What Is a Hall Sensor: A Switch With No Moving Parts
A Hall sensor turns a magnetic field into a voltage. Nothing moves, so nothing wears out.
It replaces the mechanical switch wherever a product must know whether a lid is shut. The same part watches a piston reach its end, or a shaft turn.
The part itself is a small chip with three pins: supply, ground and signal. A magnet comes near, the signal changes state, and a control board reads it.
That simplicity hides a buying problem. Four different parts share that three-pin footprint, and swapping one for another stops the product working.
The Short Answer
A Hall sensor is a solid-state device that measures a magnetic field. Most of the ones sold are not measuring anything: they are switching.
The sensor sits still and the magnet moves, or the other way round. Either way the gap between them changes, and the chip decides when that change counts.
Because there are no contacts, there is nothing to arc, bounce or corrode. A mechanical switch is rated for a million operations. A part with no contacts has no limit of that kind.
It also survives conditions that kill other sensors. A sealed part works through dust, mud and water. That is why it appears on wheels, in pumps and inside cylinders.
There are limits, and they are worth having on one page. It needs a magnet on the moving part. It senses one axis per element. And it drifts with temperature.
The rest of this article covers three things. The four families, the points where they switch, and the three lines a buyer has to fix.
A Voltage That Appears Sideways
Edwin Hall described the effect in 1879, long before anyone could build one of these into a chip. The geometry is the whole trick.
A small current is pushed along a thin strip of semiconductor. That strip is the sensing element, and it has electrodes on its two sides.
Put a magnetic field across the strip, and the moving charges feel a sideways push. Charge collects on one edge and thins on the other.
The electrodes measure the difference. The voltage that appears is proportional to the part of the field that runs across both the current and the electrodes.
Three consequences follow, and each one matters on a data sheet.
The output is small. A bare element gives microvolts, so every practical part adds an amplifier and usually a regulator in the same package.
It reads static fields. A coil-based sensor only reacts to change, so it cannot tell whether a lid is open or shut while everything is still. A Hall part can.
It has an offset. With no field at all, the output is not quite zero. Modern parts cancel that by switching the bias current on and off, then averaging. A bad part often shows as drift rather than as silence.
The chip reads one axis only, which is the other thing buyers get wrong. Two axes need two elements at right angles, and three axes need a third.
Four Families in One Footprint
The package looks the same for all of them. What changes is the circuit behind the sensing element, and that decides the job.
| Family | Output | How it behaves | Where it fits |
|---|---|---|---|
| Unipolar switch | Two states | Responds to one pole only | Lids, end of stroke |
| Omnipolar switch | Two states | Responds to either pole | Doors and covers |
| Bipolar latch | Two states | Holds its last state until the other pole arrives | Rotation counting, motor timing |
| Linear | Voltage that follows the field | Reads strength, not just presence | Position and current sensing |
The first three are switches. A comparator inside the chip turns the small signal into a clean two-state output. Most parts add a Schmitt trigger so the output does not flutter.
The omnipolar version exists for assembly reasons. If the magnet can arrive either way round, a part that accepts either pole removes a build error from the line.
The latch is the one that confuses buyers. It needs a north pole to change one way and a south pole to change back. It stays where it was put. Spin a ring of alternating poles past it and the output becomes a square wave.
The linear part is a different product in the same body. It does not decide. It reports a voltage that rises with the field, and the control board makes the decision.
It is usually ratiometric, which means its sensitivity scales with the supply. With no field near it, the output sits near half the supply.
That is convenient for a microcontroller on the same rail and awkward in a mixed-voltage design. Feed a 5 V part from 3.3 V and the readings fall short in a way that looks like a weak magnet.
Operate, Release, and the Gap Between
A switch does not turn on and off at the same field strength. Those two numbers decide whether a product works in the field.
The field at which the output changes as the magnet approaches is the operate point. The field at which it changes back as the magnet leaves is the release point.
The release point is always closer to zero than the operate point. The difference between them is the hysteresis, and it is there on purpose.
Without that gap, a magnet sitting near the switching level would chatter. The output would flicker on every vibration, and a lid switch would report the lid opening and closing all day.
The gap also sets how far the magnet can drift before the state changes. A wide gap tolerates a loose hinge or a worn mount; a narrow one is fussier about the mechanical build.
The reference material on these parts names the two points and uses them to sort the families. That sorting is the only reason the letters appear on a data sheet.
Here is the trap. Some families are sold as bipolar by one maker and as a latch by another. The behaviour and the number are comparable.
The reference material records exactly that, naming one common part that carries both labels depending on who wrote the page.
So the letters on the front of the data sheet do not settle it. The operate and release points do, and they are printed a few pages further in.
A Laptop Lid and a Motor
Both use a Hall part, and the requirements pull in opposite directions. Getting them the wrong way round is how a project pays twice.
The lid. A magnet sits in the lid, the sensor on the main board, and a few millimetres of plastic between them. The part has to stay on across a wide gap and off when the lid opens.
Nothing about that job needs precision. It needs a switch that tolerates a hinge with play, a cheap magnet, and enough hysteresis to ignore a knock.
The motor. A brushless motor has no commutator, so it cannot tell where the rotor is. Three sensors around the rotor do that job instead, as the account of brushless commutation sets out.
Each sensor reports which pole has just passed. Three sensors, each high or low, give six combinations, and the controller uses them to fire the windings in sequence.
That lands a position update every 60 degrees. Coarse is fine here, and the same page makes the point: because the motor commutates in 60-degree steps, fine resolution adds nothing.
