fft~ question / Hi there forum

Oct 26, 2006 at 10:43pm

fft~ question / Hi there forum

Hey there forum,

I’ve been using max/msp for a few years now and I’ve become quite attached to it. my favorite object(s) is the tapin~/tapout~ pair. I got into max from hearing blectum from blechdom and kit clayton on Brave New Waves (CBC radio). And I make some of my own stuff with analog synths + Max/MSP.

After a long time avoiding it, I’ve started to get into fft processing. It took me a while but I got through the tutorial understanding most of it. I even learned what the heck an imaginary number is in the process (though this gave me a headache trying to figure out why the second number is imaginary).

One question i have about phase vocoding is why is it necessary to convert the cartesian coordinates to polar to record them into the buffer… and also why the phase has to be run through framedelta~. In my head the framedelta~ processing would just cause the phase to change incorrectly if frameaccum~ is processing the same frame multiple times.

(in my head framedelta is outputing how much the phase of a specific bin changes and frameaccum is adding them up. So if frame accum is constantly getting +.3 for a bin when repeately indexing the same frame won’t it be outputing the incorrect phase?)

So yeah… why can’t you just record the x,y and play a specific frame back for phasevocoding?

Also are there any good books/tutorials/webpages out there that might help me figure out a bit more about fft processing?

Thanks in advance,

Joel

#28383
Oct 27, 2006 at 11:06am

#87051
Oct 27, 2006 at 1:35pm

#87052
Oct 27, 2006 at 1:58pm

#87053
Oct 27, 2006 at 7:11pm

Ahh… that explains it! I’d never considered that you might want the phase for a bin to be constantly moving. That also explains why you would normally convert to polar. I’m glad to hear you don’t actually have to if you want to save cpu power.

Thanks, this helps a ton. I’ll check out the literature you guys provided.

Joel

#87054
Oct 29, 2006 at 6:26am

Hey there Joel,
Cudos for the ballsy move – not an easy one to tackle.
While not being anywhere near fully grasping the subject, I can
recommend this link:

http://www.parasitaere-kapazitaeten.net/Pd/fft_und_pd_en.htm

And of course go through Tutorials 25 & 26…

Cheers,
Zax

—–Original Message—–
From: maxmsp-bounces@cycling74.com [mailto:maxmsp-bounces@cycling74.com]
On Behalf Of Joel Rich
Sent: Friday, October 27, 2006 12:44 AM
Subject: [maxmsp] fft~ question / Hi there forum

Hey there forum,

I’ve been using max/msp for a few years now and I’ve become quite
attached to it. my favorite object(s) is the tapin~/tapout~ pair. I got
into max from hearing blectum from blechdom and kit clayton on Brave New
Waves (CBC radio). And I make some of my own stuff with analog synths +
Max/MSP.

After a long time avoiding it, I’ve started to get into fft processing.
It took me a while but I got through the tutorial understanding most of
it. I even learned what the heck an imaginary number is in the process
(though this gave me a headache trying to figure out why the second
number is imaginary).

One question i have about phase vocoding is why is it necessary to
convert the cartesian coordinates to polar to record them into the
buffer… and also why the phase has to be run through framedelta~. In
my head the framedelta~ processing would just cause the phase to change
incorrectly if frameaccum~ is processing the same frame multiple times.

(in my head framedelta is outputing how much the phase of a specific bin
changes and frameaccum is adding them up. So if frame accum is
constantly getting +.3 for a bin when repeately indexing the same frame
won’t it be outputing the incorrect phase?)

So yeah… why can’t you just record the x,y and play a specific frame
back for phasevocoding?

Also are there any good books/tutorials/webpages out there that might
help me figure out a bit more about fft processing?

Thanks in advance,

Joel

#87055
Oct 30, 2006 at 7:29am

#87056

You must be logged in to reply to this topic.