Originally posted by stevebrot You can also add:
The smaller sensor effectively amplifies camera motion. I can easily hand hold at 1/15s with my FA 77/1.8 on my 35mm film cameras, but would not dream of trying that with any of my 50mm lenses on the K10D without SR. I first noticed this with the tiny sensor on my Canon G2. I don't go anywhere with that camera without a tripod.
Steve
P.S. This is just my personal experience. I am sure someone will post something with diagrams showing why this should not be the case.
I don't believe it would be the case, but I don't have diagrams :-). Every bit of logical mathematical background I have (*see below), though, suggests that shake is dependent on FOV and FOV only.
Thinking aloud:
Shake reduction works by moving the sensor to counteract this and therefore also needs to be proportional to FOV (hence the focal lengths on the SR menu). But the actual physical distance the sensor would need to be moved for a given FOV would actually depend on sensor size, I would think - smaller sensor, less distance. Less distance sounds like it's easier, but it also means you've got to be that much more precise about the movement. I'd therefore assume there is probably an optimum sensor size from the perspective of SR - bigger than that means the sensor has to move too much, small means it is harder to control the motion with enough precision. But I have no idea what that optimum would be.
* Regarding my logical and mathematical training: for some reason, this is reminding me of my greatest mathematical crisis from college, and that is when I realized that no matter how many zigzags you make while trying to approximate a diagonal line by using a series of alternating horizontal and vertical motions (eg, a "staircase" with smaller and smaller stairs), the total distance always remains the same. My intuition said it should approach the length of the diagonal as a limit as the numbers of steps approached infinity, just as the perimeter of a polygon approaches the circumference of a circle as the number of edges reaches infinity.
My experience is that issues relating to DOF and lens versus sensor resolution have the same potential to totally fly in the face of what seems logical to me at first. I have no problems believing my thinking about SR wouldn't turn out to also be wrong.