SENSORS-00355-2002

Novel Magnetic Displacement Sensors

Christian Schott Member, IEEE, Robert Racz, Fredy Betschart, Radivoje S. Popovic Member, IEEE

costly structure is implemented. For a simple small magnet the

Abstract— We describe a new magnetic displacement sensor, field decreases with 1/r3 with the distance r. Here again the raw
which measures the position of a magnet along a line or in the output signal must be linearized by data processing.
plane. The sensor is inherently non-sensitive to the magnet In this paper we present an elegant solution, which allows to
movement along the third axis, temperature variations and virtually eliminate all above drawbacks, namely temperature
aging, and does not require calibration. This is achieved by drift, ageing effects and non-linearity at the same time, without
transforming a translation of the magnet into a rotation of a adding new ones.
magnetic field, and by measuring the direction of the magnetic Essentially, the good properties of our sensor are due to the
field rather than its strength. The direction of the magnetic
fact that we measure the rotation of the field of a small magnet
field is then measured by two-axis Hall sensors.
and not the field strength as all known Hall position sensors
do.
Index Terms— Hall sensor, magnetic position sensor, contact-
less position measurement.
II. M EASUREMENT PRINCIPLE
I. INTRODUCTION
A. Single Axis Displacement Sensor

P OSITION sensors encompass a wide range of sensors,

switches and technologies that are used to determine the
The concept of the present magnetic position sensor, which
is based on our recent patent application [1], is shown in Fig.
position, speed or direction of movement of a given target. 1. The sensor consists of two parts: a cylindrical permanent
Linear position devices include capacitive, eddy current, fiber magnet, and a two-axis Hall sensor. The two parts can
optic, Hall effect, inductive, magneto-resistive, optical translatory move with respect to each other, along the three
triangulation, photoelectric, ultrasonic, and variable resistance mutually perpendicular axes x, y, z. We assume that the
technology sensors. substrate is parallel with a (x, y) plane. The permanent magnet
Hall effect magnetic position sensors are the preferred is magnetized along its rotational axis, which is positioned
solution for many applications, since they are contact-less, perpendicularly with respect to the substrate (z-direction).
small in size, robust, reliable, not sensitive to harsh and
polluted environmental conditions, and low-cost. flux lines
But the known Hall effect based position sensors are limited
in accuracy by offset, noise, temperature dependence and
ageing effects. This is the reason why traditionally they are moving
mainly used as contact-less switches detecting just the z y
permanent
presence of a magnet or magnetized body and not as linear N magnet
sensing devices. Through the co-integration of Hall element(s) x
S
and electronics on the same chip, offset and 1/f noise can be
effectively reduced by the spinning current technique.
two-axis
However, to compensate such a position sensor for drift with B Hall sensor
temperature and ageing effects of the Hall sensor and the Prited C
magnet, generally more involved and expensive treatment is irc uit Boa
rd
required, such as calibration and digital data processing.
Another important drawback of magnetic position sensors is Fig. 1 A two-axis Hall sensor is positioned in the vicinity underneath a
the strongly non-linear decrease of the magnetic field strength small permanent magnet with round shape. The magnetization axis of
the magnet is perpendicular to the sensor plane. When the magnet
with distance from the field source, unless a very big and
moves along the x-Axis, the magnetic field at the sensor rotates.

Manuscript received xxxxxx. (Write the date on which you submitted

your paper for review.)
C. Schott, R. Racz and F. Betschart are with Sentron AG, CH-6300
Zug, Switzerland phone: +41 21 693 6774; fax: +41 21 693 8780; e-
mail: christian@sentron.ch.
R.S. Popovic is with the Swiss Federal Institute of Technology When the magnet moves along the x-axis, the component of
Lausanne (EPFL), email: radivoje.popovic@epfl.ch the magnetic field parallel with the (x, y) plane, seen by the Hall

