Available online at www.sciencedirect.

com
Sensors and Actuators A 141 (2008) 558564
An LVDT-based self-actuating displacement transducer
Shang-Teh Wu

, Szu-Chieh Mo, Bo-Siou Wu

Department of Mechanical Engineering, National Yunlin University of Science & Technology,
Touliu, Yunlin 640, Taiwan
Received 23 June 2007; received in revised form 14 September 2007; accepted 8 October 2007
Available online 18 October 2007
Abstract
A novel linear actuator with inherent sensing capabilities is developed from a linear variable differential transformer (LVDT). By superposition
of dc currents upon high-frequency excitation signals, the coils of the LVDT are capable of exerting a push/pull force on the armature while
retaining the original displacement-sensing function. LC resonant circuits and active bandpass lters are designed to effectively isolate the sensing
signals from the dc energizing sources. Linearity and sensitivity of the device are largely preserved compared to the original LVDT. For actuation,
it is found that the magnetic force on the armature can be maximized by energizing a specic combination of the three coils according to armature
positions. A simple control algorithm is developed for closed-loop control. It is implemented with an 8-bit microcomputer. Real-time testing shows
that rest-to-rest control can be achieved with positioning errors of 0.020.04 mm and settling time below 0.2 s.
2007 Elsevier B.V. All rights reserved.
Keywords: Linear actuator; Displacement transducer; Linear variable differential transformer (LVDT); Self-sensing; Self-actuating
1. Introduction
In a servomechanism displacement or velocity transducers
are essential components. However, installment of an external
sensing device not only incurs expenses but also takes up space
that may not be readily available. Various techniques of sen-
sorless or self-sensing actuation have been developed for the
purpose of reducing cost, space, and increasing system robust-
ness with fewer components. For electric motors, velocity of
the rotor can be estimated from the back emf by observing the
voltagecurrent relations. Displacements can then be obtained
by integration of velocities. Although most sensorless control
methods are designed for rotary machines, self-sensing tech-
niques have also been developed for linear actuators [15] and
the related loudspeakers [68]. Self-sensing control is also an
active research topic in magnetic levitated systems [912]. It
involves estimating air gaps of the magnetic bearing by mea-
suring the coils inductance, which is inversely proportional
to the size of gap. To better estimate the inductance, a high-
frequency test signal could be injected into (i.e. superposed
upon) the input power amplier, generating a high-frequency

Corresponding author. Tel.: +886 5 5342601 4111; fax: +886 5 5312062.

