Eddy current sensors on Printed Circuit Board for compact mecha-

tronic application

Eng. Olivier Sosnicki, Cedrat Technology, Meylan, France

Eng. Gregory Michaud, Cedrat Technology, Meylan, France
Dr. Frank Claeyssen, Cedrat Technology, Meylan, France

Abstract
In a context of always smaller and smarter mechatronic devices, the need of more integrated sensors becomes critical.
Particularly, small mechanisms using small actuators like piezo actuators require compact sensors, with performances
that measure up to the actuators characteristics. Eddy current position sensors (ECPSs) are widely used for non contact
position, displacement, and proximity measurement with very good performances (bandwidth resolution, linearity, ro-
bustness in dirt environment). Operating on the principle of magnetic induction, these detectors can precisely measure
the position of a metallic target. With the performances increase in PCB manufacturing, a specific design of this kind of
sensors on Printed Circuit boards allows a compact device which could be coupled with its electronic conditioner. The
modeling of the magnetic behavior of the probe and the coupling with a dedicated electronic conditioning are studied in
this paper. Tests are involved on a X piezoelectric stage mechanism to verify the final performances. PCB-based eddy
current sensors allows less space offering reduced weight with more controllability in integrated mechatronic devices.

In compact contactless Nano positioning, only the capaci-

tive sensors or the Eddy Current Position Sensors (ECPSs)
1 Introduction could be used in industrial application. The optical meas-
The mechatronic demands more and more precise motion urement is not competitive because of the cost and the al-
in compact volume. From several years, actuators [1],[2] located volume of the final solution. The strain gages are a
are designed to answer to these problematic of integration contact sensor which have several drawbacks especially
and accuracy but the added sensors to help this last point when small and fragile parts must be measured.
are quite large and reduce the overall capability to scaling
down the actuator-sensor pair. From this aspect, we have chosen to design a new Eddy
In the high resolution/accuracy domain like piezoelectric Current Position Sensor able to sense nano-metric position
actuators, only few technologies are in competition where in the sub-millimetric range able to be integrated on com-
the ratio between the range of few microns compared the pact and embedded mechatronic mechanisms. The chal-
embedded body of the sensor is very small (see Figure lenge is to reduce the overall volume and so to implement
1.1). the ECPS on a standard Printed Circuit Board in respect to
the optimal performances.
Sensor Advantages Drawbacks
Type 1.1 Physical phenomena
Eddy current Simple structure, low Measuring signal will Eddy-current sensors have been studied for a number of
cost, light weight, dura- change with material prop- years as a method of non-destructive testing and non-
ble, harsh environment, erties, calibration is neces- contact measurement.
bandwidth is high sary with specific material Based on Faraday's law of electromagnetic induction,
target.
dΦ
Capacitive Simple structure, resolu- Expensive, sensitive to ε =− (1)
tion and accuracy are fine change in the dielectric dt
sensor gap where ε is induced emf (electromotive force), and dΦ/dt is
Optical High precision, direct Expensive, installation the rate of change of the magnetic flux, the physical model
measurement of the posi- complexity and susceptibil- of measurements (see Figure 1.2) consists of the target ob-
tion ity to optical contamina- ject and the main component of the sensor that is an induc-
tion; rather for ground tion coil. When an alternating voltage or current is applied
based laboratory testing to the induction coil, it generates an oscillating magnetic
than harsh environments. field, which induces eddy currents on the surface of the
Strain gage Direct measurement of Accuracy, difficult to be conductive target, according to the principle of eddy cur-
stress in-situ, mounted on small part. rent induction [3]. Eddy currents circulate in a direction
Figure 1.1 Comparison of high accurate/small range con- opposite to that of the coil, reducing the magnetic fluxes in
tact less sensors in mechatronic systems. the coil and thereby its inductance. Eddy currents also dis-
sipate energy, and therefore lead to an increase in the resis-
tance of the coil. For high-precision measurements, prefer- arity. The largest impedance changes per unit of displace-
able applications should make use of nonmagnetic conduc- ment occur within the first 1/3 of the standard measuring
tive target materials like Aluminium, steel: The less resis- range and is very closed to an exponential function. The
tive the target, the more eddy currents are produced in the design should maximise the variation of the Q(x) factor on
target. This improves sensitivity and signal to noise ratio, the specified range. We will see later the importance of
ultimately increasing resolution [5]. this exponential form to electronically improve the linear-
ity.
Probe As rule of the thumb, the diameter of the sensor should be
designed to be 3 times greater than the range. In the Fig-
ure 1.3, the range is 500µm for a 1.5mm probe diameter.

