MATLAB Central Newsreader - FFT

FFT

guj — Wed, 01 Jul 2009 06:02:02 -0400

I was doing an experiment, and find that FFT and NDFT are giving little different results when i implement it on uniform sampled sin signal with 101 samples.

Although results are same. Only difference is the some noise near the spike. Noise is zero in case of NDFT but little noise can be seen in FFT Should i expect the same results from both of them

Any comments will be appreciated. I have some images for what i have done, if some body interested in seeing. Drop me a mail at pinkfloydindia@yahoo.com

Re: FFT

Greg — Sat, 04 Jul 2009 19:10:26 -0400

On Jul 1, 2:02 am, "guj " <gulatiaks...@gmail.com> wrote:
> I was doing an experiment, and find thatFFTand NDFT are giving little different results when i implement it on uniform sampled sin signal with 101 samples.
>
> Although results are same. Only difference is the some noise near the spike. Noise is zero in case of NDFT but little noise can be seen inFFT Should i expect the same results from both of them

Since floating point calculations are used, you will have to
quantify
your statement. What is the normalized mean-squared error

mean(abs(X2-X1).^2)/mean(abs(X1-mean(X1)))

Hope this helps.

Greg

> Any comments will be appreciated. I have some images for what i have done, if some body interested in seeing. Drop me a mail at pinkfloydin...@yahoo.com

If your error is sisnificant and I have some time, I will look at it.

Re: FFT

guj — Mon, 06 Jul 2009 15:58:01 -0400

Can any one tell me difference between these notations

Matlab DFT
for K=1:N
for n=1:N
x(w)=sum( n=0 n=N-1) x(n) exp ( 2 pi (k-1) (n-1) /N)

NDFT
for k=1:N
x(w)=sum( n=0 n=N-1) x(n) exp ( 2 pi freq(k) time(n))

So here is my question, i was trying to derive the DFT notation of matlab, which is given in help..in that derivation i need to assume that its is equally spaced or my twiddle factor vectors are equally spaced. so DFT notation and NDFT notatin should have some different, also both give different results . DFT Notation of matlab give same result as FFT but NDFT notation give different results

Greg <heath@alumni.brown.edu> wrote in message <496ce8c4-8e49-4b17-9d27-efff652e77ea@b14g2000yqd.googlegroups.com>...
> On Jul 1, 2:02?am, "guj " <gulatiaks...@gmail.com> wrote:
> > I was doing an experiment, and find thatFFTand NDFT are giving little different results when i implement it on uniform sampled sin signal with 101 samples.
> >
> > Although results are same. Only difference is the some noise near the spike. Noise is zero in case of NDFT but little noise can be seen inFFT?Should i expect the same results from both of them
>
> Since floating point calculations are used, you will have to
> quantify
> your statement. What is the normalized mean-squared error
>
> mean(abs(X2-X1).^2)/mean(abs(X1-mean(X1)))
>
> Hope this helps.
>
> Greg
>
> > Any comments will be appreciated. I have some images for what i have done, if some body interested in seeing. Drop me a ?mail at pinkfloydin...@yahoo.com
>
> If your error is sisnificant and I have some time, I will look at it.

Re: FFT

Greg — Tue, 07 Jul 2009 05:09:10 -0400

On Jul 6, 11:58 am, "guj " <gulatiaks...@gmail.com> wrote:
> Can any one tell me difference between these notations
>
> Matlab DFT
> for K=1:N
> for n=1:N
> x(w)=sum( n=0 n=N-1) x(n) exp ( 2 pi (k-1) (n-1) /N)

You need to proof read your post. Taken literally,
what you have written is garbage.

Try again.

> NDFT
> for k=1:N
> x(w)=sum( n=0 n=N-1) x(n) exp ( 2 pi freq(k) time(n))

If you do things correctly you should get EXACTLY the same result.
when freq and time are uniformly spaced.

> So here is my question, i was trying to derive the DFT notation of matlab, which is given in help..in that derivation i need to assume that its is equally spaced or my twiddle factor vectors are equally spaced. so DFT notation and NDFT notatin should have some different, also both give different results . DFT Notation of matlab give same result as FFT but NDFT notation give different results

It shouldn't you made a mistake some where.

See my last post in "FFT,IFFT, and NDFT,NFFT "

Hope this helps.

Greg

-----SNIP