A ferrite datasheet usually presents core loss under sinusoidal excitation at selected frequencies, flux densities, and temperatures. A switched-mode high frequency transformer, however, is commonly driven by rectangular, quasi-square, phase-shifted, or pulse-width-modulated voltage. Applying a sine-wave chart directly to that waveform can understate loss, overstate loss, or miss a DC-flux imbalance that is more serious than the nominal AC calculation.
The starting point is Faraday’s law. The change in flux density is determined by the voltage-time area applied to each winding turn and the effective core area. Frequency matters because it changes the available time per cycle, but frequency alone does not determine flux swing. A converter with the same frequency and bus voltage can operate at very different flux density when duty cycle, topology, or turns count changes.
Calculate flux from the actual winding voltage
Use the voltage measured or simulated directly across the transformer winding, not only the DC bus value. Semiconductor drops, resonant transitions, dead time, clamp action, leakage spikes, and modulation alter the waveform. Integrate each voltage interval over time and divide by turns and effective core area to obtain the flux-density trajectory.
In a symmetrical full bridge, positive and negative volt-seconds should balance. In a push-pull or half-bridge converter, timing mismatch, unequal device drops, gate-drive differences, current-sense offsets, or capacitor imbalance can create flux walking. A small per-cycle error can accumulate until the magnetizing current or core loss rises sharply. Peak-current protection does not always identify the underlying asymmetry before thermal stress develops.
Why a sine-wave Steinmetz estimate has limits
Traditional Steinmetz equations fit empirical loss data for a material over a defined frequency, flux, and temperature range. They are useful, but their coefficients do not automatically remain valid for arbitrary pulse shapes. Generalized methods attempt to account for instantaneous rate of flux change and waveform segments. Even then, accuracy depends on the material data, minor-loop behavior, DC bias, and whether the operating point stays within the fitted range.
For engineering decisions, BaoHui Tech treats analytical loss as a screening tool. Candidate materials and core sizes are compared with a consistent method, then the assembled transformer is tested at representative waveforms and temperatures. This matters because winding loss heats the core, core loss heats the winding, and the final equilibrium is electrothermal rather than magnetic only.
Duty cycle, dead time, and minor loops
A waveform can have the intended peak-to-peak flux swing but still produce unexpected loss. Extended zero-voltage intervals, resonant reversals, and phase-shift modulation can create minor hysteresis loops. At light load, burst mode groups switching cycles into packets and changes both average temperature and local peak flux. In an LLC converter, frequency varies with operating point, so the relationship among magnetizing current, resonant current, applied voltage, and core loss also changes.
Dead time should be included in the winding-voltage waveform. Depending on current direction and device capacitance, the transformer may see a resonant voltage transition rather than an ideal flat zero interval. This transition contributes to flux trajectory and can differ between load conditions.
Temperature is not a simple derating factor
Ferrite loss varies nonlinearly with temperature. Some materials show a loss minimum within a particular temperature range and higher loss on either side. Saturation flux density generally falls as temperature rises. A design checked only at room temperature can therefore miss both hot loss and reduced saturation margin.
The thermal calculation should include core geometry, exposed surface, bobbin, winding, gap or mating surfaces, mounting hardware, airflow, nearby heat sources, and enclosure temperature. Core surface temperature is not necessarily the core hot spot, and a thermocouple placed between core halves can disturb the magnetic joint.
Practical validation method
- Capture the winding voltage and current at minimum line, maximum line, full load, light load, startup, and transient conditions.
- Integrate winding voltage to verify flux swing and positive/negative volt-second balance.
- Measure total transformer loss by a suitable electrical or calorimetric method, accounting for instrument phase error at high frequency.
- Estimate or separately characterize copper loss so core loss is not inferred from temperature alone.
- Repeat at stabilized temperatures and across production-relevant material tolerance.
- Inspect light-load burst and fault-recovery waveforms, where control behavior may differ most from nominal operation.
Information to give a transformer manufacturer
- Topology, bus-voltage range, switching-frequency range, and modulation method.
- Actual or simulated primary voltage waveform, including dead time and transients.
- Maximum duty cycle, expected imbalance, startup behavior, and fault duration.
- Power, RMS and peak currents, cooling conditions, and temperature limits.
- Efficiency target, available core volume, insulation requirements, and lifetime profile.
Frequently asked questions
Does doubling frequency always double transformer core loss?
No. Frequency, flux swing, waveform, material, temperature, and bias interact. If turns and voltage remain fixed, increasing frequency usually reduces flux swing, so the final loss change must be calculated from the new operating point.
Can temperature rise identify core loss by itself?
No. Copper loss, core loss, terminals, and nearby components share heat paths. Temperature is an essential validation result, but separating loss sources requires additional measurement or modeling.
How does BaoHui Tech select a ferrite material?
BaoHui Tech compares material data at the application’s frequency, flux, waveform, and temperature, checks saturation and dimensional constraints, and validates the finished high frequency transformer under representative converter conditions.