Hall-Effect Sensor

What a Hall-effect sensor actually does

Almost every sensor problem in engineering eventually reduces to "I have a magnetic field and I want a number." A magnet glued to a rotating shaft, the field around a busbar carrying 400 amps, the stray field of a permanent-magnet rotor, the Earth's 50-microtesla background — all of these carry information, and all of them are invisible to an ordinary voltmeter. The Hall-effect sensor is the cheapest, most rugged way to read that field directly and turn it into a voltage.

The trick is contactless. There is no wiper, no mechanical contact, no coil that has to see the field change. A small current flows through a thin plate, a magnetic field crosses it, and a voltage appears across the plate's two side edges. Nothing touches, nothing wears, and the reading is valid even when everything is standing still. That combination — solid-state, zero-speed-capable, and dirt cheap once integrated onto silicon — is why Hall sensors ship by the billions and hide inside crankshaft sensors, brushless motors, current clamps, joysticks, and the lid-closed detector in your laptop.

The mechanism — Lorentz force and charge pile-up

Start with a thin rectangular plate carrying a bias current I in the x-direction. The current is moving charge carriers; in an n-type semiconductor those are electrons drifting with velocity v. Now apply a magnetic field B in the z-direction, perpendicular to the plate face. Each moving carrier feels the Lorentz force:

F = q · (v × B)

With v along x and B along z, the cross product points along y — sideways. So the carriers are pushed toward one edge of the plate. They accumulate there, leaving the opposite edge positively charged. That charge separation sets up a transverse electric field E_y across the plate, and the field grows until the electric force on a carrier exactly cancels the magnetic force:

q · E_y = q · v · B      →      E_y = v · B

The steady-state potential across the two side contacts — the Hall voltage — is that field times the plate width w. Substituting the drift velocity in terms of current density gives the canonical result:

V_H = (I · B) / (n · q · t)        =  R_H · (I · B) / t

where  R_H = 1 / (n · q)   is the Hall coefficient
       I   = bias current
       B   = field component normal to the plate
       n   = carrier density
       q   = carrier charge (1.6e-19 C)
       t   = plate thickness

Three things fall straight out of that one equation. First, the signal is linear in both the field B and the drive current I — double either and you double the output, which is what makes the device a clean analog transducer. Second, the sign of V_H flips with the field direction, so a Hall sensor distinguishes a north pole from a south pole, something a coil or a reed switch cannot do natively. Third, and most usefully for engineers, the signal scales as 1/(n·t): you want a thin plate of a low-carrier-density material.

Why semiconductors, not copper — the carrier-density numbers

Edwin Hall's original 1879 experiment used a gold leaf, and the effect was so small he needed a sensitive galvanometer to see it at all. The reason is carrier density. A good metal like copper has about 8.5e28 conduction electrons per cubic metre — an enormous reservoir — so its Hall coefficient is minuscule and a copper plate gives microvolts even in a strong field. Doped silicon has a carrier density a billion-fold lower, around 10^21 to 10^22 per cubic metre, so the same geometry gives millivolts.

Work a concrete example. Take an integrated silicon Hall plate with n = 1e22 /m³, thickness t = 1 µm, biased at I = 1 mA, in a field of B = 1 T:

V_H = (I · B) / (n · q · t)
    = (1e-3 · 1) / (1e22 · 1.6e-19 · 1e-6)
    = 1e-3 / 1.6e-3
    ≈ 0.6 V   per tesla   (idealised plate)

Real integrated plates run thicker and lower-current,
giving a raw sensitivity of roughly 5–10 mV per tesla,
which the on-chip amplifier multiplies by 100–1000×.

That is why the right material matters so much. The headline materials in practice are: doped silicon (cheap, CMOS-integrable, the default for switches and motor commutation); GaAs and InSb (very high electron mobility, used for high-sensitivity discrete field probes and gaussmeters); and InAs thin films for low-temperature-coefficient industrial sensors. The semiconductor's high mobility also means the carriers reach a given drift velocity at low electric field, which keeps power dissipation down.

From a bare plate to an integrated Hall IC

A raw Hall plate is almost useless on its own: the signal is a few millivolts, it drifts hard with temperature, and it carries an offset voltage of comparable size even at zero field. Every commercial device is therefore an integrated circuit that wraps the plate in supporting electronics:

• Regulated bias source. Holds the plate current (or voltage) constant so the sensitivity does not wander with supply voltage. Ratiometric devices instead scale the output to V_supply so an ADC sharing the same rail cancels supply error.
• Chopper-stabilised amplifier. Gains the millivolt Hall signal up to a usable range while rejecting the amplifier's own 1/f offset and drift.
• Spinning-current bias (dynamic offset cancellation). The bias is rotated electronically through the four contacts of the plate in sequence; averaging the readings cancels the geometric and stress-induced offset. This single technique is what turned the Hall sensor from a lab curiosity into a precision part — residual offset drops to a few microtesla.
• Temperature compensation. Silicon mobility falls roughly 0.05 percent per kelvin; an on-chip reference and a calibration polynomial flatten the sensitivity over the automotive -40 to +150 °C range.
• Output stage. Either a comparator with hysteresis (a Hall switch, digital high/low), a linear ratiometric analog output, or a PWM / SENT / I²C digital interface for programmable devices.

The output stage is where the device taxonomy lives. A switch trips on when the field exceeds an operate threshold B_OP and trips off below a release threshold B_RP — the gap between them is hysteresis that prevents chatter. A latch needs an opposite-polarity field to reset, which is perfect for counting alternating motor poles. A linear Hall outputs a continuous voltage centred at half-supply, swinging up for north and down for south, ideal for proportional position and current sensing.

