CQT Experiment: Tossed in a branch for later

This experiment with Constant Q Transform was interesting, and it's nice
to see that it actually functions, but it's just too damn slow to really
be useful for visualization purposes. It uses nearly a full core on an
M1 processor, and I'd hate to see what it does on an Intel machine.
Stashing this in a branch and discarding it from the main tree, in case
somebody finds this code useful some day.

Signed-off-by: Christopher Snowhill <kode54@gmail.com>
CQTexperiment
Christopher Snowhill 2022-05-21 21:50:30 -07:00
parent bf54c45242
commit 024e36963a
10 changed files with 235 additions and 430 deletions

View File

@ -67,8 +67,6 @@
835FAC5F27BCA14D00BA8562 /* BadSampleCleaner.m in Sources */ = {isa = PBXBuildFile; fileRef = 835FAC5D27BCA14D00BA8562 /* BadSampleCleaner.m */; };
83725A9027AA16C90003F694 /* Accelerate.framework in Frameworks */ = {isa = PBXBuildFile; fileRef = 83725A7B27AA0D8A0003F694 /* Accelerate.framework */; };
83725A9127AA16D50003F694 /* AVFoundation.framework in Frameworks */ = {isa = PBXBuildFile; fileRef = 83725A7C27AA0D8E0003F694 /* AVFoundation.framework */; };
8377C64C27B8C51500E8BC0F /* fft_accelerate.c in Sources */ = {isa = PBXBuildFile; fileRef = 8377C64B27B8C51500E8BC0F /* fft_accelerate.c */; };
8377C64E27B8C54400E8BC0F /* fft.h in Headers */ = {isa = PBXBuildFile; fileRef = 8377C64D27B8C54400E8BC0F /* fft.h */; };
8377C65227B8CAD100E8BC0F /* VisualizationController.h in Headers */ = {isa = PBXBuildFile; fileRef = 8377C65027B8CAD100E8BC0F /* VisualizationController.h */; };
8377C65327B8CAD100E8BC0F /* VisualizationController.m in Sources */ = {isa = PBXBuildFile; fileRef = 8377C65127B8CAD100E8BC0F /* VisualizationController.m */; };
8384912718080FF100E7332D /* Logging.h in Headers */ = {isa = PBXBuildFile; fileRef = 8384912618080FF100E7332D /* Logging.h */; };
@ -76,6 +74,8 @@
839366681815923C006DD712 /* CogPluginMulti.m in Sources */ = {isa = PBXBuildFile; fileRef = 839366661815923C006DD712 /* CogPluginMulti.m */; };
8399CF2C27B5D1D5008751F1 /* NSDictionary+Merge.h in Headers */ = {isa = PBXBuildFile; fileRef = 8399CF2A27B5D1D4008751F1 /* NSDictionary+Merge.h */; };
8399CF2D27B5D1D5008751F1 /* NSDictionary+Merge.m in Sources */ = {isa = PBXBuildFile; fileRef = 8399CF2B27B5D1D4008751F1 /* NSDictionary+Merge.m */; };
83D44DC02839C60A00D4DD10 /* cqt.h in Headers */ = {isa = PBXBuildFile; fileRef = 83D44DBE2839C60A00D4DD10 /* cqt.h */; };
83D44DC12839C60A00D4DD10 /* cqt.c in Sources */ = {isa = PBXBuildFile; fileRef = 83D44DBF2839C60A00D4DD10 /* cqt.c */; };
83F18B1E27D1E8EF00385946 /* CDSPHBDownsampler.h in Headers */ = {isa = PBXBuildFile; fileRef = 83F18ADF27D1E8EF00385946 /* CDSPHBDownsampler.h */; };
83F18B1F27D1E8EF00385946 /* pffft_double.h in Headers */ = {isa = PBXBuildFile; fileRef = 83F18AE127D1E8EF00385946 /* pffft_double.h */; };
83F18B2027D1E8EF00385946 /* pf_neon_double_from_avx.h in Headers */ = {isa = PBXBuildFile; fileRef = 83F18AE327D1E8EF00385946 /* pf_neon_double_from_avx.h */; };
@ -199,8 +199,6 @@
835FAC5D27BCA14D00BA8562 /* BadSampleCleaner.m */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.objc; name = BadSampleCleaner.m; path = Utils/BadSampleCleaner.m; sourceTree = SOURCE_ROOT; };
83725A7B27AA0D8A0003F694 /* Accelerate.framework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.framework; name = Accelerate.framework; path = System/Library/Frameworks/Accelerate.framework; sourceTree = SDKROOT; };
83725A7C27AA0D8E0003F694 /* AVFoundation.framework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.framework; name = AVFoundation.framework; path = System/Library/Frameworks/AVFoundation.framework; sourceTree = SDKROOT; };
8377C64B27B8C51500E8BC0F /* fft_accelerate.c */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.c; path = fft_accelerate.c; sourceTree = "<group>"; };
8377C64D27B8C54400E8BC0F /* fft.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = fft.h; sourceTree = "<group>"; };
8377C65027B8CAD100E8BC0F /* VisualizationController.h */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.h; path = VisualizationController.h; sourceTree = "<group>"; };
