverter to frequency lock just above a resonant fre- current source that pulls or pushes charge off the static input capacitance of
the transformer and also the parasitic capacitances at the MOSFET inverter
quency. This is explained in the sequel. node. Since the MOSFET’s are relieved from supplying the power to do this,
Capacitors and form a capacitive voltage divider switching losses are averted and efficiency is increased relative to a hard-
to allow sampling the large output voltage without overloading switched configuration.
this very high impedance terminal. The phase shifter network,
- adjusts the phase of the output signal, which is then fed tude scales up, the effective gain of the loop is reduced. Thus,
to a comparator to provide the inverter drive signal. The use of from a Nyquist stability point of view, if the amplitude is per-
the equivalent circuit model for the piezoelectric transformer, turbed upward, the gain is reduced, making the system “more
as previously introduced and shown in Fig. 7, is most conve- stable” and thus resulting in an amplitude reduction. An analo-
nient for understanding the operation. In this model, the capac- gous argument applies if the amplitude is perturbed downward.
itor, , represents the compliance, with state corre- A prototype of the circuit of Fig. 6 was constructed and used
sponding to strain. The inductor, , represents the mass, with a Rosen transformer of the type shown in Fig. 1(b). The ob-
with state corresponding to mechanical velocity. From inspec- jective was only to verify operation of the soft-switching drive
tion of the electrical circuit representation and corresponding circuit, and not to exercise the piezo transformer to its maximum
waveforms, it is clear that a steady state oscillation above res- power density. The piezoelectric material was PZT-4, and overall
onance is consistent with the description. Roughly, lags the dimensions were 48 mm 8 mm 2 mm. The second longitu-
fundamental of by more than 90 due to operation above dinal extension mode resonance at 70 kHz yielded best results.
resonance. The signal lags by an amount provided by the This was also noted in [12]. A small cold cathode fluorescent
phase shifter, such that lags by 180 degrees. The power lamp was used as the load. Due to the relatively thick PZT sample,
train (gate drivers and inverter) provides the final 180 inversion a relatively high voltage of about 45 V for the inverter supply
function to effect a net 360 of delay around the loop. was used to obtain a power output of about 500 mW. Thinner
An heuristic argument for stability can be motivated by the transformers can operate efficiently with lower input voltages.
Nyquist criteria and the describing-function analysis method For the circuit implementation of Fig. 6, experimental pri-
[11]. For a steady-state oscillation, a loop gain of unity and mary voltage and charge waveforms are shown in Fig. 8(a). The
phase shift of some integral multiple of is required. The waveform at top is , the primary voltage applied by the in-
phase condition is consistent with the indicated waveforms. The verter, at 20 V/div and 2 s/div. Note the trapezoidal shape,
loop gain condition is achieved courtesy of the limiting behavior indicative of the soft-switching action. The lower waveform in
of the comparator nonlinearity. As the comparator input ampli- Fig. 8(a) is , which is representative of the charge on the pri-
FLYNN AND SANDERS: FUNDAMENTAL LIMITS ON ENERGY TRANSFER AND CIRCUIT CONSIDERATIONS 13

(a) (b)
Fig. 8. (a) The trapezoidal waveform at the top is the voltage applied to the transformer primary, labeled v in Fig. 7. Note that during the deadtime, DT , the
voltage ramps up due to the inertia of the transformer causing a change in the charge-voltage state even when the primary is undriven. The vertical scale is 20 V/div
and the horizontal scale is 2 s/div. The lower trace represents the charge at that node, measured as the voltage (labeled v in Fig. 6) across a 1 F capacitor placed
in series with the primary. The vertical scale of the lower trace is 50 mV/div, or equivalently, 50 nC/div. (b) Here, the two traces of (a) are plotted against each
other with transformer primary charge, v , on the y -axis and primary voltage, v , on the x-axis, thus tracing out the electrical domain work cycle at the primary
of the transformer. In comparing to Fig. 2(a), note that hard-switching was assumed there, whereas soft-switching is achieved here. The change in charge during a
low-to-high voltage transition is approximately 1=3 of a division decrease. That is, charge comes off the transformer primary to charge up the parasitic capacitance
at the inverter node and swing the voltage up.

