Examples and Results

This page focuses on practical backend runs with expected outputs and validation criteria.

Runnable Python examples for new features

The May 9–10 roadmap (Fase 0–4) ships a set of runnable Python scripts covering every new user-facing feature. They live under examples/python/ and each one is a self-contained file with a TL;DR docstring at the top, sensible defaults, and an optional matplotlib plot guarded by PULSIM_EXAMPLE_NOPLOT=1 so it runs cleanly in headless CI:

Script	Demonstrates	Doc page
01_ac_sweep_rc.py	Analytical AC sweep + Bode (validates -3 dB / -45° corner).	AC Analysis
02_fra_vs_ac_sweep.py	Empirical FRA overlaid on the analytical AC sweep.	FRA
03_buck_template.py	pulsim.templates.buck auto-design + transient.	Converter Templates
04_boost_buckboost_templates.py	Boost + buck-boost templates side by side.	Converter Templates
05_codegen_pil.py	Codegen → C99 → gcc → PIL parity diff.	Code Generation
06_fmu_export.py	FMI 2.0 CS .fmu export + 13/13 callback symbol check.	FMI Export
07_parameter_sweep_lhs.py	64-point LHS over (L, C) with P5/P50/P95 metrics.	Parameter Sweep
08_monte_carlo_yield.py	256 MC draws under component tolerance, yield histogram.	Parameter Sweep
09_robustness_wrapper.py	pulsim.run_transient(robust=True) retry loop.	Robustness Policy
10_linear_solver_cache.py	Per-key cache hit rate via BackendTelemetry.	Linear-Solver Cache
11_yaml_ac_analysis.py	Load circuit from YAML + dispatch AC sweep.	AC Analysis

Header-only features (magnetic models, motors, three-phase grid, robustness profile) don't have Python bindings yet — their runnable counterparts live in examples/cpp/.

Legacy notebook examples

Example 1: RC Step (sanity baseline)

PYTHONPATH=build/python python3 benchmarks/benchmark_runner.py \
  --only rc_step \
  --output-dir benchmarks/out_rc

Expected behavior:

• V(out) rises exponentially.
• very low analytical error for RC waveform.
• low rejection count and no unstable fallback loop.

Example 2: Buck Converter (switching transient)

PYTHONPATH=build/python python3 benchmarks/benchmark_runner.py \
  --only buck_converter \
  --output-dir benchmarks/out_buck

Checkpoints:

• output settles around expected duty-dependent value.
• ripple remains physically plausible for chosen L/C and switching frequency.
• telemetry does not show runaway fallback escalation.

Example 3: Closed-Loop Buck (controller + PWM path)

PYTHONPATH=build/python python3 benchmarks/benchmark_runner.py \
  --only buck_mosfet_nonlinear \
  --output-dir benchmarks/out_cl_buck

Checkpoints:

• duty command remains bounded in [0, 1].
• output tracks reference without persistent divergence.
• control path remains deterministic between runs.

Example 4: Electro-Thermal Scenario

PYTHONPATH=build/python python3 benchmarks/benchmark_runner.py \
  --benchmarks benchmarks/electrothermal_benchmarks.yaml \
  --output-dir benchmarks/out_electrothermal

Checkpoints:

• device temperatures increase consistently with losses.
• thermal_summary.max_temperature remains within design envelope.
• efficiency and loss totals are coherent with operating point.

Output Artifacts for Automation

Main files used in CI/regression tooling:

• results.json: benchmark metrics and telemetry summaries
• results.csv: tabular benchmark export
• parity_results.json: external simulator parity metrics
• stress_results.json: tiered stress evaluation

Minimal results.json shape

{
  "benchmark_id": "buck_converter",
  "status": "passed",
  "runtime_s": 0.42,
  "steps": 19876,
  "max_error": 0.0018,
  "telemetry": {
    "newton_iterations": 53211,
    "timestep_rejections": 21,
    "linear_fallbacks": 3
  }
}

Notebook Coverage

Reference notebooks are under examples/notebooks and include:

• first-contact setup
• converter design scenarios
• thermal modeling
• benchmark and validation workflows

See Notebooks for execution details.