Skip to content

Simulation overview

The Simulink work was a design-time aid — it informed the topology, modulation, and gate-drive choices before any silicon was committed. It was not a live model and is not part of the runtime control loop.

Headline results

Metric Value Source
Predicted THD (pre-filter, LS-PWM) 4.9 % LS-PWM model output, validated against the IEEE 519-2022 < 8 % limit
Predicted efficiency > 95 % Pre-filter, ideal switches, design-target load (≈ 400 W)
Predicted gate-drive failure mode IR2110 floating reference The simulation that killed the IR2110 path and forced the TLP250 choice

What was simulated

Three Simulink models are kept in simulation/simulink/:

Model What it covers
chb-5level-v1.slx First-pass 5-level CHB with IPD LS-PWM, ideal switches, resistive load. The model that produced the 4.9 % THD figure.
chb-5level-v2.slx Revised topology with the gate-drive sweep (IR2110 vs. TLP250) added. The IR2110 path is where the floating-reference failure showed up.
chb-5level-rl-nospike.slx LC-filtered variant with an RL load — used to characterize the wavy-current artefact and tune the snubber.
Simulink — output at elevated switching frequency
Simulink output at an elevated switching frequency — used during the parameter sweeps that informed the final 5 kHz choice. Higher fsw sharpens the output and pushes the ripple energy up where a smaller LC filter can attenuate it, at the cost of switching loss.

What the model does not capture

  • Parasitic inductance. Gate-loop and DC-bus parasitics that cause real-world voltage spikes are not in the ideal-switch model. The snubber design (22 Ω 2 W + 2.2 nF / 630 V) was specified to bound these on the bench, not in simulation.
  • Dead time. Simulink uses ideal complementary switches; the firmware enforces 3 µs dead time on the IRFB4110. Shoot-through margin is verified on the scope, not in the model.
  • Sensor noise + isolation. The MCP3201 + 6N137 chain has SPI line inversion and rail-stuck detection in the firmware; none of that is reflected in the simulation's idealized sensor model.
  • Thermal coupling. The model assumes both bridges at room temperature; the bench validation actually measured per-bridge MOSFET case temperatures.

The bench-validated numbers are what go into the final report. The simulation explains the why of the topology and gate-drive choices.

See also