A robot joint motor drives rotation at each articulation point — shoulder, elbow, wrist, hip, knee. The technology has shifted hard toward BLDC and frameless torque motors over the last five years, and the shift is accelerating.
Most online guides will give you a tidy comparison table and tell you to "match the motor to your application." That is not wrong, but it skips the part that actually matters: understanding why certain motor architectures dominate specific robot classes, and where the real trade-offs hide once you move past the datasheet.
This guide covers the motor types, but it also covers the decisions that trip up engineering teams during actual integration - the stuff that does not show up in a product catalog.
What Makes a Joint Motor Different from a Regular Motor
If you have designed industrial automation systems before, you might assume a joint motor is just a small servo in a tight space. It is not.
A joint motor sits inside a kinematic chain. Every gram of motor mass at the elbow gets multiplied by the moment arm when the shoulder has to accelerate it. Every millinewton-meter of cogging torque becomes a disturbance that the joint-level impedance controller has to suppress. Every millimeter of extra motor length pushes the next link outward and changes the robot's reach-to-footprint ratio.
The constraints that matter, in order of how often they kill a motor selection:
Torque density. Not peak torque — torque per unit mass and per unit volume, at the continuous thermal limit. A motor that hits 10 Nm peak but can only sustain 2 Nm before thermal shutdown is not a 10 Nm motor for joint design purposes.
Cogging and torque ripple. If you are building a position-only robot (point A to point B, no contact), cogging matters less. If you are building anything with force control — cobots, humanoids, surgical — cogging torque below 0.5% of rated is table stakes. Above that, your impedance controller fights the motor instead of controlling the environment.
Form factor and hollow shaft. Cables need to route through the joint center. If the motor does not have a hollow bore, you are routing cables externally, which limits joint rotation range, adds failure points, and makes the joint ugly. "Ugly" sounds trivial until your customer's integration team refuses to buy a robot with external cable bundles on every joint.
Encoder integration. The motor needs to mate tightly with the feedback device. Misalignment between the encoder and the motor's electrical zero creates a torque estimation error that scales with current — exactly when you need accurate force sensing the most.
Motor Types — With Honest Trade-Off Commentary
Brushed DC Motors
Still in production. Still cheap. Still used in educational kits and cost-driven consumer robots.
For anything that ships to a customer who cares about reliability or precision, brushed DC is a dead end. Brush wear creates particle contamination inside the joint, cogging is high, and torque density is the lowest of any option. If a motor vendor is pushing brushed DC for a commercial robot joint in 2026, find a different vendor.
Standard Servo Motors
The "servo motor" label covers a wide range: small hobby servos with plastic gears, all the way up to industrial AC servos with integrated drives and multi-turn encoders. The common thread is that they come as a sealed package — motor, gearbox, driver, encoder, all in one housing.
Good for: AGV steering joints, simple pick-and-place arms in structured environments, applications where you need a joint working in weeks rather than months.
Bad for: custom joint geometry, weight-sensitive designs, anything requiring force control. The sealed package means you inherit the manufacturer's bearing arrangement, gear ratio, and thermal path whether they suit your joint or not.
BLDC Motors
Brushless DC is the modern default for professional robotics. Electronically commutated, no brush wear, better thermal performance, and roughly 30–50% higher torque density than equivalent brushed designs.
Within the BLDC family, the distinction that matters most is inner-rotor vs. outer-rotor:
• Inner-rotor designs spin the magnet assembly inside the stator. Higher speed capability, lower rotor inertia, faster acceleration. This is what you want for drive wheels, high-RPM applications, and anywhere the motor runs through a high-ratio gearbox.
• Outer-rotor designs wrap the magnets around the outside of the stator. Higher torque at lower RPM, larger diameter for a given torque. You see these in direct-drive or low-reduction applications — lawnmower robots, some quadruped leg joints, gimbal drives.
For robot arm joints specifically, inner-rotor BLDC paired with a harmonic or cycloidal reducer is the most common architecture. But the real performance jump comes when you remove the motor housing entirely — which brings us to frameless.
Frameless Torque Motors — Where the Industry Has Landed
A frameless motor is just a stator and a rotor. No housing, no bearings, no shaft. You bond the stator into your joint housing and the rotor onto your output shaft, and the joint's own cross-roller bearing handles all loads.
This is not a niche approach. It is the architecture inside Universal Robots cobots, Franka Emika Panda arms, and effectively every humanoid arm program that has reached prototype stage — Tesla Optimus, Figure, Apptronik. If you are designing a cobot or humanoid joint today and not using frameless, you are probably carrying 30–40% more joint mass than you need to be.
