FFT, windowing & hop

Hi,

I' m using an FFT with a Kaiser window and a hop of 50 %. 50 
percent seemed to be theoretically optimal to me.

But recently I read that my idea is wrong and that the optimal 
hop depends on the shape of the window. The goal is that the 
contribution delivered by any individual sample to the final 
result does not depend on the random points of time where the 
window begins resp. ends.

So, how can I set the hop correctly? Are there ready-to-use 
formulas?

-- 
Rolf Keller 

Hi,

in my understanding the contribution delivered by any individual sample is
the same if the hopsize is chosen in a way, so that the overlapping analysis
windows sum to unity.
This is a goal which is especially important for the synthesis of sound
using the overlap-add-method, because otherwise you will produce an audible
amplitude modulation in the synthesized sound.
For analysis only ( to my knowledge ) this criterion is not that important.

I want to point you to this page which also gives some formulas ;-)
http://ccrma.stanford.edu/~jos/parshl/Choice_Hop_Size.html

( It also points out, that there is no "perfect hopsize" for the Kaiser
window ( other than 1 )  )

Karin

"Rolf Keller" <keller-news@t-online.de> schrieb im Newsbeitrag 
news:d5q5lv$edl$04$3@news.t-online.com...
> Hi,
>
> I' m using an FFT with a Kaiser window and a hop of 50 %. 50 percent 
> seemed to be theoretically optimal to me.
>
> But recently I read that my idea is wrong and that the optimal hop depends 
> on the shape of the window. The goal is that the contribution delivered by 
> any individual sample to the final result does not depend on the random 
> points of time where the window begins resp. ends.
>
> So, how can I set the hop correctly? Are there ready-to-use formulas?
>
> -- 
> Rolf Keller