# Background

So I wrote this lil' python script to obtain the camera matrix (intrinsics times extrinsics) from my scene and its active camera. It's been working great, under the following settings.

Render resolution: x = 1000; y = 1000

Sensor size: width = 50; height = 50

Pixel aspect ratio (PAR): x / y = 1

Today, I used my script under a new setting,

Render resolution: x = 1920; y = 1080

Sensor size: width = 50; height = 18

PAR: x / y = 1

and my camera matrix projects a point (which appears inside the camera as per the viewport, i.e., the "numpad0 view") to a negative y pixel value -- outside the frame!

I thought about it, and it makes sense. The (imaginary) physical sensor is shorter in height (18) than in width (50), and I am (unreasonably) demanding unproportional resolutions and a PAR of 1. Inevitably, some of the 1080 pixels in the y direction fall outside of the physical sensor!

# Questions

1. Is my analysis above correct?
2. If so, why does Blender still show me the complete rendering (instead of the partial photo truncated in the y direction), even if I'm being unreasonable?
3. An even higher-level question: why aren't these three quantities---resolution, pixel aspect ratio, and sensor size---coupled in Blender? If a ridiculous user like myself is demanding

Render resolution: x = 1920; y = 1080

PAR: x / y = 1

shouldn't the sensor size be constrained to a fixed, computable value already?