Magnet Tolerances & Acceptance Criteria: What's Achievable, What It Costs

Why magnet tolerances are expensive

Sintered NdFeB starts as pressed powder and shrinks 15–20% in the furnace — and not perfectly uniformly. That means as-sintered dimensions are inherently loose, and precision comes from grinding afterward: a slow, abrasive operation on a material that chips like ceramic. The cost structure follows directly:

• As-pressed/as-sintered surfaces are cheap and loose. If a face doesn't touch anything in your assembly, letting it stay as-sintered saves a grinding operation.
• Every ground surface is a paid operation. Tolerance tightening below standard means more grinding passes, tighter wheel dressing, more inspection — and more scrap, because brittle material chips at edges under aggressive material removal.
• Scrap scales nonlinearly. Halving a tolerance doesn't double the cost; past the standard window, yield loss accelerates and the price curve bends upward fast.

The design conversation this enables: tolerance the interfaces, relax everything else. A magnet with one functional face ground flat and everything else at standard tolerance is dramatically cheaper than the same part toleranced tightly all over — and works identically.

Dimensional tolerances: standard vs. tight

Typical guidance for sintered NdFeB (ceramic runs looser; exact capability varies by geometry and supplier — confirm on the quote):

Magnetic tolerances & lot spread

Magnetic properties have tolerance bands just like dimensions — the difference is that nobody prints them on the drawing, so they arrive as surprises:

• The grade is a window, not a number. N42 defines a Br range roughly 3% wide; add measurement uncertainty and normal process variation and ±3–5% part-to-part and lot-to-lot field variation is normal and conforming. A design that can't tolerate that spread needs either a field-selected (sorted) supply agreement — at a premium — or a design change; the sensor guide shows how field-tolerant architectures sidestep the problem entirely.
• Specify the property you can measure on the finished part. Br and Hcj live on the material cert (per melt); the finished-part checks are total moment (Helmholtz coil) or flux density at a defined point in a defined fixture. Choose one as the acceptance property and put numbers on it.
• Angle tolerances exist too. Magnetization axis alignment to the geometry is typically within a few degrees as standard; diametric parts needing a clocked pole axis must say so with a tolerance ("pole axis to flat: 0° ±2°"), per the directions guide.
• Anchor limits to real hardware. The clean method: measure the approved FAI/sample parts in the production test fixture, set acceptance limits around those values, and reference the fixture in the criteria. Limits derived from theory invite fixture-to-fixture disputes.

Coating & visual criteria

• Thickness: specify the coating system and thickness range (e.g., Ni-Cu-Ni 15–25 µm total), measured by XRF on defined surfaces.
• Adhesion & integrity: reference a test — tape test, thermal shock cycle, or salt spray hours for corrosion-exposed service. "Good adhesion" is not a criterion.
• Visual standards need pictures. Chips, edge flaking, plating nodules, and discoloration are the top rejection causes at incoming — and the top dispute generators, because "minor chip" means different things in two QA departments. A one-page visual standard with photos and a maximum chip dimension (e.g., no chip >0.5 mm on functional faces, >1.0 mm elsewhere) prevents 90% of the arguments.
• Marking: if you require pole marking or lot marking, define method, location, and permanence — marker dots wipe off in ultrasonic cleaning.

Writing enforceable acceptance criteria

An acceptance criterion is enforceable when a third party, given only the drawing, could reproduce the supplier's measurement and reach the same pass/fail verdict. The recurring failures and their fixes:

Every row follows the same pattern: property + number + method + condition. If a line on your drawing is missing one of those four, it's an implied requirement — and implied requirements lose disputes, as the documentation guide puts it.

Sampling plans & incoming inspection

• Define the plan on the PO, not at receiving. A standard AQL scheme (ANSI/ASQ Z1.4, general inspection level II is the common default) with stated AQLs — e.g., 1.0 for dimensional/magnetic, 2.5 for cosmetic — tells the supplier exactly what "tested" means and tells your dock exactly what to pull.
• Match your incoming check to the supplier's outgoing check. Same property, same method, ideally correlated fixtures — the fastest way to prevent "passes here, fails there."
• Weight the checks by failure cost. A dimensional miss is caught at assembly; a magnetic miss is caught in the field. Magnetic sampling deserves the tighter AQL even though it's the harder measurement.
• Handle with care during inspection — measurement itself chips parts when magnets snap together on steel surface plates. Non-magnetic tooling and spacers are part of the inspection plan, per the handling guide.

When a lot fails: handling it well

• Quarantine and re-measure first. Confirm the failure with a second operator and, for magnetic disputes, cross-check fixtures against the retained FAI/reference samples — fixture drift causes a meaningful share of "failures."
• Distinguish nonconforming from unusable. A cosmetic chip outside the visual standard on a hidden face may merit a documented deviation; a Br miss on a field-critical part never does. Deviations are fine when they're deliberate, documented, and priced.
• Demand containment plus root cause, not just replacement. Replacement parts from the same uncorrected process are the same defect on a later ship date. The corrective-action question from the supplier qualification guide exists for exactly this moment.
• Feed it back into the drawing. Most first-year disputes trace to a criterion that was vague. Every resolved dispute should tighten a line on the drawing — that's the loop that makes year two quiet.

Related: how magnets are tested · RFQ preparation guide