+86-574-87520996
sales@fzmag.com
Subtitle: From radial flux to axial flux, from geared reduction to direct drive — the engineering path to miniaturization and high torque density in joint motors
The spatial constraint of a robot joint is the first premise of all joint-motor design. Robot joints — especially in humanoid robots, collaborative robots, and quadruped robots — have strictly limited installation space.
The joint is usually confined to a roughly spherical or short-cylindrical volume: the diameter can be somewhat large, but the axial length is extremely precious. The reason is simple: joints are connected in series, and the axial length of each joint directly determines the length, bending radius, and weight distribution of the entire leg or arm. If one joint grows 20 mm axially, a four-joint leg grows 80 mm, and the robot's center of mass, inertia, and power consumption all shift accordingly.
The core contradiction of a joint motor is therefore very clear: output as much torque as possible within a limited diameter.
The traditional solution is "cylindrical motor + harmonic drive." The cylindrical motor itself has limited torque density, so it must rely on the harmonic drive to drop speed and multiply torque by 50–100×. But the reducer brings three problems: first, volume expansion (the reducer itself is often thicker than the motor); second, backlash and return error (affecting positioning accuracy); third, wear and efficiency loss (harmonic drive efficiency is typically 70–90%, meaning 10–30% of torque is consumed internally).
The flat rotor (axial-flux / disc) approach takes a different path: trade diameter for axial length. Make the motor a flat disc; the large-diameter disc structure outputs high torque directly, eliminating the reducer. The diameter can be allowed to grow somewhat, but the axial length shrinks dramatically — exactly the dimension joints are shortest on.
The flat rotor, called a pancake rotor or disc rotor in English, corresponds to the axial-flux motor topology. Its structural feature: the rotor is a flat disc, with permanent magnets mounted on a disc-shaped steel back-iron (usually in a Halbach array), the stator is a flat disc winding (or PCB winding), and the air gap is distributed along the axial direction.
Compared with the traditional cylindrical rotor (radial-flux):
· Cylindrical rotor: Long cylindrical shape, magnets wound on the inner wall of the rotor circumference, flux enters and exits along the radial direction, air gap is radial. Typical length-to-diameter ratio L/D = 1–3.
· Flat rotor: Disc shape, magnets mounted on the disc face, flux crosses the air gap along the axial direction, air gap is axial. Typical L/D = 0.1–0.3.
This geometric difference is the root of all the flat rotor's advantages. When L/D drops from 2 to 0.2, the motor goes from a "long rod" to a "flying disc" — at the same outer diameter, it sacrifices axial length in exchange for a larger disc area and a longer effective lever arm.
That flat rotors achieve "smaller volume, higher torque" is essentially the superposition of five physical mechanisms:
1. Large diameter = large torque arm. The mathematical definition of torque is T = F × r. In a motor, the electromagnetic force F is determined by the air-gap flux density and current, while the lever arm r approximates the effective radius of the rotor. A flat rotor's outer diameter can reach 80–300 mm, with the effective radius approaching half the diameter; a cylindrical motor's effective radius is also limited by the outer diameter, but the cylindrical motor's axial dimension is stretched and its diameter compressed, so its overall lever-arm utilization is inferior to the flat rotor. Under the same outer-diameter constraint, the flat rotor gets a longer usable lever arm.
2. Axial flux + Halbach array = high air-gap flux density. Traditional radial-flux motors lose half their magnetic flux crossing through air back to the magnet back-side, with poor self-shielding. The Halbach array, through a special magnetization-direction arrangement, concentrates almost all flux on the air-gap side, with the back-iron side nearing zero field. This raises air-gap flux density by 25–40%. The flat rotor's air-gap area = π × (D/2)² — when L << D, this area is far larger than the cylindrical motor's air-gap area (about 2π × r × L). Larger area × higher flux density — the product is a larger total flux and larger torque.
3. High pole count = high torque density. The flat disc geometry easily accommodates 16–48 magnetic poles, while cylindrical motors typically have only 4–12. More poles means higher average torque per electrical cycle, with torque ripple averaged out over more cycles, and significantly higher continuous torque density per unit volume. Empirically, flat-rotor direct-drive torque density can reach 20–50 kN·m/m³, with top-tier Halbach solutions hitting 50–100 kN·m/m³, while cylindrical motors are typically 5–15 kN·m/m³.
