Skip to content

Commit de3a9bc

Browse files
author
Local Development
committed
Update session log with vhs_decode workflow copy details
1 parent 297b33b commit de3a9bc

1 file changed

Lines changed: 194 additions & 0 deletions

File tree

‎SESSION_LOG_OSCILLOSCOPE_FIXES.md‎

Lines changed: 194 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,194 @@
1+
# ld-analyse Oscilloscope Y/C Fixes - Session Log
2+
3+
**Date**: 2025
4+
**Branch**: `ld-analyse/improvements`
5+
**Status**: Completed and Pushed
6+
7+
---
8+
9+
## Summary
10+
11+
Fixed scanline oscilloscope display issues for dual-source (Y+C) TBC files, ensuring correct Y (luma) and C (chroma) channel display in all viewing modes. Added binary build GitHub Actions workflow for automated artifact generation.
12+
13+
---
14+
15+
## Issues Addressed
16+
17+
### 1. Y/C Data Extraction in BOTH_SOURCES Mode
18+
**Problem**: When viewing dual-source TBC files (separate Y and C), the scanline oscilloscope was displaying incorrect luma (Y) traces. The oscilloscope was using the mono decoder's luma output instead of the chroma decoder's luma, which meant the Y trace didn't reflect the selected PAL/NTSC chroma decoder settings.
19+
20+
**Solution**: Modified `tbcsource.cpp::getScanLineData()` to use the chroma decoder's luma output (`chromaDecoder.getYData()`) instead of mono decoder when in `BOTH_SOURCES` mode, ensuring Y trace matches the selected decoder.
21+
22+
**Files Modified**:
23+
- `tools/ld-analyse/tbcsource.cpp`
24+
25+
**Commit**: `96bdba2e` - "Fix scan-line oscilloscope Y/C data extraction in BOTH_SOURCES mode"
26+
27+
---
28+
29+
### 2. Chroma (C) Trace Mapping and Offset Handling
30+
**Problem**:
31+
- In YC mode (combined Y+C display), the oscilloscope showed only luma, not the composite signal
32+
- In C-only mode, the chroma trace was shifted vertically out of place
33+
- The chroma offset (32767) was being incorrectly removed in the data extraction, but the oscilloscope expected raw chroma values to calculate centering itself
34+
35+
**Root Cause**: The chroma signal needed to be mapped as "composite minus luma" for correct vertical positioning in the oscilloscope, matching the "combine" feature logic used elsewhere in the codebase.
36+
37+
**Solution**:
38+
1. Modified `getScanLineData()` to keep raw chroma values (with offset intact) for oscilloscope's own centering calculation
39+
2. For YC mode in BOTH_SOURCES: Generate synthetic composite by adding `(chroma - CHROMA_OFFSET)` to luma
40+
3. Modified `oscilloscopedialog.cpp` to render C trace as `composite - luma` for accurate positioning
41+
4. Preserved CHROMA_OFFSET constant and proper offset handling throughout the pipeline
42+
43+
**Technical Details**:
44+
- `CHROMA_OFFSET = 32767` centers chroma around zero when subtracted
45+
- Oscilloscope calculates midpoint IRE for chroma centering at line 201
46+
- Composite generation for Y+C sources: `composite = luma + (chroma - CHROMA_OFFSET)`
47+
- C trace display: `C = composite - Y` (computed in oscilloscope dialog)
48+
49+
**Files Modified**:
50+
- `tools/ld-analyse/tbcsource.cpp` - Scanline data extraction and composite generation
51+
- `tools/ld-analyse/oscilloscopedialog.cpp` - Chroma trace rendering calculation
52+
- `tools/ld-analyse/tbcsource.h` - Validated (staged and committed)
53+
54+
**Commit**: `de5d6747` - "Fix scanline oscilloscope Y+C and C mapping for dual-source TBC"
55+
56+
---
57+
58+
## Binary Build Workflow Addition
59+
60+
### Phase 1: Initial Binary Build Workflow
61+
**Problem**: Repository only had test workflows, no automated binary artifact generation for distribution and testing.
62+
63+
**Solution**: Created comprehensive GitHub Actions workflow for building release binaries on multiple platforms.
64+
65+
**Workflow Features**:
66+
- **Linux Qt5 Build**: Ubuntu 22.04, full Qwt support
67+
- **Linux Qt6 Build**: Ubuntu 22.04, no Qwt (unavailable for Qt6 on this platform)
68+
- **macOS Build**: Latest macOS runner, Qt5 with Qwt
69+
- **Triggers**: Push to `ld-analyse/improvements` and `main` branches, PRs, and releases
70+
- **Artifacts**: Compressed tarballs (.tar.gz) with 30-day retention
71+
- **Build Type**: Release (optimized)
72+
73+
**Files Created**:
74+
- `.github/workflows/binary-builds.yml`
75+
76+
**Commit**: `badb93b6` - "Add binary build GitHub Actions workflow"
77+
78+
### Phase 2: Copy Production Workflows from vhs_decode Branch
79+
**User Request**: Copy the complete, production-tested build workflows from the `vhs_decode` branch that include Windows, Linux, and macOS builds with proper packaging (AppImage, DMG, etc.).
80+
