Why the cup does so much
A bare magnet in air sends most of its flux on a long journey through nothing. Air has a permeability about the same as vacuum, so that return path has enormous reluctance, and the magnet is forced to work hard against its own geometry just to push flux around it. Wrap the back and sides in steel and you replace most of that path with a material some hundreds to thousands of times more permeable.
The gain in holding force is typically a factor of two to three over the same magnet bare, and on short, wide magnets — the ones that self-demagnetize hardest — it can be more. That is a large return for a turned or stamped low-carbon steel part.
When a holding design comes up short, the reflex is to specify a larger or higher-grade magnet. Adding or thickening the flux return is nearly always cheaper per newton, and it improves the thermal margin as a side effect by raising the working point. Exhaust the steel before you buy grade.
The common forms
Flux return shows up in several standard shapes. They differ in how completely they close the circuit and therefore in how much they gain.
| Form | Steel covers | Typical gain | Suits |
|---|---|---|---|
| Backing plate | Rear face only | 1.3–1.8× | Flat mounting, thin envelopes, where sides must stay clear |
| Pot magnet / cup | Rear face and full circumference | 2–3.5× | The default holding assembly; threaded, countersunk or plain |
| Channel assembly | Rear and two long sides | 2–3× | Long rectangular holding, mounting rails, door latches |
| Shallow pot | Rear and a short skirt | 1.6–2.2× | Thin assemblies where a full cup will not fit |
| Two-pole face plate | Rear plus a central or outer pole extension | Varies widely | Bringing both poles to one working face, thin-gap holding |
| Yoke / horseshoe | A full return arm to a second pole face | Large | Where the target spans two poles — separators, clamps |
The two-pole arrangement is worth understanding, because it changes the character of the assembly rather than just its magnitude. Bringing both poles to the same working face gives a very short flux path through the target, which makes the assembly extremely strong in direct contact and very weak across even a small gap. That is ideal for a latch on a clean flat surface and wrong for anything that must reach.
Sizing the steel so it does not saturate
This is the part that gets left to whoever draws the cup, and it is where cup assemblies quietly underperform. Steel carries flux beautifully right up until it saturates, at which point its permeability collapses toward that of air and the return path stops working. The wall has to be thick enough to carry the flux the magnet produces.
Asteel ≥ Φ / Bsat,design
Bm = flux density at the magnet working point, T — not Br
Am = magnet pole area, m²
B_sat,design = 1.2–1.5 T for low-carbon steel, leaving margin below true saturation
Apply it at each place the flux is funnelled. For a round pot magnet the two that matter are the base and the wall.
A worked cup
A 20 mm diameter, 6 mm thick N42 disc, working point around 0.6 T in the assembled circuit.
Note that halving the wall to 1 mm does not halve the performance — it does considerably worse than that, because once the wall saturates the additional flux has nowhere useful to go and the working point falls with it. Thin walls are a false economy in a part that already costs cents.
A saturated cup looks exactly like a correct one. There is no noise, no heat and no visual difference — only a holding force perhaps 30% below what the magnet should deliver, which is easily mistaken for a weak magnet and leads to a warranty conversation with the wrong party. If an assembly underperforms, measure the bare magnet first to establish which component is actually short.
Which steel, and which ones quietly fail
The flux return has to be ferromagnetic, and a surprising number of specifications are not. This is the single most common material error in magnetic assembly design.
| Material | Magnetic | Saturation | Notes |
|---|---|---|---|
| Low-carbon steel (1006–1020) | Yes — excellent | ~2.0–2.1 T | The default. Cheap, high permeability, easy to machine and stamp. Needs plating |
| Electrical / silicon steel | Yes — excellent | ~1.9–2.0 T | Higher resistivity; worth it only where flux alternates |
| 430 stainless (ferritic) | Yes — good | ~1.5 T | The corrosion-resistant option that actually works magnetically |
| 416 stainless (martensitic) | Yes — moderate | ~1.4 T | Machinable, magnetic, moderate corrosion resistance |
| 304 / 316 stainless (austenitic) | No | — | Essentially non-magnetic. A very common and costly mistake |
| Aluminium, brass, copper | No | — | No flux return. Also conductive, which matters if anything moves |
| Nickel-iron (mu-metal) | Yes — very high μ | ~0.7–0.8 T | For shielding, not for carrying working flux — saturates early |
The 304 error is worth dwelling on because it survives review so easily. The drawing says stainless steel, the corrosion requirement is satisfied, the part looks right, and the assembly is 60% weaker than intended with no obvious cause. If corrosion resistance is genuinely required, 430 is the answer; it is ferritic, it is magnetic, and it costs little more.
Mu-metal is the mirror-image mistake. Its very high permeability makes it excellent at diverting stray field — which is what shielding is for — but its low saturation means it is the wrong choice for a return path that must carry the magnet's full working flux. Shielding material and flux-return material are different jobs.
Heavily drawn or stamped low-carbon steel has reduced permeability compared with annealed material, and a deep-drawn cup can measure noticeably below a machined one of identical geometry. Where the last few percent matter, specify an anneal after forming.
What actually sets the holding force
Once the steel is right, performance is dominated by the interface rather than by the magnetics. This ordering surprises people, and it is why measured assemblies come in below calculated ones.
The shear figure catches out latch and fixture designs regularly. A pot magnet rated at 200 N normal will slide sideways off a smooth painted surface at perhaps 30–50 N, because what resists the slide is friction — the magnetic force only supplies the normal load. If the application loads the joint in shear, either design in a mechanical location feature or size on the shear figure, and treat any published shear number as specific to the surfaces it was measured against.
The coating point deserves the same attention. A 25 µm nickel plating on both the magnet face and the target adds 50 µm of gap before anything else in the tolerance stack, and on a short assembly that is a measurable loss. Where maximum holding matters, consider whether the working face needs its full coating thickness, and check the choice against the coatings comparison so corrosion protection is not traded away carelessly.
Building the assembly
Getting the magnetics right still leaves a mechanical problem: holding a brittle magnet inside a steel cup, permanently, through temperature and vibration.
| Method | Holds against | Watch for |
|---|---|---|
| Structural adhesive | Normal and shear service loads | Surface prep, cure control, and thermal range — the bond, not the magnet, is usually the limit |
| Adhesive plus mechanical lip | Everything, including adhesive failure | Preferred wherever the assembly is safety-relevant or rotates |
| Press fit | Sustained load | Hoop stress will crack a sintered magnet; press the steel, not the magnet |
| Potting | Vibration and impact | Cure shrinkage can pre-stress a brittle part; check the resin |
| Crimp or stake | High load, high volume | Tooling cost; risk of chipping at the crimp if the magnet is unprotected |
Two practices matter more than the choice between them. Assemble unmagnetized wherever the volume justifies it, because placing an inert magnet in a steel cup is straightforward and placing a magnetized one is a fight against a part that wants to slam into the wall and chip. And never rely on adhesive alone in a rotating or safety-relevant assembly — give the magnet a lip, a pocket or a retaining feature, with the adhesive positioning rather than carrying the load. The retention design guide covers the calculation, and make versus buy covers whether the magnetizing step belongs in your building at all.
A cup assembly concentrates flux at one face, so the completed part grabs ferrous objects considerably harder than the loose magnet did. Assemblies that were manageable as components become genuinely hazardous once built, particularly in the larger sizes. Store finished assemblies separated and face-down on a non-ferrous surface, and follow the handling guidance.