Page 1 of 8
SENSORS-00355-2002

sensor, rotates and changes strength (Fig. 2). Of course, field

rotation also happens for a movement along the y-axis, hence, Both of these curves look quite non-linear, but nevertheless,
for a single-axis displacement sensor Y has to be kept a very linear position information can be obtained from them.
constant. Interestingly, if the magnet moves along its axis (z), For this it is only necessary to build the ratio Vx/Vy as shown
the magnetic field direction in the (x, y) plane does not change. in Fig. 4. This figure also gives the calculated error between
This is because the cylindrical magnet produces a radial the mechanically measured position and the position obtained
magnetic field in any (x, y) plane not identical to its plane of from the magnetic sensor. We can see, that the linear range is
symmetry. This stays so in spite of eventual changes of the limited to about 2.5mm. However, within this range the
strength of the magnetic field due to temperature or ageing absolute error is less than 25µm. This corresponds to 1%
effects. absolute accuracy within this range.

Field Rotation
1.50 0.15

Vx/Vy
1.00 0.1
error [mm]
B
0.50 0.05

ratio Vx / Vy

error [mm]
0.00 0

-0.50 -0.05

-1.00 -0.1

-1.50 -0.15
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
displacement X [mm]

Displacement of Magnet Fig. 4 Linear displacement signal and position error obtained by
Fig. 2 Top view of the single-axis displacement sensor. The magnet division of the two sensor output voltages.
moves parallel with the x-axis and the xy component of the magnetic
field at the place of the sensor changes strength and rotates. To understand this result, we now give a short analytical
explanation.
The question is now: How do we obtain accurate
In our approach we use two-axis Hall sensors for the
displacement information for the magnet from this field
measurement of Bx and By. They measure the in-plane
rotation?
components BX and BY of the magnetic field vector B. The
To solve this, we first have a look at the signals measured
corresponding Hall output voltages are VX and VY. The
by the sensor as plotted in Fig. 3. As the magnet moves, the
sensors are very linear in the used magnetic field range, so that
field component in motion direction Bx is first negative, then
V X = S ⋅ B X , VY = S ⋅ BY (1)
crosses zero at center position and increases to positive
values. For both, far negative and far positive positions, the where S is the sensor sensitivity, S ≠ B.
field becomes weaker as the distance between magnet and From the similarity of the corresponding triangles in Fig. 5
sensor increases. The By field component is always positive we deduce the following:
BX X
with a maximum about zero position. = (2)
BY Y
0.20

B By
0.15

X
Output voltage [mV]

0.10
Bx
0.05

Y
0.00
Vx
α
-0.05 Vy

z
-0.10
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
x
displacement X [mm]
y
Fig. 3 The two sensor output voltages as obtained with a passing by of
the cylindrical magnet. Fig. 5 The top view of the displacement sensor helps us to establish
the relationship between the measured magnetic field components and
the position of sensor and magnet.

