Electronics Guide

Magnetostatics

Magnetostatics is the study of the magnetic fields that steady electric currents produce and of the forces those fields exert. It is the physics inside every inductor, transformer core, relay, loudspeaker, motor, and Hall-effect sensor, and it sets the loop inductance that governs switching noise on a circuit board.

Strictly, magnetostatics assumes currents that do not change, but, like electrostatics, it stays accurate for structures much smaller than a wavelength, where the field at each instant has nearly its static shape. The inductance of a via, a short trace, or a winding is therefore a magnetostatic calculation even when the current switches in nanoseconds; the voltage that a changing current induces belongs to Faraday's law of induction, which marks where magnetostatics ends. This article is the companion to Electrostatics and follows a parallel path. Quantities are in SI units, and the text states the direction of each vector quantity in words.

The Lorentz Force and Magnetic Flux Density

In 1820 Hans Christian Oersted reported that a compass needle near a current-carrying wire swings toward a position at right angles to the wire. That autumn, Jean-Baptiste Biot and Félix Savart found that the force of a long straight current on a magnetic pole falls inversely with distance from the wire, the result behind the law that bears their names. In the same year André-Marie Ampère showed that parallel wires attract when their currents flow in the same direction and repel when the currents are opposed.

The magnetic flux density B is defined by the force it exerts on a moving charge. A charge q moving with velocity v through an electric field E and a magnetic flux density B experiences the Lorentz force

F = q(E + v × B)

The magnetic part, the cross product qv × B, has magnitude qvB sin θ, where θ is the angle between v and B, and it acts at right angles to both. With the fingers of the right hand pointing along v and curled toward B, the thumb points along the force on a positive charge; a negative charge is pushed the opposite way.

The SI unit of flux density is the tesla: a charge of 1 C moving at 1 m/s at right angles to a 1 T field feels a force of 1 N. One tesla is one weber per square meter, and the older CGS unit, the gauss, is exactly 10−4 T. Earth's magnetic field at the surface is roughly 25 to 65 µT, according to NOAA's National Centers for Environmental Information.

Because the magnetic force is always perpendicular to the velocity, it does no work: it bends a charge's path without changing its speed.

The Biot–Savart Law

The Biot–Savart law gives the flux density that a steady current produces. A short element of wire of length dl carrying current I contributes, at a point a distance R away,

dB = μ0I dl × aR / (4πR2)

where aR is a unit vector pointing from the element toward the field point. The contribution is perpendicular both to the element and to the line joining it to the point, and it falls as the inverse square of distance. Contributions add as vectors, so the field of a circuit is the integral of this expression around it; for current spread through a volume, I dl becomes J dv, where J is the current density in amperes per square meter. Unlike a point charge, a current element cannot exist on its own, because steady currents flow in closed circuits, so the law has physical meaning only when integrated around a complete circuit.

The constant μ0 is the vacuum magnetic permeability. Before the 2019 revision of the SI it was exactly 4π × 10−7 H/m, a consequence of the old definition of the ampere. It is now a measured quantity, tied to the fine-structure constant: its CODATA 2022 value, published by NIST, is 1.25663706127 × 10−6 H/m, with a relative standard uncertainty of 1.6 × 10−10. That is about 1.3 parts in 1010 below the old value, within the uncertainty, so engineering calculations may use either. The electric and magnetic constants are linked by μ0ε0 = 1/c2, where c is the speed of light.

A Straight Wire

At a point on the perpendicular bisector of a straight segment of length L, a distance r from it, integrating along the segment gives

B = μ0IL / (2πr√(L2 + 4r2))

As L grows without limit, this becomes the field of a long straight wire:

B = μ0I / (2πr)

The field lines are circles centered on the wire; with the right thumb pointing along the current, the fingers curl in the direction of B. A wire carrying 10 A produces 200 µT at 1 cm and 20 µT at 10 cm, where it is comparable to Earth's field. A segment only twice as long as the distance r already gives 71 percent of the long-wire value.

A Circular Loop

On the axis of a circular loop of radius R, at a distance z from its center, the field components perpendicular to the axis cancel by symmetry and the axial components add:

B = μ0IR2 / [2(R2 + z2)3/2]

The field points along the axis in the direction the right thumb points when the fingers curl with the current. At the center, B = μ0I/(2R), so 1 A in a loop of 1 cm radius produces about 63 µT, and N closely wound turns multiply that by N.

