How core loss is measured: the Epstein frame, single-sheet tester and IEC 60404 standards

Every W/kg figure on an electrical steel datasheet comes from one specific test: strips cut to 30 mm width, stacked into a square, magnetised with perfectly sinusoidal flux at about 23 °C. Knowing that test tells you exactly what the number means, and what it leaves out.

In why electrical steel has core loss we split loss into hysteresis, eddy-current and excess components. All three depend on the flux waveform, the direction of magnetisation, mechanical stress and temperature. Measure the same sheet two different ways and you get two different answers. So the industry fixes the method in a standard, and every datasheet number is only meaningful relative to that method.

The Epstein frame: the reference test

The standard method for measuring core loss is the 25 cm Epstein frame, described in IEC 60404-2. It turns a set of strips into a closed magnetic circuit that behaves, as nearly as possible, like a transformer core with no air gap.

The key parameters, from IEC 60404-2:

1. Strips: 30 mm ± 0.2 mm wide, 280–320 mm long, all equal in length to within ±0.5 mm. The number of strips is a multiple of four, with at least 240 g of active mass for 280 mm strips.
2. Direction: for non-oriented steel, half the strips are cut along the rolling direction and half across it. The result is an average of the two directions.
3. Windings: 700 turns (1,000 is also allowed) on both primary and secondary, spread over the four limbs.
4. Path length: the effective magnetic path length is conventionally assumed to be 0.94 m. Flux crowds and turns at the corners, so the true length is not exactly known. Everyone agrees to use 0.94 m so that results are comparable.
5. Conditions: ambient temperature 23 ± 5 °C; frequencies up to 400 Hz. IEC 60404-10 extends the method to medium frequencies, up to 10 kHz.

How the measurement works

The frame is a transformer whose core is the specimen. The primary current sets the magnetic field H; the voltage induced in the secondary tells you how fast the flux is changing, and therefore the polarisation J; the product of the two is the power flowing into the core.

Step 1: cross-section from mass

Measuring the thickness of dozens of thin strips is error-prone, so the standard derives the cross-section from the specimen mass instead:

This is why every grade has a conventional density in its specification: 7,650 kg/m³ for a 0.35 mm, 2.70 W/kg grade. It is not a precise measured value; it is the number the test agrees to use.

Step 2: polarisation from the secondary voltage

By Faraday’s law, the average rectified secondary voltage is proportional to the peak polarisation:

This only holds if the flux is sinusoidal, so the standard requires the form factor (RMS divided by average rectified value) of the secondary voltage to be 1.111 ± 1%, the value for a pure sine wave. Steel near saturation distorts the magnetising current, so modern test benches use feedback control on the amplifier to keep the flux waveform sinusoidal.

Step 3: field strength from the primary current

The pairs of peak J and peak H, measured at 50 Hz, are the magnetisation curve printed on the datasheet. That is why the datasheet curve joins only the tips of the hysteresis loops.

Step 4: specific total loss

The wattmeter multiplies primary current by secondary voltage and averages over a cycle. Using the secondary voltage rather than the primary excludes the copper loss in the primary winding. After subtracting the small power drawn by the instruments, the loss is divided by the active mass, the part of the specimen inside the mean path:

Worked example: M270-35A at 1.5 T, 50 Hz

Take sixteen strips of M270-35A, 280 mm × 30 mm × 0.35 mm. At 7,650 kg/m³ each weighs 22.5 g, so the specimen is 360 g. (Twelve strips would not be enough: their active mass would be only 227 g, below the 240 g minimum.)

The cross-section checks out: four strips per limb × 30 mm × 0.35 mm = 42.0 mm². The 1,700 A/m comes from SURA’s M270-35A magnetisation table, and 2.47 W/kg is its datasheet loss at 1.5 T and 50 Hz. The whole test measures less than one watt.

