Radial Multi-pole Ring Magnets for Robot Joint Motors: What to Put on the Drawing

Pole count is a motor decision that arrives on the magnet drawing, and it is worth writing down where it came from. Electrical frequency is pole pairs times mechanical speed, so a 14-pole ring at 3,000 rpm asks the drive for 350 Hz where an 8-pole ring asks for 200 Hz. More poles shortens the flux path in the back iron and lets the yoke get thinner, which is why joint actuators for robotics run high pole counts in short axial packages.

More poles also narrows the pole arc, and past a point the transition zone between poles stops being a small fraction of that arc. Production tooling covers 4, 6, 8, 10, 12 and 14 poles, on outside diameters from 15 mm to 250 mm, wall thickness 2 mm to 10 mm and axial length 5 mm to 80 mm. If the electromagnetic design is still open, ask which pole counts already have fixtures at your diameter before committing to one. A pole count nobody has tooled is a tooling charge and a lead time, and neither of those shows up on a datasheet.

Radial is one of three ways to orient a ring, and the choice moves airgap flux further than a grade step does. A benchmark on identical 10-pole geometry, Ø36.6 × Ø29.8 × 17 mm, measured 415.7 mT peak airgap flux from a conventional radial-oriented ring in N45SH.

A segmented ring built in the higher N50SH grade measured 521.4 mT. A one-piece sintered Halbach ring in the same N45SH measured 629.0 mT, 51% above the radial ring with no change of material. Waveform separated the same way. The radial ring measured trapezoidal, the segmented ring measured square with saddle peaks at the pole edges, and only the Halbach ring produced a near-sinusoidal back-EMF without skew.

That does not make radial the wrong pick. Radial rings tool cheaper, magnetise more easily and come from more plants, and a drive running field-oriented control will live with a trapezoidal waveform. It does mean the orientation line has to be settled before the grade line, because changing it later changes the motor.

Four numbers carry most of the mechanical risk: outside diameter, bore, wall thickness variation and length. Standard production radial-oriented multi-pole rings are held to ±0.05 mm on dimensions, coaxiality 0.1 and cylindricity 0.04, and that is what to design against. The ±0.02 mm figure that circulates for magnet rings comes from one-piece sintered Halbach features; on a radial ring drawing it turns a standard part into a quoted special without buying anything the motor needed.

Wall thickness variation is the number that reaches the balancing machine. The magnet sits at the largest radius of anything on the shaft, so a gram of asymmetry there costs more residual unbalance than the same gram in the lamination stack. ISO 1940-1 gives the whole rotor one allowance to share: at balance grade G2.5 and 30,000 rpm the permissible residual specific unbalance works out near 0.8 µm of mass eccentricity, covering shaft, stack, magnet, adhesive and retention sleeve together.

Put the magnet's share of that budget on the drawing instead of discovering it in the balancing machine. The balancing guide works through the arithmetic.

Most robot joint actuators land on the SH class. N42SH and N45SH both carry a 150 °C rating with HcJ at or above 20 kOe, and N45SH gives Br of 13.2 to 13.7 kGs where N42SH gives 12.9 to 13.3 kGs. N45UH takes the rating to 180 °C and HcJ to 25 kOe and above at the same Br band, and it is produced by a small number of qualified plants, so lead time and supplier qualification both get harder.

Specify minimum HcJ at the working temperature rather than at 20 °C. The room-temperature figure is the one everybody quotes and the one nobody operates at, and demagnetisation risk in a joint motor is a knee-point question at the hottest point of the duty cycle, under the worst current the drive can deliver into it.

The sintered range runs N30 to N45AH across the F, N and D series, so if the thermal model says the SH class is more than the design needs, there is room below it. The grade pages carry the property tables.

Where the ring feeds a position sensor or an encoder, deviation angle becomes its own line item. It measures how far the actual field direction at a point on the surface departs from the direction the drawing asked for, and a sensor reading absolute position off that field inherits the error directly.

Sensor-grade rings are held under ±3° deviation angle and inspected 100% rather than sampled. Both halves matter. A sampling plan applied to a characteristic that drifts with orientation-fixture wear will pass a lot containing exactly the parts that fail in the field. Surface flux variance ring to ring runs under 1% on production multi-pole rings, and asking for the per-batch waveform record costs nothing at quotation time. Where the ring only produces torque and commutation runs off a separate encoder, none of this applies.

Four things move the number, roughly in this order. Heavy rare earth loading comes first: the temperature class sets how much dysprosium or terbium goes into the alloy, and the SH and UH classes are where the metal cost sits. Tooling comes second, charged once per pole count and diameter, which is the argument for reusing an existing fixture wherever the electromagnetic design allows it.

Volume comes third, and the step is real: prototype batches of 50 pieces ship in 2 to 3 weeks, first production lots start around 500 pieces at 6 to 8 weeks, and the piece price at 5,000 a month is a different conversation from the piece price at 500. Inspection comes fourth, and it is what buyers most often leave undefined.

Where the ring arrives as part of a rotor assembly rather than a loose part, the tolerance stack and the bonding yield move to the supplier, which changes what incoming inspection you have to run yourself.