ELSEVTER Sensors and Actuators A 68 ( 1998) 286-289 A PHYSICAL

Portable fhxgate magnetometer

Pave1 Ripka *, Petr Kaspar Czech Technical University, Electrotechnicul Facult): Department OfMeasurement, 166 27 Prague 6, Czech Republic

Abstract

A low-power three-axis fluxgate magnetometer has been developed for portable applications. The instrument has 300 mW consumption from a Jr 6 V source and a range of + 100 &T with O,Ol% linearity. The effective resolution is 1 nT and the response time to a large field step is 4 ms. Innovative sealed ring-core flnxgate sensors made of etched rings ensure high resistance against vibrations and mechanical shock. 0 1998 Elsevier Science S.A. All rights reserved.

Keywords: Fluxgate sensors; Ivlagnerometers

1. Introduction

Three-axis vector portable d.c. magnetometers with a full- scale range of about 100 p.T and 1 nT resolution are required for geophysical research, geological and archeological pros- pecting, material testing, car detection, search and navigation applications.

The fluxgate sensor is traditionally used for the above- mentioned applications, as since the 1930s until now it has been the most sensitive solid-state magnetic-field sensor [ 11. The most popular ring-core geometry is found to be advan- tageous for low-noise zero-stable sensors even if the large demagnetization factor reduces the field sensitivity. For the same reason the performance is dramatically deteriorated with decreasing sensor size; a bar shape is more advantageous for sensors with size below 8 mm. Fluxgates made in planar and CMOS technology were recently reported in Refs. [ 2,3].

The traditional fluxgate uses an analog synchronous detec- tor to measure the field-dependent second-harmonic com- ponent in induced voltage. A fluxgate magnetometer using short-circuited output current processed by a digital signal processing (DSP) board was described in Ref. [ 41. Current- output fluxgates require a low number of pick-up coil turns, which may be advantage in the case of miniature sensors. The present technology of digital signal processing gives a magnetometer noise level of 300 pTp-p, which is comparable to the noise of low-power analog magnetometers (the mag- netometer presented here has 200 pT p-p noise), while the

* Corresponding author. Tel.: + 422-243-52188; Fax: + 422-31 l-99-29; E-mail: ripka@feld.cvut.cz

power consumption of fast analog to digital converters ( ADCs) and DSP is considerably higher.

The most competitive sensors to fluxgates are anisotropic and giant magnetoresistance (GMR) magnetoresistors; they may have smaller power consumption, smaller sensor size and easier interfacing compared to fluxgates and they may also be much faster, but they still have larger noise, worse long-term offset stability and larger temperature coefficients of both offset and sensitivity IS]. The ideai instrument for scalar measurements with 1 nT absolute precision is the Over- hauser magnetometer. A portable instrument with 3 W power consumption is described in Ref. [6].

The design of portable fluxgate magnetometers is always a trade-off between power consumption and achieved noise and offset stability as the deep saturation ofthe magnetic core improves the performance of the fluxgate sensor; the low- noise and fast electronics also consumes more power [ 71.

2. The magnetometer design

The low-cost fluxgate sensors developed for this magne- tometer have a ring core made of low-magnetostriction, high- permeability molybdenum permalloy. The core consists of eight 50 pm thick rings with 18 mm inner and 22 mm outer diameter. The excitation winding has 150 turns of 0.28 mm copper wire. The ring is mounted inside the pick-up coil having 1500 turns. The sensor is used in classical voltage- output mode and is tuned to second harmonics by a parallel capacitor. The quality factor of the resonant circuit has to be artificially lowered as the undamped sensor is unstable due

P. Ripku, P. Kaspar/Sensors and Actuators A 68 (1998) 286-289 287

Fig. 1. Block diagram of the magnetometer.

to large parametric amplification (for related stability anal- ysis, see Ref. [ 81) .

Three orthogonally mounted single-axis fluxgate sensors form the sensor head; each axis has its individual feedback. The spherical feedback coil described in Ref. [9] is more precise, as all sensors work in magnetic vacuum and so the cross-field sensitivity is suppressed; the disadvantage of the spherical compensation coil is that it increases the magneto- meter dimensions and cost.

