Claw-Pole Alternator on 3D FEM Flux Maps: Rectifier and Regulator
Simulate a 16-pole, delta-wound claw-pole alternator from 3D FEM flux maps. Explore rectifier ripple, field regulation, torque, copper losses and energy balance.
This claw pole alternator simulation follows a 16-pole machine after D. Bilyi, V. Bilyi and D. Gerling (2016). It uses WiredWhite’s own three-dimensional finite-element flux maps of the authors’ model, restored in a commercial 3D FEM program, with a solid rotor where the authors’ project laminates it. This is the three-phase variant, with one delta winding and a six-diode bridge. The paper’s stator has six phases arranged as two delta sets displaced by 30 electrical degrees.
The 3D view and the flux-path section colour the iron by |B| from WiredWhite’s 3D field solution at the default operating point. Changing the inputs changes the circuit’s results, not the field picture. Little of the claws’ flux reaches the stator (about 0.1–0.2 T in its teeth): with the restored geometry’s 2 mm gap, most of it closes between neighbouring claws.
Start with speed, load resistance and regulator set voltage. Open Details to change winding turns, field supply, diode characteristics, hysteresis, voltage filtering or time resolution. The alternator field current regulator closes the field switch below the lower threshold and opens it above the upper threshold, using the filtered bus voltage. The freewheel diode carries the field current while the switch is open. Play the final electrical period in the circuit picture to follow the solved currents and the conducting diodes through time.
The automotive alternator rectifier model includes phase resistance, magnetic saturation, armature reaction, delta circulating current, diode commutation and the field winding’s changing magnetic energy. The browser integrates the coupled winding and circuit equations from the stored maps; it does not rerun a 3D field solver. Changing either winding’s turns rescales flux and current at equal ampere-turns, with copper resistance proportional to turns squared.
At the defaults, 3000 rpm into 2 Ω with a 12 V setpoint and a 12 V field supply, the mean bus voltage is about 12.05 V and the field takes about 4.8 A, switched on about 44% of the time. With no smoothing capacitor, the three-phase bridge has about 1.8 V peak-to-peak ripple. Use the last-period charts to inspect phase voltages and winding current, and the full-time charts to inspect regulation, torque and field current. Means and rms values use the last whole electrical periods selected. Long or fast full histories use time-bin means with min/max bands; the last-period charts retain individual cycles.
The rotor here is solid, as real claw-pole rotors are; the authors’ original model laminates it along its axis with stacking factor 0.95. The field’s flux runs along the axis through the rotor core and claws, where such laminations carry flux poorly: with that rotor the same point needs about 34 A of field current and 1.2 kW of field copper loss, and the efficiency falls from about 60% to under 6%. A solve of the reference model with a solid rotor confirms these maps at one loaded point. With the 12 V field supply the regulator holds 12 V at 3000 rpm up to about 9.9 A of load, and into 2 Ω down to about 2100 rpm; beyond that the field switch stays on, the bus voltage falls and a warning says so.
As an alternator efficiency calculator, it counts both shaft power and field-supply power as inputs. It reports copper, diode, switch and stator iron losses and the magnetic-energy change separately. Speed and winding temperature are fixed; eddy currents in the solid rotor and friction are excluded. The map-extrapolation counter flags excursions beyond the solved current grid. Check it after increasing field voltage, reducing field turns or imposing a heavy load. The commercial 3D FEM magnetostatic reference (solid rotor), independent no-load EMF reference, power balance and finer-time-step checks accompany the tool.
How it is calculated
The 2,032-row Flux maps table supplies nonlinear winding flux linkages, torque and coenergy. machine_maps interpolates in field current, d/q current and rotor angle; turns rescale at equal ampere-turns. The Circuit table couples one delta stator set, six bridge diodes, R load, field switch, freewheel diode and filtered hysteresis regulator. Adaptive implicit BDF2 resolves switching events. Means and rms values use the last whole electrical periods; power balance includes magnetic-energy change. The picture follows the solved conduction states.
Assumptions
- Three-phase variant of D. Bilyi, V. Bilyi and D. Gerling (2016); their original has two delta sets 30° electrical apart. Sixteen poles; 160 phase turns and 100 field turns in the base maps.
