The magnet as an engineered component
The habit that produces good magnet designs is treating the magnet the way you treat a bearing or a spring: specified by the function it performs, verified by measurement, toleranced where it matters. In practice that means the primary requirement is usually field at a working point — B at a defined distance from a defined face, or total moment, or pull force at a gap — not a grade name and a size copied from a catalog.
- Design the field, derive the part. Geometry drives field-at-distance far more than grade steps do — a grade step buys only a few percent, per the grades chart math. Size the face and thickness for the gap first; pick the grade second.
- Design for the worst corner, not the bench. Hot magnet, maximum gap, minimum-strength lot — if the function works there, it works. The pull force guide covers why catalog forces and real-assembly forces diverge.
- Decide the magnetization direction deliberately. Axial, diametric, true radial, multipole — it changes the tooling, the supplier pool, and sometimes the architecture; definitions and drawing language in the directions guide. Sensor systems get their own treatment in the sensor magnet guide.
Material & grade: windows, not numbers
The material families divide the design space quickly — the full trade study is in the material comparison:
| material | reach for it when | watch |
|---|---|---|
| Sintered NdFeB | Maximum energy in minimum volume — the default for motors, sensors, holding | Temperature sensitivity, corrosion without coating, brittleness, rare earth supply exposure |
| Bonded NdFeB | Complex shapes, fine multipole patterns, over-molding onto hubs | Roughly half the flux of sintered — the trade mapped in sintered vs. bonded |
| SmCo | High temperature, thermal stability (3–4× less drift), corrosion resistance | Cost, brittleness even beyond NdFeB |
| Ferrite | Cost-driven volume, corrosion immunity | Coercivity falls when cold — demag check at minimum temperature, not maximum |
| Alnico | Extreme temperature, legacy sensitivity curves | Very low coercivity — easily demagnetized by design |
Then internalize the fact that shapes everything downstream: a grade is a property window, not a value. N42 defines a Br range; add process variation and ±3–5% part-to-part and lot-to-lot field spread is normal and conforming. The design choice is binary — either the architecture tolerates that spread (angle-based sensing, zero-crossing schemes, holding designs with margin), or you pay for sorted supply forever. Choose deliberately; the tolerant architecture is almost always the better product.
Geometry, tolerances & DFM
Sintered magnets are pressed powder, shrunk unpredictably in a furnace, then ground to precision like a ceramic. The cost rules fall straight out of that process — detail in the tolerances guide:
- Tolerance the interfaces; relax everything else. Every ground surface is a paid operation. One functional face ground flat with the rest at standard (±0.05 mm) is dramatically cheaper than tight-all-over — and works identically.
- Always allow an edge break. Sharp edges chip and shed plating; a 0.1–0.4 mm chamfer is free reliability.
- Mind thin-part limits. Thickness-to-diameter ratios below roughly 1:10 bring flatness caveats, handling scrap — and a magnetic penalty covered in section 04.
- The coating lives inside your dimension. Standard Ni-Cu-Ni adds up to ~0.05 mm on a diameter — an entire tolerance band. State whether dimensions apply after coating (convention: yes), and never design a press-fit to bare-magnet numbers.
- No drilling, no threading, no sharp-cornered pockets. Sintered material cracks under point loads and hoop tension; countersinks are pressed in tooling, not machined later. Retention strategy comes from the bonding & mounting guide — shear-loaded joints, pockets, capture features.
Temperature & the load line
Heat is the number-one field failure mode, and the spec that prevents it has two halves — both from the temperature guide:
- The class rating is conditional. “80 °C” or “150 °C” assumes a favorable permeance coefficient. A thin, wide disc in open air self-demagnetizes hard and can lose flux below its rated class; the same material in a closed circuit can exceed it. For anything above ~60 °C, ask for the load-line check with the real geometry and circuit — it's a standard calculation.
- Coercivity falls 4–6× faster than flux. A hot magnet isn't just weaker; it's fragile against demagnetizing fields — the reason motor rotors and opposing-magnet designs need the class checked against the worst-case field, not just the worst-case temperature.
- Don't buy letters you don't need. Temperature classes are the most expensive characters in the business — the premium structure is in pricing explained. Re-justifying an over-specified class is also the #1 cost-down lever in the cost reduction guide.
- Sequence process heat. Reflow exceeds every NdFeB class; cure ovens and autoclaves need checking. Options: magnetize after the hot step, mount downstream, or SmCo. Calibrated devices get thermal stabilization specified on the drawing.
Coating & environment
Uncoated NdFeB corrodes; the coating is a design variable, not a finish. Selection logic — full menu in the coatings guide:
Match the coating to the real environment — humidity, condensation, chemicals, washdown, immersion — and to the assembly method: if the joint is bond-critical, the coating choice from the bonding guide's compatibility table matters as much as corrosion resistance. Then spec it enforceably: system, thickness range, and a referenced test (XRF thickness, salt-spray hours), not “nickel plated.”
The drawing callouts that prevent disputes
Every requirement on a magnet drawing needs the four-part pattern — property + number + method + condition — or it's an implied requirement that loses disputes. The callouts that separate a quiet production year from a loud one:
- The functional magnetic requirement: “B ≥ X mT on-axis at Y mm from the marked face, at 25 °C, fixture per FAI reference” — or a Helmholtz moment limit. Measurable, enforceable, and it communicates intent better than grade + dimensions ever will; methods in how magnets are tested.
- Direction and clocking: the magnetization axis, its tolerance to a geometric feature (“pole axis to flat: 0° ±2°”), and pole marking method if assembly needs it.
- Grade + material cert requirement: the grade name plus “material cert with demag curve per melt lot” — the paper trail that makes the grade real, per the documentation guide.
- Coating system, thickness, and test — and whether dimensions apply post-coating.
- Visual/chip limits with numbers: maximum chip size by face class beats “no defects” every time.
- Anchor limits to hardware: approved FAI samples measured in the production fixture become the reference standard — limits derived from theory invite fixture-to-fixture arguments.
Designing for sourceability
Design decisions set the supply chain's degrees of freedom for the life of the product:
- Check the stock catalog before cutting a custom. A design nudged to a stock size ships in days forever, skips tooling, and dodges the MOQ math in the MOQ & lead-time guide. The nudge is usually free at concept stage and impossible after tooling.
- Spec function, not one factory's recipe. A field-at-point requirement with a grade window travels between suppliers; a spec built around one supplier's exact material data effectively sole-sources the part — the core problem in the second-sourcing guide.
- Mind the heavy rare earth content. SH/UH/EH classes carry Dy/Tb — the most price-volatile and export-control-exposed ingredients, per the supply chain guide. A design that genuinely needs the class needs it; a design that inherited it from a template is carrying supply risk for free.
- Leave the acceptance criteria portable: methods referenced to standards and FAI hardware you own, not to equipment only one factory has.