81+
**Solution**: Fetched and copied all build workflows from `fork/vhs_decode` branch.
82+
83+
**Workflows Copied**:
84+
- **Linux Tools**: `build_linux_tools.yml` - Builds AppImage with linuxdeploy, Qt5, includes FlaLDF and scripts
85+
- **Linux Decode**: `build_linux_decode.yml` - Builds Python decode tools
86+
- **Windows Tools**: `build_windows_tools.yml` - Builds Windows executables with Qt6 via vcpkg, includes FFmpeg, FLAC, FlaLDF
87+
- **Windows Decode**: `build_windows_decode.yml` - Builds Python decode tools for Windows
88+
- **macOS Tools**: `build_macos_tools.yml` - Builds DMG for both x86_64 and ARM64, includes FlaLDF (x86_64 only)
89+
- **macOS Decode**: `build_macos_decode.yml` - Builds Python decode tools for macOS
90+
- **PyPI Publishing**: `push_pypi.yml` - Workflow for publishing to PyPI
91+
92+
**Trigger Method**: All workflows use `workflow_dispatch` for manual triggering, allowing on-demand testing.
93+
94+
**Commit**: `d6807011` - "Copy build workflows from vhs_decode branch"
95+
96+
---
97+
98+
## Testing and Validation
99+
100+
**User Validation**:
101+
- User confirmed oscilloscope now correctly displays Y, C, and YC traces
102+
- Chroma positioning "much more accurate"
103+
- Y trace correctly reflects PAL/NTSC decoder selection in Y+C mode
104+
- All view modes (field/frame/split) working correctly with oscilloscope
105+
106+
**Code Quality**:
107+
- No compilation errors
108+
- Consistent with existing codebase patterns
109+
- Proper offset handling maintained throughout signal chain
110+
- Integration with main window modes validated
111+
112+
---
113+
114+
## Git Push Summary
115+
116+
**Pushed to**: `fork/ld-analyse/improvements` (https://github.com/harrypm/ld-decode.git)
117+
118+
**Commits Pushed**:
119+
1. `96bdba2e` - Fix scan-line oscilloscope Y/C data extraction in BOTH_SOURCES mode
120+
2. `de5d6747` - Fix scanline oscilloscope Y+C and C mapping for dual-source TBC
121+
3. `badb93b6` - Add binary build GitHub Actions workflow
122+
4. `d6807011` - Copy build workflows from vhs_decode branch (+ session log)
123+
124+
**Push Stats**:
125+
- First push: 24 objects, 5.12 KiB delta (oscilloscope fixes + initial workflow)
126+
- Second push: 12 objects, 8.45 KiB delta (vhs_decode workflows + session log)
127+
128+
---
129+
130+
## Next Steps
131+
132+
### Testing Workflows
133+
1. Navigate to: https://github.com/harrypm/ld-decode/actions
134+
2. Manually trigger workflows to test:
135+
- **Build Linux tools** - Test AppImage generation
136+
- **Build Windows tools** - Test Windows Qt6 executable with bundled dependencies
137+
- **Build macOS tools** - Test DMG creation for both architectures
138+
3. Download and validate artifacts from successful workflow runs
139+
4. Test binaries on respective platforms
140+
141+
### Oscilloscope Testing
142+
1. Test oscilloscope behavior with various TBC file types:
143+
- Single-source composite files
144+
- Dual-source Y+C files (S-Video style)
145+
- Chroma-only source files
146+
- PAL vs NTSC decoder outputs
147+
2. Verify Y trace follows selected chroma decoder (PAL/NTSC)
148+
3. Verify C trace positioning is accurate relative to Y and composite
149+
150+
---
151+
152+
## Technical Notes
153+
154+
### Signal Flow for Oscilloscope
155+
```
156+
TBC File(s) → ldDecodeMetaData → TBCSource
157+
↓
158+
getScanLineData() [tbcsource.cpp]
159+
↓
160+
┌───────────────┴───────────────┐
161+
↓ ↓
162+
BOTH_SOURCES ONE_SOURCE
163+
(Dual Y+C files) (Composite file)
164+
↓ ↓
165+
chromaDecoder.getYData() monoDecoder.getYData()
166+
chromaDecoder.getCData() (no chroma)
167+
↓ ↓
168+
Composite = Y + (C - 32767) Composite = raw composite
169+
↓ ↓
170+
└───────────────┬───────────────┘
171+
↓
172+
OscilloscopeDialog::updateScope()
173+
↓
174+
Render Y, C, and YC traces
175+
(C trace = Composite - Y)
176+
```
177+
178+
### Key Constants
179+
- `CHROMA_OFFSET = 32767` - Centers chroma signal around zero
180+
- Oscilloscope expects raw chroma with offset for internal centering
181+
- Composite generation removes offset: `Y + (C - CHROMA_OFFSET)`
182+
183+
---
184+
185+
## References
186+
187+
- Main repository: https://github.com/happycube/ld-decode
188+
- User fork: https://github.com/harrypm/ld-decode
189+
- Improvement branch: `ld-analyse/improvements`
190+
- Related issues: Scanline oscilloscope Y/C display, chroma decoder integration
191+
192+
---
193+
194+
**Session completed successfully. All changes committed and pushed.**

0 commit comments

Comments
 (0)