A block diagram of the magnetometer is shown in Fig. 1. The electronics consists of the timing, excitation and signal- processing circuits. The basic crystal oscillator and digital timing circuits generate 4 kHz excitation frequency and 8 kHz reference with digitally adjustable phase. The basic fre- quency was selected as a compromise: increasing the exci- tation frequency improves the frequency response, but the power consumption also increases. The use of higher working frequency also requires the use of faster and more power- consuming amplifiers. The excitation circuit consists of a totem-pole Hexfet transistor pair with 0.1 R on-resistance. The symmetry of the excitation waveform is important, as any second-harmonic component in the excitation current is coupled to the sensor output thropgh its imbalance and causes a false signal which is sensed as the sensor offset. The exci- tation circuit is tuned to resonance by a serial capacitor C,. This technique increases the amplitude of the excitation cur- rent peaks to 1 A p-p, whichgnsures the deep saturation of the core during each excitation cycle while the excitation power is kept low. The use-of a serial capacitor is rather unusual, as the classical configuration of the excitation tank is parallel; it was experimentally found as more suitable for voltage-output excitation power end, while the parallel tank requires more complicated current excitation.

The sensor signal iS a.c.;preamplified and directly fed (without any filtering) to a phase-sensitive detector made of CMOS switches. The follo&g integrator ensures the open-

0 100 200 3w 400 500

time (ps) Fig. 2. Output voltage of the detector for a measured field of 10 p,T and open feedback loop (lower trace) ; reference signal (upper trace).

loop gain at low frequencies and stabilizes the loop. The feedback is closed through the sensing resistor; the compen- sation current is fed back into the pick-up coil to null the measured d.c. magnetic field (see Fig. 2). The magnetometer output is derived from the feedback current so that the sta- bility of the sensitivity factor depends mainly on the dimen- sional stability of the pick-up/feedback coil. The a.c. open-loop gain at frequency f was adjusted to 7600/f which ensures loop stability and low dynamic error.

3. The magnetometer parameters

3.1. Linearity

The sensor was fixed in a 50 cm Helmholtz coil pair. A 10 000 nT field step was repetitively applied in the z-axis, while the basic d.c, field was changed by turning both Helm- holtz coils and sensor in the horizontal plane. An example of the magnetometer response is shown in Fig. 3. Differential sensitivity was calculated by averaging to eliminate noise and interference.

The achieved non-linearity error was + 0.01%.

3.2. Temperature coejjicient of offset

The measurements were made in 60 cm long, 18 cm inner diameter six-layer permalloy shielding in the temperature range - 13 to 75°C. The cylindrical fiberglass sensor cham- ber was heated by bifilar resistive wire supplied with a

5cml

4cw

3oMx) _ - . . _- -_

2m--- *oooo. --- _---

IIT 0 IOOOOIlT - - - _ _-~__.-- _

-10000~ ml -3cml

Fig. 3. Magnetometer response for the linearity test.

288 P. Ripka, P. Kaspar/Sensors and Actuators A 68 (1998) 286-289

40 kHz sinewave (up to 2 A) and cooled by a bifilarly wound plastic spiral with circulating liquid medium.

The estimated individual sensor temperature coefficients were 0.3, 1 and 0.3 nT ‘C-’ for the X, y and z axis, respec- tively. The temperature coefficient of the electronics is 0.2 nT “C-l. The achieved practical offset stability is 1 to 3 nT for several days.

3.3. Sensitivily

The sensitivity of individual sensors varies by 1%. This difference is mainly caused by uncertainty in the geometry of the pick-up coil caused by the low-cost technology used and it is software corrected. The temperature stability of the sensitivity was tested by heating the sensor was heated 65°C and monitoring the response to a +_ 10 000 nT0.05 Hz square- wave during cooling to room temperature. The measured variation of the response was between 20 005 and 19 957 nT in the 25-65°C range. Other sensitivity test were performed using a 7 Hz sinewave calibration field and DSP lock-in amplifier to suppress the magnetic noise in the laboratory. The most precise sensitivity calibration was performed with a 120 cm diameter Helmholz pair supplied from a d.c. cali- brator. Both the calibration current and magnetometer response were measured by 6 l/2 digit voltmeters; therepeat- ability achieved in a thermostatted room was 10 ppm. The typically achieved temperature constant of sensitivity was40 ppm ‘C--l, corresponding to the coefficient of linear thermal expansion of glass-filled epoxy, which is the constructional material for the pick-up coil support.

3.4. Orthogonality

A sensor orthogonality test was performed using a non- magnetic theodolite. The sensor head was rotated in a hori- zontal plane for three perpendicular positions. The orthogonality error was calculated after the corrections to individual sensor offsets. The measured errors were 0.5” between the x and y axes, and 0.6” and 0.3” for the x-z and y-z axes, respectively.

These errors are caused by machining and mounting imper- fections and also by non-perfect positioning of the sensor cores and their magnetic non-homogeneity. The orthogonai- ity error can be software corrected; the improvement of the long-term stability of such a correction from the present 0.1” to well below 0.01” is the subject of present efforts.