Page 2 of 8
SENSORS-00355-2002

On the other hand: A B2

By2
BX VX B1 B y1
= = tan(α ) (3)
BY VY
X
Bx1 Bx2
Then we can calculate from the above equations:
X = Y ⋅ tan(α ) (4) Y
Here tan(a) is obtained by calculating the ratio of the two α1 α2
Hall voltages Eq. (3). These values are independent of the
absolute field strength and also of temperature effects and z
x
ageing as long as these effects are similar for both sensor axes.
This concept allows measuring a position along an axis with y
the great advantage of virtual immunity to the sensor and
Fig. 7 By relating magnetic field components to angles and positions,
magnet tolerances as well as to temperature and aging effects. the equations leading to the two -axis displacement sensor are set up.
Moreover, the measurement result is theoretically independent
to the mutual distance between the magnet and the Hall sensor Then the two sensors measure the in-plane components of
along the third axis (z). the two magnetic field vectors B1 and B2, which are BX1, BY1,
BX2, and BY2. The corresponding Hall output voltages are VX1,
B. Two Axis Displacement VY1, VX2, and VY2. Here also we use the geometrical relations of
After the application of a two-axis Hall sensor for single-axis corresponding triangles to solve the position equations.
displacement sensing, we now extend the principle to two-axis From Fig. 7 we can write:
displacement sensing. Whereas up to now we were using the BX 1 X BX 2 A − X
magnetic field direction in one point to determine one position = , = (5)
BY 1 Y BY 2 Y
coordinate (x) of the magnet, we now want to determine the
coordinates of the magnet in the plane (x and y) and we Then, expressing the ratio of field components by the ratio
therefore need to measure the magnetic field direction in two of output voltages we find:
BX 1 VX 1 BX 2 V X 2
points. This can be realized by using a first and a second two- = = tan(α1) , = = tan(α 2) (6)
axis Hall sensor, as shown in Fig. 6. BY 1 VY 1 BY 2 VY 2
Again, when the magnet moves in the (x, y) plane, the We now can calculate the two coordinates X and Y of the
components of the magnetic field parallel with this plane, seen magnet center with respect to the left Hall sensor to:
by the two Hall sensors (B1, B2), rotate. tan(α 1) A
X = A⋅ ,Y= (7)
tan(α 1) + tan(α 2) tan(α1) + tan(α 2)
flux lines
As before, tan(a1) and tan(a2) are obtained by calculating
the ratio of the two Hall voltages for each of the two sensors
Eq. (6).
moving Here again, due to the field direction measurement rather
z y
permanent than field strength measurement, the two-axis displacement
N magnet sensor is virtually insensitive regarding absolute field strength
x
S of the magnet, temperature effects, ageing and z-axis position.
Moreover, the two-axis displacement sensor does not
B1 require a motion parallel to a given axis, since any position in a
B2 (reasonable) distance from the magnet can be measured. The
only requirement is that the distance between the two Hall
sub str two-axis
a te sensors A is fix.
Hall sensor

Fig. 6 The principle of the two-axis displacement sensor is based on

the measurement of the magnetic field direction in the plane in two
separate spots. This allows to obtain two equations for two unknown,
the position X and the position Y of the magnet. III. EXPERIMENTAL SETUP
Now we will develop the equations defining the magnet For the experimental setup we used two-axis Hall sensors
position with respect to the sensors. Let us assume that two 2D-VH-11 of SENTRON AG [2]. Such sensors are based on
such sensors are placed at a distance A from each other, Fig. 7. vertical Hall devices described in [3] and they are sensitive
parallel with the chip surface. The angular error between the
two sensor axes is smaller than 0.1°.

Page 3 of 8
SENSORS-00355-2002

Two such sensors were mounted on a ceramic substrate at a functions (6), and the magnet position co-ordinates (7), are
distance of A=2mm. A samarium-cobalt cylindrical permanent calculated using a microprocessor.
magnet with a diameter of 6mm and a length of 2.5mm was The position sensor was mounted on precise micrometric
placed at Y=3mm from a line through the sensors and at z=0.5 translation stages (see Fig. 8).
mm above them. All measurements were performed at room temperature,
The Hall elements are biased with the constant voltage of without any temperature stabilization measures.
5V. Except of cancelling offset voltages, no adjustments or
calibrations were made. The Hall output voltages are amplified
using conventional integrated differential amplifiers and then
converted into digital signals. The directional tangent

High precision table with

micrometers Digital electronic with AD-
converter and microcontroller

X Z Y

Sensors on ceramic and axial

magnetized magnet

Amplification circuit (ca. 26)

Display with distance in X-

and Y-direction

MAGNET
Y

Fig. 8 Photograph of the measurement and test setup. The magnetic position sensor is tested with reference to precise micrometric translation
stages. The test setup allows to place the sensor with high precision on x, y and z coordinates with an error of less than 2 µm.

Page 4 of 8
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

IV. SIMULATION OF LIMITATIONS DUE TO NOISE

Before we start with the actual displacement measurements,
we want to quantify the limits of measurement accuracy given
by noise. For this reason we have established a simulation

error [mm]
model of our magnet and sensors where we can calculate the
influence of sensor noise on the displacement value (Fig. 9).
We measured the sensor noise before to a value of ±10 µT
over the frequency range from 0.01 to 100Hz.

displacement x
[mm]
Fig. 11 Simulation result data of the influence of sensor noise onto the
displacement accuracy

A summary of all simulation results within the displacement

range of ±1mm in x-direction is given in Table I.

