This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal

IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2020 – REV5.1.0 1

A Stray-Field-Immune Magnetic Displacement

Sensor with 1% Accuracy
Nicolas Dupré, Yves Bidaux, Olivier Dubrulle, Gael F. Close, Senior Member, IEEE

Abstract— We present a new Hall sensor design for the accurate and robust measurement of linear displacement. Implemented
in CMOS, the sensor is based on a novel gradient measurement concept combining Hall elements with integrated magnetic
concentrators. In typical applications with practical Ferrite magnets, the peak output voltage of the Hall transducers is
only around 1.7 mV at the maximum operating temperature of 160℃, and thus requires high-performance low-offset readout
electronics. Over its 15-mm linear displacement range, the sensor’s total error is 1% including manufacturing tolerances, trimming
accuracy, temperature, aging effects, and practical magnet constraints. In addition, the sensor is immune to magnetic stray
fields up to 5 mT, complying with the most stringent automotive norm.
Index Terms— Automotive electronics, Hall effect, Magnetic sensors

I. I NTRODUCTION present, limits the rejection [7].

This leaves us with point-like magnetic field sensors
A BOUT 6 billions semiconductor magnetic sensors
are shipped each year [1]. Magnetic sensors cover a
range of technologies: Hall, AMR, GMR, TMR, magneto-
based on Hall effect [8] or magnetoresistive effects. They
provide point-like field measurements, a proxy for the
mechanical position of the magnet. The magnetoresistive
inductive and others. They serve numerous applications
sensors are especially attractive thanks to their high
such as compass, position (angle, linear displacement),
intrinsic sensitivity [9]. They work directly in the angle
and current sensing. Automotive applications accounts for
domain, which is an advantage in angle sensing. However,
about half of the market [1]. Due to the electrification of
this complicates largely the rejection of stray fields [10].
vehicles, magnetic stray fields are increasingly present. For
The stray field rotates in a non-linear way the sensed angle,
example in [2], magnetic fields of several hundreds of µT
and this error cannot simply be canceled by differential
were measured in electric vehicles, and traced to traction
sensing.
currents. Such level of stray field would corrupt any
accurate magnetic sensor measurement if left unmitigated. We previously demonstrated an angle sensor immune to
We focus here on compact linear displacement sensor with stray fields [11] based on differential field sensing. This is
a full-stroke above 10 mm. This is by contrast to close not a complete immunity, just a rejection up to the degree
proximity sensors where the full-scale range is limited to specified by international standard [12]. Angle sensing play
a few millimeters, but with sub-nanometer resolution [3]. a major role in engine control, steering, and numerous
other applications (pedal, wipers, ...). Besides angle sens-
We target space-constrained applications requiring
ing, linear displacement sensors are also of interest for a
”PCB-less” point-like sensors. There is no physical space
variety of automotive applications, for example shift-by-
for a long distributed transducer covering the range of
wire and brake booster with physical space at a premium.
motion. This excludes magnetostrictive and magneto-
inductive sensors. With both of these technologies, the In this paper, we generalize our previous work by
magnet position is transduced into a localized change developing a compact Hall-based linear displacement sen-
of a material property. And this is sensed electrically sor suitable for operation in harsh automotive environ-
either as a change of pulse delay [4], or as a change of ment. We give an overview of our concept in section II.
impedance [5], [6]. The active transducer length is then Section III describes the electronic design. Section IV
commensurate with the range of motion: the magnet slides provides implementation details. The experimental results
along the transducer, and the transducer is distributed and are presented in section V, and discussed in section VI.
not point-like. For the same reason inductive sensors are
not suitable despite being robust to low-frequency stray II. M AGNETIC SENSING CONCEPT
fields by the virtue of their operation at several MHz. The A. Mechanical stroke and magnetic fields
nonlinearity of the B-H curve of the ferromagnet core, if
The magnetic concept is illustrated in Fig. 1. We
Manuscript submitted in February 2020. consider a general curvilinear displacement of a two-pole
N. D., Y. B. and G. F. C. are with Melexis, 2022 Be- magnet along an arc above a Hall sensor. In this specific
vaix, Switzerland (e-mail: [email protected], [email protected] and example, we have rarc = 20 mm for the arc radius and
[email protected]).
O. D. is with Melexis, 8900 Ieper, Belgium ∆θin = ±22° for the mechanical angle range. In terms
(email:[email protected]). of linear displacement d, the range is then drange =