The frequency response of the sensor is a like a resonant

filter composed of the R and L components and the para-
site capacitance C (cable, winding,…) with
Target size ω res = 1 / Leddy xC parasite .
Figure 1.2 Physical principle of an ECPS.

From this physical phenomena, it is possible to define an

inductor coil and an electronic conditioner to transform the
Eddy current effect in electrical signal compatible with the
position sensing application, i.e a position sensor with a
voltage output and with a sensitivity in µm/V.

1.2 Concept
As explained, an ECPS is composed in a eddy current gen-
erator , i.e an inductor coil, and an electronic conditioner
able to transform the impedance variation due to the posi-
tion sensing in voltage. Figure 1.4. Frequency response of the ECPS

The different relations to design this kind of sensors are The working frequency is chosen to be 3 times lower than
well established in [6] and [7]. The complex impedance the resonant frequency Fres (see Figure 1.4) by adding an
varies with the displacement between the inductor coil and external capacitor.
the target. At fixed frequency and infinity target, the qual-
ity factor Q(x) can be explained as: In the industry, two main topologies are used based on an
L( x) xω alternative current supply: The driving oscillator (or Col-
Q( x) = (2) pitts oscillator) or the bridge oscillator.
R( x)
The first one can be more embedded because only one in-
With ω, the pulsation of the current/voltage, L(x) et R(x)
ductive coil is used. In the Colpitts oscillator circuit, the
respectively the inductance and the resistance over the
sensor acts as the resonating coil for the oscillator. The
range x (see Figure 1.3).
proximity of the targets to the sensor coil pulls the oscilla-
tor, changing its frequency (small change not used) and
amplitude (large change) of modulation. The signal is half-
wave rectified (AC to DC) and filtered to obtain an analog
voltage proportional to the target position. The main draw-
back is to implement a non linear function able to improve
the intrinsic linearity.
A derived topology uses a Wheastone bridge : The bridge
is tuned to be near resonance so that slight changes in in-
ductance in the sensor coil (caused by changes in the elec-
tromagnetic coupling between the sensor and the target)
provide shifts in the output of the bridge. Target movement
relative to the sensor causes impedance change in the sen-
Figure 1.3 R(x), L(x) and Q(x) variations over the range: sor coil. The impedance of the coil is detected (measured)
500µm and Fexcitation=8MHz. by the RF control circuitry. The dual coil has an active and
an inactive reference coil for improved temperature stabil-
The design of a position sensor is intimately connected to ity.
the optimisation of the Signal to Noise Ratio and the line-
In the differential bridge, the coils of a pair of sensors form Years Line Manufacturing Substrates
the opposite legs of a balanced bridge circuit. the main ad- width
vantage is the suppression of the anti-Log function and so 1985 500µm Single sided board, Rigid,
improve the drift and the resolution by eliminating electri- 1985-90 250µm Dble sided board, Flex,
cal noise from this electronic function. 1995-00 150µm Multilayers board, Rigid, Flex,
When two coils are used, the possibility towards a compact 2000-05 100µm Microvia High 3D
solution is slightly reduced. (see Figure 1.5-a, b, c). For 2005-08 75-50µm Density Intercon-
this reason, the Colpitts oscillator is chosen to compact the 2008-10 30µm nect board.
final solution of the inductor coil. Figure 1.6 PCB technology improvement since 1985 [10].

2 Design of the Inductor coil

The inductor coil is the core of the ECPS. The design of
this part mixes modelling and manufacturing process man-
agement.

2.1 Modelling
The aim of the modelling work is to find the best coil con-
figuration which maximise the eddy current effect to im-
prove the sensitivity of the sensor on the range.
The modelling is based on the analysis of the constitutive
lumping element of the coil. The effective impedance de-
pends on the complex interaction among several variables
(target, distance, shield) and the FEM analysis gives a
more efficient way to handle the design issues (see Figure
2.1). The modelling will be used to optimise the Q(x)
variation and so the sensitivity of the ECPS. Flux 2D [8]
and INCA3D [9] were performed to analyse the critical
behaviours of the ECPS. The first one gives the behaviour
of the R(x) and L(x) with the range and the second one is
used to analyse the capacitive behaviour to place the reso-
nant frequency much higher than the working frequency.
a- Isolvalues and Equiflux

Figure 1.5 Concept of electronic conditioner for the

ECPS a- Colpitts oscillator configuration , b-Wheastone
configuration, c- Differential configuration.

