==========================================================================================
==========================================================================================
******************************************************************************************
QRealFourier
******************************************************************************************
==========================================================================================
==========================================================================================
A Qt-based C++ library for Fast Fourier Transform (FFT) of real samples.
Last Release: 12 June 2013
Latest version: 0.3.0
Developer: Christoph Stallmann, University of Pretoria
==========================================================================================
About
==========================================================================================
QRealFourier is a Qt library for calculating FFTs on real samples. This library is based
on FFTReal designed and created by Laurent de Soras. I take no credit and on the work done
in FFTReal. This purpose of this library is to create a simple-to-use interface for novice
FFT users in Qt. A single class (QFourierTransformer) provides a number of function for
handling forward and inverse FFTs and rescaling. A lot of the work is done during compile-
time which creates a more efficient FFT implementation. The library is slightly larger and
takes up a couple of extra bytes, but this is a fair trade-off for simplicity.
==========================================================================================
License
==========================================================================================
QRealFourier is distributed under the Lesser GNU General Public License (LGPL) version 3
(LGPLv3). Copyright © 2012 Christoph Stallmann, University of Pretoria.
Please refer to License.txt and Copyright.txt for further information.
==========================================================================================
Requirements
==========================================================================================
QRealFourier requires CMake (2.8.9 or later) for makefile generation and Qt (5.0 or later).
Any native compilers should do, recommended are g++ for Linux and MinGW for Windows.
==========================================================================================
Compiling
==========================================================================================
The simplest way of compiling QRealFourier is to run the following command:
<sh> build/os/os-build-architecture options
where os is your operating system (Linux, Windows or Mac) and architecture is 32bit or
64bit. You can specify any of the following options in any order:
1. configure: configure cmake
2. build: compile QRealFourier
3. install: install the library
4. package: create a DEB RPM or STGZ package (if possible - Linux only)
Linux example: "sh build/linux/linux-build-32bit configure build install package"
Or you can manually build as follows:
With Qt installed in the default location
1. cd code
2. mkdir build
3. cd build
4. cmake ..
5. make
With Qt installed in a custom location
1. cd code
2. mkdir build
3. cd build
4. ccmake ..
5. press 'c', 'c' and then 'g' (change Qt path if needed)
6. make
==========================================================================================
Usage
==========================================================================================
There are three examples under the example directory.
The first one illustrates a forward FFT, followed by an inverse FFT and the necessary
rescaling. The second one adds a Hamming window function to the execution. The last
example illustrates a FFT with a none-default size. Additionally the complex conjugate is
taken in the last example.
You can execute the examples by using the command:
./exampleN
where N indicates the number of the example (1, 2 or 3).
There also is a standalone example that uses CMake to find the installed QRealFourier
library and linking to it.
==========================================================================================
Fixed Sizes
==========================================================================================
QRealFourier is optimized for the following sizes:
8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192 and 16384
The transformation will have a better performance with chunks with one of these sizes.
Any other sizes will work, but pre-calculations have to be done during run-time which
reduces the performance.
==========================================================================================
Real & Imaginary
==========================================================================================
When doing the following:
transformer.forwardTransform(samples, fft);
the fft array will contain the following:
Output array | Values
---------------+----------------------
f [0] | Real (bin 0)
f [...] | Real (bin ...)
f [length/2] | Real (bin length/2)
f [length/2+1] | Imaginary (bin 1)
f [...] | Imaginary (bin ...)
f [length-1] | Imaginary (bin length/2-1)
The static function QFourierTransformer::toComplex(float input[])
can be used to convert this array into a QComplexVector. QComplexVector is a QVector
with QComplexNumbers, each number has a real and imaginary part.
==========================================================================================
The Main Class
==========================================================================================
QRealFourier is the interface that should be used for all transformations.
1. QFourierTransformer(int size = -1, QString functionName = "");
Constructor with optional parameters.
The first parameter specifies the fixed size of the FFT. The default is -1 which will
result in a slower variable-sized FFT.
The second parameter specifies the window function to use. By default a rectangular
window will be used.
2. Initialization setSize(int size);
Changes the size to one of sizes given in the list above. This is the same as
the parameter in the constructor. Can be changed during runtime.
If the size is one of the default fixed sizes, FixedSize will be returned. If the
size is not in the default list, but still a valid FFT size (power of 2),
VariableSize will be returned. If the size is an invalid FFT size, InvalidSize is
returned.
3. bool setWindowFunction(QString functionName);
Sets the window function to be used. If this function is never called, nor a name
passed to the constructor, a rectangular window will be used, which by default will be
NULL, and not have any changes on the input data.
4. QStringList windowFunctions();
Returns a list of strings for all the supported window functions.
5. void forwardTransform(float *input, float *output, QWindower *windower);
Calculates the FFT of the input samples and stores them in output. The output array
must be allocated beforehand and has the same size as the input array.
If a window function is required, you can pass one to the function. The input array
will automatically be adjusted according to the window function. If NULL is passed
no windowing will be done.
6. void inverseTransform(float input[], float output[]);
Calculates the inverse FFT from the input and stores it as real samples in the output
array. The output array must be allocated beforehand with the same size as input.
7. void transform(float input[], float output[], QWindower *windower = NULL,
Direction direction = QFourierTransformer::Forward);
Calls one of the function in 5 or 6 above, depending on the direction parameter.
This function is generally not recommended since it uses an extra if-statement.
Rather use 3 or 4 instead.
8. void rescale(float input[]);
After the inverse FFT was calculated, you have to rescale the values of the samples.
9. void conjugate(float input[]);
Conjugates the imaginary part of the input. Hence the imaginary part's sign is changed
Eg: input[length/2 + 1] = -input[length/2 + 1]
10. QComplexVector toComplex(float input[]);
Converts the FFT calculated with 6 or 7 above to a QVector of complex numbers
as explained in the previous section.
==========================================================================================
Window Functions
==========================================================================================
The following window functions are supported:
1. Rectangular
2. Hamming
3. Hann
==========================================================================================
History
==========================================================================================
*******************
12 June 2013
Version 0.3.0
*******************
QRealFourier now requires Qt5.
Memory leak fixed with a virtual destructor.
*******************
28 May 2012
Version 0.2.0
*******************
Entire restructuring of the library.
Separate running threads removed.
Conjugation support added.
Separate calculations (forward FFT, inverse FFT and rescaling) moved to the same class.
Variable and fixed FFTs now be done with the same functions.
Window functions called via their name in QFourierTransformer.
*******************
22 May 2012
Version 0.1.0
*******************
Initial release.
==========================================================================================
Contact
==========================================================================================
QRealFourier is part of the Visore project:
http://www.visore.org
http://github.com/visore/Visore
http://sourceforge.net/projects/visore
*******************
QRealFourier:
Christoph Stallmann
University of Pretoria
Department of Computer Science
qrealfourier@visore.org
qrealfourier@gmail.com
http://github.com/visore/QRealFourier
http://sourceforge.net/p/qrealfourier
*******************
FFTReal:
Laurent de Soras
laurent.de.soras@free.fr
http://ldesoras.free.fr
visore/QRealFourier
A Qt-based C++ library for Fast Fourier Transform (FFT) of real samples.
C++NOASSERTION