mary, obtained by sampling across a 1 F capacitor inserted in

series with the primary. Scale for this waveform is 50 mV/div,
or equivalently, 50 nC/div. Fig. 8(b) provides a charge-voltage
phase portrait for these same waveforms, by displaying charge
on the vertical -axis and voltage on the horizontal -axis. Note
the close resemblance to the electrical work cycle of Fig. 2(a).
The main difference is that in Fig. 8(b), as the voltage increases,
the charge actually decreases. The charge decrease is exactly
the charge required to swing the parasitic capacitance associated
with the MOSFET inverter output. By inspection of Fig. 8(b),
it is evident that this charge is about 15 nC, consistent with the
charge needed to swing a parasitic output capacitance of about
330 pF through 45 V. Fig. 9. The trapezoidal waveform, v , is shown again, as in Fig. 8, for
We note that if a hard-switched square-wave is applied at the reference. The smooth sine waveform is 1=400 of the voltage at the transformer
primary, the switching loss is significant. A total capacitance secondary. That is, this waveform is the voltage fed into the feedback circuit
from the capacitive voltage divider attached to the transformer secondary. The
of about 1 nF needs to be switched—about 300 pF associated vertical scale of this feedback signal is 1 V/div. The horizontal axis is again 2
with our inverter and approximately 700 pF associated with the s/div. The third waveform represents the current through the lamp, measured
static input capacitance of the transformer. The resulting across a 1 k
resistor (shown in Fig. 6) placed in series with the lamp. The
vertical scale of the third waveform is 1 V/div, or equivalently, 1 mA/div.
switching loss is then around 140 mW for a 45 V input voltage,
and operation at 70 kHz. This comprises more than a 30% loss! TABLE II
Fig. 9 shows the secondary electrical waveforms along with EFFICIENCY MEASUREMENTS
the primary voltage [as in Fig. 8(a)]. The sinusoid-like wave-
form is , a scaled version of the output voltage obtained
through a 400:1 capacitive divider. This is shown on a 1 V/div
scale, indicating an output voltage of about 1200 V peak. Note
that the phase of this output voltage is related to the primary
drive voltage, as shown in Fig. 5, and previously discussed.
The third waveform represents the lamp current, sampled with
a 1 k resistor. This waveform is shown on a 1 V/div, or equiva-
lently 1 mA/div, scale. Lamp current is thus about 2.4 mA peak.
The phase shift between the lamp current and the transformer
secondary output voltage is due to a small compensation capac-
itor of about 4 pF placed in series with the lamp. replacing the fluorescent lamp with a resistive load in series with
We note that accurate estimation of the output power from the 1 k current sense resistor. Table II shows the measured re-
these waveforms is difficult. Consequently, we measured effi- sults of output power dissipated in this load and the measured
ciency for the system (transformer and converter circuitry) by dc input power. The efficiency was determined to be 83%.
14 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 17, NO. 1, JANUARY 2002