8377C65127B8CAD100E8BC0F /* VisualizationController.m */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.objc; path = VisualizationController.m; sourceTree = "<group>"; };
8384912618080FF100E7332D /* Logging.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; name = Logging.h; path = ../../Utils/Logging.h; sourceTree = "<group>"; };
@ -208,6 +206,8 @@
839366661815923C006DD712 /* CogPluginMulti.m */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.objc; path = CogPluginMulti.m; sourceTree = "<group>"; };
8399CF2A27B5D1D4008751F1 /* NSDictionary+Merge.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; name = "NSDictionary+Merge.h"; path = "../../Utils/NSDictionary+Merge.h"; sourceTree = "<group>"; };
8399CF2B27B5D1D4008751F1 /* NSDictionary+Merge.m */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.objc; name = "NSDictionary+Merge.m"; path = "../../Utils/NSDictionary+Merge.m"; sourceTree = "<group>"; };
83D44DBE2839C60A00D4DD10 /* cqt.h */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.h; path = cqt.h; sourceTree = "<group>"; };
83D44DBF2839C60A00D4DD10 /* cqt.c */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.c; path = cqt.c; sourceTree = "<group>"; };
83F18ADF27D1E8EF00385946 /* CDSPHBDownsampler.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = CDSPHBDownsampler.h; sourceTree = "<group>"; };
83F18AE127D1E8EF00385946 /* pffft_double.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = pffft_double.h; sourceTree = "<group>"; };
83F18AE327D1E8EF00385946 /* pf_neon_double_from_avx.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = pf_neon_double_from_avx.h; sourceTree = "<group>"; };
@ -400,7 +400,6 @@
isa = PBXGroup;
children = (
83F18ADE27D1E8EF00385946 /* r8brain-free-src */,
8377C64A27B8C51500E8BC0F /* deadbeef */,
835C88AE279811A500E28EAE /* hdcd */,
835C88A22797D4D400E28EAE /* lvqcl */,
17D21DC40B8BE79700D1EBDE /* CoreAudioUtils */,
@ -482,20 +481,13 @@
name = Frameworks;
sourceTree = "<group>";
};
8377C64A27B8C51500E8BC0F /* deadbeef */ = {
isa = PBXGroup;
children = (
8377C64D27B8C54400E8BC0F /* fft.h */,
8377C64B27B8C51500E8BC0F /* fft_accelerate.c */,
);
path = deadbeef;
sourceTree = "<group>";
};
8377C64F27B8CAAB00E8BC0F /* Visualization */ = {
isa = PBXGroup;
children = (
8377C65027B8CAD100E8BC0F /* VisualizationController.h */,
8377C65127B8CAD100E8BC0F /* VisualizationController.m */,
83D44DBE2839C60A00D4DD10 /* cqt.h */,
83D44DBF2839C60A00D4DD10 /* cqt.c */,
);
path = Visualization;
sourceTree = "<group>";
@ -561,6 +553,7 @@
17D21CA10B8BE4BA00D1EBDE /* BufferChain.h in Headers */,
17D21CA50B8BE4BA00D1EBDE /* InputNode.h in Headers */,
17D21CA70B8BE4BA00D1EBDE /* Node.h in Headers */,
83D44DC02839C60A00D4DD10 /* cqt.h in Headers */,
83F18B3427D1E8EF00385946 /* r8butil.h in Headers */,
8399CF2C27B5D1D5008751F1 /* NSDictionary+Merge.h in Headers */,
17D21CA90B8BE4BA00D1EBDE /* OutputNode.h in Headers */,
@ -607,7 +600,6 @@
835C88B1279811A500E28EAE /* hdcd_decode2.h in Headers */,
8EC1225F0B993BD500C5B3AD /* ConverterNode.h in Headers */,
8384912718080FF100E7332D /* Logging.h in Headers */,
8377C64E27B8C54400E8BC0F /* fft.h in Headers */,
83F18B5027D1E8F000385946 /* r8bconf.h in Headers */,
835FAC5E27BCA14D00BA8562 /* BadSampleCleaner.h in Headers */,
83F18B2327D1E8EF00385946 /* pf_sse2_double.h in Headers */,
@ -712,11 +704,11 @@
17D21DC80B8BE79700D1EBDE /* CoreAudioUtils.m in Sources */,
83F18B2727D1E8EF00385946 /* pffft_double.c in Sources */,
8328995327CB511000D7F028 /* RedundantPlaylistDataStore.m in Sources */,
8377C64C27B8C51500E8BC0F /* fft_accelerate.c in Sources */,
839366681815923C006DD712 /* CogPluginMulti.m in Sources */,
835C88AA2797D4D400E28EAE /* lpc.c in Sources */,
17D21EBE0B8BF44000D1EBDE /* AudioPlayer.m in Sources */,
17F94DD60B8D0F7000A34E87 /* PluginController.mm in Sources */,
83D44DC12839C60A00D4DD10 /* cqt.c in Sources */,
17A2D3C60B8D1D37000778C4 /* AudioDecoder.m in Sources */,
8328995827CB51B700D7F028 /* SHA256Digest.m in Sources */,
17C940240B900909008627D6 /* AudioMetadataReader.m in Sources */,

