IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO.

1, FEBRUARY 2007 25

A Fluxgate Magnetic Sensor: From PCB to

Micro-Integrated Technology
Andrea Baschirotto, Senior Member, IEEE, Enrico Dallago, Member, IEEE, Piero Malcovati, Senior Member, IEEE,
Marco Marchesi, and Giuseppe Venchi

Abstract—In this paper, a double-axis micro Fluxgate magnetic

sensor is presented. The device represents an evolution of a PCB
dual axis sensor previously realized. In the PCB version the ex-
perimental results exhibit excellent agreement with the simulations
performed using a tool based on the finite element method. Using
the same design approach a version of double-axis Fluxgate struc-
ture is here proposed to be realized in IC technology. The accurate
study of the magnetic field distribution allows a 75% area saving
for the IC version with respect to a direct scaling of the PCB ver-
sion. Imposing an external magnetic field of about 60 T, the sim-
ulated differential output voltage of the micro-integrated Fluxgate
sensor achieves a peak value of 1 mV with 5 mA sinusoidal excita-
tion current peak at 100 kHz. The integrated microstructure shows
a linearity error of about 1.15% of the full scale, in the range of
50 T, with a sensitivity of about 0.45 mV T.
Index Terms—Amorphous magnetic core, finite element method, Fig. 1. Schematic of the PCB double-axis Fluxgate realized.
magnetic sensors, planar fluxgate.

I. INTRODUCTION

HE Fluxgate structure [1], [2] appears to be the most fea-

T sible solution for realizing precise vector magnetic field
sensors to be used in Earth’s magnetic field measurements (i.e.,
with 60 full scale). Fluxgates are preferred with respect to
other magnetic sensors for their capability to operate in harsh
environments where endurance against magnetic, thermal and
mechanical shocks is important [3]–[5]. The main drawback of
Fluxgate magnetic sensors is the complex construction of the
core and the coils. In recent years, new topologies of planar in-
tegrated micro-fluxgate [6] have been presented, which allow
small dimensions and low power consumption to be achieved
together with a simplification of the integration process. When Fig. 2. Photograph of the double-axis planar Fluxgate magnetic sensor pro-
posed; the ferromagnetic material core is glued over the excitation coil with an
small dimensions and low power consumption are not vital, sim- adhesive.
ilar topologies can be fabricated in PCB technology [7], [8]. The
choice in favor of PCB technology derives primarily from its re-
duced cost and efficient fabrication. case, a study to foresee the characteristics of the micro-inte-
On the other hand, since low power consumption and small grated sensor can be very useful. Since the simulation results
dimensions are indeed the fundamental requirements for a mag- obtained with the software Flux3D [8] for the PCB structure
netic field sensor to be embedded in portable devices, a micro- showed a good matching with measurements, the same tool and
integrated version is mandatory for these applications. In this the same design approach have been used to develop a micro-
integrated version of the double-axis Fluxgate to be realized in
Manuscript received June 15, 2005; revised September 15, 2006. IC technology.
A. Baschirotto is with the Department of Innovation Engineering, University
of Lecce, 73100 Lecce, Italy (e-mail: [email protected]). II. BASIC SENSOR STRUCTURE
E. Dallago, P. Malcovati, M. Marchesi, and G. Venchi are with the Depart-
ment of Electrical Engineering, University of Pavia, 27100 Pavia, Italy (e-mail: The schematic of the PCB structure realized for the double-
[email protected]; [email protected]; [email protected]. axis Fluxgate is shown in Fig. 1. It consists of an excitation coil
it). and four sensing coils. A photo of the actual implementation is
Color versions of Figs. 1–4, 6–8, and 10–17 available online at http://ieeex-
plore.ieee.org. shown in Fig. 2. The planar excitation coil is characterized by
Digital Object Identifier 10.1109/TIM.2006.887218 30 thickness, 30 turns and 400 pitch. The current in
0018-9456/$20.00 © 2007 IEEE
26 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO. 1, FEBRUARY 2007

