Sensor shift
In a parallel rig, the way to set convergence during the shoot is to shift each sensor horizontally off the lens axis. The image is not cropped; full resolution is preserved.
Amount of shift
For the point at the convergence distance to fall at the center of both images, the shift of each sensor is:
where: h the shift of each sensor (mm), f focal length, b interaxial, C convergence distance
Because the lens inverts the image, the sensors are shifted outward from the axis. The sum of the two shifts is the on-sensor equivalent of horizontal image translation (HIT) in post-production: f · b / C.
Image circle
The corners of the shifted sensor must stay within the image circle of the lens:
where: Ø the diameter of the image circle, Wc and Hc the width and height of the sensor
Example: 35 mm lens, Super 35 sensor (diagonal 31.1 mm). With a 65 mm interaxial and 3.5 m convergence, the shift is 0.33 mm and the required circle is 31.6 mm. As the convergence comes closer, to 1 m, the shift rises to 1.14 mm and the required circle to 33.0 mm.
In practice
- Shift lens: tilt-shift lenses leave room for the shift with a large image circle. When this is selected as the lens type in the calculator, the required shift is compared with the shift limit of the lens.
- Virtual camera: in computer graphics, an asymmetric viewing frustum does the same job without loss.
- Post-production: if the shift was not made during the shoot, it is applied with HIT; the cost is cropping at the edges. See Parallel rig.