View File

@ -1,40 +0,0 @@
/*
DeaDBeeF -- the music player
Copyright (C) 2009-2021 Alexey Yakovenko and other contributors
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
#ifndef FFT_H
#define FFT_H
#ifdef __cplusplus
extern "C" {
}
#endif
void fft_calculate(const float *data, float *freq, int fft_size);
void fft_free(void);
#ifdef __cplusplus
}
#endif
#endif

View File

@ -1,95 +0,0 @@
/*
DeaDBeeF -- the music player
Copyright (C) 2009-2021 Alexey Yakovenko and other contributors
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
#include "fft.h"
#include <Accelerate/Accelerate.h>
static int _fft_size;
static float *_input_real;
static float *_input_imaginary;
static float *_output_real;
static float *_output_imaginary;
static float *_hamming;
static float *_sq_mags;
static vDSP_DFT_Setup _dft_setup;
static void
_init_buffers(int fft_size) {
if(fft_size != _fft_size) {
fft_free();
_input_real = calloc(fft_size * 2, sizeof(float));
_input_imaginary = calloc(fft_size * 2, sizeof(float));
_hamming = calloc(fft_size * 2, sizeof(float));
_sq_mags = calloc(fft_size, sizeof(float));
_output_real = calloc(fft_size * 2, sizeof(float));
_output_imaginary = calloc(fft_size * 2, sizeof(float));
_dft_setup = vDSP_DFT_zop_CreateSetup(NULL, fft_size * 2, FFT_FORWARD);
vDSP_hamm_window(_hamming, fft_size * 2, 0);
_fft_size = fft_size;
}
}
void fft_calculate(const float *data, float *freq, int fft_size) {
int dft_size = fft_size * 2;
_init_buffers(fft_size);
vDSP_vmul(data, 1, _hamming, 1, _input_real, 1, dft_size);
vDSP_DFT_Execute(_dft_setup, _input_real, _input_imaginary, _output_real, _output_imaginary);
DSPSplitComplex split_complex = {
.realp = _output_real,
.imagp = _output_imaginary
};
vDSP_zvmags(&split_complex, 1, _sq_mags, 1, fft_size);
int sq_count = fft_size;
vvsqrtf(_sq_mags, _sq_mags, &sq_count);
float mult = 2.f / fft_size;
vDSP_vsmul(_sq_mags, 1, &mult, freq, 1, fft_size);
}
void fft_free(void) {
free(_input_real);
free(_input_imaginary);
free(_hamming);
free(_sq_mags);
free(_output_real);
free(_output_imaginary);
if(_dft_setup != NULL) {
vDSP_DFT_DestroySetup(_dft_setup);
}
_input_real = NULL;
_input_imaginary = NULL;
_hamming = NULL;
_sq_mags = NULL;
_dft_setup = NULL;
_output_real = NULL;
_output_imaginary = NULL;
}

View File

@ -19,7 +19,7 @@ NS_ASSUME_NONNULL_BEGIN
- (void)postSampleRate:(double)sampleRate;
- (void)postVisPCM:(const float *)inPCM amount:(int)amount;
- (double)readSampleRate;
- (void)copyVisPCM:(float *)outPCM visFFT:(float *)outFFT;
- (void)copyVisPCM:(float *)outPCM visCQT:(float *)outCQT;
@end

View File

@ -8,7 +8,7 @@
#import "VisualizationController.h"
#import <Accelerate/Accelerate.h>
#import "fft.h"
#import "cqt.h"
@implementation VisualizationController
@ -32,7 +32,7 @@ static VisualizationController *_sharedController = nil;
}
- (void)dealloc {
fft_free();
cqt_free();
}
- (void)postSampleRate:(double)sampleRate {
@ -59,10 +59,10 @@ static VisualizationController *_sharedController = nil;
}
}
- (void)copyVisPCM:(float *)outPCM visFFT:(float *)outFFT {
- (void)copyVisPCM:(float *)outPCM visCQT:(float *)outCQT {
@synchronized(self) {
cblas_scopy(4096, visAudio, 1, outPCM, 1);
fft_calculate(visAudio, outFFT, 2048);
cqt_calculate(visAudio, sampleRate, outCQT, 4096);
}
}

