Thermal Demagnetization Is the Most Common Magnet Failure Mode Nobody Plans For
Most motor thermal analysis focuses on the winding. The copper loss is calculated, a thermal resistance network is built, and the design is declared safe if the winding temperature stays below the insulation class limit. The magnet temperature gets noted as a secondary output of the same model - typically 10–20°C below the winding - and checked against the maximum operating temperature on the datasheet. If it's below the datasheet limit, the analysis stops.
This approach misses two things. First, the magnet's maximum operating temperature on the datasheet is not a demagnetization safety limit - it's the temperature beyond which the magnet's properties degrade permanently even without an external field. Second, even below that temperature, the combination of elevated temperature and armature reaction field during peak torque events creates demagnetization risk that can't be evaluated from a steady-state temperature alone. This article covers the actual calculation methodology.
1. Why Motor Magnets Run Hotter Than Most Designers Expect
In a surface-mount PM motor, the magnets are bonded to the rotor and separated from the winding copper by the air gap, the slot liner, the tooth iron, and the housing. At first glance they look well-isolated from the primary heat source. In practice, the rotor heats from two sources: eddy current losses in the magnets themselves (significant in segmented magnets with high slot harmonics) and conducted heat from the stator through radiation and convection across the air gap.
Eddy current losses in solid NdFeB magnets scale with the square of the flux density harmonic amplitude and with the square of frequency. In a 6-pole motor running at 3000 RPM with a concentrated winding (high harmonic content), eddy current losses in the magnets can reach 15–30% of total rotor loss. This is not a minor correction. In a 500W motor, 10W of magnet eddy current loss with a thermal resistance of 3°C/W between the magnet and ambient produces 30°C of additional magnet temperature rise above the rotor iron. Combined with conducted heat from the stator, magnet temperatures in continuously-loaded concentrated winding motors regularly exceed winding temperatures by 5–15°C rather than being below them.
The thermal path matters. In an axial flux motor with a double-rotor configuration, the magnets are on the outer surfaces of the rotor discs - they have excellent convective cooling in most geometries and run cooler than the stator. In a deep-slot radial flux motor with a long stack and limited end-turn cooling, the magnet temperature gradient along the axial length can reach 20°C from end to center. The hottest spot in the magnet is not visible to the thermal sensors embedded in the stator, and it's the hottest spot that determines the demagnetization margin.
2. The Correct Demagnetization Safety Calculation
The demagnetization safety margin has two components: the thermal margin (distance from the operating temperature to the temperature at which the knee crosses the operating field) and the field margin (distance from the armature reaction demagnetizing field to the knee at the operating temperature). Both must be calculated.
Start with the load line analysis from the magnetic circuit geometry (covered in the previous article in this series). This gives the operating point (Bop, Hop) in the absence of armature reaction. For a well-designed radial flux motor with PC = 6, Hop is approximately −Br/PC ≈ −0.22 T / μ₀ ≈ −175 kA/m. This is the ambient-temperature, no-load operating point.
Add the armature reaction field. At peak current (typically 3× continuous in a servo drive), the demagnetizing field at the magnet surface is Id × Ld / (μ₀ × Am × Lm), where Id is the d-axis current component and Ld is the d-axis inductance. For a typical servo motor with 10A peak and Ld = 2mH, the armature reaction field at the magnet surface can reach 200–400 kA/m depending on motor geometry. The total demagnetizing field is the sum: −175 + (−300) = −475 kA/m in this example.
Now check the knee position at the magnet operating temperature. For N45 (Hcj = 1380 kA/m at 20°C, temperature coefficient −0.55%/°C), at 120°C the Hcj drops to approximately 1380 × (1 − 0.55% × 100) = 621 kA/m. The knee of the demagnetization curve for N45 at 120°C occurs at approximately 0.5 × Hcj = 310 kA/m in demagnetizing direction. The total field −475 kA/m exceeds the knee field −310 kA/m. This design will experience partial irreversible demagnetization during peak torque events at operating temperature. The calculation shows the failure mode before it appears in test.
The fix is not always a grade upgrade. In this example, the options are: (1) upgrade to N45H (Hcj ≈ 2000 kA/m at 20°C, knee at 120°C ≈ 650 kA/m - safe); (2) increase PC by reducing the air gap, which moves Hop closer to 0 and reduces the sensitivity to armature reaction; (3) reduce peak current via current limiting in the drive; or (4) accept the operating profile and use a lower-grade magnet with better reversible temperature coefficient. Options 2–4 don't require more expensive magnets. Option 1 is the easiest to spec but not always the cheapest.

3. Estimating Magnet Temperature Without FEA
In early design stages, full thermal FEA isn't always available. A simplified lumped-parameter thermal network gives adequate accuracy for initial decisions.