V. CONCLUSION [10] Y. Fuda, K. Kumasaka, M. Katsumo, H. Sato, and Y. Ino, “Piezoelectric

transformer for cold cathode fluorescent lamp inverter,” Jpn. J. Appl.
Piezoelectric transformers offer potentially impressive power Phys., vol. 86, pp. 3050–3052, May 1997.
density. We calculated the field-limited maximum power den- [11] A. Gelb and W. E. V. Velde, Multiple-Input Describing Functions and
Non-Linear System Design. New York: McGraw-Hill, 1968.
sity and found that it was stress-limited. The peak power den- [12] T. Zaitsu, T. Inoue, O. Ohnisi, and A. Iwamoto, “2 MHz power converter
sity for a PZT-5H sample was calculated to be 330 W/cm at with piezoelectric ceramic transformer,” in Proc. IEEE INTELEC Conf.,
100 kHz. 1992, pp. 430–437.
[13] R. L. Steigerwald, “A comparison of half-bridge resonant converter
We also showed that it is possible to effect soft-switching of topologies,” IEEE Trans. Power Electron., vol. 3, pp. 174–182, Apr.
a piezoelectric transformer in an inverter circuit without any 1988.
magnetic components and to achieve efficiencies on the order
of 83%. Without soft-switching, it would be impossible to
attain efficiencies above 70%. Significantly, we experimentally
verified that the entire driver for a cold cathode fluorescent Anita M. Flynn received the B.S.E.E., M.S.E.E.,
lamp can be implemented without the need for any magnetic and Ph.D. degrees from the Massachusetts Institute
of Technology, Cambridge, in 1983, 1985, and 1995,
components. The manufacturing process for piezoelectric respectively.
transformers can be much simpler than for electromagnetic She spent two years at the University of California,
transformers as no windings or assembled cores are required. Berkeley (UC Berkeley), as a Lecturer/Post-doctoral
Assistant from 1995 to 1997, researching piezo-
The thin, flat form factor of the piezoelectric transformer electric transformers. She is presently President
further enhances this packaging consideration. of MicroPropulsion Corporation, specializing in
consulting/contracting work in the areas of circuit
design, power electronics, and MEMS. She was a
REFERENCES Visiting Professor in the Mechanical Engineering Department, UC Berkeley,
[1] H. W. Katz, Ed., Solid State Magnetic and Dielectric Devices. New from 1999 to 2000.
York: Wiley, 1959.
[2] C. Rosen, “Electromechanical Transducer,” U.S. Patent 2 974 296, Mar.
7, 1961.
[3] T. Zaitsu, T. Shigehisa, M. Shoyama, and T. Ninomiya, “Piezoelectric
transformer converter with PWM control,” in Proc. 1996 IEEE Appl. Seth R. Sanders (M’88) received the S.B. degree in
Power Electron. Conf. (APEC’96), 1996, pp. 279–283. electrical engineering and the S.B. degree in physics,
[4] S. Kawashima, O. Ohnishi, H. Hakamata, S. Tagami, A. Fukuoka, T. the S.M. degree in electrical engineering, and the
Inoue, and S. Hirose, “Third order longitudinal mode piezoelectric ce- Ph.D. degree in electrical engineering, all from the
ramic transformer and its application to high-voltage power inverter,” in Massachusetts Institute of Technology, Cambridge,
Proc. IEEE Ultrason. Symp., Cannes, France, 1994, pp. 525–530. in 1981, 1985, and 1989 respectively.
[5] Piezo Systems Technical Literature [Online]. Available: From 1981 to 1983, he was a Design Engineer at
http://www.piezo.com the Honeywell Test Instruments Division, Denver,
[6] B. Jaffe, W. R. Cook, and H. Jaffe, Piezoelectric Ceramics. New York: CO. Since 1989, he has been on the faculty of the
Academic, 1971. Department of Electrical Engineering and Computer
[7] T. Ninomiya, M. Shoyama, T. Zaitsu, and T. Inoue, “Zero-voltage- Sciences, University of California, Berkeley, where
switching techniques and their application to high-frequency converter he is presently Associate Professor. His research interests are in high frequency
with piezoelectric transformer,” in Proc. 20th Ann. IEEE Conf. Ind. power conversion circuits and components, in variable speed ac motor drive
Electron., Bologna, Italy, 1994, pp. 1665–1669. systems, and in nonlinear circuit and system theory as related to the power
[8] M. Shoyama, K. Horikoshi, T. Ninomiya, T. Zaitsu, and Y. Sasaki, electronics field. During the 1992 to 1993 academic year, he was on industrial
“Steady-state characteristics of the push-pull piezoelectric inverter,” leave with National Semiconductor in Santa Clara, CA.
in Proc. Conf. Rec.28th Ann. IEEE Power Electron. Spec. Conf. Dr. Sanders received the NSF Young Investigator Award in 1993 and is a
(PESC’97), St. Louis, MO, 1997, pp. 715–721. co-author on papers awarded prizes by the IEEE Power Electronics Society and
[9] P. J. M. Smidt and J. L. Duarte, “Powering neon lamps through piezo- the IEEE Industry Applications Society. He has served as Chair of the IEEE
electric transformers,” in Proc. Conf. Rec. 27th Ann. IEEE Power Elec- Technical Committee on Computers in Power Electronics, and as a Member-At-
tron. Spec. Conf. (PESC96), Baveno, Italy, 1996, pp. 310–315. Large of the IEEE PELS Adcom.