## Curves & math

Jun 18 2006 | 8:35 am

• Jun 18 2006 | 11:47 am
On 18 Jun 2006, at 09:35, f.e wrote:
I'm not sure I understand this: you're dividing all the sample amplitudes by something called "fadeTimeL". Is that what you intend?
What exactly is it that you're trying to achieve? Periodicity? One- shot fades?
-- N.
nick rothwell -- composition, systems, performance -- http:// www.cassiel.com
• Jun 18 2006 | 4:18 pm
Doesn't matter, but for the records : in a "for i=0 to " loop, i'm multiplying (which the part i'm fading) by i/ length>. What i do is a linear fade in (in this case). Perfectly working (great live looping tool). But i wonder what are the ways to do other fades curve. And it's always the time i remember maybe i shouldn't have spent my math classes playing pinball.
cheers
f.e
• Jun 18 2006 | 4:42 pm
On 18 Jun 2006, at 17:18, f.e wrote:
> but for the records : in a "for i=0 to " loop, i'm > multiplying (which the part i'm fading) by i/.
Makes sense (although to be finicky, it should probably be fade length - 1).
> But i wonder what are the ways to do other fades curve.
I tend to do this kind of thing as a normalisation (which you're doing, I guess: fade goes from 1.0 to 0.0) then just pass the value through a function providing the fade curve. (I can't remember offhand what an equal-power curve looks like, for instance - just a bit of sine/cosine as I recall, so map your 0.0..1.0 into 0.0..pi/2 radians.)
-- N.
nick rothwell -- composition, systems, performance -- http:// www.cassiel.com
• Jun 18 2006 | 5:50 pm
That's precisely what you don't remember that i need. If you get a hand on this, don't forget to drop a mail.
cheers
f.e
• Jun 18 2006 | 6:18 pm
maybe the msp examples on spatialization can help?
-thijs
• Jun 18 2006 | 9:28 pm
hi fe, i would suggest to scale your fade values between 0 and 1 (or 1 to 0 for fade out) and raise to a desired exponent. an exponent of 1 would give you a linear fade. an exponent smaller than 1 will give you a faster fade in (slowing down) - useful for crossfades. an exponent bigger than 1 will give you a slower fade in (which is "accelerating") - useful for fade in/out.
for sinusoidal fades scale your values between 0 and pi/2 and take sine or cosine resp. and there are a lot more, i guess. hth, volker.