Skip to content

rfx Gallery

A short, visual tour of what rfx can do. Each case shows four things: what the structure is, what the fields do as they evolve, how close the result lands to the exact textbook answer, and — because rfx is a differentiable FDTD code — the gradient of that result, taken with jax.grad through the supported time-domain workflow. Pick one to dive in.

Dielectric Slab

Reflection and transmission of a dielectric slab vs the exact Fresnel curves

A plane wave at normal incidence on a lossless dielectric slab. Reflection and transmission track the exact transfer-matrix curves to ~1%, and jax.grad returns dT/dd matching the closed-form derivative. See it →

Patch Antenna

Patch antenna return loss with a deep matched dip at resonance

A 2.4 GHz microstrip patch on FR4. A tuned inset feed gives a deep matched return-loss dip at the TM010 resonance, cross-checked with a Harminv ring-down, plus a differentiable substrate-permittivity sweep. See it →

Rectangular Waveguide

TE10 mode propagating down a WR-90 rectangular waveguide

The TE10 dominant mode down a WR-90 X-band guide. |S11|/|S21| match the exact matched empty-guide answer at the CPML floor, and jax.grad differentiates the transmission through a dielectric fill. See it →

Anti-Reflection Coating Design

X-band-mean reflection cost converging to the exact transfer-matrix optimum over 60 gradient-descent iterations

A gradient-designed 3-layer anti-reflection coating on a high-permittivity substrate. Gradient descent through the FDTD (jax.grad) drives the X-band-mean reflection cost to about 0.99x the exact transfer-matrix optimum in 60 iterations. See it →

Computed with rfx · 2026-06-30