Skip to content

rfx Documentation

rfx is a JAX-based differentiable FDTD simulator for RF and microwave engineering.

SurfaceWhat to expect
Recommended defaultuniform Cartesian Yee RF workflows: cavity, waveguide, patch-style resonance, probes, Harminv, selected S-parameter workflows, and benchmarked far-field workflows
Bounded port-family envelopeslumped/wire, microstrip-line, rectangular waveguide, and coaxial-line workflows each use their matching calculator and stated limits
Curated public APIuser-facing builders, observables, and reporting helpers; lower-level symbols are supporting reference
Other repository codemay exist in the repository, but is not a documented public workflow unless a guide and support entry cover it

Start with the recommended default lane unless a public guide explicitly routes you to a bounded envelope.

GuideDescription
InstallationPython/JAX install, GPU notes, dev setup
Quick StartFirst simulation with the current high-level API
Your First Patch AntennaFirst end-to-end resonance workflow
GuideDescription
Simulation APISimulation, Result, materials, sources, probes, ports, and NTFF helpers
Materials & GeometryMaterial library, Debye/Lorentz models, CSG shapes, and PCB stackup basics
Sources & PortsSoft sources, lumped/wire ports, microstrip-line ports, waveguide ports, and bounded coaxial-line reflection
Probes & S-ParametersDFT probes, S-matrix helpers, Harminv, de-embedding, and exports
Memory ReductionHow to reduce FDTD/AD memory without crossing validation boundaries
Waveguide PortsModal waveguide excitation and S-matrix extraction
GuideDescription
ValidationPublic validation overview and support status
Benchmark TableCurrent A/B cross-validation cases and reproduce commands
Convergence StudyMesh-refinement workflow
Far-Field & RCSNTFF radiation patterns and scattering workflows
Visualization & AnalysisPlots, exports, post-processing, and result interpretation
GuideDescription
Autodiff and Adjoint BackgroundGradient concepts for microwave engineers, mapped from Meep adjoint terminology to rfx autodiff workflows
Inverse DesignGradient-based optimization with proxy objectives and validation caveats
Gradient BehaviorWhere gradients are strong, weak, or noisy
Parametric SweepsSequential sweeps and jax.vmap batch evaluation
Patch Antenna DesignPractical rectangular patch workflow
  • Examples — recommended public runnable paths
  • Validation — support and validation overview
  • API — curated public API contract