IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO.

X, MONTH 2024 1

Magnetic hysteresis cycle measurements

with the needle probes method
Patrick Fagan, Abdellahi Abderahmane, Mathieu Domenjoud and Laurent Daniel
Université Paris-Saclay, CentraleSupélec, CNRS, Laboratoire de Génie Électrique et Électronique de Paris, 91192
Gif-sur-Yvette, France
Sorbonne Université, CNRS, Laboratoire de Génie Électrique et Électronique de Paris, 75252 Paris, France
Email: patrick.fagan@centralesupelec.fr

Abstract—Different magnetic needle probe setups are pre- The needle probe method can be used for the non-
sented, and their strengths and weaknesses are explored. A destructive evaluation of the magnetic properties of a fer-
new setup, combining the advantages of the other setups, is romagnetic sample. The wrapping coil is replaced by two
presented. Finite element simulations are implemented to show
the effect of the excitation yokes and residual stresses on the conductive probes touching the surface of the sample. Such
magnetic flux symmetry. The error sources affecting the precision setup is equivalent to a wrapping coil of one turn [7], [8].
of the experimental measurements are analysed and quantified. It only requires electrical contact between the measurement
Solutions to minimize them are also proposed. By minimizing probes and the sample surface. As a result, it can be used
error sources, the obtained experimental measurements closely on large samples instead of drilling holes. However, the
follow the reference hysteresis loop obtained from a wrapping
coil in the unidirectional case. The proposed measurement device main disadvantage is that the signal-to-noise ratio can not be
notably provides enhanced accuracy for surface measurements improved as easily as for the wrapping coil method. Hence,
when wrapping coils are not implementable. the needle probe method requires low-noise amplifiers and
Index Terms—Magnetic needle probe measurements, non- a precisely defined experimental setup in order to obtain a
destructive testing, magnetic hysteresis. reliable measurement.
This paper presents different needle probe setups and their
main assumptions. An important assumption concerns the
I. I NTRODUCTION
magnetic flux symmetry inside the sample. The error sources

F ERROMAGNETIC materials exhibit a hysteresis behav-

ior under magnetic excitation. Such behavior can be
defined through several properties (coercive field, remanent
on the experimental setup are studied in order to minimize
them, and an example of measurements following the pre-
scribed protocol shows the correspondence between wrapping
induction, loop surface) that play a crucial role in a great coil and needle probe measurements in a specific case with
number of industrial applications [1], [2]. For example, the very satisfying agreement. The paper is structured as follows:
area of the magnetic hysteresis loop is linked to the magnetic
• the first section describes different needle probe setups
losses in the material, and hence to its performance under
found in the literature and presents a novel setup allowing
alternating currents. A precise experimental measurement of
for more precise measurements;
the magnetic hysteresis loop is therefore required for a great
• the second section provides finite element simulations to
number of applications.
shed light on the critical factors for needle probe design;
The wrapping coil is one of the standard magnetic mea-
• the third section gives the equations linking the measured
surement methods due to its simple experimental setup and
voltages to the magnetic flux variations inside the sample;
good signal-to-noise ratio [3]. A coil whose number of turns
• the fourth section is dedicated to the analysis of error
is known is wound around a sample, usually of rectangular
sources in the proposed setup and their correction;
or circular cross-section. The time variation of magnetic flux
• the fifth section describes the practical implementation of
inside the sample generates an induced voltage in the coil.
the experimental setup;
Such voltage is proportional to the number of turns; hence,
• the sixth and last section presents and discusses the
if a stronger signal is needed one can wind a longer coil
experimental measurements.
around the sample. However, the coil is sensitive only to the
component of the magnetic flux density that is perpendicular
to the coil cross-section. Hence, the measurement is sensitive II. D IFFERENT NEEDLE PROBE SETUPS
only to one direction in space. Additionally, coil measurements
assume an uniform magnetic flux density across the cross- The principle of the needle probe method is derived from
section, so standard samples typically feature a small cross- Maxwell-Faraday’s law. Given a surface S of boundary ∂S,
section to ensure this assumption holds true. While drilling ⃗ on ∂S is opposite to
the line integral of the electric field E
holes allow for the deployment of wrapping coils on large the surface integral of the scalar product between the surface
samples, they alter both magnetic and mechanical states [4]– ⃗ across
normal ⃗n and the derivative of the magnetic induction B
[6]. the surface S:
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 2

I ZZ ⃗
∂B
E ⃗ =−
⃗ · dl · ⃗ndS (1)
∂S S ∂t
Hence, by connecting two conductive probes to the sample
surface, the induced voltage between the two will allow to
measure the variation of the magnetic flux in the sample.
Since the probes only need to touch the sample, the needle
probe method allows for flexible setups and non-invasive
multidimensional measurements.
This method has been patented in the 1950s by three differ-
ent authors [7]–[9]. The very weak amplitude of the measured
voltages requires strong amplifications to precisely measure
the magnetic flux variation in the material. Theoretical studies
have been carried out [10]–[15] to better evaluate uncertainty
sources and alternative setups [12], [16] have been devised
to minimize their effect on the experimental measurements.
The magnetic needles method has since been employed with
good results in a great variety of materials [17], [18] for the
measurement of rotational losses [19] and defect detection Fig. 1: The four different methods for needle probe mea-
[20]. surements described in this paper: a. Original Setup (OS); b.
Four different setups have been identified in the literature, Single-Side Original Setup (SSOS); c. Air Flux Compensation
and each of them is described in this paper. These setups are Setup (AFCS); d. Transverse Needle Probe Setup (TNPS).
shown in Fig. 1. They are denoted as follows:
• the Original Setup (OS), described by Czeija and of the setup and for the scanning of the entire sample surface,
Zawischa [7] and Werner [8] in their patents, in but at the price of a greater air surface.
which a pair of needle probes is placed on each side
of the sample, hence giving two voltages Vtop and Vbottom .
A. Preamble: magnetic fields definition
• the Single-Side Original Setup (SSOS), patented by For this section, the general definition for the magnetic
⃗ in the material is considered:
induction vector B
Czeija and Zawischa [7] and Stauffer [9], [10], which
has then been employed in the subsequent decades [21].  
Bx (t)
This setup is a simplified OS, in which the measurement ⃗
B(t) = By (t) (2)
on the bottom face has been removed. As a result,
Bz (t)
the change in magnetization within the sample can be
measured using a single voltage V , under the assumption To enhance the readability of the equations, the integral
that the magnetic flux in the sample is symmetrical. form of the magnetic flux is substituted with Φ as follows:
ZZ
∂B⃗
• the Air Flux Compensation Setup (AFCS) [16], Φ(S) = · ⃗ndS (3)
derived from the OS by adding a pair of needles that S ∂t
are short-circuited, but isolated from the sample. Two Given a surface S of boundary ∂S, the induced voltage
voltage measurements are required to describe the Vind on ∂S can be calculated through Maxwell-Faraday’s law
magnetic flux variation in one direction: the voltage in as follows:
the sample Vair and the air voltage Vsample . I
Vind = E ⃗ = −Φ(S)
⃗ · dl (4)
• the Transverse Needle Probe Setup (TNPS) [12], in ∂S
which two pairs of needle probes are placed in the The surface S can be separated into two parts Sair and
same fashion as in the OS. This setup is equivalent to Ssample depending on whether there is air or ferromagnetic
a differential measurement. material in a precise point of S. Such decomposition is
Fig. 1 shows that each needle probe setup will be affected by illustrated in Fig. 2 to 6 and the resulting equation is given
magnetic flux variation in the air between the needle probes. below:
Such effect must be minimized to avoid unwanted drift in the
reconstructed hysteresis cycle. One way to achieve this is to
Φ(S) = Φ(Sair ) + Φ(Ssample ) (5)
print the needle probes directly on the sample surface [18].
Such probes would be fixed on the sample, which could be a An assumption that can reasonably be made is that the
limit since only one position is allowed. Needle probe setups measurement device has an infinite impedance, hence no
can be realized with spring contact probes to ensure the contact current flows on the cables connecting the needle probes to
between probes and sample, allowing for an easy positioning the measurement device.
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 3

B. Original Setup
Considering a magnetic field B ⃗ normal to the cross-section
Ssample , Fig. 2 shows the OS and the associated Maxwell-
Faraday’s surfaces Ssample , Sair t and Sair b .

