Where magnets appear in a robot
A modern articulated robot or cobot contains permanent magnets in four distinct roles, often several per joint.
| Role | Typical magnet | What governs it |
|---|---|---|
| Joint motor rotor | Arc segments or a magnet ring on a frameless rotor | Torque per unit mass and volume; thermal duty inside a sealed joint |
| Position feedback | Diametric disc on the shaft end, or a multipole ring for off-axis and hollow shafts | Angular accuracy, pole placement, mounting concentricity |
| Holding brake | Permanent magnet in a power-off engaged brake | Holding torque with zero power; release current; thermal stability |
| End effector | Switchable or electro-permanent assemblies, pot magnets | Holding force on real workpieces, release reliability, safety |
Why mass compounds
Every kilogram added at the wrist must be carried, accelerated and decelerated by every joint beneath it. A heavier wrist motor needs a stronger elbow motor, which needs a stronger shoulder motor, which needs a heavier base. This is why robot joints reach for the highest practical energy density rather than the cheapest adequate magnet — the saving from a lower grade at the wrist is repaid several times over in mass further down the arm. It is one of the few applications where specifying up is genuinely the cost-optimal choice.
Joint motors and torque density
Robot joints are almost universally permanent magnet synchronous machines, increasingly supplied as frameless kits that the joint housing integrates directly — the robot builder becomes the motor builder, and the magnet specification lands on their desk.
What drives the magnet choice
- Torque density. Torque scales with the airgap flux and the rotor radius, so a joint wants the highest usable remanence in the available annulus. This is the case for high energy product — but only up to the point where coercivity becomes limiting.
- Thermal duty is harsher than it looks. A joint module is a sealed aluminium housing containing a motor, a gearbox, a brake and drive electronics, often with no active cooling. Winding heat has nowhere convenient to go, and the rotor sits in the middle of it. Rotor temperatures well above ambient are normal even at modest duty cycles.
- Coercivity, not energy product, usually sets the grade. The combination of elevated rotor temperature and the stator’s demagnetizing field at peak torque is what determines the required class. Specifying N52 for maximum torque and discovering it demagnetizes during a hard stop is a real and expensive failure. Run the load-line check at maximum rotor temperature — the method is in the temperature guide.
- Cogging torque matters more than in most machines. Cobots doing force-controlled work, hand guiding, or precision assembly are sensitive to torque ripple that a factory robot would never notice. Pole arc, magnet skew and pole-to-pole consistency all feed into it — and pole arc ratio is a magnet dimension, specified on the drawing.
Rotor construction
| Topology | Magnet form | Suits |
|---|---|---|
| Surface mounted | Arc segments bonded to the rotor, sometimes sleeved | The common choice — moderate speeds, simple construction, good torque constant |
| One-piece magnet ring | Multipole magnetized ring | Small joints where assembly simplicity and consistency beat maximum performance |
| Interior permanent magnet | Flat blocks in rotor slots | Higher speeds and field weakening; the steel bridges carry retention |
| Halbach arrangement | Segments with rotating orientation | Slotless and high-precision designs; concentrates flux without back iron |
Retention deserves attention even at modest joint speeds, because a robot joint reverses constantly and a bonded segment sees cyclic shear for millions of cycles. Adhesive alone is a fatigue question; geometric capture or a retaining sleeve is not. The mechanical treatment is in assembly and retention design.
Position feedback and the error stack
Every joint needs to know where it is, and increasingly needs to know both where the motor is and where the output shaft is — dual feedback across the gearbox, which doubles the magnet count and doubles the opportunity for error.
Angular error stacks; it does not average
An arm is a chain. An angular error at the shoulder is multiplied by the full reach of the arm by the time it reaches the tool, while an identical error at the wrist contributes far less. A tenth of a degree at a shoulder joint on a 1.3 m reach is roughly 2 mm at the tool centre point. Feedback accuracy requirements should therefore be allocated by joint position, not applied uniformly — and the magnet contribution to that error is real.
What actually causes magnet-related angle error
- Eccentricity of the magnet on the rotation axis. The dominant contributor in end-of-shaft diametric sensing. The field direction the sensor reads is distorted by off-axis mounting, and it appears directly as angle error. This is a mounting and concentricity problem, not a magnet strength problem.
- Tilt and axial gap variation, which do the same thing more subtly.
- Pole placement error on multipole rings. Where a ring encodes many pole pairs, the angular spacing of the transitions is the scale. Specify pole count, pole-to-pole balance and maximum transition angle error, and require a pole scan on the first article.