The Magnetic Dipole

Far from a small current loop, its field depends only on its magnetic dipole moment m = NIA, in ampere square meters, where A is the loop area; m points along the loop's axis by the same right-hand rule. On the axis, at a distance r much greater than the loop's size,

B = μ0m / (2πr3)

pointing along m. In the plane of the loop the field has half this magnitude and points opposite to m. As with the electric dipole, the field falls as the inverse cube of distance: the 1 cm loop above, carrying 1 A, produces about 63 nT on its axis at 10 cm and 8 nT at 20 cm. Small current loops on a circuit board are dipoles of this kind, so magnetic coupling between them grows with their areas and falls quickly with separation.

Magnetic Flux and Gauss's Law for Magnetism

The magnetic flux through a surface S is

Φ = ∫S B · dS

where dS is a small patch of the surface oriented along its normal. The unit is the weber, one tesla square meter; the CGS maxwell is exactly 10−8 Wb. Gauss's law for magnetism states that the net flux out of any closed surface S is zero:

S B · dS = 0

In differential form,

div B = 0

Electric flux begins and ends on charge, but no isolated magnetic pole, or monopole, has been confirmed experimentally, so magnetic flux has no sources or sinks. Lines of B close on themselves or continue without end. Like its electric counterpart, this law holds for time-varying fields and is one of Maxwell's equations.

One consequence underlies inductance. Two surfaces bounded by the same closed path together form a closed surface, so the same flux passes through both, and the flux linking a circuit depends only on the circuit's path. For a coil of N turns, each linked by flux Φ, the flux linkage is Λ = NΦ.

Ampère's Circuital Law

The Biot–Savart law handles any steady current in a uniform medium, but Ampère's circuital law is faster when the geometry is symmetric, as Gauss's law is in electrostatics. It uses the magnetic field intensity H, also called the magnetic field strength, in amperes per meter. In vacuum, H = B/μ0; the section on magnetization explains the relationship in materials.

The Integral Form

For steady currents, the line integral of H around any closed path C equals the free current passing through any surface bounded by the path:

C H · dl = Ienc

With the fingers of the right hand curling along the direction of travel around C, the thumb points in the direction of positive current. A coil of N turns threading the path contributes NI, which is why magnetomotive force is counted in ampere-turns. The law holds for every path, but it yields H directly only when symmetry provides a path along which H is either tangent with constant magnitude or perpendicular to the path.

For a long straight conductor of radius a carrying current I, a circle of radius r outside it encloses the whole current, so H × 2πr = I and H = I/(2πr), matching the Biot–Savart result. A circle inside encloses the fraction r2/a2 of a uniform current, so H = Ir/(2πa2), and that internal field gives the wire its internal inductance.

The Solenoid and the Toroid

A long, tightly wound solenoid with n turns per meter has a nearly uniform axial field inside and almost none outside. A rectangular path with one long side inside the coil, parallel to its axis, and the other outside encloses nI for each meter of its inside side, so inside the coil

B = μ0nI

independent of position across the bore; 1,000 turns per meter carrying 1 A give about 1.26 mT.

In a toroid of N turns, a circle of radius r within the core encloses the current NI, so

H = NI / (2πr)

and B = μH, where μ is the core's permeability. The field is strongest at the inner radius, where a toroidal core saturates first. A circle outside an ideal toroid encloses no net current, so the field there is zero, which is why toroidal inductors produce little stray field.

The Coaxial Cable

A coaxial cable carries current I on an inner conductor of radius a and returns it on a shield with inner radius b and outer radius c. With the current spread uniformly over each conductor, circular paths give H in four regions:

  • Inside the inner conductor, with r less than a: H = Ir/(2πa2).
  • Between the conductors: H = I/(2πr).
  • Within the shield: H = I(c2 − r2) / [2πr(c2 − b2)].
  • Outside the shield, with r greater than c: the enclosed current is zero, so H = 0.

The field circles the axis in the direction set by the inner conductor's current. When the shield carries the full return current, the cable produces no external magnetic field. If part of the return current flows elsewhere, such as through a chassis, a path around the cable encloses a net current, and the cable produces that current's field.

