Frameless Servo Motor Thermal Design: Why Air Gaps Kill Continuous Torque

When engineers select a frameless servo motor, they usually look at the torque-speed curve first. They see a "Continuous Torque" value, check if it meets their requirement, and move to mechanical CAD.

This is the most common integration failure mode in direct-drive robotics.

A frameless motor has no outer housing. It has no cooling fins. It has no fan. The motor does not have a continuous torque rating—the entire thermal assembly has a continuous torque rating. The stator is merely the heat source; the customer's housing is the heat sink. If the thermal path between the two is flawed, a motor rated for 2.0 N·m on paper will overheat at 1.0 N·m in reality.

This guide explains the real-world thermal bottlenecks of frameless integration, how to calculate thermal derating, and what OEM buyers must review before freezing a CAD design.

The Bottleneck: I²R Losses vs. Thermal Resistance

In a permanent magnet motor, the primary heat source during continuous operation is copper loss (I²R losses) in the stator windings. The heat must flow from the copper coils $\rightarrow$ through the electrical insulation $\rightarrow$ into the stator lamination stack $\rightarrow$ across the mounting interface $\rightarrow$ into the machine housing $\rightarrow$ and finally to the ambient environment.

The stator generates heat. The housing acts as the sink. An air gap or poor potting limits continuous torque by creating a thermal bottleneck.

As seen in the diagram above, the most critical chokepoint is the stator-to-housing interface.

Even if you use aerospace-grade aluminum for your housing, if the stator is loosely fitted and separated by a microscopic air gap, the thermal resistance skyrockets. Air is a terrible thermal conductor (approx. 0.024 W/m·K).

Derating Matrix: How Housing and Fit Impact Continuous Torque

Catalog continuous torque values are typically measured under ideal conditions: the stator is pressed into a massive, actively cooled aluminum fixture in a 25°C ambient room.

Your application is likely a compact, sealed robotic joint or a closed medical actuator. You must derate the catalog value. Use the following diagnostic table as a baseline before you have physical prototypes.

Calculating the Thermal Stack-Up

To mathematically predict if your motor will overheat, you need to understand the thermal resistance network.

The temperature rise of the winding (ΔT_winding) is calculated as:

ΔT_winding = P_loss × (R_th_internal + R_th_interface + R_th_housing)

• P_loss: Total power dissipated as heat (Watts).
• R_th_internal: Thermal resistance from coils to the outer diameter of the stator (Provided by the motor supplier).
• R_th_interface: Thermal resistance of your mounting gap (This is under your control).
• R_th_housing: Thermal resistance of your housing to ambient air.

If ΔT_winding + T_ambient exceeds the insulation temperature limit (often 130°C or 155°C), the motor will burn out.

The Math Behind the Air Gap

Why is an air gap so destructive? Let's look at the thermal conductivity ($k$):

• Aluminum: 167 W/m·K
• Thermal Epoxy: 1.0 - 2.5 W/m·K
• Static Air: 0.024 W/m·K

An air gap of just 0.05 mm has the same thermal resistance as 3.5 mm of standard thermal epoxy, or nearly 350 mm of solid aluminum! If your slip-fit tolerance is too loose and you don't backfill it with thermal resin, the heat has nowhere to go.

Three Actionable Design Strategies

If your continuous torque requirements are close to the catalog limit, you must optimize the interface.

1. The Shrink-Fit (Interference Fit)

This is the gold standard for robotics. The aluminum housing is heated to expand, the room-temperature stator is dropped in, and as the housing cools, it clamps down on the stator.

• Advantage: Near-zero thermal resistance at the interface. Maximum torque density.
• Risk: If the interference is too high, the compressive stress can deform the stator laminations, altering the magnetic properties and increasing cogging torque. For most frameless motors in the 60-120mm OD range, the recommended interference is 0.01-0.03mm. Always request the supplier's recommended compression limits before machining the housing bore.

2. The Potted Slip-Fit

The stator is machined with a slight clearance fit (e.g., +0.05mm). The gap is injected with a high-conductivity thermal epoxy.

• Advantage: Zero compressive stress on the stator. Much easier assembly and lower tooling cost.
• Risk: Potting compounds can have bubbles (voids). Voids are microscopic air gaps that create localized hotspots, leading to premature winding failure. Vacuum potting is highly recommended. We have seen bubble-related winding burnout on a 110mm motor where standard gravity-fill potting left a 3mm void pocket near the hottest coil.

3. Housing Surface Area Optimization

The heat doesn't just stop at the housing—it must escape to the air. A smooth, compact cylinder has very little surface area.

• Action: If you cannot add cooling fins due to space constraints (e.g., a cobot arm), you must ensure the housing is mechanically coupled to larger, thermally conductive machine structures (like the robot's main links) to act as extended heat sinks.

The following chart shows what happens in practice when you change only the mounting method while keeping everything else identical:

Same 80mm OD motor, same 1.2 N·m continuous load, same aluminum housing. Only the mounting method changes. The set-screw mount reaches thermal shutdown in under 12 minutes.

The data comes from bench tests on an 80mm OD frameless motor running at 1.2 N·m continuous. The shrink-fit assembly stabilized at 78°C with comfortable margin below the 130°C Class-B insulation limit. The potted slip-fit reached 95°C—still within spec but with less margin for higher ambient temperatures. The set-screw mount hit thermal shutdown at 130°C in under 12 minutes, at the same load the other two methods handled easily.

How to Verify Thermal Performance with a Prototype

You do not need a thermal simulation lab to validate the supplier's claims. A K-type thermocouple, a handheld data logger, and 30 minutes of testing can reveal whether the integration is thermally sound.

Prototype thermal validation procedure:

1. Attach a K-type thermocouple to the housing outer surface directly above the stator center, using thermal paste and Kapton tape.
2. If possible, route a second thermocouple into the end-winding area (between the stator and the housing endcap). This gives you the most critical temperature reading.
3. Run the motor at 100% of the target continuous torque at the target speed.
4. Log temperature every 30 seconds for at least 20 minutes, or until the temperature stabilizes (rate of change less than 1°C per 5 minutes).
5. Record the steady-state housing temperature and the steady-state winding temperature (if accessible).

How to interpret the results:

The most important number is the winding-to-housing temperature difference. If the housing is 45°C but the winding is 120°C, you have a 75°C delta across the interface—this means the thermal resistance between stator and housing is unacceptably high, regardless of what the housing surface looks like.

What to Request from Your Frameless Motor Supplier

When submitting an RFQ, do not just send a peak torque target. Send a thermal profile.

Ask the supplier for:

1. R_th (Thermal Resistance): What is the resistance from winding to stator OD? A good supplier will provide a number in °C/W. Typical values for frameless motors range from 0.5 to 3.0 °C/W depending on motor size and construction.
2. Test Conditions: Was the catalog continuous torque measured on an aluminum plate? What size? What was the ambient temperature? Most catalog values assume a 300mm x 300mm x 20mm aluminum plate at 25°C—far better than most real housings.
3. Maximum Housing Temperature: At continuous load, what should the temperature of the external housing be? (This allows you to verify the interface quality during prototype testing using the thermocouple method above).

At Frameless Servo, we provide application-specific thermal derating assistance. If you are designing a sealed actuator and need to know the true continuous torque boundary, share your housing material and ambient conditions via our Contact / RFQ page before you lock your CAD design.