Fig. 3: Air and sample surfaces in the Single-Side Original

Setup (SSOS).

D. Air Flux Compensation Setup

The Maxwell-Faraday surfaces for the AFCS are shown in
Fig. 4.

Fig. 2: Air and sample surfaces in the Original Setup (OS).

To obtain the magnetic flux variation inside the sample, two

extra assumptions are made to remove parasitic effects:
• the effective vertical electric field is negligible:

V14 + V32 ≈ 0 (6)

• the effective air flux term is negligible:
Φ(Sair t ) + Φ(Sair b ) ≈ 0 (7)
In this case, it is shown in appendix I-A that the magnetic
flux variation is given by the sum of the two measured voltages
V1′ 2′ and V3′ 4′ :
Fig. 4: Air and sample surfaces in the Air Flux Compensation
ZZ ⃗
∂B Setup (AFCS).
· ⃗ndS ≈ V1′ 2′ + V3′ 4′ (8)
Ssample ∂t
The only difference of this setup compared to the SSOS
C. Single-Side Original Setup is that the ”no air flux term” assumption is modified in
In the case of the SSOS, only one of either V1′ 2′ or V3′ 4′ ”negligible air flux outside the air loop”:
is available, while the same Maxwell-Faraday surfaces are
defined (Eq. (27a) and (27c)). The setup is shown in Fig. 3. Φ(Sair + ) − Φ(Sair - ) ≈ 0 (11)
To be able to solve such system, one must assume that
the electric field at the surface of the sample (and hence the It is shown in appendix I-B that the magnetic flux variation
magnetic flux inside the sample) is symmetric with respect to can be obtained from this setup by subtracting the ”inner”
the center of the cross-section: voltage V5′ 6′ from the ”outer” voltage V1′ 2′ :
⃗
( RR
∂B
V12 = V34 , V23 = V41 (9) Ssample ∂t
· ⃗ndS ≈ 2(V1′ 2′ − V5′ 6′ )
RR ⃗
∂B
(12)
Thanks to this assumption, Eq. (33) in appendix I-A shows Sair - ∂t
· ⃗ndS = V5′ 6′
that the magnetic flux variation is proportional to one voltage
The AFCS is equivalent to the SSOS, but with an effective
measurement:
air surface Sair = Sair + − Sair - . This setup allows to decrease
ZZ ⃗
∂B the air flux term by decreasing the effective surface Sair , but the
· ⃗ndS ≈ 2V1′ 2′ (10) vertical electric field can still be an issue, and the symmetry
Ssample ∂t
assumption requires a precise positioning of the needle probes.
The main disadvantages of the SSOS compared to the OS
is the halved gain in Eq. (10) and the additional assumption
regarding the symmetry of the problem, while keeping the E. Transverse Needle Probe Setup
sensitivity to the magnetic flux variation in the air. Fig. 5 shows the Maxwell-Faraday surfaces for the TNPS.
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 4

Fig. 5: Air and sample surfaces in the Transverse Needle Probe

Setup (TNPS).

It is shown in appendix I-C that such setup is equivalent

to the difference of two SSOS, defined by their measured
voltages V1′ 4′ and V2′ 3′ :
ZZ ⃗
∂B
· ⃗nds ≈ V1′ 4′ − V2′ 3′ (13) Fig. 6: Air and sample surfaces in the Complete Needle Probe
Ssample ∂t
Setup (CNPS).
Such setup doubles the gain for the magnetic flux variation
compared to the SSOS and is less impacted by the vertical
field (since the two contributions are subtracted). It also allows Table I compares the five different setups, showing the
to decrease the air surface (even if not as much as the strengths and weaknesses of each one.
AFCS). As a result, the TNPS can be employed for precise
measurements. However, the needle probes need to be aligned TABLE I: Comparisons between the different needle probe
on both sides of the sample, hence the TNPS can be hard setups studied above.
to implement on plates and less practical than the AFCS for Setup Number of measures Requires symmetry Air flux effect
quick measurements.
SSOS 1 Yes Great
OS 2 No Great
F. Complete Needle Probe Setup AFCS 2 Yes Small
TNPS 2 No Great
The ideal setup would allow to reduce both the air flux CNPS 4 No Small
and the vertical electric field contributions to negligible levels.
Hence, one could try to combine the AFCS and the TNPS to
get the advantages of both. An improved system, the Com-
plete Needle Probe Setup (CNPS), should then give accurate III. F INITE ELEMENTS MODEL FOR THE NEEDLE PROBE
measurements. The setup is shown in Fig. 6. MEASUREMENTS
With the assumptions defined in appendix I-D, four mea- A. Definition of the modeled problem
surements allow to define the magnetic flux variation inside
the sample: The experimental system was modeled using the finite
element (FE) method in COMSOL Multiphysics™ . The
setup is presented in Fig. 7. The measurement region size
ZZ ⃗
∂B
· ⃗nds ≈ [V1′ 2′ − V5′ 6′ ] + [V3′ 4′ − V7′ 8′ ] (14) (20×20×6 mm3 ) and location (at the center of the sample)
Ssample ∂t fulfill the criteria (derived from [6]) for reliable magnetic
measurements:
The advantage of the CNPS is that it can be simplified to
• Uniformity criterion: H ⃗ and B⃗ fields are uniform (di-
obtain five different measurements simultaneously (OS and
AFCS / SSOS for both sample surfaces) simply by choosing rection and magnitude) in their measurement regions;
• Correspondence criterion: the measured H ⃗ field corre-
which measured voltages to analyze, without moving the
sponds to the one giving rise to the measured B ⃗ field;
needle probes. Therefore, a single measurement file can be
• Direction criterion: H ⃗ and B ⃗ fields’ directions are
thoroughly investigated to identify dissymmetries in magnetic
flux and address any issue related to the air flux term. However, known.
it requires to measure four voltages per axis (and, hence, to The sample was meshed using a swept quadrangle mesh
carefully position four couples of needle probes). Thus, the in the Z-axis and a boundary layer mesh in the Y-axis (with
experimental setup must be carefully designed. four elements in the skin depth). The rest of the geometry
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 5

electric field on the cross-section is horizontal. This implies

that the vertical voltages Vl and Vr are negligible and opposite
(Vl + Vr ≈ 0), hence the voltage measured on a virtual
wrapping coil following the path on Fig. 8 is obtained by
summing Vt and Vb :