186
Audio/Visualization/cqt.c Normal file
View File

@ -0,0 +1,186 @@
//
// cqt.c
// CogAudio Framework
//
// Created by Christopher Snowhill on 5/21/22.
//
#include "cqt.h"
#include <stdlib.h>
#include <Accelerate/Accelerate.h>
struct CQTState {
float m_minFreq;
float m_maxFreq;
int m_bins;
float m_sampleFreq;
// enum WindowFunction; // Always Hamm
float m_Q;
FFTSetup m_fftSetup;
int m_fftLength;
int m_fftLogLength;
DSPSplitComplex m_fftLengthMatrix;
DSPSplitComplex m_kernel;
int m_K;
DSPSplitComplex m_KVector;
};
static struct CQTState m_state = { 0 };
static void *malloc_aligned(size_t size) {
void *ret = NULL;
if(posix_memalign(&ret, 8, size)) {
return NULL;
}
return ret;
}
static int dspsplit_alloc(DSPSplitComplex *out, size_t size) {
size_t realSize = size * sizeof(float);
out->realp = malloc_aligned(realSize);
out->imagp = malloc_aligned(realSize);
if(!out->realp || !out->imagp) return -1;
bzero(out->realp, realSize);
bzero(out->imagp, realSize);
return 0;
}
static void dspsplit_free(DSPSplitComplex *in) {
if(in->realp) free(in->realp);
if(in->imagp) free(in->imagp);
in->realp = NULL;
in->imagp = NULL;
}
static int cqtstate_alloc(struct CQTState *state, float minFreq, float maxFreq, int binsPerOctave, float sampleFreq /*, enum WindowFunction */) {
state->m_minFreq = minFreq;
state->m_maxFreq = maxFreq;
state->m_bins = binsPerOctave;
state->m_sampleFreq = sampleFreq;
/* state->m_windowFunction = func; */
const int K = (int)(ceilf(binsPerOctave * log2f(maxFreq / minFreq)));
state->m_K = K;
float Q = state->m_Q = 1 / (powf(2.0, 1.0f / binsPerOctave) - 1.0);
const int fftLogLength = state->m_fftLogLength = (int)(ceilf(log2f(Q * sampleFreq / minFreq)));
const int fftLength = state->m_fftLength = (int)(powf(2.0, fftLogLength));
if(dspsplit_alloc(&state->m_kernel, K * fftLength) < 0)
return -1;
state->m_fftSetup = vDSP_create_fftsetup(fftLogLength, FFT_RADIX2);
if(!state->m_fftSetup)
return -1;
const int maxN = (int)(ceilf(Q * sampleFreq / minFreq));
DSPSplitComplex window;
if(dspsplit_alloc(&window, maxN) < 0)
return -1;
DSPSplitComplex exponents;
if(dspsplit_alloc(&exponents, maxN) < 0) {
dspsplit_free(&window);
return -1;
}
for(int k = K; k > 0; --k) {
const int N = (int)(ceilf(Q * sampleFreq / (minFreq * pow(2.0, (float)(k - 1) / binsPerOctave))));
assert(N <= maxN);
bzero(window.imagp, sizeof(float) * N);
vDSP_hamm_window(window.realp, N, 0);
for(int i = 0; i < N; ++i) {
float inner = 2 * M_PI * Q * (float)(i) / (float)(N);
exponents.realp[i] = inner;
exponents.imagp[i] = inner;
}
vvcosf(exponents.realp, exponents.realp, &N);
vvsinf(exponents.imagp, exponents.imagp, &N);
const float nFloat = (float)(N);
vDSP_vsdiv(exponents.realp, 1, &nFloat, exponents.realp, 1, N);
vDSP_vsdiv(exponents.imagp, 1, &nFloat, exponents.imagp, 1, N);
DSPSplitComplex complexPointer;
complexPointer.realp = &state->m_kernel.realp[(k - 1) * fftLength];
complexPointer.imagp = &state->m_kernel.imagp[(k - 1) * fftLength];
vDSP_zvmul(&window, 1, &exponents, 1, &complexPointer, 1, N < fftLength ? N : fftLength, 1);
vDSP_fft_zip(state->m_fftSetup, &complexPointer, 1, fftLogLength, FFT_FORWARD);
}
dspsplit_free(&window);
dspsplit_free(&exponents);
vDSP_mtrans(state->m_kernel.realp, 1, state->m_kernel.realp, 1, fftLength, K);
vDSP_mtrans(state->m_kernel.imagp, 1, state->m_kernel.imagp, 1, fftLength, K);
vDSP_zvconj(&state->m_kernel, 1, &state->m_kernel, 1, fftLength * K);
const float lengthFloat = (float)(fftLength);
vDSP_vsdiv(state->m_kernel.realp, 1, &lengthFloat, state->m_kernel.realp, 1, fftLength * K);
vDSP_vsdiv(state->m_kernel.imagp, 1, &lengthFloat, state->m_kernel.imagp, 1, fftLength * K);
if(dspsplit_alloc(&state->m_fftLengthMatrix, fftLength) < 0)
return -1;
if(dspsplit_alloc(&state->m_KVector, K) < 0)
return -1;
return 0;
}
static void cqtstate_free(struct CQTState *in) {
if(in) {
dspsplit_free(&in->m_KVector);
dspsplit_free(&in->m_kernel);
dspsplit_free(&in->m_fftLengthMatrix);
if(in->m_fftSetup) vDSP_destroy_fftsetup(in->m_fftSetup);
in->m_sampleFreq = 0.0;
}
}
void cqt_calculate(const float *data, const float sampleRate, float *freq, int samples_in) {
if(!sampleRate) {
bzero(freq, 88 * sizeof(float));
return;
}
if(sampleRate != m_state.m_sampleFreq) {
cqtstate_free(&m_state);
if(cqtstate_alloc(&m_state, 27.5f, 4434.92f, 12, sampleRate) < 0) {
cqtstate_free(&m_state);
bzero(freq, 88 * sizeof(float));
return;
}
}
const int fftLength = m_state.m_fftLength;
const int K = m_state.m_K;
if(samples_in < fftLength) {
bzero(m_state.m_fftLengthMatrix.realp, sizeof(float) * fftLength);
}
cblas_scopy(samples_in > fftLength ? fftLength : samples_in, data, 1, m_state.m_fftLengthMatrix.realp, 1);
bzero(m_state.m_fftLengthMatrix.imagp, sizeof(float) * fftLength);
vDSP_fft_zip(m_state.m_fftSetup, &m_state.m_fftLengthMatrix, 1, m_state.m_fftLogLength, FFT_FORWARD);
vDSP_zmmul(&m_state.m_fftLengthMatrix, 1, &m_state.m_kernel, 1, &m_state.m_KVector, 1, 1, K, fftLength);
vDSP_zvmags(&m_state.m_KVector, 1, freq, 1, K);
vvsqrtf(freq, freq, &K);
}
void cqt_free() {
cqtstate_free(&m_state);
}

