Datasheet Curves Are Not Your Operating Point
Core loss figures on a datasheet are measured on a small toroid, at a specific flux density and frequency, with sinusoidal excitation and a defined temperature. Your inductor runs at a different flux swing, a different frequency, a different temperature, and often a non-sinusoidal waveform. Treating the datasheet number as the design value is one of the most common sources of thermal surprises in switching converters.
This note covers the four practical methods for getting usable core loss data, and the correction factors that must be applied to any of them.
Four Ways to Get Real Loss Data
| Method | How it works | Best use | Limitation |
|---|---|---|---|
| Datasheet interpolation | Fit the vendor’s loss curves at your flux and frequency | Early sizing, first-pass comparison | Assumes sinusoidal excitation and the vendor’s test core |
| Steinmetz equation with waveform correction | Pv = k · fα · Bβ, refined by iGSE for non-sinusoidal waveforms | Converter design with triangular flux | Coefficients are material- and condition-specific; needs correct α and β |
| B-H loop measurement | Drive a sample winding, integrate the voltage to get flux, measure current for H | Verifying a material or a supplier’s claim | Requires care with phase error and winding capacitance |
| Calorimetric measurement | Measure the temperature rise of a thermally isolated sample in steady state | Validating a finished part in its real waveform | Slow; needs a well-insulated enclosure and stable ambient |
For most design work, the practical path is: datasheet for first sizing, iGSE correction for the actual triangular waveform, then a calorimetric check on the first article. That catches both the material assumption and the construction assumption in one measurement.
The Corrections That Change the Answer
Waveform. A triangular flux wave at the same peak and frequency has a different loss from a sinusoid, typically 10–30 % higher for ferrite depending on duty cycle. The improved Generalized Steinmetz Equation accounts for this by integrating over the flux derivative rather than using peak values.
Temperature. Ferrite loss usually has a minimum near 80–100 °C and rises on both sides; powder cores often rise monotonically with temperature. A design evaluated at 25 °C can be 40 % off at 100 °C — in either direction.
DC bias. A biased core has reduced incremental permeability, and loss at a given AC flux can change substantially. Powder core datasheets provide bias curves; ferrite datasheets usually do not, so measurement is the only reliable route.
Gap and fringing. Loss near a gap can be several times the bulk value because fringing flux enters the core laterally. Distributed gaps avoid this; a single large gap does not.
Setting Up a Usable Measurement
- Wind a test winding with enough turns that the exciting current is small relative to the current you can measure accurately.
- Drive with the actual waveform (a bridge or a buck stage), not a sine source, if the design is a converter.
- Measure the voltage across the winding with a compensated probe and integrate it; phase error between voltage and current is the dominant error source.
- Hold the sample at a stable temperature, and record loss at two or three temperatures, not one.
- Report the result as loss per unit volume at a stated flux, frequency, waveform, and temperature — never as a single unqualified number.
Related reading
- Core Loss vs Copper Loss Optimization
- How to Read a Magnetic Core Datasheet
- How to Test Magnetic Components
- Air Gap Design
Our Magnetic Core data sheets state the test waveform, temperature, and core size behind every loss curve, and we can supply measured loss on your actual waveform for a first article. Send the flux swing, frequency, duty cycle, and temperature to sales@sirui-tech.com.