extracting individual frequencies

Aug 3, 2008 at 4:00pm

extracting individual frequencies

Hi,

I’m looking to extract various different frequencies from a signal, process them independently, then put them back together. What should I look into for doing this?

Also, is there an object out there like meter~ that shows levels at different frequency bands?

Jay

#39115
Aug 3, 2008 at 4:09pm

#137287
Aug 3, 2008 at 4:20pm

>> I’m looking to extract various different frequencies from a signal,
>> process them independently, then put them back together. What
>> should I look into for doing this?

Something like FFT tutorials?

J-F.

#137288
Aug 3, 2008 at 4:48pm

in the pfft~ tutorials there are examples for processing certain bins which will do what you want with some modification.

#137289
Aug 3, 2008 at 5:09pm

If you only want to use the frequencies with greatest amplitude, you could
also try strategies that use, for example, Tristan Jehan’s analyzer~ to
determine the frequencies available and an array of reson~ objects or other
filters to isolate the bands around the selected frequencies. You can
process the bands and then remix with the source after that. If you want to
emphasize the processed content, notch the source at the same frequencies as
the extracted bands.

– Paul

On Sun, Aug 3, 2008 at 11:48 AM, fairesigneaumachiniste
wrote:

>
> in the pfft~ tutorials there are examples for processing certain bins which
> will do what you want with some modification.
>


—– |(*,+,#,=)(#,=,*,+)(=,#,+,*)(+,*,=,#)| —–

#137290
Aug 3, 2008 at 5:49pm

I have been able to get through all of the max/msp tutorials just fine, with the exception of fft~ and pfft~. What is it that these can offer that Ignotus’s reson~ approach cannot?

Also, does anyone know of a good ‘beginner’s guide to fft’ out there? I’d really like to understand this stuff.

#137291
Aug 3, 2008 at 6:24pm

#137292
Aug 3, 2008 at 7:56pm

Roughly, fft~ and pfft~ give you access to all of the frequencies in a
signal. There are limits due to sampling frequency and windowing, but all
the frequency-domain information needed to recreate the time-domain signal
is present in the transform. Analyzer~ just gives you its best approximation
of what the principal frequencies are in a signal. With some signals it can
be very good (especially with pitched sounds, as you might guess), with
others it is something of a guessing game. It will not give you enough
information to recreate the signal, but can be used for driving certain
kinds of resynthesis that don’t necessarily reproduce a sound so much as
approximate it or imitate certain qualities. Analyzer~ may not be your best
tool for real time synthesis, though some of the related objects (Miller
Puckette’s fiddle~) can work well in real time (or so I recall).

There are all kinds of marvelous things you can do with Fourier Transforms,
such as convolving two sounds together to create new ones, time-stretching,
frequency-shifting, phase vocoders, etc. Which strategy you use depends what
you want to do, how computationally intensive your patches can be, whether
you need to run in real time, etc.

analyzer~ and other objects by Tristan Jehan
http://maxobjects.com/?v=authors&id_auteur=4&requested=analyze&operateur=AND&id_plateforme=0&id_format=0
port of fiddle~, a few others

http://maxobjects.com/?request=fiddle%7E&operateur=AND&id_format=0&id_plateforme=0&Submit=OK

HTH,

– Paul

On Sun, Aug 3, 2008 at 12:49 PM, Jay Bodley wrote:

>
> I have been able to get through all of the max/msp tutorials just fine,
> with the exception of fft~ and pfft~. What is it that these can offer that
> Ignotus’s reson~ approach cannot?
>
> Also, does anyone know of a good ‘beginner’s guide to fft’ out there? I’d
> really like to understand this stuff.
>


—– |(*,+,#,=)(#,=,*,+)(=,#,+,*)(+,*,=,#)| —–

#137293
Aug 4, 2008 at 5:45am

#137294

You must be logged in to reply to this topic.