By Sarah Xiao 
In a dexterous hand, the motor is not a commodity you pick off a catalogue. A hollow-cup brushless motor — no iron teeth, low torque ripple, low rotor inertia, fast response — is what makes a finger able to close on a glass without cracking it, and it is the reason a premium hand costs what it costs.
The important thing to understand before specifying one: in this motor, the magnets sit on the stator and the winding cup is the rotor. The magnet sets the electromagnetic ceiling, but what you actually receive in volume is decided by the cup — its geometry and, above all, how the wire was laid.
What the cup has to do
A conventional slotted motor holds its windings in stator teeth. That is cheap and stiff, and it produces cogging torque: the rotor prefers certain positions, so the motor stutters at start-up and vibrates in motion. For precision drives that is disqualifying.
A hollow-cup machine removes the teeth and puts the winding in a self-supporting cylinder. There is nothing holding the coils except the wire’s own form, adhesive and a carrier. That single change creates both the benefit and the manufacturing problem: the cup has no iron to lean on, so every geometric error becomes a torque error.
A complete motor is a stator assembly, the cup assembly, a Hall sensor assembly, the rotor, end caps and bearings. Good designs are coaxially integrated, with switch-type Hall elements sensing a magnet that rotates with the main rotor; the cup position and the sensor angle are then adjusted so the Hall signals land where the commutation expects them. Get that alignment wrong and no amount of software fixes the speed regulation.
Within the cup assembly there are three things to hold: circumferential symmetry of the three-phase coils, the stability of the bond between winding and carrier, and the accuracy of the lead-out weld. Any of the three drifting produces torque imbalance and vibration.
Three winding types, three different bets
| Type | Construction | Strength | Cost of getting it wrong |
|---|---|---|---|
| Straight (concentrated) | Ring wound on a former, then laid flat to the winding diagram, ends glued and cured | Highest fill factor, best torque density | Manual laying stacks and crosses wires; end turns stay bulky |
| Skewed (honeycomb) | Effective sides angled to the axis, wound continuously on a pinned mandrel | Very small end turns, compact | Crossed wire cuts fill factor, so torque ceiling is low |
| Diamond (pre-formed) | Self-bonding wire shaped on a dedicated die, then laid in two regular layers | Regular layers, controllable dimensions, automatable | Dies and forming cycles are the whole investment |
The diamond form is what most humanoid and medical programmes now choose, because it is the only one that is simultaneously good enough on fill factor and compatible with automated volume production. Straight winding survives where the envelope is generous and torque matters more than size. Skewed winding survives in the narrow band where the envelope is the binding constraint and torque is not.
The process is the moat
Three production models exist, and they are not interchangeable.
Manual. Skilled operators insert pins, wind and lay. Precision is poor, but it is fine for samples.
Semi-automatic. The dominant domestic approach: wind onto a diamond mandrel, remove and flatten into a wire plate, then roll into a cup. Many manual interventions, so quality drifts batch to batch.
One-shot automatic forming. Wire placement, forming and curing in a single cycle with no operator in the loop. Expensive tooling, but it is the only way to control tension and pitch at speed — and therefore the only way to stop fine wire from being scraped or broken.
A hollow-cup motor runs to roughly 30 process steps, from winding and insulation shaping through bearing, shaft and support-ring assembly to lead welding, encapsulation and test. Winding is the centre of it, because a small pitch error, a tension fluctuation or a curing deviation does not show up as a reject — it shows up as vibration, noise and efficiency loss in the customer’s hand, three months later.
That is the mechanism behind the industry’s familiar complaint: impressive samples, disappointing volume. It is rarely a design failure. It is the process not being under control.
What to ask before you approve a supplier
- Winding type, named, with the fill factor achieved. Not the theoretical maximum — the figure on their production parts, with the measurement method.
- Which of the three production models they run, and on which line. A supplier with one automatic line and three semi-automatic ones will quote you from the automatic line and ship from the semi-automatic one.
- Tension and pitch control data. Setpoint, tolerance and the Cpk across a shift. This is the parameter that predicts your field failure rate.
- Enamel damage and break rate. Ask for the scrap rate at the winding station specifically, not the plant average.
- Hall alignment method and its tolerance. How the sensor angle is set and verified, and what the resulting commutation error is in electrical degrees.
- End-of-line test content. Cogging torque, no-load current, back-EMF symmetry, insulation resistance — and the accept limits, not just the test names.
Two commercial notes. First, the cup is where the price is: a fully automatic line carries a capital charge that only amortises above a few tens of thousands of units a year, so a supplier quoting a very low price at low volume is either subsidising you or running semi-automatic and hoping. Second, dual-sourcing a wound cup is harder than dual-sourcing a housing — the second supplier has to reproduce the winding, not just the drawing. Budget a qualification cycle, not a sample order.
Related reading
- Robot tactile sensors compared: flexible arrays, magnetic Hall, vision-based — the other half of a dexterous hand’s bill of materials.
- Robot component sourcing: from prototype to production — the qualification sequence this motor has to pass.
- Robotics shutdowns 2026: funding is not delivery capability — what happens to a programme when the process was never under control.
Sources: motor architecture, the three winding types and the process sequence follow 人形机器人精密制造 (7 September 2026). The supplier qualification questions, the cost-amortisation argument and the dual-sourcing note are Latest Robot News editorial.