Vcoil ≈ Vt + Vb (15)
If the setup is not symmetrical (as detailed in the next
paragraphs), then Vt will not be identical to Vb , hence an
asymmetry factor ϵASYM can be defined:
v
uR T
u [Vt (t) − Vb (t)]2 dt
ϵASYM (%) ≈ 100t 0 R T (16)
2 (t)dt
Vcoil
0
Fig. 7: Simulated structure on COMSOL. Dimensions are in
Simulations are carried out for a duration T = 4f1exc , fexc
mm.
being the excitation frequency. This implies that only the
first magnetization curve of the material is simulated to keep
uses tetrahedral elements. The coils are modeled using a reasonable simulation times.
homogenized multi-turn conductor model.
The material properties are given in Tab. II. Note that other B. Geometrical asymmetry
configurations with different geometries (coils on the yoke One way to introduce asymmetry in the through-thickness
legs, thinner sample, shorter yokes) and material properties magnetic flux is to break the symmetry of the excitation system
were studied. However, the conclusions drawn from such (coils and yoke). Several examples have been tested with
studies are identical to the ones presented here. The sample fexc = 1 Hz:
material is an Iron-Cobalt alloy, whose magnetization function • Case 0: two excitation yokes (perfectly symmetrical),
M (H) and parameters are taken from [22]. chosen as reference;
• Case 1: one excitation yoke;
TABLE II: Electromagnetic properties used in the FE model.
• Case 2: two excitation yokes, with one yoke generating
Properties Yoke Coil Sample 10% more magnetic flux compared to the other;
Electric conductivity (MS/m) 1 6 3 • Case 3: two excitation yokes, one yoke being moved
Relative permittivity 1 1 1 5 mm in the X direction;
Relative magnetic permeability 104 1 Non-linear [22] • Case 4: two excitation yokes, one yoke being moved
5 mm in the Y direction (normal to the sample surface);
Simulated needle probe measurements are obtained by inte- • Case 5: two excitation yokes, one yoke being moved
grating the simulated electric field following the paths in Fig. 5 mm in the Z direction;
8 counter-clock-wise. • Case 6: two excitation yokes, one yoke being rotated of
10° compared to the other one.
Results are shown in Table III.
TABLE III: Asymmetry factor in different simulation cases.
Case 0 1 2 3 4 5 6
ϵASYM (%) 0.16 8.41 0.61 0.27 2.92 0.25 0.23

It can first be noticed that the asymmetry coefficient is not

zero in the symmetric case. This is due to residual dissymetries
in the meshes. The value of 0.16% for ϵASYM can be taken as
the reference for negligible asymmetry. Among the studied
cases, a strong asymmetry is generated by an asymmetric
magnetic excitation, with either one yoke or two yokes with a
detectable liftoff for one of them. A current asymmetry (as in
Case 2) should not have a great effect on the asymmetry if the
Fig. 8: Integration paths employed for the needle probes and two excitation currents are similar enough, and a misalignment
wrapping coil voltages. for the two yokes in the sample surface directions should not
be noticeable.
The simulated system is symmetrical and the needle probes Hence, using two identical yokes with similar excitation
are far enough from the sample edges to assume that the coils (in terms of coil turns and surface) connected in series
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 6

and ensuring that liftoff is minimal should reduce the flux

asymmetry to acceptable levels.

C. Material asymmetry
An asymmetry in the material properties of the sample
(e.g. due to residual stresses) can cause an asymmetry in the
through-thickness magnetic flux, even if the excitation setup is
perfectly symmetric. Various through-thickness uniaxial stress
profiles were studied. These are similar to those introduced by
a rolling process [23]. The analytical magneto-elastic model
[22] was used to compute anhysteretic B(H) curves as a
function of uniaxial stress (see Fig. 9). The simulated stress
profile is given in Fig. 10. Fig. 11: Simulated wrapping coil (Vt + Vb ) and top (2Vt )
and bottom (2Vb ) Single-Side Original Setup (SSOS) mea-
surements on a sample with residual stress.

IV. DATA TREATMENT: FROM THE MEASURED VOLTAGES

TO THE MAGNETIC INDUCTION FOR THE PROPOSED SETUP

This section describes the different steps required to obtain

the magnetic flux variation from the measured voltages for the
CNPS (similar equations can be found for the simpler setups).
The complete setup is shown in Fig. 12, and the Maxwell-
Faraday surfaces are shown in Fig. 13.

Fig. 9: Anhysteretic B(H) curves for the simulated ferromag-

netic material for various applied uniaxial mechanical stresses.

Fig. 12: Proposed setup with the geometric dimensions and

the analog treatment circuits.

The connection wires for the measured signals are grouped

in a thick coaxial cable. The voltage ei is the voltage measured
in the loop containing the points Ti and Ti’ for the ”top” face,
Fig. 10: Continuous and discretized stress profiles for the FE and Bi and Bi’ for the ”bottom” face. Each voltage is also
modelling. amplified by an analog circuit of transfer function Gi (jω)
(with ω the angular frequency).
For this configuration, the simulated needle probes voltages The general equation for this setup is developed in appendix
are given in Fig. 11. The asymmetry between Vt and Vb II and is simplified thanks to the following assumptions:
is clearly visible, with a second lobe visible on the bottom
• Assumption 1: the amplification circuits on all inputs are
measurement (the compression being maximal near the bottom
perfect, each one with a gain Gi for all frequencies:
surface). For this particular case, ϵASYM = 28.6%. However,
such asymmetry strongly depends on the stress profile, hence Vi = Gi ei (17)
no simple recommendations can be given to ensure the mag-
netic flux symmetry, unlike for the geometric factors seen • Assumption 2: input impedances are infinite (or, in other
previously. words, no current flows in the connections between the
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 7

placing two pairs of needle probes perpendicular to each other

(unless specified otherwise). The magnetic flux is supposed
uniform on the entirety of the cross-section and along the Z
direction.
Different error sources have been analyzed in appendix III:
• The needle probe position errors include the position
errors between the ”top” and ”bottom” sample needle
probes due to a misalignment of the needles support.
Such position errors can be a translation error (∆x and
∆z) and a rotation error (β). Only the latter has an effect
on the experimental measurement if the magnetic flux is
uniform. The resulting error can be ignored if |β| < 14°,
Fig. 13: Maxwell-Faraday surfaces in the proposed setup. All which should be the case in the great majority of cases;
surfaces beyond the coaxial cable are ignored. • The air needle probes liftoff concerns the presence of
a liftoff δ between the sample surface and the air needle
probes. The analytical expression of this error can not
needle probes and the amplification circuit): be easily calculated, since it heavily depends on Bn (y).
However, since Bn (y) should quickly converge towards 0
Ii = 0 (18) far enough from the sample surface, even small values of
• Assumption 3: the liftoff δ between the ”air” needle δ will significantly decrease the measured air flux voltage.
probes and the sample surface is negligible: Hence, the resulting B(H) loops will be affected by an
increased drift at saturation if the air needle probes liftoff
δ→0 (19) is not minimized;
• the air-sample angle mismatch concerns the position
• Assumption 4: the magnetic flux far from the sample is
negligible compared to the magnetic flux near the surface: errors between the ”sample” and the ”air” needle probes
on the same sample surface. As with the needle probes
∂Bn ∂Bn position errors, only the angular error γ affects the ex-
(|y| → h + hair ) ≪ (|y| → h) (20)
∂t ∂t perimental measurements and, like the air needle probes
hair + ≪ hair (21) liftoff, it will result in an increased drift at saturation for
the B(H) loops. This error evolves in 1 − cos(γ), hence it
• Assumption 5: the magnetic flux is uniform on the X should be considered only in presence of major alignment
axis between the two ”sample” needle probes (T1 T1′ and issues on the experimental setup;
B3 B3′ ): • the gain mismatch ∆G and the non-orthogonality angle
Z h ∆α include the gain differences for the amplification
∂Bn
Φ(Ssample ) = Lneedles (y)dy (22) circuits and the alignments issues between the X and the
−h ∂t
Z hair Z needle probes. Such effects can be corrected in post-
∂Bn
Φ(Sair int top ) = Lair (y)dy (23) treatment with two different measurements.
h+δ ∂t
Z −h−δ As long as the experimental setup is manufactured with
∂Bn precision (such as by 3D printing the needle probes support),
Φ(Sair int bottom ) = Lair (y)dy (24)
−hair ∂t all these errors should be either negligible or easily corrected
By combining these assumptions, the magnetic flux varia- in post-treatment.
tion in the sample is given by:
VI. E XPERIMENTAL SETUP
   