- Rotor solid (isotropic M330-35A), as real claw-pole rotors are — not the original’s: its authors laminated the rotor along its axis at 0.95, as they did the stator (which stays so here). The field’s flux runs along the axis through the rotor core and claws, where such sheets carry at most about 20 times air’s permeability: with that rotor the default point takes about 34 A of field current instead of 4.8 A, 1.2 kW of field copper loss instead of about 23 W, and the efficiency falls from about 60% to under 6%. The reference model solved again with a solid rotor checks these maps at one loaded point.
- No magnets between the claws: the authors’ model has none (the paper’s machine shows them, without dimensions).
- The 2 mm air gap is the authors’ model’s, not confirmed for a production machine; a smaller gap would carry more flux into the stator.
- Field supply 12 V by default, as the paper’s field source: through 1.0 Ω at most 12 A (1200 A-turns). At 3000 rpm it holds the bus at 12 V up to about 9.9 A of load (1.2 Ω); into 2 Ω down to about 2100 rpm. Past that the field switch stays on, the bus falls below the set voltage and a warning says so.
- Winding resistances are estimates from the coil windows; the authors’ model gives none. Estimated field resistance: 1.0 Ω at 75 °C for 100 turns from the coil window, ×(N_f/100)². Phase resistance: 0.5948 Ω at 75 °C and 160 turns, ×(N_s/160)².
- Stator iron loss from WiredWhite’s 3D FEM at no load (table Stator iron loss), at the mean field current, added to the drive power: about 1.6 W at the default point, under 2% — the stator works at low flux density across the 2 mm gap (about 0.4 T in a typical tooth). Not included: eddy currents in the solid rotor’s claws, wheels and core (likely this machine’s larger iron loss; they need an eddy-current solver), loss from axial flux in the sheets, from cutting and stamping, minor loops, the load current’s harmonics, friction and windage. Constant speed and resistance; no bus capacitor or battery; field switch 10 mΩ, freewheel diode 0.7 V + 10 mΩ; field pre-excited at U_f/R_field, stator currents start at zero.
Where it applies
WiredWhite’s 0.83-million-tetrahedron 3D solves against a mesh twice as fine (1.75 million): phase flux linkages within 0.6%, the field’s within 0.2%, torque within 0.6%. Against the commercial program’s adaptive solve of the same model with its rotor solid at its loaded point P1: torque −0.2%, field flux linkage −0.9%, phase flux linkages within about 5% of the largest. Six loaded rotor angles resolve slot ripple of about 2% peak to peak, included in T_rip. No-load EMF is independently checked against dψ/dt of the raw FEM samples. The maps hold 13 field currents from 0 to 51 A, eight below 13 A where the solid rotor’s core saturates (knee near 4–5 A). There the regulated field current is the result most sensitive to the maps: leaving any one level out moves it by at most 4%, bus voltage and torque by at most 0.1%, so with all levels in it is estimated within about 1%; no-load points between the levels lie within 0.06% of the flux. Outside the mapped currents interpolation becomes linear extrapolation: Steps past the flux maps counts excursions. Increase simulated time if a changed operating point has not settled; the initial field is already excited. The default 60 ms run retains 24 electrical periods at 3000 rpm, averaging the last 20; it shortens the displayed history without coarsening time steps. Halving the maximum step changes mean voltage by 0.016% and torque by 0.025% at this point. Full histories adapt to bridge frequency and switching-event density. Above 20,000 required samples they show means and densely sampled min/max bands in up to 400 whole-ripple-period bins, with a final partial bin as needed. Means integrate the solver’s steps; extrema are sampling estimates. Last-period charts retain 600 samples.
References
- D. Bilyi, V. Bilyi, D. Gerling, “FEM Based Model Development and Co-Simulation of Automotive Multi-Phase Claw-Pole Alternator and Rectifier”, IEEE ITEC, 2016. ISBN 978-1-5090-0403-4. [Publication record](https://athene-forschung.unibw.de/92548?change_language=en&show_id=129887).
- [WiredWhite](https://wiredwhite.com/), own 3D FEM solves of the same model with a solid rotor, 2026-11-17; embedded Flux maps (0.83 M tetrahedra) and stator iron loss.