Fig. 3. Mesh view near the ferromagnetic core: to create a good mesh near the
core a box has been introduced to simplify the mesh generation.
Fig. 4. Simulated magnetic induction in the structure with a 600–mA excitation
current.

the excitation coil should be enough to saturate the ferromag-

netic material glued over the PCB (see Fig. 2). This material is
shaped as a cross in order to make the device sensitive to the
two components of a magnetic field coplanar with the PCB. For
each component of the magnetic field the output voltage is ob-
tained from two sensing coils placed in differential configura-
tion, having 17 thickness, 21 turns and 400 pitch. The
excitation and sensing coils are realized on two different metal
layers of the multilayer PCB structure, at a distance of 50
from each other. The total device size is 57.3 mm 58.1 mm.
The ferromagnetic material core is the amorphous alloy Vit-
rovac 6025 [9]. This material was chosen because of its ex-
tremely high relative permeability and low losses.
Its saturation magnetic induction is 0.55 T. Two types of Vit-
rovac 6025 (6025X, with a thickness of 25 mm, and 6025Z with
a thickness of 20 mm) were used. The difference between them
is essentially due to the shape of the hysteresis loop. In the case
of 6025Z, a suitable heat treatment (in the presence of an ex- Fig. 5. Comparison between the simulation and experimental result.
ternal magnetic field) allows the shape of the hysteresis loop to
be adjusted. This operation can be used to modify the peak of
the excitation current needed for the saturation of the ferromag- 600 mA is necessary to saturate the Vitrovac, as shown in
netic material. Fig. 4.
The core has been modeled in cross shape using a scissor; The second simulation, performed considering the geomet-
therefore, the orthogonality between the two axes of sensi- rical model proposed in Fig. 1 with a 10–kHz, 600–mA peak
tivity will be tightly correlated to the precision of the manual amplitude sinusoidal excitation current together with a field of
approach. 20 coplanar to the PCB and parallel to the ferromagnetic
material, was aimed to determine the transient response of the
III. PCB STRUCTURE device. The results of this transient magnetic simulation, for
one axis of sensitivity, are shown in Fig. 5, which reports also
A. Simulation Model the comparison with the experimental result obtained from the
The software based on the finite element method used for the prototype. The simulation shows a differential output voltage in
analysis is Flux3D [10]. The main issue in the setup of the model agreement with the Fluxgate principle. The output voltage ob-
was to create a good mesh after taking into account the geo- tained has 15–mV amplitude, which demonstrates the good de-
metrical configuration of the ferromagnetic material, which is a vice sensitivity.
cross sheet with dimensions of about
on the diagonal (Fig. 2). The critical dimension is the B. Experimental Results
thickness of only 25 , which leads to a large number of ele- Imposing an external magnetic field with a pair of Helmholtz
ments in the mesh (285985 volume elements). A particular view coils, the differential output voltage of the PCB structure has
of the mesh obtained is shown in Fig. 3. been evaluated. The comparison between the simulation and ex-
The first simulation performed was devoted to evaluate the perimental results, shown in Fig. 5, gives an error of about 10%
minimum peak value of the excitation current that guarantees on the amplitude of the output voltage. This error is extremely
the saturation of the ferromagnetic material. Using different low in consideration of the simulation conditions adopted to re-
magnetostatic analyses we verified that a peak current of about duce the computation time.
BASCHIROTTO et al.: FLUXGATE MAGNETIC SENSOR 27

Fig. 6. Transfer characteristic of the PCB Fluxgate obtained with a spectrum Fig. 7. Angle response for the two measurement axes (X and Y).
analyzer.