V1 Lneedles V2 V3 Lneedles V4 The easiest way to compare the needle probes and the
Φ(Ssample ) ≈ − − + − wrapping coil measurements is through a classical Single-
G1 Lair G2 G3 Lair G4
(25) Sheet Tester (SST) measurement. In this case, the magnetic
field is generated through one pair of yokes and hence both
This equation is the general case for all magnetic needle ⃗ and B ⃗ are supposed 1D. The experimental setup is shown
H
probe setups except the TNPS, and can be simplified for each
in Fig. 14. Two types of samples have been tested: DC04 steel
elementary setup seen previously.
[24] and Galfenol [25].
As in the FE model, two identical iron-silicon yokes have
V. E RROR SOURCES IN THE 2D MAGNETIC INDUCTION been employed to obtain a symmetrical magnetic excitation
MEASUREMENT ( UNIAXIAL EXCITATION CASE ) (the two excitation yokes have 106 ± 3 coil-turns). Both
The relationship between magnetic induction change and excitation coils are connected to a Kepco BOP72-14MG
measured voltages seen above can also be employed to detect bipolar power source (with a maximal current of 14 A and
several error sources for the hysteresis loop reconstruction in a maximal voltage of 72 V). The magnetic field is measured
the 2D measurements. Such measurements are obtained by using a Projekt Elektronik AS-V3DM Hall sensor, allowing
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 8

Fig. 14: Experimental setup employed for needle probes and

wrapping coil measurements.

a 3D measurement with an estimated error of 4 A/m. The

sampling frequency is 5 kHz. More details about the excitation
generation and control can be found in [26]. The geometric Fig. 16: Schematics of the needle probe holder and the probe
dimensions and the error parameters are given in Table IV. connections.
The magnetic induction is measured with the proposed
needle probe setup and compared to the signal of a wrapping TABLE IV: Experimental setup parameters.
coil of 25 (for the DC04 sample) or 30 (for the Galfenol
Parameter Value
sample) coil turns. A specific sensor has been designed and
2h (mm) 2
3D-printed in order to read the magnetic induction through the
2L (mm) 20
needle probes at the same place as the wrapping coil. A photo δ (mm) 2
and a drawing of the holder are shown in Fig. 15 and 16. The hair (mm) 35
”inner” needle probes are short-circuited above the wrapping hair + (mm) 5
coil to measure the air flux, while the ”outer” needle probes ∆G <1%
touch the sample. ∆α (°) <2
Linner (mm) 20 ± 0.5
Louter (mm) 25 ± 0.5
β (°) 45

clear drift after the magnetic saturation. The results obtained

from the asymmetric setups (SSOS and AFCS) are also differ-
ent depending on the chosen face (A or B), such effect being
more noticeable on the dB/dt measurements. Such asymmetry
is linked to the material, since the magnetic excitation system
is symmetric. It is not attempted here to explain the presence
of this asymmetry, an asymmetrical residual stress distribution
being one possible reason, as seen in Section III-C. However
the proposed setup gives access to this asymmetry, allowing for
a more complete measurement than the wrapping coil setup.
The asymmetry coefficient ϵASYM for each measurement is
Fig. 15: Needle probe holder mounted on the sample between given in Table V.
the yokes.
TABLE V: Experimental asymmetry coefficients for the mea-
surements.
VII. E XPERIMENTAL MEASUREMENTS Sample Setup ϵASYM (%)
The measured hysteresis loops for both DC04 and Galfenol DC04 AFCS 4.44
samples have been carried out with fexc = 1 Hz. Fig. 17 DC04 SSOS 8.76
shows the experimental hysteresis loops through the symmetric Galfenol AFCS 3.36
setups (CNPS, OS and TNPS) and the measured dB/dt for the Galfenol SSOS 4.85
asymmetric setups (SSOS and AFCS). For both figures, all
signals are compared to the measurements from the wrapping For more quantitative comparisons between the curves, the
coil. Relative Euclidean Difference (RED) is calculated. This factor
The air flux term is clearly noticeable on all experimental will be 0 if two curves are identical, and the greater its value,
measurements for the chosen needle probes support, since all the greater the differences. Given two arrays (the reference x
setups without air flux compensation (OS and TNPS) show a and the test curve y) of size N , their RED will be:
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 9

v
u PN
u (x[i] − y[i])2
RED(%) = 100t i=1
PN (26)
2
i=1 x[i]

Table VI shows the RED between the wrapping coil data

(B(H) and dB/dt(H)) and the measured data with each needle
probe setup.
TABLE VI: Relative Euclidean Difference for each setup for
the two samples.
DC04 Galfenol
Setup B(H) (%) dB/dt(H) (%) B(H) (%) dB/dt(H) (%)
CNPS 1.07 3.74 0.93 5.67
TNPS 18.63 12.05 13.26 9.21
(a) DC04, B(H). OS 21.09 13.70 13.80 9.10
SSOS A 20.05 14.45 13.22 9.53
SSOS B 22.25 15.55 13.30 11.85
AFCS A 2.09 7.76 2.10 6.25
AFCS B 4.13 6.45 6.12 9.39

For both materials, the hysteresis loop given by the proposed

setup closely matches the wrapping coil measurement, while
the drift generated by the air flux heavily impacts the setup
without air flux compensation (TNPS, OS and SSOS). The
AFCS gives similar results to the proposed setup, but with a
noticeable variability of the measure depending on the chosen
face for the measurement (A or B).
(b) DC04, dB/dt(H).
VIII. C ONCLUSION
The needle probe technique is a good candidate for magnetic
induction measurements for several reasons:
• the position and direction of the measurement can be
easily changed, for example by moving the needle probe
holder;
• there are less restrictions on the shape of the sample to
allow for needle probe measurements;
• bi-dimensional measurements are available.
This paper presents several needle probe setups, some
already described in the literature and whose strong and weak
(c) Galfenol, B(H). points are known, but also presents a novel setup. Even if a
greater number of inputs is required to acquire all the useful
signals, this setup combines the advantages of the previous
setups, namely the reduction of the air flux term and the fact
that flux symmetry is not required for further data treatment.
Moreover, its versatility allows for measurements that are
sensitive to flux asymmetries.
Furthermore, several uncertainty sources have been studied
through both numerical and analytical approaches. Dual yokes
are required (and their liftoff minimized) to minimize flux
asymmetries generated by the setup, and simple algorithms
are available to fix the other error sources once they have
(d) Galfenol, dB/dt(H). been measured by a calibration measurement. After correction
of the uncertainty sources, the proposed setup gives the same
Fig. 17: Experimental hysteresis loops and dB/dt(H) measured measurements in a 1D-configuration as the wrapping coil with
through different needle probe setups, and comparison to the a relative error smaller than 1.2%.
wrapping coil hysteresis loop. This study allows defining guidelines for the implementation
of the needle probe technique for magnetic field measure-
ments. It is notably shown that the following aspects should
be considered carefully:
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 10

