review of fluxgate sensors

Sensors and Actuators A, 33 (1992) 129- 141 129

Invited Review

Review of fluxgate sensors

Pave1 Ripka Czech Technical University, Electrotechnicai Faculty, Department of Measurement, Technicka 2, 166 27 Prague 6 (Czechoslovakia)

(Received Novemt~r 19, 1991; accepted February 25, 1992)

Abstract

Since the 193Os, fluxgate sensors have heen used for measuring d.c. magnetic fields up to I mT with a maximum resolution of 10 pT. In the sensor core the flux is gated by the excitation field. The preferable sensor geometry is a ring-core; both crystalline and amorphous ferromagnetic materials can be used for the core. Although a lot of fluxgate magnetometer types have appeared, the classical type with detection of the second harmonics by a phase-sensitive detector is the most popular. Fluxgate sensors are reliable and rugged and their applications range from space research to submarine detection.

Introduction

A fluxgate sensor is a solid-state device for measuring the magnitude and direction of the d.c. or low-frequency a.c. magnetic field in the range lo-” to lop4 T. The basic sensor configuration is shown in Fig. 1. The soft magnetic material of the sensor core is periodically saturated by the excitation field, which is produced by the excitation current I,,,. Hence the core permeability changes and the d.c. flux caused by the measured d.c. magnetic field Z& is modulated. A voltage Vind proportional to the measured field intensity is induced in the sensing (pick-up) coil at the second (and also higher) harmonics of the excitation frequency. The same principle was used in magnetic modulators, d.c. transformers, d.c. current transformers and magnetic amplifiers, but in all these cases the measured variable is a d.c. electrical current flowing through the primary coil.

sensors for submarine detection were developed during World War II. Fluxgate magnetometers were used for geophysical prospecting, airborne field mapping and later for space applications: since Sputnik 3 in 1958, hundreds of fluxgate magnetometers (most of them three-axis) have been launched. Since the 198Os, magnetic variation stations with a fluxgate supported by a proton magnetometer [7j have been replacing classical mechanical registration variometers for observing the changes of the Earth’s magnetic field [8]. Flux- gate compasses are extensively used for aircraft and vehicle navigation. FGrster [9] started to use the fluxgate principle for the non-destructive testing of ferromagnetic materials. Compact fluxgate magnetometers are used for detection and search operations.

A comprehensive bibliography of fluxgate papers was collected by Primdahl [l-3]. The first patent on the fluxgate sensor (applied for in 193 1) is credited to Thomas [4]. The subsequent development of a sensor that would be able to measure weak fields was very fast; in 1936 0.3 nT resolution was reported by Aschenbrenner and Goubau [5]. According to Geyger’s book [6], they had been working with fluxgate sensors since 1928. Sensitive

Vi.d

m,l

N

Fig. 1, The basic sensor configuration. The Sensor core is excited by an a.c. current I_, in the excitation winding so that the core permeability p(t) is modulated with twice the excitation frequency. BO is the measured d.c. magnetic field and B(t) the corresponding field in the sensor core. V,,, is the voltage induced in the pick-up (excitation) winding with N turns.

0924-4247/92/$5.00 @ 1992 ~ Elsevier Sequoia. All rights reserved

130

Fluxgate sensors are solid-state devices without any moving parts. They are reliable and rugged and they may have very low energy consumption. They can reach 10 pT resolution and 1 nT long- term stability; 100 pT resolution is standard in commercially produced devices.

This paper presents recent developments after the previous review articles by Gordon and Brown [lo] and Primdahl [ 111. The main emphasis is laid on the properties of the sensor core and the electronics that interface the sensor and improve its characteristics.

The most important source of information about the development of fluxgate sensors in Rus- sia is the book written by Kolachevski [ 121. The book was compiled from a number of research reports and internal papers that were not openly circulated. Therefore some of the Russian articles are cited indirectly through ref. 12.

Other types of magnetic-field sensors are described in the review by Lenz [ 131. A recently published book on magnetic sensors, especially the chapter covering applications [ 141, is another valu- able source of information.

Basic core configurations

Figure 1 shows the configuration of the most widely used parallel type of sensor, for which both the measured and the excitation fields have the same direction.

There is also another type called an orthogonal sensor, with the excitation field perpendicular to the sensitive axis of the sensor, which is identical to the ideal axis of the sensing coil. A miniature planar sensor described recently by Seitz [ 151 is of the latter type. Another orthogonal sensor made from permalloy film electrodeposited on a solid copper rod was reported by Gise and Yarbrough ]161.

A mixed orthogonal-parallel sensor was constructed by Schiinstedt from the helical core formed by a tape wound on a tube [ 171. The hairpin sensor developed by Nielsen et al. [ 181 uses the same principle, but the anisotropy of the sensor core is not caused by its shape or stress; the helical anisotropy is induced by annealing the tape under torsion.

This paper is devoted to the parallel type, which generally has better noise parameters.

The basic single-core design is used in simple devices such as the one described by Rabinovici et al. [ 191. This configuration is compulsory for magnetometers using time-domain detection, the devices described by Sonoda and Ueda [20] and Heinecke [21] being typical examples, and is often used for auto-oscillation or magnetic multivibrator sensors, e.g., [22]. Most magnetometers use the conventional method of evaluating the second harmonics of the output signal. In such a case many difficulties may arise from the presence of a large signal at the excitation frequency caused by the transformer effect between the excitation and sensing windings. A large part of this spurious signal is eliminated in a two-core sensor consisting of symmetrical halves excited in opposite directions so that the mutual inductance between the excitation and measuring coils is near zero.

Figure 2 shows the most widely used ring-core sensor in which the two half-cores are the parts of the closed magnetic circuit. The core usually consists of several turns of thin tape made of soft magnetic material. The ring-core sensor design was used as early as 1936 by Aschenbrenner and Goubau [5]. Sensors made from sheets in the shape of flat rings or race-tracks have also been reported [23]. The ring-core sensor geometry was found to be the best for low-noise sensors [lo] for several possible reasons: (1) it allows fine balan- cing of the core symmetry by rotating the core with respect to the sensing coil; (2) the possible tension in the core is uniformly distributed; (3) the open ends, usually associated with regions of increased noise, are absent.

