Fluxgate magnetometers are important tools for geophysics and space physics,
providing high-precision magnetic field measurements. Fluxgate magnetometer
noise performance is typically limited by a ferromagnetic element that is
periodically forced into magnetic saturation to modulate, or gate, the local
magnetic field. The parameters that control the intrinsic magnetic noise of
the ferromagnetic element remain poorly understood. Much of the basic
research into producing low-noise fluxgate sensors was completed in the
1960s for military purposes and was never publicly released. Many modern
fluxgates depend on legacy Infinetics S1000 ring cores that have been out of
production since 1996 and for which there is no published manufacturing
process. We present a manufacturing approach that can consistently produce
fluxgate ring cores with a noise of
Fluxgate magnetometers (e.g., Primdahl, 1979) are important tools for geophysics, solar–terrestrial and space physics, space exploration, and monitoring space weather. They provide high-precision measurements of Earth's magnetic field that can be used to image downward into the Earth, resolving subsurface conductivity via magnetotellurics, and upward into near-Earth space, inferring the currents and waves coupling the ionosphere to the magnetosphere. Fluxgate magnetometers deliver a magnetic field measurement by modulating (gating) the local magnetic field by periodically saturating a piece of ferromagnetic core material – often in the form of a ring (Fig. 1). The ferromagnetic core alone would act as a magnetic-flux concentrator, but combined with a toroidal drive winding to periodically drive it into magnetic saturation it acts as a magnetic-flux modulator – or a fluxgate. The addition of a solenoidal sense winding completes the fluxgate sensor as the, now modulated, magnetic field induces a current or voltage that can be conditioned and digitized.
A fluxgate sensor has three primary components: a ferromagnetic core, a drive winding to periodically force the ferromagnetic core into magnetic saturation, and a sense winding to pick up the modulated (gated) field. Reproduced from Miles et al. (2017).
The instrumental noise floor of the sensor is typically limited by the intrinsic magnetic noise of the ring core as it is driven in and out of magnetic saturation. Despite their widespread use, the parameters that control the limiting intrinsic magnetic noise of a fluxgate ring core remain poorly understood. The key research and manufacturing process for low-noise fluxgate ring cores are insufficiently documented (Narod, 2014) for comparable ring cores to be reproduced. Here we document a ring-core manufacturing process that has been developed from the limited historical documentation that exists, much of which dates to the 1960s. This process yields fluxgate ring cores comparable to those produced historically and serves as a baseline for further low-noise fluxgate ring-core development.
A preferred ferromagnetic material used in fluxgate sensors is 6-81 permalloy containing 6 % molybdenum, 81.3 % nickel, and the remainder iron. The 6-81 permalloy is visible in the ring core in Fig. 2 as the glossy grey metal within the black supporting bobbin. The earliest known reference to 6-81 permalloy is by Odani (1964), who examined magnetic properties for 5.3 %–6.8 % molybdenum permalloy processed into thin foils and heat-treated. Pfeifer (1966) undertook similar but wider-ranging work that was introduced to the English language literature by English and Chin (1967). Pfeifer and Boll (1969) explored the magnetic properties of similar alloys for applications such as transformers and magnetic amplifiers. The US Naval Ordinance Laboratory was aware of the utility of 6-81 permalloy as early at 1965 (Scanlon, 1966); however, few details of their research are available to the public.
An S1000-compatible fluxgate ring core. The red enameled wire forming the drive winding has been partially removed for the photograph, exposing the ferromagnetic sense element (silver) and the supporting bobbin (black).