• the air flux term adds an important drift to the experimen- [14] H. Pfutzner and G. Krismanic, “The needle method for induction tests:
tal measurements, hence its correction is necessary, either Sources of error,” IEEE Transactions on Magnetics, vol. 40, no. 3, pp.
1610–1616, 2004.
by minimizing the effective air surface or by directly [15] G. Crevecoeur, L. Dupré, L. Vandenbossche, and R. Van de Walle,
measuring the induced voltage in that surface; “Local identification of magnetic hysteresis properties near cutting edges
• the gain and dephasing brought by analog amplifiers and of electrical steel sheets,” IEEE Transactions on Magnetics, vol. 44,
no. 6, pp. 1010–1013, 2008.
filters must be precisely measured to minimize distor- [16] A.-E. Abdallh, P. Sergeant, G. Crevecoeur, L. Vandenbossche, L. Dupré,
tions; and M. Sablik, “Magnetic material identification in geometries with
• high-frequency measurements affect the magnetic flux non-uniform electromagnetic fields using global and local magnetic
measurements,” IEEE Transactions on Magnetics, vol. 45, no. 10, pp.
uniformity through the skin effect and bring both heat and 4157–4160, 2009.
mechanical vibrations, hence caution must be taken when [17] Y. A. Tene Deffo, P. Tsafack, B. Ducharne, B. Gupta, A. Chazotte-
comparing the needle probe technique to the reference Leconte, and L. Morel, “Local measurement of peening-induced residual
stresses on iron nickel material using needle probes technique,” IEEE
wrapping coils; Transactions on Magnetics, vol. 55, no. 7, pp. 1–8, 2019.
• the contact between the needle probes and the sample [18] B. Ducharne, Y. A. Tene Dieffo, P. Tsafack, and S. H. Ngued-
surface must be ensured at all times. jang Kouakeuo, “Directional magnetic barkhausen noise measurement
using the magnetic needle probe method,” Journal of Magnetism and
Provided that a careful implementation is performed, the Magnetic Materials, vol. 519, p. 10, 2020.
needle probe technique is a very powerful tool for accurate [19] W. Brix, K. Hempel, and F. Schulte, “Improved method for the investi-
gation of the rotational magnetization process in electrical steel sheets,”
magnetic induction measurements, with limited restrictions on IEEE Transactions on Magnetics, vol. 20, no. 5, pp. 1708–1710, 1984.
the geometry of the inspected specimens. [20] S. Imamori, S. Aihara, H. Shimoji, A. Kutsukake, and K. Hameyer,
“Evaluation of local magnetic degradation by interlocking electrical steel
sheets for an effective modelling of electrical machines,” Journal of
ACKNOWLEDGEMENTS Magnetism and Magnetic Materials, vol. 500, p. 166372, 2020.
[21] M. Enokizono, I. Tanabe, and T. Kubota, “Localized distribution of
This work was supported by the French Agence Nationale two-dimensional magnetic properties and magnetic domain observation,”
de la Recherche (ANR) under Grant ANR-22-CE42-0029-01 Journal of Magnetism and Magnetic Materials, vol. 196-197, pp. 338–
340, 1999.
(project ENCORE). [22] L. Daniel, “An analytical model for the effect of multiaxial stress on the
magnetic susceptibility of ferromagnetic materials,” IEEE Transactions
on Magnetics, vol. 49, no. 5, pp. 2037–2040, 2013.
R EFERENCES [23] M. Mahmoodi, M. Sedighi, and D. Tanner, “Investigation of through
thickness residual stress distribution in equal channel angular rolled al
[1] N. Soltau, D. Eggers, K. Hameyer, and R. W. De Doncker, “Iron losses 5083 alloy by layer removal technique and x-ray diffraction,” Materials
in a medium-frequency transformer operated in a high-power DC–DC & Design, vol. 40, pp. 516–520, 2012.
converter,” IEEE Transactions on Magnetics, vol. 50, no. 2, pp. 953–956, [24] M. Domenjoud and L. Daniel, “Effects of plastic strain and reloading
2014. stress on the magneto-mechanical behavior of electrical steels: Exper-
[2] Y. Li, X. Yan, C. Wang, Q. Yang, and C. Zhang, “Eddy current loss iments and modeling,” Mechanics of Materials, vol. 176, p. 104510,
effect in foil winding of transformer based on magneto-fluid-thermal 2023.
simulation,” IEEE Transactions on Magnetics, vol. 55, no. 7, pp. 1–5, [25] M. Domenjoud, A. Pecheux, and L. Daniel, “Characterization and
2019. multiscale modeling of the magneto-elastic behavior of galfenol,” IEEE
[3] S. Tumanski, “Induction coil sensors—a review,” Measurement Science Transactions on Magnetics, vol. 59, no. 11, pp. 1–5, 2023.
and Technology, vol. 18, no. 3, pp. R31–R46, 2007. [26] M. Domenjoud, E. Berthelot, N. Galopin, R. Corcolle, Y. Bernard, and
[4] S. Zurek and T. Meydan, “Rotational power losses and vector loci under L. Daniel, “Characterization of giant magnetostrictive materials under
controlled high flux density and magnetic field in electrical steel sheets,” static stress: influence of loading boundary conditions,” Smart Materials
IEEE Transactions on Magnetics, vol. 42, no. 10, pp. 2815–2817, 2006. and Structures, vol. 28, no. 9, p. 095012, 2019.
[5] S. Nguedjang Kouakeuo, Y. Tena Deffo, B. Ducharne, L. Morel,
M. Raulet, P. Tsafack, J. Garcia-Bravo, and B. Newell, “Embedded
printed magnetic needle probes sensor for the real-time control of the
local induction state through a laminated magnetic core,” Journal of
Magnetism and Magnetic Materials, vol. 505, p. 166767, 2020.
[6] A. Abderahmane and L. Daniel, “Measurement criteria for the magnetic
characterization of magnetic materials,” IEEE Transactions on Instru-
mentation and Measurement, vol. 72, pp. 1–18, 2023.
[7] E. Czeija and R. Zawischa, “Vorrichtung zum messen des wechselin-
duktionsflusses oder der flußänderung in ferromagnetischen materialien
aus der induktionsspannung,” patent, 1955.
[8] E. Werner, “Einrichtung zur messung magnetischer eigenschaften von
blechen bei wechselstrommagnetisierung,” patent AT191 015B, 1957.
[9] L. H. Stauffer, “Methods of and device for determining the magnetic
properties of specimens of magnetic material,” patent 2 828 467, 1958.
[10] R. E. Tompkins, L. H. Stauffer, and A. Kaplan, “New magnetic core loss
comparator,” Journal of Applied Physics, vol. 29, no. 3, pp. 502–503,
1958.
[11] T. Yamaguchi, K. Senda, M. Ishida, K. Sato, A. Honda, and T. Ya-
mamoto, “Theoretical analysis of localized magnetic flux measurement
by needle probe,” Le Journal de Physique IV, vol. 08, pp. Pr2–717–Pr2–
720, 1998.
[12] G. Loisos and A. Moses, “Critical evaluation and limitations of localized
flux density measurements in electrical steels,” IEEE Transactions on
Magnetics, vol. 37, no. 4, pp. 2755–2757, 2001.
[13] M. De Wulf, L. Dupré, D. Makaveev, and J. Melkebeek, “Needle-probe
techniques for local magnetic flux measurements,” Journal of Applied
Physics, vol. 93, no. 10, pp. 8271–8273, 2003.
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 11