E-mail address: [email protected] (S.-T. Wu).
current whose magnitude is inversely proportional to the coil
inductance [10].
Instead of adding sensing capability to an electromagnetic
motor or actuator, this paper explores an alternative approach
that equips a displacement sensor with self-actuating capabil-
ities. A linear variable differential transformer (LVDT) will be
converted into a transducer that functions as a linear actuator
as well as a displacement sensor. The LVDT is a robust, high-
resolution displacement sensor since it has no electrical contact
between the moving element and the transducers coils. LVDTs
are usually installed on a linear motor or actuator for servo con-
trol. A sub micron positioning system can be achieved by an
LVDT-equipped piezoactuator [13]. One drawback for a system
to install an LVDT is the extra space needed for the sensor. By
making it function as an integrated sensor and actuator, the size
and cost of a servomechanism can be reduced.
The original LVDT operates by the principle that a harmonic
excitation signal applied to a primary coil will induce voltages
on two secondary coils. The voltage differential varies linearly
with the displacement of an armature moving inside the coils.
The weakharmonic excitationhas a negligible force onthe arma-
ture, a magnetically permeable iron core. The idea behind the
proposed LVDT actuator is to energize the coils by superim-
posing dc (or low-frequency) voltages on the high-frequency
excitation signals. The primary and secondary coils are
0924-4247/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.sna.2007.10.027
S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564 559
separately energized to obtain a force of the desired orienta-
tion. While the magnetic force on the armature is not as strong
as a conventional actuator, the resolution of sensing is largely
preserved. Using an LVDT signal conditioning IC, the armature
displacement can be readily measured.
This paper presents the design of the LVDT actuator. A pro-
totype is built for experiments. Linearity and sensitivity of the
device as a displacement sensor are tested. Actuation forces
exerted on the armature are also measured by a high-resolution
pushpull meter. It is found that the force can be maximized
by energizing a specic combination of the primary coil and
the two secondary coils according to armature positions. A
switching strategy for maximum force is established. A simple
control algorithmis then developed and implemented by an 8-bit
microcomputer. It is shown that closed-loop control is achiev-
able with settling time below 0.2 s and positioning errors of
0.020.04 mm.
2. Design of the LVDT actuator
2.1. LVDT as a displacement sensor
An LVDThas three solenoidal coils wound around a cylindri-
cal tube, as shown in Fig. 1. A sinusoidal signal is applied at the
primary coil, inducing voltages on the two secondary coils. As
the armature moves inside the tube, the induced voltage associ-
ated with one secondary coil increases while the other decreases.
The armature displacement canthus be determinedbymeasuring
the differential of the two induced voltages. In practice a signal
conditioningICsuchas AD598(byAnalogDevices, Inc.) is used
to convert the sinusoidal differential to a dc voltage proportional
to the armature displacement.
Dimensions of the LVDTexperimented in this paper are indi-
catedinFig. 1. The dimensions are close tothose of the Schaevitz
E200 (by Measurement Specialties, Inc.), an LVDT with a lin-
ear range of 5 mm. Note that the three coils are of the same
dimension, and the length of the armature is about the length
of two coils. When the armature is at the leftmost position (i.e.,
when the left end of the armature reaches the left end of Sec-
ondary coil 1), mutual inductance between the primary coil and
Secondary coil 1 is maximal while mutual inductance between
the primary coil and Secondary coil 2 is minimal. The opposite
is true when the armature is moved to the rightmost position.
2.2. Superposition of sensing signals and actuating
currents
Fig. 2 shows the basic conguration of the proposed LVDT
actuator, where a set of inductors and capacitors are used to
isolate the high-frequency sensing signal from the dc actuating
voltages. The excitation signal v
s
will be set at a relatively high
frequencybetween10and20 kHz. Note that the impedance of an
inductor with inductance L is equal to jL, where is the radian
frequency of the signal, and the impedance of a capacitor with
capacitance Cis equal to 1/jC. Hence the inductor appears as a
high impedance to high-frequency ac signal but a lowimpedance
to the dc sources (u
1
, u
2
and u
3
). The harmonic signals are
blocked away from the dc sources while the dc currents are
passed through the inductor to energize the coils. In other words,
if the inductance is large enough, the high-frequency excitation
Fig. 1. Linear variable differential transformer: the coils (top left), the armature which moves inside the coils (middle left), the casing (bottom left), and photo of the
experimental setup (right).
560 S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564
Fig. 2. Basic conguration of the LVDT actuator: v
s
is a harmonic excitation
signal; v
1
and v
2
are the induced signals; u
i
s are the actuating voltages.
signal and the induced signals on the secondary coils will not
leak into the dc sources. On the other hand, the capacitors
appear as a low impedance to high-frequency signals but a high
impedance to the dc sources, so that the induced harmonic volt-
ages (v
1
and v
2
) are passed through to reach the measurement
circuits while the dc currents are blocked away. Note that by dc
we mean at frequencies much lower than that of the excitation
signal.
Fig. 2 illustrates the basic idea but has some problems in
practice. First it needs a large inductor to effectively block the
ac excitation signal. For example, to have an impedance of 100 k
to a 10 kHz signal, an inductor of 1.59 Hhas to be used. Large
inductors are undesirable because they take up space and slow
down response. The second problem is with the coupling capac-
itors. While the actuating voltages (u
1
, u
2
, and u
3
) are primarily
of lowfrequencies, high-frequency switching occurs in the tran-
sient state. The transient actuating voltage may have frequency
components close to or larger than that of the ac excitation signal.
A single coupling capacitor can hardly distinguish the sensing
signal from the transient voltage.
Exploiting the fact that the excitation frequency is constant
and known, more efcient stop/pass circuits are constructed, as
shown in Fig. 3. In this rened design, an LCpair is used instead
of the single inductor to block a constant-frequency signal. This
Fig. 4. Frequency response of the bandpass lter. Note the unity (0 dB) gain at
15.9 kHz for the dashed curve (calculated response); solid curve is the measured
response.
is because the impedance of the LC pair is
jL
1
2
LC
. (1)
From Eq. (1) the LC pair appears as an innite impedance for a
signal of 1/

