Forums > MaxMSP

convolution reverb

May 4, 2009 | 12:37 am

hi list,

after stumbling across this really interesting thread from 2 years ago

http://www.cycling74.com/forums/index.php?t=msg&goto=61070

and pescadoo’s site http://www.pescadoo.net/annexe/max/index.html

i had a go at building the convolution reverb he outlines. it kinda works but the output jumps around like the each block buffer is wrapped round and offset slightly

when i built the proof of concept patch i noticed that hanning window in the output is not allways offset by the same amount as the dirac pulse but rather it varies every time i whack loadbang. i know this has something to do with the vector size, when i set it to the same as the signal size i seem more likely to get the hanning window in the right place.

can anyone shed any light on this ?

– Pasted Max Patch, click to expand. –

pconvblock-forum.maxpat

– Pasted Max Patch, click to expand. –

May 10, 2009 | 6:22 am
crx091081gb wrote on Mon, 04 May 2009 02:37
after stumbling across this really interesting thread from 2 years ago

http://www.cycling74.com/forums/index.php?t=msg&goto=61070

and pescadoo’s site http://www.pescadoo.net/annexe/max/index.html

i had a go at building the convolution reverb he outlines. it kinda works but the output jumps around like the each block buffer is wrapped round and offset slightly

Unfortunately, he probably just "outlined" as much as he understood, and lost interest when it didn’t work. And put all his beginners attempts on his site to confuse others…
The patch you posted doesn’t make much sense, you create windowing outside of pfft~, but it would be easier inside. (thats what pfft~ is made for in the first place) Inside everything would be synchronized anyway.

There is the buffir~ external which does convolution up to 256 samples (covers a bit more than 5 ms). There is a reason why its not more. It is not a beginners task to reinvent a convolution reverb, as I said already 2 years ago…


May 10, 2009 | 8:42 pm

thanks for all your help, if I could get the offset problem sorted it would actally work, it’s so close it seems a shame to stop trying, I’ll trying doing more of it inside pfft.

I know about buffir but 5 ms is not long enough to build interesting things out of.


May 10, 2009 | 9:34 pm

might want to google convolution max msp ..

just a shot in the dark.


May 10, 2009 | 10:04 pm

the only workable solution of IR in Max/MSP I ever found was Thomas Resch’s tconvolution~

http://www.zippernoise.net/data/index.php?option=com_content&task=view&id=22&Itemid=59

(mac only)

haven’t checked it for a while, but I remember it quite cpu hungry, but I got some interesting results from it.

S


Viewing 5 posts - 1 through 5 (of 5 total)