A PPENDIX I In the case of the SSOS, only one of either V1′ 2′ or V3′ 4′
D ETAILED CALCULATIONS FOR EACH NEEDLE PROBE is available, while the same Maxwell-Faraday surfaces are
SETUP defined (Eq. (27a) and (27c)). To be able to solve such system,
A. Original Setup (OS) and Single-Side Original Setup (SSOS) one must assume that the electric field is symmetric:
Fig. 18 shows the OS and the Maxwell-Faraday’s surfaces
V12 = V34 , V23 = V41 (34)
Ssample , Sair t and Sair b .
Hence, Eq. (27a) and (27c) become:

2V21 = −Φ(S) − 2V23 (35)

V2′ 1′ = V21 − Φ(Sair t ) (36)

1
V2′ 1′ = − Φ(Ssample ) − V23 − Φ(Sair t ) (37)
2
Finally, with the same assumptions defined in (6) and (32):
ZZ ⃗
∂B
· ⃗ndS ≈ 2V1′ 2′ (38)
S ∂t

Fig. 18: Air and sample surfaces in the Original Setup (OS). B. Air Flux Compensation Setup (AFCS)
The Maxwell-Faraday surfaces for the AFCS are shown in
The magnetic flux variation in each surface Sair t , Sair b and Fig. 19. This setup being an extension of the SSOS, Eq. (28)
Ssample can be linked to the measured voltages as follows: and (34) are supposed true for this setup.

 V2′ 1′ + V1′ 1 + V12 + V22′ = −Φ(Sair t )
 (27a)
V4′ 3′ + V3′ 3 + V34 + V44′ = −Φ(Sair b ) (27b)

V21 + V14 + V43 + V32 = −Φ(Ssample ) (27c)


To further simplify these equations, it is assumed that

there is no voltage drop between the needle probe and the
acquisition device:

Vii′ = 0 (28)
Hence, by combining Eq. (27a), (27c) and (28):
Fig. 19: Air and sample surfaces in the Air Flux Compensation
Setup (AFCS).
V2′ 1′ + V12 + V4′ 3′ + V34 = −Φ(Sair t ) − Φ(Sair b ) (29)
The measured voltages V1′ 2′ and V3′ 4′ define two Maxwell-
Faraday surfaces S + Sair + and Sair - :
V2′ 1′ + V4′ 3′ = − Φ(Sair t ) − Φ(Sair b )
− Φ(Ssample ) − (V14 + V32 ) (30) V2′ 1′ + V14 + V43 + V32 = −Φ(S) − Φ(Sair + ) (39)
To obtain the magnetic flux variation inside the sample, two V6′ 5′ = −Φ(Sair - ) (40)
extra assumptions are made to remove parasitic effects: ⇒ V2′ 1′ − V6′ 5′ =
• the residual vertical electric field is negligible: 1
− (V32 + V14 ) − Φ(Sair + ) + Φ(Sair - ) − Φ(Ssample ) (41)
2
V14 + V32 ≈ 0 (31)
To simplify Eq. (41), the ”no air flux term” assumption is
• the air flux term is negligible: modified in ”negligible air flux outside the air loop”:
Φ(Sair t ) + Φ(Sair b ) ≈ 0 (32)
Φ(Sair + ) − Φ(Sair - ) ≈ 0 (42)
In this case, the magnetic flux variation inside the material
for the OS setup is given by: In this case:
⃗
( RR
∂B
ZZ ⃗
∂B · ⃗ndS ≈ 2(V1′ 2′ − V5′ 6′ )
· ⃗ndS ≈ V1′ 2′ + V3′ 4′ (33) RRSsample∂ B⃗∂t (43)
Ssample ∂t Sair - ∂t
· ⃗ndS = V5′ 6′
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 12


 V2′ 1′ + V14 + V4′ 3′ + V32 = ΣΦ (48a)

 ΣΦ = −Φ(S
air t + ) − Φ(Ssample ) − Φ(Sair b + )



 V6′ 5′ = −Φ(Sair t - ) (48b)

V8′ 7′ = −Φ(Sair b - )

(48c)
Combining these equations gives a first link between the
four measured voltages and the magnetic flux variation in the
sample:
Fig. 20: Air and sample surfaces in the Transverse Needle
Probe Setup (TNPS). V2′ 1′ + V14 + V4′ 3′ + V14 − V6′ 5′ − V8′ 7′ =
= −[Φ(Sair t + ) − Φ(Sair t - )] − Φ(Ssample )
C. Transverse Needle Probe Setup (TNPS) − [Φ(Sair b + ) − Φ(Sair b - )] (49)
Fig. 20 shows the Maxwell-Faraday surfaces for the TNPS. With the assumptions (6) and that the air surfaces are
The two measured voltages V1′ 4′ and V2′ 3′ can be exploited identical (Sair t + = Sair t - , Sair b + = Sair b - ):
for a differential measurement:
ZZ ⃗
∂B
V4′ 1′ = −V14 − Φ(Sair + ) − Φ(Ssample + ) (44) · ⃗nds ≈ [V1′ 2′ − V5′ 6′ ] + [V3′ 4′ − V7′ 8′ ] (50)
V3′ 2′ = −V23 − Φ(Sair - ) − Φ(Ssample - ) (45) Ssample ∂t
⇒ V4′ 1′ − V3′ 2′ =
− (V14 − V23 ) − [Φ(Sair + ) + Φ(Sair - )] − [Φ(Ssample + ) − Φ(Ssample - )] A PPENDIX II
(46) D ETAILED CALCULATIONS FOR THE CNPS
If Eq. (6) and (32) are verified: This section details the calculations for the CNPS to obtain
the equation linking the magnetic flux variation in the sample
ZZ ⃗
∂B Φ(Ssample ) and the measured voltages ei . Fig. 22 and 23 de-
· ⃗nds ≈ V1′ 4′ − V2′ 3′ (47)
S ∂t scribe the setup and the associated Maxwell-Faraday surfaces.

D. Complete Needle Probe Setup (CNPS)

The Maxwell-Faraday surfaces for the CNPS is shown in
Fig. 21.

Fig. 22: Proposed setup with the geometric dimensions and

the analog treatment circuits.