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal
2 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2020 – REV5.1.0

Fig. 1. Magnetic concept. (a) Curvilinear motion of a two-pole magnet along an arc over the sensor. (b) Magnetic field components as
function of the displacement. (c) Gradient components of interest. (d) Resulting sensor transfer curve from the mechanical displacement to
the sensed electrical angle.

rarc · ∆θin ≈ ±7.5 mm. The magnetic field lines above

the sensor are arching in a U-shaped pattern. When the
sensor is centered under the axis of symmetry of the
magnet, the magnetic field lines are in the sensor plane,
and the field component Bx dominates. As the magnet
moves sideways, an out-of-plane component Bz arises.
The field components as a function of the magnet motion
are shown in Fig. 1b. These plots were calculated by a
magnetic simulation tool we discussed before [13]. In order
to reject any potential interfering stray field, the sensor
measures instead the gradient component ∂Bx /∂x and
∂Bz /∂x. They are plotted in Fig. 1c. Unlike in the classical
angle sensor case, the waveforms are not simple sine
waves. The electrical angle extracted from these two raw
components is then a non-linear function of the mechanical
angle. The overall transfer function of the sensor from the
linear displacement d to the electrical output angle θout is
shown in Fig. 1d. The 15-mm linear displacement range
is mapped, non linearly, to an electrical angular range Fig. 2. Geometry and magnetic field lines for a cube magnet.
of 300°,electrical . We use the suffix electrical to distinguish
this electrically-sensed angle from the mechanical angle
defining the position along the arc. The overall non-
linearity is a static error, and can be readily compensated |B|
B/∆B = √( ) ( z )2 . (1)
in the embedded software by a calibration. The residual ∂Bx 2
∂x + ∂B
∂x · 1 mm
dominant errors are due to thermal and lifetime drift.
Longer magnets naturally allow longer range. However,
these longer magnets increasingly generate a uniform field
B. Error model with limited gradient. In other words, for longer range
In order to gain insight into the trade-off, we derive the ratio B/∆B increases. At the center position, the
a simplified analytical model. Consider a Ferrite cubic common-mode field Bx increasingly dominates over the
magnet of length L = 10 mm with magnetization along signal gradient. In the presence of sensitivity mismatch
the x-axis (Br = 350 mT). Fig. 2 shows the magnetic field σSM M , the common-mode field leaks into the signal path,
pattern from such magnet as calculated by the Magpylib yielding a standard error equal to σSMM Bx . In addition,
library [14]. The displacement range drange is limited to there is also an offset error impacting each Hall element
about 1.5 times the magnet length: drange = 1.5 · L (HE) with standard deviation σOffset = 50 µT . Both error
to preserve some margin against the angle wrap-around sources are to be summed in root sum square (RSS)
behavior. Assuming an airgap of 5 mm, the field Bx at the fashion. The total error relative to the signal amplitude
sensor location is Bx ≈ 24 mT and its gradient ∂Bx /∂x ≈ yields the angular error. As there is a calibration in
6 mT/mm. Considering the characteristic dimension of the final module at room temperature, what matters
the chip, we can define the following dimensionless ratio is the drift after this calibration. Fig. 3 illustrates the
B/∆B: trade-off between range and accuracy. The angle drift is

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2020) 3

combination. For differential sensing, two disks, one on

each side of the sensor, are used. The left and right disk
measurements are then combined differentially to measure
∂Bx /∂x and ∂Bz /∂x. Fig 4a depicts the situation when
the magnet is centered. The field lines are along the x-
axis. This common-mode field component is rejected by
the differential sensing scheme. On the contrary, when
the magnet is off-centered close to the extremes of the
displacement range, the field lines are curved (Fig 4b),
and a differential signal appears.
By contrast: measure the gradient X. Han, Q. Cao,
and M. Wang. Long range, Hall measuring field in highly
uniform gradient to deduce linear position. Need special
coils arrangement, no just a cuboid magnet. Could be
good to constrast our solution. Specific Coil arrangement
Fig. 3. Trade-off curve. Angle drift as a function of the displacement unform field gradient to deduce the position. Here sim-
range assuming different sensitivity mismatch. ple cuboid magnets. Small ”Angle”, suitable for linear
dispalcement and curvilinear ratiometric measurement.
Differential fields to reject SF.

