kamiyo / constantq

cpp implementation of a constantQ filter generator.

Contents

• General
• Constant Q
• Finite Impulse Response filter generator (windowed sinc)

General

Eigen library must be present for compilation to work. You can download at http://eigen.tuxfamily.org/index.php?title=Main_Page

Constant Q

This folder contains C++ code for generating the Constant Q transform matrix.

Compile with gcc or clang, for example:

clang++ -Ipath_to_Eigen_library -std=c++11 ./constantQ.cpp -o constantQ

Fastest executable I've created is with:

clang++ -Ipath_to_Eigen_library -std=c++14 -march=native -O3 -fopenmp ./constantQ.cpp -o constantQ

Make sure to install libomp-dev for OpenMP to work.

Program takes three versions:

./ConstantQ samplerate binsPerOctave minFrequency maxFrequency
./ConstantQ binsPerOctave minFrequency maxFrequency
./ConstantQ binsPerOctave maxFrequency

The version without samplerate will calculate filters for all common samplerates: 44100, 48000, 96000, 192000.

The version without samplerate or minFrequency will calculate optimum minFrequency

Example:

./ConstantQ 12 4434

.dat file header and format.

constantQ outputs the matrix in compressed sparse format. The .dat file is structured as follows.

Size in bytes	Field
31	Header
variable	Value
variable	Inner pointers
variable	Outer pointers

To get the size of the variable sections, see the header structure:

Header: 32 bytes

Byte offset	Type	Field
0-3	uint_32t	Sample rate (Hz)
4-7	uint_32t	Bins per octave
8-11	float	Lowest frequency
12-15	float	Highest frequency
16-19	uint_32t	# of rows
20-23	uint_32t	# of columns
24-27	uint_32t	size of inner pointers (# values)
28-31	uint_32t	size of outer pointers

Finite Impulse Response Filter (windowed sinc)

Compile with gcc or clang, for example:

clang++ -Ipath_to_Eigen_library -std=c++11 ./fir.cpp -o fir

Program takes 5 versions

./fir
./fir numCrossings
./fir numCrossings samplesPerCrossing
./fir numCrossings samplesPerCrossing cutoffCycles
./fir numCrossings samplesPerCrossing cutoffCycles kaiserBeta

The first variant will generate fir.dat with 7 crossings, 32 samples per crossing, a cutoff of 0.5, and a beta of 7.0. For the other variants, arguments that are not supplied will default to 512 samples per crossing, 0.5 cutoff, and 7.0 beta.

.dat file header and format.

Header:

Byte offset	Type	Field
0-3	uint_32t	Number of crossings
4-7	uint_32t	Samples per crossing
8-11	float	Cutoff cycles
12-15	float	Kaiser beta

Body:

• filterSize is SamplesPerCrossing * (NumCrossings - 1) - 1

Bytes	Type	Field
filterSize * sizeof(float)	float	Filter Coefficients
filterSize * sizeof(float)	float	Delta Coefficients (filter[i + 1] - filter [i])