With the chosen notation, three surfaces are considered for

Maxwell-Faraday: Ssample + Sair tot top + Sair tot bottom , Sair int top
and Sair int bottom :


Fig. 21: Air and sample surfaces in the Complete Needle Probe  e1 − e3 = −Φ(Ssample ) − Φ(Sair tot top ) − Φ(Sair tot bottom ) (51a)

Setup (CNPS). e2 = −Φ(Sair int top ) (51b)

−e4 = −Φ(Sair int bottom ) (51c)

By applying Maxwell-Faraday’s law to the surfaces
Sair t + + Ssample + Sair b + , Sair t - and Sair b - : Eq. (51b) can be simplified thanks to assumptions 3 and 5:
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 13

∂Bn
Φ(S) = S (63)
∂t

4
∂Bn X
(t) ≈ αk Vkk′ (t) (64)
∂t
k=1
1

 α1 = − SG1

 α = Lneedles 1

2 Lair SG2
1
(65)
 α3 = SG
3


α4 = − LLneedles 1

air SG4

Fig. 23: Maxwell-Faraday surfaces in the proposed setup. All

A PPENDIX III
surfaces beyond the coaxial cable are ignored.
U NCERTAINTY SOURCES ON 2D MAGNETIC INDUCTION
MEASUREMENTS AND THEIR DETERMINATION
The relationship between magnetic induction change and
ZZ
∂Bn measured voltages seen above can also be employed to detect
e2 = − dS (52) several error sources for the hysteresis loop reconstruction in
Sair int top ∂t
Z hair the 2D measurements. Such measurements are obtained by
∂Bn placing two pairs of needle probes perpendicular to each other
e2 = −Lair dy (Assumption 5) (53)
δ ∂t (unless specified otherwise). The magnetic flux is supposed
hair
uniform on the entirety of the cross-section and along the Z
Z
∂Bn
e2 = −Lair dy (Assumption 3) (54) direction.
0 ∂t
Applying assumptions 4 and 5 to Φ(Sair tot top ) yields:
A. Linear displacements and angle shift between the ”top”
ZZ
∂Bn and the ”bottom” needle probes
Φ(Sair tot top ) = − dS (55)
Sair tot top ∂t This section develops the equations linking the measured
Z hair + flux error to the linear displacements ∆x and ∆z and to the
∂Bn
Φ(Sair tot top ) = −Lneedles dy (Assumption 5) angular displacement β. The setup is described in Fig. 24.
0 ∂t
(56)
Z hair
∂Bn
Φ(Sair tot top ) ≈ −Lneedles dy (Assumption 4)
0 ∂t
(57)
Hence, by combining Eq. (54) and (57):

Lneedles
Φ(Sair tot top ) ≈ e2 (58)
Lair
Applying the same reasoning to the bottom surface and
applying Assumption 1 gives:

Lneedles
Φ(Sair tot bottom ) ≈ e4 (59)
Lair
Fig. 24: Lengths and angle definition for the angle shift error.
Finally, injecting Eq. (58) on Eq. (51a) gives:
The position of each probe is:
Lneedles
e1 − e3 ≈ −Φ(S) + (e2 − e4 ) (60)
Lair      
  ∆x + Lneedles cos(β) ∆x − Lneedles cos(β)
Lneedles Lneedles  , T1′ = 
Φ(S) ≈ − e1 − e2 + e3 − e4 (61) T1 =  h h 
Lair Lair ∆z − Lneedles sin(β) ∆z + Lneedles sin(β)
   
V1 Lneedles V2 V3 Lneedles V4 (66)
Φ(S) ≈ − − + −
G1 Lair G2 G3 Lair G4 
−∆x + Lneedles cos(β)
 
−∆x − Lneedles cos(β)

(62) −h  , B3′ =  −h
B3 =  
In the classical case in which the magnetic flux is supposed −∆z + Lneedles sin(β) −∆z − Lneedles sin(β)
uniform inside the sample: (67)
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 14

It is assumed that the effective cross-section S can be

approximated by the superposition of many elementary seg- Z 1 Z 1
ments, whose angles are linearly dependent on the depth. Their Φz = B⃗z · n⃗z dadb
−1 −1
parametric description (of parameters a and b) is: Z 1 Z 1
= Bz hL cos(bβ)dadb
−1 −1
 Z 1
 x = aLneedles cos(bβ) + b∆x
 = 2hLBz cos(bβ)db
−1
P (x, y, z) ∈ S → y = bh (68)  1
 sin(bβ)
z = −aLneedles sin(bβ) + b∆z = 2hLBz

β −1
a ∈ [−1, 1], b ∈ [−1, 1]
Φz = 4hLBz sinc(β) (72)
As expected, the magnetic flux in the ideal case (for which
The normal ⃗n for each elementary surface can be calculated β = 0, ∆x = ∆z = 0) is the product of S = 2h · 2L and
by: Bz . In the case where the magnetic flux is entirely along the
Z direction, the relative error ϵ depends only on β and is
described as:
∂ P⃗ ∂ P⃗ RR
⃗ · ⃗ndS
⃗n = × B
∂a ∂b ϵ=1− S = 1 − sinc(β) (73)
Lneedles cos(bβ)
 
−aβLneedles sin(bβ) + ∆x
 4hLBz
= 0 × h 
−Lneedles sin(bβ) −aβLneedles cos(bβ) + ∆z B. Air needle probes liftoff

hLneedles sin(bβ)
 If Assumption 3 (see Eq.(19)) is not verified, the magnetic
⃗n = −aβL2needles + Lneedles [∆z cos(bβ) + ∆x sin(bβ)] flux crossing the air needle probes loop will be reduced
hLneedles cos(bβ) compared to the total air flux. If all other assumptions are
(69) verified:
Z eair
∂Bn
Hence the value of the magnetic flux Φ crossing the surface Φ(Sair t - ) ≈ Lair (y)dy (74)
∂t
S is: Z∆z
eair
∂Bn
Φ(Sair t ) ≈ Lneedles (y)dy (75)
0 ∂t
Z 1 Z 1
Since the magnetic flux in the air diminishes as a function of
ΦX = BX⃗ · n⃗X dadb the distance from the sample, even a small δ can significantly
−1 −1 reduce the measured air flux.
Z 1 Z 1 The expression for Bn (y) strongly depends on the magnetic
= BX Lh sin(bβ)dadb flux distribution inside the sample. However, it can be expected
−1 −1
 1 that the error will quickly increase as a function of δ. Hence,
cos(bβ)
= 2hLBX − it is crucial to minimize the air needle probes liftoff.
β −1
ΦX = 0 (70) C. Air-sample angle mismatch
In the case the needle probes holder is not precisely built,
there could be a mismatch between the ”air” and the ”sample”
needle probes. As seen before, only the angular mismatch is
Z 1 Z 1 able to affect the measurement. The configuration associated
ΦY = B⃗Y · n⃗Y dadb to this error is shown in Fig. 25.
−1 −1
Z 1Z 1
= BY Lneedles −βaLneedles
−1 −1
+ ∆z cos(bβ) + ∆x sin(bβ)dadb
Z 1
= 2BY Lneedles ∆z cos(bβ) + ∆x sin(bβ)db
−1
 1
sin(bβ) cos(bβ)
= 2BY Lneedles ∆z − ∆x
β β −1 Fig. 25: Lengths and angle definition for the air-sample angle
ΦY = 4BY Lneedles ∆z sinc(β) (71) mismatch.
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 15

