FFT Convolution


D
Nov 24, 2006 at 10:41pm

FFT Convolution

Hi, Im relatively new to max and I was looking through the MSP tutorial 26. I know a little about the effects convolution can have but when putting two signals through this it is very noisy and not very pleasing to the ear. I was wondering if they are any techniques i can use to improve my results. I have considered normalising or silencing any bins with amplitudes below a cetrain threshold

Any advice on this would greatly be appreciated.

Regards

AD

#28873
Nov 26, 2006 at 1:49am

> Hi, Im relatively new to max and I was looking through the MSP tutorial 26. I know a little about the effects convolution can have but when putting two signals through this it is very noisy and not very pleasing to the ear. I was wondering if they are any techniques i can use to improve my results. I have considered normalising or silencing any bins with amplitudes below a cetrain threshold
>
> Any advice on this would greatly be appreciated.
>
> Regards
>
> AD
—————————————————-

Hi AD-

Obviously, the output of convolution has a lot to do with the inputs. I try to use one “noisy” sound and one no so noisy.

You can make a quick and dirty noise gate like this:

#P window setfont “Sans Serif” 9.;
#P window linecount 1;
#P newex 480 182 54 196617 deltaclip~;
#P newex 479 160 49 196617 vectral~;
#P newex 310 178 27 196617 *~;
#P user ezdac~ 400 64 444 97 0;
#P newex 310 81 52 196617 cycle~ 2.;
#P user scope~ 310 241 440 371 256 3 128 -1. 1. 0 0. 0 0. 102 255 51 135 135 135 0;
#P user scope~ 160 110 290 240 256 3 128 -1. 1. 0 0. 0 0. 102 255 51 135 135 135 0;
#P newex 327 150 33 196617 >~ 0.5;
#P comment 471 132 100 196617 see also:;
#P connect 4 0 2 0;
#P connect 4 0 6 0;
#P connect 4 0 1 0;
#P connect 6 0 3 0;
#P connect 1 0 6 1;
#P window clipboard copycount 9;

But, it probably won’t sound very good. (I think there’s a similar example in the tutorial).

You should look at the vectral~ object for a more sophisticated behavior.

mz

#89035
Nov 26, 2006 at 6:45pm

>> Obviously, the output of convolution has a lot to do with
>> the inputs. I try to use one “noisy” sound and one no so
>> noisy.

the problem is always that
tighter (no overlap) or of a nicer sound (higher overlap).

so you can try things like combining different settings of
the fft stuff and mix them against each other, for example
accoring to the power of the carrier input.

#89036
Nov 26, 2006 at 7:11pm

While I’m not sure that I’d contrast ‘tighter’ with ‘nicer
sound’ (that’s a pretty subjective way of characterizing a fairly
complex set of trade-offs) it is true that there is a number of
choices to be made in selecting parameter settings based on input
material and desired output. In general, the most important choices
come down to time resolution vs. frequency resolution (number of fft
frequency bins), windowing type/shape, window size (length), and
window overlap. There’s no easy way to state the various trade-offs
in terms of ‘how to choose settings’ for general use, but a decent
discussion of the overall situation can be found in such books as
Curtis Roads’ ‘Computer Music Tutorial’ (MIT Press), and Dodge &
Jerse’s ‘Computer Music’ (Schirmer Boooks), among other sources.

On Nov 26, 2006, at 11:45 AM, Roman Thilenius wrote:

>
>
>>> Obviously, the output of convolution has a lot to do with
>>> the inputs. I try to use one “noisy” sound and one no so
>>> noisy.
>
>
> the problem is always that
> tighter (no overlap) or of a nicer sound (higher overlap).
>
> so you can try things like combining different settings of
> the fft stuff and mix them against each other, for example
> accoring to the power of the carrier input.
>

—-
Steven M. Miller

Associate Professor of Contemporary Music
College of Santa Fe
Contemporary Music Program
1600 St. Michaels Drive
Santa Fe NM 87505
http://pubweb.csf.edu/~smill
(505) 473-6197

SFIFEM Atrium Sound Space OVOS
http://sfifem.csf.edu http://atrium.csf.edu http://pubweb.csf.edu/
~smill/ovos.html

#89037
Nov 27, 2006 at 1:00pm

> In general, the most important choices
> come down to time resolution vs. frequency resolution (number of fft
> frequency bins), windowing type/shape, window size (length), and
> window overlap. There’s no easy way to state the various trade-offs
> in terms of ‘how to choose settings’ for general use

not for general use, what i am suggesting is to offer
flexible settings as user parameters, not finding the
perfect settings.

#89038

D
Nov 27, 2006 at 4:37pm

thanks for suggestions, the input is appreciated

AD

#89039
Nov 27, 2006 at 5:25pm

I’m not sure that I follow your reply, but as far as I do, it seems
like we’re saying the same thing here. My point is simply that
‘perfect settings’ (or some close approximation) are probably what
most folks would like to find for the work they are doing, and these
will vary on a case-by-case basis – again depending on the
characteristics of the input file and the desired output.

On Nov 27, 2006, at 6:00 AM, Roman Thilenius wrote:

>
>> In general, the most important choices
>> come down to time resolution vs. frequency resolution (number of fft
>> frequency bins), windowing type/shape, window size (length), and
>> window overlap. There’s no easy way to state the various trade-offs
>> in terms of ‘how to choose settings’ for general use
>
>
> not for general use, what i am suggesting is to offer
> flexible settings as user parameters, not finding the
> perfect settings.

—-
Steven M. Miller

Associate Professor of Contemporary Music
College of Santa Fe
Contemporary Music Program
1600 St. Michaels Drive
Santa Fe NM 87505
http://pubweb.csf.edu/~smill
(505) 473-6197

SFIFEM Atrium Sound Space OVOS
http://sfifem.csf.edu http://atrium.csf.edu http://pubweb.csf.edu/
~smill/ovos.html

#89040
Nov 27, 2006 at 6:28pm

Quote: smill wrote on Mon, 27 November 2006 10:25
—————————————————-
> I’m not sure that I follow your reply, but as far as I do, it seems
> like we’re saying the same thing here. My point is simply that
> ‘perfect settings’ (or some close approximation) are probably what
> most folks would like to find for the work they are doing, and these
> will vary on a case-by-case basis – again depending on the
> characteristics of the input file and the desired output.

those are the _presets in this scenario.

one for “snaredrum” and one for “atmopshere”

:)

#89041

You must be logged in to reply to this topic.