aamir.nedian wrote:
> i am working on a FFT processor. i need to know that if i need to take the
> 1024 point FFT of a voltage time series then what should i do. i mean
> which one is correct way.
>
> 1) wait for 1024 samples and take their FFT. then wait for another 1024
> samples and take their FFT and so on....
>
> 2) wait for 1024 samples for the first time only and take their FFT. then
> after the computation of this FFT, immediately start taking FFT of the
> currrently available 1024 samples in the input buffer and so on..
>
> please tell me which on is the right way of taking FFT of a voltage time
> series. plz reply early
>
> Regards
> Aamir

The second aproach is more convenient if you want to obtain a good
spectral estimation, but maybe is not necessary to make another FFT for
every new sample. You can search about STFT (Short Time Fourier
Transform) in google because there are lots of techniques depending on
the aplication.

"aamir.nedian" <aamir.nedian@gmail.com> wrote in message 
news:hu2dnXA0_fppUzPZnZ2dnUVZ_oGdnZ2d@giganews.com...
>i am working on a FFT processor. i need to know that if i need to take the
> 1024 point FFT of a voltage time series then what should i do. i mean
> which one is correct way.
>
> 1) wait for 1024 samples and take their FFT. then wait for another 1024
> samples and take their FFT and so on....
>
> 2) wait for 1024 samples for the first time only and take their FFT. then
> after the computation of this FFT, immediately start taking FFT of the
> currrently available 1024 samples in the input buffer and so on..
>
> please tell me which on is the right way of taking FFT of a voltage time
> series. plz reply early
>
> Regards
> Aamir
>


They are both valid approaches and it really depends on the application. The 
second approach (sliding FFT) gives you greater resolution but at a much 
higher processing cost.1022 times. 

aamir.nedian wrote:
> i am working on a FFT processor. i need to know that if i need to take the
> 1024 point FFT of a voltage time series then what should i do. i mean
> which one is correct way.
> 
> 1) wait for 1024 samples and take their FFT. then wait for another 1024
> samples and take their FFT and so on....
> 
> 2) wait for 1024 samples for the first time only and take their FFT. then
> after the computation of this FFT, immediately start taking FFT of the
> currrently available 1024 samples in the input buffer and so on..
> 
> please tell me which on is the right way of taking FFT of a voltage time
> series. plz reply early

It all depends on what is to be done with the result.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

i am working on a FFT processor. i need to know that if i need to take the
1024 point FFT of a voltage time series then what should i do. i mean
which one is correct way.

1) wait for 1024 samples and take their FFT. then wait for another 1024
samples and take their FFT and so on....

2) wait for 1024 samples for the first time only and take their FFT. then
after the computation of this FFT, immediately start taking FFT of the
currrently available 1024 samples in the input buffer and so on..

please tell me which on is the right way of taking FFT of a voltage time
series. plz reply early

Regards
Aamir