Skip to content

Changelog

All notable changes to RFlect are documented here. The format follows Keep a Changelog and the project adheres to Semantic Versioning.

Full, detailed per-release notes live in RELEASE_NOTES.md; this file is the concise summary.

[6.1.0] — 2026-07-30

3D radiation-pattern rendering correctness & robustness. The shared scaling machinery (equal symmetric axis limits + set_box_aspect([1,1,1])) was verified to render true proportions — a sphere renders as a sphere in both the passive and active routes — and these changes fix an active-save data-selection bug and harden the degenerate-input paths.

Fixed

  • Active 3D saves now use per-polarization data. save_to_results_folder passed the total-power arrays into the H-pol and V-pol active 3D plot calls, so the saved 3D_TRP_hpol / 3D_TRP_vpol images rendered the total-power pattern under a polarization label. They now receive h_power_dBm_2d / v_power_dBm_2d (matching the already-correct GUI display path).
  • Passive 3D rendering hardened against degenerate input. A NaN-containing or constant-gain (max == min) interpolated pattern no longer NaN-poisons / divides-by-zero into a blank surface; NaN-robust min/max + a zero-radius fallback now mirror the active route.
  • 3D axis setup guards degenerate extents. _setup_3d_axes no longer lets a flat (extent 0) or all-NaN pattern collapse the bounding box to lim=0 / lim=NaN (which blanked the axes); it falls back to a unit box.

Changed

  • process_data (3D render helper) now returns only the interpolated grid and its axes (data_interp, theta_interp, phi_interp); it previously also built Cartesian X/Y/Z and a db_to_linear radius that every caller discarded.
  • Removed dead code in plot_passive_3d_component: the unused gain_normalized local and the vestigial shadowing_enabled / shadow_direction parameters. Human-shadow fading is applied to the gain data upstream; these parameters had no effect inside the plot function (the shadow-cone overlay was never wired up). Re-add them if/when that overlay is implemented.

Tests

  • tests/test_3d_scaling_fixes.py: regression coverage for the active per-pol save, passive NaN/constant-gain robustness, _setup_3d_axes degenerate extents, and the process_data return contract.

Packaging

  • Windows executable hardened against AV/EDR false positives (Cylance and similar were blocking installs). RFlect.spec: disabled UPX compression (upx=False — a well-known AV/EDR heuristic trigger) and added a PE VERSIONINFO resource (version_info.txt) embedding Company/Product/File version metadata, previously absent. Also dropped a stale anthropic hidden-import left over from the removed AI/LLM stack. Code signing (so IT can allow-list RFlect by publisher certificate instead of per-build file hash) is scoped as follow-up work, not included here.

[6.0.0] — 2026-05-28

Consolidated release bundling the v5.1 (correctness & quick wins), v5.2 (RF analysis expansion) and v6.0 (automation & platform) milestones on top of v5.0.0 — 46 issues. MCP tool count grew to 61.

Added — new RF methods & analysis (v5.2 / v6.0)

  • Advanced RF methods in plot_antenna/rf_methods.py, exposed as MCP tools: axial-ratio / CP sense, 3-antenna absolute-gain method (#48), uniform-linear array factor with electronic steering / HPBW / SLL / grating-lobe flag (#46), planar near-field→far-field via 2D FFT (#47), S-parameter time-gating + port-extension de-embedding (#43), CTIA TIS + OTA test-plan templates (#42).
  • Optional trapezoidal TRP quadrature with pole/phi-wrap half-cells (#35).
  • Multiport Touchstone .s3p/.s4p + mixed-mode S-parameters (#31).
  • Measurement-uncertainty budgets / error bars on TRP & gain (#30); regulatory spec-mask checks (FCC Part 15 / ETSI) (#29) with automated PASS/FAIL limit-lines in reports (#28); statistical pattern averaging across repeat measurements (#32); n-antenna Min/Max VSWR/Eff/Gain comparison tables (#7, #34).
  • Exact Rice/Marcum-Q fade-margin model (#23); main-lobe guard on sidelobe detection (#22); compute_group_delay_dispersion() group-delay variance/std (#8).

Added — automation & platform (v6.0)

  • SCPI VNA control + chamber-positioner automation via a driver-Protocol + in-memory mock backend (CI-testable) + optional pyVISA/pyserial backends (#44, #45).
  • Cal-drift: threshold alerts on new runs (#4), gain-standard recertification reminders (#3), scheduled monitoring (#33), passive-calibration support (#6), cable-loss .s2p history tracking (#5).
  • Off-the-Tk-thread rendering helper (async_render) and report generation (#40, #24); determinate batch progress bars + dialog keyboard nav (#26).
  • pip install rflect[mcp] extra; [instruments] extra for live hardware (#39).

Changed — engineering & refactors (v5.1 / v6.0)

  • Decomposed the giant modules into cohesive submodules (zero API change via re-export): advanced_plots, extrapolation, advanced_analysis_config, docx_helpers; plotting.py shrank 3825→2986 lines (#36, #37, #38, #41).
  • Documented + test-locked the string return contract for the analysis/bulk MCP tools; cal_drift_report returns a dict and never raises (#9, #10).
  • Slimmer install: openai → [ai] extra; dropped unused keyring/cryptography; pyinstaller out of runtime reqs; numpy ceiling raised to <3.0.0 (#16, #17, #18).
  • Cross-platform GUI fonts + themed dialogs (#25); GUI errors routed to log/messagebox instead of print() / silent except…pass (#15, #19).

Fixed

  • Clarified TRP docstring (input is EIRP; the 1/4π factor is correct) and added a golden-reference regression test locking the TRP/gain core (#11, #12).

Docs

  • Keep-a-Changelog CHANGELOG.md; expanded glossary / measurement-types / MCP overview (#20, #21); v6.0 RF-method example-figure gallery + reproducible generator (#27); test coverage for previously-untested modules (#13, #14).

[5.0.0] — 2026-05-28

Zero-dependency, MCP-first relaunch. Removed the entire in-app AI/LLM stack (chat, AI report generation, provider abstraction, API-key store) — RFlect makes no LLM/API calls and needs no key or subscription. generate_report is deterministic by default with an optional agent-authored narrative. Added 6 MCP tools (compare_antennas, analyze_s11, analyze_group_delay, estimate_link_budget, analyze_mimo_diversity, generate_active_cal); MCP tool count 35 → 41. Renamed ai_analysis.py → analysis_engine.py.

[4.3.0] — 2026-05-28

Multi-angle iperf throughput-validation MCP tool (analyze_iperf_angle_sweep).

[4.2.0] — 2026-05-12

process_folder single-call folder-orchestration MCP tool; MkDocs Material docs site.

[4.1.9] — 2026-04-14

Patch: restored the Active Chamber Calibration routine.

Earlier releases (4.1.8 and prior) are detailed in RELEASE_NOTES.md.