this repo has no description
update CLAUDE.md
+36
+36
CLAUDE.md
+36
CLAUDE.md
···
524
524
525
525
The profiling code has been added to measure actual performance on each platform.
526
526
527
+
## Current CQT Implementation Status (December 2024)
528
+
529
+
### What's Working:
530
+
- **16K FFT implemented** - provides excellent low-frequency resolution
531
+
- **cqt(bin)** function working correctly - returns raw CQT magnitude for bin 0-119
532
+
- **Frequency detection accurate** - 440Hz correctly maps to bin 54, etc.
533
+
- **Kernels properly generated** - sparse storage working, good frequency selectivity
534
+
- **Performance excellent** - Runtime on M1 Pro: 0.368ms total (0.255ms FFT + 0.113ms kernels)
535
+
536
+
### Runtime Performance Expectations:
537
+
- **M1 Pro MacBook**: 0.368ms total CQT processing (~2.2% of frame budget)
538
+
- **ThinkPad i5-1130G7 (extrapolated)**:
539
+
- Performance mode: ~0.426ms (2.6% of frame budget)
540
+
- Power save mode: ~1.273ms (7.6% of frame budget)
541
+
- Runtime is ~3.8x slower than synthetic benchmarks due to real audio data, cache effects
542
+
543
+
### Critical Implementation Details:
544
+
1. **FFT_SIZE temporarily changed** in fftdata.h from 1024 to 8192 to support CQT
545
+
- This breaks FFT resolution but enables CQT
546
+
- Must be reverted when separate audio buffer is implemented
547
+
548
+
2. **Key fixes that were essential**:
549
+
- Kernel phase calculation must use full FFT position: `2π * (f/fs) * (idx - N/2)`
550
+
- NO scaling factor after kernel FFT - this was critical!
551
+
- Normalization by windowLength before FFT
552
+
553
+
3. **Test scripts created**:
554
+
- `test_cqt_spectrum_v2.lua` - visual spectrum analyzer
555
+
- `test_cqt_a4.lua` - 440Hz tone generator
556
+
- `test_cqt_stable.lua` - controlled testing
557
+
558
+
### Resolution Characteristics with 16K FFT:
559
+
- **20Hz**: Q≈7.4 (window truncated to 16384 samples, but much better than 6K's Q≈1.86)
560
+
- **30Hz**: Q≈11.2 (good resolution)
561
+
- **45Hz+**: Full Q≈17 (perfect - no truncation)
562
+
527
563
## Phase 3: Next Steps
528
564
1. ~~Implement 16K FFT based on benchmark results~~ (COMPLETE)
529
565
2. Add remaining API functions: `cqts()`, `cqto()`, `cqtos()`