Table I

Position Direction X [µm] Direction Y [µm]

1 ± 0.6 ± 2.0

Fig. 9 Simulation model of two-axis displacement sensor 2 ± 0.25 ± 0.5

The y and z coordinates for which we simulated a displacement 3 ± 0.2 ± 0.25

in x-direction are given in Fig. 10. The distances are from 0.5mm ± 0.6 ± 2.0
4
to 1.5mm in vertical z direction and from –1.1mm to +1.1mm
from the magnet edge in y-direction. 5 ± 0.3 ± 0.6

6 ± 0.2 ± 0.25
Z
7 ± 0.75 ± 2.0

8 ± 0.35 ± 0.75

9 ± 0.35 ± 0.5

The sensor noise causes a corresponding X-position error

of ±0.75µm, and in Y-position error of: ±2µm. (for a movement
1 2 3
0.5mm in x-direction). This corresponds to the maximum obtainable
0.5mm
4 5 6 accuracy of such a position sensor. In order to further
0.5mm
7 8 9 Y increase the accuracy, the magnetic field has to be increased,
for example by using a bigger magnet or a different magnet
1.1mm 1.1mm material.

Fig. 10 Points of operation which have been examinated using the V. EXPERIMENTAL RESULTS
simulation model
A first measurement result of this setup was already given in
Now we performed a simulation for displacement in x- Fig. 3, showing the measured output signals of one of the Hall
direction and we calculated the uncertainty at each position sensors for a displacement of the magnet along the x-axis.
due to the sensor noise. According to (7) we can quantify this We shall now test the equation (7) of the position sensor
error for the varying coordinate X as well as for the constant system. For a displacement along any one of the three axes x, y
coordinate Y. and z at a time and by keeping the other two coordinates
Fig. 11 shows such displacement error simulation result for constant, we should once receive a signal proportional to X,
Y. The error is smallest in the center, where the signal to noise once proportional to Y and then it should be constant if we
ratio is highest and increases towards the outside. move along z. To investigate the obtainable accuracy of the

Page 5 of 8
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

sensor, we then plot for each of the three measurements the

Movement in Y-direction (X = 0um, Z = 500um)
difference between the measured position and the mechanical
5000
reference position in µm. 4500
Fig. 12 shows the calculated X and Y position for a 4000 measured X
measured Y

measured distance [um]

displacement of ±2mm around zero on the x-axis. We notice 3500
3000
that the measured X value is very linear with the position and 2500
that the measured Y value is nearly constant. 2000
1500
1000

Movement in X-direction (Y = 3000um, Z = 500um) 500

0
4000
-500
1000 1500 2000 2500 3000 3500 4000 4500
3000
Reference distance Y [um]
measured distance [um]

2000
measured X
1000 measured Y Fig. 14 Position signals X and Y calculated from the Hall sensor
voltages for travel along y-axis
0

-1000 When we now again plot the errors between measured and
-2000
mechanical position (Fig. 15), we can see, that the accuracy of
the measured X position is about ±5 µm (±0.5%) and for Y
-3000
-2000 -1500 -1000 -500 0 500 1000 1500 2000 about ±10 µm (±1%) in the range from 1..3mm. For larger
Reference distance X [um]
distances both errors increase due to the strongly decreasing
field, causing the Hall voltages to become very small.
Fig. 12 Position signals X and Y calculated from the Hall sensor The experimental setup is optimized for a Y-position of 2..3
voltages for travel along x-axis
mm and shows up to Y = 3.8 mm an accuracy of 1% for
In Fig. 13 we give the calculated remaining error between the displacement along x.
measured values and the reference position of the translation
stages.
Measurement errors for movement in Y-direction
(X = 0um, Z = 500um)
40
Measurement errors for movement in X-direction
30 Xerr
(Y = 3000um, Z = 500um)
40 Yerr
20
Error signals [um]