Fields from Ampère's circuital law
Geometry Field intensity H Conditions
Long straight conductor, current I I/(2πr) Outside the conductor
Inside a round conductor of radius a Ir/(2πa2) Uniform current density
Long solenoid, n turns per meter nI, along the axis Inside, far from the ends
Toroid, N turns NI/(2πr) Inside the winding
Coaxial cable, between the conductors I/(2πr) Zero outside when the shield carries the full return current

The Differential Form and Its Limits

Applied to a vanishingly small loop, Ampère's law becomes

curl H = J

where curl H measures the circulation of H per unit area; in Cartesian coordinates its z component, for example, is ∂Hy/∂x − ∂Hx/∂y.

The divergence of any curl is zero, so the law requires div J = 0, the field form of Kirchhoff's current law. Where charge builds up, as on the plates of a charging capacitor, the law fails. James Clerk Maxwell repaired it by adding the displacement current density, the rate of change of the electric flux density D, to give curl H = J + ∂D/∂t.

Faraday's law marks the other boundary: a changing flux induces a circulating electric field, curl E = −∂B/∂t, which couples the electric and magnetic fields. Magnetostatic results stay useful while a structure is much smaller than a wavelength and while eddy currents in nearby conductors are negligible or handled separately. At high frequency, where the skin effect crowds current toward conductor surfaces, inductance can still be found magnetostatically by placing the currents on those surfaces.

Magnetization, Permeability, and Hysteresis

Magnetization and the Two Magnetic Fields

Electrons carry magnetic moment from their orbital motion and their spin. In most materials these moments cancel or point randomly, but an applied field partly aligns them. The magnetic dipole moment per unit volume is the magnetization M, in amperes per meter. Aligned atomic moments act like tiny current loops: where M is uniform, their currents cancel inside the material and leave a bound current circulating on the surface, with surface density M × n, where n is the outward normal. Where M varies, a bound current density curl M also flows inside. Flux density responds to free and bound currents alike:

B = μ0(H + M)

so that curl H = J counts only free current, the current in windings that an engineer controls. The split mirrors D = ε0E + P in electrostatics: the ampere-turns fix the circulation of H, while B, which exerts force and induces voltage, also depends on the material. In a linear, isotropic material, M = χmH, where χm is the magnetic susceptibility, and

B = μ0(1 + χm)H = μ0μrH = μH

The relative permeability μr = 1 + χm is the factor by which a material multiplies the flux, and so the inductance, of a winding whose field it completely fills. Older literature gives H in oersteds; NIST Special Publication 811 lists 1 Oe as about 79.58 A/m, which equals 1,000/(4π) A/m.

Diamagnetic, Paramagnetic, and Ferromagnetic Materials

Most materials respond so weakly that μr differs from 1 by parts per hundred thousand. Diamagnetic materials develop moments that oppose the applied field, a response every material has, while paramagnetic materials, whose atoms carry permanent moments, align slightly with it. Georgia State University's HyperPhysics tabulates χm at 20 °C as −1.0 × 10−5 for copper and −0.91 × 10−5 for water, against +2.2 × 10−5 for aluminum. For field calculations, copper, aluminum, and ordinary insulators are therefore nonmagnetic. A superconductor, which expels the field from its interior, is the extreme case: a perfect diamagnet, with χm = −1.

In iron, nickel, cobalt, and many of their alloys, a quantum-mechanical exchange interaction aligns neighboring atomic moments spontaneously within regions called domains. In an unmagnetized sample the domains largely cancel. An applied field enlarges the domains aligned with it and rotates the others, producing a magnetization many orders of magnitude larger than a paramagnet's. Above the Curie temperature, about 1043 K (770 °C) for iron, thermal agitation destroys the ordering. Ferrites are ferrimagnetic: two sublattices with opposing but unequal moments leave a net magnetization, and their high electrical resistivity suits them to high frequencies.

Hysteresis and Saturation

A ferromagnet's B is neither proportional to H nor a single-valued function of it. Raising H from zero in an unmagnetized sample traces the initial magnetization curve, along which the ratio B/(μ0H) gives the initial relative permeability near the origin and the maximum relative permeability where that ratio peaks. At high H nearly all the moments align, and B then rises only as fast as μ0H: the material saturates. Reducing H to zero leaves a remanent flux density Br, driving H negative to the coercive field Hc returns B to zero, and cycling H traces the closed hysteresis loop.