To characterize the PCB structure, the external magnetic

field has been varied in the range of . The differential
output voltage from the pick-up coils was analyzed with a
3562A Hewlett Packard Dynamic Signal Analyzer to extract
the value of the second-order harmonic (at double the frequency
of the excitation signal). This voltage is plotted in Fig. 6 against
the external magnetic field. The sensor shows a linearity error of
about 2.6% full scale in the range of with a sensitivity
of 0.37 .
To evaluate the possibility of using the sensor as a two–
dimensional (2–D) vector magnetic field detector, a suitable
read-out circuit has been realized. Different analyses have been
realized as reported in [11]. These analyses show that when the
Vitrovac 6025Z was glued over the sensors instead of the Vit-
rovac 6025X, the peak of excitation current required to saturate
the core was reduced to 450 mA. This lower value of excitation
current is mainly due to the different shape of the hysteresis
loop and the smaller thickness (20 ) of the ferromagnetic Fig. 8. Measured PSD of the proposed solution.
material 6025Z. This means that by using the Vitrovac 6025Z
the power consumption can be reduced. To further improve the
power consumption it is possible to 3.5 and a measured noise PSD of ,
• increase the number of turns per unit of area of the coils the noise PSD referred to the input magnetic field results to be
(actual pitch is 400 ); equal to 7.4 , thus, leading to a signal-to-noise ratio
• reduce the core thickness; of about 68 dB with the full scale Earth’s magnetic field for a
• perform a periodical measurement of the external magnetic 10–Hz bandwidth.
field (duty-cycle operation).
To characterize the sensors as a compass, the voltage output IV. INTEGRATED MICRO STRUCTURE
of the device was measured while rotating it in the horizontal Using the experience with the PCB prototype, an IC version
plane, as shown in Fig. 7. The angle error is smaller than 6 and of the double-axis Fluxgate sensor for low power applications
includes signal nonlinearity, hysteresis, and noise. This value of requiring small dimension has been developed. The design fea-
angle error is suspected to be mainly due to manual imprecision tures have been assumed in agreement with the IC realization
in cutting the ferromagnetic material into a cross shape as well of the two coils in a 0.5– CMOS technology with a postpro-
as to inaccuracy in the orientation of the sensor in the measure- cessing for the deposition of the ferromagnetic material.
ment setup. To establish the best geometrical dimensions for the IC struc-
Finally, Fig. 8 shows the measured output noise power ture, it was necessary to evaluate for each sensing coil the in-
spectral density (PSD) of the sensor evaluated by using the HP fluence on the induced voltage of the number of turns and the
3562A dynamic signal analyzer. Considering a sensitivity of active area (i.e., the flux linkage) of the coils as well as the fre-
about 1.25 , an overall gain in the read-out circuit of quency of the excitation current. According to the Faraday law,
28 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO. 1, FEBRUARY 2007

Fig. 10. Simulation model of the integrated version of the PCB structure.

Fig. 9. Frequency dependence of the relative magnetic permeability of the Vit-

rovac 6025.

the induced voltage for a coil wrapped around the core is given
by

(1)

where is the number of turns of the sensing coil, is

the number of turns of the excitation coil, S is the cross section
of the sensing coil, is the length of the excitation coil, is the
absolute magnetic permeability, and is the
sinusoidal excitation current at frequency . In the case of a Fig. 11. Distribution of the perpendicular component of the magnetic induction
planar coil, a scale factor could be introduced to adjust the for- in a plane coplanar with the sensing coil.
mulas in the particular case. Since all the parameters related to
the area in the IC version are scaled down, the desired sensitivity
should be achieved by In order to optimize the device area occupation, a number of
• increasing the excitation frequency ; magnetostatic analyses were performed by varying the area of
• increasing the number of turns of the sensing coil ; the sensing coil. The simulations showed that the component
• increasing the cross section of the ferromagnetic mate- of the magnetic induction associated with the ferromagnetic
rial (S). material perpendicular to the plane of the sensing coil is con-
However, the increase of the excitation frequency is limited centrated in a small area under each of the outer ends of the core
by the loss in the relative permeability of the material due to the branches (Fig. 11). Therefore, the sensing coils were centered in
eddy current effect. The frequency dependence of the relative these areas, and their size was reduced to the minimum possible.
magnetic permeability for Vitrovac 6025 is shown in Fig. 9 This gives a 75% area saving with respect to the direct scaling
where a cut-off frequency of 10 kHz can be observed. The of the PCB prototype, without any performance loss. The geo-
number of turns in the sensing coils can be increased according metrical parameters of the sensing coils based on the used tech-
to the minimum pitch of the used technology, but a limit on nology are 1 thickness, 1.4 width and 1.6 spacing.
the overall area exists. Finally, if a larger cross section of the The total area for the sensing coils, consisting of 75 turns, is
material is used, a higher current is required to saturate the .
ferromagnetic material and the power consumption increases. The Vitrovac 6025 has been introduced in the simulation in
The model created in Flux3D is shown in Fig. 10: it con- cross shape with the dimensions of on the
sists of an excitation coil and four sensing coils, a pair for each diagonal (Fig. 10).
axis of sensitivity. The planar excitation coil is characterized The first analysis step performed was aimed to evaluate the
by 5.5 thickness, 95 turns and 9 pitch (5 width minimum thickness for the ferromagnetic material that can be
and 4 spacing). The total area for the excitation coil is saturated with a peak current of about 5 mA (in order to limit
. the power consumption). Using different magnetostatic analyses
BASCHIROTTO et al.: FLUXGATE MAGNETIC SENSOR 29

