Forums > MaxMSP

fiddle question

February 12, 2010 | 8:34 pm

hi board

the last outlet of miller puckettes fiddle~ outputs sinusoidal peaks and amplitudes… i was wondering if there was any object that did the same thing but at sample rate? i need those numbers to be fast!

also, fiddle~ only seems to think 6 – 8 peaks are important enaugh to output… after that the output is near to zero in amplitude. is there a way to make fiddle~ more sensitive to less dominating peaks?

thank you.



pid
February 12, 2010 | 9:10 pm

try sigmund~ – mr puckette’s own successor to fiddle~ – if memory serves me yada yada, you can choose to output more peaks…


February 12, 2010 | 9:42 pm

I second trying [sigmund~] (it gives more accurate results than [fiddle~].

Also, you can’t get peaks more than once per hop size – say your hopsize is 512 samples, and your sample rate is 44.1kHz – then the speed of the results is about once every 11ms, for which max messages are fine.

Finally, when you say fiddle~ thinks 6-8 peaks are enough, and that the others have a low amplitude it sounds like you’re not identifying an actual problem with fiddle – it can only output the amplitudes as they are in the input signal. The amplitude’s are measured values, and it’s best for them to be accurate – many sounds will not have that many strong peaks, and also, you should expect many peaks to "near to zero" in amplitude – as amplitude is a linear measurement, whereas your perception is logarithmic. If you use [atodb] on the amplitudes you may get more manageable values to use – depends on what you want them for though

A.


February 12, 2010 | 10:36 pm

thanks for the tips

what i want to do is make something that resynthesizes an input signal with an x amount of sinewaves. i want to be able to go from 1 oscillator to maybe 256…

any further thoughts on this?


February 12, 2010 | 10:37 pm

Have you looked at oscillators~ from CNMAT? That and SDIF formatted analysis files may help.


February 12, 2010 | 11:23 pm

The [sigmund~] help patch has an example. You want to use the track functionality of [sigmund~] which tries to track partials between frames, as well as identifying peaks…..

The CNMAT stuff is also worth checking out


February 15, 2010 | 6:37 pm

feh, CNMAT stuff is lame.

Look at the iana~ and transcribe~ objects also.

mz


Viewing 7 posts - 1 through 7 (of 7 total)