Magnets for R&D & Prototyping Teams

The fast loop: iterate with stock

• Sweep geometry with a size ladder, not a single part. Field-at-distance is dominated by dimensions, not grade — a grade step buys only ~3–4% (grades chart math), so order a spread of diameters and thicknesses around the estimate and let the bench find the answer. A stock-size ladder costs less than a meeting.
• Vary one thing at a time, magnetically: same grade across the size sweep, then a grade step at the winning size if needed. Mixing both variables in one sweep produces the classic uninterpretable prototype notebook.
• Diameter-to-thickness ratio is the shape lever: wide faces throw field further and fall off slower; tall parts concentrate it near the pole. If the working gap is the problem, change the ratio before the grade.
• Know the ladder's limits: stock parts are overwhelmingly axially magnetized discs, blocks, and rings. Diametric, true radial, and multipole patterns are where the catalog thins and section 06 begins — patterns defined in the directions guide.

Picking the first magnet

Bench measurement without a magnetics lab

• A handheld gaussmeter is a comparison tool, not an acceptance tool. Readings swing with probe position, angle, and technique — fine for A-vs-B on your own bench with a jig, misleading as an absolute number. Build a simple probe-positioning jig on day one; every reading without one is a vibe.
• Measure field at the working point, in a fixture. A 3D-printed cradle holding probe and magnet at the real gap turns the gaussmeter into a repeatable instrument — and that fixtured field-at-point measurement is exactly the language the eventual production spec will use, per how magnets are tested.
• Cheap force measurement is legitimate: a luggage scale or force gauge pulling a magnet off the real target material, through the real gap, averaged over pulls — crude, honest, and directly relevant in a way catalog numbers aren't.
• Map, don't spot-check, for sensing work: a Hall breakout on an XY grid (even hand-positioned) reveals the field shape that a single reading hides — and field shape is what sensor architectures live on, per the sensor guide.
• Log the part identity with every measurement: size, grade, lot/order, temperature. The prototype notebook that says “the magnet” is unrepeatable by definition.

Prototyping the attachment, not just the magnet

• Hot glue and CA are rig materials, not results. They answer “does the concept work” and nothing else — strength, temperature behavior, and lifetime all change with a real adhesive system. Before the design review, rebuild the joint per the bonding & mounting guide: prepped surfaces, structural adhesive, controlled bondline.
• Prototype the release and capture early: if the product must let go (grippers, closures) or must never let go (anything overhead or near people), those mechanisms are the hard part — and they change the magnet choice. Mechanical capture is a design feature, not a production detail.
• Never machine the magnet on the bench. Drilling, grinding, or sanding sintered NdFeB cracks it, breaches the plating, and makes flammable dust — the universal shop rule. Need a hole? Buy countersunk or pot-mounted formats. Need a shape? Section 06.
• Respect the bench hazards at prototype scale: pinch injuries from palm-size parts, dead hard drives and stripe cards, and the mystery of the demagnetized part that spent an afternoon next to the soldering station — small-lab versions of real rules.

Results that survive production

The gap between a working prototype and a manufacturable product is where magnet programs stumble — the R&D-side preview of the traps:

• Your samples are from the middle of the distribution. Production delivers a ±3–5% part-to-part and lot-to-lot window on field — normal and conforming. Before declaring victory, ask: does the design still work at −5%? Architectures that measure field direction or a zero-crossing shrug at the spread; absolute-threshold designs inherit it as error.
• Add tolerance sensitivity to the bench plan: deliberately shim the gap, offset the alignment, and swap magnets between units. A design that only works with that magnet in that slot is a demo, not a product.
• One hot-soak test early beats a thermal surprise late: an oven hour at worst-case service temperature followed by a re-measure exposes both reversible drift and any irreversible knock-down — the two loss types in the temperature guide — while the design can still absorb the answer.
• Write requirements in transferable language: the deliverable of a magnet prototype isn't a part number, it's a spec — field at a defined point with a tolerance, temperature range, envelope, and attachment interface. That's the form a supplier can quote and a factory can hold, per the tolerances guide's enforceability logic.

When the prototype needs a custom

• Exhaust the near-custom options first: a stack of stock discs approximates a taller part; a machinist's fixture repositions a stock magnet; bonded material handles odd shapes at lower field. Each keeps the loop measured in days.
• Prototype customs are usually machined from blanks, not pressed with production tooling — fast and legitimate for geometry and field questions, but the surface, edges, and coating differ from production parts. Label the results accordingly.
• Patterns are the true custom frontier: clocked diametric, true radial, and multipole magnetization need fixtures — this is where a supplier's prototype service replaces the bench, and where budget-and-weeks conversations start; the cost structure is the MOQ & tooling guide.
• Bring the supplier in one iteration before you think you need to. An application engineer looking at the sweep data will often relocate the answer — a different geometry, a stock part you missed, a pattern that simplifies the sensing — and the conversation is free, unlike the tooling it can save. That's also the moment the RFQ guide's checklist starts paying.

The handoff package

When the project graduates — to your own NPI process or a contract manufacturer — the magnet handoff is one page plus data:

Related: neodymium grades chart · how magnets are tested · MOQs, lead times & tooling