30 Xerr 10
Yerr
20 0
Error signals [um]

10 -10

0 -20

-10 -30

-20 -40
1000 1500 2000 2500 3000 3500 4000 4500
-30
Reference distance Y [um]
-40
-2000 -1500 -1000 -500 0 500 1000 1500 2000
Reference distance X [um] Fig. 15 Position error between Hall sensor measurement and mechanical
reference for travel along y-axis
Fig. 13 Position error between Hall sensor measurement and mechanical The last case to investigate is whether the position sensor is
reference for travel along x-axis
really insensitive to a movement of the magnet along its axis (z)
The error on X is less than 10 µm (0.5% full scale) and on Y or not. Fig. 16 shows that for this case indeed the output
about 40 µm (2% full scale). However, tests with other magnet signals x-position and y-position stay virtually constant.
samples have shown that these errors arise from non-
uniformity in the permanent magnetic material and not from the
sensor inaccuracy.
In the next, step we investigate the output signals for a
movement along y-axis. Fig. 14 shows the position data
calculated from the sensor outputs. It can be noticed, similar to
the first case, that the calculated position for the movement (Y)
is very linear, whereas the other value (X) is nearly constant.

Page 6 of 8
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

single-axis, and even more for the two-axis displacement

Movement in Z-direction (Y = 3000um, X = 0um)
sensor, several ratios have to be calculated to obtain a signal
3500
proportional to the displacement value. This can be performed
3000
by a microcontroller or any other advanced digital signal
2500
measured distance [um]

treatment. Those components increase the system price and

measured X
2000
measured Y the arithmetic operations are often very time consuming.
1500
For this reason we have developed a system architecture
1000
which allows to perform the ratio of the two Hall voltages Vx
500
and Vy in an analog way already on the chip-level.
0
Fig. 18 shows a circuit schematic for a magnetic
-500
0 500 1000 1500 2000 2500 3000 3500
displacement sensor based on a two-axis Hall sensor which
Reference distance Z [um] directly outputs a voltage proportional to the ratio of the two
perpendicular magnetic field components.
Fig. 16 Position signals X and Y calculated from the Hall sensor
C1
voltages for travel along z-axis R3

For the z-position range, from z = 0.5..2mm, we find an x- R1

Vref
position error of = 10 µm, which corresponds to ±0.5% of R2 Op3

±1mm.and a y-position error of = 10 µm, which also

corresponds to ±0.5% of ±1mm (Fig. 17). I I
Here again, it can be observed that the amplitude of the Vy
Op1 H1 H2 Op2
sensor signals and so the signal quality decreases fast with Vout
increasing Z-distance.
Fig. 18 Electronic Circuit implemented for building the ratio between
Measurement errors for movement in Z-direction the two magnetic field components Bx and By.
(Y = 3000um, X = 0um)
40 The circuit contains a Hall element H1 for measuring the
30 Xerr component By and a Hall element H2 for measuring the
Yerr
20 component Bx. The output voltage of H1 is amplified by Op1
Error signals [um]

10 and the one of H2 by Op2. Op3 controls the bias current

0 through H1 and H2, in such a way, that
-10 VY = Vref for R1 = R2 (8)
-20
with Vref as a predefined reference voltage.
-30
On the other Hand, the Hall voltages are given as product of
-40
0 500 1000 1500 2000 2500 3000 3500 biasing current, sensitivity and magnetic field:
Reference distance Z [um] Vout = V X = S IX ⋅ I X ⋅ B X and VY = S IY ⋅ I Y ⋅ BY (9)
Since the two Hall elements are similar and biased in parallel
Fig. 17 Position error between Hall sensor measurement and mechanical
by the same voltage, we may write
reference for travel along z-axis
I X = I Y = I and S IX = S IY = S I (10)
From these measurements we conclude that indeed,
This finally leads us to an expression of the output voltage
according to our expectations, we can measure accurately and
Vout = Vref ⋅ B X BY (11)
independently x and y position of a small rotation symmetric
magnet using our position sensor. The distance between the The division of the measured components of the magnetic
magnet and the sensor plane has virtually no influence on the field takes place analogously because with this circuit the
position measurement as long as it does not exceed a value current flowing through H2 is proportional to the Hall voltage
where the field emanating from the magnet becomes very small. of H1. Therefore the circuit delivers an output signal Vout,
Reasonable values for the measurement range in the plane which is proportional to the ratio Bx/By.
seem to be from 0.5 to 1.5 times the magnet radius in Y and The feed-back loop for the supply current control is
from –1.5 to +1.5 times the magnet radius in X. Within this area implemented via a PI controller, so that the control loop is at
the field is strong enough to allow for a signal to noise ratio of the same time fast and stable. First experimental results of this
better than 100. architecture for the application as displacement sensor will be
available in early 2003.
VI. A NALOG INTEGRATED DISPLACEMENT SENSOR
Considering equations (3) and (6,7) we notice, that for the

