Second-Generation Stereo Rig

As with the first stereo rig tutorial, here we present a follow-on iteration of stereo rig. You should be familiar with the first rig as the discussion of the second will omit details already described there.

This second rig is designed to allow the cameras to be approximately aligned during setup, so that less zoom will be required in post-production to correctly align the images from the two cameras. Unlike the first rig, which had a human-sized inter-ocular distance, this rig creates an extremely large inter-ocular distance. You can easily change the inter-ocular distance during construction by changing the length of the main horizontal beam, down to a minimum value set by the width of the fold-out LCD monitor display on the right-hand camera.

Here is a first picture of the rig.

The rig is constructed from assorted pieces of Dexion-type angle iron, totaling one long piece and four small angle bracket pieces that are only approximately an inch wide. A number of additional screws are required to mount the cameras and assemble the pieces. With the pieces of already-cut Dexion that were handy, the inter-ocular distance is 11 inches.

There are two key ingredients to the design: first, the brackets are assembled such that the cameras can be tilted vertically and panned horizontally, then locked in place; and second, the camera LCD monitors have a small overlay attached that will be used carefully. Note that, improving on the first rig, you can access the monitors, see both images, and change batteries or SDHC cards while the rig is assembled.

Here's the detail of the individual camera mount:

The first angle bracket on each side attaches to the main crosspiece and creates a vertical surface, to which the second bracket attaches. After the 90 degree bend of the second bracket, we're back to a horizontal surface to which the camera can mount. The connection between the first and second brackets occurs via a bolt; this bolt is the axis about which the camera and second bracket rotate to produce the tilt. The camera pan motion is created by rotating the camera about its main mount hole.

The point of the ability to aim the cameras is to allow us to approximately align them prior to shooting, to minimize the post-production zoom required. To understand this, you should review the tutorial on lens centering In short, cameras are not assembled very accurately or repeatably from unit to unit, so the optic axis of the camera is generally not at the center of the image, as we require for general use.

As an approximate way to align the cameras despite the mis-centering, align them first with the cameras zoomed as tight as possible. Then zoom each camera back out to the operating point: perhaps not all the way wide, to minimize the amount of lens distortion. (The transition between 1x and 2x zoom can be located approximately on these cameras by watching the LCD display during a slow zoom in, but it is not a very accurate method.)

To align the cameras, we put a small overlay over the camera monitor display. You can see it in the pictures above. The overlay was constructed in Adobe Illustrator and the resulting centering grid PDF printed on transparency film at FedEx Office (Kinkos) for $0.65 or so, then cut exactly to size. This grid is particular to these Panasonic cameras; you would probably need to produce your own design for other cameras.

These grids are designed to match the center point of the camera display, which is not the optic center point, of course. Part of the point of the zooming tight, then zooming-out procedure is to mitigate the effect of using the wrong centering point. Roughly, if you have zoomed in by a factor of ten to align, then the error due to using the wrong center will be reduced by a factor of ten once you get out to the actual working zoom level. Alternatively, if you are very careful, you might be able to create customized centering grids that match the actual optic center location, as determined by the process described in the centering tutorial That is probably over-optimistic, however.

The overall procedure on-set goes like this:

Assemble the cameras onto the rig, get them turned on and running.
Pick a particular visible spot on the set to be the alignment point.
With the rig fixed on a tripod, align each camera to center the alignment point in the set at the cross-hair on the camera's LCD display, then tighten the screws to lock it in place.
Shoot! You can move the camera rig all around now, but don't knock the cameras out of alignment on the rig.
In post-production, take the footage from the cameras and do a final alignment on it.

The distance from the camera rig to the alignment point is an important value. It is the vergence distance of the cameras: you are converging them at 10 feet, 20 feet, whatever. This will be the distance at which there will be no disparity between the left and right images. Most of the useful content should be in front of this reference point; the reference point is effectively the movie theater screen, and the significant 3-D content will come out in front. So generally the distance should be the comparable to the furthest-away features.

We show how to achieve this in a separate post-production alignment tutorial.

Despite the pictures above, you should not use this rig for objects close to the camera, it will produce too much disparity. Instead, it is better suited for producing a sense of depth in scenes where objects are further away.

Below is an aligned anaglyph image from this rig. Due to the extreme inter-ocular distance, the floating water bubbles in the foreground have very large disparities that make them hard to watch.