4. Direct drive, no reducer. Flat rotors are naturally suited to low-speed high-torque design. They usually pair with direct-drive or low-speed high-torque operation, needing no harmonic drive. This not only saves the reducer's volume but also eliminates backlash, wear, and efficiency loss — equivalent to merging "motor + reducer" into one component, so total volume is actually smaller.
5. Higher surface-area-to-volume ratio = better cooling. The flat structure's outer surface area relative to volume is larger (a disc is more "spread out" than a column), making cooling easier. For high-power-density motors, cooling capacity often determines how long peak torque can be sustained. The flat rotor's steel back-iron serves as both magnetic circuit and heat sink, further strengthening thermal management.
6. Hollow-shaft design = routing space. The center of a flat rotor can be made a hollow shaft (through-bore), letting power cables, signal wires, and air tubes pass through the joint center. This not only simplifies robot wiring but also avoids cable fatigue from repeated bending during joint motion. Cylindrical motors can also have hollow shafts, but the flat rotor's hollow-shaft proportion is more natural and wider.
The comparison is most intuitive with same-torque-level data. Suppose the target output torque is 30 N·m in both cases:
· Cylindrical motor + harmonic drive: Motor diameter 80 mm, axial 60 mm; reducer axial 40 mm; flange 20 mm; total axial about 120 mm.
· Flat-rotor direct drive: Diameter 100 mm, axial 40 mm, no reducer, total axial 40 mm.
Diameter increases 25%, but axial length drops 67%, and total volume decreases by about 40%.
For robots, axial length is often more precious than diameter. Saving 80 mm of axial length per joint lets the entire leg be shorter, lighter, and with a smaller bending radius. This is why humanoid and collaborative robots increasingly adopt flat-rotor solutions — even if the diameter is larger, the joint can "sit" inside the body rather than "stick out."
The flat rotor's high torque is not a single-point breakthrough but a coordinated optimization across three levels — magnetic, thermal, and structural:
Magnetic level:
· Halbach array: air-gap flux density +25–40%
· High-grade magnets: N48/N50SH/N40UH, higher remanence Br and better temperature stability
· Multi-pole design: 16–48 poles, torque density multiplied
· Fractional-slot concentrated windings: shorter end-turns, higher slot fill factor, lower copper loss
Thermal level:
· Flat structure's high surface-area-to-volume ratio, good natural cooling
· Steel back-iron serves as both magnetic circuit and heat sink, short thermal path
· Direct drive has no reducer loss, fewer heat sources (the harmonic drive alone eats 10–30% of torque)
Structural level:
· Hollow through-bore shaft, integrated routing
· Integrated magnetic encoder (no separate position sensor, further shorter axial)
· Modular stacked-disc design (multi-disc stacking compensates axial length, torque linearly accumulates)
Measured torque density comparison (typical):
· Cylindrical motor: 5–15 kN·m/m³
· Flat-rotor direct drive: 20–50 kN·m/m³
· Top-tier flat rotor (Halbach + high grade): 50–100 kN·m/m³
The trade-offs among three joint-motor approaches:
Cylindrical rotor + harmonic drive: High torque density (reducer multiplies), but long axial, has backlash, has wear, medium efficiency. Suited for scenarios extremely sensitive to diameter but not to axial length (such as finger and wrist small joints).
Cylindrical rotor direct drive: No backlash, high efficiency, but medium torque density, no volume advantage. Suited for small-to-medium torque, high-precision scenarios.
Flat-rotor direct drive: Short axial, high torque density, no backlash, good cooling, but large diameter and high initial cost. Suited for large-torque main joints like hip, knee, shoulder.
There is no "best" solution, only "most appropriate." In fact, a single humanoid robot often mixes all three: hip and knee use flat-rotor direct drive, elbow and wrist use cylindrical direct drive, fingers use miniature cylindrical or flat.
