FFT related

[d0153030]

I just want to know that does any one of you experts have used FFT analysis in their programming.

I have to do FFT analysis using: one dimensional real data i.e. float.
No. of points is 4096

I have tried on of the implementation taken from NET named as FFTReal. I have to compare the outputr result with output from mathematical programming language out put

Ultimate aim is to display Power Spectrum of the signal on time basis.

Any help and suggestion are highly welcomed and i am looking after.

Sandeep Arya

[sbubis]

Sorry, it is not clear (for me at least) what exactly is your problem?

[Marc G]

Me neither...
... What exactly is your question?

[d0153030]

SOryy for not attaching the file.

Well here is the image file which shows the difference between the FFT function result obtained from mathematical programming lang DAOS and with RealFFT alogo

I hope the fig. illustrates the problem faced.

[ReorX]

Actually, both look the same to me - can anyone else see a problem (except perhaps scaling, but that's a minor issue)?

[Marc G]

I still don't get the problem...
Both graphs look the same to me...

[Tom Frohman]

There are 4096 pts in your fft. Visually it looks like the ratio of the amplitudes is approximately 4096*4096. I've similar scaling differences when I compared power spectra from MATLAB and MTS software. Each FFT routine seems to differ by a multiplicative constant.
Matlab defines the DFT as

X(k) =sum x(n)*exp(-j*2*pi*(k-1)*(n-1)/N) , 1 <= k <= N.


another book I have defines the DFT as

X(k) =(sum x(n)*exp(-j*2*pi*(k-1)*(n-1)/N)) / T , 1 <= k <= N.


where T is the sampling interval

[d0153030]

Thanks Tom, MarcG, ReorX

for putting effort to reply back

Well the problem is clearly visible in the figure which i am enumerating as following:
1.0 Y axis values:
In fft calculation from my routine i am getting values in power of 10pow+3 with magnitude of max value upto 7000 , while for other routine from DAOS values are in the range of 10pow-1 and value of max value about 4*10-3

This is remarkable difference

2.0 On the X axis, it can be viewed that point of max in my routine calculation occurs at position 16th while with DAOS it comes to be about 4.

So these are main reason of my problem.

I had even tried the code from Numerical Reciepe in C, but it too provides similar outputs as FFTReal routine, which i am using.

Kindly provide your kind inputs.

Thanking You

Sandeep Arya