dspark::FFTReal< T > Class Template Reference

FFT optimised for real-valued input signals (the common audio case). More...

#include <FFT.h>

Public Member Functions
	FFTReal (size_t size)
	Constructs a real FFT processor.
size_t	getSize () const noexcept
	Returns the number of real input samples (N).
size_t	getFrequencyDomainSize () const noexcept
	Returns the frequency-domain buffer size in elements (N + 2).
size_t	getNumBins () const noexcept
	Returns the number of frequency bins (N/2 + 1) including DC and Nyquist.
void	forward (const T *timeData, T *freqData) noexcept
	Forward transform: real time-domain -> complex frequency-domain.
void	inverse (const T *freqData, T *timeData) noexcept
	Inverse transform: complex frequency-domain -> real time-domain.
void	computeMagnitudes (const T *freqData, T *magnitudes) const noexcept
	Computes the magnitude of each frequency bin.
void	computePhases (const T *freqData, T *phases) const noexcept
	Computes the phase angle of each frequency bin.
void	computePowerSpectrum (const T *freqData, T *power) const noexcept
	Computes the power spectrum (magnitude squared) of each bin.

Static Public Member Functions
static T	binToFrequency (size_t binIndex, double sampleRate, size_t fftSize) noexcept
	Returns the frequency in Hz corresponding to a given bin index.
static size_t	frequencyToBin (double frequency, double sampleRate, size_t fftSize) noexcept
	Returns the bin index closest to a given frequency.

Detailed Description

template<typename T>
class dspark::FFTReal< T >

FFT optimised for real-valued input signals (the common audio case).

Uses a half-size complex FFT internally, saving ~50% computation.

Frequency domain layout: N+2 elements (interleaved complex).

• Bins 0 to N/2 inclusive -> (N/2 + 1) complex values -> (N + 2) floats.
• data[2*k] = real part of bin k.
• data[2*k+1] = imaginary part of bin k.
• Bin 0 = DC, bin N/2 = Nyquist.

In-place use is supported: timeData and freqData may point to the same buffer as long as it holds getFrequencyDomainSize() = N+2 elements.

Template Parameters

T	Sample type (float or double).

Definition at line 418 of file FFT.h.

Constructor & Destructor Documentation

FFTReal()

template<typename T >

dspark::FFTReal< T >::FFTReal ( size_t  size )

inlineexplicit

Constructs a real FFT processor.

Parameters

size	Number of real samples. Must be a power of two and >= 4.

Exceptions

std::invalid_argument

if size is invalid (with DSPARK_NO_EXCEPTIONS: asserts and degrades to the minimal valid size instead).

Definition at line 427 of file FFT.h.

Member Function Documentation

binToFrequency()

template<typename T >

static T dspark::FFTReal< T >::binToFrequency ( size_t  binIndex, double  sampleRate, size_t  fftSize  )

inlinestaticnoexcept

Returns the frequency in Hz corresponding to a given bin index.

Parameters

binIndex	Bin index (0 to N/2).
sampleRate	Sample rate in Hz.
fftSize	FFT size (N).

Returns

Center frequency of the requested bin.

Definition at line 527 of file FFT.h.

computeMagnitudes()

template<typename T >

void dspark::FFTReal< T >::computeMagnitudes ( const T *  freqData, T *  magnitudes  ) const

inlinenoexcept

Computes the magnitude of each frequency bin.

Parameters

freqData	Input: frequency-domain data (N+2 elements).
magnitudes	Output: N/2+1 magnitude values. Must be pre-allocated.

Definition at line 480 of file FFT.h.

computePhases()

template<typename T >

void dspark::FFTReal< T >::computePhases ( const T *  freqData, T *  phases  ) const

inlinenoexcept

Computes the phase angle of each frequency bin.

Parameters

freqData	Input: frequency-domain data (N+2 elements).
phases	Output: N/2+1 phase values in radians (-pi to +pi).

Definition at line 495 of file FFT.h.

computePowerSpectrum()

template<typename T >

void dspark::FFTReal< T >::computePowerSpectrum ( const T *  freqData, T *  power  ) const

inlinenoexcept

Computes the power spectrum (magnitude squared) of each bin.

Parameters

freqData	Input: frequency-domain data (N+2 elements).
power	Output: N/2+1 power values.

Definition at line 510 of file FFT.h.

forward()

template<typename T >

void dspark::FFTReal< T >::forward ( const T *  timeData, T *  freqData  )

inlinenoexcept

Forward transform: real time-domain -> complex frequency-domain.

Parameters

timeData	Input: N real samples.
freqData	Output: N+2 elements (interleaved complex, N/2+1 bins).

Definition at line 450 of file FFT.h.

frequencyToBin()

template<typename T >

static size_t dspark::FFTReal< T >::frequencyToBin ( double  frequency, double  sampleRate, size_t  fftSize  )

inlinestaticnoexcept

Returns the bin index closest to a given frequency.

Parameters

frequency	Frequency in Hz.
sampleRate	Sample rate in Hz.
fftSize	FFT size (N).

Returns

Integer index of the nearest bin.

Definition at line 539 of file FFT.h.

getFrequencyDomainSize()

template<typename T >

size_t dspark::FFTReal< T >::getFrequencyDomainSize ( ) const

inlinenoexcept

Returns the frequency-domain buffer size in elements (N + 2).

Definition at line 440 of file FFT.h.

getNumBins()

template<typename T >

size_t dspark::FFTReal< T >::getNumBins ( ) const

inlinenoexcept

Returns the number of frequency bins (N/2 + 1) including DC and Nyquist.

Definition at line 443 of file FFT.h.

getSize()

template<typename T >

size_t dspark::FFTReal< T >::getSize ( ) const

inlinenoexcept

Returns the number of real input samples (N).

Definition at line 437 of file FFT.h.

inverse()

template<typename T >

void dspark::FFTReal< T >::inverse ( const T *  freqData, T *  timeData  )

inlinenoexcept

Inverse transform: complex frequency-domain -> real time-domain.

Parameters

freqData	Input: N+2 elements (interleaved complex, N/2+1 bins).
timeData	Output: N real samples.

Definition at line 466 of file FFT.h.

---

The documentation for this class was generated from the following file:

• Core/FFT.h