Properties of selected ferromagnetic materials
Material and treatment Initial μr Maximum μr Coercive field Hc Remanence Br
Iron, 99.8 percent pure, annealed1505,00080 A/m1.3 T
Iron, 99.95 percent pure, annealed in hydrogen10,000200,0004 A/m1.3 T
78 Permalloy, annealed and quenched8,000100,0004 A/m0.7 T
Superpermalloy, annealed in hydrogen with controlled cooling100,0001,000,0000.16 A/m0.7 T
Alnico 5, cooled in a magnetic field4Not listed46 kA/m1.25 T

The table converts HyperPhysics values from oersteds and gauss. Soft magnetic materials, with small coercive fields and narrow loops, suit cores and shields, and their highest permeabilities appear only after careful annealing. Hard materials suit permanent magnets because they resist demagnetization: the coercive field of Alnico 5 is more than 10,000 times that of 78 Permalloy.

The energy supplied per unit volume as B changes is the integral of H dB, so the area of the loop, in joules per cubic meter, is the energy converted to heat in each cycle, the hysteresis part of core loss. Permeability also varies with flux level, frequency, temperature, and mechanical stress, so a single value of μr is an idealization. Magnetic Materials covers practical core and magnet materials and their loss and temperature behavior.

Boundary Conditions

Where two materials meet, the magnetic fields obey two rules that parallel those of electrostatics:

  • Normal B is continuous. A pillbox straddling the boundary encloses no net flux, so Bn1 = Bn2.
  • Tangential H jumps by the surface current. A thin rectangular path straddling the boundary gives Ht1 − Ht2 = K, where K is the free surface current per unit width and the components compared lie along the surface at right angles to K; with the unit normal n pointing from material 2 into material 1, the jump points along K × n. Real conductors carry current through a finite thickness rather than a sheet, so tangential H is normally continuous.

At an interface without surface current, the rules bend flux lines. With angles measured from the normal,

tan θ1 / tan θ2 = μ1 / μ2

Flux crossing from a high-permeability material into air therefore leaves almost perpendicular to the surface. If μr = 1,000 and the flux inside meets the surface at 89° from the normal, the flux emerges into air only 3.3° from the normal. As permeability grows, tangential H at the surface falls toward zero, so the pole faces of a core behave like surfaces of constant magnetic potential, and their shape sets the field in the gap between them.

Images in Iron and in Conductors

The method of images works here too. A long wire carrying current I at height h, parallel to the flat surface of a material with relative permeability μr, produces above the surface the field of itself plus an image current I(μr − 1)/(μr + 1) at depth h, flowing in the same direction, so the wire is drawn toward the iron. A thick, highly conductive plane does the opposite when the wire carries high-frequency alternating current: induced eddy currents keep the field out, the image current is equal and opposite, and the wire is repelled. That second case describes a circuit-board plane at high frequency, as the applications section shows.

Inductance from Geometry

The self-inductance of a circuit is its flux linkage per unit current:

L = Λ / I

The unit is the henry, one weber per ampere. In linear materials L depends only on geometry and permeability, and a procedure parallel to the one for capacitance finds it:

  1. Assume a current I in the circuit.
  2. Find H, using Ampère's law where the geometry is symmetric or the Biot–Savart law otherwise.
  3. Find B = μH and integrate it over a surface bounded by the circuit, counting each part of the flux once for every turn it links.
  4. Divide the flux linkage by I; the assumed current cancels.

Alternatively, find the stored energy W and use L = 2W/I2, which also handles flux that links only part of a winding, such as the field inside a conductor.

Solenoids and Toroids

For an air-core solenoid of N turns, length l much greater than its radius, and cross-sectional area A, every turn links the flux μ0NIA/l, so

L = μ0N2A / l

Inductance grows as the square of the turns because each added turn both adds field and links it. The formula ignores the weaker field near the ends and so overestimates short coils. Harold Wheeler's approximation for single-layer air-core coils, published in the Proceedings of the IRE in 1928, is equivalent to within 0.3 percent to replacing l with l + 0.9r, where r is the coil radius; Wheeler gave its accuracy as about 1 percent for coils longer than 0.8r. For a coil of 100 turns, 5 mm radius, and 50 mm length, the long-solenoid formula gives 19.7 µH and the corrected one 18.1 µH.