Standard solutions to realise the inductor coil use wired

coil but are quite large due to electrical connections. Our
solution is to use the latest improvements in PCB technol-
ogy (See Figure 1.6) to obtain a sensor with a small area
b- Equiflux and Eddy current density

and a low profile. Several substrates like Low Temperature

Cofired Ceramic, Printed Circuit Board can be used. The
chosen substrate is FR4: The laminate is constructed from
glass fabric impregnated with epoxy resin (known as "pre-
preg")and copper foil. The copper foil is partly etched
away, and the remaining copper forms a network of elec-
trical connections. For several reasons as a topology not
enough complex (only two connections), as the PCB tech-
nology is improved since several years (see Figure 1.6), as
the standard temperature will not reach more than 100°C,
the classic FR4 substrate can be used as an industrial solu-
tion. Figure 2.1 FEM analysis of the ECPS: a-Equiflux, Iso-
value; b-Equi-flux, Eddy current density.
The results of the FEM analysis are extracted in the Figure fine substrates (core + pregs) to improve the mutual cou-
2.2 a, b to trace the variation of the R(x), L(x) and Z(x) pling between each layer.
parameters and to obtain the Q(x) variation for the measur- Finally, the High Density Interconnects PCB technology
ing range. was chosen with a standard FR4 based multilayer with a
2mm of height. The HDI PCB design follows the IPC-
2226 - level A. [10].
a- R(x), L(x) and Z(x) variations

For more integrated sensor with small range (see later), the
HDI PCB is more challenging but substrates building with
same number of layers were validated with height below
1mm and 2mmx2mm area following the IPC-2226-Level
C [10].

Figure 2.2 3D High Density Interconnects PCB of the in-

ductor coil.
b- Q(x) and ∆ Q(x)

2.3 Scaling down

The presented solution for 2mm range used a 5mmx5mm
multilayer planar coil. A scaling down of the overall di-
mension of the sensor will reduce the target size and fo-
Figure 2.2 R(x), L(x) and Z(x) variations- Q(x) and cuses the field on a smaller area which improves the sensi-
∆Q(x) for 2mm range at 1MHz tivity. In [11], it is shown the relations between the Q(x)
variation, the scaling factor and the frequency. Compared
On a 1.5mm measuring range, the variation of the quality to a wired solution a scaling down of a planar coil without
factor is more than 130%. modification of the intrinsic characteristics of the planar
coil accomplishes a reduction of the Q(x) variation. From
[12] in a multilayer planar coil, the way to improve Q(x) is
to increase the mutual and the self inductances and to re-
duce the resistance of the wire which is difficult in the
same area.

Several simulations were performed to analyse the behav-

iour of the sensitivity with the physical dimension of the
coil in Figure 2.3.

Figure 2.3 Complex impedance of the inductor coil ver-

sus frequency

The modelling impedance shows a resonant frequency near

11.85MHz, which permits to work around the MHz by
a- Q (x) variations

adding parallel capacitance. The parasite capacitances

from the PCB are near 6.3pF.

2.2 PCB manufacturing process

As the optimal performances are reached with a high den-
sity PCB, the solution has been found in the phone cellular
technology. The High Density Interconnects PCB technol-
ogy is a good candidate because of combining buried mi-
crovias, high pitches and small width of lines coupled with
 b- ∆ Q(x) 3 Electronics for ECPS

Figure 2.3 a- R(x), L(x) and Q variations, b- Variation of

∆Q(x) for 500µm range with different physical dimen-
sions.