A. W. Geyger (1962) and W. A. Geyger (1962) suggested the use of
thin foils as a way to suppress eddy currents when constructing fluxgate
magnetometers. However, it seems unlikely that, at the time, the researchers
understood how the choice of foil thickness was impacting parameters such as
grain size that appear to have controlled the magnetic noise of the sensors
they were constructing (e.g., later work by Pfeifer and Kunz,
1977; Pfeifer and Radeloff, 1980). The potential of 6-81 permalloy to be
used in magnetic field instruments was established in a seminal work by
Gordon et al. (1968). Two other groups are known to have
constructed fluxgate sensors from 6-81 permalloy. The Themis
(Auster et al., 2008)
ring cores result from years of research in Germany
(Müller et al., 1998) and achieved noise better than 5 pT Hz
It is noteworthy that none of the works of this era published sufficient details for their process to be understood unambiguously. Much subsequent work has involved rediscovering the manufacturing details that controlled parameters such as homogenization and recrystallization. To fully understand these materials, it may be necessary to document the complete manufacturing history from melt to final heat treatment. The approach to sensor construction presented here is based on this historical research and attempts to develop a process that can produce comparable results.
Somewhat remarkably, virtually all the permalloy used in North American
fluxgate magnetometers appears to have been manufactured from a single batch
of 6-81 permalloy, likely by the Hamilton Watch Company, in or around 1969.
This permalloy was then rolled into 12 and 3
In some cases, the providence of the ring cores is complicated and difficult
to know for certain. For example, the outboard sensor on Double Star was
manufactured by Ultra Electronics (Carr et al.,
2005), the inboard sensor having been developed at the Technical University
in Braunschweig, Germany
(Auster
et al., 2008; Fornacon et al., 1999). However, there is significant, albeit
circumstantial, evidence that the Ultra Electronics sensor was manufactured
from Infinetics S1000 ring cores. Around 1993 there was a management buyout
at Ultra Electronics that appears to have included Domain Magnetics–Dowty
Aerospace, who are known to have held a significant stock of Infinetics
ring cores. The Double Star sensor has comparable geometry to the
Cassiope–e-POP (Wallis et al., 2015) sensor, suggesting that its ring cores are at least geometrically
similar to the Infinetics S1000. Ultra Electronics sensors also have similar
noise performance of
From this historical context, it becomes apparent why many magnetometers in use worldwide have nearly identical sensors (see Fig. 3); they all appear to be using legacy Infinetics S1000 ring cores. For this reason, the ring cores developed under this project were designed to be compatible with the S1000 form factor.
Example fluxgate sensors believed to be based on the Infinetics
S1000 ring core.
The physical construction of the fluxgate ring core consisted of several steps illustrated in Fig. 4. A circular bobbin (a) established the geometry of the ring core and provided mechanical strength. The ferromagnetic element, used to modulate the magnetic field in the sensor, was constructed by spiral winding a thin strip of cold-rolled permalloy foil coated with an insulator (b). The assembled bobbin and strips were heat-treated in a reducing atmosphere to optimize their magnetic properties (c). The ring core was then electrically isolated using polyimide tape (d) and a toroidal drive winding was applied (e).
Major steps in the production of an S1000-compatible ring core.
The bobbins in the ring cores described here had the geometry of the common S1000 bobbin, whose dimensions are shown in Fig. 5. The bobbin defined the geometry of the ring core and provided mechanical support to prevent the permalloy from experiencing mechanical strain after heat treatment. Even minor deformations or stresses applied to the ferromagnetic element, such as when pushing on the ring core to turn it within the sense winding, were found to significantly increase magnetic noise. The ferromagnetic element was spiral-wound into a groove machined into the outer circumference of the bobbin. This allowed a toroidal drive winding to be subsequently wound onto the bobbin without contacting or imparting stress onto the permalloy strip.
Schematic of a bobbin compatible with the S1000 geometry (mm).
The bobbins were manufactured from Inconel x750 that was selected as being
nonmagnetic and providing high rigidity even at the elevated temperatures
of the heat treatment required to optimize the magnetic properties of the
ring core. However, Inconel x750 was an imperfect match to the permalloy
sense element in terms of linear thermal expansion (12.6 ppm
A custom 4 kg ingot of 6-81 permalloy was created at the Canadian government
CANMET lab. Following the suggestion of CANMET staff, a vacuum arc furnace
was used to create a 50–50 alloy of molybdenum and nickel and then melting
in the remaining constituents in a conventional furnace. Subsequent work has
shown that it is possible to dissolve molybdenum into nickel at
The complete alloy was intended to be reduced by hot rolling. However, the
process was stopped almost immediately as the alloy began to develop severe
surface cracking (Fig. 6). This may have resulted
from partial melting near the surface due to inhomogeneity in the mix or the
alloy being non-eutectic. The cooled ingot was instead machined into 3 mm
thick stock using a milling machine. The 3 mm stock was then heat-treated at
1100
A 3 mm thick stock machined from the 6-81 permalloy ingot showing the surface cracking that resulted from an attempt at hot rolling.
