Temperature Rise And Demagnetization of Frameless Torque Motors

1. A Counterintuitive Phenomenon: Why Is Slower More Dangerous?

For many engineers, the intuition is that the faster a motor runs, the hotter it gets. But in the world of frameless torque motors, this logic is exactly reversed.

A frameless torque motor has no housing, no bearings, and no output shaft. It is essentially a power core consisting of a "bare rotor + stator" embedded directly into a mechanical structure. It is packed into extremely cramped cavities such as robot joints, collaborative robot arms, and direct-drive rotary tables, where heat dissipation conditions are far worse than those of conventional housed motors. And its most common operating scenario—low-speed joint holding and high-torque output—is precisely the working condition most likely to "cook" the magnets.

The reason is simple: at low speed, the back EMF is very low, so the drive must continuously output a large current to maintain torque. Copper loss is proportional to the square of the current, and this heat all accumulates in the stator windings, while the magnets on the rotor have almost no direct heat dissipation path. The result is that the lower the speed, the hotter the magnets. In a 5-km continuous walking test of a humanoid robot, the joint temperature measured 80°C, already approaching the high-temperature failure limit of most greases.

Temperature rise is not an isolated problem. It simultaneously threatens the winding insulation (accelerated aging), bearing lubrication (performance degradation), and, most critically, the permanent magnets themselves. Once irreversible demagnetization occurs, the motor's torque capability is permanently degraded.

2. The Two Faces of Demagnetization: One That "Comes Back," One That Never Will

NdFeB (neodymium iron boron) is currently the mainstream magnet material for frameless torque motors. Its Curie temperature is about 310–370°C, which appears far away from the motor's operating range. But what really determines how long a magnet can be used is not the Curie temperature, but the temperature coefficient of coercivity—typically between -0.55%/°C and -0.65%/°C.

The meaning of this coefficient is that for every 1°C increase in temperature, the magnet's ability to resist an external demagnetizing field drops by 0.55% to 0.65%. From 20°C to 80°C, coercivity has already lost about 36%. At 100°C, the loss is close to half. When coercivity falls below the demagnetizing field generated by the armature reaction, the magnetic domain alignment is permanently damaged.

Demagnetization falls into two categories:

Reversible demagnetization is a "false alarm." When temperature rises, magnetic force decreases; when temperature falls, it recovers. The motor exhibits "higher current when hot, recovery when cold." It seems fine, but in reality it is already hovering at the critical edge.

Irreversible demagnetization is the real trouble. When the temperature exceeds what the magnet can withstand, or when a sufficiently strong reverse magnetic field is superimposed at high temperature, part of the magnetic domain alignment is permanently damaged, and the magnetic force cannot fully recover even after the temperature returns to normal.

Irreversible demagnetization also has "partial" and "overall" forms. Partial demagnetization often starts at the corners of the magnet—eddy currents concentrate at the corners, where the critical temperature is breached first. It manifests as a 3%–5% drop in the back-EMF constant KE; at the same speed, higher current is required to maintain torque, the temperature rises further, and a vicious cycle forms. Overall demagnetization is the consequence of the entire machine exceeding 120°C for a long time, with KE dropping by more than 10%. The motor must "add 10% throttle" to do the same work, entering a "thermal-magnetic" positive-feedback death spiral.

For commonly used NdFeB magnets, the critical temperature for irreversible demagnetization is usually between 80°C and 150°C, depending on the magnet grade. High-temperature grades can push the threshold close to 150°C or even higher, but the price is a trade-off between magnetic performance and cost.

3. Where Does the Heat Come From, and Where Does It Get Trapped?

The heat sources of frameless torque motors are more complex than those of conventional motors and can be roughly grouped into four categories:

Copper loss is the most basic. The I⊃2;R loss generated by winding resistance under continuous current is proportional to the square of the current. The problem is that, in pursuit of high torque density, frameless motors often use designs with a high slot fill factor and thin wire, so the winding resistance itself is relatively large. Even a small current produces considerable baseline heat and forms a positive feedback loop of "heating → resistance increase → more heating."

Iron loss is particularly prominent in high-pole-count designs. The commonly used 16–32 pole designs in frameless torque motors make the magnetic field alternating frequency much higher than that of ordinary servo motors. Even at low speed and no load, hysteresis and eddy current losses in the iron core are already considerable.

Rotor eddy current loss is an easily overlooked hidden heat source. The stator cogging effect and PWM carrier harmonics induce eddy currents on the surface of the rotor magnets. This heat is generated directly inside the magnets and has almost no path to conduct outward.

Additional losses introduced by control strategy are the most concealed category. FOC field-orientation error, continuous activation of zero-speed high-frequency injection, and continuous corrective currents caused by improper PID parameters—these "invisible currents" continuously generate copper loss even under no-load conditions, which is the root cause of many engineers finding that "it is also hot at no load."

After these four heat paths are superimposed, the actual magnet temperature is usually 8–12°C higher than the housing surface temperature. When you read 70°C on the housing surface with an infrared thermometer, the magnets may well have stepped onto the 80°C red line.

4. Three Hard Lines in Selection

Once the temperature-rise mechanism is understood, the logic of selection becomes clear. Temperature selection for frameless torque motors can be summarized into three hard lines, corresponding to three different levels of protection targets.

The first line: magnet grade determines the "bottom line"

NdFeB magnets are divided into multiple grade levels according to intrinsic coercivity, with temperature resistance increasing step by step:

The N series is the standard grade, suitable for room-temperature applications, with a temperature resistance of about 80°C. The H series withstands 120°C, the SH series 150°C, and the UH series 180°C. From N to UH, heavy rare earth elements Dy and Tb are added to improve the temperature coefficient of coercivity, but the cost also rises by a factor of 1.5 to 3.

For typical frameless torque motor applications—robot joints and collaborative robot arms—SH grade or higher is the recommended starting point. The reason is that frameless motors inherently have harsh heat dissipation conditions, and the "hidden differential" between magnet operating temperature and housing temperature can reach 15°C. The N series may look sufficient on paper, but in actual operation it can easily cross the line. The first-generation knee-joint motor of a humanoid robot, for example, did not verify demagnetization under stall conditions. After repeated fall impacts, the magnets suffered irreversible demagnetization, and the output torque permanently dropped by more than 15%.

The second line: insulation class determines the "ceiling"

Winding insulation class is another temperature constraint, and it often reaches its limit earlier than the magnets. Class F insulation has a rated temperature of 155°C, which is currently the mainstream configuration for frameless torque motors. Class B insulation (130°C) appears in some cost-sensitive or extremely space-constrained products.

There is thermal coupling between magnet temperature and winding temperature: winding heat is conducted through the stator core to the rotor and ultimately affects the magnets. Therefore, the 155°C winding insulation limit actually sets an indirect ceiling for magnet temperature as well. The engineering verification logic is that, under thermal steady state, the winding temperature must not exceed the insulation class limit, while the magnet temperature must not exceed its demagnetization critical temperature—the latter usually becomes the bottleneck earlier.

In industry practice, the maximum continuous operating temperature of NdFeB SH-grade magnets is usually set at 120°C. Above this value, the motor begins to enter the demagnetization risk range.

The third line: current limit determines the "dynamic boundary"

In addition to steady-state temperature, there is another dynamic factor that is often overlooked: demagnetization risk under peak current.

Torque motors have peak torque and stall torque. Peak torque is the starting torque for a short time, usually lasting only 10–15 seconds; stall torque is the value that can be output continuously. When peak current is too large, the demagnetizing field generated by the armature reaction may directly trigger irreversible demagnetization under high-temperature conditions.

The maximum current of a direct-drive torque motor limited by permanent magnet demagnetization is its "demagnetization limit point." In some products, after high-performance magnets are used, torque is limited only by maximum temperature rise, and peak torque can be greatly increased. But if the magnet grade is insufficient, demagnetization becomes the first constraint triggered.

During selection, it is necessary to verify not only the temperature rise under continuous current, but also whether the current under stall and transient overload conditions exceeds the magnet's demagnetization limit at that temperature. This limit is usually given by the magnet supplier's demagnetization curve and must be determined in combination with armature reaction calculations in the motor design.

5. Engineering Approaches to Cooling

The essence of temperature-rise control is to establish a balance between heat generation and heat removal. For frameless torque motors, there are several mature engineering paths.

Potting is currently the most mainstream means of enhancing heat dissipation. TQ Robodrive uses epoxy encapsulation and potting technology, and Kinco pioneered a frameless potting process. The core logic of both is to fill the internal gaps of the motor with high-thermal-conductivity potting material, so that winding heat is conducted more quickly to the stator core and mounting flange. The thermal conductivity of potting material can reach 1.4–1.6 W/mK, far higher than that of air.

Water-cooling channels are suitable for compact joints and other scenarios with high heat flux density. Helical channels are cut into the stator sleeve or housing, and coolant is passed through them, which can directly reduce temperature rise by 40–60°C. The cost is increased system complexity and sealing requirements.

Control strategy optimization is a zero-hardware-cost path. Recalibrating the magnetic pole zero point, optimizing PID parameters, managing zero-speed high-frequency injection, and enabling minimum-current control can reduce the static holding current by 50%–80% without changing the hardware. For systems that already have no-load heating problems, control optimization often produces more immediate results than adding heat dissipation measures.

Built-in temperature sensors are the safety bottom line. NTC (100k) and PT1000 are currently the standard temperature measurement configurations for frameless torque motors, monitoring winding temperature in real time and triggering derating or protection when approaching the limit. But note that the sensor measures winding temperature, not magnet temperature—the difference between the two depends on the design quality of the heat dissipation path, and sufficient safety margin must be left during selection.

Conclusion

Temperature selection for frameless torque motors is essentially a balance among three constraints: the magnet grade determines how high a temperature you can withstand, the insulation class defines the winding ceiling, and the current limit defines the boundary of dynamic operating conditions. The three lines are coupled to one another, and if any one is breached, it can trigger a chain reaction from demagnetization to burnout.

The most pragmatic approach is: at the selection stage, assume the harshest heat dissipation conditions, move up one magnet grade, leave ample temperature-rise margin, and make demagnetization verification under stall and overload conditions a mandatory item. After all, demagnetization is irreversible—it will not give you a second chance.