Here the parts want to be latches rather than wide-gap switches. The rotor poles alternate as they sweep past, so the sensor must change on one pole and change back on the next.
The same three-pin part therefore does two very different jobs, and only the numbers behind the output tell them apart.
What It Cannot Do
Five limits separate a Hall sensor from the job people imagine. Four of them show up as a support ticket, not as a failure.
It needs a magnet. A ferromagnetic target helps, but a plain aluminium or plastic part has to carry a magnet of its own. That is a part, a supplier and an assembly step.
Darkness and dirt are irrelevant, distance is not. The field falls away quickly as the gap opens. Most designs live within a few millimetres, and a sealed housing adds to that gap.
It reads one axis. Two axes need two elements; three need three. A joystick is a different part from a lid switch, not a mounting choice.
Stray fields count. Nearby wiring, a motor and the earth itself all add to what the element sees. The standard answer is a shield, or a differential pair of elements that cancels what both of them share.
It drifts. Output moves with temperature and, slowly, with age. A part that reads 2.5 V today may read a little differently after a year at 60 °C in a case.
Bandwidth is the quiet limit. Common silicon parts run to tens of kilohertz. That is far more than a lid needs, and often not enough for a fast current loop.
One more comparison is worth keeping. Against an optical encoder, a Hall sensor gives absolute position less easily. It is cheaper and dirt-proof, and that is the trade.
Three Lines a Buyer Has to Fix
Two parts that look identical can differ in all three of these, and a factory will ship whichever it has unless the order says otherwise.
| Line | What to write | What goes wrong without it |
|---|---|---|
| Type and polarity | Switch, latch or linear, and which pole actuates it | A latch arrives where a switch was specified, and the output stays put |
| Sensitivity | Operate point, release point, and the gap between them | The magnet has to be repositioned on a finished assembly |
| Supply and output | Voltage range, and whether the output is open-drain or push-pull | An open-drain part with no pull-up reads as dead on the bench |
The first line is the one that stops a line. A latch on a lid never releases, so the device believes the lid is shut for ever.
The second line decides the mechanics. Fix the operate point and the release point, and the mounting distance becomes a calculation instead of a trial fit.
The third line decides whether the control board can read it at all. An open-drain output pulls low and needs a resistor to pull it high; a push-pull output drives both ways.
Two more lines belong on the same sheet: the package, and the temperature range over which those numbers have to hold.
Sealing is a separate conversation, and the second digit of an ingress rating decides whether the housing protects the part in a wet machine.
For a comparison across suppliers, the answers need to line up. One specification sheet does that better than four data sheets in four layouts.
How to Check a Batch
The bench test takes a minute and it separates the families without a data sheet. All it needs is a supply, a meter and a magnet.
Power the part across supply and ground, then measure the signal pin against ground. With no magnet near it, note the level it sits at.
Bring one pole slowly towards the part. A switch or latch will snap to the other level, and the gap at which it does so is roughly the operate point.
Take the magnet away, then bring it back with the other pole. A switch ignores the second pole; a latch changes state; an omnipolar part behaves either way.
That is enough to catch a mixed box. The appliance-repair write-up on these parts describes the same test on a motor. It works there because a spinning rotor is just a magnet arriving and leaving.
It does not measure anything. For the numbers you need a Helmholtz coil or a calibrated gap jig, which is a factory question rather than a bench one.
Before the order grows, it is cheaper to look at stock than to argue about a photo. The market where these parts are sold has bins of them, and a magnet in your pocket sorts the families in an afternoon.
Six Lines for the Order
Put the three lines above into six sentences, and add the two that decide whether the part survives the product.
One. Name the family and the polarity that actuates it, not just the package size.
Two. Give the operate point, the release point and the gap between them, in the unit the supplier uses.
Three. Fix the supply range and say whether the output is open-drain or push-pull.
Four. State the temperature range over which those switching points must hold.
Five. Specify the package and the sealing, including whether the part is overmoulded or potted.
Six. Agree the incoming test: which magnet, which gap, and which meter reading counts as a pass.
None of that replaces a look at the line. A check before the first batch should include the switching test, because a mixed reel passes a visual inspection every time.
Sensors also carry their own marks, and the paperwork a component needs to clear customs is a separate file from the data sheet.
Send me the Hall sensor data sheet you were quoted and I will read the switching lines. I will say whether the family matches your mechanism, whether the operate and release points suit your gap, and which of the six order lines is missing. No cost, no obligation.
Check My PartQuestions Buyers Ask
Does a Hall sensor need a magnet?
Usually yes. It senses a magnetic field, so the moving part must carry a magnet or be made of a ferromagnetic metal. Aluminium and plastic need an added magnet.
What is the difference between a switch and a latch?
A switch follows the magnet and returns when it leaves. A latch keeps its last state and only changes when the opposite pole arrives, which suits rotation counting.
Why does the output switch on and off at different distances?
That gap is deliberate. It stops the output fluttering when the magnet sits near the switching level. It also decides how much mechanical play the design can take.
Can a Hall sensor measure current?
Yes. A current makes a field around its conductor, so a linear part placed near that conductor reads the field and the board converts it to current, without a shunt.
Why does my open-drain sensor read as dead?
It can only pull the signal low. Without a pull-up resistor to the logic supply, the meter sees nothing. A push-pull part drives both levels and does not need one.
Are Hall sensors affected by temperature?
Yes. The output drifts with temperature and slowly with age, which is why the switching points have to be specified over the range the product will see.
Questions & Comments
We read every one and reply within 24 hoursIf a sensor arrived and the product behaves as if the magnet is always there, tell me the family you were sent. A latch in a lid design does exactly that.