What the Giunzioni Team’s Low-Frequency Solver Means for Balun Modelling
What the Giunzioni Team’s Low-Frequency Solver Means for Balun Modelling
A stabilized full-wave boundary-integral formulation is a real advance. Turning it into a predictive HF ferrite-transformer model still requires material, winding, port, thermal and bench evidence that the paper does not claim to provide.
Anyone who has wound a 49:1 or 64:1 transformer, changed the turns, shifted the compensation and measured again knows why better modelling matters. Viviana Giunzioni, Alberto Scazzola, Adrien Merlini and Francesco P. Andriulli have solved an important numerical problem on that road. Their work is not yet a balun or unun model—and that distinction makes the research more useful, not less.
Joeri’s short version: the paper makes one difficult class of low-frequency full-wave surface-integral calculations better conditioned. It does not turn a toroidal test geometry into a ferrite core with windings, nor does it replace the equivalent circuit, material characterization, thermal model and measured prototype.
The Exact Paper and the Exact Advance
The work is Low-Frequency Stabilizations of the PMCHWT Equation for Dielectric and Conductive Media: On a Full-Wave Alternative to Eddy-Current Solvers, published by the four named authors in IEEE Transactions on Antennas and Propagation, volume 73, issue 8, pages 5725–5740. An open author preprint exposes the full formulation and numerical tests; the Politecnico di Torino record identifies the final paper and DOI.
The paper addresses the Poggio–Miller–Chang–Harrington–Wu–Tsai equation: a surface-integral formulation used for penetrable dielectric and conductive bodies. Its standard discretization can become ill-conditioned or lose numerical accuracy as electrical size and frequency fall, mesh density rises, or conductivity changes. Giunzioni and colleagues rescale quasi-Helmholtz current components and precondition the system so the formulation remains stable across their low-frequency regimes. They also retain useful inductive and capacitive magnetic-frill excitations and support multiply connected geometries.
That is a specific statement about a PMCHWT boundary-element formulation. It is not evidence that every Method of Moments, finite-element or boundary-element solver becomes unusable at HF. Other formulations, low-frequency stabilizations and quasi-static or eddy-current solvers remain valid tools when their assumptions fit the problem.
“Quasi-Static” Does Not Mean One Current Mechanism
The paper separates three asymptotic cases. Its labels must travel with their definitions:
| Paper regime | Physical boundary in the formulation | What not to infer |
|---|---|---|
| Quasi-static regime (QSR) | The object is electrically small and, for this defined limit, conduction current inside the material is negligible relative to displacement current. | That displacement current dominates every electrically small HF component or every problem commonly called quasi-static. |
| Eddy-current-free regime (ECFR) | The object is electrically small and conductive, but skin depth is large relative to its characteristic size. | That conductor loss, geometry or inductive coupling can be ignored. |
| Skin-effect-dominated regime (SEDR) | The object is electrically small and conductive, with skin depth small relative to its characteristic size. | That a surface-current solution automatically describes ferrite hysteresis, temperature or a complete wound transformer. |
The boundary is precise: displacement current dominates only the authors’ defined QSR limit. The same paper deliberately treats conduction-dominated eddy-current and skin-effect limits too. “Low frequency” here describes electrical scale and numerical asymptotics, not a universal material mechanism.
What the Numerical Tests Demonstrate
The tests are carefully chosen to exercise the formulation, its conditioning and its asymptotic accuracy. They are not disguised transformer qualification tests.
| Test in the paper | Demonstrated result | Boundary |
|---|---|---|
| Torus and sphere conditioning sweeps | The stabilized system avoids the reported low-frequency and conductivity-related condition-number failures over the tested cases. | Canonical homogeneous penetrable bodies, not a core-and-winding assembly. |
| Conducting toroidal body with a magnetic-frill excitation | Calculated resistance and inductance follow the stated circuit-theory comparison over the conductivity sweep. | The torus is itself the homogeneous conducting body. It is not ferrite carrying a multi-turn winding. |
| Connected parallel-plate capacitive structure | The formulation recovers the expected capacitive behaviour within the reported numerical comparison. | A canonical capacitor geometry, not measured inter-winding capacitance in a transformer. |
| Highly conductive square-wire structure | The surface solution resolves current penetration as skin depth changes. | Conductor skin effect under the stated excitation, not nonlinear magnetic or thermal behaviour. |
The paper therefore demonstrates a stable numerical building block. It does not report a ferrite material model, a wound toroid, dielectric supports around an HF winding, a transformer turns ratio, common-mode or differential-mode ports, a kilowatt excitation, a thermal solution or a measured balun comparison.
The Bridge to a Real HF Transformer Model
I still see a credible bridge. A full-wave formulation that remains stable when the assembly is tiny relative to wavelength can help connect local conductor, dielectric and field behaviour to a circuit model. But the bridge needs additional spans.
| Model layer | Inputs that must be declared | Evidence it should produce |
|---|---|---|
| Geometry and conductors | Actual core dimensions, winding route, spacing, wire construction, joints, enclosure, dielectric supports and nearby conductors | Current distribution, leakage field, conductor loss and parasitic coupling at stated ports |
| Magnetic material | Complex permeability versus frequency, field amplitude, bias and temperature; anisotropy or manufacturing spread where material data support it | Magnetising impedance and local magnetic loss without treating a catalogue initial-permeability value as a power model |
| Equivalent circuit | Leakage inductance, winding capacitance, loss branches, electrical delay and port definitions | A compact model that can be fitted to and challenged by measured complex data |
| Modes and reference planes | Single-ended, differential and common-mode port transforms, calibration planes and fixture model | Insertion, reflection and mode-conversion terms that correspond to the installed use |
| Power and temperature | Waveform, accepted power, duty cycle, ambient, enclosure, airflow, mounting and temperature-dependent properties | Loss density coupled to a thermal model, followed by measured temperature and drift under the same boundary conditions |
| Nonlinearity | Field-dependent magnetic behaviour, hysteresis, saturation, conductor heating and any bias | Compression, harmonic or intermodulation behaviour when a linear frequency-domain model is no longer sufficient |
Fair-Rite’s technical catalogue is a useful reminder of the evidence burden: ferrite data are frequency-, geometry-, drive- and temperature-dependent, and broadband-transformer design includes both low- and high-frequency limits. One initial-permeability number is not enough to predict RF power loss. TDK’s ferrite application data likewise publish complex permeability, amplitude behaviour and loss versus temperature for specific materials and conditions.
