I’m having some problems with math related to the perspective camera in Blender. I need to know the camera model perfectly so I can estimate each pixel’s pointing vector for some work I am doing, but I see some discrepancies in my calculations.

I set up a 1920x1080 camera with a 35mm focal length. Blender appears to create the focal plane size as 32x18mm. Both 1920/1080 and 32/18 = 1.77…, so the aspect ratio of the image size aligns with that of the focal plane size.

With some math, I can estimate the instantaneous field of view of the camera’s pixels in the X and Y directions. I create a right triangle from the center of the image to the side for the X direction, and from the center of the image to the top in the Y direction. Using simple calculation across and along half the focal plane, I estimate the IFOV as:

Across: atan((32/2) / 35) / (1920/2) = 0.255908

Up/Down: atan((18/2) / 35) / (1080/2) = 0.267051

From this calculation, it seems the pixels are not exactly square - they have a pixel aspect ratio of about 1.044 (or 0.9583, depending on how the ratio is defined). But Blender GUI’s pixel aspect ratio value is 1.000. Should I be setting up my camera model with rectangular pixels even though Blender says the ratio is 1.000?