23
Audio/Visualization/cqt.h Normal file
View File

@ -0,0 +1,23 @@
//
// cqt.h
// CogAudio Framework
//
// Created by Christopher Snowhill on 5/21/22.
//
#ifndef cqt_h
#define cqt_h
#ifdef __cplusplus
extern "C" {
#endif
void cqt_calculate(const float *data, const float sampleRate, float *freq, int samples_in);
void cqt_free(void);
#ifdef __cplusplus
}
#endif
#endif /* cqt_h */

View File

@ -63,19 +63,13 @@ extern NSString *CogPlaybackDidStopNotficiation;
[self colorsDidChange:nil];
BOOL freqMode = [[NSUserDefaults standardUserDefaults] boolForKey:@"spectrumFreqMode"];
ddb_analyzer_init(&_analyzer);
_analyzer.db_lower_bound = LOWER_BOUND;
_analyzer.min_freq = 10;
_analyzer.max_freq = 22000;
_analyzer.peak_hold = 10;
_analyzer.view_width = 64;
_analyzer.fractional_bars = 1;
_analyzer.octave_bars_step = 2;
_analyzer.max_of_stereo_data = 1;
_analyzer.freq_is_log = 0;
_analyzer.mode = freqMode ? DDB_ANALYZER_MODE_FREQUENCIES : DDB_ANALYZER_MODE_OCTAVE_NOTE_BANDS;
[[NSNotificationCenter defaultCenter] addObserver:self
selector:@selector(colorsDidChange:)
@ -222,11 +216,7 @@ extern NSString *CogPlaybackDidStopNotficiation;
}
- (void)drawAnalyzer {
if(_analyzer.mode == DDB_ANALYZER_MODE_FREQUENCIES) {
[self drawAnalyzerDescreteFrequencies];
} else {
[self drawAnalyzerOctaveBands];
}
}
- (void)drawRect:(NSRect)dirtyRect {
@ -248,25 +238,15 @@ extern NSString *CogPlaybackDidStopNotficiation;
if(stopped) return;
float visAudio[4096], visFFT[2048];
float visAudio[4096], visCQT[88];
[self->visController copyVisPCM:&visAudio[0] visFFT:&visFFT[0]];
[self->visController copyVisPCM:&visAudio[0] visCQT:&visCQT[0]];
ddb_analyzer_process(&_analyzer, [self->visController readSampleRate] / 2.0, 1, visFFT, 2048);
ddb_analyzer_process(&_analyzer, [self->visController readSampleRate] / 2.0, 1, visCQT, 88);
ddb_analyzer_tick(&_analyzer);
ddb_analyzer_get_draw_data(&_analyzer, self.bounds.size.width, self.bounds.size.height, &_draw_data);
[self drawAnalyzer];
}
- (void)mouseDown:(NSEvent *)event {
BOOL freqMode = ![[NSUserDefaults standardUserDefaults] boolForKey:@"spectrumFreqMode"];
[[NSUserDefaults standardUserDefaults] setBool:freqMode forKey:@"spectrumFreqMode"];
_analyzer.mode = freqMode ? DDB_ANALYZER_MODE_FREQUENCIES : DDB_ANALYZER_MODE_OCTAVE_NOTE_BANDS;
_analyzer.mode_did_change = 1;
[self repaint];
}
@end