A predicted “1 kW hot spot” requires local electromagnetic loss, nonlinear and temperature-dependent material data, thermal conductivity and contact, enclosure and cooling boundaries, and a converged electrothermal solution. A 1 kW generator setting by itself supplies none of that. Likewise, common-mode behaviour needs explicit modal ports and the complete winding, feed and return geometry. The stabilized PMCHWT paper supplies neither result.
Keep the Field Model and the Equivalent Circuit
A field solution and an equivalent circuit answer different but connected questions. The field solution can expose where current crowds, which dielectric region stores energy and where a material dissipates power. The circuit model can show how those effects appear at declared ports, how they cascade with the source and load, and which measured parameters identify them.
Neither should be granted authority by appearance alone. Fit the compact model to one dataset, then test it against different frequencies, complex loads, drive levels and temperatures. If one set of fitted values explains only the fixture used to derive it, it is not yet a predictive transformer model.
A Validation Ladder for a Balun or Unun
- Characterize the material boundary. Use manufacturer data only inside its stated frequency, flux, bias and temperature conditions. Where power prediction matters, measure representative cores or coupons from the actual material batch.
- Define the device ports. State the winding terminals, common reference, differential and common-mode definitions, source and load impedances, and the plane at which each quantity is reported.
- Calibrate to the useful plane. Characterize or de-embed leads and fixtures only over the bandwidth where that model remains passive and repeatable. Save complex data, not only an SWR or magnitude trace.
- Measure modes separately. Mixed-mode S-parameters distinguish differential transmission, common-mode transmission and mode conversion. A good differential match does not prove high common-mode impedance, and the reverse is also true.
- Challenge the small-signal model. Test representative complex loads and compare reflection, insertion, phase, transformation and mode conversion against simulation.
- Raise drive under control. Record accepted power, waveform, duty cycle, temperature, time to equilibrium and drift. Stop on rapid heating, arcing, odour, unstable impedance or unexpected spectral products.
- Validate the installed boundary. The enclosure, connectors, coax route, counterpoise, antenna and nearby conductors can change common-mode current and voltage stress even when the bench transformer is unchanged.
Keysight’s balanced-measurement documentation shows why the mode definition matters: a VNA derives differential and common responses from calibrated complex single-ended measurements and a declared port topology. Its de-embedding guidance makes the other limit just as clear—the instrument measures at calibrated reference planes, while fixture removal is only as trustworthy as the fixture characterization.
Primary and Authoritative References
- Giunzioni, Scazzola, Merlini and Andriulli — Low-Frequency Stabilizations of the PMCHWT Equation for Dielectric and Conductive Media, IEEE TAP
- Open author preprint — full formulation, asymptotic regimes and numerical examples
- Fair-Rite — Technical Information and broadband-transformer design guidance
- TDK Electronics — Ferrite material data: complex permeability, drive and temperature dependence
- Keysight — Balanced and mixed-mode S-parameter measurements
- Keysight — Calibration reference planes and fixture de-embedding
Joeri’s Bottom Line
Giunzioni, Scazzola, Merlini and Andriulli have produced something worth watching: a better-conditioned full-wave surface-integral route through difficult low-frequency and conductivity limits. That can become part of a stronger modelling chain for ferrite transformers.
But today the demonstrated chain stops before the ferrite core, winding, supports, modal ports, nonlinear material, thermal boundary and measured RF assembly. The honest next step is not to replace wind-and-test with a colourful field plot. It is to connect stable field mathematics to material data, an equivalent circuit, calibrated mixed-mode measurements and controlled power testing—then make every layer predict evidence it did not fit.
Mini-FAQ
- What did the Giunzioni team actually stabilize? A PMCHWT surface-integral formulation for penetrable dielectric and conductive bodies at low frequency, including defined eddy-current and skin-effect regimes.
- Does the paper prove that conventional HF solvers are unusable? No. It identifies breakdown and accuracy problems in a standard PMCHWT discretization and demonstrates one stabilization. Other stabilized full-wave and appropriate quasi-static methods remain available.
- Does the paper model a ferrite toroid with windings? No. Its toroidal example is a homogeneous conducting body driven by a magnetic-frill excitation, not a ferrite core carrying a multi-turn winding.
- Can the method already predict 1 kW hot spots? Not from the published tests. That prediction requires local loss, nonlinear and temperature-dependent material data, thermal boundaries and validation at the same power and duty conditions.
- Could this research still help balun and unun modelling? Yes. Stable electrically-small full-wave calculations could become one layer in a model that also includes real geometry, ferrite data, modal ports, an equivalent circuit, electrothermal coupling and bench validation.
- What should I measure on a real transformer? At minimum, measure calibrated complex reflection and transmission, differential/common-mode behaviour, representative complex loads, temperature and drift at declared power, waveform, duty and ambient conditions.