NdFeB Magnet Tolerances: Dimensional Specs, Surface Finish, and What to Put on Your Drawing

Sintered NdFeB production starts with hydrogen decrepitation, jet milling, pressing in a magnetic field, vacuum sintering at 1050-1080 C, and annealing. The resulting blank shrinks unevenly: 20-25% along the press axis (which becomes the magnetization direction) and 10-12% perpendicular. This differential shrinkage is why as-sintered tolerances are loose. After sintering, blanks are ground to final dimensions using diamond or CBN wheels with water-soluble coolant. Each process stage determines what tolerances are achievable.

• •As-sintered (no grinding): +/-0.1 to +/-0.15 mm. Suitable for non-critical dimensions where the magnet is adhesive-bonded into a pocket with clearance.
• •Standard grinding: +/-0.05 mm. The default for motor magnets. One pass per face on a surface grinder with in-process gauging.
• •Precision grinding: +/-0.02 to +/-0.03 mm. Requires slower feed rates, finer grit wheels, and 100% gauging. Adds 15-30% to machining cost.
• •Wire EDM: +/-0.01 to +/-0.02 mm. Used for complex profiles (arcs, slots, chamfers) where grinding geometry is impractical. Slow and expensive but highly accurate.
• •Lapping: +/-0.005 to +/-0.01 mm. Reserved for sensor magnets and scientific instruments. Rarely justified in motor applications.

The press axis during compaction becomes the magnetization direction (the axis along which the crystallographic easy axes are aligned). This axis experiences the most shrinkage during sintering because the powder particles are compacted most densely in this direction. The result is that dimensional control along the magnetization axis is inherently worse than perpendicular axes. A block pressed to 10 mm in the magnetization direction might sinter to 7.5-8.0 mm, while a block pressed to 10 mm perpendicular sinters to 8.8-9.0 mm. The larger and less predictable shrinkage along the press axis means more grinding stock is needed and the as-sintered variation is wider. For motor magnets where the magnetization direction corresponds to the magnet thickness (the dimension between the rotor surface and the back iron), this is usually the least critical dimension geometrically. Specify the widest acceptable tolerance on this dimension. If your design needs +/-0.05 mm on width and length, consider whether +/-0.08 or +/-0.1 mm is acceptable on thickness. The cost difference is meaningful at volume.

Ground sintered NdFeB surfaces typically measure Ra 0.8 to 1.6 microns (32 to 63 microinches). This finish is adequate for most coating adhesion and assembly bonding. Finer finishes are achievable: precision grinding reaches Ra 0.4 to 0.8 microns, and lapping reaches Ra 0.1 to 0.2 microns. The practical question is whether finer finish improves your application. For NiCuNi plating, the electroless nickel strike layer bonds well to Ra 1.6 micron surfaces. For epoxy coating, a slightly rougher surface (Ra 1.6 to 3.2) actually improves adhesion through mechanical keying. For adhesive bonding of magnets into rotor slots, Ra 1.6 is optimal for structural adhesives like Loctite AA 326 or 3M DP460. Specifying Ra 0.4 on a surface that will be epoxy-coated or adhesive-bonded adds cost without functional benefit.

Coatings add material to every surface and must be included in the final dimensional tolerance analysis. NiCuNi plating adds 10-20 microns (0.01-0.02 mm) per side, so a magnet ground to 10.00 mm becomes 10.02-10.04 mm after plating. Epoxy coating adds 15-25 microns (0.015-0.025 mm) per side. Zinc plating adds 5-10 microns per side. Parylene conformal coating adds 10-25 microns per side. The coating thickness variation itself has tolerances: NiCuNi is typically +/-5 microns, epoxy is +/-10 microns. For tight-fitting rotor slot applications, specify the final coated dimension with tolerance, not the bare magnet dimension. Let the supplier back-calculate the grinding target from your coated specification. This avoids the common mistake of specifying a bare magnet tolerance that, after coating, pushes the part out of the assembly envelope.

• •NiCuNi: +10 to +20 microns per side (+/-5 micron variation)
• •Epoxy: +15 to +25 microns per side (+/-10 micron variation)
• •Zinc: +5 to +10 microns per side (+/-3 micron variation)
• •Parylene: +10 to +25 microns per side (+/-5 micron variation)

A complete magnet drawing should include six elements. First, dimensions with tolerances on every dimension, including chamfers and radii. Mark which dimensions are critical (tight tolerance) and which are reference (wide tolerance). Second, magnetization direction shown with a clear arrow and labeled (e.g., 'Magnetized through thickness'). Third, coating specification including type, minimum thickness, and whether the final dimension is pre-coating or post-coating. Fourth, magnetic property requirements: minimum Br, minimum Hcj, minimum BHmax, and the test temperature if not 20 C. Fifth, inspection criteria: what gets measured, at what frequency (100%, AQL sampling, SPC), and the measurement method (caliper, micrometer, CMM). Sixth, marking and packaging requirements if applicable. Drawings missing the magnetization direction or magnetic property specs force the supplier to make assumptions and often result in requotes after clarification rounds that add 1-2 weeks to the timeline.

The most expensive tolerance on a magnet drawing is the one that is tighter than necessary. Three common patterns drive unnecessary cost. First, specifying +/-0.02 mm on all dimensions when only one dimension (usually the width that fits a rotor slot) actually needs it. A drawing with +/-0.02 on width, +/-0.05 on length, and +/-0.1 on thickness costs 20-30% less to produce than +/-0.02 on all three. Second, specifying a surface finish tighter than Ra 0.8 on surfaces that will be coated or bonded. Third, calling out flatness or parallelism on thin magnets (under 3 mm) without understanding that sintered NdFeB has a natural bow of 0.02-0.05 mm per 25 mm of length after grinding, and flattening it requires either lapping or adhesive fixturing during grind. Each of these adds cost and lead time without improving the magnet's performance in the assembly.