Picking a BLDC motor for a robot arm or a mobile platform starts with the mechanical load, not the nameplate. Voltage and wattage say nothing about low-speed torque or the thermal duty cycle.
Key takeaways
- The actuator is a system: motor, inverter, current sensors, encoder, gearbox, brake and thermal path
- An arm joint needs controllable torque, a mobile drive needs efficiency and traction
- Six-step, sinusoidal and FOC commutation are three different behaviours, not synonyms
- FOC splits current into d and q components via the Clarke and Park transformations
- Regenerative braking can blow up a system sized correctly for acceleration
The actuator is not a motor but a chain. Every link caps the final result.
Load first, catalogue second
For a geared joint, output torque is motor torque times gear ratio times gearbox efficiency, minus losses. Speed scales with the ratio, and power is torque times angular velocity.
Before choosing a controller you must settle continuous and peak torque, reflected inertia, duty cycle and thermal resistance. Energy returning to the DC bus under regenerative braking is a separate item.
An arm and a platform are two different jobs
An industrial or collaborative arm joint needs controllable torque, smooth low-speed motion, position accuracy and backdrivability?Backdrivability: The drive's ability to yield under external force. Essential when a robot must respond safely to human contact.. A motor-side encoder reports rotor position accurately while the joint output still carries transmission error.
A mobile drive shifts emphasis to efficiency, starting torque, traction and channel synchronisation. An outdoor UGV runs under heavy load far longer than a light indoor AMR.
Commutation: three methods, three behaviours
| Method | How it works | When it makes sense |
|---|---|---|
| Six-step | six inverter states by rotor position, Hall sensors suffice | simplicity and cost, acceptable ripple, moderate to high speed |
| Sinusoidal | continuously varying current references instead of discrete states | lower ripple and noise, outcome depends on construction and tuning |
| FOC | Clarke and Park transformations, independent d and q axis control | smooth low-speed motion, fast torque response, force control |
FOC transforms measured phase currents into a rotating reference frame. The Clarke transformation reduces three phases to two stationary axes, the Park transformation rotates them into d and q components. In a PMSM the q component is tied to torque and the d component to field weakening.
Why it matters
The humanoid market is driving demand for torque-dense joints, and the conversation is dominated by control models. Meanwhile a robot's ceiling is set by a chain nobody writes about: encoder, inverter, thermal path and gearbox. A VLA policy will not execute a motion the actuator cannot survive thermally.
Picking a drive is a decision that lasts years, because reworking a joint means redesigning the whole arm.
What's next
- The next decision after commutation is feedback architecture: Hall sensors, incremental and absolute encoders, resolvers or sensorless estimation
- Application notes from Texas Instruments and STMicroelectronics cover six-step control and FOC for BLDC and PMSM motors
Sources
- The Robot Report — How to choose a BLDC motor for robotic arms and mobile robots
- Texas Instruments — Sensorless Field Oriented Control of 3-Phase Permanent Magnet Synchronous Motors (SPRABQ3)