Q(x)_1,2,3 are issued from a same geometry and with fre-

quencies 1,2 and 4 times than the nominal working fre-
quency. Q(x)_4,5,6 are issued from a same geometry and Figure 3.1 Principle of the electronics of the ECPS
with less inductances from the 1, 2, 3 geometries (0.65 fac-
tor but with the same area and layers)) and with frequen- As seen in the previous chapter, the ECPS conditioner is
cies 1,2 and 4 times than the nominal working frequency. not only an oscillator. The electrical circuit is more com-
Q(x)_7,8 are issued from a same geometry and with less plex with a demodulation function able to extract the am-
inductances from the 1, 2, 3 geometries (0.3 factor but with plitude of the modulated signal and a linearity function
the same area and layers)) and with frequencies 1 and 4 able to reduce the intrinsic non linearity of these kind of
times than the nominal working frequency. position sensor over a certain measuring range. A analogue
based antilog function can be used to correct the non linear
From Figure 2.3, we extract the best solution from the signal. This correction can be perform with a digital func-
modelling steps to be compatible with the oscillator. The tion (After analogue to digital conversion) able to add
solution Q(x)_2 is chosen. some other corrections like active temperature correction.
Finally, a picture of the designed electronic conditioner is
given in the Figure 3.2.

Figure 2.4 R(x), L(x) and Q variations.

The variation of Q(x) is maximised to reach more than

100% in the 500µm range by increasing the frequency by
2.
By reducing the area of the probe, the height can be re-
duced by decreasing the number of layers. In this case the Figure 3.2 Eddy current position sensor conditioner.
total inductance is reduced but the resonant frequency is
increased allowing a higher working frequency. This board can include until 3 synchronised channels to be
integrated on 3 axis mechanisms.

4 Practical tests on a single ECPS

A single planar coil based on a 2mm range is manufac-
tured following the design rules of the chapter2 (see Fig-
ure 4.1) to check the resolution, the accuracy and thermal
drift of the overall chain.
 4.2 Test results

a- Linearity- offset 500µm and stroke 1000µm

Figure 4.1 2mm range manufactured ECPS

4.1 Test setup

The philosophy of the first test bench is to compare the
real time measurement of a 1000µm stroke piezoactuator

b- thermal drift results -20°C +70°C

with a capacitive sensor which is more accurate than the
eddy current sensor (linearity, accuracy and resolution).
The second one is a test bench to measure the thermal drift
of a accurate distance. An Invar + Alumine structure is
used to minimise the thermal drift of the distance between
the probe and the target. A thermal sequence between -
20°C to +70°C is applied on the Device Under Test (see
Figure 4.2 a&b).

Capacitive Aluminium
probe target
Figure 4.3 a- b- Thermal behaviour results on a -20°C to
a- Test bench - capacitive comparison

+70°C temperature range

The linearity is based on an offset of 500µm and a dy-

namic stroke of the piezoactuator of 1000µm. In this
range, the linearity is under +/-1% and is basically function
of the electronic correction. In the same time, the measured
resolution is near 50nmRms on [10Hz-8kHz] bandwidth.

Piezoelectric Band Voltage Power Position Rms

ECPS probe actuator
Noise
10-110Hz 24.91 nVrms² 34nm
110-1710Hz 42.4 nVrms² 32nm
b- Test bench forthermal drift beahaviour

1.71kHz-8kHz 10n Vrms² 16nm

Figure 4.4 Resolution issued from noise measurement on
Invar the voltage output of the sensor

The temperature has a non negligible influence on the sen-

sor behaviour. Fortunately, the inductance remains almost
Alumine constant, a small increase is possible due to the coil ther-
mal expansion, but this effect is negligible. On the other
hand, the sensor electrical resistance increases dramatically
Aluminium with the temperature. It is more than 30% larger at 100°C
target ECPS probe than at room temperature. With the HDI PCB integration,
it’s possible to add a thermal sensor near the inductor coil
to compensate the temperature drift with a look-up table.
Figure 4.2 a- Test bench for resolution/accuracy testing of Finally a thermal drift of the probe around 800nm/°C can
the single Eddy current position sensor. b - Tests bench for be given by deducting the thermo-mechanical drift of the
thermal drift testing of the probe. stiffener.
All these parameters can be computed in a table to be
compared with a standard wired coil eddy current position
sensor (see Figure 4.5).