The size of the core affects the sensor sensitivity, as analysed in ref. 24. Although the problem is complex due to the demagnetization effect and non-linearity, the sensitivity generally increases

pick-up coil

xxcitation coil

Fig. 2. Ring-core fluxgate sensor. When the measured field is non- zero, the half-core fluxes Q, , a)>. are not equal and the output voltage is induced.

131

with the sensor diameter. With a given diameter there is always an optimum of the other dimensions for maximum performance, but it must usually be found experimentally. The core dimensions and numbers of excitation and measuring turns

- also play a crucial role in matching the excitation and interfacing electronic circuits. A typical low- noise sensor size is a 25 mm diameter ring-core wound with 4 to 16 turns of 1 to 2 mm wide tape of 25 pm thickness. (a)

The principle of operation

The classical description of the fluxgate principle given in ref. 24 is based on the idealized magnetization characteristics B(H) of the sensor core and excitation field waveform H(t). Such an analysis for a single-core sensor was recently pre- sented in this journal by Seitz [ 151. The validity of such a semigraphical description was verified by Primdahl [25] observing the B-H curves on a two-core sensor. A similar description for a ring- core sensor was given in ref. 26.

A ring-core model including hysteresis is shown in Figs. 3 and 4. An idealized hysteresis loop of one half of the sensor (@, versus H,,,) is identical with the characteristics of the magnetic material as the magnetic circuit is closed. When the external measured d.c. magnetic field is present, the characteristic is distorted as shown in Fig. 3(b): for some critical value of He,,, that half of the core in which the excitation and measured fields have the same direction is saturated. At this moment the magnetic resistance of the circuit rapidly increases (as it is magnetically ‘broken’) and the effective permeability-of the other half-core decreases. In addition, the whole characteristic is shifted along the H axis. The characteristic for the second half-core is symmetrical with respect to the Q axis. By sum- ming up these two loops, we obtain the transfer function, i.e., the @ versus He,, characteristics shown in Fig. 4(a).

The height of the transfer function (which cor- responds to the peak-to-peak (p-p) change of the measuring coil flux) increases with the measured field. The mentioned dependence is linear up to high field intensities, for which the whole sensor becomes saturated. This principle has also been used for evaluation of the sensor output by inte- grating the induced voltage and measuring the

(W I Fig. 3. Magnetization characteristics of the half-core for the closed- core sensor. (a) Idealized hysteresis loop without external field. (b) Distorted loop in non-zero measured field: for critical field H,, one half of the core is saturated and the permeability of the other half rapidly decreases.

p-p value of the waveform obtained in this way. Figure 5 (from [ 261) shows the actual shape of the dynamic hysteresis loop and the transfer function measured at 1 kHz with an oval-shaped core of amorphous material.

If we know the excitation field waveform and transfer function, we may construct the flux waveform and by taking its derivative we obtain the waveform of the induced voltage. Further analysis by Fourier transformation may follow to calculate the sensor sensitivity, as was done by Burger [27] and recently by Nielsen et al. [28].

The basic analytical description may start from the Faraday law:

vi = d@/dt = d(NAp&t)H(t))/dt (1)

where Vi is the voltage induced in the measuring coil having N turns, @ is the magnetic flux in this coil, @ = BA as we neglect the air flux; A is the core cross-sectional area; H is the magnetic field in the sensor core and ,u(t) is the sensor core relative permeability.

132

ta

p/.:

H exe

(b) I

l-v-r t (4

Fig. 4. Derivation of the sensor output flux. (a) Excitation field waveform. (b) Transfer function (sensor flux 8 vs. H,,,), which is the sum of two half-core magnetization characteristics. (c) Total core flux.

Thus we may write the general equation for induction sensors:

v; = NA,u,,p dH(t)/dt + NhpH dA(t)/dt

+ NAp,H d&t)/dt (2)

Basic induction or ‘search’ coils are based on the first term of eqn. (2), the middle term describes rotating coil sensors where A(t) is the effective area in the plane perpendicular to the measured field and the last term is the basic fluxgate equation. The time dependence of the core permeability is caused by the excitation field.

In the design of a real sensor, the demagnetization effects cannot be neglected. Due to this, H is substantially lower than the measured field H,,, outside the sensor core. Therefore we must write for the flux density within the core

B = ,~&x,/[ 1 + D(P - 1)l (3)

where D is the effective demagnetizing factor. Thus the equation for the output voltage be-

comes more complex:

dB (1 -D) vi = NA dt = NAPoH~~ [1 + D(P _ f)]’ dP(t)/dt ’

(4)

(a)

1-i-I I I I I i-Y-1 (b)

Cc)

Fig. 5. Real sensor waveforms. (a) Dynamic hysteresis Transfer function. (c) Time dependences of excitation field, flux and Induced voltage (from [26]).

loop. (b) total core

The detailed analysis of the demagnetization effect is beyond the scope of the present paper. Readers are referred to [29] and the references listed there.

Non-selective detection methods

Although detection of the second harmonic component of the sensor output voltage, performed by means of a phase-sensitive detector usually preceded by a bandpass filter, is the most usual method, several other detection methods have appeared that process the output signal in the time domain.

The peak detection method is based on the fact that with increasing measured field, voltage peaks at the sensor output are increasing in one polarity and decreasing simultaneously in the opposite polarity. The difference between the positive and negative peak values is zero for the null field and may be linearly dependent on the measured field

133

within a narrow interval. Probably the best magnetometer based on this principle was constructed by Marshall in 1971 [30]. He used 4-79 MO permalloy ring core and reached 0.1 nT resolution with a portable instrument.

The principle of the phase-delay method is illus- trated in Fig. 6. We assume simple magnetization characteristics without hysteresis and a triangular waveform of the excitation current. For the time intervals t, , t2 between two succeeding output voltage pulses, we may write

t, = T/2 + 2 At = T/2 + THo/H,,, (5)

t2 = T/2 - 2 At = T/2 - THJH,,, (6)

where T is the period of the excitation current, H, is the excitation field maximum value and Ho is the measured field.