Fig. 12. Simulated magnetic induction in the structure with a 5–mA peak of
m
excitation current and 1  core thick. Fig. 14. Photograph of the micro-integrated Fluxgate realized (with the ferro-
m
magnetic core of 1- thick).

Fig. 13. Differential output voltage simulated with Flux3D.

Fig. 15. Differential output voltage of the micro-integrated Fluxgate with the
we established that the complete saturation condition is obtained Earth’s magnetic field parallel to the sensor.
with a thickness of 1 , as shown in Fig. 12.
The second analysis was performed considering the geomet-
rical model proposed in Fig. 7 with a 100–kHz, 5–mA peak The characterization of the sensor has been realized using
amplitude sinusoidal excitation current together with a 60– a universal source (HP3245A) to supply the excitation coil.
magnetic field coplanar to the micro-integrated structure and Based on the unexpected worsening of the magnetic properties
parallel to the ferromagnetic material. The results of this tran- of the core, due to dc-magnetron sputtering process deposition,
sient magnetic simulation, for one axis of sensitivity, are shown to improve the performance of the sensor, the frequency of
in Fig. 13. The simulation shows a differential output voltage in excitation has been fixed to 100 kHz (instead of the 10 kHz
agreement with the Fluxgate principle. The output voltage ob- used in simulation), and the excitation current has been set
tained has an amplitude of about 1 mV, sufficiently large to be to 18 for the complete core saturation (instead of the
processed by the readout circuit. 5 mA used in simulation). The differential output voltage of the
sensor has been evaluated in the Earth magnetic field (about
20 ). The two cases of Earth magnetic field coplanar to the
V. MEASUREMENT AND RESULTS
sensor in the orthogonal and parallel direction, with respect to
Fig. 14 shows a photo of the micro-Fluxgate sensor realized one axis of sensitivity, have been evaluated. Fig. 15 shows
with the geometrical configuration specified earlier. the differential output voltage when the Earth’s field is in
30 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO. 1, FEBRUARY 2007

VI. CONCLUSION
A PCB double-axis Fluxgate magnetic sensor has been re-
alized and analyzed with a software tool based on the finite
element method. The simulations showed a good agreement
with the experimental results; therefore, the same software has
been used to evaluate the performance of a micro-integrated
version of the double-axis Fluxgate. The simulation results
predict that the micro-integrated sensor will have enough sen-
sitivity for sensing the Earth’s magnetic field, with a power
consumption two orders of magnitude lower than the PCB ver-
sion. The appropriate use of the simulation tools allows 75%
area saving for the IC version with respect to the direct scaling
of the PCB version, as well as a strong reduction of the power
consumption.
The performance of the integrated microstructure has been
evaluated, even if not comparable with the simulation results
because of the worst magnetic properties of the deposited core.
Fig. 16. Differential output voltage of the micro-integrated Fluxgate with the The experimental results have shown a linearity error of about
Earth’s magnetic field orthogonal to the sensor. 1.15% of the full scale in the range of with a sensitivity
of about 0.45 . At the moment an opportune front-end
circuitry is under construction, together with the evaluation of
possible postprocessing steps, to improve the magnetic prop-
erties of the deposited core (e.g., annealing the core after the
deposition).