- Commercial 3D FEM program: adaptive magnetostatic solves by the WiredWhite team, 2026-11-14; embedded reference points XX25 (the authors’ laminated rotor, for the d-axis) and P1Solid (solid rotor).
Inputs
| Input | Symbol | Unit | Default |
|---|---|---|---|
| Speed | n | rpm | 3000 |
| Load resistance | R_L | ohm | 2 |
| Regulator set voltage | U_set | V | 12 |
| Turns of a phase | N_s | 160 | |
| Turns of the field winding | N_f | 100 | |
| Field supply voltage | U_f | V | 12 |
| Diode threshold voltage | V_f | V | 0.85 |
| Diode on-resistance | R_on | mohm | 4 |
| Regulator hysteresis | band | V | 0.1 |
| Voltage filter time constant | t_filt | ms | 0.2 |
| Simulated time | t_end | ms | 60 |
| Periods averaged at the end | window | 20 | |
| Minimum steps per electrical period | s_per | 200 | |
| Playback position 0° starts the final electrical period. | theta_play | deg | 273.6 |
| Show the stator Off: the rotor alone, to see the field in the claws. | show_stator | true |
Results
With the default inputs above, the calc gives:
| Result | Value |
|---|---|
| Bus voltage | 12.051 V |
| Load current | 6.03 A |
| Field current | 4.78 A |
| Field switch on | 43.5 % |
| Drive torque, with the stator's iron loss | 0.3097 N·m |
| Torque ripple, peak to peak | 0.0373 N·m |
| Bus voltage ripple, peak to peak | 1.84 V |
| Phase voltage, rms | 11.27 V |
| Phase current, rms | 2.6 A |
| Circulating current in the delta, rms | 0.17 A |
| Drive power | 97.3 W |
| Load power | 72.8 W |
| Diode losses | 12.5 W |
| Copper losses, stator and field | 35 W |
| Stator iron loss, at the mean field current | 1.6 W |
| Efficiency without the rotor's eddy and friction losses | 59.5 % |
| Electrical frequency | 400 Hz |
| Power balance residual | 0.311 % |
| Steps past the flux maps | 0 |
| Field supply power | 25 W |
| Stator copper losses | 12.1 W |
| Field copper losses | 22.8 W |
| Field switch losses | 0.1 W |
| Rate of stored magnetic energy change | 0.0021 W |
Checked against sources
Each case below was calculated by this calc and agrees with the value from its source (within the tolerance where one is shown).
Commercial 3D FEM, magnetostatic, rotor solid: torque and field flux linkage at P1
Source: The table Reference 3D FEM points: the reference model's adaptive magnetostatic solve in a commercial 3D FEM program with its rotor solid, loaded as at its point P1 (field 25.5 A; phases −14, −5.12 and 19.12 A). The machine of this calc's maps at those currents and that rotor angle, the reference model's d-axis from its no-load flux at XX25 (its authors' laminated rotor: the same geometry, so the same axis). Torque within 3 % (−0.2 %; the maps' mesh alone is worth about 0.6 %), field flux linkage within 2 % (−0.9 %).
Inputs: the default inputs above.
| Result | Source | This calc |
|---|---|---|
| Torque at reference point P1, rotor solid | -3.007 ± 0.09021 N·m | -3.014 N·m |
| Field flux linkage at P1, rotor solid | 0.2028 ± 0.004055 Wb | 0.201 Wb |
No-load EMF from independent raw FEM flux differentiation
Source: WiredWhite 3D FEM solves, 2026-11-14: the 48 no-load Flux maps rows at 25 A, 7.5° electrical spacing over 360°. An independent discrete Fourier transform of raw phase-A flux, with dψ/dt = ω_e dψ/dθ_e and harmonics 1–23, gives 23.8406681892 V rms at 3000 rpm (160 phase turns, 100 field turns). Compare the separate open-winding time-domain circuit, tolerance 0.5%; no delta circulating current in this verification circuit.
Inputs: the default inputs above.
| Result | Source | This calc |
|---|---|---|
| No-load phase EMF at 3000 rpm, 25 A | 23.841 ± 0.1192 V | 23.841 V |
Related Resources
Related articles
Related courses
Induction motor simulation using RMXPRT tool of ANSYS MAXWELL
IEC Electrical Schematics and Industrial Control Panels

Responses