These relations were used by Heinecke in the digital magnetometer described in ref. 21. The time intervals t, , t2 were measured by a counter with the reference frequency n times higher than the excitation oscillator frequency. If the time (t,-tz) is equal to N periods of the reference oscillator, then we may write

t, - t2 = 2TH,,IH, = NT/n (7)

H,, = H,,, N/2n (8)

The resolution of this method is limited by the maximum counter frequency. Heinecke reached 2.5 nT resolution with a 10 MHz basic oscillator. The resolution was improved to 0.1 nT by sum- ming 100 time intervals, but this caused a remarkable limitation of sensor dynamic response as the

+

b!xc

(a)

Fig. 6. The principle of the phase-delay method. (a) Excitation field waveform with d.c. shift of Ho. (b) Sensor core magnetization characteristics. (c) Flux waveform. (d) Induced voltage.

excitation frequency was only 400 Hz. The author of the present paper reproduced Heinecke’s results, but complications with noise from fast digital signals of the counter made him almost sure that this method cannot replace classical ones, as was expected by the authors of ref. 31. A magnetometer working on a similar principle but with analog output was described by Rhodes [32].

Auto-oscillation magnetometers are considered as a separate group, although most of them are similar to the previously mentioned ones. The magnetic multivibrator constructed by Takeuchi and Harada [22] consists of a single-core sensor, capacitor and operational amplifier forming the oscillating circuit. The multivibrator duty cycle depends on the amplitude of the measured field. 0.1 nT resolution of this very simple device was reported. Another auto-oscillation magnetometer design has the oscillator frequency as the output variable.

A sampling method was used by Son [33]. The instantaneous value of the excitation current at the time of zero-crossing of the core induction depends on the measured field. In an ideal case the sensitivity is not dependent on the excitation frequency, amplitude or waveform. Son reached 0.1 nT resolution and 5 uV/nT sensitivity using two open 34 mm long cores made from amorphous water-quenched Vitrovac 6030. A similar principle was used by Sonoda and Ueda in their single-core field sensor [20].

None of the methods described in this Section has reached the parameters usual for low-noise and long-term stable magnetometers based on the conventional second-harmonic principle. Never- theless, they may find application in simple low- cost and low-power instruments used for various indication and search purposes.

Second-harmonic fluxgate magnetometers

In this Section a block diagram of a typical magnetometer working on the second-harmonic principle is described and the crucial parts of the magnetometer electronics are discussed.

As the sensor itself has a linear range limited to typically 100 nT, fields larger than this have to be compensated. Figure 7 shows the block diagram of the feedback magnetometer design. In this case, the analog feedback loop has such a large

134

I

@ ‘d

I -‘feedback coi I I

Fig. 7. Feedback magnetometer. The sensor core is excited by the generator G. The output voltage is fed through the bandpass filter BP to the phase-sensitive detector PSD. The output of the magnetometer is the current into the feedback coil. The digital current source Id increases the dynamic range of the instrument. The integrator in the feedback loop increases the loop gain for small static error.

amplification that the sensor works only as a zero indicator. The output variable is then the current of the compensation coil and the sensor non-linearity and the instability of its sensitivity are sup- pressed. (The term sensitivity is used here to mean the output signal per unit measured field.)

The sensor is excited by the generator working at the frequency J The generator circuits also produce the 2fsquarewave signal as a reference for the phase-sensitive detector PSD. A phase shift of this signal must also be performed. The pick-up or detection coil has typically 2000 turns. The first harmonic signal at the sensor output is usually too high for the PSD, so pre-amplification and filtra- tion have to be performed in the BP. To filter the a.c. component of the PSD output and to obtain sufficient amplification, an integrator is introduced into the feedback loop. The loop signal is fed back via the voltage-to-current amplifier (which may sometimes be replaced by a large resistor) to the feedback coil. An additional feedback current generated by the digitally controlled current source may be added to increase the dynamic range of the instrument. Thus the typical range of the basic analog feedback magnetometer is 1000 nT, while the digital electronics may increase the range of the whole instrument to about 100 pT, which makes it possible to measure variations of the Earth’s field with 0.1 nT resolution.

The excitation generator produces a sinewave or squarewave of frequency between 400 Hz and 100 kHz, 5 kHz being typical for crystalline materials. An increase of the frequency increases the sensitivity to the point where eddy currents become important. Thus sensors made from thin tape with large electrical resistance may be operated at higher

frequencies. Increasing the excitation frequency accelerates the dynamic performance of the sensor. 1 kHz bandpass of the sensor itself may be reached with a 10 kHz driving frequency, as shown in [34]. At high frequencies problems with electronic circuits, such as limited switching speed and slew rate or signal cross-talk, may appear.

The excitation current has to be free from any distortion by the second harmonics, as this may leak into the sensor output through the inductive coupling, caused by non-ideal balance of the sensor, or capacitance coupling, and cause a false signal. The amplitude of the excitation current must be large enough to deeply saturate the sensor core in each cycle in order to remove any rema- nent effect. As the excitation field is attenuated in the central part of the core by the eddy currents, the excitation current peak value has to be 10 to 100 times higher than that required for ‘technical’ saturation. This may be achieved by using resonant circuits matching the sensor impedance to the amplifier output. In the case of voltage output, the serial capacitor may be used for this purpose. It is primarily used for removing any d.c. component of the driving current (which may again cause a false signal through sensor residual imbalance), as described in [ 111. In the case of current output (or when the resistor or linear inductor is in series) the parallel tuning capacitor may be used according to Acuna [35]. A similar configuration was described by Berkman et al. [36].

Increasing the number of turns of the pick-up coil increases the sensor sensitivity, but sometimes there are limitations of the physical size of the coil, so an optimum must be found for the wire diameter and number of turns to match the coil impedance to the input amplifier. If a single coil is used as the measuring (pick-up) and feedback coils simultaneously, the input amplifier has to be cou- pled by the serial capacitor to prevent d.c. saturation. The feedback current source should be of large output impedance to prevent short-circuiting of the sensor output. The pick-up coil should be close to the sensor core to keep the air flux low. The feedback field should be homogeneous and therefore a large feedback coil is required. Thus in precise magnetometers the two coils are separate. But even in this case, the impedance of all the connected circuits must be kept high to prevent a decrease of the sensitivity, as the interaction between the two coils is high. On the contrary, the

135

mutual inductance between the excitation and the pick-up coil is very low, so the output impedance of the excitation generator is not critical.