For a closely wound toroid of N turns on a linear core of rectangular cross-section with height h, inner radius a, and outer radius b, integrating B = μNI/(2πr) across the section gives

L = μN2h ln(b/a) / (2π)

Transmission Lines and Wire Pairs

For a coaxial line with inner radius a and shield radius b, integrating μI/(2πr) from a to b gives the flux per meter between the conductors, and so the external inductance per unit length:

L′ = (μ/2π) ln(b/a)

The polyethylene-filled line used as an example in Electrostatics, with a = 0.45 mm and b = 1.5 mm, has L′ ≈ 241 nH/m. For any two-conductor line in a uniform medium, with currents on the conductor surfaces, the external inductance and the capacitance per unit length obey

L′C′ = με

so a line's inductance and capacitance are two views of one cross-section. With that line's 104 pF/m, the characteristic impedance √(L′/C′) is about 48 Ω, and the propagation speed 1/√(L′C′) is two-thirds of the speed of light.

Inductance of lines and loops
Geometry Inductance Conditions
Coaxial line, radii a and b (μ/2π) ln(b/a) per meter Currents on conductor surfaces
Two parallel wires, radius a, center spacing D (μ/π) cosh−1(D/(2a)) per meter Currents on conductor surfaces; near (μ/π) ln(D/a) when D is much larger than a
Wire of radius a, axis at height h above a ground plane (μ/2π) cosh−1(h/a) per meter Currents on conductor surfaces; near (μ/2π) ln(2h/a) when h is much larger than a
Parallel strips, width w, spacing h μh/w per meter Width much greater than spacing; fringing neglected
Circular loop, radius R, wire radius a μ0R[ln(8R/a) − 2] R much greater than a; current on the wire surface
Straight segment, length l, radius a (partial inductance) 0l/2π)[ln(2l/a) − 1] l much greater than a; current on the wire surface

Spacing matters logarithmically for wires and linearly for strips. Two wires 0.5 mm in diameter with centers 5 mm apart have about 1.20 µH/m, and moving them to 1 mm apart lowers that only to 0.53 µH/m, while strips 10 mm wide separated by 0.1 mm of insulation have 12.6 nH/m. That is why high-current converters use laminated busbars.

Internal Inductance, Loops, and Partial Inductance

A nonmagnetic round wire carrying uniformly distributed current also links flux inside its own metal, which adds an internal inductance of μ0/(8π), or 50 nH/m, whatever the radius. The skin effect removes it at high frequency; with uniform current, the bracketed terms in the loop and segment formulas become ln(8R/a) − 7/4 and ln(2l/a) − 3/4.

A single-turn loop 20 mm in diameter, made of wire 0.5 mm in diameter, has about 47 nH at high frequency. Strictly, inductance belongs only to a closed circuit, because only a closed path bounds a surface for the flux. Circuit-board and package designers nonetheless assign inductance to pieces of a loop, such as vias, bond wires, and trace segments, using partial inductance, a method that A. E. Ruehli set out in the IBM Journal of Research and Development in 1972. The partial self-inductances of the pieces, plus the partial mutual inductances between every pair, sum to the loop inductance. A straight segment 10 mm long and 0.5 mm in diameter has a partial self-inductance of about 7 nH, but a loop containing it may have more or less, depending on where the return current flows.

Mutual Inductance as a Field Quantity

The mutual inductance of two circuits is the flux linkage in one per unit current in the other, M21 = Λ21/I1. For thin circuits in a uniform medium of permeability μ, Franz Neumann's formula gives it as a double line integral around both circuits:

M = (μ/4π) ∫C1C2 (dl1 · dl2) / R

Here R is the distance between the elements dl1 and dl2. The formula is symmetric in the two circuits, so M12 = M21. M depends only on geometry and permeability, its sign depends on the current directions chosen as positive, and the coupling coefficient k = M/√(L1L2) never exceeds 1 in magnitude. Circuit models of coupled windings and transformers build on this quantity.

