GUI design: adventures in Blender.

db is relative, so there is no absolute point at which clipping will start, that's dependent on the main volume setting in Reaktor (and possibly even driver settings and hardware on your system?).

Probably the best option is to look at how some of the poly factory synths do it. I guess just have a low enough level per voice that clipping doesn't start to early, and tweak the main volume.

In general, a good rule of thumb for balancing the output level on uncorrelated voices is scaling the output by 1/(sqrt(N) where N is the number of voices. That means as N gets larger, each additional voice has less of an impact. So for 8 voices, the scaling factor is ~0.35, while for 16 voices its 0.25. I use a variation of this for my granular engine. I wouldn't use this in a dynamic way on a polysynth, but it's useful to bear in mind when setting the per voice levels.

(If the voices are correlated - so non-detuned unison, synced, phase locked octaves etc., then its 1/N)

Probably makes sense to stick some sort of soft saturation after the mixer, so any time the output does clip, you don't get nasty digi-clipping, instead something a bit warmer. Or a limiter of some sort (rather than compression?)

In the end I think its mostly down to the user to set the main volume at a level where they don't get clipping on the loud parts, and thy can add a compressor if they want.

You're very welcome. Here's a circuit fragment that demonstrates what I was talking about.

I don't use the P input at all. I tie it to a large negative number so the internal P>F calculation results in a minimal frequency.

There are four modulation inputs. From the top:

The first is the frequency cutoff control, and ranges from midi note numbers 0 to 127, or about 8 Hz to 12.5 kHz.

The second is keyboard scaling, as done on the Nord Modular. When set to zero there is no modulation from the keyboard. When set to 1, the filter tracks the keyboard perfectly, with midi note number 64, which is E4, as the center point.

The third is an LFO whose depth can range from 0 to +/- one octave.

The fourth is an envelope generator, whose modulation depth can range from 0 to five octaves. In practice you might want the envelope depth to be scaled by note number, too, so that low notes can get a larger modulation amount. But I haven't included that.

That's nothing to worry about, you'll find out what keys to play when you play in a band. More of less of anything is all part of music.

I like all the sequencers which offer a geometric and graphical approach to repetitive patterns like node-e and that circular one (forgot the name). I also like probability parameters and nodal systems. Finally, a sequencer which repeatedly bursts into little spreads of notes like the particle engine on Noir are also very fascinating.

Basically everything which takes you away from the highly controlled details of piano rolls or regular step sequencing