XIANYUAlloy · Changzhou
Technical note · 2026-11-02

No constantan grade holds resistance better than the others

There is no constantan grade that is more stable than the rest, because 6J40 and CuNi44 are one alloy written in two standards systems. What actually varies between suppliers is the coefficient band you buy and the temperature window it is guaranteed over. Across 20 to 100 °C the standard band and the selected band differ by roughly 0.32 % against 0.08 % of resistance change, and on a DC shunt a 40 µV/K thermal EMF against copper usually matters more than either.

The two names are one alloy, not two grades

Constantan is copper-nickel with a little manganese, and it appears in two standards systems under two designations. GB/T 6145 lists it as 6J40 with 39 to 41 % nickel. The European and American systems list CuNi44, material 2.0842 and UNS C72150, with 43 to 45 % nickel. The nickel range in the two documents is not the same, which is the first reason two certificates for the same material never agree line for line.

Constantan is specified at about 0.49 µΩ·m. That is the number that decides whether the design is practical: high enough that a compact element reaches the target value, low enough that a one milliohm part is still a manageable piece of metal rather than a foil.

So when a buyer asks which constantan grade is most stable, the honest reading is that the answer is not in the grade name at all. Two suppliers shipping the same designation can hand over parts at opposite ends of the permitted band, and both are conforming.

Stability is a band you buy, not a grade you name

ISOTAN, the CuNi44 grade Isabellenhütte supplies, is sold to a coefficient of −80 to +40 ppm/K between 20 and 105 °C, which is the range DIN 17471 permits, with ±10 ppm/K available as a special selection. Those are not two alloys. They are two sorting decisions taken on the same melt practice.

The difference over an 80 K span is easy to size. A part sitting on the standard band can move 3200 ppm, which is 0.32 % of its value. A selected part moves 800 ppm, or 0.08 %. If your bridge balances to 0.02 %, that gap is the whole error budget.

There is a third number that explains why constantan has the reputation it does. On an ideal curve, constantan's coefficient between 20 and 60 °C is close to 2 ppm/K, which is flatter than MANGANIN's ±10 ppm/K. The alloy is genuinely capable of it. Whether you receive that behaviour is a purchasing question, not a metallurgical one.

Why the flatter alloy still loses on a DC shunt

Constantan's thermal EMF against copper is about −40 µV/K. Manganin's is −1.00 µV/K. That is the whole argument, and Isabellenhütte states it without hedging: because of the high EMF against copper, constantan is not suitable for electrical precision resistors, and the datasheet points to MANGANIN, NOVENTIN or ZERANIN instead.

Put a number on it. A 10 K gradient between the two ends of a constantan element generates 400 µV. On a 60 mV shunt that is 0.67 % of reading, roughly eight times the coefficient error the alloy was chosen to avoid. It also does not average out with repeated readings, because a thermal gradient is a DC error and looks exactly like current.

This is the part that surprises people who have read that constantan is the more stable of the two resistance alloys. It is more stable in coefficient, over a wider temperature range, and still the worse choice for a low-level DC measurement. Two different questions got the same answer for fifty years because they were never separated.

So when is constantan the right call?

Three cases hold up. AC measurement, where the thermal EMF appears as a DC offset that the instrument can reject. Elements that run hot: constantan is rated at least to 400 °C, Isabellenhütte rates ISOTAN to +600 °C for resistance and heating duty, and Chinese 6J40 sheets stop at 500 °C, so the spread in those three figures is itself a warning against treating any single ceiling as final. And strain gauges, where the gauge factor and the insensitivity of the alloy are the entire point.

A test that settles the question in an afternoon

Clamp the element at the current it will carry, let it stabilise, and measure its resistance at two temperatures you have actually recorded rather than assumed. Compare that slope with the band on the certificate. If the measured slope is inside the band and the residual error is still too large, the coefficient was never the problem and a tighter grade will not fix it. If the slope is outside the band, you have a supplier conversation, not a design problem.

The rule of thumb worth keeping

Below about 60 °C, manganin wins on both counts that matter for DC: a ±10 ppm/K band and 1 µV/K against copper. Between 60 and 140 °C, constantan's wider window starts to pay and its EMF starts to hurt. Above 140 °C manganin is out of the question, and the EMF problem has to be solved in the circuit rather than in the purchase order.

Questions we get asked
Is 6J40 the same as CuNi44?
Compositionally close, not identical on paper. 6J40 is written with 39 to 41 % nickel and CuNi44 with 43 to 45 %, so the two certificates will not match digit for digit. Both are sold as constantan and both behave as constantan; the difference is a standards question, and it is the reason to buy against a stated designation plus a measured certificate.
Which constantan is best between 20 and 100 °C?
None of them is best by name. Ask for a measured coefficient band of ±10 ppm/K or tighter over the actual window, and get the certificate to state the range it was measured over. A grade name without a band and a range tells you nothing you can design against.
Why do suppliers still recommend constantan for precision work?
Usually because the application is AC, or because the element runs hot, or because the recommendation is about a strain gauge rather than a shunt. Constantan's coefficient is genuinely low. Its thermal EMF against copper is the reason it loses to manganin in DC current sensing.
Can I use constantan for a 100 A shunt?
It will work and it will drift with any temperature gradient along the element. At 100 A the element is hot, so the gradient is real, and 40 µV/K turns a 10 K difference into 400 µV. If the sense amplifier sees 60 mV at full scale, that is 0.67 % before any other error.
Citable facts
Sources

Send the specification — diameter, grade, temper, quantity.

We come back with a confirmed lead time and the documentation we can issue for the shipment. 100 g sample at cost; orders from 1 kg.

Request a quotation Order a 100 g sample