Performances Wired coil Planar coil

Range – Full 1.25mm 2mm
Scale
offset 250µm 250µm
Dimensions (mm) Dia 5, length 18 5.8x5.8x1.8
Resolution 0.009% of FS 50nm rms or
0.0025% of FS
Linearity +/-0.25% +/-1%
Bandwidth 80kHz 15kHz
Drift 0.04% of FS at 800nm/°C or
mid-range 0.04%/°C of FS
Figure 4.5 Comparison with standard wired coil Eddy
current position sensor [13]. Figure 6.1 Fives ECPS on a 50mm diameter HDI PCB in
contactless Force sensor application.
The different characteristics are very similar: The resolu-
tion and the linearity are directly given from the electronic
unit. On the volume point of view, the main advantage of a 7 Conclusion
planar coil is the drastic reduction of the length of the sen- In this paper, we have shown the new possibilities offered
sor (i.e. 1.8mm versus 18mm for a standard wired coil). by improving the manufacturing process of PCBs to design
This allows a compact integration of the sensing function. compact eddy current position sensors with low profile.
The chosen technology allows similar performances with
6 Integration in piezoelectric wired coils used in the industry. The scaling down process
can be done until parasite effects modify the performance
mechanisms sof the sensor: a 1500µm range sensor in a
An example of a possible integration of this kind of sensor 5.8mmx5.8mm*1.8mm volume has been designed and
is shown in the following figure to sense the stroke be- tested to validate the performance and a 500µm range sen-
tween two (see Figure 6.1). The main possibility of this sor has been simulated to validate the intrinsic probe per-
design is found in the PCB including five probes to meas- formance in small volume 2mmx2mmx1mm. This last
ure the real displacement of a aluminium target. The me- sensor will be implemented in applications where very
chanical integration is simpler than five autonomous small volume are required.
probes. The principle of the Contactless force sensor is to
create an measurement of a torque by sensing the dis-
placement of two mechanical parts (part A and part B see
8 Acknowledgements
Figure 6.1)-[14]. By combining several measurement, it’s The work leading to those results was carried out in the
possible to minimise the parasite offsets and to read the frame of the OSEO SRC project supported by the French
torque and the angle of the shaft. As the mechanism is very Agency.
small, standard eddy current sensors are not a good solu-
tion due to their lengths. The benefits of a printed circuit
board can be directly used to be implemented inside di- 9 References
rectly close to mechanical parts. The dynamic stroke to be [1] Claeyssen, Le Letty, Barillot, Sosnicki: Amplified
sensed is near 20µm with an offset near 200µm. Piezoelectric Actuators: Static & Dynamic Applica-
tions, Ferroelectrics, 2007
The PCB doesn’t show the electrical interface with the [2] Le Letty, Barillot, Fabbro, Claeyssen, Guay,
electronic unit but it can be integrated on the rear face of Cadiergues: Miniature Piezo Mechanisms for Optical
the PCB to pre-amplify the signal before the final treat- and Space Applications, Actuators, 2004
ment away. [3] “Non-contact eddy current displacement and distance
measuring systems”, Website of MICRO-EPSILON
Messtechnik GmbH & Co.KG, www.micro-
epsilon.com.
[4] Dowell, Sylvester et al, "Progress in turbomachinery
prognostics and health management via eddy-current
sensing," Proceedings of Aerospace conference IEEE,
2000, pp. 133-143.
[5] Target Material, “the good, the bad and the hugly”:
website of Kaman Precision, www.kamansensors.com
[6] Roach: Designing and Building an Eddy Current Posi-
tion Sensor, Sensors Magazine, 1998
[7] Welsby, Hitz: True Position Measurement with Eddy
Current Technology, Sensors Magazine, 1997.
[8] FLUX3D v11 user manual, Ed. Cedrat, Meylan (F),
2009. www.cedrat.com
[9] InCa3D® software, Ed. Cedrat, , Meylan (F) 2009,
www.cedrat.com
[10] Holden and al.: The HDI handbook, 1st edition.
[11] Passeraub, Besse, Popovic: Scaling down an induc-
tive proximity sensor, Transducer 95, pp 249-252.
[12] Greenhouse: Design of Planar rectangular Microelec-
tronic Inductors.
[13] Web site of Lion Precision , www.lionprecision.com
[14] Benoit et Al : Contactless Torque Sensor based on
shaft torsion measurement, com-patible with existing
shafts, to be published in Sensoren und Messsysteme
2010
[15] Ardon, Chadebec, et Al: Modelisation d’un capteur à
courant de Foucault via un couplage de méthodes in-
tégrales adaptées , to be published in CEM confe-
rence, 2010, May 2010.