The stability of the feedback coil size and shape is important for the constant sensitivity of the whole magnetometer. Materials with high dimensional stability, such as glass and ceramics, are used for supporting structures. For a single sensor a simple solenoid is mostly used. In the case of a three-axis magnetometer, a large three-axis feedback system is preferred to three separate coils. In this case mutually orthogonal Helmholtz coils are frequently used with both circular and rectangular shape. Primdahl and Jensen [37] have constructed a spherical coil system for rocket applications, where the size of the sensor is strictly limited. Each coil consists of nine sections approximating the ideal spherical coil, which generates a uniform field. Although the outer diameter of the coil system is just 90 mm, the homogeneity proved to be sufficiently high for 40 mm long sensors. The main advantage of keeping all three orthogonal sensors of the magnetometer in the centre of the three-dimensional feedback system is that the sensors are kept in a very low field. This is important for long-term stability and low-noise operation. The measuring axes are only defined by the feedback coil system and thus may be easily deter- mined and kept very stable [ 111.

The first stages of the input amplifier must be constructed from low-noise circuits as the typical sensor sensitivity is 20 uV/nT. Some designers use the parametric amplification at the sensor output. Although the resonating circuit is very simple (it may be just a single capacitor parallel to the sensor pick-up coil), the analytical description is complex and the effect is non-linear. Since the first study by Serson and Hannaford [38], parametric amplification and its stability were analysed by many authors, recently by Russel et al. [39], Gao and Russel [40] and Player [41]. Primdahl and Jensen [ 421 have demonstrated the disadvantages of parametric amplification: increased sensitivity on the sensor core parameters may degrade the temperature and long-term stability, and additional noise may be introduced. Nevertheless, a low-quality-factor resonating circuit at the sensor output causes moderate parametric amplification and may enhance the output signal without sig- nificant degradation of the sensor stability. Al- though parametric amplification is still often used,

especially in simple magnetometers, recent developments of low-noise operational amplifiers show that the electronic method of amplification is preferable.

Selective filter design must guarantee stable frequency characteristics not only in amplitude but also in phase, as the synchronous demodulator is phase sensitive (therefore it is often called a PSD). RC-active bandpass filters with very small sensitivity on the component value variation are preferred.

The phase-sensitive detector is usually of the switching type. A reference square-wave signal of twice the excitation frequency with very short transients is necessary. This signal may be generated in a frequency doubler implemented by a phase-sensitive loop. An easier approach is to use a crystal-controlled oscillator for a basic generator of the reference signal and to obtain the excitation frequency by simply dividing by two. This configuration allows even the phase delay betweeen excitation and reference to be generated digitally. On the other hand, it is difficult to generate a spec- trally pure excitation waveform in this case. For proper operation of the PSD, it is important to use fast switches and high slew-rate amplifiers to keep the dynamic distortion low. Second-harmonic distortion of the relatively large spurious signal at the excitation frequency may cause a false signal at the detector output. A low-pass filter at the output of the switching circuit is necessary for removing the higher-frequency products of mixing. The sensor output is amplitude modulated by the measured field and the PSD demodulates it back to d.c. or near-zero frequency.

Short-circuited fluxgate

In a conventional fluxgate magnetometer the output of the pick-up coil is connected to the amplifier with large input impedance so that the v&age induced into this coil forms the output of the sensor. Primdahl et al. [43] introduced another method of coupling: they short-circuited the pick- up coil by the current-to-voltage converter with very low input impedance and used the current- output mode of operation. The amplitude of the current pulses was shown to depend linearly on the measured field and to be theoretically indepen- dent of the excitation parameters (assuming that the amplitude of the excitation current is large

136

enough to completely saturate the sensor core). The low input impedance of the electronics elimi- nates problems with the stray capacitance of the coil and cable and the design of the low-noise input amplifier is simplified.

A fluxgate magnetometer working on this principle is described in ref. 28. Rather than direct evaluation of the p-p value by means of a peak detector (which is in general a noisy device), Nielsen et al. use the sampling circuit. Primdahl et al. [44] have shown that using the controlled rec- tifier of the switching type, the maximum sensitivity is achieved for a specific phase delay and width of the reference voltage when the shape of the pulse is best fitted. Because this principle uses the information from all the even-harmonic compo- nents (the weight of each being given by the spectrum of the reference), it is not possible to use a classical bandpass input filter. Switching filters with comb characteristics were shown not to be useful for fluxgate applications [45]. Therefore the demands on electronic circuits are very high in this case, as very small signals have to be processed in the presence of large overcoupled disturbing signals.

The sensitivity of the short-circuited fluxgate increases with increasing sensor length and cross section (the latter dependence saturates for thick sensors because of the demagnetization), as is usual for a voltage-output sensor, but decreases with an increasing number of turns. The sensitivity of the 17 mm toroidal core sensor was about 40 nA/nT, compared with about 20 uV/nT sensitivity that may be reached for an untuned voltage- output sensor of the same dimensions.

Although the short-circuited fluxgate sensors have shown a lot of advantages, with present knowledge it is not possible to decide if they are in general more advantageous than the traditional voltage-output design.

Marshall [46] has also used a short-circuited coil for measuring very short impulses with a conventional low-speed sensor. He stored the energy of the impulse in the shorted coil with a large time constant which was wound around the sensor. Thus the sensor does not measure the primary magnetic-field impulse, which may be of a dura- tion even shorter than one half period of the excitation frequency. This impulse is effectively compensated by the shorted coil and the magnetometer measures the overshoot field caused by the

slowly delaying short-circuited coil current. The integral value of field impulses caused by lightning flashes was measured by this ballistic method to investigate charge transfer. The principal difference from a short-circuited fluxgate sensor is that the latter uses a coil of very short time-constant to follow the signal on the harmonics of the excitation and measures the pick-up coil current directly, while Marshall measured the field inside the shorted coil by means of a conventional fluxgate magnetometer.

Noise and stability of the zero

The most important factor affecting the resoiu- tion of the fluxgate magnetometer is the stability of the sensor zero, as the offset may be nulled by turning the ring core with respect to the sensing winding. The changes of the offset may, for conve- nience, be divided up (although they may be partially caused by similar effects): noise as a relatively fast variation and the long-term instability of the sensor.

Sensor noise has been investigated by many authors: a remarkable study was performed by Scouten [47] and Burger [48]. It was demonstrated that the noise level may be decreased using the high peak value of the excitation current [49]. This was interpreted as if there existed small volumes inside the material which are more difficult to magnetize than the rest, so that they are not necessarily saturated during each period of the excitation field. The uncertainty of magnetization of these hypothetical regions may be one of the sources of sensor noise and offset.

