Buying NdFeB Magnets from China: The Letter After the Grade Is the Whole Argument

Every magnet argument I have been pulled into starts the same way. The buyer specifies N42, the supplier quotes N42, the parts arrive, and eighteen months later a customer in Arizona or a customer in a motor application reports that half the units have lost twenty percent of their flux. Nobody lied. N42 was delivered. The problem is that N42 describes how strong the magnet is when new and says nothing about how much heat it can take before that strength starts to disappear.

Two numbers in one grade name

A sintered neodymium grade has a number and usually a letter. The number, N35 through N52 or so, is the maximum energy product — roughly, how much field you get per unit of magnet. The letter after it is the coercivity class, and coercivity is what resists demagnetisation. The industry classes run from plain N through M, H, SH, UH, EH and up to AH, and each step roughly corresponds to a minimum intrinsic coercivity of about 12, 14, 17, 20, 25, 30 and 33 to 35 kOe respectively.

The nominal maximum operating temperatures that get quoted alongside those classes are around 80 °C for plain N, 100 for M, 120 for H, 150 for SH, 180 for UH and 200 for EH, with AH reaching into the 220s. Treat those as rules of thumb, not specifications. They assume a reasonable magnet shape with a permeance coefficient around one or better and no significant opposing field. A thin disc in an open circuit is not that shape, and it will lose flux at temperatures well below the table number.

Why the suffix costs money

Getting from N to SH is not a matter of better process control. It is done with dysprosium and, in some cases, terbium — grain boundary diffusion, alloying, or both. Heavy rare earths are the expensive part of the magnet, and they are also the part with export and supply volatility that periodically makes purchasing departments nervous. When dysprosium gets tight, the price difference between an N42 and an N42SH is no longer a rounding error.

Which means you should expect a supplier to ask why you need SH. If the honest answer is "because our application actually reaches 140 °C at the magnet surface", fine. If the honest answer is "because the customer's old drawing says SH", you may be paying a substantial premium for headroom you cannot use. Conversely, if the honest answer is "we are not sure how hot it gets", that is the moment to measure it, because the answer determines a large part of your bill of materials.

What actually happens in a demagnetised magnet

Reversible loss is straightforward — the magnet recovers when it cools. Irreversible loss does not. Above a certain temperature, part of the magnet's alignment is lost permanently, and the flux never comes back even at room temperature. The temperature at which this starts depends on the coercivity class, the permeance of the circuit, and the opposing field from any nearby coil or adjacent magnet. In a motor with a strong armature field, the effective limit can be considerably lower than the open-circuit table value.

So the specification that matters is not "max operating temperature 150 °C". It is the demagnetisation curve at your working temperature and your working permeance, which a competent supplier can produce and a trading company cannot. If you cannot get that curve, you are buying on faith.

Coating, dimensions and the tolerance nobody reads

NiCuNi is the default coating and is fine for most indoor use. For salt exposure, or for medical and food-adjacent equipment, epoxy or a zinc-based system comes up, and you should ask for salt spray hours with the test condition stated. Bare NdFeB corrodes quickly, and a pinhole in a plating bath is enough to start a blister.

Dimensionally, sintered NdFeB shrinks during sintering, so magnet dimensions are always cut or ground to final size and tolerance costs money. A general tolerance of ±0.05 mm is normal for small parts; ±0.02 mm or better is grinding territory and the price reflects it. More importantly, check whether the drawing tolerance is on the magnet or on the assembly. A 0.1 mm magnet tolerance stacked into a housing with its own 0.1 mm tolerance can put your air gap out by a fifth of a millimetre, and in a small motor that shows up as a change in the current draw.

I would also insist on the magnetisation direction on the drawing. Axial, diametral, multipole — getting this wrong is the single most expensive mistake in this category, because a correctly manufactured magnet magnetised the wrong way cannot be fixed. Some factories will re-magnetise, but only if the original direction was one they can still access, and a diametrally magnetised ring is not going back.

Incoming inspection that is worth the effort

A gauss meter on the bench and a retained sample from each lot is the minimum. Measure flux in a jig that reproduces your assembly's geometry, not in free air, and record it per lot. When a field failure appears two years later, that record is the only thing that tells you whether the magnets changed or the application did. Dimensional checks on the critical face, a coating thickness spot check, and a visual for chipping at the edges — sintered parts chip easily and a chipped corner is a weak point for corrosion and for cracking.

For anything going into a regulated product, also confirm the material composition declaration and any RoHS or REACH documentation at the lot level, since the alloy composition varies by grade and by supplier's dysprosium strategy.

Dealing with the market

Ningbo and the wider Zhejiang area, Jiangxi, Shanxi and Inner Mongolia are the centres you will encounter, with a mix of listed producers, specialist sintered magnet factories and a large layer of trading intermediaries. The trading layer is not automatically a problem — some of them are excellent and handle export logistics and documentation better than a factory — but a trader cannot give you a demagnetisation curve and cannot change a coercivity class without going back to the plant. For anything thermally demanding, get to the factory.

Verify claims about any specific supplier against their own official documentation and current listings rather than a marketplace page, and ask for the grade as "N42SH, Hcj minimum 20 kOe, coating NiCuNi 15 µm" in the purchase order. Written that way, there is nothing left to interpret at the port.