Maxwell

6.1  Maxwell Solver Types

Ansys Learning Hub: Maxwell Getting Started ↗ · 8 hr · E-learning

Three solver types cover power electronics electromagnetic analysis:

Magnetostatic

Solves the static (DC) magnetic field. Returns flux density distribution, inductance, force, and torque. Does not compute AC losses or eddy currents. Use for: inductance extraction, core flux density mapping, force on conductors or magnetic components, DC resistance distribution.

Eddy Current (AC frequency domain)

Solves the steady-state response at a single frequency. Returns ohmic losses including skin effect and proximity effect, core losses, and impedance. Use for: transformer winding losses at switching frequency, busbar AC resistance, skin depth verification. This is the starting point for most power magnetics loss estimation — 10–100× faster than transient.

Transient (time domain)

Solves the full time-varying field at arbitrary waveform. Most accurate for switching converters but the slowest solver by a large margin. Use when: actual waveform shape is critical (duty cycle effects, non-sinusoidal excitation), saturation occurs during the switching cycle, or inrush behavior is the concern.

Starting point recommendation: use Eddy Current at the fundamental switching frequency for transformer and inductor loss estimation. It produces a reliable estimate in a fraction of the transient solve time and is sufficient for most design decisions.

6.2  Geometry and Setup

Air region

The simulation domain must be bounded by an air or vacuum region. Create a box enclosing the device with 3× clearance in each principal direction as a starting point. Verify by comparing results at 3× and 5× clearance — if inductance or total loss changes by more than 2%, use 5×. The field must approach zero at the region boundary; too small an air region compresses field lines and produces wrong inductance and loss values.

Material assignment

Conductors: assign bulk conductivity (σ, S/m) at the operating temperature. Copper conductivity varies significantly with temperature (58 MS/m at 20°C, 48 MS/m at 75°C) — enter the value at operating temperature for accurate loss calculations.

Core material: assign relative permeability (μ_r) and core loss coefficients (Steinmetz parameters). Core loss coefficients must come from the specific core material datasheet — not from generic ferrite approximations in the material library. Using incorrect Steinmetz parameters for a different core material produces errors of 2–5× in core loss predictions.

Excitation

Current excitation: specifies total ampere-turns through a winding cross-section. The most common setup for inductors and transformers. Voltage excitation: specifies voltage across a winding terminal. Use when the voltage waveform is the controlled variable.

6.3  Maxwell Adaptive Meshing

Maxwell's meshing approach is fundamentally different from Mechanical. The solver manages the mesh automatically through an adaptive refinement process — the engineer sets convergence targets, not element sizes.

The adaptive process: the solver generates an initial coarse mesh, solves, and computes the energy error — the fractional change in total field energy between successive passes. Where the error is highest, the mesh is refined. This repeats until the energy error falls below the target or the maximum number of passes is reached.

Key settings^[1]

• Target energy error (Percent Error): set to 1% as a starting point. Smaller values produce more accurate but slower solutions.
• Maximum Number of Passes: stopping criterion if the energy error target is not reached. Default is 10. Increase to 15 or 20 if the solution does not converge within the default pass count.

Do not attempt to manually control Maxwell mesh density the way you would in Mechanical. The adaptive engine identifies where refinement is needed. If the solution fails to converge after 15 passes:

1. Check for geometry problems — thin gaps between conductors, near-zero-thickness insulation layers, or very small radii at conductor edges.
2. Add a manual mesh seed on the critical geometry (faces near windings or air gaps) to help the initial mesh capture those regions.
3. Increase the maximum passes to 20.

6.4  DC Conduction and Joule Heating

Use the DC conduction solver (or Eddy Current at near-zero frequency) to compute current distribution and ohmic losses in conductors under steady DC excitation.

Setup

Assign voltage boundary conditions at the conductor terminals: 0 V at one terminal, the applied DC voltage at the other. Assign bulk conductivity to all conductor bodies. The solver computes the current density field throughout the conductor volume.

Key results

• Current density (J, A/m²): shows where current concentrates. High-current-density regions have proportionally higher local power dissipation.
• Joule loss density (W/m³): volumetric power dissipation. Integrates to total ohmic loss. Compare the integrated total loss to the expected I²R for a sanity check before using the result.
• DC resistance: computed from total loss and applied current. Compare to the hand-calculated DC resistance.

Current density is not uniform in a conductor with varying cross-section — it concentrates at corners and at necked-down regions. The volumetric loss map (W/m³) can be exported directly as a distributed heat source in Icepak or Mechanical Thermal. This is more accurate than applying a uniform heat flux to the conductor surface.

6.5  Transformer Winding and Core Losses

Ansys Learning Hub: Electronic Transformer Simulation ↗ · 8 hr · E-learning

Use the Eddy Current solver at the switching frequency. For non-sinusoidal waveforms (square wave, trapezoidal), decompose the excitation into harmonics and run the solver at each significant harmonic; sum the losses.

Winding losses

Assign each winding layer as a separate conductor with its current excitation (ampere-turns, in the correct phase and direction for primary vs. secondary). The solver returns the current density distribution across each conductor cross-section.

Proximity effect: the magnetic field from adjacent winding layers induces eddy currents that add to each conductor's self-induced skin-effect currents. In multi-layer windings at high frequencies, proximity effect loss can exceed skin-effect loss by 5–10×. This is not visible in a simple resistance calculation. Maxwell shows it as a current density plot across the winding cross-section — high-density on the face closest to the adjacent layer is the proximity effect signature. This is the result that informs litz wire or winding interleaving decisions.

Core losses

The Steinmetz parameters (k, α, β) for the specific core material must be entered in the material definition. The solver returns the core loss density distribution, which shows whether loss is uniform in the core or concentrated at specific regions (indicating flux crowding at corners or in narrow sections).

6.6  Exporting Maxwell Losses to Icepak

Ansys Learning Hub: Maxwell Control Program ↗ · 1.5 hr · E-learning

After an Eddy Current or Transient solve, Maxwell produces a spatial distribution of ohmic losses (W/m³) across every conductor and core body in the model. This loss map can be imported into Icepak as a distributed heat source.

Export and import workflow

In Workbench: connect the Maxwell component to the Icepak component using a loss link (drag from Maxwell Solution cell to Icepak Setup cell). This connection transfers: geometry, material assignments, and the loss map. When Maxwell is re-solved, the connection propagates the updated loss map to Icepak automatically.

In Icepak: the imported loss map assigns different power density to different regions of the conductor body — high at surfaces (skin effect) and at faces adjacent to neighboring conductors (proximity effect). The Icepak solve uses this spatial distribution as the volumetric heat source.

The practical difference from uniform power assignment is largest where loss concentration is high: high-frequency windings, transformer cores with non-uniform flux density, and busbars with cross-section changes. For a copper winding at 200 kHz, the current and loss are concentrated in a skin depth of approximately 0.15 mm from the conductor surface — a significant fraction of the total conductor volume generates almost no heat, while the surface layer generates nearly all of it. For the 20 kHz transformer in the worked example (section 6.5), the exported loss map passed directly to Icepak, where the heatsink solve then used this spatially distributed transformer loss rather than a uniform 1.32 W smeared across the entire transformer body.