# Recommended Configuration

Channel: release; package: 1.8.0; source: 7d1e97331656dd74ad0f5471653ac0c5effdff70

Page: https://remilab.ai/rfx/versions/v1.8.0/validation/reference-lane/
Source: https://github.com/bk-squared/rfx/blob/7d1e97331656dd74ad0f5471653ac0c5effdff70/docs/public/validation/reference-lane.mdx

This page is the reference lane (its URL slug): the configuration to start from
and the calculations whose limits are narrower than it.

## Start with the uniform grid

Use the **uniform Cartesian Yee** solver for ordinary RF examples and public
reports. The commonly documented setup includes:

- PEC, CPML, or UPML boundaries within their stated limits;
- point/current sources and the port type appropriate to the structure;
- probes, flux monitors, Harminv resonance extraction, selected S-parameter
  calculations, and the documented far-field workflows; and
- convergence checks and an analytic or independent solver comparison suited
  to the reported observable.

The [support matrix](https://remilab.ai/rfx/versions/v1.8.0/api/support-boundaries/) gives the general limits.
For port calculations, also read the
[S-parameter guidance](https://remilab.ai/rfx/versions/v1.8.0/guide/probes-sparams/).

## Calculations with narrower limits

| Configuration | Current use and restriction |
|---|---|
| rectangular waveguide S-matrix | Use `compute_waveguide_s_matrix(...)` for the documented rectangular-guide modes, frequency bands, normalization, reference planes, and port placement. Magnitude evidence is broader than phase and arbitrary-junction evidence. |
| microstrip-line S-matrix | Use `compute_msl_s_matrix(...)` for the documented quasi-TEM configurations. Exclude the whole frequency slice `S[:, :, k]` wherever `reliable[p, k]` is false (S is solved jointly across drives, #507); a true entry is not an accuracy guarantee. |
| lumped or wire-port S-parameters | Use `run(compute_s_params=True)` or the documented differentiable S11 path. These discrete lumped/wire feed models are not modal transmission-line ports. |
| coaxial transmission-line reflection | Use `compute_coaxial_line_reflection(...)` only with float32 precision, a nonperiodic 3D second-order uniform Yee grid, CPML on all six boundary faces with positive thickness on both z faces, `cpml_axes="z"`, and exactly one `face="top"` coaxial port. It builds its own line, source, probes, and termination, is not a general multi-port coaxial-network solver, and has no shared automatic passivity guard. See [Sources and Ports](https://remilab.ai/rfx/versions/v1.8.0/api/sources-ports/). |
| nonuniform graded-z mesh | Check the limit for each source, port, and observable. Regression or finite-gradient coverage alone does not establish RF accuracy. |
| differentiable design | Treat the optimization quantity as a proxy objective. Validate the final design independently with the applicable RF observable. |

## S-parameter API selection

- `run(compute_s_params=True)` computes lumped/wire-port S-parameters.
- `compute_msl_s_matrix(...)` computes the microstrip-line result.
- `compute_waveguide_s_matrix(...)` computes the rectangular-waveguide result.
- `compute_coaxial_line_reflection(...)` computes the one-port line-reflection
  result for the float32, nonperiodic 3D second-order uniform Yee and all-face
  CPML setup described above. Separately registered geometry, circuit elements,
  monitors, and coaxial termination helpers are rejected.
- Field sources, TFSF, probes, DFT planes, and flux monitors do not define a
  port impedance or S-matrix reference plane.

Do not substitute one calculator for another port family. A completed run or
an absent warning does not establish convergence, passivity, reference-plane
correctness, or agreement with an independent RF reference. Examples and
regression tests do not expand those validated limits either: a committed test
locks the behaviour it was written against, at the mesh, band and configuration
it was written for.