View File

@ -27,34 +27,14 @@
#include <stdlib.h>
#include <string.h>
#define OCTAVES 11
#define STEPS 24
#define ROOT24 1.0293022366 // pow(2, 1.0 / STEPS)
#define C0 16.3515978313 // 440 * pow(ROOT24, -114);
#pragma mark - Forward declarations
static float
_get_bar_height(ddb_analyzer_t *analyzer, float normalized_height, int view_height);
static void
_generate_frequency_labels(ddb_analyzer_t *analyzer);
static void
_generate_frequency_bars(ddb_analyzer_t *analyzer);
static void
_generate_octave_note_bars(ddb_analyzer_t *analyzer);
static void
_tempered_scale_bands_precalc(ddb_analyzer_t *analyzer);
static float _bin_for_freq_floor(ddb_analyzer_t *analyzer, float freq);
static float _bin_for_freq_round(ddb_analyzer_t *analyzer, float freq);
static float _freq_for_bin(ddb_analyzer_t *analyzer, int bin);
static float
_interpolate_bin_with_ratio(float *fft_data, int bin, float ratio, int fft_size);
@ -67,21 +47,16 @@ ddb_analyzer_alloc(void) {
ddb_analyzer_t *
ddb_analyzer_init(ddb_analyzer_t *analyzer) {
analyzer->mode = DDB_ANALYZER_MODE_FREQUENCIES;
analyzer->min_freq = 50;
analyzer->max_freq = 22000;
analyzer->view_width = 1000;
analyzer->peak_hold = 10;
analyzer->peak_speed_scale = 1000.f;
analyzer->db_lower_bound = -80;
analyzer->octave_bars_step = 1;
analyzer->freq_is_log = 1;
analyzer->bar_gap_denominator = 3;
return analyzer;
}
void ddb_analyzer_dealloc(ddb_analyzer_t *analyzer) {
free(analyzer->tempered_scale_bands);
free(analyzer->fft_data);
memset(analyzer, 0, sizeof(ddb_analyzer_t));
}
@ -97,11 +72,7 @@ void ddb_analyzer_process(ddb_analyzer_t *analyzer, int samplerate, int channels
channels = 2;
}
if(!analyzer->max_of_stereo_data) {
channels = 1;
}
if(analyzer->mode_did_change || channels != analyzer->channels || fft_size != analyzer->fft_size || samplerate != analyzer->samplerate) {
if(channels != analyzer->channels || fft_size != analyzer->fft_size || samplerate != analyzer->samplerate) {
analyzer->channels = channels;
analyzer->fft_size = fft_size;
analyzer->samplerate = samplerate;
@ -114,16 +85,7 @@ void ddb_analyzer_process(ddb_analyzer_t *analyzer, int samplerate, int channels
memcpy(analyzer->fft_data, fft_data, fft_size * channels * sizeof(float));
if(need_regenerate) {
switch(analyzer->mode) {
case DDB_ANALYZER_MODE_FREQUENCIES:
_generate_frequency_bars(analyzer);
break;
case DDB_ANALYZER_MODE_OCTAVE_NOTE_BANDS:
_generate_octave_note_bars(analyzer);
break;
}
_generate_frequency_labels(analyzer);
}
}
@ -137,15 +99,7 @@ void ddb_analyzer_tick(ddb_analyzer_t *analyzer) {
float *fft_data = analyzer->fft_data + ch * analyzer->fft_size;
ddb_analyzer_bar_t *bar = analyzer->bars;
for(int i = 0; i < analyzer->bar_count; i++, bar++) {
float norm_h = _interpolate_bin_with_ratio(fft_data, bar->bin, bar->ratio, analyzer->fft_size);
// if the bar spans more than one bin, find the max value
for(int b = bar->bin + 1; b <= bar->last_bin; b++) {
float val = analyzer->fft_data[b];
if(val > norm_h) {
norm_h = val;
}
}
float norm_h = fft_data[bar->bin];
float bound = -analyzer->db_lower_bound;
float height = (20 * log10(norm_h) + bound) / bound;
@ -182,9 +136,7 @@ void ddb_analyzer_get_draw_data(ddb_analyzer_t *analyzer, int view_width, int vi
draw_data->bar_count = analyzer->bar_count;
}
if(analyzer->mode == DDB_ANALYZER_MODE_FREQUENCIES) {
draw_data->bar_width = 1;
} else if(analyzer->mode == DDB_ANALYZER_MODE_OCTAVE_NOTE_BANDS) {
{
if(analyzer->fractional_bars) {
float width = (float)view_width / analyzer->bar_count;
float gap = analyzer->bar_gap_denominator > 0 ? width / analyzer->bar_gap_denominator : 0;
@ -212,12 +164,6 @@ void ddb_analyzer_get_draw_data(ddb_analyzer_t *analyzer, int view_width, int vi
draw_bar->xpos = bar->xpos * view_width;
draw_bar->peak_ypos = _get_bar_height(analyzer, bar->peak, view_height);
}
memcpy(draw_data->label_freq_texts, analyzer->label_freq_texts, sizeof(analyzer->label_freq_texts));
for(int i = 0; i < analyzer->label_freq_count; i++) {
draw_data->label_freq_positions[i] = analyzer->label_freq_positions[i] * view_width;
}
draw_data->label_freq_count = analyzer->label_freq_count;
}
void ddb_analyzer_draw_data_dealloc(ddb_analyzer_draw_data_t *draw_data) {
@ -239,113 +185,20 @@ _get_bar_height(ddb_analyzer_t *analyzer, float normalized_height, int view_heig
return height;
}
static void
_generate_frequency_labels(ddb_analyzer_t *analyzer) {
float min_freq_log = log10(analyzer->min_freq);
float max_freq_log = log10(analyzer->max_freq);
float view_width = analyzer->view_width;
float width_log = view_width / (max_freq_log - min_freq_log);
// calculate the distance between any 2 neighbour labels
float freq = 64000;
float freq2 = 32000;
float pos = width_log * (log10(freq) - min_freq_log) / view_width;
float pos2 = width_log * (log10(freq2) - min_freq_log) / view_width;
float dist = pos - pos2;
// generate position and text for each label