Why frameless wins at the joint level:
The joint's bearing is the only bearing. No redundant bearing sets stacking tolerances. The motor sits coaxially with the harmonic reducer, eliminating coupling misalignment. The hollow bore routes all cabling through center. And because there is no motor housing adding thermal mass between the windings and the joint structure, thermal management is actually better — the joint housing itself becomes the heat sink.
The downside is integration effort. You are responsible for bonding, alignment, bearing selection, and thermal path design. For a first-time integration, budget 2–3 engineering iterations to get the stator bonding interface and encoder alignment right.
What to look for in a frameless motor supplier:
- Cogging torque spec — get the actual number in % of rated torque, not just "low cogging." Below 0.3% is excellent; below 0.5% is acceptable for cobots; above 1% is a problem.
- Custom winding availability — your driver's voltage and current envelope is not the same as everyone else's. A supplier that only offers catalog windings will force you to compromise on the electrical match.
- Quality system — if you are sourcing for production, not prototyping, ask about traceability. IATF 16949 (automotive quality) is increasingly the bar that humanoid and medical robotics OEMs require from motor suppliers.
For precision robotic joints — cobots, humanoids, surgical, semiconductor handling - Twirl manufactures frameless BLDC motors engineered specifically for these applications. 200+ patents in precision motor design, IATF 16949 certified, with custom winding and encoder integration as standard engineering services. If you are past the research phase and into active motor sourcing, their BLDC motor has the technical detail you need for an informed selection.
Harmonic-Integrated Joint Modules
Motor + harmonic reducer + encoder + driver, sometimes plus brake and torque sensor, all in one connectorized package. Plug it into your robot structure, wire power and comms, and you have a working joint.
The gold standard for time-to-prototype. Backlash is near zero, repeatability is in arcseconds, and the integration risk is close to zero because the manufacturer has already solved the mechanical and electrical integration.
The catch: cost is 3–5× a discrete frameless motor plus separate harmonic, and you lose the ability to optimize the joint geometry for your specific robot. Every joint on your arm has the same module envelope, even if your wrist could be 40% smaller.
For production robots shipping in volume, most teams start with integrated modules to de-risk the prototype, then switch to discrete frameless motor + chosen reducer for production — where the per-unit savings justify the integration engineering.
QDD (Quasi-Direct-Drive) Actuators
A high-torque BLDC motor paired with a low-ratio planetary reducer, typically 6:1 to 10:1. The low ratio makes the joint backdrivable — an external force can push the joint without breaking anything.
QDD is the technology that made dynamic quadrupeds possible (Boston Dynamics Spot, Unitree Go2, ANYmal) and is now standard for humanoid leg joints where impact resilience matters more than positioning accuracy.
The trade-off is explicit: you sacrifice output torque and positioning precision for transparency and bandwidth. A QDD hip joint might achieve ±0.5° positioning versus ±0.01° for a harmonic-integrated joint. For a leg that needs to absorb ground impact at walking speed, that trade-off is correct. For a wrist joint doing pick-and-place at 0.05 mm repeatability, it is not.
How to Choose — A Framework That Actually Works
Most selection guides give you a three-step process that looks clean on paper but falls apart when you sit down with real requirements. Here is what the process actually looks like.
Start With the Joint You Are Most Worried About
Do not try to select motors for all six joints at once. Identify the joint with the tightest constraints — usually the shoulder (highest torque, most thermal stress) or the wrist (tightest volume, finest motion requirements) — and solve that one first. The other joints are usually easier once you have established a motor platform and supply relationship.
Get the Four Numbers That Eliminate 80% of Options
1. Maximum continuous torque at the joint output — not peak, not instantaneous. The continuous number at worst-case duty cycle, including gravity loading. Add 30% margin.
2. Maximum joint speed in rad/s — this determines your reducer ratio range. Most cobot joints sit between 1–3 rad/s at the output; humanoid joints can be faster.
3. Allowable joint OD and length — this is a hard geometric constraint. Measure your available envelope honestly. The motor vendor cannot change physics.
4. Positioning accuracy requirement — arcmin for cobots, arcsec for precision industrial. This number drives your reducer selection and encoder resolution.
These four numbers, combined, will narrow your search to 2–3 motor families at most.
Decide: Custom Integration or Drop-In Module?
| Factor | Frameless Motor + Discrete Reducer | Integrated Joint Module |
|---|---|---|
| Per-unit cost at volume | Lower (often 50–70% of module cost) | Higher |
| Engineering time to prototype | 3–6 months | 2–6 weeks |
| Mechanical flexibility | Full — you own the joint design | Limited to module envelope |
| Supply chain control | You manage motor, reducer, encoder separately | Single supplier |
| Best for | Production programs, weight-critical designs | Prototyping, time-critical programs |
Most serious robotics programs end up doing both: modules for the prototype phase, discrete integration for production. Plan for this transition from the start rather than treating it as a surprise at SOP.