LC rad/s (1/2

LC Hz), which is the resonant

of the LC circuit. The inductance and capacitance are chosen
to be 10 mH and 0.01 F, respectively, so that the resonant fre-
quency is 15.9 kHz. The harmonic excitation signal is set at this
frequency so that it will be well isolated from the dc sources.
The inductance chosen for the LC pair is much smaller than is
required for a single inductor.
In place of the coupling capacitors of Fig. 2, a narrow-band
active lter is used in Fig. 3 to pass the induced harmonic
signals. It is a second-order tuned-circuit lter that can be con-
veniently built using an active-lter ICsuch as UAF42 (by Texas
Instruments, Inc.). The center frequency is set at 15.9 kHz, same
as the excitation signal. Signals of other frequencies, lower or
higher, are effectively attenuated. Such a design prevents high-
frequency transient signals (due to switching of the dc power
input) as well as low frequency actuating currents from contam-
Fig. 3. Rened design of the LVDT actuator: v
s
is a signal of frequency 15.9 kHz, same as the resonant frequency of the LC pairs; center frequency (f
s
) of the
band-pass lters is also set at 15.9 kHz.
S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564 561
Fig. 5. Output voltage of the signal conditioner versus armature displacement
for various dc energizing voltages (a), and the close-up view (b)
Fig. 6. Force measurement by a digital pushpull meter.
inating the LVDT signal conditioning circuits. Fig. 4 shows the
frequency response of the band-pass lter constructed with an
UAF42.
3. Sensitivity and forcedisplacement curves
The sensitivity curves of the LVDT actuator are measured
with the signal conditioning IC (Fig. 3). The harmonic excita-
tion signal for the primary coil is also generated by this IC. Fig. 5
shows the curves of the output voltage from the signal condi-
tioner versus armature positions. The measurements are repeated
for different dc energizing voltages from 0 to 5 V. It is seen that
the lines shift slightly upward with the applied voltage. While
this is an undesirable phenomenon, for high-precision applica-
tions the measured signals can be calibrated against the applied
voltage, which is a controlled quantity and is constant in steady
state. For instance, in rest to rest operation the applied voltage
will tend toward zero.
Let linearity of the sensing curves be dened to be
max
i
|(x
i
)|
S
where (x
i
) denotes the voltage deviation of the ith measured
data fromthe least-squares line, andSdenotes the range of output
voltage. In our device the output voltage varies from 0 to 5 V
for a stroke of 5 mm, so that S = 5 V. The linearity for the
data curves shown in Fig. 5 is calculated to be between 0.25 and
0.33%.
Fig. 7. Three different methods of energization (for rightward force). In Type 3 Secondary coil 1 is energized in one direction and the other two coils in another.
562 S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564
Fig. 8. Force versus armature displacement for an applied dc voltage of 15 V: (a)
rightward force; (b) leftward force. Solid curves are the quadratic least-squares
t of the measured data.
Next is measurement of magnetic forces on the armature
(Fig. 6). Each of the primary and secondary coils can be inde-
pendently energized according to the position of the armature
and the intended force orientation. Three different energizing
methods are tested. The rst one is energizing a single coil (one
secondary coil on either side); the second method is energizing
two neighboring coils (the primary coil and one secondary coil);
the third is energizing two neighboring coils in one direction and
the remaining coil in the opposite direction (Fig. 7). Figs. 8 and 9
compare the forcedisplacement curves for the three different
energizing methods (types). It is seen that, depending on the
position of the armature, either Type 2 or Type 3 yields the
maximum force.
Refer to Fig. 7 and denote displacement of the center of the
armature by x. Fig. 9 indicates that, when x > 2.5 mm, the
primary coil and Secondary coil 2 should be energized in one
directionandSecondarycoil 1be energizedinthe opposite direc-
tion (Type 3) in order to generate maximumpositive forces. (For
negative forces the process is reversed.) This can be explained
by the phenomenon that energizing the two coils on one side
of the armature not only pulls the armature toward them but
Fig. 9. Least-squares tted curves for voltages from 14 to 16 V: (a) rightward
force; (b) leftward force.
also polarizes the armature, so that energizing the third coil in
the opposite direction helps push the armature. In the design of
the closed-loop controller, the switching law will be based on
the maximum force curves. That is, the coil(s) to be energized
and the directions will be determined from a table based on the
crossing points shown in Fig. 9. Note that an average will be
taken for the crossing points since they vary slightly with the
voltages: For positive force the crossing is set at 2.5 mm and
for negative force it is set at 2.3 mm.
4. Closed-loop control
Employing both the sensing and actuating capabilities, one
can devise a closed-loop controller to move the armature of the
LVDT actuator to a desired position. A modied proportional
and derivative control algorithm is designed as follows:
u = k
p
e k
d
x (2)

