Transformer Guide
Transformer Core Loss: Waveform, Frequency, Flux Density, and Thermal Design
Turn a magnetic operating point into a defensible loss and temperature estimate without treating a fitted material equation as a universal transformer model.
19 min · Advanced · Updated October 15, 2026
Core Loss Starts With the Excitation Waveform
The Transformer Flux Density Calculator establishes a sinusoidal RMS magnetic operating point. For a switched waveform, derive peak-to-peak flux excursion from volt-seconds instead. Frequency labels alone are insufficient: duty cycle, dead time, reset interval, imbalance, and DC bias can change the actual excursion.
Formula reference
Magnetic operating point
Sine wave: Bmax = Vrms/(4.44 f N Ae)General winding law: v(t) = N Ae dB/dtFlux change: ΔB = (1/(N Ae)) ∫v(t)dtBalanced square-wave first pass: ΔB = V ton/(N Ae)Variable definitions
- Vrms and Bmax use the stated sinusoidal convention
- N
- active winding turns
- Ae
- effective magnetic area
- The integration interval and voltage polarity must match the real waveform
Separate Magnetic Loss Mechanisms
| Mechanism | Primary dependence | Engineering implication |
|---|---|---|
| Hysteresis | Material loop area, frequency, flux excursion, temperature | Material grade and operating point matter; a universal coefficient is not valid. |
| Eddy current | Conductivity, geometry, frequency, flux waveform | Laminations or high-resistivity ferrites reduce circulating-current loss. |
| Residual/excess loss | Domain dynamics and empirical material behavior | Manufacturer characterization is usually more reliable than mechanism-by-mechanism reconstruction. |
| Gap/fringing effects | Gap geometry, local field, nearby conductor position | Can add localized winding and core heating not represented by bulk loss density. |
| DC bias | Shifted B-H operating point and permeability | Loss and inductance can differ from unbiased small-signal data. |
Use Steinmetz Models Only Inside Their Valid Range
Formula reference
Empirical core-loss model
Pv = k f^α Bpk^βPcore = Pv Veor Pcore = Pmass mcoreVariable definitions
- Pv
- volumetric loss density using the coefficient's stated units
- k, α, and β are material-fit coefficients
- Ve
- effective magnetic volume
- Pmass
- mass-specific loss from a material curve
- Bpk may need replacement by a stated flux excursion convention
The equation is useful for interpolation, sensitivity, and design iteration only when the coefficient set matches material, waveform, temperature, frequency, and flux convention. Do not mix coefficients expressed in mW/cm³ with SI geometry without conversion. Do not extrapolate through saturation or far beyond the fitted frequency range.
Non-sinusoidal waveforms may require generalized or improved Steinmetz methods, loss maps, or direct manufacturer tools. Those methods are not interchangeable, especially when minor loops, DC bias, asymmetric duty cycle, or relaxation intervals matter.
Frequency and Flux Density Trade Against Each Other
Increasing frequency can reduce turns for a fixed volt-second limit, but empirical loss rises approximately with frequency to α and flux density to β. Reducing B may offset some frequency increase, yet more turns increase winding length, DCR, leakage, capacitance, and window use. Optimize total transformer loss and manufacturability, not core loss in isolation.
Frequency doubled
At unchanged Bpk, an empirical model predicts a core-loss-density multiplier of 2^α. Recalculate AC copper and switching losses too.
Flux reduced 20%
At unchanged frequency, the empirical multiplier is 0.8^β. Use the fitted β only within characterized data.
Turns increased
Flux falls inversely with turns for fixed voltage, frequency, area, and waveform, while winding resistance and fill usually rise.
Minimum frequency event
A brownout, startup, or control fault can increase volt-seconds and flux even when nominal operation is acceptable.
Build a Complete Real-Loss Budget
Formula reference
Transformer loss and efficiency
Pcu,dc = Σ Irms² RhotPtotal = Pcore + Pcu,dc + Pcu,ac + Pstray + Pdielectricη = Pout/(Pout + Ptotal)First-pass ΔT = Ptotal RθVariable definitions
- Use true RMS winding currents
- Rhot must correspond to winding temperature
- Pcu,ac includes skin, proximity, and leakage-field effects
- Rθ must represent the tested assembly and cooling condition
Enter a material-based core-loss estimate into the Transformer Efficiency & Loss Calculator, then keep it separate from copper and other loss terms. The Winding Resistance Calculator supports the DC copper branch; it does not predict AC resistance.
Reactive magnetizing current contributes winding RMS current and source VA, but V × Im is not automatically real core loss. Measure or model phase and waveform before converting no-load apparent power into watts.
Thermal Iteration and Hotspots
A single case-to-ambient resistance is only a first pass. Core and winding losses occur in different locations, transfer heat through bobbins, insulation, impregnation, clamps, PCB copper, airflow, and enclosure surfaces, and change with temperature. Core loss may rise or fall with temperature depending on material and operating point; copper resistance generally rises.
Iterate loss at the estimated hot condition, then check core hotspot, inner winding, insulation system, nearby components, ambient range, airflow tolerance, and abnormal modes. Thermal equilibrium can take much longer than electrical stabilization.
