| Author |
Message |
Eugene
Guest
|
 Posted:
Thu Dec 09, 2004 7:25 pm Post subject:
FFT build with DSP |
|
|
thank you for your answer,
I want to say, that I have a sine Signal. I can show it with "compose
studio"
It's a sine signal.
Mike, you have right, that if I build a complex FFT, that I should
have two peaks of spectrum, but this peaks are symmetric.
On the site from mathworks.de I found that I should cut my cpectrum,
so that I have only half from it.
I've done it.
so, now should I have only one peak of FFT spectrum.
My problem is, that I don't know, whether it is calculated in complex
or real numbers.
i use the dsplib function rfft() and cfft()
I assume, that the cfft() work with complex number, but I have only
one array, how can differentiate wich numbers are real and wich imag.
?
thanks in advance |
|
| Back to top |
|
 |
Stephan M. Bernsee
Guest
|
| Posted:
Thu Dec 09, 2004 7:38 pm Post subject:
Re: FFT build with DSP |
|
|
On 2004-12-09 15:25:14 +0100, johnbesel@web.de (Eugene) said:
| Quote: | It's a sine signal. Mike, you have right, that if I build a complex
FFT, that I should
have two peaks of spectrum, but this peaks are symmetric.
|
Stephan, not Mike. Nevermind.
| Quote: | On the site from mathworks.de I found that I should cut my cpectrum,
so that I have only half from it.
|
A real signal has a symmetric Fourier transform so, yes, you can ignore
the negative frequencies.
| Quote: | i use the dsplib function rfft() and cfft() I assume, that the cfft()
work with complex number, but I have only
one array, how can differentiate wich numbers are real and wich imag.
?
|
I would guess from the names that rfft() calculates the real-input
Fourier transform, so I would go with that. How complex numbers are
represented depends on the particular complex fft(), usually they are
stored in adjacent elements in an array, ie. as data[2*index]=re;
data[2*index+1]=im; etc.
Have a look at my "DFT à Pied" article
(http://www.dspdimension.com/html/dftapied.html) to get started with
the Fourier transform in general, or post here for more info.
--
Stephan M. Bernsee
http://www.dspdimension.com |
|
| Back to top |
|
|
Shawn Steenhagen
Guest
|
| Posted:
Fri Dec 10, 2004 12:04 am Post subject:
Re: FFT build with DSP |
|
|
The TI document SPRA480 will tell you all you need to know about using
rfft( ), for example: rfft(xin,512,0), including the format of the fft
output and special requirements about the size of xin and how it must be
align in memory to get the proper result.
I didn't see the original post, but I would make sure you are using the
rfft( ) function properly and interpreting its results properly.
The output of the rfft( ) is a complex result in the form:
y(0)Re y(nx/2)Im (DC and Nyquist)
y(1)Re y(1)Im
y(2)Re y(2)Im
etc.
-Shawn
"Eugene" <johnbesel@web.de> wrote in message
news:6484110f.0412090625.f07a211@posting.google.com...
| Quote: | thank you for your answer,
I want to say, that I have a sine Signal. I can show it with "compose
studio"
It's a sine signal.
Mike, you have right, that if I build a complex FFT, that I should
have two peaks of spectrum, but this peaks are symmetric.
On the site from mathworks.de I found that I should cut my cpectrum,
so that I have only half from it.
I've done it.
so, now should I have only one peak of FFT spectrum.
My problem is, that I don't know, whether it is calculated in complex
or real numbers.
i use the dsplib function rfft() and cfft()
I assume, that the cfft() work with complex number, but I have only
one array, how can differentiate wich numbers are real and wich imag.
?
thanks in advance |
|
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in