Successive cold rolling reduced the permalloy from 3 mm to the final foil
thickness of
The thinnest results to date were achieved using a Cavallin bench rolling
mill for plate–strip model L.80/44-044 adapted with a reducing gear and a
735 W electric motor. The 44 mm diameter rollers and solid bronze bushings
allow more force to be applied to the foil than other, nominally comparable,
rolling mills. The current rolling mill setup
(Fig. 7a) reduced 3 mm permalloy stock
(Fig. 7b) to 100
The permalloy was then cut into 1.57 mm wide strips to fit within the groove cut into the external face of the bobbin used in S1000 form factor ring cores (see Fig. 5). The cold-rolled permalloy, despite work hardening, remained sufficiently ductile that it tended to fold rather than cut. A sharp guillotine shear could cut the permalloy foil if supported by a sacrificial brass sheet, but electro-discharge machining (EDM) and water jetting, while requiring more expensive infrastructure, were found to provide superior cutting results.
The permalloy strips were coated with magnesium oxide (an electrical
insulator) to prevent the formation of spot-welds between layers when the
strip was attached by electrical discharge welding and to prevent the
tightly wound layers from fusing with each other during heat treatments.
This insulator was created using an established process (Bill Billingsley
Sr., personal communication, 2008) from milk of magnesia,
The ferromagnetic element of the ring core was made up of insulated
permalloy strip spirally wound onto the Inconel bobbin. Depending on the
foil available, some sensors used one continuous strip, while others used
several strips welded together to be long enough for six turns on the
Inconel bobbin (Fig. 9a). The end of the
permalloy strip was spot-welded to the bottom of the channel cut into the
outer circumference of the bobbin. The strip was cut to length such that the
start and end of the strip were aligned. The ends of the permalloy strip
created an unwanted magnetic asymmetry in the ring core that manifested in
the output of the sensor. Aligning the start and end of the strip localized this
asymmetry, allowing the sensor output to be tuned for maximum symmetry by
rotating the ring core within the sense winding. In the case of double-wound
sensors, this procedure could be carried out for the first (inner) sense
winding prior to affixing the second (outer) sense winding. The spiral
winding was terminated by scraping away a small amount of the oxide layer
(Fig. 9b) and spot-welding the end of the strip
down to the layer immediately underneath (Fig. 9c). A similar method of assembling rings is described in Musmann (2010). In this case, the foil was wound by hand using a simple jig
and the tension was not finely controlled. The 50 cm long permalloy strip
used in a six-layer core had a resistance of about 1.8
The assembled fluxgate ring cores were heat-treated to produce high-permeability, low-coercivity, and repeatable re-magnetization properties in the ferromagnetic material that the authors hypothesize produced a relatively stress-free crystalline structure and therefore low magnetic noise. This approach was guided by the theory of the origin of fluxgate magnetic noise developed in Narod (2014). The heat treatment was intended to develop the largest possible grains in the given thickness of the permalloy foil without developing undesirable fabric in which the easy axes directions are misaligned with respect to the desired magnetizing direction (e.g., Major and Martin, 1970; Odani, 1964).
In contrast, Gorobei and Gorobei (1981) showed experimentally
that a low-temperature (800
Figure 10a shows a generic temperature profile for a ring core showing a rapid heating from ambient temperature (A–B), a high-temperature dwell (B–C), a controlled ramp-down (C–D), and a slow ramp through the disordering range and down to ambient temperature (D–E). The furnace had a maximum thermal slew rate due to its heating power, thermal mass, and potential for thermal-shock failure of the work tube that prevented it from rapidly heating from ambient to the dwell temperature. To circumvent this, the furnace was programmed to execute a slower temperature increase, and once the furnace had reached the desired peak dwell temperature, the ring cores were rapidly inserted into the hot zone using a mobile loading plate as illustrated in Fig. 12. Rapid heating had the effect of maximizing grain growth due to primary recrystallization. A factor as much as 10 times larger in average grain size volume is achievable by rapid compared to slow heating.
