Here is my new Cello patch (tweaked cello from my first post). Articulation is very basic yet (velocity, modwheel and X1 from XY1. Also LFO4 add some pitch floating for better ensemble stacking, its amplitude is for tweaking).

Short sound demo:
https://dl.dropbox.com/u/57065234/Strings/cello.mp3

and patch:
https://dl.dropbox.com/u/57065234/Strings/dkcello.zip

Now it is time to revisit violin patches and contrabass and do heavy tweaking with my new experience.

That's a great patch! I wouldn't call it spot on, it's a little reedy for me, but I like it. It's definitely composition worthy. It's great for chords.

Also, good choice of music, I love Bach.

Thank you. Yes, it is quite reedy, and also sounds like accordeon playing chords.
I think I know why. I'll try to fix it.

Here is my new cello model.

I have redesigned patch so it is using more physical modelling approach. Osc plus noise works as bow. Comb 1 acts as string model. Lower modules including master mixer acts as cello body. Also used Zebra reverb and Logic limiter.

Demo:
https://dl.dropbox.com/u/57065234/Strings/cello_bowed.mp3

and patch:
https://dl.dropbox.com/u/57065234/Strings/dk_cello_bowed.zip

P. S. Try to move XY2!

Great work. That would sound great doubled. Lots of rich body.

Hey, sound very good!
Try to fine-tune the envelopes, attack in particular maybe.

Thank you Gonga, 3ee.
Yes, envelope is not as good as before - such aggressive comb string model smears envelope (higher comb feedback - more inertia of string exciting/damping). It is need to adjust attack. Also there are double use of envelope: at the stage of bowing (noise+osc) and in the end of chain (mixer). That was needed for my previous models which was using non-linear shaping. But for current model (string (comb) exciting by bow (osc+noise)) theoretically only one envelope needed - I think at the exciting stage only.

I think I prefer the previous, unsmeared version.

Something I've been wondering - are there any practical quantitative analytic techniques that might be used to match the spectral content of, eg, a Zebra patch to a given source?

Wouldn't it be nice to automatically "evolve" a synth sound towards a given target? (I've been thinking that since DX7 days).

Yes, I've had exactly the same thought myself!

1) start with a seed, 2) breed n random variations, 3) test and select variation with best fit (closest dynamic+freq response to a given source), 4) repeat 2) for how ever many generations necessary.

... and a linked thought - there is a hypothetical hyper-dimensional universe of all possible Zebra patches, most of which will be of no practical use. But there must also be countless undiscovered superb sounds amongst them... if there were some way to map this universe and sort the galaxies and stars from the interstellar dust, as it were.

What about - for now, here on earth - a simple Preset to preset morph?

Can we ask Urs?
A morph mod with to or more presets?
It shoud be possible to automate all parameters (in a controllable way, via envelope or Mw.)

Well we can do it the hard way already.
2 instances. Pick patches. Assign automations using track outputs.

In more simple ways, the modwheel can act as a go between between 2 very different patches, by using the modmatrix for assigning extreme parameter changes all set to the modwheel.
I do this with many of my patches.

I guess you mean snapshot to snapshot. Hmmm...

BTW: Preset to preset isn't even difficult. It's impossible.

How do you morph between patches with different modules in the voice/effects grids?