I still have a goal of totally simplifying stereoscopic 3D rendering. The simpler, the better.
I’m getting closer every minute. 
If you use the default settings in this “stereo base” calculator, you can end up with 3.3% (1/30) net deviation (NetD), which, not coincidentally, is my “Optimum”:
http://www.stereoeye.jp/software/sbcalc_e.html
Good news: Stereoscopy is Relative. Also, anything you discover in CG programs theoretically can be applied to all forms of stereo graphics, although perhaps not as easily.
You can substitute “units”, like those used in CG programs such as Blender, in the stereo base calculator.
I created a grid that exactly matches the Blender units and parented it to my 3D camera rig. I made one Blender unit equivalent to one meter in the calculator. Also, since Stereoscopy is Relative, things do not have to be “to scale”.
The best news is you can “eyeball” the camera settings. As long as you get “close enough”, most people will not notice small errors in NetD, i.e. anything less than .5% or so. I simply create a “near point plane” and a “far point plane”, then eyeball the calculations. I am finding that I can consistently end up with less than .5% errors in NetD.
Without writing a chapter for a book, here, I will add that I usually use a 50mm FL (focal length), simply because the farther you stray from that, the more challenges you introduce (e.g. cardboarding). Also, Blender has x-axis camera image plane shift, so the stereo window edges can be set before capture. This eliminates potentially confusing calculations for cropping non-stereo overlap, which changes the amount of NetD.
Anywhoo, maybe someone can find this info useful. I have only found one stereo camera script for Blender, and it was very weird and complicated (plus, it did not work for me).
My setup is extremely simple… I like it that way.
Have a Great Day!
Thanks,
Roger Maddy
http://rogermaddy.com/