Specifically on the robot body, the distribution of flat rotors follows a clear pattern:
Hip joint: Requires large torque and medium speed; the most typical application of flat-rotor direct drive. Humanoid robot single-hip torque often needs 30–80 N·m; a cylindrical motor + reducer is too bulky, so flat-rotor direct drive is the mainstream choice.
Knee joint: Quadruped robot knees commonly use disc Halbach multi-disc stacked structures (our used outer diameter Ø50–100 mm, 16–28 pole solution), with extremely high torque-density requirements, because the knee must bear the whole machine's weight and impact.
Shoulder joint: Similar to hip — large torque, multiple DOF, flat-rotor direct drive common.
Elbow joint: Medium torque (15–40 N·m), flat or cylindrical direct drive both work.
Wrist joint: Small torque (≤5 N·m), small diameter (≤50 mm), miniature flat or cylindrical more appropriate.
Finger joint: Ultra-small torque (0.2–1.5 N·m), miniaturized, usually Ø15–30 mm cylindrical or ultra-thin flat.
You can see: the closer to the torso and the larger the torque, the more the joint leans toward flat rotors; the closer to the end-effector and the smaller the torque, the more cylindrical or miniature solutions can be used.
Flat rotors are not without cost. Their engineering challenges concentrate in four areas:
Axial-flux manufacturing difficulty. Disc winding manufacturing is far more complex than cylindrical winding, especially the disc arrangement of fractional-slot concentrated windings and the insulation and fixation of end-turns. PCB windings (putting the winding on a printed circuit board) are one solution, but current-carrying capacity and cooling are bottlenecks.
Thermal management. The flat structure cools well, but the power density is also higher, so the heat-flux density is actually larger. Instantaneous temperature rise under peak torque, steady-state rise under continuous torque — both need precise thermal simulation. The steel back-iron as heat sink is a double-edged sword — it conducts heat well but also draws heat near the magnets, requiring control of the magnet operating point not to exceed its temperature ceiling.
Cogging torque. Multi-pole motors naturally have large cogging torque (the periodic attraction between rotor magnets and stator teeth during rotation), which causes low-speed jitter. Compensation means such as skewed magnets, fractional slots, magnet eccentricity, and tooth chamfering are needed. The flat rotor's multi-pole nature makes this problem more prominent.
Bearing and precision. Large-diameter disc rotors demand high bearing load capacity and runout precision, especially under hollow-shaft design where the bearing inner diameter is large and the wall thin — rigidity is a challenge. Dynamic balance grade usually requires G1.0 or better.
Cost control. Initial cost is higher than the "cylindrical + reducer" solution. The Halbach array's magnet arrangement and disc winding manufacturing are both more expensive than traditional solutions. But as production ramps up, this gap is shrinking fast.
Flat rotors are still evolving rapidly:
1. Higher-grade magnets. N55, higher-temperature SH/UH/AH grades, continuing to raise same-volume torque.
2. Low-cost Halbach. Current Halbach magnet arrangement relies on precision magnetization and multi-arc-segment assembly, with high cost. New technologies like compression-molded bonded Halbach and 3D-printed magnets are lowering the barrier.
3. 3D-printed windings. Complex disc winding structures can be formed in one additive-manufacturing step, raising slot fill, shortening end-turns, improving cooling.
4. Integrated drive. Motor + driver + encoder + bearing in one encapsulated unit, further shortening axial length and raising power density.
5. Rare-earth-free exploration. Rare-earth-free permanent magnets like ferrite and iron nitride have shown potential in the lab, but it will be difficult to replace NdFeB in torque density in the short term; commercialization still needs time.
The flat-rotor solution for robot joint motors is essentially an engineering choice of "geometry for performance" — trading diameter for axial length, direct drive for reducer, Halbach for flux density. It lets the joint output larger torque in smaller axial space, and is one of the key technology paths for humanoid, collaborative, and quadruped robots moving toward lightweight and high-torque design.
As high-grade magnets, low-cost Halbach, and 3D-printed windings mature, the torque density of flat rotors will keep rising and the cost will keep falling.
For flat-rotor magnetic-circuit design, Halbach array arrangement, custom sizing, or sample requests, please feel free to reach out — FAIZEAL can provide technical support.