Match Motor Architecture to Robot Class
| Robot Type | What Works | Why |
|---|---|---|
| Industrial 6-axis (palletizing, welding) | Servo motors with harmonic or RV reducers | Proven, available, well-supported by industrial drive ecosystems |
| Collaborative robot (cobot) | Frameless BLDC + harmonic + torque sensing | Torque sensing at the joint is non-negotiable for human-safe interaction |
| Humanoid (upper body) | Frameless BLDC + harmonic | Same as cobot, with tighter weight and volume constraints |
| Humanoid (legs / hips) | QDD with low-ratio planetary | Backdrivability for ground contact and impact absorption |
| Quadruped | QDD | Same reasoning as humanoid legs |
| Surgical / medical | Frameless torque motor + high-resolution encoder | Precision and traceability requirements exceed other categories |
| AGV / AMR | BLDC drive motors + servo steering | Different problem — continuous rotation, not articulated positioning |
Common Specification Mistakes
These are errors we see repeatedly in RFQs from engineering teams, including experienced ones:
Sizing to peak torque instead of continuous. Your motor needs to survive the thermal duty cycle, not just the worst-case instantaneous load. A motor sized to peak will overheat in the first hour of real-world operation.
Ignoring reflected inertia. A high-ratio reducer amplifies motor inertia by the square of the ratio. A 100:1 harmonic makes a 10 g·cm² motor rotor look like 100 kg·cm² at the joint output. This directly affects your bandwidth and disturbance rejection.
Specifying encoder resolution without specifying encoder type. "19-bit" means very different things for a magnetic encoder versus an optical encoder. Magnetic encoders at 19 bits often have effective resolution closer to 14–15 bits due to noise floor. Specify both the resolution and the technology.
Forgetting about thermal path. A frameless motor's thermal performance depends entirely on how well the joint housing conducts heat away from the stator windings. If your joint housing is thin aluminum with poor surface area, the motor will derate significantly below its catalog spec.
Frequently Asked Questions
What is the difference between a "servo motor" and a "BLDC motor"?
In common industry usage, "servo motor" describes a packaging concept — motor plus driver plus encoder, sold as a closed-loop system. "BLDC motor" describes the underlying motor technology. Most modern servo motors are BLDC motors inside. The distinction matters for purchasing conversations: if you ask for a "servo," you will get quoted a complete package. If you ask for a "BLDC motor," you may get just the motor.
How much torque does a robot joint motor need?
It depends entirely on joint position and payload. Rough ranges for a 6 kg payload cobot: shoulder 40–80 Nm continuous (after reduction), elbow 20–50 Nm, wrist 5–15 Nm. A humanoid hip might need 100–300 Nm. Always calculate from your actual kinematic model — these numbers are sanity checks, not specifications.
Why do cobots and humanoids use frameless motors instead of standard servos?
Weight, integration density, and force control. A standard servo adds its own bearings and housing on top of the joint structure — dead weight that does not contribute to the robot's payload capacity. Frameless motors eliminate that overhead. They also enable coaxial harmonic integration, which is mechanically cleaner and gives better force sensing accuracy through motor current estimation.
Is IATF 16949 certification actually important when sourcing robot motors?
For prototyping and research — no, any reputable motor works. For production programs intended to ship commercially, especially humanoids, medical robots, or robots entering regulated markets — yes, it matters. IATF 16949 guarantees lot traceability, statistical process control on critical dimensions, and documented corrective-action processes. Several major humanoid OEMs now require it during supplier qualification. It is not a marketing badge; it is a supply-chain risk management tool.
What is a QDD actuator and when should I use one?
QDD pairs a high-torque BLDC motor with a low-ratio (6:1 to 10:1) planetary. The low ratio preserves backdrivability — the joint moves freely when pushed by an external force. Use QDD for legs, hips, and any joint that needs to absorb impacts or provide transparent force feedback. Do not use it where positioning accuracy below ±0.1° is required.
Next Steps
If you are still deciding on architecture — motor type, reducer approach, integration depth — this guide gives you the framework. Write your requirements against the four-number method above before contacting any vendor.
If you already know your joint specs:
For general industrial, AGV, and service robotics: contact us through this site for standard BLDC and gearmotor options, including custom winding and gearbox configurations.
For precision robotics — cobots, humanoids, surgical, semiconductor handling: go to Twirl for frameless torque motors, harmonic-integrated joint modules, and QDD actuators. IATF 16949 certified, 200+ patents, and engineering support for custom integration. Send your joint spec — position, torque, speed, encoder, volume, timeline — and expect a technical proposal within two business days.
