The degree of sensor movement would be very different at high magnifications & extension tubes don't normally tell the camera how long they are (or even that they are present)...
To break this down a bit,
- Neither OIS or IBIS assists with z stabilization (the optical axis). AF-C doesn't provide stabilization and doesn't apply to manual lenses but, in autofocus cases, can provide some compensation for z axis movement.
- Roll is magnification invariant.
- IBIS correction for xy body translation (movement in the image plane) increases linearly with magnification. Since the the correction is 1:1 with translation at 1x magnification the displacements are especially easy to calculate. Because it doesn't know the focus distance, five axis IBIS accepting a focal length input for manual lenses has to assume some magnification to correct for and probably it assumes something low, such as for a 10 m focus distance, so I suspect these IBIS axes are pretty much disabled in this scenario.
- The xy sensor translation, in mm, to compensate for pitch and yaw is -S tan(Δθ/AOV), S being the sensor size, Δθ the angular displacement, and AOV the angle of view provided by the lens(es) used. Within the constraints of IBIS motion and, for native lenses, the image circle it's probably reasonable to use the small angle approximation -S Δθ/AOV and assume the angle of view is calculated exactly from the focal length given since the body lacks focus distance, focus breathing, and other information.
Assuming a naïve IBIS system working only from accelerometer information, the above implies
- IBIS will probably compensate more accurately if given a focal length longer than the rear lens in the coupled pair (close up lens and manual lens in this case) rather than the focal length of the lenses.
- The optimal focal length to enter will vary based on magnification and the photographer's stability and, since IBIS is probably trying to compensate for both translation and rotation based on rotation only mechanisms, may not ever be that successful.
Neither of these hold if the IBIS is smart enough to learn the system characteristics by detecting and tracking features in the image but, despite tracking AF having been available for some years now, I've not heard of any manufacturer taking this obvious step. So, presumably, naïve IBIS assumptions apply.
I don't have either an IBIS body or manual rear lenses that are useful for close up or macro photography so haven't been able to test these hypothesis. But if someone's inclined to try them I'd be curious of the results.
VERY illuminating.
