==============================================================================
FFTReal
Version 2.00, 2005/10/18
Fourier transformation (FFT, IFFT) library specialised for real data
Portable ISO C++
(c) Laurent de Soras <laurent.de.soras@club-internet.fr>
Object Pascal port (c) Frederic Vanmol <frederic@fruityloops.com>
==============================================================================
1. Legal
--------
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Check the file license.txt to get full information about the license.
2. Content
----------
FFTReal is a library to compute Discrete Fourier Transforms (DFT) with the
FFT algorithm (Fast Fourier Transform) on arrays of real numbers. It can
also compute the inverse transform.
You should find in this package a lot of files ; some of them are of interest:
- readme.txt: you are reading it
- FFTReal.h: FFT, length fixed at run-time
- FFTRealFixLen.h: FFT, length fixed at compile-time
- FFTReal.pas: Pascal implementation (working but not up-to-date)
- stopwatch directory
3. Using FFTReal
----------------
Important - if you were using older versions of FFTReal (up to 1.03), some
things have changed. FFTReal is now a template. Therefore use FFTReal<float>
or FFTReal<double> in your code depending on the application datatype. The
flt_t typedef has been removed.
You have two ways to use FFTReal. In the first way, the FFT has its length
fixed at run-time, when the object is instanciated. It means that you have
not to know the length when you write the code. This is the usual way of
proceeding.
3.1 FFTReal - Length fixed at run-time
--------------------------------------
Just instanciate one time a FFTReal object. Specify the data type you want
as template parameter (only floating point: float, double, long double or
custom type). The constructor precompute a lot of things, so it may be a bit
long. The parameter is the number of points used for the next FFTs. It must
be a power of 2:
#include "FFTReal.h"
...
long len = 1024;
...
FFTReal <float> fft_object (len); // 1024-point FFT object constructed.
Then you can use this object to compute as many FFTs and IFFTs as you want.
They will be computed very quickly because a lot of work has been done in the
object construction.
float x [1024];
float f [1024];
...
fft_object.do_fft (f, x); // x (real) --FFT---> f (complex)
...
fft_object.do_ifft (f, x); // f (complex) --IFFT--> x (real)
fft_object.rescale (x); // Post-scaling should be done after FFT+IFFT
...
x [] and f [] are floating point number arrays. x [] is the real number
sequence which we want to compute the FFT. f [] is the result, in the
"frequency" domain. f has the same number of elements as x [], but f []
elements are complex numbers. The routine uses some FFT properties to
optimize memory and to reduce calculations: the transformaton of a real
number sequence is a conjugate complex number sequence: F [k] = F [-k]*.
3.2 FFTRealFixLen - Length fixed at compile-time
------------------------------------------------
This class is significantly faster than the previous one, giving a speed
gain between 50 and 100 %. The template parameter is the base-2 logarithm of
the FFT length. The datatype is float; it can be changed by modifying the
DataType typedef in FFTRealFixLenParam.h. As FFTReal class, it supports
only floating-point types or equivalent.
To instanciate the object, just proceed as below:
#include "FFTRealFixLen.h"
...
FFTRealFixLen <10> fft_object; // 1024-point (2^10) FFT object constructed.
Use is similar as the one of FFTReal.
3.3 Data organisation
---------------------
Mathematically speaking, the formulas below show what does FFTReal:
do_fft() : f(k) = sum (p = 0, N-1, x(p) * exp (+j*2*pi*k*p/N))
do_ifft(): x(k) = sum (p = 0, N-1, f(p) * exp (-j*2*pi*k*p/N))
Where j is the square root of -1. The formulas differ only by the sign of
the exponential. When the sign is positive, the transform is called positive.
Common formulas for Fourier transform are negative for the direct tranform and
positive for the inverse one.
However in these formulas, f is an array of complex numbers and doesn't
correspound exactly to the f[] array taken as function parameter. The
following table shows how the f[] sequence is mapped onto the usable FFT
coefficients (called bins):
FFTReal output | Positive FFT equiv. | Negative FFT equiv.
---------------+-----------------------+-----------------------
f [0] | Real (bin 0) | Real (bin 0)
f [...] | Real (bin ...) | Real (bin ...)
f [length/2] | Real (bin length/2) | Real (bin length/2)
f [length/2+1] | Imag (bin 1) | -Imag (bin 1)
f [...] | Imag (bin ...) | -Imag (bin ...)
f [length-1] | Imag (bin length/2-1) | -Imag (bin length/2-1)
And FFT bins are distributed in f [] as above:
| | Positive FFT | Negative FFT
Bin | Real part | imaginary part | imaginary part
------------+----------------+-----------------+---------------
0 | f [0] | 0 | 0
1 | f [1] | f [length/2+1] | -f [length/2+1]
... | f [...], | f [...] | -f [...]
length/2-1 | f [length/2-1] | f [length-1] | -f [length-1]
length/2 | f [length/2] | 0 | 0
length/2+1 | f [length/2-1] | -f [length-1] | f [length-1]
... | f [...] | -f [...] | f [...]
length-1 | f [1] | -f [length/2+1] | f [length/2+1]
f [] coefficients have the same layout for FFT and IFFT functions. You may
notice that scaling must be done if you want to retrieve x after FFT and IFFT.