The single-sheet tester

The second standard method, IEC 60404-3, clamps a single sheet at least 500 mm long between two U-shaped yokes made of grain-oriented steel. The yokes close the magnetic circuit, so no corners are formed in the specimen itself.

The two methods do not give identical numbers for the same steel, which is why a datasheet should state which one it used. SURA, for example, states that its losses are measured on a 25 cm Epstein frame to IEC 60404-2, and IS 648 specifies the Epstein frame for acceptance testing.

The standards map

Two kinds of standard work together. Measurement standards define how to test. Product specifications define the grades, and the maximum loss each grade must meet using that test.

Reading the grade specification

The grade name encodes the test result. Under EN 10106 and IEC 60404-8-4, M270-35A means a maximum specific total loss of 2.70 W/kg at 1.5 T and 50 Hz, a nominal thickness of 0.35 mm, and A for non-oriented, fully processed steel. India’s IS 648 reorders the same information: 35C270 is 0.35 mm, cold rolled, 2.70 W/kg.

The specification then sets the limits every coil must meet. For this grade, IS 648:2022 requires:

Three things to take from this table:

1. Guaranteed is not typical. The grade name is a ceiling. SURA’s typical value sits about 9% below it. Design with typical values, but know that any delivered coil may be anywhere up to the limit.
2. J, not B. Standards specify magnetic polarisation J = B − μ₀H, the part of the flux density that comes from the material. At 10,000 A/m the difference is μ₀H = 0.013 T, negligible below saturation but worth knowing when comparing curves.
3. Only one point is guaranteed. The specification fixes loss at 1.5 T and 50/60 Hz. Everything else on a manufacturer’s datasheet, including the high-frequency loss tables, is typical data, provided for information.

Anisotropy, in IS 648, is (P_⊥ − P_∥) / (P_⊥ + P_∥) × 100, where the two losses are measured across and along the rolling direction. “Non-oriented” does not mean identical in every direction: up to ±17% is allowed for this grade.

What the test leaves out

The Epstein test is designed to rank materials fairly, not to predict the loss in your motor. Each condition that makes it repeatable is one your motor does not share:

That last row is easy to check. Here is every point in SURA’s M270-35A loss table, with three example motors:

At 50 Hz the table runs to 1.8 T. At 400 Hz it stops at 1.5 T, and at 1 kHz at 1.1 T. Suppliers rarely test there: at high frequency and high polarisation the specimen absorbs tens of watts per kilogram and heats up during the measurement. High-speed motors operate exactly in that missing corner, so their core loss is always an extrapolation. How far a loss model fitted to the table can safely be stretched is the subject of core loss models compared.

What this means for motor design

1. Check which test produced the numbers. Epstein and single-sheet results differ; so do 50 Hz and 60 Hz ratings. Compare grades from the same method.
2. Use the guaranteed limit for worst-case losses and the typical value for expected performance.
3. Plot your operating points on the loss table. If they sit outside it, ask the supplier for data at your frequency and flux density, or treat the loss as an estimate with a wide margin.
4. Add a manufacturing allowance. Punching, stacking and fitting add loss the datasheet cannot know about. Narrow teeth and small motors are affected most.
5. Measure your own core when it matters. A ring specimen (IEC 60404-6) punched with your production tool, or a test on the finished stator, captures cutting damage that datasheet strips do not.

Summary

Datasheet core loss comes from a standard test, most often the 25 cm Epstein frame of IEC 60404-2: 30 mm strips, half along and half across the rolling direction, sinusoidal flux with a form factor of 1.111, a conventional path length of 0.94 m, at 23 °C. The secondary voltage gives J, the primary current gives H, and a wattmeter gives the loss per kilogram of active mass. Product standards such as EN 10106, IEC 60404-8-4 and IS 648 turn one of those measurements, the loss at 1.5 T and 50 Hz, into the grade name and its guaranteed limit. Everything else on the datasheet is typical data measured under ideal conditions, so the loss in a real motor, with its punched edges, stresses, rotating flux and PWM harmonics, will be higher.