Worked Engineering Checks
| Check | Calculation | Interpretation |
|---|---|---|
| Loss-density conversion | 250 mW/cm³ × 8 cm³ = 2.0 W | Both source data and effective volume use compatible units. |
| Frequency sensitivity | If α = 1.5, doubling f at fixed B gives 2^1.5 = 2.83× Pv | A sensitivity result, not permission to extrapolate coefficients. |
| Flux sensitivity | If β = 2.4, reducing B to 80% gives 0.8^2.4 = 0.585× Pv | Core loss drops, but extra turns may raise copper loss. |
| Total loss | 1.8 W core + 1.2 W primary copper + 0.8 W secondary copper + 0.4 W other = 4.2 W | All terms must describe the same operating point. |
| First-pass temperature | 4.2 W × 12 °C/W = 50.4 °C rise | Verify distributed hotspots and the actual assembly experimentally. |
| Round-trip power | At 50 W output and 4.2 W loss, η = 50/54.2 = 92.25% | Input/output power measurement should correlate with the loss breakdown. |
Design and Verification Workflow
| Step | Action | Required evidence |
|---|---|---|
| 1 | Define excitation | Record winding voltage waveform, frequency, duty cycle, DC bias, operating modes, and worst credible volt-second condition. |
| 2 | Calculate flux | Use the correct Faraday or volt-second relation and actual integer turns, effective core area, and waveform convention. |
| 3 | Select material data | Use loss curves or coefficients for the exact material and a relevant temperature, frequency, and flux range. |
| 4 | Estimate core loss | Interpolate inside characterized data; avoid unsupported extrapolation and keep volume, mass, and density units explicit. |
| 5 | Add winding loss | Use RMS current, hot DCR, and frequency-dependent AC resistance for every winding. |
| 6 | Close the thermal loop | Convert total real loss to temperature, update temperature-sensitive properties, and iterate. |
| 7 | Check transients | Review startup, minimum frequency, saturation, burst operation, imbalance, and protection response separately. |
| 8 | Verify hardware | Measure waveforms, input/output power, core temperature, winding temperature, and operating margin in the final assembly. |
Common Mistakes
- Using a sinusoidal 4.44 relation for an arbitrary switched waveform.
- Treating one Steinmetz coefficient set as universal.
- Mixing peak, peak-to-peak, and RMS magnetic conventions.
- Extrapolating loss curves beyond their characterized range.
- Ignoring core-loss temperature dependence.
- Counting reactive magnetizing VA as real loss.
- Ignoring AC winding and gap-fringing loss.
- Optimizing core loss while making copper loss or window fill unacceptable.
- Using nominal frequency while missing startup or minimum-frequency volt-seconds.
- Assuming average surface temperature equals the winding or core hotspot.
- Comparing electrical loss at inconsistent reference planes.
- Skipping waveform and temperature validation in the final assembly.
Scope Boundaries
This guide owns material-data-based core-loss estimation, waveform validity, total-loss integration, thermal iteration, and validation. The Transformer Magnetic Design Guide owns volts per turn, flux density, magnetizing current, and window utilization. The Transformer Winding Design Guide owns conductor selection, fill, insulation, and manufacturability. Critical designs require manufacturer material data, appropriate field/circuit models, safety review, prototypes, and measurement.
Support reference
FAQ
What causes transformer core loss?
Core loss combines hysteresis, eddy-current, and other frequency-dependent magnetic loss mechanisms. Their relative importance depends on material, frequency, flux excursion, waveform, temperature, and construction.
Is transformer core loss the same as copper loss?
No. Copper loss is primarily winding current squared times hot resistance, with additional AC winding effects. Core loss occurs in the magnetic material and is driven mainly by excitation conditions.
How is transformer core loss estimated?
Use manufacturer loss curves or validated material coefficients at the actual frequency, peak flux density, waveform, and temperature. Multiply volumetric or mass-specific loss by the effective core volume or mass using consistent units.
Can I use one Steinmetz equation for every waveform?
No. A basic Steinmetz fit is normally valid only over its stated material, frequency, flux-density, temperature, and waveform range. Non-sinusoidal excitation needs an appropriate waveform method or manufacturer data.
Why does a small flux-density increase raise loss so much?
Empirical core-loss models usually contain a flux-density exponent greater than one. Loss can therefore rise much faster than flux density, especially near the edge of the fitted data range.
Does higher switching frequency always reduce transformer size?
Higher frequency can reduce required turns or core area for a given volt-second condition, but it also increases core loss, AC winding loss, switching loss, EMI, and parasitic sensitivity.
How do I combine transformer losses?
Add core loss, primary and secondary copper loss, AC winding loss, and other measured or modeled losses at the same operating point. Do not count reactive magnetizing VA as real loss.
How is transformer temperature rise estimated?
A first pass is temperature rise equals total real loss times an applicable thermal resistance. Real transformer thermal paths are distributed and temperature dependent, so verify winding and core temperatures in the final assembly.
Does the flux-density calculator predict core loss?
No. It establishes the magnetic operating point under its stated waveform convention. Core-loss estimation additionally requires material-specific loss data and thermal validation.
How should core loss be measured?
Use a method appropriate to frequency and power level, subtract fixture and winding effects consistently, verify voltage-current phase accuracy, and correlate electrical loss with stabilized core and winding temperatures.
Summary
Start with the real excitation waveform, calculate flux using the correct convention, estimate core loss only from applicable material data, and combine it with hot copper and other real losses. Iterate temperature-sensitive properties and verify electrical loss and thermal hotspots in the completed transformer and enclosure.