Sensor noise depends mostly on the core material, but the geometrical aspect is also important, as the demagnetization factor determines how the material noise contributes to the effective noise value on the sensor input [29]. The dependence of the noise level on excitation frequency for permalloy cores was discussed by Afanasiev and Berk- man [50]. They have shown that the effective noise level decreases up to a certain limit for increasing frequency. This effect may also be fully explained by changing the sensor sensitivity.

Decrease of the sensor noise with increasing temperature was first observed for low Curie- point alloys [51]. With increasing temperature the core permeability increased and the saturation

magnetization decreased simultaneously. A similar effect was observed for low-noise permalloys. The effect of decreasing the room-temperature noise with the Curie temperature observed for amorphous alloys is mentioned in the next para- graph.

The techniques used for measuring the spectrum of the sensor noise are described by Snare and McPherron [52]. Measurement of the noise spectral density may be performed by fast Fourier transformation (FFT) on time-domain data. From the noise spectra the r.m.s. value of the noise in a given frequency band may be calculated. The peak-to-peak level, although more illustrative, is not a parameter suitable for comparison of the sensor noise levels. Primdahl et al. [44] have found that the p-p level is typically approximately six times higher than the r.m.s. value in the same frequency band. The typical noise power spectrum P(f) of the fluxgate has l/f character. Thus the noise level may also be expressed just by the power spectral density P( 1) at 1 Hz (or another given frequency), as P(f) = P( 1)/f (in units of nT’/Hz). The r.m.s. level of the noise, N,,,, in the frequency band from fL to fH is then given by the expression

137

f” _ . .

&, = (S p(f) d,>‘12 = (P(l)ln(.WW2 (9)

J-L

Figure 8(a) shows the noise spectrum for a sensor with a permalloy low-noise core produced by Infi- netics for NASA (type SlOOO-C3 l-JC-2239C). The measurement was performed by an HP 3566A digital spectrum analyser using overlapping aver- aging from 1000 samples to smooth the plot and remove the distortion caused by the time window. The spectral density was 3.8 pT/(Hz) ‘I2 and the r.m.s. value calculated from the measured spectrum was 8.76 pT (64 mHz - 10 Hz). The time plot of the last one of the 1000 (overlapping) 32 s time intervals is shown in Fig. 8(b). The r.m.s. value calculated from the estimated P( 1) using formula (9) is 8.81 pT, which is close to the measured value.

Zero offset and its changes may be partially caused by some of these factors:

(1) magnetically hard regions already mentioned;

(2) thermal and mechanical stresses; (3) inhomogeneities of the core and winding.

56.234

*

562.34

mx

Marker x: 1.03125 HZ Y: 3.6417 *

Start: 31.26 ml-iz PSD Ghan 1

1Hz Stop: 12.6 HZ RM6:looo

Real 10

,CK

PT 1 I I I I I I I I I I start: 0 a stop: 31.969 e

('9 Time Chen +

Fig. 8. Ring-core fluxgate sensor noise. (a) Power spectral density from 31 mHz to 12.5 Hz. P( 1 Hz) is 3.8 pT/Hz, r.m.s. value 8.76 pT (64 mHz- 10 Hz). (b) 30 s time plot of the sensor output.

138

In addition, there are contributions from changes of the magnetic properties of the core material, the parameters of the excitation field and temperature.

To increase the temperature and long-term stability of the sensor, it is useful to match the thermal expansion coefficient of all the sensor parts. The non-magnetic metal Incotel X-750 or machinable ceramics such as Macor are used for the bobbin.

Gordon and Brown [ lo] have measured zero offset changes less than + 50 pT during 24 h and a temperature shift less than 100 pT from -40 to + 70 “C using 81 MO permalloy for the sensor core. These results are probably the best ever reported. More recently, the long-term stability of magnetometers on magnetic observatories was compared during the geomagnetic observatory workshops [53,54]. A drift less than 1 nT/year may be achieved using presently available instruments in a thermostatted room. The temperature coefficient of the compensation field may be as low as 2 ppm/“C using quartz coil frames. This corre- sponds to z 0.1 nT/“C temperature dependence while measuring the Earth’s field.

For the measurement of the sensor noise and offset a place with very low magnetic field is required. Shielding of the ambient field (geomagnetic field of about 50 000 nT, its variations up to 500 nT and magnetic noise of human origin, rang- ing from 10 nT at magnetically silent locations to 1 mT in industrial environment) may be performed by various devices. Ferromagnetic shielding [55] is usually made in the form of a multi-layer cylinder of permalloy. The superconducting shield described by Brown [56] has an attenuation factor of the order of lop7 in the transverse and lop9 in the radial direction. Compensation of the field by means of a three-dimensional system of Helmholtz coils may also serve as the source of a magnetic vacuum [ 571.

The absolute value of the sensor offset may be measured by rotating the sensor through 180” in the low residual field of the shielding. A similar technique of flipping the sensor was used at the AMPTE Subsatellite [58] for in-flight offset calibration. Some satellites and rockets rotate with respect to their axis to stabilize the trajectory. This rotation modulates the component of the field perpendicular to the axis, and so it may be used for the same calibration purpose.

Sensor core material

It is difficult to discuss the selection of the core material generally, as it depends on the type and geometry of the sensor, the type of evaluation of the output signal and even the excitation parameters and measuring range. But still there are general rules for the material properties:

( 1) high permeability (permeability may be further reduced by thermomagnetic treatment);

(2) low coercive force; (3) non-rectangular shape of the magnetization

curve (points (2) and (3) are equivalent to the smallest possible area of the B-H loop);

(4) low magnetostriction; (5) low Barkhausen noise; (6) low number of structural imperfections and

low internal stresses; (7) smooth surface; (8) uniform cross section. From the literature already mentioned in a pre-

vious text, the sensor core material properties are discussed in refs. 11, 24, 47 and 48.

At present the material most frequently used for the sensor cores is permalloy in the form of a thin tape. Electrodeposited permalloy [ 15, 161 and ferrite [ 111 were also used for the sensor core. Other promising materials for sensor cores are magnetic glasses, the amorphous materials produced by rapid quenching.

Electrodeposition technology allows planar sensors of very small dimensions to be produced [ 151. Thin-film sensors made by vacuum deposition were reported by Hoffmann [59]. The only advantage of the ferrite core is that it allows the use of a very high excitation frequency, which may be important for magnetometers measuring very weak short impulses.