Example heat profile for a ring core showing a rapid heating from ambient temperature (A–B), a high-temperature dwell (B–C), a controlled ramp-down (C–D), and a slow ramp through the disordering range and down to ambient temperature (D–E).
At the dwell temperature, the mechanical stress embedded in cold-rolled
permalloy strips is believed to enhance primary recrystallization of the
heat-treated material. The ring cores were heat-treated in a protective
(oxygen-free) atmosphere produced by a gas mixture of 5 % hydrogen and
95 % argon continuously injected at 100 mL min
The heat profile shown in Fig. 10 was drawn
loosely from that given by Gordon et al. (1968). The dwell
temperature and duration were determined empirically based on test
ring cores constructed from 100
Ring-core heat treatment was undertaken using a modified Carbolite Gero STF160 tube furnace configured for operation with a hydrogen atmosphere. The furnace was adapted, as shown schematically in Fig. 10a, by the addition of a loading chamber where ring cores could be staged at room temperature while the furnace heated and then rapidly inserted into the hot zone without violating the controlled atmosphere. A complete heat-treatment system setup consisted of a furnace, gas cabinet, ring-core loading chamber, transport system, and data acquisition system. Figure 11 shows a diagram and a photograph of the heat-treatment system used at the University of Alberta.
Diagram
The furnace hot zone comprised a work tube of high-purity (99.8 %) alumina that is impermeable and inert to hydrogen. The adaptation for rapid ring-core insertion is shown schematically in Fig. 12. A high-purity alumina Dee tube (Coorstek, AD-998) was inserted into the alumina work tube to create a flat surface on which parts could move. The heat plug at the end of the work tube adjacent to the loading chamber was modified to create an opening aligned with the surface of the Dee tube through which the ring cores could be inserted. In the loading chamber, aluminum tubes coplanar with the surface of the Dee tube created the staging area for the ring cores.
Exploded view of the internal components of the furnace and the loading chamber. Ring cores were staged in the loading chamber for rapid insertion after the furnace reached the nominal dwell temperature.
The ring cores were placed on a molybdenum loading plate (Fig. 13) that could be pulled into the hot zone using tungsten wire threaded through the length of the work tube. Alumina powder was used to prevent the ring cores from welding to the loading plate during the heat treatment.
Ring cores on a molybdenum loading plate in a loading chamber. Alumina powder prevented the ring cores from welding to the plate.
The heat treatment had significant effects on the crystalline structure of permalloy strips as shown in Fig. 14. The width (horizontal dimension) of the permalloy strip in each photograph is about 1.57 mm. The correlation between grain size and the magnetic noise of permalloy ring cores in fluxgate sensors continues to be investigated. Larger grains are believed to produce less magnetic noise but only if they grow through primary recrystallization (Herzer, 1997). Further research on this topic is needed.
Optical image of permalloy grains
The effect of the heat treatment on the magnetic properties of the permalloy
was tested by examining the
The heat treatment reduced the coercivity of the permalloy strip by a factor of about 6. It seems likely that heat treatments can be further optimized in terms of dwell temperature, dwell time, and cooling rates; this is still being investigated. The ideal and optimized thermal profile for heat treatment is also likely to depend on the permalloy thickness.
The new ring cores were integrated into a fully working fluxgate
magnetometer to assess their noise performance. Each ring core was given a
toroidal drive winding of
Test jig used to rapidly characterize multiple ring cores. Each ring core has a toroidal drive winding applied and is mounted on a small printed circuit board. The mounted ring core inserts through a modified sense winding and the entire assembly is placed in a solenoid–magnetic shield for testing.