Actually, IFFT (FFT (x)) = x * length(x). This is a not a problem because
most of the applications don't care about absolute values. Thus, the operation
requires less calculation. If you want to use the FFT and IFFT to transform a
signal, you have to apply post- (or pre-) processing yourself. Multiplying
or dividing floating point numbers by a power of 2 doesn't generate extra
computation noise.
4. Compilation and testing
--------------------------
Drop the following files into your project or makefile:
Array.*
def.h
DynArray.*
FFTReal*.cpp
FFTReal*.h*
OscSinCos.*
Other files are for testing purpose only, do not include them if you just need
to use the library ; they are not needed to use FFTReal in your own programs.
FFTReal may be compiled in two versions: release and debug. Debug version
has checks that could slow down the code. Define NDEBUG to set the Release
mode. For example, the command line to compile the test bench on GCC would
look like:
Debug mode:
g++ -Wall -o fftreal_debug.exe *.cpp stopwatch/*.cpp
Release mode:
g++ -Wall -o fftreal_release.exe -DNDEBUG -O3 *.cpp stopwatch/*.cpp
It may be tricky to compile the test bench because the speed tests use the
stopwatch sub-library, which is not that cross-platform. If you encounter
any problem that you cannot easily fix whil
Qt-5.15.5+vs2019+32位+windows版
需积分: 19 84 浏览量
2022-09-19
09:10:22
上传
评论
收藏 468.17MB ZIP 举报
Qt-5.15.5+vs2019+32位+windows版 （2000个子文件） 
torrentclient.cpp 53KB xform.cpp 41KB pictureflow.cpp 40KB volumetric.cpp 32KB textedit.cpp 32KB engine.cpp 27KB mainwindow.cpp 27KB main.cpp 26KB colorswatch.cpp 25KB pathstroke.cpp 25KB weatherinfo.cpp 24KB scriptdebugger.cpp 24KB context2d.cpp 23KB gradients.cpp 22KB spreadsheet.cpp 22KB mainwindow.cpp 22KB peerwireclient.cpp 22KB pathdeform.cpp 21KB browserwindow.cpp 21KB dialog.cpp 21KB arthurstyle.cpp 20KB composition.cpp 20KB main.cpp 20KB mainwindow.cpp 19KB environment.cpp 19KB window.cpp 19KB regularexpressiondialog.cpp 19KB glwidget.cpp 18KB player.cpp 18KB mainwindow.cpp 18KB pieview.cpp 18KB xmlconverter.cpp 18KB widgetgallery.cpp 17KB mainwindow.cpp 17KB mainwindow.cpp 17KB appmodel.cpp 17KB mainwidget.cpp 16KB glwindow.cpp 16KB graphmodifier.cpp 16KB main.cpp 16KB classwizard.cpp 16KB mainwindow.cpp 16KB musicplayer.cpp 16KB mainwidget.cpp 16KB mainwindow.cpp 16KB window_multithreaded.cpp 15KB waveform.cpp 15KB pingpong.cpp 15KB themewidget.cpp 15KB widget.cpp 15KB filemanager.cpp 15KB remoteselector.cpp 15KB hoverpoints.cpp 15KB variantdelegate.cpp 14KB addressbook.cpp 14KB mainwindow.cpp 14KB d3d11squircle.cpp 14KB padnavigator.cpp 14KB camera.cpp 14KB mainwindow.cpp 14KB audiorecorder.cpp 14KB xorblender.cpp 14KB imageviewer.cpp 13KB filetree.cpp 13KB customitemgraph.cpp 13KB mainwindow.cpp 13KB main.cpp 13KB addressbook.cpp 13KB mainwindow.cpp 13KB toolbar.cpp 13KB mainwindow.cpp 13KB window_singlethreaded.cpp 13KB httpwindow.cpp 13KB main.cpp 13KB audioinput.cpp 12KB cborconverter.cpp 12KB arthurwidgets.cpp 12KB document.cpp 12KB mainwindow.cpp 12KB qtlogo.cpp 12KB device.cpp 12KB licensewizard.cpp 12KB browser.cpp 12KB tabletcanvas.cpp 12KB calculator.cpp 12KB norwegianwoodstyle.cpp 12KB main.cpp 12KB mainwindow.cpp 12KB d3d12renderer.cpp 12KB audiodevices.cpp 12KB histogramwidget.cpp 12KB window.cpp 12KB states.cpp 12KB glwidget.cpp 11KB hellowindow.cpp 11KB mainwindow.cpp 11KB devicehandler.cpp 11KB window.cpp 11KB main.cpp 11KB encodingdialog.cpp 11KB共 2000 条
- 1
- 2
- 3
- 4
- 5
- 6
- 20
评论0