The 81.6 Ni 6 MO permalloy developed by the Naval Ordnance Laboratory [60] for a low-noise NASA magnetometer is still the superior material with the lowest known noise. Afanasiev and Gorobej [61] reported 12 pT p-p noise (1 mHz- 1 Hz) using 81 NMA permalloy of Soviet produc- tion. The author of the present article tried to use the same material in a race-track sensor, but the measured noise level was 36 pT r.m.s. at the same frequency band, a higher level than that of industrial-standard 79 Ni permalloy.

Amorphous magnetic materials started to be used for fluxgate cores from the early 1980s. The

139

properties of these alloys as sensor materials were discussed by Mohri [62]. A study concerning the noise of the amorphous magnetic materials was performed by Shirae [63]. He found that:

(1) low magnetostriction Co-based alloys are suitable for fluxgate applications;

(2) room-temperature noise decreases with Curie temperature;

(3) annealing of the tape may further decrease the noise level.

Another study was performed by Narod et al. [64]. They tested a number of alloys and observed the llfcharacteristics of the noise spectrum as was already known for crystalline materials. The mini- mum noise level for annealed Co-based alloy was 10e4 nT*/Hz at 1 Hz. The measured sensor was a 23 mm ring-core type. Nielsen et al. have used stress annealing of the tape to reduce the sensor noise [65]. For a 17 mm diameter ring-core sensor the lowest noise level was 11 pT r.m.s. (64 mHz- 12 Hz), corresponding to 1.8 x lop5 nT2/Hz at 1 Hz [44].

Applications

Fluxgates have advantages over other types of field sensors in a certain area of measured field intensities and frequencies.

Fluxgate sensors are about five orders of magnitude more sensitive than other solid-state devices like magnetoresistors and Hall-effect sensors. Un- like the induction magnetometers, which register field changes only, fluxgate sensors may be used for absolute measurements. Fluxgates also measure the direction of the field, compared with the scalar character of nuclear resonance magnetometers, which are in general better in long-term stability. The fluxgate sensors have no moving parts (unlike rotation coil magnetometers) and they are more sensitive to the measured field and less sensitive to vibrations and thermal changes compared to optical-fibre magnetometers. Fluxgate sensors are much cheaper than (more sensitive) SQUIDS and need no liquid helium.

The main fields of application are: geophysical measurements; space research; identification, location and compasses; and measurement of electrical current.

Three-axis fluxgate magnetometers are widely used for monitoring variations of the Earth’s field

(often together with a proton magnetometer) at magnetic observatories [53, 541, or remote locations [66, 671. Portable instruments are used for field and airborne measurements of local magnetic-field anomalies in mineral prospecting. Ap- plications in archeology are described in [68].

Fluxgate sensors are used for the measurement of rock magnetism [69], as the null sensor in coercivity measurement instruments and in nondestructive material testing [ 51.

Applications in space research have been described elsewhere [ 10, 35, 581. The Giotto experi- ment [70] and Magsat mission [ 711 were the most important of the recent fluxgate magnetometer launches.

The Czechoslovak satellite MAGION, which was launched during the Interkosmos program, had a three-axis fluxgate magnetometer and one- axis variometer on board, both developed in Ro- mania by Cobanu [72]. The variometer range was + 154.6 nT, resolution 50 pT; the three-axis magnetometer ranges were 50 048 nT (resolution 16 nT) and 6256 nT (resolution 2 nT). The power consumption of the whole system was 1 W.

Other applications include sensing of the magnetic ink [ 151 and magnetic marks on steel ropes [73], location of ferromagnetic bodies [74], detection of submarines and vehicles [75] and missile navigation. Fluxgate sensors are extensively used in compasses for automobiles [76] and aircraft.

A fluxgate sensor was also used for indirect measurement of electrical currents in pipelines [77]. D.c. magnetic current comparators, which work on the same principle, are used for precise measurement of large d.c. currents and as stan- dards of the current ratio, with an error of the order of lop7 [78].

Acknowledgements

The author would like to thank 0. V. Nielsen of the Department of Electrophysics, The Technical University of Denmark and F. Primdahl of the Danish Space Research Institute for providing their knowledge and their collection of literature, and also for suggestions concerning this review. Other thanks to K. Zaveta of the Institute of Physics, Czech Academy of Sciences, for comments on the manuscript.

140

References 26 P. Ripka, Contribution to the ring-core fluxgate theory, Phys. Ser., 40 (1989) 544547.

I F. Primdahl, Bibliography of Fluxgate Magnetometers, Publica- tions of the Earth Physics Branch, Vol. 41, 1970. No. 1.

2 P. H. Serson and F. Primdahl, Bibliography of Magnetometers, Publication of the Earth Physics Branch, Vol. 43, 1912, No. 8.

3 F. Primdahl and R. L. Coles, A bibliography of magnetometers, Appendix I, Proc. Int. Workshop on Magnetic Observatory Instru- ments, Ottawa, Canada, July 30-Aug. 9, 1986.

4 H. P. Thomas, Direction responsive system, US Patenr No.

27

28

29

J. R. Burger, Theoretical output of a ring core fluxgate sensor, IEEE Trans. Magn., MAG-8 (1972) 791-796. 0. V. Nielsen, J. R. Petersen, A. Fernandez, B. Hernando, P. Spisak, F. Primdahl and N. Moser, Analysis of a fluxgate magnetometer based on metallic glass sensor, Meas. Sci. Technol., 2 (1991) 435-440. F. Primdahl, B. Hernando, 0. V. Nielsen and J. R. Petersen. Demagnetising factor and noise in the fluxgate ring-core sensor, J. Phys. E: Ser. Instrum., 22 (1989) 1004- 1008. S. V. Marshall, A gamma-level portable ring-core magnetometer, IEEE Trans. Magn., MAC-7 (1971) 1X3-185. W. Bornhofft and G. Trenkler, in R. Boll and K. J. Overshott (eds.), Magnetic Sensors, Sensors, Vol. 2, VCH, Veiden, Germany, 1989, Ch. 5, pp. 1533205. M. H. Rhodes, Magnetic field detection by differential phase lag, US Patent No. 4 321 536 (March 1982). D. Son, A new type of fluxgate magnetometer using apparent coercive field strength measurement, fEEE Trans. Magn., MAG-25 ( 1989) 3420-3422.