ACKNOWLEDGMENT
The authors would like to thank STMicroelectronics,
Cornaredo Italy, for technological support, Cedrat, Grenoble
France, for allowing our use of their FEM simulator and Vacu-
umschmelze, Hanau Germany, for Vitrovac samples.

REFERENCES
[1] P. Ripka, Magnetic Sensors and Magnetometers. Boston, MA:
Artech House, 2001.
[2] M. Vopàlensky, P. Ripka, and A. Platil, “Precise magnetic sensors,”
Sens. Actuators A, vol. 106, pp. 38–42, 2003.
Fig. 17. Voltage output from the sensor imposing different value for the ex- [3] P. Ripka, “New directions in fluxgate sensors,” J. Magn. Magn. Mater.,
ternal magnetic field. pp. 215–216, 2000.
[4] F. Kaluza, A. Grüger, and H. Grüger, “New and future applications of
fluxgate sensors,” Sens. Actuators A, vol. 106, pp. 48–51, 2003.
[5] P. Ripka, “Advances in fluxgate sensors,” Sens. Actuators A, vol. 106,
parallel direction. The differential output voltage achieves a pp. 8–14, 2003.
[6] S. O. Choi, S. Kawahito, Y. Matsumoto, M. Ishida, and Y. Tadokoro,
peak-peak value of about 3.5 mV. By contrast, Fig. 16 shows “An integrated micro fluxgate magnetic sensor,” Sens. Actuators A, vol.
the differential output voltage when the Earth’s field is in 55, pp. 121–126, 1996.
orthogonal direction. As expected, the output voltage in this [7] P. Kejik, L. Chiesi, B. Janossy, and R. D. Popovic, “A new compact 2D
case is significantly lower than in the previous case although planar fluxgate sensor with amorphous metal core,” Sens. Actuators,
vol. 81, pp. 180–183, 2000.
not zero, because of the nonideality of both the device and the [8] A. Baschirotto, E. Dallago, P. Malcovati, M. Marchesi, and G.
experimental setup. Venchi, “Fluxgate magnetic sensor in PCB technology,” in Proc.
The fundamental component of the differential output Instrum. Meas. Technol. Conf. (IMTC-2004), Como, 2004, vol. 2,
voltage, after an amplification of about 33 dB, and imposing pp. 808–812.
[9] Vacuumschmelze. Hanau, Germany, Amorphous Metals VIT-
and external magnetic field with a pair of Helmholtz coils, has ROVAC [Online]. Available: www.vacuumschmelze.de
been then evaluated using a spectrum analyzer (HP3589A). [10] CEDRAT: Flux3D User’s guide Cedrat. Meylan, France, 2002 [On-
This voltage is plotted in Fig. 17 against the external magnetic line]. Available: www.cedrat.com
[11] A. Baschirotto, E. Dallago, P. Malcovati, M. Marchesi, and G.
field. The sensor shows a linearity error of about 1.15% of the
Venchi, “Precise vector-2D magnetic field sensor system for electronic
full scale in the range of with a sensitivity of about compass,” in Proc. IEEE Sens. 2004, Wien, Oct. 24–27, 2004, pp.
0.45 . 1028–1031.
BASCHIROTTO et al.: FLUXGATE MAGNETIC SENSOR 31