int index = 0;
while(freq > 30 && index < DDB_ANALYZER_MAX_LABEL_FREQS) {
analyzer->label_freq_positions[index] = pos;
if(freq < 1000) {
snprintf(analyzer->label_freq_texts[index], sizeof(analyzer->label_freq_texts[index]), "%d", (int)round(freq));
} else {
snprintf(analyzer->label_freq_texts[index], sizeof(analyzer->label_freq_texts[index]), "%dk", ((int)freq) / 1000);
}
pos -= dist;
freq /= 2;
index += 1;
}
analyzer->label_freq_count = index;
}
static void
_generate_frequency_bars(ddb_analyzer_t *analyzer) {
float min_freq = analyzer->min_freq;
float min_freq_log;
float view_width = analyzer->view_width;
float width;
if(analyzer->freq_is_log) {
min_freq_log = log10(analyzer->min_freq);
float max_freq_log = log10(analyzer->max_freq);
width = view_width / (max_freq_log - min_freq_log);
} else {
min_freq = analyzer->min_freq;
width = view_width / (analyzer->max_freq - min_freq);
}
float minIndex = _bin_for_freq_floor(analyzer, analyzer->min_freq);
float maxIndex = _bin_for_freq_round(analyzer, analyzer->max_freq);
int prev = -1;
analyzer->bar_count = 0;
if(analyzer->bar_count_max != analyzer->view_width) {
free(analyzer->bars);
analyzer->bars = calloc(analyzer->view_width, sizeof(ddb_analyzer_bar_t));
analyzer->bar_count_max = analyzer->view_width;
}
for(int i = minIndex; i <= maxIndex; i++) {
float freq = _freq_for_bin(analyzer, i);
// FIXME: only int position!
int pos;
if(analyzer->freq_is_log)
pos = width * (log10(freq) - min_freq_log);
else
pos = width * (freq - min_freq);
if(pos > prev && pos >= 0) {
// start accumulating frequencies for the new band
ddb_analyzer_bar_t *bar = analyzer->bars + analyzer->bar_count;
bar->xpos = pos / view_width; // normalized position
bar->bin = i;
bar->ratio = 0;
analyzer->bar_count += 1;
prev = pos;
}
}
}
static void
_generate_octave_note_bars(ddb_analyzer_t *analyzer) {
analyzer->bar_count = 0;
_tempered_scale_bands_precalc(analyzer);
if(analyzer->bar_count_max != OCTAVES * STEPS) {
if(analyzer->bar_count_max != 88) {
free(analyzer->bars);
analyzer->bars = calloc(OCTAVES * STEPS, sizeof(ddb_analyzer_bar_t));
analyzer->bar_count_max = OCTAVES * STEPS;
analyzer->bars = calloc(88, sizeof(ddb_analyzer_bar_t));
analyzer->bar_count_max = 88;
}
int minBand = -1;
int maxBand = -1;
ddb_analyzer_bar_t *prev_bar = NULL;
for(int i = 0; i < OCTAVES * STEPS; i += analyzer->octave_bars_step) {
ddb_analyzer_band_t *band = &analyzer->tempered_scale_bands[i];
if(band->freq < analyzer->min_freq || band->freq > analyzer->max_freq) {
continue;
}
for(int i = 0; i < 88; i += analyzer->octave_bars_step) {
if(minBand == -1) {
minBand = i;
}
@ -354,30 +207,11 @@ _generate_octave_note_bars(ddb_analyzer_t *analyzer) {
ddb_analyzer_bar_t *bar = analyzer->bars + analyzer->bar_count;
int bin = _bin_for_freq_floor(analyzer, band->freq);
int bin = i;
bar->bin = bin;
bar->last_bin = 0;
bar->ratio = 0;
// interpolation ratio of next bin of previous bar to the first bin of this bar
if(prev_bar && bin - 1 > prev_bar->bin) {
prev_bar->last_bin = bin - 1;
}
analyzer->bar_count += 1;
// get interpolation ratio to the next bin
int bin2 = bin + 1;
if(bin2 < analyzer->fft_size) {
float p = log10(band->freq);
float p1 = log10(_freq_for_bin(analyzer, bin));
float p2 = log10(_freq_for_bin(analyzer, bin2));
float d = p2 - p1;
bar->ratio = (p - p1) / d;
}
prev_bar = bar;
}
for(int i = 0; i < analyzer->bar_count; i++) {
@ -385,44 +219,6 @@ _generate_octave_note_bars(ddb_analyzer_t *analyzer) {
}
}
static float _bin_for_freq_floor(ddb_analyzer_t *analyzer, float freq) {
float max = analyzer->fft_size - 1;
float bin = floor(freq * analyzer->fft_size / analyzer->samplerate);
return bin < max ? bin : max;
}
static float _bin_for_freq_round(ddb_analyzer_t *analyzer, float freq) {
float max = analyzer->fft_size - 1;
float bin = round(freq * analyzer->fft_size / analyzer->samplerate);
return bin < max ? bin : max;
}
static float _freq_for_bin(ddb_analyzer_t *analyzer, int bin) {
return (int64_t)bin * analyzer->samplerate / analyzer->fft_size;
}
// Precalculate data for tempered scale
static void
_tempered_scale_bands_precalc(ddb_analyzer_t *analyzer) {
if(analyzer->tempered_scale_bands != NULL) {
return;
}
analyzer->tempered_scale_bands = calloc(OCTAVES * STEPS, sizeof(ddb_analyzer_band_t));
for(int i = 0; i < OCTAVES * STEPS; i++) {
float f = C0 * pow(ROOT24, i);
float bin = _bin_for_freq_floor(analyzer, f);
float binf = _freq_for_bin(analyzer, bin);
float fn = _freq_for_bin(analyzer, bin + 1);
float ratio = (f - binf) / (fn - binf);
analyzer->tempered_scale_bands[i].bin = bin;
analyzer->tempered_scale_bands[i].freq = f;
analyzer->tempered_scale_bands[i].ratio = ratio;
}
}
static float
_interpolate_bin_with_ratio(float *fft_data, int bin, float ratio, int fft_size) {
return bin < fft_size ? (bin + 1 < fft_size ? (fft_data[bin] + (fft_data[bin + 1] - fft_data[bin]) * ratio) : fft_data[bin]) : 0.0;

View File

@ -27,19 +27,9 @@
extern "C" {
#endif
#define DDB_ANALYZER_MAX_LABEL_FREQS 20
typedef struct {
float freq;
float ratio;
int bin;
} ddb_analyzer_band_t;
typedef struct {
// interpolation data
int bin;
int last_bin;
float ratio;
// normalized position
float xpos;
@ -58,35 +48,17 @@ typedef struct {
int bar_count;
ddb_analyzer_draw_bar_t *bars;
float bar_width;
// freq label drawing positions in view space
float label_freq_positions[DDB_ANALYZER_MAX_LABEL_FREQS];
char label_freq_texts[DDB_ANALYZER_MAX_LABEL_FREQS][4];
int label_freq_count;
} ddb_analyzer_draw_data_t;
typedef enum {
DDB_ANALYZER_MODE_FREQUENCIES,
DDB_ANALYZER_MODE_OCTAVE_NOTE_BANDS,
} ddb_analyzer_mode_t;
typedef struct ddb_analyzer_s {
// Settings
float min_freq;
float max_freq;
ddb_analyzer_mode_t mode;
/// Set to 1 after changing the @c mode or @c octave_bars_step at runtime, to refresh the internal state
int mode_did_change;
/// Generate fractional bar positions and width. Default is 0.
int fractional_bars;
/// Process 2 channels, if available, and use the max value. Default is 0.
int max_of_stereo_data;
/// How to calculate the gap between bars. E.g. 10 means bar_width/10. Default is 3.
/// If this value is 0, no gap will be created.
/// If the gap is >=1, a gap of at least 1px will be created, unless the bar width itself is 1px.
@ -112,15 +84,6 @@ typedef struct ddb_analyzer_s {
int channels;
int fft_size;
float *fft_data;
/// Calculated label texts and frequencies
float label_freq_positions[DDB_ANALYZER_MAX_LABEL_FREQS];
char label_freq_texts[DDB_ANALYZER_MAX_LABEL_FREQS][4];
int label_freq_count;
/// Tempered scale data, precalculated from fft pins
ddb_analyzer_band_t *tempered_scale_bands;
} ddb_analyzer_t;
ddb_analyzer_t *ddb_analyzer_alloc(void);