3.5. Remnnence

The sensor offset change after magnetic-field shock is

below 2 nT. The remanence of the heavily magnetized elec- tronics board is 20 nT at 5 cm distance. This indicates that the minimum distance between the sensor head and electron- ics board is 20 cm.

600

400

200

0 IO 20 30 40

time (ms)

Fig. 4. Magnetometer response to a field step of 20 @T.

50

3.6. Noise

Sensor noise was tested in six-layer permaIloy cylindrical shielding with a typical remanent field of 1 nT in the radial direction and 5 nT in the axial direction. The measured sensor noise density is 40 pTHz- *‘2 at 1 Hz, and the p-p noise value is 200 pT. The overall analog magnetometer noise is below 1 nT p-p.

3.7. Dynamic response

The dynamic response was tested in 25 cm Helmhoiz pair having only 20 turns to minimize the influence of the coil self-capacity. The large field step response is 4 ms to 90%; the 3 dB bandwidth is 100 Hz. The magnetometer response to a field step of 20 PT is shown in Fig. 4.

4. Conclusions

The described analog part of the magnetometer has 1 nT resolution and 100 Hz bandwidth at 100 000 nT range. The magnetometer analog output should be connected to a wide- band high-resolulion AJX (such as a sigma-delta converter) to use the high dynamic range of the instrument. The noise and thermal stability parameters are a result of trade-offs which were necessary to achieve the low cost and 300 mW power consumption.

Acknowledgements

This work was supported by the Grant Agency of the Czech Republic (grant no. 102/96/ 1251).

References

[l] P. Ripka, Review of fluxgate sensors, Sensors and Actuators A 33 (1992) 129-141.

[2j S. Kawahito, H. Satoh, M. Sutoh, Y. Tadokoro, High-resolution micro-ff lrxgate sensing eIements using closely coupled coil structures, Sensors and Actuators A 54 (1996) 612-617.

P. Ripka, P. Kaspar/ Sensors and Actuators A 68 (1998) 286-289 289

[31

[41

[51

I61

[71

t81

[91

B. Sauer, R. Gottfried, T. Haase, H. Kuck, CMOS-compatible inte- gration of thin ferromagnetic films, Sensors and Actuators A 41-42 ( 1994) 582-584. J. Piil-Hendriksen, J.M.G. Merayo, O.V. Nielsen, H. Petersen, J.R. Petersen, F. Primdahl, Digital detection and feedback fluxgate mag- netometer, Meas. Sci. Technol. 7 (1996) 897-903. P. Ripka, Noise and stability of magnetic sensors, J. MagnetismMag. Mater. 157 ( 1986) 424-427. D. Duret, J. Bonzom, M. Brochier, M. Frances, L.M. Leger, R. Odru, C. Salvi, T. Thomas, Overhauser magnetometerforthe Danish Oersted Satellite, IEEE Trans. Magn. 3 1 ( 1995) 3 197-3 199. P. Ripka, Magnetic sensors for industrial and field applications, Sen- sors and Actuators A 41-42 (1994) 394-397. Z.C. Gao, R.D. Russel, Fluxgate sensor theory, IEEE Trans. Geosci. Remote Sensing 25 ( 1987) 862-870. P. Ripka, F. Primdahl, O.V. Nielsen, J.R. Petersen, A. Ranta, A.c. magnetic field measurement using the fluxgate, Sensors and Actuators A46-47 (1995) 307-311.

Biographies

Pavel R&AZ received an Ing. degree in 1984, a C.Sc. (equiv- alent to Ph.D.) in 1989 and a Dot. degree in 1996 at the Czech Technical University, Prague, Czech Republic. He

works at the Department of Measurement, Faculty of Elec- trical Engineering, Czech Technical University, as a lecturer, teaching courses in electrical measurements and instrumen- tation, sensors and transducers, contactless measurements and engineering magnetism. In 1992 and 1993 he was a vis- iting researcher at the Danish Technical University. His main research interests are magnetic sensors, especially fluxgate and magnetic measurements. He is a member of Electra Soci- ety, Czech National IMEKO Committee, CzechMetrological Society and the Eurosensors Steering Committee.

Petr Kaspar received an Ing. degree in 1980 and a CSc. (equivalent to Ph.D.) in 1985 at the Czech Technical Uni- versity, Prague, Czech Republic. He works at the Department of Measurement, Faculty of Electrical Engineering, Czech Technical University, as a lecturer, teaching courses in elec- trical measurements and instrumentation, magnetic measure- ment and engineering magnetism. His main research interests are magnetic measurements. He is a member of Czech National IMEKO Committee.