5

6

7

8

H. Aschenbrenner and G. Goubau, Eine Anordung zur Reg- istrierung rascher magnetischer Storungen, Hochfrequenztech. Elektroakust., 47 (1936) 177-181. W. A. Geyger, Magnetic-amplifier Circuits, McGraw-Hill, New York, 2nd edn., 1957, Ch. 17, p. 271. W. F. Stuart, Earth’s field magnetomctry, Rep. Progr. Phys., 35 (1972) 803-881. E. R. Niblett and R. L. Coles, Scientific requirements for geomagnetic observatory data, Proc. Int. Workshop on Magnetic Obserm- tory Instruments, Ottawa, Canada, July 30- Aug. 9, 1986, pp. 11-15.

9

10

11

12

13

F. Forster, A method for the measurement of d.c. field differences and its application to nondestructive testing, Non Destr. Test. (Guildford, Engl.), 13 (1955) 31-41. D. I. Gordon and R. E. Brown, Recent advances in fluxgate magnetometry, IEEE Trans. Magn., MAG-8 (1972) 76-82. F. Primdahl, The fluxgate magnetometer, J. Phys. E: Sci. Instrum., 12 (1979) 241-253. N. N. Kolachevski, Fhdctuacionnye processy v feromagnetikach (Fluctuation Processes in Ferromagnetic Materials), Nauka, Moscow, USSR, 1985. J. E. Lena, A review of magnetic sensors, Proc. IEEE, 78 (1990) 973-989.

14

15

16

17

18

19

20

21

22

23

M. R. J. Gibbs and P. T. Squire, in R. Boll and K. J. Overshott (ed.), Magnetic Sensors, Sensors, Vol. 2, VCH, Veiden, Germany, 1989, Ch. 12, pp. 477-485. T. Seitz. Fluxgate sensor in planar microtechnology, Sensors and Actuators, A21-A23 (1990) 799-802. P. E. Gise and R. B. Yarbrough, An improved cylindrical magnetometer sensor, ZEEE Trans. Magn., MAC-12 (1977) 1104-1106. E. 0. Schiinstedt, Saturable measuring device and magnetic core thereof, US Patent No. 2 916 696 (1959). 0. V. Nielsen, J. Gutierrez, B. Hernando and H. T. Savage, A new amorphous ribbon fluxgate sensor based on torsial-creep-induced anisotropy, IEEE Trans. Magn., MAG-26 (1990) 2766280. R. Rabinovici, 8. Z. Kaplan, Y. Elyakim and I. Abargil, Low power portable DC magnetometer, IEEE Trans. Magn., MAG-20 (1989) 3411-3413. T. Sonoda and R. Ueda, Distinctive features of magnetic field controlled type magnetic field sensor, IEEE Trans. Magn., MAG- 25 (1989) 3393-3395. W. Heinccke, Magnetfeldmessung mit Saturationskernsonden, Tech. Mess., 48 (1981) 6. S. Takeuchi and K. Harada, A resonant-type amorphous ribbon magnetometer driven by an operational amplifier, IEEE Trans. Magn., MAG-20 (1984) I7233 1725. P. Ripka, F. JireS and M. Mach&k, Fluxgate sensor with increased homogeneity, IEEE Trans. Magn., MAG-26 (1990) 2038-2041.

24 D. I. Gordon, R. H. Lundsten and R. A. Chiarodo, Factors affecting the sensitivity of gamma-level ring-core magnetometers, IEEE Trans. Magn., MAG-I (1965) 330-337.

25 F. Primdahl, The fluxgate mechanism, Part 1, IEEE Trans. Magn., MAG-6 (1970) 376-383.

2016977 (1935). 30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

P. Ripka, Improved fluxgate for compasses and position sensors, J. Magn. Magn. Mater., 83 (1990) 543- 544. M. H. Acuna, Fluxgate magnetometers for outer planet explo- ration, IEEE Trans. Magn., MAG-10 (1974) 519-523. R. Ja. Berkman and B. L. Bondaruk, Feroresonansnyj rezhim vozbuzhdenia magnitnych modulatorov i ferozondov (Ferroreso- nance mode of excitation of magnetic modulators and fluxgate sensors), Geojz. Appara. (50) (1972) 20. F. Primdahl and P. A. Jensen, Compact spherical coil for fluxgate magnetometer system, J. Phys. E: Sci. Inu~rum., 15( 1982) 221-226. P. H. Serson and W. L. W. Hannaford, A portable electrical magnetometer, Can. J. Tech., 34 (1956) 232-243. R. D. Russel, B. 8. Narod and F. Kollar, Characteristics of the capacitively loaded fluxgate sensor, IEEE Trans. Magn., MAG-19 (1983) 126-130. 2. CH. Gao and R. D. Russel, Fluxgate sensor theory: stability study, IEEE Trans. Geosci., GE-25 (1987) 124- 129. M. A. Player, Parametric amplification in fluxgate sensors, J. Phys. D; Appl. Phys., 21 (1988) 147331480. F. Primdahl and P. A. Jensen, Noise in the tuned fluxgate, J. Phys. E; Sci. Instrum., 20 (1987) 637-642. F. Primdahl, J. R. Petersen, C. Olin and K. H. Andersen, The short-circuited fluxgate output current, J. Phys. Ser. E; Sci. Instrum., 22 (1989) 349-354. F. Primdahl, P. Ripka, J. R. Petersen and 0. V. Nielsen, The short-circuited fluxgate sensitivity parameters, Meas. Sci. Technol., 2 (1991) 1039-1045.

45 S. Dada, S. PodolLk and P. Ripka, Synchronous filters for measurement technology, Pror. Em&on 1989 Conf., Prague, Czechoslovakia, July 4 6, 1989 (in Czech).

46. S. V. Marshall, Impulse response of a Ruxgate sensor-application to lightning discharge location and measurement, IEEE Trans. Magn., MAG-9 (1973) 235-238.

47 D. C. Scouten, Noise in low-level flux-gate magnetometers, IEEE Trans. Magn., MAG-8 (1972) 223-231.

48 J. R. Burger, Comments on [31], IEEE Tram. Magi?., MAG-8 (1972) 797.

49 R. Ja. Berkman, Sobstvennye schummy ferrozondov i metodika ich issledovania (Own noises of the fluxgate sensors and methods of their observation), GeojIz. Priborostr., (7) (1960) 25.

