Fix NEC takeoff analysis for 0.43.0

Author: bubakbubak500

README.md

@@ -49,8 +49,9 @@ confirmation.
   replay, versioned provenance, blocked validation, and separate NEC,
   observed-coverage, and propagation-normalized layers;
 - built-in NEC2 wire-modeling workbench with five templates, validation,
-  3D geometry, impedance/current/pattern results, reproducible baselines, and
-  an optional separately installed OpenNEC calculation process;
+  3D geometry, impedance/current/pattern results, explicit take-off elevation,
+  practical ionospheric-hop geometry, reproducible baselines, and an optional
+  separately installed OpenNEC calculation process;
 - ADIF/ADI and CSV import, plus CSV export;
 - local SQLite storage and diagnostic export;
 - English and Czech application UI;

docs/RELEASE_NOTES_0.43.0.md

@@ -0,0 +1,38 @@
+# Antenna Pattern Lab 0.43.0
+
+## NEC candidate-grid reliability
+
+- Fixes assisted height/ground grids failing before OpenNEC with an ASCII
+  encoding error when the generated candidate name contained `·`, `Δ`, Czech
+  text, or another Unicode character.
+- Stores the model name in the ASCII NEC deck as an escaped `CM NAME-JSON`
+  comment. Import restores the exact Unicode value, so the supported deck
+  round trip and model SHA-256 remain lossless.
+- Verifies the complete Yagi candidate path against standalone OpenNEC 2.2.0.
+
+## Clear take-off angle and practical interpretation
+
+- Converts NEC `theta` (measured from zenith) to elevation above the horizon
+  (`90° − theta`) and presents the vertical cut on an explicit `0–90°` axis.
+- Adds a horizon/ground plane, azimuth references, and a marked peak ray to the
+  rotatable 3D far-field surface.
+- Adds a CZE/ENG practical panel with peak take-off elevation and azimuth,
+  sampled upper-hemisphere power shares, radio horizon, and spherical-Earth
+  one-hop geometry for representative E (110 km) and F2 (300 km) virtual
+  heights.
+- Clearly separates antenna-pattern geometry from propagation prediction:
+  current MUF/foF2, absorption, refraction, terrain, polarization, link budget,
+  and ground-wave coverage are not inferred.
+
+## Interface and verification
+
+- Prevents the workbench toolbar and tabs from clipping at the minimum width,
+  including native Classic.
+- Validates Monitor Light, Monitor Dark, and Classic in Czech and English at
+  1180×720, 1366×850, and 1920×1080, plus 125%, 150%, and 200% scaling.
+- Includes focused regression tests for Unicode NEC decks, spherical
+  ionospheric-hop geometry, radio horizon, pattern-power integration, the
+  corrected elevation axis, and the candidate grid.
+
+OpenNEC remains an optional, separately installed MIT-licensed executable.
+Antenna Pattern Lab does not bundle or link the solver.

docs/USER_GUIDE.md

@@ -207,7 +207,27 @@ the temporary directory. The stored result contains the exact model revision,
 engine path/version, UTC time, command parameters, input/output SHA-256, and
 normal NEC output. The result tabs show R/X and 50 Ω SWR over frequency, wire
 currents, peak and front/back values, selectable absolute/relative azimuth and
-elevation cuts, and a mouse-rotatable 3D far-field surface.
+elevation cuts, and a mouse-rotatable 3D far-field surface. NEC reports
+`theta` from the zenith; the workbench converts it to the operational take-off
+elevation `90° − theta`, where `0°` is the horizon and `90°` is straight up.
+The 3D view draws the horizon plane, azimuth references, and the peak ray so a
+low or high lobe is not inferred from the viewing perspective.
+
+**Practical interpretation** reports the peak elevation/azimuth, approximate
+relative power shares in the `0–10°`, `10–30°`, and `30–90°` upper-hemisphere
+bands, and the radio horizon from the highest modeled point. It also gives
+spherical-Earth one-hop geometry for assumed virtual heights of 110 km
+(representative E region) and 300 km (representative F2 peak). The layer ranges
+are consistent with the [NOAA ionospheric-region
+definitions](https://www.ngdc.noaa.gov/stp/IONO/ionostru.html); this simple
+geometry is deliberately not an [ITU-R P.533 HF propagation
+prediction](https://www.itu.int/rec/R-REC-P.533/en).
+
+The displayed hop distance does **not** establish that the selected frequency
+will be supported. It excludes current foF2/MUF, D-region absorption, refraction
+details, terrain, receiver height, polarization mismatch, and link budget.
+The NEC far-field surface also does not calculate ground-wave coverage, which
+depends on polarization, frequency, ground conductivity, terrain, and loss.
 
