Validation
How TRAFOLO results compare with measurements and with other tools. Each page gives the case, the comparison in numbers and its limits.
| Case | Compared with | Result |
|---|---|---|
| Leakage Inductance of Medium-Frequency Transformers Compared with Measurements: two 50 kW shell-type transformers, ferrite and nanocrystalline core, 1 to 10 kHz | LCR-meter measurements (Eslamian et al., IEEE Access 2021) | Leakage inductance 4.9 % (ferrite) and 6.7 % (nanocrystalline) above the measurement |
| Inductance of a Gapped E-Core Inductor Compared with Measurements: E 55/28/21 in N27 ferrite with gaps of 1.0 to 2.0 mm in all three legs | Measurements and the classic gap formula (Mühlethaler et al., ECCE Asia 2011) | Inductance 4.6 to 7.4 % above the measurement; the classic gap formula 31 to 43 % below |
| TEAM Problem 7: Eddy Currents in a Plate Compared with Measurements: aluminium plate with a hole under a coil, 50 and 200 Hz (Compumag benchmark) | Measurements of TEAM Problem 7 (Fujiwara and Nakata, COMPEL 1990), three printing errors corrected | Bz above the plate within 2.5 to 5.4 %, Jy in the plate within 1.6 to 6.5 %; two small 200 Hz curves left out as unusable |
| Modelling Litz Wire using Homogenization: PQ65 reactor with two parallel litz wires, DC to 100 kHz | AC resistance measured by Otomo et al. | Within 12 % |
| Planar transformer from an APEC 2024 paper, 1 MHz (TRAFOLO on LinkedIn) | A commercial 3-D FEM tool | Winding loss within 6 % once both tools used the same core-loss data and a refined mesh |
Comparing your own model with a measurement or another tool
- Measure well below the self-resonant frequency, about a third of it or less, or remove the winding capacitance from the measured values first: towards the self-resonance it raises the apparent inductance and resistance.
- Calibrate the network analyzer at the end of the test cable (short, open, load, thru) and check cables and connectors. At high Q, a small phase error turns reactance into apparent resistance.
- Add the resistance of leads and terminals that the model leaves out before you compare DC or low-frequency values.
- Give both sides the same material data. Different core-loss fits of the same ferrite are the largest source of disagreement between tools: in the planar case above, the first comparison differed by a factor of 2 in core loss.
- Include the core loss when you compare AC resistance or Q. Above a few hundred kilohertz it can be most of the resistance of a ferrite inductor.
- Refine the mesh where the field is strongest, on winding faces next to gaps, until the compared quantity stops changing.
- Compare like with like: a 2D axisymmetric model is exact only for a round part. For E cores and rectangular coils it gives a range, depending on how the core is mapped into 2D.
- For temperatures, compare the rise above the ambient air, with the same losses, conductivities, base contact and radiation setting on both sides.