50 M. P. Afanasiev and R. Ja. Berkman, Vlianie konstruktivnych i elektricheskich parametrov na uroven sobstvennych schumov magnitnych modulatorov (Effect of constructional and electrical parameters on the magnetic modulator own noise level), Otbor i peredacha informacii, (42) ( 1974) 61.

141

51 A. S. Abramov, Experimentahtoe issledovanie nizkcchastotnych chumov magnitnych materialov v zavisimosti ot temperatury i rezhimov peremagnichivania (Experimental observation of the low-frequency noise of magnetic materials dependance on temperature and magnetisation regime, Candidare (Ph.D.) Thesis, cited in [la.

52 R. C. Snare and R. L. McPherron, Measurement of instrument noise spectra at frequencies below 1 Hertz, IEEE Trans. Magn., MAC-9 (1973) 232-235.

53 R. L. Coles (ed.), Proc. Inf. Workshop on Magnetic Observatory Instruments, Ottawa, Canada, July 30-Aug. 9, 1986, Geological Survey of Canada, paper 88 17.

54 K. Kauristie, C. Sucksdorff and H. Nevanlinna, Proc. lat. Work- shop on Geomagnetic Observatory Data Acquisition and Processing,

Nurm&zrvi, Finland, May 15-25, 1989, Finnish Meteorological Institute Geophysical Public.ation. No IS, 1990.

55 T. Rikitake, Magnetic and Electromagnetic Shielding, Terra, Tokyo, 1987.

56 R. E. Brown, Device for producing very low magnetic fields, Rev. Sci. Instrum., 39 (1968) 547-550.

57 K. Pfihoda, M. Krs, B. Pegina and J. Blaha, MAVCS-a new system for creating a non-magnetic environment for paleomag- netic studies, Geol. Ibrrica, 12 (1988-89) 223.

58 D. J. Southwood, W. A. C. Mier-Jedrzejowicz and C. T. Russel, The fluxgate magnetometer for the AMPTE UK subsatellite, IEEE Trans. Geosci., GE-23 (1985) 301-304.

59 G. R. Hoffman, Some factors affecting the noise in a thin-film inductance variation magnetometer, IEEE Trans. Magn., MAG-

17 (1981) 3367-3369. 60 D. I. Gordon, R. H. Lundsten, R. A. Chiarodo and H. H. Helms

Jr., A fluxgate sensor of high stability for low field magnetometry, IEEE Trans. Magn., MAC-4 (1968) 397-405.

61 Ju. V. Afanasiev and V. N. Gorobej, Schumy ferrozondov i magnitnych usilitelej (Noises of fluxgate sensors and magnetic amplifiers), Izmereniu, Kontrol, Automarizacia, (4) (1981) (38) 37.

62 K. Mohri, Review of recent advances in the field of amorphous sensors and transducers, IEEE Trans. Magn., MAC-20 (1984) 942-947.

63 K. Shirae, Noise in amorphous magnetic materials, IEEE Trans. Magn., MAC-20 (1984) 1299-1301.

64 B. B. Narod, J. R. Bennest, J. 0. Strom-Olsen, F. Nezil and R. A. Dunlap, An evaluation of the noise performance of Fe, Co, Si and B amorphous alloys in ring-core fluxgate magnetometers, Can. J. fhys., 63 (1985) l468- 1472.

65 0. V. Nielsen, J. R. Petersen, B. Hernando, J. Gutierrez and F. Primdahl, Metallic glasses for fluxgate applications, An. Fis., 886 (1990) 271-276.

66 L. V. Medford, C. G. MacLennan, P. E. Rosenfeld, L. J. Lanze- rotti and M. H. Acuna, Low-power portable geophysical data acquisition system and its use in geomagnetic measurements, IEEE Trans. Geophys., GE-19 (1981) 122-128.

67

68

69

A. White, A sea floor magnetometer for the continental shelf, Marine Geophys. Res., 4 (1979) lO5- 114. A. J. Clarc, Archeological geophysics in Britain, Geophys., 51

( 1986) 1404 I41 7. M. Kono, Y. Hamano, T. Nishitani and T. Tosha, A new spin- ner magnetometer, Geophys. J. R. Astron. Sot., 67 ( 1981) 217- 227.

70

71

72

73

74

75

F. M. Neubauer et al., The Giotto magnetic-field investigation, Space Sci. Rev., 23 (1988) 250. R. Langel, Cl. Ousley. J. Berbert, J. Murphy and M. Settle, The Magsat mission, Geophys. Res. Lett., 9 ( 1982) 243-245. M. Cobanu, Project “Aktivnyj” --Instruments SG-R6 and SG- R7, Tech. Rep., Institute of Nuclear Sciences, Bucharest, Roma- nia, 1987. D. C. Lewis and H. Ormondroyd, Magnetic striping of steel ropes, Mining Techno/., (June) ( 1978) 220-227. B. G. Steward, I. A. Rollo and J. F. L. Simmons, Magnitude determination of magnetic fields from three orthogonal measurements, J. Phys. E; Sci. Instrum., 22 ( 1989) 457-461. 1. F. Scarzello and G. W. Usher, A low power magnetometer for vehicle detection, IEEE Trans. Magn., MAC-13 (1977) 1101 1102.

76 T. J. Peters, Automobile navigation using a magnetic flux-gate compass, IEEE Trans. Veh. Technol., VT-35 (1986) 41-47.

77 W. H. Campbell and J. E. Zimmerman, Induced electric currents in the Alaska oil pipeline measured by gradient fluxgate and SQUID magnetometers, IEEE Trans. Geosci., GE-18 (1980) 244-

250.

78 N. L. Kusters and W. J. Moore, The compensated current com- parator, IEEE Trans. Instrum. Meas., IM-14 (1965) 178-190.

Biography

Pave1 Ripka received an Ing. degree in electro- technical engineering in 1983 and a CSc. degree (equivalent to Ph.D.) in 1989, both at the Electro- technical Faculty, Czech Technical University, Prague, Czechoslovakia. In 1990/91 he was visiting researcher at the Danish Technical University. At present he works as a researcher and lecturer at the Department of Measurement of the Electro- technical Faculty Czech Technical University. His current research interests cover magnetic materials measurements, fluxgate magnetometers and coher- ent methods of signal processing.

review of fluxgate sensors

Documents