Where Hall sensors actually show up

• Engine crankshaft and camshaft position. A toothed wheel passes a Hall switch; the controller counts teeth and one missing-tooth gap to know crank angle to within a degree, even at a dead stop. Hall replaced variable-reluctance sensors here precisely because it works at zero rpm — essential for stop-start and for synchronising fuel injection on the first crank.
• Brushless DC motor commutation. Three Hall sensors at 120 electrical degrees read the rotor magnets and output a 3-bit code that cycles through six states per electrical revolution. The controller decodes the state to energise the correct two phases. Found in PC fans, disk-drive spindles, drone ESCs, and EV traction motors.
• Current sensing. An open-loop sensor reads the field a primary current makes in a gapped ferrite core; a closed-loop (zero-flux) transducer nulls that field with a feedback winding for sub-0.5 percent accuracy. Both give galvanic isolation — the measured high-voltage circuit never touches the low-voltage output. Core of clamp meters, inverter phase-current measurement, and EV battery current monitoring.
• Proximity and lid-closed detection. A magnet on a laptop lid, phone flip cover, or appliance door trips a Hall switch to wake or sleep the device. Nanopower devices here draw under a microamp average by sampling the plate at a few hertz.
• Joysticks and throttle/pedal position. A magnet on the moving member and a linear or 3D Hall IC give a wear-free, contactless analog position — the reason modern Hall-effect gamepad sticks do not develop "stick drift" the way resistive potentiometers do.
• Flow and speed measurement. A magnet embedded in a turbine wheel or paddle trips a Hall switch once per revolution; pulse frequency gives flow rate or wheel speed (bicycle computers, water meters, ABS wheel-speed sensors).
• Gaussmeters and magnetic field mapping. A calibrated discrete GaAs or InSb probe reads absolute field strength for characterising magnets, MRI fringe fields, and electromagnet uniformity.

Failure modes and design trade-offs

• Offset voltage and zero drift. The biggest error. Contact misalignment and mechanical stress create a voltage that mimics a real field. Untreated it is millivolts; spinning-current cancellation drops it to microtesla-equivalent, but it never reaches zero. Always specify the residual quiescent output over temperature, not just at 25 °C.
• Temperature coefficient of sensitivity. Carrier mobility falls with temperature, so an uncompensated plate loses sensitivity as it heats. Integrated ICs correct this in firmware; discrete plates need external compensation or a ratiometric scheme.
• Piezo-Hall (mechanical stress) effect. Packaging stress, PCB flex, and solder-reflow shrinkage shift the offset. Precision parts use low-stress packages and on-chip stress compensation; this is why a reflowed sensor must be re-zeroed in the application.
• Limited resolution. The noise floor sits around 1–50 µT/√Hz, so a Hall sensor cannot resolve the geomagnetic field's small variations or detect a current of a few milliamps. Below roughly a microtesla you move to magnetoresistive (AMR/GMR/TMR) or fluxgate technology.
• Air-gap sensitivity. Output falls steeply with distance because a magnet's field drops as 1/distance³ in the near field. Doubling the gap between magnet and sensor can quarter the signal — mechanical tolerance of the gap is often the dominant system error in position sensing.
• Stray-field interference. A nearby busbar or motor adds an unwanted field. Differential (gradiometer) Hall layouts read the difference between two plates so a uniform stray field cancels, leaving only the local gradient from the intended target — the standard defence in automotive current and angle sensors.

How the Hall sensor compares to alternative field sensors

Property	Hall effect	Reed switch	Variable reluctance	Magnetoresistive (GMR/TMR)
Sensing principle	Lorentz deflection → voltage	Magnetic mechanical contact	dΦ/dt induced EMF	Resistance change with field
Works at zero speed	Yes	Yes	No — output dies at low speed	Yes
Moving parts	None (solid-state)	Yes (contacts wear)	None	None
Analog field output	Yes (linear)	No (on/off only)	Speed-dependent	Yes (high sensitivity)
Resolution floor	~1–50 µT/√Hz	Threshold only	Coarse	~1–100 nT/√Hz
Bandwidth	DC to ~1 MHz	~kHz (bounce-limited)	Speed-dependent	DC to MHz
Galvanic isolation	Yes	Yes	Yes	Yes
Cost (integrated)	Very low	Low	Low	Medium
Typical use	Position, BLDC, current	Door/lid switch	Legacy crank/wheel speed	Angle, low-field, hard-drive heads

Common pitfalls when designing with Hall sensors

• Forgetting field direction. Only the field component normal to the plate produces a signal. Mount the sensor with the magnet's flux lines passing through the package face, not edge-on, or you lose most of the output.
• Picking a switch when you need a latch (or vice versa). A unipolar switch will not count a ring of alternating magnet poles cleanly; a latch needs to see both polarities. Match the device's magnetic response to the magnet arrangement before laying out the wheel.
• Ignoring the operate/release hysteresis window. Choose magnet strength and gap so the field comfortably crosses B_OP and drops below B_RP over the full temperature range, or the switch will chatter or stick.
• Treating the analog output as absolute. Linear Hall output is ratiometric to supply on many parts; reference the ADC to the same rail or the reading drifts with the regulator.
• Underestimating air-gap tolerance. Because field falls off so fast with distance, mechanical play in the gap dominates the error budget. Specify and control the gap; do not trust nominal CAD dimensions.
• Letting stray fields in. A power conductor a centimetre away can swamp a weak target field. Use a differential layout, shield with high-permeability mu-metal, or concentrate the wanted field with a flux concentrator.