So make the savonius half the diameter of the main turbine, then its blades will spin at approx 2x the wind speed. Maybe.
A savonius (say, the Sandia design rather than the classic offset split-cylinder) runs at a tsr of about .8 Darrieus runs at a TSR of, what, maybe 6?
So make your Savonius about 1/7.5 or 1/8 the diameter of your darrieus and you won't need a clutch. (Or adjust that number if your Darrieus runs at some other TSR.) It will still be driving when the darrieus is up to speed. No clutch required: It won't be producing a net drag until its TSR is near 1. So even if the darrieus pushes it beyond its optimal power-collection speed it will never act as a brake.
As a drag-type turbine, the savonius provides MORE torque at lower TSR, so when the rotor is stopped it will be providing max start torque.
How much power will it take to overcome the drag of the savonius itself plus the drag of our main turbine (bearings, aerodynamic,...)? So how tall does the savonius need to be?
I think that will depend almost entirely on your bearing friction and genny's cogging, and from drag asymmetries if you have an odd number of blades. As I recall an N-blade Darrieus starts having some forward drive as soon as it's moving enough to go through 1/Nth of a rotation. (That's why 3-bladers will self-start if the friction is low enough: The asymmetric drag due to two blades on one side of the mast gets 'em moving enough to start getting drive, if I have this right.) So if the Savonius gets it going at all it should spin it up to speed.
A darrieus has pretty low drag even with one extra blade on the downwind side being driven backward, due to the thinness of the blades and their orientation at right angles to the radius. So it shouldn't take much of a Savonius to do the startup. I'd try making one that ought to break cogging and bearing friction in 80% of the cutin windspeed if it were the only thing on the shaft and give that a try. If it doesn't do the job at cutin windspeed, make it taller until it does. B-)
We can't really put it inside the main turbine because it will choke out much needed air flow through the turbine and may not even be high enough.
For sure. It would also increase the flexing stresses on the blades by magnifying the windshadow of the central shaft.
But if it needs to be even a fifth the height of the darrieus I'd be shocked, and wonder what you're doing for bearings and what happened to the design of the genny.
(By the way: I'd modify that housing in the bottom-middle of the rotors, especially the blue one - bringing in the corners and replacing the rectangular casing with a cylindrical one, as narrow as possible.)
Put it on top and as the central shaft traces a circular path due to forces on the blades of the main turbine (the central shaft tends to wander during operation), the savonius starts a rockin' in a major way.
"rockin'" won't bother a savonius. It will get a litte more power when moving toward the wind, a little less when moving away. Side-to-side will do nearly zilch - averaging to a slight boost across the various orientations. It's not like a lift-type, where angle-of-attack errors cause stalling, so you can "shake the wind out of" the rotor.
Another option might be to redesign the struts using spring-loaded airfoils that are open at low wind speeds and close once the turbine begins to lift on its own, essentially operating as drag turbine at low speeds and lift at higher speeds.
I wouldn't attempt that. Moving parts and joints are a disaster. (It's bad enough that we have to have the rotor in the first place, and that the blades flex.) Same argument applies to the clutch: Design with the right diameter ratio so you don't need it, making the rotor one solid structure that does its job just by being the right shape.
Large scale commercial HAWT's (and VAWT's) are very much computer controlled, direction, speed, pitch, pretty much everything. Why can't it be done for a small VAWT?
It can be done. But it's a far larger percentage boost in the cost. With a large turbine you have a lot more machinery to build and a lot more power being collected, so the cost of the control is a smaller fraction of the resources. Also, a large one may NEED serious process control to work properly and safely. And a small percentage improvement in efficiency can pay for a LOT of control hardware and programming on a big machine, while on a small machine it's a drop in the bucket. (Just like with fossil fuel plants vs. automobile engines...)
Now perhaps your mill will need the computer and motor control to furl properly, or otherwise to control and/or monitor it. If so, it might not be that much extra expense and effort to make it also funciton as an electric starter. In that case it there's little to be gained by making the rotor self-starting and it makes more sense to use the whole cross-section for the darrieus and K.I.S.S.
Also: Computers are dirt cheap these days, and servo control is getting there, too. Perhaps the cost will not be an issue.
And I'm sure there are other factors I'm not taking into account.
Anyhoo, it's your design and your call. Good luck.