Originally posted by m42man That's

*completely* incorrect, I'm afraid. I'll give you a quck example: say you have a 50-200mm zoom, your compromise FL value would be 125mm. With the zoom set at 200mm, the camera will move the sensor to follow the movement of the image due to shake, but will only move it 125/200 as much as it needs, so the blur is reduced to 1 - 125/200, or 37.5% of the amount you'd get with SR turned off.

Now you set the zoom to 50mm. The camera will move the sensor 125/50 or 2.5 times as much as it needs. What this means is that the camera has overshot by 150%, so it's actually

**added 50% extra blur**, over and above what you'd get with SR off.

It's crucial you fully understand this, otherwise we can proceed no further. I'll give you this link, because it illustrates what's going on in these situations:

https://www.pentaxforums.com/forums/pentax-dslr-discussion/132496-shake-reduc...ml#post1380845
Those equations appearing in post 10 of

https://www.pentaxforums.com/forums/pentax-slr-lens-discussion/178019-tamron-...ml#post1859602
reflect the model I've described. If you don't fully understand the model, then you won't understand the equations.

As I've said before, my aim is to ensure some degree of SR effectiveness whether you're at the min or max end of the zoom range.

*Having the same SR effect at both ends is a good way of ensuring you don't run into problems at the wide end, and therefore you know you'll always be better off than you'd be by turning SR off.* But it isn't necessarily the optimum way of decreasing error on the image, which is our goal.

I have completely followed your links and understand what you say; you are finding how the SR parameter chosen affects the fractional error with SR compared to no SR relative to the end points of the span. Then you find the SR parameter that gives the same error at both ends of the span.

But the goal should be to

**minimize the absolute error on the sensor**, not to minimize the fractional error in the SR engine. Say for example we have a 10% error at both ends of a range from 50 to 150mm. The motion of the image on the sensor at 150mm is 3x larger than at 50mm, although the percentage error is the same at both 50 and 150.

Regarding your example,

Quote: I'll give you a quck example: say you have a 50-200mm zoom, your compromise FL value would be 125mm. With the zoom set at 200mm, the camera will move the sensor to follow the movement of the image due to shake, but will only move it 125/200 as much as it needs, so the blur is reduced to 1 - 125/200, or 37.5% of the amount you'd get with SR turned off.

It will move the image on the sensor in proportion to (200-125) = 75mm

Quote: Now you set the zoom to 50mm. The camera will move the sensor 125/50 or 2.5 times as much as it needs. What this means is that the camera has overshot by 150%, so it's actually added 50% extra blur, over and above what you'd get with SR off.

It will move the image on the sensor in proportion to (125-50) = 75mm

Exactly the same image shift in each case - the % of what it would have been had shake reduction been off is not the question (at the low end the % error is larger because the denominator is smaller) - but that's not the concern.

What we want for a sharp photo is for the absolute motion of the image on the sensor to be a minimum given a random selection of focal lengths within the range. That's different from equating the fractional error at both ends of the focal length range.

I believe your point that when the focal length range is large enough, there is range where no SR is better than SR. Therefore there is a strategy where, if the chosen focal length is short enough, one should shut off SR - but that means you know what the focal length is, in which case you are best off to just tell SR the known focal length and have no error.

Dave