[sdiy] Multiple oscillator detuning

[rburnett at richieburnett.co.uk]

Hi Tom and all,

>>> For a start, I could either use linear detuning (Hertz)...

This tends to give a nice even chorusing effect over high and mid pitch 
ranges but progressively sounds more and more out of tune as you decend 
into the bass register.  This is because a fixed Hz offset becomes a 
larger proportion of the fundemantal frequency for lower notes and the 
tuning error in cents increases rapidly as you go down in pitch!  (PWM 
of a single pulse oscillator by triangle LFO is equivalent to two 
sawtooth oscillators detuned by a fixed amount in Hz.  Anyone who's 
played with PWM on a single VCO monosynth knows that LFO speeds which 
sounds lush for midrange and high notes can make bass notes sound 
horribly out of tune.)

> or exponential (cents).

This is the natural choice since we perceive pitch exponentially in Hz. 
Although a fixed detuning in cents can cause some pretty high-speed 
beating at the high end of the keyboard that can quickly start to sound 
out of tune :-(

The best choice might depend on what pitches you intend to play, or 
might consist of a combination of Hz and cents detuning methods?!?!

>>> That adds another query - how much detune is reasonable? I suppose 
>>> various "rave hoover" sounds suggest the answer to that one is "more 
>>> than you think is reasonable!"

You can get away with a lot more detuning than you might think provided 
the spectrum is "filled out".  What I mean by this is, provided the 
space between the two most extremely detuned oscillators is filled with 
other oscillators it doesn't sound so out of tune.  Two individual 
oscillators seperated by 50 cents detuning sounds very out of tune.  But 
10 oscillators spread over a wider range of 75 cents sounds thick and 
menacing.

Notice in that paper how the detune control law is not linear.  You get 
lots of fine control over the slight detunings which are more muscially 
useful, and then it kicks in with more severe detuning in the last part 
of the potentiometer's range for your hoover leads ;-))  (Why the author 
tried to fit a 10th order polynomial to data that's clearly piecewise 
linear is beyond me though!)

It's probably also a good idea to randomise the amplitudes of all the 
oscillators slightly too, so that you don't get really deep combing when 
any two sawtooths pass each other in phase.

-Richie,