e = e +e (3)
e = x x
d
(4)
S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564 563
Fig. 10. Driving circuits for energizing the three coils. The control input is from
an 8-bit microcontroller (05 V). Four analog switches (controlled by digital out-
put pins sw
1
sw
4
) are used to select the coil(s) to be energized and determine
the current directions.
where x
d
is the desired displacement of the armature, k
p
, k
d
and are positive constants. The velocity of the armature ( x) is
estimated by calculating the difference of two consecutive xs.
Since the initial transient currents (due to the proportional
gain) may penetrate the band-pass lter and contaminate the
displacement signals, a low-pass lter of cutoff frequency (Eq.
(3)) is included to smooth out the proportional control input. The
lter slows down the initial response but does not affect steady-
state performance. In the experiments the sampling time is 4 ms,
k
p
= 3, k
d
= 0.068, and = 62.5.
The simple control algorithm is implemented by an 8-bit
microcontroller (PIC16F877 by Microchip, Inc.). The IC is
equipped with 10-bit A/Dconverters and two PWMoutput chan-
nels. The latter are used as D/A converters with an external RC
low-pass lter. That is, the output from the microcontroller is a
0 to 5 volt analog signal, which is directed to three power opera-
Fig. 11. Closed-loop response: (a) armature displacements; (b) control input as
a percentage of the full-scale voltage.
Fig. 12. Responses for target displacements of 1, 2, 3, and 4 mm.
tional ampliers (OPA547) that in turn drive the three coils. As
shown in Fig. 10, a set of analog switches are used to select the
coil(s) to be energized (sw
1
and sw
3
) and determine the direc-
tion of currents (sw
2
and sw
4
). The selection is based on the
maximum force curves discussed in the previous section. The
voltages from the microcontroller are amplied ve times, so
that the maximum driving voltage on each coil is 25 V.
Fig. 11 shows the response of the armature moving from
0 to 1 mm and the corresponding control command. The set-
tling time is 0.11 s and the steady-state error is about 0.02 mm.
Fig. 12 shows the responses for target displacements of 14 mm
using the same control parameters. The settling time varies from
0.093 to 0.196 s, and the steady-state errors are between 0.02 and
0.04 mm.
5. Conclusions
By superposition of dc voltages on the original high-
frequency excitation signal of an LVDT, a self-actuating
linear-displacement sensor is devised. With the additional power
currents, linearity and sensitivity of the original LVDT are
largely preserved. It is found that energizing two consecutive
coils in one direction and the third coil in the reverse direc-
tion yields maximum force for most of the moving range of
the armature. Rest-to-rest positioning is achievable with a mod-
ied PD control algorithm, that can be implemented by an 8-bit
microcomputer.
The experimental results also indicate that there is room for
improving in the prototypes steady-state precision, speed of
response, and the magnetic force. This may involve a renement
of the ltering circuits and the control algorithm. To raise the
magnetic force, different dimensions of coils and armature may
be explored.
Acknowledgment
This research was supported by the National Science Council,
Taiwan, ROC, under grant number NSC95-2221-E-224-031-