Page 7 of 8
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

VII. SOME A PPLICATIONS The absolute accuracy can be increased by calibration. To

this end, we may simply use a look-up table in a micro
A. Contactless Adjustment Sensor for Machinery controller. By doing so, the accuracy will be limited by the
The magnetic displacement sensor is a very robust and system noise and offset instability.
cheap sensor to adjust parts of machinery in a plane without To increase the mutual alignment and the precision of the
requiring expensive and pollution-sensitive optical systems. distance A between the two two-axis Hall sensors, they can
B. Thickness Monitoring both be integrated on the same silicon chip. This becomes
possible with the technology of our new integrated CMOS Hall
In this application the magnetic displacement sensor
angular position sensor [4]. Moreover, by integration with
monitors the thickness of items like sheets or wires passing by
analog electronics, the division of the two Hall voltages can
during production. The advantage of using a magnetic device
already be performed on the chip, so that a linear displacement
here, is that it is very insensitive to dust and fibers and that it
signal is generated.
cannot be destroyed like mechanical sensors by a sudden
stroke.
REFERENCES
C. Long Range Incremental Displacement Sensor [1] European Patent Application EP1243898
Here the magnetic displacement sensor is used as part of a [2] For details on commercially available two -axis Hall sensors, see:
linear position encoder for displacement ranges much larger www.sentron.ch.
than sensor or magnet. A plurality of equally spaced sensors is [3] R. S. Popovic, “Non-plate-like Hall magnetic sensors and their
arranged in a line and a magnet is attached to the part moving applications”, Sensors and Actuators A85 (2000), 9-17.
parallel to the line. Our measurements showed that the signal
[4] R.S. Popovic, C. Schott, P.M. Drljaca, R. Racz, “A new CMOS Hall
of one sensor is very linear in a range of about 1.5 times the angular position sensor”, Technisches Messen, tm Vol 6, June
magnet diameter. This means that for a magnet of 6mm diameter 2001, Pages. 286-291
on sensor every 9mm is sufficient to obtain an accurate
displacement value over the entire sensor array. The decision
which sensor has the valid signal can be obtained by a very
simple comparator operation.
D. Vibration Sensor
A typical application for displacement sensors is to detect
slightly increased vibration of defective rotating machinery
before failure. For this application the displacement sensor can
be used as a small module in combination with a small magnet
which is mounted on a flexible suspension. Suspension and
magnet are matched to meet the requirements of vibration
amplitude and frequency. Such a solution can even be
imagined as a totally self contained micro-electro-mechanical
system (MEMS).

VIII. CONCLUSION A ND OUTLOOK

With the combination of two two-axis Hall sensors and a
single axially magnetized cylindrical magnet displacement in a
plane can be accurately measured. The displacement
information is virtually immune to the sensor and magnet
tolerances. From this fact we infer that the measurement result
is also immune on temperature and ageing effects.
Our demonstrator measures linear position with a linearity of
better than 1% and resolution of about 2µm in the position
range of > ±1mm.
The accuracy of the presented position sensor is limited by
the following factors: a rapid decrease in the magnetic field
with an increase in distance; irregular magnetization;
inaccuracy in the mutual position of the two 2-Axis -Hall
sensors; and offset and noise.

Page 8 of 8