- Thermal drift, which affects magnitude-based architectures far more than angle-based ones. Angle sensing that reads field direction rides out remanence drift; anything reading magnitude inherits it.
The architecture, sensor pairing and specification detail is covered fully in magnets for sensors. Two robotics-specific points are worth adding:
- Hollow shafts drive ring selection. Robot joints route cables and services through the joint centre, so the shaft end is frequently unavailable. That pushes feedback to an off-axis ring at the outer diameter — and where uniform field through 360° matters, to a one-piece true radial ring rather than a segmented assembly whose glue joints appear as periodic error.
- Stray field immunity. The feedback magnet sits centimetres from a motor rotor producing a far stronger field. Differential and gradiometric sensor topologies exist for exactly this reason; the magnet specification does not change, but the sensor selection must account for the neighbourhood.
Holding brakes and the safety case
A robot arm holding a payload with the power off is being held by a brake, and in most designs that brake is engaged by a permanent magnet and released by a coil. The magnet is therefore a component in a safety function.
Treat brake magnets as safety-relevant parts
Where a brake contributes to a functional safety claim, the magnet inside it inherits that status. In practice that means a tighter acceptance limit than a general industrial part, a documented magnetic acceptance test rather than a certificate alone, lot traceability, and a change-control clause preventing a silent grade or process substitution. The methods are in incoming inspection and requalification triggers.
Magnetic end effectors
Magnetic gripping suits ferrous parts that vacuum handles badly — perforated, oily, hot, irregular or porous stock. It is fast, needs no compressed air, and has no moving jaws to obstruct the part.
| Type | How it releases | Trade-off |
|---|---|---|
| Electromagnet | Power off | Simple and fully controllable, but continuous power, continuous heat, and the part drops on power loss |
| Electro-permanent | Current pulse switches the magnetic circuit | Zero holding power, holds through power loss, pulse to release. The dominant choice in robotic handling |
| Mechanically switchable | Lever or actuator shunts the flux | No electrical supply at the tool; needs an actuator and adds cycle time |
| Fixed permanent + stripper | Mechanical push-off | Simplest and cheapest; release force must exceed hold, which limits capacity |
Rated hold is not the hold you get
Published holding force assumes a thick, clean, flat, low-carbon steel plate in direct contact and pulled straight off. Real parts fail every one of those assumptions:
- Thin material does not saturate, so a sheet-metal part may develop a fraction of the rated force.
- Surface condition creates an air gap. Paint, scale, oil film or rust each cost force, and field falls steeply with distance.
- Shear resistance is far below normal force. A load that will not pull off will readily slide, and a robot accelerating through a path change loads the gripper laterally.
- Alloy matters. Austenitic stainless is essentially non-magnetic; a gripper sized on mild steel will not hold it at all.
- Residual magnetism transfers to the part, which then attracts swarf, disturbs downstream sensors, or interferes with welding.
Size against the worst-case real part with a deliberate margin, and verify on hardware rather than from a table.
Collaborative operation
Where a robot shares space with people, a dropped part is a safety event rather than a scrap event. Two considerations follow: the gripper should hold through a power loss or emergency stop, which favours electro-permanent and mechanically switchable types over electromagnets; and the stray field around the tool should be assessed against nearby sensors, instruments and any implanted medical device policy the site operates. Containing that field is a flux-circuit design problem — see magnetic shielding and flux containment.
What we supply
- Arc segments and magnet rings for joint motor rotors, in H through UH classes, with specified pole arc and magnetization direction.
- Diametric discs for end-of-shaft angle sensing, with clocking to a mechanical feature where the application needs a defined electrical zero.
- Multipole and one-piece true radial rings for off-axis and hollow-shaft feedback, with pole-scan certification on first article and production lots.
- Blocks and pot assemblies for brakes and magnetic end effectors.
- Per-lot certification and acceptance testing to a defined magnetic limit for safety-relevant parts.
What to send us
- The joint envelope — rotor OD and ID, axial length, available annulus
- Continuous and peak torque, and the speed range
- Worst-case rotor temperature and the duty cycle that produces it
- For feedback: the sensor part number, the air gap, and the required angular accuracy at that joint
- Pole count and any clocking requirement
- For grippers: the actual workpiece — alloy, thickness, surface condition — and the acceleration the part must survive
- Any functional safety classification the part inherits