Energy Stored in the Magnetic Field

Building up current I in an inductor with linear materials takes work against the voltage its changing flux induces, and integrating gives the stored energy:

W = ½LI2 = ½ΛI

The energy resides in the field. For a long air-core solenoid, ½LI2 equals B2/(2μ0) times the volume Al inside the coil, and in general, for a linear material, the energy density in joules per cubic meter is

w = B2/(2μ) = ½B · H

At a given flux density, energy density is inversely proportional to permeability. Flux crosses a core and its air gap at nearly the same density, so the gap stores μr times the energy density of the core material beside it, which is why a gapped inductor stores most of its energy in the gap. Inductors and Magnetic Components turns this into a sizing rule for gap volume.

Air at 1 T holds about 398 kJ/m3, against about 40 J/m3 at 3 MV/m, near the breakdown strength of air. That factor of 10,000 is why motors, relays, and loudspeakers are magnetic, while electrostatic actuation pays off mainly at micrometer scale.

Forces and Torque on Conductors

Force on a Current-Carrying Conductor

An element dl of a conductor carrying current I feels a force dF = I dl × B. A straight conductor of length l at an angle θ to a uniform flux density B therefore feels BIl sin θ, directed along l × B, where l points along the current. Motor windings and loudspeaker voice coils work by this force; a loudspeaker's force factor is the product Bl, in newtons per ampere, where l is the length of wire in its magnetic gap.

Forces Between Parallel Conductors

Each of two long parallel wires a distance d apart sits in the other's field, B = μ0I/(2πd), so the force per unit length on each is

F/l = μ0I1I2 / (2πd)

Currents in the same direction attract, and opposite currents repel. Until May 20, 2019, the SI defined the ampere by this force, as the constant current that, maintained in two straight parallel conductors of infinite length and negligible circular cross-section placed 1 m apart in vacuum, would produce between them a force of 2 × 10−7 N per meter of length. The force grows as the square of the current, which matters during faults: two round conductors 10 cm apart carrying an instantaneous 50 kA in opposite directions push apart with 5 kN per meter, so busbar supports are designed for the peak short-circuit current.

Magnetic Pressure and Virtual Work

Forces on iron parts and on complicated windings follow most easily from energy. For a single circuit in linear materials, if a part moves a distance dx while the current I is held constant, the force on it, positive in the direction of increasing x, is

F = ½I2 dL/dx

so magnetic forces act to increase inductance: a relay armature closes its gap, a plunger is drawn into a solenoid, and a coil's turns are pushed outward and pressed together along its axis. Where flux crosses from high-permeability iron into an air gap, the field pulls the facing surfaces together, whichever way the flux points, with a pressure equal to the energy density in the gap:

p = B2/(2μ0)

At 1 T that is about 398 kPa, nearly four atmospheres. A relay pole face of 20 mm2 with 0.5 T in the gap pulls with about 2 N.

Torque on a Current Loop

A flat coil with dipole moment m = NIA in a uniform field feels no net force but a torque

τ = m × B

of magnitude NIAB sin θ, where θ is the angle between the coil's axis and the field. The torque turns the coil to align m with B, the position of lowest potential energy, −m · B. A 100-turn coil of 1 cm2 carrying 10 mA in 0.2 T feels at most 20 µN·m. Moving-coil meters balance this torque against a spring.

The Vector Potential and Field Solvers

Because div B = 0, flux density can be written as B = curl A, where A is the magnetic vector potential, in webers per meter. For currents in a uniform medium,

A = ∫ μJ dv / (4πR)

which parallels the electric potential of a charge distribution: each current element contributes potential along its own direction, falling as 1/R. The flux through a surface equals the line integral of A around its edge, Φ = ∫C A · dl, so A gives flux linkage directly. Substituting B = curl A into curl H = J, with A chosen so that div A = 0, gives, for uniform permeability, a Poisson equation for each Cartesian component:

2Ax/∂x2 + ∂2Ax/∂y2 + ∂2Ax/∂z2 = −μJx

and likewise for y and z. The magnetostatic problem thus has the mathematical form of electrostatics, with μJ in place of ρv/ε, and the analytical and numerical methods described in Electrostatics carry over. In a current-free region of uniform permeability, H can also be written as −grad Vm, where the magnetic scalar potential Vm obeys Laplace's equation.