B. Electrical signal chain

The signal chain is based on a previously-discussed
generic platform [16]. Fig. 5 shows the block diagram for
this specific sensing mode. At the maximum temperature
160 ◦C, the magnet strength has dropped (temp.coFerrite =
−2000 ppm/◦C), the gradient is only 4.5 mT/mm. The
sensed Hall voltage sum at the input of the electronic
signal chain reaches about 1.7 mV. Offsets are mitigated
thanks to a combination of 4-phase current spinning,
effectively chopping the signal, and a fine correction in
software. The offset correction is calibrated at the factory
Fig. 4. Sensor cross-section through the two disks and magnetic field
lines: (a) when the magnet is centered above the sensor, (b) when
at multiple temperatures. The software performs the
the magnet is displaced sideways to the left such that the right disk correction based on real-time readout of the chip tem-
is under a pole. perature and interpolation. Residual offsets [17] are below
5 µV (standard deviation) at 160 ◦C after accelerated life
testing. The Hall voltages are summed in the current
plotted as a function of the displacement range for various domain after a transconductance amplification stage and
sensitivity mismatches. The feasible region in terms of appropriate sign inversion ak = ±1. The resulting signal
angle drift is bounded on the high side by the specification is then amplified, filtered and converted to digital codes.
limit, and on the low side by the technological limit. The rest of the signal chain is implemented by an
We assume a typical automotive specification limit of embedded software algorithm. It performs unchopping,
1% corresponding to 3°,electrical . In the plot, we assumed applies the factory-calibrated corrections, and calculates
that the ultimate technological limit is such that σSMM = the angle. The embedded software also performs closed-
0.25%. The practical realistic mismatch is higher. Some loop gain and centering adjustments to avoid saturation.
level of residual mismatch is expected given that the A final non-linearity (NL) correction is also included.
sensitive HEs have to be spaced at least 1 mm apart to The calibration of the NL correction is performed by
sense significant field differences. This is the nature of the final user in the final assembled module (consisting
gradient sensing: it requires spatial spreading, and hence of the sensor chip and the magnet). This is because
significant mismatches are to be expected. these calibration coefficients depend on the mechanical
and magnetic application details. The corrected angle is
III. E LECTRONIC DESIGN then transmitted as an analog voltage or as a digital frame
A. Differential sensing using standard automotive digital protocols.
To sense Bx and Bz , integrated magneto concentrator
(IMC) disks are post-processed on the chip surface. Each C. Hall plate biasing and readout
IMC disk (other shapes would also work) deflects the field Given that the sensitivity mismatches δSk between HEs
lines, thereby allowing the use of a pair of horizontal HEs are critical, special attention is needed to their biasing and
to measure Bx and Bz [15] depending on the sign of the readout. Fig. 6 shows the biasing and readout circuit. A

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal
4 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2020 – REV5.1.0

Fig. 5. Sensor block diagram.

IV. I MPLEMENTATION

The design is implemented in a CMOS 0.18-µm tech-

nology. The HEs are covered by IMC disks post-processed
at the wafer level [15]. Fig. 7 shows the chip micrograph
after IMC post-processing. The chip is a standalone smart
sensor. It includes a digital microcontroller running the
embedded software. The chip is automotive qualified, and
supports harsh environment. The operating temperature
can reach up to 160 ◦C, with absolute maximum voltage
rating of 28 V. The chip complies with stringent EMC and
ESD requirements calling for a robust power supply and
output driver.
Fig. 6. Hall element bias and readout circuit.

