This seems to make intuitive sense, but is a myth. You will get keystoning when using a "converged" or "toed-in" set up. This can be corrected through digital means, but not with film. In mild cases you may not notice, or may experience some eye strain, but in severe cases the images will be unfusable. Try it and see what you think! There's a pretty good short article with diagrams here: http://www.rmm3d.com/3d.encyclopedia/keystone/keystone.html
Sorry, I don't have any digital images of this effect as I don't routinely make a habit of doing this! The vast majority of all stereo cameras ever made have parallel lenses and parallel film planes. There was a good reason for this design decision.
The reason people want to toe-in their cameras is so that the left and right image borders are "the same" for the two images. This is actually fairly important, because if the edges of the frame do not "coincide" correctly, you will get window violations (the image will appear in front of the stereo window, which most stereo enthusiasts regard as bad! James Cameron doesn't seem to mind window violations (and nor do most other film makers who followed in his wake), but I regard this as a major reason 3D movies have never gained widespread acceptance. But that's a story for another day.)
Instead of toeing-in, the correct way to remedy the problem is to crop a bit off left side of the left image and crop a bit off the right side of the right image.
For a stereo camera, this is done "automatically" by designing the camera with properly adjusted film gates, such that the image captured on the film is shifted toward the outside of the image circle produced by the lens, rather than captured directly behind the lens in the centre of the image circle. Dedicated stereo cameras such as the Realist or the Sputnik will have properly aligned film gates. But as I said above, it is not really a problem to use a regular "mono" camera that captures the centre of the image circle. All you have to do is crop a small strip off one side of each of the images. Then the stereo window will lie in front of the image. The only downside is that you have slightly less film economy. Technically the two approaches are identical.
To explain this better, I made a little graphic:
We are looking down from above onto two cameras that are parallel. The scene is at the top of the image, and the film plane is at the bottom of the image. The blue solid lines show the field of view of the left camera and the red lines are the right camera. The points where all the lens cross is the optical centre of the lens (think of it as a pinhole).
The black dotted lines show the optical axes of the two cameras. Note that the cameras are parallel and no toed-in.
The green lines show the entire image circle produced by each lens.
The coloured dashed lines show the frame boundary of the stereo image (where there is information captured by both cameras).
The red and blue lines sitting on top of the green line show where the image containing stereo information will be captured. This corresponds to where the film gates would be placed on a true stereo camera. There is no need to capture the image circle outside this area. If using a "mono" camera, the film gates are positioned directly behind the lens, and so you would just chop off the "innermost" sides of the image.
Keep in mind that when using a lens for imaging, you get an inverted image (just like a pinhole)! So to view "parallel", you need to cut the film ships apart and swap them. When you swap the chips, the "innermost" sides of the image that need to be cropped off become the outermost (i.e. as I said above, you need to remove the left side of the left image, and the right side of the right image.)
Hopefully that gives some helpful background as to why one should generally avoid toeing-in.
Last edited: