Originally posted by RGlasel You and I are probably products of the same educational system, so same caveat from me.
I do happen to know that the earth's rotation speed is not constant (I played with algorithims for calculating sunrise and moonrise to do some calendrical calculations). The earth's acceleration is very small and is negative at the moment (the earth's rotation is slowing down a bit) but the speed that objects on the surface of the earth are moving is high enough that the amount of inertia those objects have (which need to be moved in a hurry for image stabilization) will be a limiting factor. Bigger sensors have more mass, so they require more force to be stabilized and the physical dimensions and energy supply of a particular camera model might also be a limiting factor. There might be some measurement limitations of the MEMS (microelectromechanical system) gyros used for image stabilization in cameras, but when the Olympus spokesman is talking about the rotation of the earth limiting their IBIS to 6.5 stops, he is probably refering to the effect of inertia on their sensor. A smaller sensor witjh the same IS system might be capable of even greater stabilization, but who wants a Q that is the same size as an OM-D E-M1?
It's actually because the MEMS gyros are so good that this is a problem. The modern chips are now accurate down to almost 1 degree per hour so they readily sense the 15° per hour rotation of the Earth and camera.
If you put the camera on a solid tripod and read the signal from the gyro, the gyro would tell you that the camera is actually spinning once per day about an axis pointed at the North Star. And if the camera's software knew it was on a perfectly solid tripod (which don't exist in real life), the software could easily know to ignore the measurable spin of the Earth, camera, and subject.
But, in real life, the data from the gyros of a handheld camera would show all sorts of random rotations in random axis directions. Unless the camera can be confident about whether it is pointing North, East, South, or West, it doesn't know how to subtract the Earth's rotation from the gyro data to accurately estimate the rotation of the camera relative to the Earth.
The limitation is actually the flip-side of the astrophotographer's problem with star trails. On a K-1 with a 100 mm lens fixed on a tripod, the Earth's rotation means the stars move as much as 1 pixel every 1.5 seconds (depending on the pointing direction). Using the raw gyro signal for motion correction can remove the Earth's rotation but now all the Earthly subjects in the scene start moving up to 1 pixel every 1.5 seconds. That implies that shutter speeds slower than 150/F will show Earth rotation effects at the pixel peeping level. That's 7.2 stops slower than the 1/F rule of thumb for safe shutter speeds.