Planar and Axisymmetric Problems

When all currents flow perpendicular to a cross-section, as in long conductors, A has only the component Az along that direction. B then lies in the cross-section, tangent to the lines of constant Az, so those contours are flux lines, and the flux per meter of depth passing between two points equals the difference of their Az values. In axisymmetric problems A circles the axis, and contours of r times A are the flux lines. FEMM, a finite element program by David Meeker, solves linear and nonlinear magnetostatic problems in this way on two-dimensional planar and axisymmetric domains. Nonlinear B-H curves let it model saturating iron, at the cost of iterating until the permeability everywhere agrees with the local flux density.

Extracting and Checking Inductance

Solvers find inductance from stored energy, L = 2W/I2, or from flux linkage computed with A. Interconnect that no single cross-section represents needs three-dimensional extraction; FastHenry, a multipole-accelerated 3-D inductance extraction program described by M. Kamon, M. J. Tsuk, and J. K. White in IEEE Transactions on Microwave Theory and Techniques in 1994, is one such tool.

The checks described in Electrostatics apply. Also confirm that every current has a return path: a planar model of a single conductor with no return has no finite inductance per meter, so its computed value depends on where the outer boundary lies.

Magnetic Circuits

Ampère's law applied along the center line of a core with an air gap gives NI = Hclc + Hgg, where lc is the path length in the core and g the gap length. If the gap is short compared with the width of its faces, little flux fringes or leaks, and because normal B is continuous across the faces, B is nearly the same in core and gap. With B = μ0μrHc in a linear, unsaturated core and B = μ0Hg in the gap,

B = μ0NI / (g + lcr)

The gap usually dominates. With one turn carrying 10 A, a 1.5 mm gap, and a 50 mm core path with μr = 2,000, B is 8.24 mT, only 1.6 percent below the 8.38 mT the gap alone would give. Multiplying by the core's cross-sectional area A gives the flux, Φ = NI/(Rc + Rg), with reluctances Rc = lc/(μ0μrA) and Rg = g/(μ0A). This is the magnetic circuit, in which magnetomotive force drives flux through reluctances in series and parallel, as voltage drives current through resistances.

The analogy is approximate. Copper's conductivity exceeds that of a good insulator by a factor of more than 1020, but even the most permeable alloy in the table above reaches only 106 times the permeability of air, so flux leaks between core legs and fringes outward at gaps. Inductors and Magnetic Components develops the magnetic circuit into a design method, with effective core parameters, gap design, and fringing.

Applications in Electronics

Inductor and Transformer Design

Magnetostatics supplies the core equations of wound-component design: inductance from turns and reluctance, energy stored in the gap, and the peak flux density that must stay below saturation. Since flux linkage is both LI and NBA, the flux density in a core of area A is B = LI/(NA), a figure to check at the largest current the winding will carry.

Transformer leakage inductance is a field calculation too. Take two concentric windings on one core leg, each of breadth bw along the leg, with thicknesses a1 and a2, separated by insulation of thickness d, and with a mean turn length MLT. Under load their ampere-turns cancel, so by Ampère's law H is nearly zero outside the pair and equals NI/bw in the insulation, rising linearly across one winding and falling across the other. Integrating ½μ0H2 over this region and applying L = 2W/I2 gives, for a breadth much larger than the total build, the leakage inductance referred to the N-turn winding:

Llk = μ0N2(MLT)(d + (a1 + a2)/3) / bw

For 20 turns, a 60 mm mean turn, a 20 mm breadth, 0.5 mm of insulation, and windings 1 mm thick, Llk is about 1.8 µH. The levers are fewer turns, broader and thinner windings, thinner insulation, and interleaving, which halves the peak field between winding sections.

Current Sensing and Hall-Effect Sensors

Edwin H. Hall reported in the American Journal of Mathematics in 1879 that a magnetic field acting on a current-carrying conductor produces a voltage across it. The Lorentz force pushes the moving carriers toward one edge of the conductor until the charge that builds up there creates an electric field that balances the magnetic force. For a strip of thickness t carrying current I through a perpendicular flux density B, a simple model with one type of carrier gives the Hall voltage

VH = IB / (nqt)

where n is the carrier density and q the carrier charge; the polarity of the voltage reveals the carriers' sign. In copper, with 8.47 × 1028 free electrons per cubic meter, a strip 0.1 mm thick carrying 1 A in 1 T develops only about 0.74 µV. A semiconductor layer 10 µm thick with 1022 electrons per cubic meter gives 1/(nqt) ≈ 62 V per ampere-tesla, so a 1 mA bias in a 10 mT field produces about 0.6 mV. That is why Hall elements are made from semiconductors and paired with amplifiers. Basic Sensor Components describes linear Hall sensors and Hall switches.

Ampère's law makes a ring-shaped current sensor insensitive to where the conductor passes through its aperture, because the line integral of H around the ring equals the enclosed current wherever the conductor sits. In an open-loop sensor, a gapped ring of high-permeability material concentrates the field onto a Hall element in the gap: by the magnetic-circuit example, 10 A through the aperture of a core with a 1.5 mm gap gives about 8.2 mT. Core remanence, core nonlinearity, and the element's temperature drift limit its accuracy. A closed-loop, or zero-flux, sensor drives a secondary winding so that its ampere-turns cancel the primary's and the Hall element reads zero. The secondary current, smaller than the primary by the turns ratio, becomes the measurement, and the core stays near zero flux.

Coreless sensors place the element beside the conductor instead. The field 1 cm from a 10 A conductor, 200 µT, is only a few times Earth's field, so these sensors sit within millimeters of the current path or subtract the outputs of two elements, which cancels a uniform stray field. A Rogowski coil, an air-cored winding that encircles a conductor, uses the same Ampère's-law geometry, but its output is the voltage induced by the changing current, so it measures only alternating current.

Magnetic Shielding

A static magnetic field passes through copper and aluminum as if they were absent. Shielding at DC therefore relies on high-permeability material, which the boundary conditions make an easy path: flux entering the wall turns to run along it and bypasses the interior. For a long cylinder of diameter D with a thin wall of thickness t, in a field across its axis, the shielding factor, the ratio of the outside field to the inside field, is approximately

S ≈ 1 + μrt / D

when μr is large. With an assumed effective μr of 20,000, a 1 mm wall, and a 100 mm diameter, S is about 200, or 46 dB. Saturation, seams, holes, open ends, and mechanical stress after annealing all reduce it. Two shells separated by a gap attenuate far more than one shell with the same total wall thickness. At higher frequencies, eddy currents let ordinary conductors shield magnetic fields as well, the regime that Electromagnetic Shielding Fundamentals treats.

Canceling a field at its source often works better than shielding it. Two conductors carrying equal and opposite currents a distance d apart produce, at a distance r much greater than d,

B ≈ μ0Id / (2πr2)

which falls as the inverse square of distance rather than the inverse first power. A pair carrying 10 A with 1 mm spacing produces 0.2 µT at 10 cm, one-hundredth of the 20 µT from a single conductor. Twisting the pair reverses its residual field every half twist.

Loop Inductance on Circuit Boards

Every switching current flows around a loop whose inductance sets the voltage spike L di/dt. Model a trace as a thin wire at height h above a wide, solid plane, at frequencies high enough that the plane keeps the field out. The return current in the plane mirrors the trace, with a surface current density, at a lateral distance x from the point beneath the trace, of

K(x) = Ih / [π(h2 + x2)]

Half the return current flows within a distance h on either side of the trace, 80 percent within 3h, and 94 percent within 10h. This distribution minimizes the stored field energy, and with it the loop inductance, which is why high-frequency return current follows the trace; at DC it instead spreads along the paths of least resistance. Several layout rules follow:

  • Keep a solid reference plane beneath fast signals. A slot or split forces the return current to detour around it, enlarging the loop and its inductance.
  • Keep switching loops small. Place a switching regulator's input capacitor and switching transistors close together, with the return path on the adjacent layer.
  • Use thin dielectrics between power and ground planes. Like the parallel strips above, a plane pair has inductance proportional to its spacing.

Summary

Magnetostatics rests on two facts: steady currents produce flux density according to the Biot–Savart law, and magnetic flux has no sources, so div B = 0. Ampère's circuital law, the vector potential, and the description of magnetic materials follow, and from them come inductance from geometry, the energy density B2/(2μ), magnetic forces, and the magnetic circuit. Flux leaves high-permeability material almost at right angles, which lets cores guide flux and shields divert it. Loop area and the path of the return current set the inductance of every circuit, from a transformer winding to a trace over a plane.

When currents change, Faraday's law couples the magnetic field to the electric field of Electrostatics, and Maxwell's displacement current completes Ampère's law. The Electromagnetic Theory overview introduces the resulting equations and the waves they predict.

Related Topics