The fluxgate ring cores were characterized by driving and interrogating them with a fluxgate magnetometer electronics package based on a heritage design (Miles et al., 2013, 2016; Narod and Bennest, 1990; Wallis et al., 2015) utilizing a second-harmonic-tuned drive with direct digitization (Fig. 17a), which had been generalized to drive and interrogate a wide range of ring cores.
Figure 17b shows a single-axis laboratory fluxgate electronics design built for this project. The magnetometer operated on external benchtop voltage supplies, allowing the ring-core drive signal to be rapidly optimized for different ring-core designs. The sequence of large power inductors, visible on the right half of the circuit board, combined with a similar capacitor bank on the sensor fixture and custom firmware, allowed the drive circuit to be rapidly tuned to achieve the rapid and deep magnetic saturation required for low-noise operation.
Simplified functional diagram
The ring cores were driven at 5 kHz and tuned to create large-amplitude and short-duration current spikes (Fig. 18a). The sense winding was used in a short-circuit current-output configuration, creating waveforms at the output of the preamplifier as shown in Fig. 18b. The ambient magnetic field created second harmonic modulation that was sampled using synchronous digitization indicated by the vertical dashed red lines in the figure.
The ring cores were characterized for noise performance, aging effects, and
for basic thermal and vibration flight qualification. The noise floor of
each ring core was calculated by calibrating the coupled
ring–sensor–electronics against a known magnetic field using a solenoid
inside a magnetic shield. The solenoid was then switched off and 20 min
of 100 sps magnetically quiet data were recorded and processed to produce
power spectral density (PSD) using a unit-correct quantitative
implementation of Welch's method (Heinzel et al., 2002;
Welch, 1967) with a 2048-point fast Fourier transform (FFT), a 1536-point
overlap, and a Hann window function, giving a power spectral density
estimation down to
A total of 22 ring cores were manufactured using the process described herein
and heat-treated in three batches of 2, 10, and 10. The average power
spectral noise density was very similar in ring cores from all three batches,
with variability decreasing slightly in the final batch, likely because of
improved consistency in spot-welding the permalloy strips.
Figure 19 shows a histogram for the noise level of
these ring cores compared to those for historical 3 and 12
Histogram of ring-core noise (pT Hz
The number of ring cores produced by the current process was lower (22
pieces) than for the Infinetics 3
Our work towards developing a well-defined and reliable method for manufacturing low-noise fluxgate sensors for space physics and geophysics uses has generally followed on the efforts undertaken by the Naval Ordinance Laboratory in the 1960s (Gordon et al., 1968), the commonalities being the use of 6-81 Mo permalloy as the principal material, as well as its processing by cold deformation and specific heat treatment for the development of large grain structure. However, several other methodologies making use of different materials have also been used for the manufacturing of low-noise fluxgate sensors, with significant success.
Beginning in 1984 with the experimental works of Shirae (1984) and Narod et al. (1985), amorphous, high-cobalt alloys and processes were developed to create fluxgate sensors for both ground- and space-based uses. For example, in the 1990s, Otto Neilson and others at the Danish Technical University (Nielsen et al., 1995) used an amorphous alloy to create the compact spherical coil (CSC) ring-core fluxgate magnetometer aboard the Orsted satellite. Also beginning in the 1990s, Luis Benyoseph of the National Observatory of Brazil has undertaken a long-term experimental effort to improve amorphous alloys for fluxgate sensors (Benyosef, 1996; Benyosef et al., 1995, 2008). More recently, Lajos Varga at the Hungarian Academy of Sciences developed amorphous materials for magnetometry (Lajos K. Varga, personal communication, 2013).
