segeke

I want to develop the patch of fft base Reverb with Max/MSP.

Does not it rely on the third party object and does it know the development method in the basic external object of Max/MSP?

They might use the following fft algorithms.

However, these cannot calculate more than the size of the FFT frame set with pfft~ object...

They will have to use the equal segmented fft convolution method.

I want to develop the patch of fft base Reverb with Max/MSP.
Does not it rely on the third party object and does it know the development method in the basic external object of Max/MSP?

They might use the following fft algorithms.
However, these cannot calculate more than the size of the FFT frame set with pfft~ object...
They will have to use the equal segmented fft convolution method.

equal-segmented-fft-convolution

From the look of your patch, you probably wont have alot of luck with this kind of convolution when it comes to creating a reverb. You need something more of an impulse system, ala Peak/Soundhack etc.

I'd have a check of the forum, theres alot of people out there who've tried this (me included). Theres one guy in particular who wrote a few externals to come up with a block convolution method which really interested me.

Hey,
From the look of your patch, you probably wont have alot of luck with this kind of convolution when it comes to creating a reverb. You need something more of an impulse system, ala Peak/Soundhack etc.

FWIW, thomas resch, one of our students here at esb, has written a 

anybody interested in it should be able to find it at 

FWIW, thomas resch, one of our students here at esb, has written a  
convolution external (osx only).
anybody interested in it should be able to find it at  
www.zippernoise.net
volker.

I am using his external now. I want to study those mechanisms with the patch of pure Max/MSP without relying on the third party external.

volker.
I am using his external now. I want to study those mechanisms with the patch of pure Max/MSP without relying on the third party external.

Your GrainSynth patch is wonderful!
 I newly learnt many from you.
Thank you!

> I am using his external now. I want to study those mechanisms with 

> the patch of pure Max/MSP without relying on the third party external.

the complex multiplication in the patch you sent the other day was 

correct, but in order to avoid circular convolution you will have to 

zero-pad your input data (as far as i could see, you just used a 

square window). e.g. create a window with a value of 1 up to the 

middle of the window and put the rest to 0. then you can have an 

impulse response that is half the fft size.

of course you need two processes overlapping each other.

there are some ideas how to deal with that on 

it's not hard to do with the old fft~ object (limited to 4096), but 

the pfft~ based fourier transform does some really funky stuff which 

after all, i don't think it's worth spending too much time on it. an 

msp-based convolution reverb will probably be not very effective in 

aha, while it's perfectly possible that i did such a patch in the 

past, i'm not quite sure which patch you are talking about - maybe 

> I am using his external now. I want to study those mechanisms with  
> the patch of pure Max/MSP without relying on the third party external.

the complex multiplication in the patch you sent the other day was  
correct, but in order to avoid circular convolution you will have to  
zero-pad your input data (as far as i could see, you just used a  
square window). e.g. create a window with a value of 1 up to the  
middle of the window and put the rest to 0. then you can have an  
impulse response that is half the fft size.
of course you need two processes overlapping each other.
there are some ideas how to deal with that on http://www.pescadoo.net/ 
annexe/max/index.html

it's not hard to do with the old fft~ object (limited to 4096), but  
the pfft~ based fourier transform does some really funky stuff which  
i couldn't figure out how to deal with.
after all, i don't think it's worth spending too much time on it. an  
msp-based convolution reverb will probably be not very effective in  
the end.

>
> though the topic is different..
>
> Your GrainSynth patch is wonderful!
>  I newly learnt many from you.
> Thank you!

aha, while it's perfectly possible that i did such a patch in the  
past, i'm not quite sure which patch you are talking about - maybe  
it's by someone else...(blush!)
well, anyway, have fun.
volker.

the convolution part of your patch looks right to me. however the 

tricky part is, to get the two pfft~ to synchronize correctly. this 

is easily done with [fft~] but i've never managed to get it working 

maybe jean-yves bernier, the original author of the patches at 

www.pescadoo.net/ has done some further testing.

the convolution part of your patch looks right to me. however the  
tricky part is, to get the two pfft~ to synchronize correctly. this  
is easily done with [fft~] but i've never managed to get it working  
properly with [pfft~].
maybe jean-yves bernier, the original author of the patches at http:// 
www.pescadoo.net/  has done some further testing.
vb


I made them based on information on this site. Two or more pfft ? of a short FFT size might have to be used to evade latency.

[pfft~subpach.fft 256 0 1] . [pfft~subpach.fft 512 0 1] . [pfft~subpach.fft 1024 0 1] . [pfft~subpach.fft 2048 0 1]....

It delays it by delay~ before all signals are added.

I made them based on information on this site. Two or more pfft ? of a short FFT size might have to be used to evade latency.
For instance, 