 Saved independent baselines automatically appear in **Propagation
 Intelligence → Three layers**, next to raw observed coverage and the separate
@@ -216,7 +236,7 @@ height/ground combinations separately. Orientation selection uses only the
 training time blocks and always reports error on later, previously unused
 campaign blocks; it never replaces the original independent baseline.
 
-Version 0.42.0 intentionally supports a documented NEC2 subset: `GW`, `GE`,
+Version 0.43.0 intentionally supports a documented NEC2 subset: `GW`, `GE`,
 `GN`, voltage `EX 0`, series-RLC `LD 0`, linear `FR`, and the generated `RP`.
 Unsupported cards are refused instead of silently discarded. Patches,
 buildings, volumetric solids, a real coax/feed-line model, full terrain,

docs/ui-audit.md

@@ -16,6 +16,43 @@ that contain radiation data, applies equal physical axis ranges, formats table
 coordinates to six significant digits, reserves two visible wire rows at the
 minimum supported height, and localizes/resizes the candidate table.
 
+## 0.43 NEC take-off interpretation and candidate-grid correction
+
+Two operator captures from 0.42.0 established the regression baseline: a Yagi
+candidate grid stopped with an ASCII encoding exception, and the 3D far-field
+surface did not expose the horizon or explain that NEC theta is measured from
+zenith. The source failure was reproduced with the exact `·`/`Δ` candidate
+name, then exercised end to end with standalone OpenNEC 2.2.0 after the fix.
+The supplied before states are retained as
+`dialog-antenna-modeling-0.42-candidate-ascii-error-cze-dark.png` and
+`dialog-antenna-modeling-0.42-radiation-ambiguous-elevation-cze-dark.png`
+under [`docs/ui/before`](ui/before/).
+
+The first visual pass of the real revised dialog found and corrected:
+
+1. the practical interpretation form forced long Czech and English use/ground
+   wave explanations into an unreadably narrow value column;
+2. the horizon caption collided with the 270° reference while the 3D axes used
+   too little of the available plotting area;
+3. the single-row English toolbar clipped Create/Calculate controls in native
+   Classic at 1180×720;
+4. the peak direction value clipped its `dBi` suffix at the same breakpoint.
+
+The final dialog uses a two-row toolbar, non-scrolling four-tab bar, full-width
+interpretation paragraphs, a larger equal-scale upper-hemisphere surface,
+horizon plane, four azimuth references, and a marked peak ray. The 2D vertical
+cut is Cartesian with an explicit `0–90° elevation above horizon` axis and a
+peak guide. The information panel states that its E/F2 distances are spherical
+geometry for assumed virtual heights—not MUF/foF2, absorption, terrain,
+ground-wave, or link-budget predictions.
+
+Final evidence under [`docs/ui/after`](ui/after/) covers Czech Monitor Light at
+1180×720, English Monitor Dark at 1366×850 and 1920×1080, English Classic at
+1180×720, the corrected 2D results, and the populated candidate table. The same
+seven states were rendered and inspected in isolated processes at 125%, 150%,
+and 200% Qt scaling. No obvious P0/P1 clipping, overlap, theme-parity, or chart
+meaning issue remains.
+
 ## Evidence and method
 
 The current source was rendered with an isolated temporary SQLite database and