The normal to the air surface (defined by T2 and T2′ ) on the In the general case, one can return to the case αX = 0 by
XZ plane is given as [sin(γ); cos(γ)], hence: rotating the X axis, hence this case will be considered in this
section.
Φeffective (Sair t - ) = Φideal (Sair t - ) cos(γ) (76) In RUV , and taking into account ∆G and ∆α, B ⃗ becomes:
Like the liftoff, the air-sample angle mismatch will merely
add a constant coefficient to the measured air flux, which will
(
BU (t) = BX (t)
result in an increased drift on the experimental B(H) loops:
BV (t) = (1 + ∆G) [−BX (t) sin(∆α) + BY (t) cos(∆α)]
dΦmeas (80)
Vind = (77)
dt (
Φmeas = Φ(S) + [1 − cos(γ)]Φ(Sair t ) (78) BU (t) = B(t) cos(α)
BV (t)
However, in the large majority of cases, the error generated 1+∆G = −B(t) sin(∆α) cos(α) + B(t) cos(∆α) sin(α)
by the air-sample angle mismatch can be ignored, since (81)
cos(γ) > 0.99 for |γ| < 8°. Only evident alignment issues (
can have a noticeable effect on the measured flux. BU (t) = B(t) cos(α)
(82)
BV (t) = (1 + ∆G) [B(t) sin(α − ∆α)]
D. Gain mismatch and non-orthogonality angle
Two main sources of error in the 2D reconstruction have ⃗ and angle α of B
The norm ||B|| ⃗ in the two frames will
been identified: be:
• the gain mismatch ∆G between two inputs, linked to the
(
fact that two sets of each electronic component (whose ⃗ XY = |B(t)|
||B||
gain can be regulated thanks to resistors) will more often (83)
αXY = α
than not have different performances (or values, in the
case of resistors);
• the non-orthogonality angle ∆α between the two mea-
 q
 ⃗ U V = |B(t)| cos2 (α) + (1 + ∆G)2 sin2 (α − ∆α)
||B||
surement axis. This angle can have a noticeable value, h i
since the needle distance is usually small. For example,  αU V = atan (1 + ∆G) sin(α−∆α)
cos(α)
for a needle distance of 10 mm and a positioning uncer- (84)
tainty of 0.5 mm, then ∆α = atan(0.5/10) = 2.8°. Hence, if ∆G and ∆α are not zero, both ||B|| ⃗ and αU V
The measured 2D magnetic field can be described in two will change in function of the excitation angle α. Even if, in
frames of reference: the 1D case, the difference αU V will be constant since α is
• the external orthogonal frame of reference, of axis X and ⃗
fixed, in the case of rotating magnetic fields the measured B
Y: RXY ; will be distorted, which requires to measure experimentally
• the needle probes frame of reference, defined by the ∆G and ∆α to obtain an accurate induction measurement.
normal of the two pairs of probes U and V: RUV . The magnitude and angle distortions are noted ϵM and ϵA and
RUV is not orthogonal a priori, the angle between U and V are given below:
being π2 − ∆α. Moreover, several other angles can be defined:
the angle between the U and X axis αX and between the V and
 ⃗
 ϵM (α)
 = |||B(t)|
B||U V
−1
Y axis αY and the angle between the magnetic field and the  p
= cos (α) + (1 +h ∆G)2 sin2 (α − ∆α)i− 1
2 (85)
X axis (or the U axis) θ (or α). This also implies that θ = α
= αU V − α = atan (1 + ∆G) sin(α−∆α)

 ϵA (α) −α

and αY = ∆α. The different angles are shown in Fig. 26. cos(α)

These equations can not easily be studied in the general

case, but one can assume that ∆G and ∆α are ”small enough”
(for r2 > 0.98, |∆G| < 0.2 and |∆α| < 5°). In this case,
Taylor series to the first order are used to simplify ϵM and
ϵA :

(1 + x)2 ≈ 1 + 2x
sin(α − x) ≈ sin(α) − x cos(α)
Fig. 26: Angle definitions. sin2 (α − x) ≈ sin2 (α) − x sin(2α)
√ x
⃗ is assumed 1D, and can be described
The magnetic field B 1+x≈1+
2
in RXY as follows: x
( atan(α + x) ≈ atan(α) + 2
BX (t) = B(t) cos(α) α +1
(79)
BY (t) = B(t) sin(α) Hence:
IEEE TRANSACTIONS ON MEASUREMENTS AND INSTRUMENTATION , VOL. XX, NO. X, MONTH 2024 16

that αX = 0, the U and V signals can be expressed in the XY

q reference frame as follows:
ϵM ≈ cos2 (α) + (1 + 2∆G)(sin2 (α) − ∆α sin(2α)) − 1
q (
≈ 1 − ∆α sin(2α) + 2∆G sin2 (α) − 1 BU =B⃗ · e⃗U
(91)
∆α ∆G BV =B⃗ · e⃗V
≈− sin(2α) + [1 − cos(2α)]  ! !
2 2  BU’ 1
∆G 1  BU = ·


ϵM ≈ − [∆G cos(2α) + ∆α sin(2α)] (86)

BV’ 0
2 2 ! ! (92)
 BU’ − sin(∆α)
 BV = ·



BV’ cos(∆α)
ϵA ≈ atan [(1 + ∆G)(tan(α) − ∆α)] − α (
BU = BU’
≈ atan[tan(α) − ∆α + ∆G tan(α)] − α (93)
−∆α + ∆G tan(α) BV = −BU’ sin(∆α) + BV’ cos(∆α)
≈
tan2 (α) + 1 Inverting Eq. (93) gives:
≈ −∆α cos2 (α) + ∆G sin(α) cos(α) (
BU’ = BU
∆α ∆G (94)
≈−
2
(1 + cos(2α)) +
2
sin(2α) BV’ = BV +B U sin(∆α)
cos(∆α)
∆α 1 In the case αX ̸= 0, once BU’ and BV’ have been obtained,
ϵA ≈ − + [∆G sin(2α) − ∆α cos(2α)] (87)
2 2 a rotation of −αX brings the measurements to RXY :
Eq. (86) and (87) are interesting for several reasons: (
• both functions have an offset proportional to either ∆α
BX = BU’ cos(αX ) − BV’ sin(αX )
(95)
or ∆G; BY = BU’ sin(αX ) + BV’ cos(αX )
• both functions are π-periodic, which is expected given
that a rotation of π is the needles in the same position,
but with the positive and negative needles inverted;
• one function can be determined from the other thanks to
a rotation of 45° and an offset:
∆G  π  ∆α
ϵM (θ) − = ϵA θ + + (88)
2 4 2
Their values for specific θ are given below.

TABLE VII: Values of the error functions ϵM and ϵA .

θ(°) 0 45 90 135
1 1
ϵM 0 2
(∆G − ∆α) ∆G 2
(∆G + ∆α)
1
ϵA -∆α 2
(∆G − ∆α) 0 − 12 (∆G + ∆α)

The interest of these approximations is that ϵA can easily

be bounded if ∆α and ∆G are known:

 
∆α 1 p 
min ϵA (θ) + =− |∆α| + ∆α2 + ∆G2 (89)
2 2
 
∆α 1 p 
max ϵA (θ) + = |∆α| + ∆α2 + ∆G2 (90)
2 2
In most cases, ∆α should be negligible compared to ∆G:
for example, a gain mismatch of 10% will be matched by
∆α ≈ 6°, which can be easily avoided by carefully manufac-
turing the needle probes support.

E. Signal treatment for non-orthogonality correction

If the needle probe normals are not orthogonal (the case
∆α ̸= 0 seen previously), the calculation of the 2D com-
ponents of B ⃗ on the XY reference frame is not merely a
rotation of the signals on the UV reference frame. Assuming