[sdiy] String synth (solina) and Hammond organ questions

[Tom Wiltshire]

On 11 Feb 2017, at 14:08, Gordonjcp <gordonjcp at gjcp.net> wrote:

> On Sat, Feb 11, 2017 at 11:42:18AM +0000, Tom Wiltshire wrote:
>> 
>> On 11 Feb 2017, at 11:03, Gordonjcp <gordonjcp at gjcp.net> wrote:
>> 
>>> What you absolutely must do is ensure that all the notes you generate start in phase.
>> <snip>
>>> This is something you'd need to do on a tonewheel emulation, too.
>> 
>> 
>> Why? The Hammond tone generator is free-running and the pitches are fed straight to the busbars. While the frequencies themselves might have a fixed phase relationship, it's not related to note onset in any way.
>> 
>> Or am I misunderstanding what you meant?
>> 
> 
> Okay, pull out the 8' bar and play C3.  Holding that down, play C4, on and off a few times.  You'll hear it key on and off, great.
> 
> Now pull out the 4' bar as well, and play C4 again.  What's the phase relationship between the fundamental of C4 and the 4' stop of C3?  Can you see a situation where you might end up with the two cancelling?
> 
> If you were generating all 91 tones, you'd just be gating them on and off.  But in a typical software synth you'd generate each voice individually with fixed polyphony.  If you're not careful with it you could easily end up with some notes out of phase with others and odd random shifts in tone colour as a result.

Aaah! Yes, ok, absolutely! I see what you mean. Thanks Gordon.

This is a very good point he's making - since the Hammond uses the same tone generator for every/any note, there's no danger of phase cancellation. The same is true in the TOG organs/synths. Any "voice based" implementation needs to ensure that it copies this effect. This isn't too difficult if all the voices are produced on the same DSP, but it makes processor-per-voice implementations a bit awkward to do.

I'd be inclined to run NCOs simulating the correct Hammond pitches for the driving/driven gear combinations, and then use shifted versions of the phase accumulator to generate the waveform indexes for the required number of octaves. With only a dozen notes, that's not a lot of work.
To adjust the pitches for the standard TOG division frequencies would then be a case of tweaking the twelve frequency increments.

Tom