>On 29 Jun, 20:24, "westocl" <cwest...@hotmail.com> wrote:
>> Hello all.
>>
>> Is it feasable to do an fft in 'real time' on different 'blocks' o
data?
>
>You need to look at the times:
>
>- How often do you want / need a result?
>- How long does the computations take?
>- What latency is acceptable?
>
>If you have a sampling rate of 1/T, then the limitig factor
>is how much work your CPU can get done in T seconds. You
>need to get the job done, shuffle the result away and
>be ready to start filling the buffer again by the time
>the next sample is available T seconds after the buffer is
>full.
>
>Rune


I need an output at every 'time' instance. Is there any 'pipelined' typ
FFT algoritms available that you know of?

I am going to have new data at every sample (im assuming i can run th
clock much faster than the sampling rate)... but at any rate, im thinkin
NlogN is still too many calculations unless i have some kind of pipeline
type structure.