V. R ESULTS
reference current is generated by a reference bias voltage A. Drift of key parameters
imposed across a reference HE. This reference current
is then mirrored into the active sensing HEs, possibly About 100 production samples were subjected to
with digital fine tuning. The variability associated with AECQ-100 qualification tests to emulate the lifetime mis-
the resistance spread is then mitigated. Alternatively, the sion profile. The measured drift of the two key parameters
switch S0 when open, disables the current mirrors and is plotted in Fig. 8. The offset software correction, which
instead transistors M0 ...M3 operate as closed switches, was factory calibrated, remains effective even at the end
applying the maximum voltage to the HEs. of lifetime. The standard deviation of the residual offset
The Hall voltages are converted to currents by matched after embedded software correction at 160 ◦C was below
transconductance amplifier stages Gk , and summed in the 5 µV (by contrast: 20 µV without software correction). To
current domain. Note that the mismatch is dominated by characterize the sensitivity mismatches δSk , and capture
the HEs and not the amplifiers Gk . Consequently, dynamic how much common-mode leaks into the signal, we define
element matching techniques to swap the Gk stages would a common mode leakage ratio CMLR.
only provide a marginal improvement. As alluded before, ∑
some level of residual mismatch is naturally to be expected ∑k δSk ak
CMLR = . (2)
given that the sensitive HEs have to be spaced apart to k Sk
sense field difference. Over the relevant scale, chip bending
[18] and stress gradient contribute to residual mismatches. It can be viewed as the inverse of the traditional common-
mode rejection ration. Over the whole sample population
(N = 96) the drift of CMLR remained < 1%.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2020) 5

Fig. 9. Experimental validation. (a) Simulated angle drift curves

by Monte Carlo technique for 100 chips at 135°C. (b) Corresponding
simulated distribution. (c) Empirically measured angle drift curves of
Fig. 7. Chip photograph. The blue disks are the IMC disks that 6 sample chips at 135°C assembled in final application modules. The
have been micro-machined on top of the CMOS wafers. common horizontal red lines denotes the ±3° specification limits.

TABLE I
E RROR BUDGET FOR THE TOTAL STATIC ERROR

Error source Limits Distri. Divisor Std. err.

√
Residual NL after cal. ±0.2° Uniform 3 0.11°
Stray-field error @5mT ±0.8° Normal 3 0.27°
Drift1 ±3.0° Normal 3 1.00°
Total std. error 1.04°
Total error (3σ) 3.12°
1 Thermal and lifetime drift.

not ideal. Practically, there is a residual error of up to

Fig. 8. Empirical probability density functions extracted from ±0.2°. We assume, conservatively, that the probability
96 chips after 408 hours of operation at 175 ◦C to emulate the distribution is uniform. Second, there could be a stray field
lifetime (the y-axis units are such that the area under the curves of up to 5 mT [12]. Just like the common-mode field, it
are 1). (a) Input-referred offset voltage at 160 ◦C before and after
software correction. (b) Change, over temperature and lifetime, of leaks into the signal path via sensitivity mismatches. The
the common-mode leakage ratio of the HE sensitivities. The actual stray-field-induced error was experimentally characterized
96 measurements at 160 ◦C are showed as marks along the x-axis in a handful of samples. The error remained below 0.2°.
(rug plot).
To understand the probability distribution limits, we again
ran Monte Carlo simulations including mismatches. Error
B. Angle drift in application of up to ±0.8° were observed in simulation (3 sigma,
Normal distribution assumed as this effect is related to
Production samples have been characterized using mismatches). Table I summarizes the static error budget
multi-pole electromagnets with almost zero common-mode expressed in electrical degree. The errors are all standard-
field. Although this gives insight into the IC intrinsic ized, and combined in RSS fashion to obtain the final total
lab performance, this is not representative of practical static error of ±3.1° (Normal distribution, 3 sigma). This
applications. To account for the common-mode field, the is equivalent to 1% for the assumed range of 300°. For
offset and mismatch statistics measured above were re- dynamic application, noise should be included. The noise
injected into a Monte Carlo simulation including realistic is dominated by the thermal noise arising from the HE
magnetic stimulus. Fig. 9a depicts the simulated angle resistance ≈ 20 kΩ (corresponding to about 2 µV,rms in
drift curves at 135 ◦C for a configuration like in Fig. 1. a 10 kHz bandwidth) at the maximum temperature, and
The corresponding simulated probability density function not by the electronic readout chain. Noise adds an extra
is shown in Fig. 9b. The expected 3-sigma limit is about standard error of 0.15°,rms .
1%. To validate these simulations, 6 random samples were Table II compares our results against other Hall-based
assembled in a module with a 2-pole magnet like in Fig. 1. automotive linear displacement sensors. The first one is
Fig. 9c shows the measured angle drift at 135 ◦C. The angle another dual pixel sensor, but based on a combination of
drift on all 6 samples remained well below 1%. horizontal and vertical Hall devices. Historically, vertical
Hall devices have suffered from significantly lower sensitiv-
VI. D ISCUSSION ity and larger offsets [19], although state-of-the art devices
To finalize the error budget, additional secondary errors are narrowing the gap [20]. The larger offset degrades
need to be accounted for. First the user calibration is the accuracy at low magnetic field amplitude (like the