MY2.
564 S.-T. Wu et al. / Sensors and Actuators A 141 (2008) 558564
References
[1] D. Pawelczak, H.-R. Trankler, Sensorless position control of electromag-
netic linear actuator, in: Proceedings of the 21st IEEE Instrumentation and
Measurement Technology Conference, IMTC 04, 2004.
[2] B. Hanson, M. Levesley, Self-sensing applications for electromagnetic
actuators, Sens. Actuators A 116 (2004) 345351.
[3] Y.-R. Chen, N.C. Cheung, J. Wu, Sensorless drive of permanent magnet lin-
ear motors using modied Kalman lter, in: Proceedings of PESC2001, the
32nd Annual IEEE Power Electronics Specialists Conference, Vancouver,
2001, pp. 20092013.
[4] J. Hirai, T-W. Kim, A. Kawamura, Position-sensorless drive of linear pulse
motor for suppressing transient vibration, IEEE Trans. Ind. Electron. IE-47
(2) (2000) 337345.
[5] R. Leidhold, P. Mutschler, Sensorless position-control method based on
magnetic saliencies for a Long-Stator Linear Synchronous-Motor, in: Pro-
ceedings of IECON 2006, the 32nd IEEE Annual Conference on Industrial
Electronics, 2006, pp. 781786.
[6] Y. Li, G.T.-C. Chiu, Control of loudspeakers using disturbance-observer-
type velocity estimation, IEEE/ASME Trans. Mechatronics 10 (1) (2005)
111117.
[7] M.R. Bai, H. Wu, Robust control of a sensorless bass-enhancedmoving-coil
loudspeaker system, J. Acoust. Soc. Am. 105 (1999) 32833289.
[8] D.J. Leo, D. Limpert, A self-sensing technique for active acoustic attenu-
ation, J. Sound Vib. 235 (2000) 863873.
[9] L. Li, T. Shinshi, A. Shimokohbe, State feedbackcontrol for active magnetic
bearings based on current change rate alone, IEEE Trans. Magnet. 40 (6)
(2004) 35123517.
[10] K.K. Sivadasan, Analysis of self-sensing active magnetic bearings working
on inductance measurement principle, IEEE Trans. Magnet. 32 (2) (1996)
329334.
[11] C. Choi, K. Park, Self-sensing magnetic levitation using a LC resonant
circuit, Sens. Actuators A 72 (1999) 169177.
[12] B.Z. Kaplan, G. Saraan, Employing chaos tools and methods in
magnetic levitation, IEE Proc. Sci. Meas. Technol. 146 (1) (1999)
1520.
[13] S. Salapaka, A. Sebastian, J.P. Cleveland, M.V. Salapaka, Design, identi-
cation and control of a fast nanopositioning device, in: Proceedings of the
2002 American Control Conference, 2002, pp. 19661971.
Biographies
Shang-Teh Wu was born in Changhua, Taiwan, in 1963. He received the BS
and MS degrees in mechanical engineering from National Taiwan University,
in 1985 and 1989, respectively, and the PhD degree in mechanical engineer-
ing from Massachusetts Institute of Technology in 1993. He joined National
Yunlin University of Science & Technology since 1993, and is currently a
professor in the Department of Mechanical Engineering. His research inter-
ests include vibration and control, design of electromechanical sensors and
actuators.
Szu-Chieh Mo was born in Taipei, Taiwan, in 1983. He received the BS degree
from Ta Hua Institute of Technology in 2005 and the MS degree in mechan-
ical engineering from National Yunlin University of Science & Technology
in 2007. He is currently working at Foxlink Image Technology Inc., Taipei,
Taiwan.
Bo-Siou Wu was born in Taichung, Taiwan, in 1982. He received the BS and MS
degrees in mechanical engineering from National Yunlin University of Science
&Technology, in 2004and 2006, respectively. He is currently working at Avision
Inc., Hsinchu, Taiwan.