Hello,

I'm relatively new to PD, and am enjoying it

I notice that fft examples have been condensed into the audio examples in 0.43..
perhaps somebody can expliain to me how to use 'overlap' in the 'I02.Hann.window.pd' example

I understand that 'block 512' should be changed to 'block 512 4'
What else needs to change? The hanning window doesn't seem to look right?

Thanks .. I'd like to understand about 'windowing' more thoroughly.
Perhaps somebody could suggest a 'site ref' on this topic .. I've been googling for a while, and have still not got to the bottom of it.

Hi,

I would recommend Hann windowing before and after FFT, with overlap 4. This way, you can do the extremest of filtering in frequency domain without artifacts. Of course if you use different FFT size than the example patch you need to calculate a window accordingly. Also it is important to get the normalization right, this is dependent on overlap and window type. It's discussed on this forum more than once, find a normalization demo patch in this thread:

http://puredata.hurleur.com/sujet-5266-fft-overlap-normalization-factor

Katja

@katjav said:

Hi,

I would recommend Hann windowing before and after FFT, with overlap 4
. . .
->http://puredata.hurleur.com/sujet-5266-fft-overlap-normalization-factor<-
Katja

Thanks very much for that! It will take me a while to absorb .. it's very deep
the normalisation is beginning to dawn on me .. a couple of days..

really no one wanted to talk about FFT with me or try to understand more
but thank you for your help katjav your so smart : )

Ah! .. playing around with katjav's normalisation patch and the vanilla examples, I think I see where I was getting confused..

I see now that it is VERY IMPORTANT whether the array/window is in the sub or parent patch .. the same goes for the tabwrite/read etc.

So, putting all the 'hann window objects' in a default 64 patch, rather than in the [block~ 1024 4] one, removes the need for any modification of [tabwrite~ $0-hann]
ie. no need to use cold inlet

Also, any [tabwrite~ spectral-magnitude] must be included on the subpatch in order to use the correct block size etc.

Am I getting it, now?

Thanks again