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2020.2998289, IEEE Sensors
Journal
6 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2020 – REV5.1.0

TABLE II [4] A. Affanni, A. Guerra, L. Dallagiovanna, and G. Chiorboli,

C OMPARISON BETWEEN THIS WORK AND OTHER H ALL - BASED “Design and characterization of magnetostrictive linear dis-
DISPLACEMENT AUTOMOTIVE SENSORS . placement sensors,” in Proc. of IEEE Instrumentation and
Measurement Technology Conf., 2004.
This work [22] [21] [5] M. Jagiella, S. Fericean, R. Droxler, and A. Dorneich, “New
Nr. of sensor chips 1 1 2 magneto-inductive sensing principle and its implementation in
Hall technology Horiz. and IMC Horiz. Vert. and horiz. sensors for industrial applications,” in Proc. of IEEE Sensors,
Bz offset drift 66 µT 500 µT 440 µT pp. 1020–1023 vol.2, Oct. 2004.
Range 15 mm n/a 30 mm [7] M. Martino, A. Danisi, R. Losito, A. Masi, and G. Spiezia,
Accuracy 1% n/a 1% “Design of a linear variable differential transformer with high
Stray-field imm. Yes Yes No rejection to external interfering magnetic field,” IEEE Trans.
Magn., vol. 46, pp. 674–677, Feb. 2010.
[8] R. S. Popovic, Hall Effect Devices, Second Edition. Taylor &
Francis, 2003.
field amplitude generated by a small Ferrite magnet). [9] L. Jogschies, D. Klaas, R. Kruppe, J. Rittinger, P. Taptimthong,
The solution in [21] uses two separate chips with just Bz A. Wienecke, L. Rissing, and M. C. Wurz, “Recent developments
sensing. Angle calculation is then deferred off-chip. This of magnetoresistive sensors for industrial applications,” Sensors,
vol. 15, pp. 28665–28689, Nov. 2015.
arrangement has the advantage of offering larger ranges,
[10] K. Rohrmann, P. Meier, M. Sandner, and M. Prochaska, “A
because the sensitive spots are spread much further on novel methodology for stray field insensitive xMR angular
two different chips. In addition to requiring two chips, position sensors,” in Proc. of IEEE Sensors, 2018.
this solution is not stray-field immune, as the field Bz is [11] N. Dupré, O. Dubrulle, S. Huber, J. W. Burssens, S. Christian,
measured directly. Hence any stray field along the z-axis is and G. F. Close, “Experimental demonstration of stray-field
immunity beyond 5 mT for an automotive-grade rotary position
interpreted as a legitimate signal, and there is no rejection sensor,” in Proc. Eurosensors, vol. 2, p. 763, 2018.
factor. [12] International Standards Organization, “ISO 11452-8 road vehi-
cles: part 8, immunity to magnetic fields,” June 2015.
VII. C ONCLUSIONS [13] Z. Lázár, Y. Bidaux, M. Roos, and G. F. Close, “Model-based
engineering of magnetic position sensors,” in Proc. Int. Conf.
A new concept for a linear displacement Hall sensor was Synthesis, Modeling, Analysis and Simulation Methods and
presented. It consists of two IMC disks, horizontal HEs Applications to Circuit Design (SMACD), June 2019.
measuring the left-vs-right change of magnetic field, the [14] M. Ortner and L. G. C. Bandeira, “Magpylib: Python package
