The new Olympus flagship SLR, the E-3, like the preceding E-510, has a body-based image stabilization system, as briefly described in my E-3 review. Olympus claims the system brings 2 to 5 EV of improvement; i.e., all other things equal (lens, condition, operator's shooting habits) a photographer may use with IS shutter speeds 2 to 32 times longer than without, expecting the same degree of image motion blur. (We are assuming, of course, that the subject is static.) Two to 32 times is a wide margin, but this cannot be avoided: the effectiveness of the system depends on many factors; one of them is the lens focal length.

I have learned never to take similar claims at the face value, especially if they are not supported by any hard data and/or references to method used. This is why I had to make an attempt to check these.

As the whole image blur process is of stochastic nature (after all, how your hands shake is not a very repeatable process), I had to devise a method which would be reproducible in the aspects which could be made such, and statistically sound in others. I think my method holds water in both these areas.

Update of 2008/03/02: a note on how not to evaluate IS benefits.

Experiment design and result interpretation

Results

Now, time for the moment of truth. All raw and semi-processed numbers are, again, in the Appendices; here I'm showing only the data points and best-fit lines describing the behavior of goodness versus shutter speed. Remember: for each focal length the value of Δv₅₀, or the horizontal spacing between lines at half-height, is the measure of the gain we are getting from image stabilization.

Conclusions

The results of my small experiment may be summarized as follows:

• The improvement in handholdable shutter speed range due to image stabilization in the E3 varies from about 1 EV at F=12 mm to slightly more than 2 EV at F=150 mm.
• There is no evidence that this gain is greater than it was in the E-510.

It is possible (I would say: thinkable) that under more adverse conditions, when handholdable shutter speeds are higher (less steady hands, longer lenses), the improvement may be more visible. This is hinted by the results for different focal lengths, obtained for both cameras I've tried. I have, however, no hard data to support or dispute such a hypothesis. When I get the new 70-300 mm ZD lens, I will try it out in this aspect.

Last but not least, a comparison of v₅₀ for both cameras at similar focal lengths shows that not only the improvement, but also handholdability threshold itself is the same for both cameras in spite of differences in weight and size of the lens/camera combinations.

Note: How not to evaluate IS benefits

I have seen this error committed so many times, that it I feel it deserves a separate explanation.

Somebody says: "With IS I'm getting sharp images at 50 mm (100 mm EFL) and 1/10 s; at this focal length the normal handholdable exposure is 1/100 s, therefore I see IS benefit to be 10× or 3.3 EV".

Wrong. This means nothing. Turn the IS off; you'll be probably having good pictures at 1/40 s or so. Does that mean that a disabled IS gives an improvement of 2.5× (1.3 EV)? Just because your camera has an "IS" sticker on it?

The only right way is to do a with/without comparison, keeping all other things equal, and then you have to catch the region when the change occurs (i.e., the pictures start getting blurred). Anything else is old wives' tales.

Appendix One: Raw data

These are the raw "bad-medium-good" frame counts for each series of frames, with my "goodness score: shown below, as described in the E-510 IS article.

Appendix Two: The v coefficients

More interesting: the values of v₀ and v₁₀₀, as described in the other article.