[sdiy] A potential problem with a voice-assignable organ

[Tom Wiltshire]

Hi All,

I've been thinking about a digital model of the Vox Continental  
(hence earlier questions). This is purely because it amuses me; I  
don't really think the world needs more digital organ emulations.  
Anyway, Clavia nailed it for me.

However...!

One way to do such an instrument would be to have individual voice  
processors that are sent pitch information by the voice assigner.  
This is pretty standard stuff.
The voice processors would then generate a virtual master pitch using  
an NCO, and then send it through virtual dividers to generate the  
different footages, and finally through one-tap IIR filters to  
simulate the RC response of the original's analog filters. This part  
is all in Hal Chamberlin's book, pg 485. As he says about putting  
virtual square waves through the digital filter; "In fact, the  
samples of the square wave response are *exactly* the same as ADC  
samples from the response of an analog RC lowpass filter would be".
So far so good.

What is bothering me is the following:

In the original organ, each type of note (say, all the Cs) derive  
their frequencies from the *same* divider chain. Hence all the C  
notes are locked in an unchanging phase relationship. You get no  
phasing whatsoever playing an octave. But you also get no phase  
cancellation.

My voice-assignable clone, however, uses different processors for  
each note, so if you played an octave, you'd get two different  
processors playing two different Cs.
Given that the 16' waveform of the higher note is the same pitch as  
the 8' waveform of the lower note, what's to stop them being  
completely out of phase?
If this happened, large parts of the sound would disappear.

Since we're talking about a digital system, this isn't going to be  
rich, lush beating between different oscillators - just two out of  
phase signals cancelling.

How would one get around such a difficulty? I'm finding I can't let  
go of the idea until I've worked it out completely...

Thanks,
Tom