Rapidly quenched amorphous alloys are limited by both their minimum and
maximum thicknesses compared with crystalline permalloys. Also, the
rapid cooling creates an anisotropy in the cooling direction across the
thickness. Both properties may limit their possible performance in
magnetometers. Their low Curie temperatures, 200
In the 1990s investigators in Braunschweig, Berlin, and Dresden developed a
low-noise permalloy, again having 6.0 % molybdenum and about 81 % nickel
(Müller et al., 1998). The processing of this material
was significantly different from our present material. From conversations
with Karl-Heinz Fornacon (Karl-Heinz Fornacon, personal communication, 2017) we
understand that their heat treatments were performed in a furnace that did
not permit the rapid insertion of specimens into the hot zone and that specimens
were required to warm up slowly as the furnace was heated. Such a heating
curve would eliminate all possibility of grain growth by primary
recrystallization occurring, leaving all grain growth to occur via secondary
recrystallization (Pfeifer and Radeloff, 1980). It has been
believed for a long time that secondary recrystallization is undesirable for
the development of magnetic properties (Odani, 1964). Their overall
result was that their best material for fluxgate sensors had the smallest
grain size of about 10
Significant work with various permalloys has been published in the Russian language literature (e.g., Afanas'ev, 1986; Afanas'ev et al., 1977; Afanassiev et al., 1980; Musmann, 2010). However, limited details of the heat treatments and grain sizes from the work completed around the 1970s are available in the English language literature.
If one takes as common knowledge that low coercivity and high initial
permeability are desirable traits for sensor materials, then there should be
merit is seeking materials of very small grain. Herzer (1990,
1992) found that such desirable properties occur at both ends of the grain
size spectrum, that is, for grain sizes much smaller than 1
The process described herein has been developed so that eventually all the essential steps (melting, rolling, cutting, assembling, heat-treating, calibrating, and testing) should be achievable in-house and in small quantities. This capability is intended to enable future studies to probe the many different design choices that may affect fluxgate performance (metallurgical composition, alloying process, homogenization, reduction method, work hardening, machining and heat affect, geometry, heat-treatment profile, etc.). The authors would like to encourage other researchers to publish as many of these details as is practical to facilitate the meaningful comparison of results between different cores manufactured by different groups using different processes to eventually expose all the critical parameters that enable low-noise, highly stable materials for fluxgate sensors.
Despite their ubiquitous use, the parameters that control the intrinsic
magnetic noise of the ferromagnetic element remain poorly understood, and no
published process can reproduce the performance of the widely used low-noise
Infinetics S1000 ring core. We described a manufacturing approach that can
reliably produce an S1000-compatible fluxgate ring core with a noise of 6–11 pT Hz
Ongoing research into optimizing this process and new low-noise materials
and sensors is being conducted with the goal of
The data and source code used in the creation of this paper can be accessed by contacting the authors.
DMM wrote the paper with contributions from all authors. The ring cores presented here were developed at the University of Alberta under a contract led by IRM from the Canadian Space Agency. DMM developed and built the experimental apparatus to characterize the new ring cores. BBN primarily completed the literature review on the ring-core physics and developed the heat treatment presented here. JRB provided guidance on the ring-core manufacturing process. MC and DB designed, constructed, and operated the heat-treatment system. AK primarily characterized ring cores. MRL completed preliminary tests of the ring-core manufacturing process. DKM primarily provided guidance on testing and characterizing ring-core performance. JL built, assembled, and tested infrastructure at the University of Iowa used to characterize and document ring cores for this paper.
B. Barry Narod operated Narod Geophysics Ltd., which manufactured fluxgate magnetometers until the company ceased production operation in 2008. John R. Bennest operated Bennest Enterprises Ltd., which manufactured a variety of custom scientific instruments and equipment, including fluxgate magnetometers, until the company ceased operations in 2018.
Work on the project was supported by the Canadian Space Agency under contract 9F063-140909/006/MTB_PT6_Ring-cores. David M. Miles was subsequently supported by faculty start-up funding from the University of Iowa. Ian R. Mann is supported by a Discovery Grant from Canadian NSERC. The authors wish to thank Richard Dvorsky, Michael Webb, Christian Hansen, and Spencer Kuhl for developing, manufacturing, and documenting the ring cores shown in this paper.
This research has been supported by the Canadian Space Agency (grant no. 9F063-140909/006/MTB_PT6_Ring-cores), the University of Iowa (faculty start-up grant), and the Natural Sciences and Engineering Research Council of Canada (Discovery Grant).
This paper was edited by Valery Korepanov and reviewed by two anonymous referees.