FFT Utilities

FFT is used everywhere. Image, video, audio and biomedical signal processing, just to name a few. FFT is a very CPU-intensive transformation and everybody wants the fastest solution. It’s here.

The Superpowered FFT processes with the same speed as Apple’s vDSP. We don’t claim we’re faster (sometimes we do), but definitely not slower.

FFT by itself is not really useful when you need to work with magnitudes and phases (such as audio), not just complex numbers. In this case, you need Polar FFT.

We made it super convenient for you: just one call to do the whole transformation. And here is the big deal: the Superpowered Polar FFT is at least 2x faster than Apple’s vDSP.

FFTComplex

METHOD

Complex in-place FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').

Parameters

Name	Type	Description
real	`Number (pointer in Linear Memory)`	Real part.
imag	`Number (pointer in Linear Memory)`	Imaginary part.
logSize	`Number`	Should be between 4 and 12 (FFT sizes 16 - 4096).
forward	`Boolean`	Forward or inverse.

Returns

None

FFTComplex
METHOD
Complex in-place FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').
Parameters
Name Type Default Description
real float * Real part.
imag float * Imaginary part.
logSize int Should be between 4 and 12 (FFT sizes 16 - 4096).
forward bool Forward or inverse.
Returns
None

Name	Type	Description
real	`float *`	Real part.
imag	`float *`	Imaginary part.
logSize	`int`	Should be between 4 and 12 (FFT sizes 16 - 4096).
forward	`bool`	Forward or inverse.

FFTReal

METHOD

Real in-place FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').

Parameters

Name	Type	Description
real	`Number (pointer in Linear Memory)`	Real part.
imag	`Number (pointer in Linear Memory)`	Imaginary part.
logSize	`Number`	Should be between 5 - 13 (FFT sizes 32 - 8192).
forward	`Boolean`	Forward or inverse.

Returns

None

FFTReal
METHOD
Real in-place FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').
Parameters
Name Type Default Description
real float * Real part.
imag float * Imaginary part.
logSize int Should be between 5 - 13 (FFT sizes 32 - 8192).
forward bool Forward or inverse.
Returns
None

Name	Type	Description
real	`float *`	Real part.
imag	`float *`	Imaginary part.
logSize	`int`	Should be between 5 - 13 (FFT sizes 32 - 8192).
forward	`bool`	Forward or inverse.

PolarFFT

METHOD

Polar FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').

Parameters

Name	Type	Description
mag	`Number (pointer in Linear Memory)`	Input: split real part. Output: magnitudes.
phase	`Number (pointer in Linear Memory)`	Input: split real part. Output: phases.
logSize	`Number`	Should be between 5 - 13 (FFT sizes 32 - 8192).
forward	`Boolean`	Forward or inverse. Inverse PolarFFT will clear (zero) the DC offset.
valueOfPi	`Number`	The function can translate pi to any value (Google: the tau manifesto). Use 0 for M_PI.

Returns

None

PolarFFT

METHOD

Polar FFT. Data packing is same as Apple's vDSP. Check the 'Using Fourier Transforms' page of Apple's vDSP documentation ('Data Packing for Real FFTs').

Parameters

Name	Type	Default	Description
mag	`float *`		Input: split real part. Output: magnitudes.
phase	`float *`		Input: split real part. Output: phases.
logSize	`int`		Should be between 5 - 13 (FFT sizes 32 - 8192).
forward	`bool`		Forward or inverse. Inverse PolarFFT will clear (zero) the DC offset.
valueOfPi	`float`	0.0	The function can translate pi to any value (Google: the tau manifesto). Use 0 for M_PI.

Returns

None