for calculating magnetic fields of magnets, currents and mo-
signal conditioning, and the needed calculation algorithm. ments.” https://magpylib.readthedocs.io/en/latest/. Accessed:
The leakage of the common-mode field dominates the 2020-1-20.
error budget. A joint understanding of the on-chip and [15] V. Hiligsmann and P. Riendeau, “Monolithic 360 degrees rotary
application non-idealities is needed to quantify the real- position sensor IC,” in Proc. IEEE Sensors, pp. 1137–1142 vol.3,
Oct. 2004.
world performance.
[16] S. Leroy, S. Rigert, A. Laville, A. Ajbl, and G. F. Close,
We showed that production parts achieve 1% accuracy “Integrated Hall-based magnetic platform for position sensing,”
with practical simple cubic Ferrite 2-pole magnet. Lower in Proc. European Solid State Circuits Conf. (ESSCIRC),
errors are achievable with more elaborate magnet designs. pp. 360–363, 2017.
The sensor meets the most stringent automotive norms [17] P. Ruther, U. Schiller, W. Buesser, R. Janke, and O. Paul,
“Thermomagnetic residual offset in integrated hall plates,” in
for standalone sensors in harsh environment (in terms of Proc. IEEE Sensors, vol. 2, pp. 763–766 vol.2, 2002.
EMC, ESD, stray field, operating temperature ...). The [18] H. Husstedt, U. Ausserlechner, and M. Kaltenbacher, “In-situ
demonstrated stray-field immunity makes this solution analysis of deformation and mechanical stress of packaged
ready for the vehicle electrification wave. silicon dies with an array of Hall plates,” IEEE Sens. J., vol. 11,
pp. 2993–3000, Nov. 2011.
[19] J. Pascal, L. Hébrard, V. Frick, J.-B. Kammerer, and J.-P.
R EFERENCES Blondé, “Intrinsic limits of the sensitivity of CMOS integrated
[1] Y. de Charentenay, “Magnetic sensors market & technologies vertical Hall devices,” Sens. Actuators A Phys., vol. 152, pp. 21–
2017,” tech. rep., Yole, 2017. 28, May 2009.
[2] A. Vassilev, A. Ferber, C. Wehrmann, O. Pinaud, and A. R. [20] C. Sander, R. Raz, P. Ruther, O. Paul, T. Kaufmann,
Ruddle, “Magnetic field exposure assessment in electric vehi- M. Cornils, and M. C. Vecchi, “Fully symmetric vertical Hall
cles,” IEEE Trans. Electromagn. Compat., vol. 57, pp. 35–43, devices in CMOS technology,” in Proc. IEEE Sensors, pp. 1–4,
Feb. 2015. Nov. 2013.
[3] S. Nihtianov, “Measuring in the subnanometer range: capac-
[21] A. Foletto and A. Friedrich, “Analysis of a Hall-effect system
itive and eddy current nanodisplacement sensors,” IEEE Ind.
with two linear sensor ICs for 30 mm displacement,” tech. rep.,
Electron. Mag., vol. 8, pp. 6–15, Mar. 2014.
Allegro MicroSystems, 2013.
[6] B. Kaviraj and S. K. Ghatak, “Magnetic field and displace-
ment sensor based on giant magneto-impedance effect,” Mater. [22] ams, “AS5403A/D/E: 3D Hall position sensor for linear and
Manuf. Process., vol. 21, pp. 271–274, May 2006. off-axis applications,” 2015.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.