01
Why magnets stress a standard QMS
- The critical characteristics are invisible. Grade authenticity, coercivity at temperature, coating porosity, saturation completeness — none is caliper-checkable. Incoming inspection as usually practiced verifies the part's geometry and misses everything that fails in the field.
- Conforming parts vary. A grade is a property window; ±3–5% part-to-part and lot-to-lot magnetic spread is normal and conforming. A quality system that treats that spread as special-cause variation generates false rejections and real disputes.
- The failure modes are substitutions, not defects. N35 shipped as N42; class N where the drawing says H (identical at room temperature, fatal at 100 °C); thin plating that blooms rust in month six; a silent factory change between lots. Each has a specific control — and none of the controls is a tighter caliper.
- The consequence: for magnets, the quality system's weight shifts from inspection of parts to verification of process — documentation, qualification, reference hardware, and targeted measurement. That's the structure of everything below, and of the supplier qualification guide upstream of it.
02
The documentation stack
Each document proves one specific thing — and only that thing. The hierarchy, from the quality documentation guide:
| document | proves | the check that makes it real |
|---|---|---|
| CoC | The supplier asserts conformance for this lot | Lot numbers present, matching cartons, traceable to a melt lot — a CoC without lot numbers proves nothing |
| Material cert + demag curve | The material is the grade | Curve matches the grade's window; grade label matches the drawing; per melt lot, not “typical” |
| FAI report | The process can make the part | Every drawing characteristic measured; magnetic acceptance property included; samples retained (section 05) |
| PPAP-style package | The process is controlled and won't silently change | The change-management commitment is the point — site, material, and process changes notified before shipment |
| RoHS / REACH / CMRT | Regulatory compliance | Onboarding + annual refresh; compliance paper, not quality paper |
Two audit habits that pay for themselves: verify ISO certificates against the registrar's database (fabricated certs are a documented phenomenon), and confirm the cert scope covers the site that makes your part — a trader's ISO 9001 says nothing about the factory.
03
What you can measure — and how
The measurable properties sort by where they live — full methods in how magnets are tested:
on the material cert (per melt) — br, hcj, bhmax
measured on a hysteresisgraph at the factory; you audit the paper, you don't re-measure the material
on the finished part — total moment
Helmholtz coil + fluxmeter; fast, geometry-tolerant, the best single incoming check for “is this the magnet we approved”
on the finished part — field at a point
gaussmeter in a defined fixture at a defined distance; the functional check — and the most fixture-sensitive, which is why limits anchor to FAI hardware
on patterned parts — pole scan
automated field mapping; verifies pole count, placement, and balance on multipole and diametric parts — also the diagnostic that shows thermal loss as a distorted pattern
on the coating — thickness, adhesion, corrosion
XRF thickness on defined surfaces; tape/thermal-shock adhesion; salt spray per ASTM B117 by hours
What a surface-field number from a handheld gaussmeter pressed on the pole is not: a repeatable acceptance property. Probe position, part geometry, and technique swamp the tolerance — the classic source of “passes here, fails there.”
04
Acceptance criteria that hold up
The enforceability test: could a third party, given only the drawing, reproduce the supplier's measurement and reach the same verdict? Every criterion needs property + number + method + condition — the vague-to-enforceable conversion table lives in the tolerances guide. The quality engineer's additions:
- Choose ONE magnetic acceptance property (moment or field-at-point in a defined fixture) and put limits on it. Two properties with independent limits will eventually disagree on a conforming lot.
- Set limits around measured FAI hardware, not theory — the mean of approved samples in the production fixture, with a band that accommodates legitimate lot spread.
- Give visual criteria numbers and pictures: max chip dimension by face class, plating flaking = reject, one page of photos. “Minor chip” means different things in two QA departments, and that difference is where disputes breed.
- Condition everything: temperature of measurement, fixture ID, sample state (stabilized or as-shipped). An uncontrolled condition is a future argument.
05
FAI & reference hardware
- FAI from production tooling and process — hand-finished golden samples qualify nothing. Every drawing characteristic measured, including the magnetic acceptance property in the fixture that will run production.
- Retain the measured samples. Documented and stored, they are the physical reference standard: fixture correlation checks, dispute arbitration, and drift detection all trace back to them. Losing the FAI set converts every future disagreement into theory-vs-theory.
- Re-FAI on the events that actually change the part: new or reworked tooling, site change, material recipe change, coating line change — the notification triggers your change-management clause defines. This is PPAP logic earning its keep outside automotive, per the documentation guide.
- Bake a set for temperature-critical parts. An afternoon at operating temperature with before/after moment measurement exposes a substituted temperature class immediately — the cheapest insurance in this entire guide, given what the temperature guide says about irreversible loss.
06
Sampling & incoming inspection
- Define the plan on the PO: ANSI/ASQ Z1.4, general inspection level II as the common default, with stated AQLs — e.g., 1.0 dimensional/magnetic, 2.5 cosmetic. The supplier's outgoing inspection should run the same scheme.
- Weight by failure cost, not by ease. 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.
- Correlate fixtures before you need to. Same property, same method, and a periodic cross-check of your fixture against the retained FAI samples — fixture drift causes a meaningful share of apparent failures.
- Engineer the inspection station for the parts. Magnets chip when they snap together on steel surface plates: non-magnetic tooling, spacers, and controlled handling are part of the inspection plan — the assembly-side rules in the bonding guide's handling section apply at the bench too. Inspection-induced damage billed as supplier quality poisons the scorecard data.
07
Nonconformance, disputes & CAPA
- Quarantine and re-measure first: second operator, fixture cross-checked against FAI reference hardware. Confirmed data converts a dispute into a conversation.
- Separate nonconforming from unusable. A cosmetic chip on a hidden face may merit a documented, priced deviation; a magnetic miss on a field-critical part never does.
- Diagnose before blaming. Uniform low readings across a lot point at fixture or method; a distorted pole pattern points at thermal or field exposure somewhere in the chain — the failure signatures in the demagnetization guide assign root cause to the right party, which is occasionally your own reflow oven.
- Demand containment + root cause, not replacement. Replacement parts from the same uncorrected process are the same defect on a later date. The corrective-action walkthrough question from the qualification guide exists for this moment.
- Close the loop on the drawing. Most first-year disputes trace to a vague criterion. Every resolved dispute should tighten a line — that's the mechanism that makes year two quiet.