For a radial flux motor, model the thermal resistance path from magnet to ambient as three series elements: magnet-to-rotor-iron thermal resistance (R1 = Lm / (km × Am), typically 0.3–0.8°C/W depending on adhesive layer thickness and conductivity); rotor-iron-to-shaft (R2, depends on rotor geometry, typically 0.5–2°C/W for a solid steel rotor); and shaft-to-ambient via bearing and housing path (R3, typically 1–5°C/W depending on cooling method). For the stator heat conduction contribution, add the air gap thermal resistance (R_gap ≈ Lg / (k_air × Ag) + correction for convection, typically 0.1–0.5°C/W for a 0.5mm gap).
With a simplified network, the magnet temperature rise above ambient is approximately:
ΔT_magnet = P_eddy × (R1 + R2 + R3) + P_stator × (R_gap + R2 + R3) / (R_stator_total)
For a 1kW brushless servo motor with: P_eddy = 5W, R1 = 0.5°C/W, R2 = 1°C/W, R3 = 3°C/W; P_stator = 80W, R_stator_to_ambient = 0.8°C/W, R_gap = 0.3°C/W. The magnet temperature rise above ambient is approximately 5 × 4.5 + 80 × 4.3/0.8 ≈ 22.5 + 430 = incorrect - the P_stator term doesn't work this way because the stator heat mostly flows out through the housing, not through the gap to the magnet.
The corrected simplified approach: the magnet temperature rise due to stator heat is approximately P_stator × (R_gap / R_stator_total) × R_gap, capped at the stator winding temperature minus a fixed gap thermal drop. For most forced-cooled motor designs, the magnet temperature is the stator temperature minus 10–25°C due to the air gap isolation. For passively cooled designs, the gap isolation is less effective and the difference shrinks to 5–10°C. If the stator winding is at 155°C (Class F limit), the magnet is at 130–145°C - which puts N45 into the danger zone calculated above. This is why Class F insulated motors with N45 magnets are a reliability problem waiting to happen in high-peak-current applications.
4. Grade Selection Logic Based on Peak Magnet Temperature
With the peak magnet temperature estimated, the grade selection becomes a direct engineering decision rather than a table lookup.
Below 80°C peak magnet temperature: standard grades (N35–N52) are appropriate. Hcj margin to the knee at 80°C is adequate for most motor applications with peak currents below 3× continuous.
80–120°C peak magnet temperature: H grades (N35H–N52H) are the baseline for anything with peak current above 2× continuous, or concentrated windings (high harmonic content). For distributed windings with peak current below 2× continuous, standard grades with verified load line margins are acceptable.
120–150°C peak magnet temperature: SH grades (N35SH–N48SH) are required for high-peak-current applications. Hcj for N42SH is approximately 2400 kA/m at 20°C, giving knee position at 150°C of approximately 550 kA/m - double the N42 standard grade at the same temperature. For distributed windings with conservative peak current limits, H grades may still be sufficient at the lower end of this range.
Above 150°C peak magnet temperature: EH or AH grades (N35EH–N38EH) for NdFeB, or transition to SmCo Grade 26–28. NdFeB EH grades provide Hcj above 3000 kA/m at 20°C but Br is limited to 1.08–1.18 T range. For applications where torque density matters more than thermal resistance (high-speed compressor drives, aerospace motor drives), SmCo at 150–200°C peak temperature is usually the correct material choice despite the 4–6× cost premium.
One common mistake: specifying an EH grade for a motor design that only needs H because someone on the team was "being safe." EH grades cost approximately 25–35% more than H grades for equivalent Br, and their Br is lower for the same grade number. If the peak magnet temperature calculation shows 130°C, N45H is the correct specification - not N40EH.
5. Verifying the Thermal Margin After Production
The calculation gives a design intent. Verification requires test data.
The standard test for demagnetization margin verification is the short-circuit current test at elevated temperature. The motor is heated to the target operating temperature, then a controlled short-circuit is applied to produce peak armature reaction demagnetizing field. Back-EMF is measured before and after. If the post-test back-EMF is within 1% of the pre-test value, the magnets survived without irreversible demagnetization. A drop greater than 2% indicates partial demagnetization occurred.
For production acceptance, this test is expensive to run on every motor. Instead, suppliers typically verify batch-level magnetic properties: Br uniformity within ±1.5% within a batch (not lot-to-lot - batch uniformity matters for demagnetization margin consistency), Hcj above the specification minimum (usually 10% above the minimum required for margin), and magnetization uniformity via Hall scan. These verify that the magnets delivered to production have the properties assumed in the design calculation.
HIMAGNET provides full demagnetization curves at application temperatures and can cross-reference your load line analysis against batch-specific measured Hcj values before shipment. For motor designs with tight thermal margins - Class F insulation, peak-to-continuous torque ratio above 3×, or operating environments above 60°C ambient - request the high-temperature batch certificate (Hcj at 100°C and 120°C measured values, not nominal) when placing your order.





