polyphonic randomisation

[HughBanton]

I'm still working on my 'Virtual Pipe Organ' generator; coming on well and I promise to put something up as soon as it's presentable

However there's one element I desperately want to add and the method is eluding me.

I'm after a high-speed random generator to apply polyphonically, such that if you are playing, say, a four-note chord, then the pitch (& amplitude) of each of the four notes will be modulated by an independent random stream.

Only a microscopic shift, you understand, to mimic the turbulence that occurs naturally in the wind supply, but it 'unlocks' each note to make it truly independent. (Actually real organ pipes have a genuine tendancy to 'snap' in and out of phase-lock with each other, but that one's for a future enhancement once I've cracked this bit!)

Worth noting that exactly the same thing applies to a quartet of human players - violins, trumpets or whatever - physically impossible for humans to blow as steadily as a fixed digital oscillator (thank goodness) so free-phase vibratos and randomisations are most important for creating naturalness.

I've found several methods of generating a random stream, but not a polyphonic version.

Can anyone point me in the right direction?

[adamszabo]

Write this code into an assembler and it will output a random stream for every voice, between 0 and 1. Or whatever number you set in the brackets.

Code: Select all

streamout out;

float rand[1]=rand(0,1);

movaps xmm0,rand;
movaps out,xmm0;

[martinvicanek]

Well, Adams method will give you a (different) random number for each voice. If you want to have uncorrelated streams of random numbers for each voice, you can use that number as a seed for a (pseudo) random number generator. I have attached a fairly efficient such white noise generator.

[HughBanton]

Many thanks Martin (and Adam), definitely heading in the right direction. With suitable attenuation this addition brings the voices to life.

I wish I knew more about assembler! (I'll re-phrase that .. I wish I knew anything about assembler ) Arduino Man, me.

I get the feeling that I could do with exactly this process but acting much more slowly, so that it produces a 'run-around' the central note frequencies rather than a rapid noise modulation. Would some kind of rate control be easy to implement?

Most grateful for your help. My basic design is finally working pretty well but these subtle modulations are clearly raising the bar.

[martinvicanek]

You could "hop" the processing, i.e. do the iteration only every n samples. Alternatively you could lowpass it. Depends on what exactly you want to use it for.

[HughBanton]

Given that I know nerrthing about assembly, I've empirically arrived at this, which I assume creates a hop 32.

It certainly seems to do the trick for what I want - have I got things in the right order?

I already have a built in octave-by-octave scaling system, so it's relatively easy to set the correct proportions of modulation, these things are always very pitch dependant.

I promise to do some serious studying

Hugh

[martinvicanek]

You are progressing fast, almost there! Try the attached.

[HughBanton]

Excellent, works well. I'm so grateful.

I've found that following it with a 'slew limiter' component, plus some attenuation, perfects the job.

So there's my Teach Yourself Assembler Lesson 1 assignment right there ..
Add the lines of code required to incorporate a variable envelope follower and a variable attenuator