All pull requests

Pull requests

• Every single-device lumped and wire port keeps a named time record, and the settling check reads only the records behind the returned S

  lane:coax-mixed-port
  #1681·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Upgrading to 2.0, rfx terms, Check the model on the grid and the accuracy bar: four 2.0 pages filled (#1171)

  lane:ci-infralane:examples-docs
  #1679·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• forward() runs a normal-incidence plane wave on a graded mesh and returns real fields for an oblique one, as run() does; the injection-plane vacuum rule holds on every entry

  lane:absorberlane:ci-infralane:nu-meshmerge-train
  #1677·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• S-parameter JSON trimmed to numbers with an in-repo witness; the JSON-prose number check runs again (#1171)

  lane:ci-infralane:examples-docs
  #1674·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• An NTFF far field is refused on a model with a PEC, PMC or zero-depth absorbing face: it would miss the face's images (a half model gave a pattern 6 dB off)

  lane:absorberlane:ci-infralane:distributedlane:nu-meshlane:waveguide-portmerge-train
  #1673·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Patch tutorial compares with a converged openEMS ladder of the same structure: rfx +0.13 to +0.33 % (#737)

  lane:examples-docsmerge-train
  #1672·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• S-parameter support matrix folded into the support matrix; parity pins moved (#1171)

  lane:ci-infralane:examples-docslane:msl-port
  #1671·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• A z-only absorber leaves x and y unabsorbed: judge the no-absorber control by the cross-trace bar, not bit equality (arm64 Mac rounding)

  lane:absorberlane:ci-inframerge-train
  #1670·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Waveguide port: run() and compute_waveguide_s_matrix() report S at one default plane, the port's own position

  lane:ci-infralane:nu-meshlane:waveguide-portmerge-train
  #1669·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Labeler: the built-in waveforms and soft sources (rfx/sources/sources.py) earn lane:performance

  lane:ci-inframerge-train
  #1667·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• A multi-device run that falls back to one device says so in Result.diagnostics; the direct runner refuses arguments it would drop

  lane:ci-infralane:distributedmerge-train
  #1666·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• run() and forward() are compared byte for byte on 39 declarations and three paths; every remaining difference is a named row (53 today, zero at 2.0)

  lane:ci-inframerge-train
  #1665·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Finite flux boxes close on both meshes: one flux weight builder (trapezoid on real nodes, distance-weighted normal H, zero on a magnetic face)

  lane:absorberlane:ci-infralane:examples-docslane:msl-portlane:nu-meshlane:waveguide-portmerge-train
  #1664·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Microstrip example back on a uniform mesh, trace drawn at its solved width (#1171)

  lane:examples-docs
  #1663·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• The matrix pencil functions are named for the method they implement; the harminv names stay as deprecated aliases

  lane:absorberlane:ci-infralane:distributedlane:examples-docslane:nu-meshlane:performancemerge-train
  #1661·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• S from a run with several excited ports is refused where it mixed the drives

  lane:ci-infralane:coax-mixed-portlane:examples-docslane:nu-meshmerge-train
  #1660·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Mixed S: refuse the entries formed from the refuted per-cell lumped/wire wave

  lane:ci-infralane:coax-mixed-portlane:msl-portmerge-train
  #1659·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Lumped elements and f0 sheets stay free of design metal; a design box over them is refused; sources under design metal are reported

  lane:absorberlane:ci-infralane:coax-mixed-portlane:nu-meshmerge-train
  #1658·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• A port on a magnetic face stands for the full symmetric object: pin it, refuse the mixed flux channel there

  lane:absorberlane:ci-infralane:coax-mixed-portlane:examples-docsmerge-train
  #1657·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Smoothing a curved high-contrast dielectric gives a resonance further off than no smoothing, without warning

  lane:absorberlane:ci-infralane:nu-meshmerge-train
  #1656·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Support page refusal tables checked cell by cell against the path table (docs-consistency, stacked on #1634)

  lane:ci-infra
  #1655·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• A float32 graded profile is refused on the concrete path but admitted when traced

  lane:nu-mesh
  #1654·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• Conformal PEC warns that it enlarges a curved metal body; known limitation (#1651)

  lane:examples-docsmerge-train
  #1653·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·

• A plane wave's wrapped seam no longer opens a slot in a full-span sheet on run()

  lane:absorberlane:ci-inframerge-prioritymerge-trainmerge-train-blocked
  #1649·rfx-agent[bot] opened Oct 12, 2026Bot·

  ·
  1

• Paper-support scripts say the paper's numbers come from tag paper-tmtt-2026 and show what main gives; lumped-port gradient check redesigned (#737)

  lane:absorberlane:examples-docsmerge-trainmerge-train-blocked
  #1647·rfx-agent[bot] opened Oct 11, 2026Bot·

  ·
  1