Andrea Baschirotto (M’95–SM’01) was born in Piero Malcovati (SM’06) was born in Milan, Italy
Legnago, Verona, Italy, in 1965. In 1989, he received in 1968. He received the “Laurea” degree (summa
the “Laurea” degreehe in electronic engineering cum laude) in electronic engineering from the Uni-
(summa cum laude) from the University of Pavia, versity of Pavia, Pavia, Italy, in 1991. He received the
Pavia, Italy. In 1994, he received the Ph.D. degree in Ph.D. degree in electrical engineering from the Phys-
electrical engineering from the University of Pavia. ical Electronics Laboratory (PEL), Federal Institute
In 1994, he joined the Department of Electronics, of Technology, Zurich (ETH Zurich), Switzerland, in
University of Pavia, as a Researcher (Assistant Pro- 1996.
fessor). In 1998, he joined the Department of Inno- From 1996 to 2001, he was an Assistant Professor
vation Engineering, University of Lecce, Italy, as an with the Department of Electrical Engineering of the
Associate Professor. Since 1989, he has collaborated University of Pavia. Since 2002, he has been an As-
with STMicroelectronics on the design of ASICs. He has been associated with sociate Professor of Electrical Measurements with the same institution. His re-
I.N.F.N. since 1991 on the design and realization of read-out channels for high- search activities are focused on microsensor interface circuits and high-perfor-
energy physics experiments and space experiments. He collaborated with SMIs mance data converters. He authored and coauthored more than 25 papers in inter-
for the design of mixed signals ASICS. He is also the Coordinator of a national national journals, more than 70 presentations at international conferences (with
project for the design of large-dynamic range gas sensors. His main research in- published proceedings), four book chapters, and three industrial patents.
terests are in the design of mixed analog/digital integrated circuits, in particular Dr. Malcovati was Guest Editor for the Journal of Analog Integrated Cir-
for low-power and/or high-speed signal processing. He has authored or coau- cuits and Signal Processing for the Special Issue on the IEEE ICECS 1999. He
thored more than 40 papers in international journals, more than 50 presenta- served as Special Session Chairman for the IEEE ICECS 2001 Conference and
tions at international conferences, three book chapters, and holds 10 industrial as Secretary of the Technical Program Committee for the ESSCIRC 2002 Con-
patents. In addition, he has coauthored more than 120 papers within research ference. He is a member of the scientific committees for several international
collaborations on high-energy physics experiments. conferences, including ESSCIRC, FDL, and DATE. He is an Associate Editor
Dr. Baschirotto was a Guest Editor for the IEEE TRANSACTIONS ON CIRCUITS of the Journal of Circuits, Systems, and Computers.
AND SYSTEMS — II for the Sspecial Issue on IEEE ISCAS 1998. He was an As-
sociate Editor of the IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS — PART
II during 2000–2003, and is now serving as an Associate Editor for the IEEE
TRANSACTIONS ON CIRCUITS AND SYSTEMS — PART I as an associate editor. Marco Marchesi was born in Castelsangiovanni
He has been the Technical Program Committee Chairman for ESSCIRC 2002 (PC), Italy, in 1975. He received the Dr. Eng.
and the Guest Editor for the IEEE JSSC for ESSCIRC 2003. He is a member of degree in electronics engineering in 2002 from the
the Technical Program Committee of ISSCC and ESSCIRC. University of Pavia, Pavia, Italy.
Currently, he is pursuing the Ph.D. degree in
electrical engineering at the same University. His re-
search activities are focused on magnetic sensors for
Enrico Dallago (M’87) was born in Bolzano, Italy, low magnetic field detection and relative electronic
in 1949. He received the Dr.Eng. degree in electrical circuitry.
engineering from the University of Pavia, Pavia, Italy,
in 1974.
Since 1975, he has been with the Department of
Electrical Engineering, University of Pavia, where
he is currently a Professor of Power Electronics.
He is founder and head of the Power Electronics Giuseppe Venchi was born in Pavia, Italy, in 1970.
Research Group. His current research activities He received the Dr.Eng. degree in electronics engi-
include circuit simulation, high-frequency switching neering in 1996 and the Ph.D. degree in electrical en-
power supplies, power integrated circuits, thermal gineering in 2000 from the University of Pavia, Italy.
analysis of electronic systems, magnetic sensors, and microelectromechanical He is currently a researcher with the Department
systems. of Electrical Engineering, University of Pavia. His
Dr. Dallago is a member of the Italian Electrical and Electronic Association research activities are device and packaging thermal
(AEI). analysis, circuit simulation, and development of inte-
grated smart-power circuits.