dibman

Hi, first post here so apologies if this is not in the right forum. I've done quite bait of searching and not managed to find the answer anywhere. I'm building an MSP external to be used within a pfft~ subpatch, and want to have no overlap. The smallest hop size seems to be windowSize/32. Does anyone know if this can be reduced further or a suitable workaround? I've looked into using fft~ but the project is well underway and works so would like to continue with pfft~ if possible!

using-pfft-with-no-overlap

If you want no overlap, your hopsize should be the same as the frame size.

Thank you for information. Is the frame size required to be a multiple of the audio driver's sample block size?

Hi Andrea - thanks. This would mean using an overlap factor of 1. Ernest - the frame size (fft size) must be a power of 2.

I think andrea is right and an overlap factor of 1 equals no overlap, how else could you have a hopsize the size of the frame size (try to picture it),

if you have an overlap of 2 then 50% of the samples are re-analysed next time (and calcluation happens every fftsize/2)

if you have an overlap of 4 then 75% of the samples are re-analysed next time (and calcluation happens every fftsize/4)

if you have an overlap of 8 then 87.5% of the samples are re-analysed next time (and calcluation happens every fftsize/8)

if you have an overlap of 1 then 0% of the samples are re-analysed next time (and calcluation happens every fftsize)

I think andrea is right and an overlap factor of 1 equals no overlap, how else could you have a hopsize the size of the frame size (try to picture it), 

if you have an overlap of 2 then 50% of the samples are re-analysed next time (and calcluation happens every fftsize/2)
if you have an overlap of 4 then 75% of the samples are re-analysed next time (and calcluation happens every fftsize/4)
if you have an overlap of 8 then 87.5% of the samples are re-analysed next time (and calcluation happens every fftsize/8)

so
if you have an overlap of 1 then 0% of the samples are re-analysed next time (and calcluation happens every fftsize)


It is true what andrea and Timo are saying. But then again - why would you want to have no overlap? Why the need for FFT data in such an useless form? Just curious...

And one more thing you probably need to know that is not written in any documentation - the sample rate inside pfft~ is overlap times bigger as defined in your MSP status due to the fact, that there is overlap times more information to process.

I never did an external in my life but probably such an information is useful...?

And one more thing you probably need to know that is not written in any documentation - the sample rate inside pfft~ is overlap times bigger as defined in your MSP status due to the fact, that there is overlap times more information to process. 

I never did an external in my life but probably such an information is useful...?


Using pfft~ with no overlap