I seriosly doubt Blenders Camera has a nodal point, other than Cameras Position, and if it does - probably not at the same place as your Graph. But thats probably not important for you?
Just out of curiosity: What does your graph show? Every Lens has other nodal Point, so this is for a specific Lens?
Do you have the Formula to get from focal length to offset? If so, i would try to add a driver to camera and then scripted expression where you can put in your formula.
If you don’t have the formula, well i’m not sure about this…
bashi’s suggestion of a driver is a good idea, however I would link the lens to the z-position of the camera, not the other way around. As far as a formula, you could just create a fake user f-curve that looks like the graph you have posted up there. Then make the script expression sample the f-curve, based upon the camera z-position to derive a new value for the lens. I’m not sure which value you need to drive, are you thinking focal length? Then when you move the camera the lens will change automatically.
Looking into drivers - I have a simple file where the focal length drives the translation of a cube - I can see that adding a modifier to the driver gets me close to what I need …if there was a way to add an f-curve as a modifier to the driver I would have it exactly…
Hmm, I wonder how the tracker accounts for nodal offset? I thought that it would only be an issue for tripod type shots, however there is a difference between DSLR type cameras and ENG or video type cameras. In the latter the nodal point is quite a way forward of the handheld position.
If I understand your example file there’s still needs to be another curve that represents the how the nodal offset (Y Location) behaves for each focal length value. I think it’s all about the driver modifiers…
BUT digging thru the docs I find this:
Example
In this example, we are going to control the size of the well-known monkey head (Suzanne) with the Y-location of the Empty driver. So, we Add Driver to the three ScaleX, ScaleY and ScaleZ channel of the Suzanne object (as usual, if there is no curve yet, it is automatically created). Note that for now, there is no curve, so Blender applies a one-to-one mapping, as if there was virtual unitary gradient linear curves (materialized as yellow dashed lines in the pictures below). This also illustrates that you can use the same driver property (here, the Y location) for several different drivers…
So I guess drivers are indeed the right approach to solving this problem BUT unless these docs are not up-to-date - I believe some other approach is needed.
BUT digging thru the docs I find this: