One of the under-the-hood refinements in the recent OM-D E-M1 Mk.II camera by Olympus was a re-work of their Image Stabilization (IS) system. Now, Olympus claims, it provides "up to 5 EV" gain in handholdable shutter speeds, which is more than the (unspecified) value for the original E-M1.

5 EV means exposure times 32 times longer, not too shabby. In situations where you needed a shutter speed of 1/500 s, now you should get sharp pictures at 1/15 s (assuming no moving subjects, of course).

The problem is that one sure thing which "up to" means is "never more than". Even if the number given is a result of some measurement, usually we don't know, how it was made and how the direct results were converted into the single magic value. And I may suspect that, given a choice between multiple measurement procedures (without a commonly accepted standard), the manufacturer will be tempted to select the one showing the product in best light.

The MZD 12-100/4.0 IS PRO Lens

While sticking to their in-body IS system, now Olympus also builds IS into some of their new lenses (just two at the moment), and both system can work together.

One of those lenses is the top-shelf MZD 12-100/4.0 ED PRO, which I am using most of the time with the Mark II. For this combination they claim a benefit of "up to 6.5 EV"!

This sounds quite sensational, as it translates into shutter speeds (almost) 100× slower. I had to come up with my own numbers.

This is no casual experience (see the second reference above). To assure a decent level of statistical accuracy, each focal length requires shooting from 200 to 300 frames (depending on how steep the transition line); then they have to be scrutinized for camera shake, and then the results computed.

At that last stage it helps to be literate in numerical methods. Doing it right improves accuracy and statistical reliability, making the procedure more resistant to flukes. Yes, it helps being a programmer, too.

Anyway, a few hours and 880 frames later, plus two bottles of the Sweet Baby Jesus porter from Bowie, Maryland, I had my samples collected and scrutinized in a blind process. And when I saw the first success rate values, I knew I had to stay up until the final numbers come up.

So, have a look for yourself.

The result for 100 mm is, by far, the highest IS effect I was able to measure ever, although my experience is quite modest, limited to just a few Olympus FT or μFT cameras and lenses. At 4.4 EV or so, this is still far from 6.5 EV used in Olympus advertising; still, I find this performance most impressive.

Imagine this: shooting freehand, with a 200 mm EFL lens, at 1/4 of a second, with a 100% success rate? Hard to believe, but real.

I don't know what Olympus exactly did here, but it must have been something very, very right.

One interesting detail. Note that all three green lines (showing the performance with IS) run quite close to each other, all starting to climb at -1 EV (2 s) and reaching 100% at 2 EV (1/4s) or a bit below. In other words, stabilization brings handholdable speeds to the same level, regardless of the focal length — within the limits of our experiment, at least. On the other hand, the non-stabilized system behaves the way we would have expected:

• At 35 mm handholdable exposures are three times longer than at 100 mm;
• At 12 mm handholdable exposures are twice as long as at 35 mm.

  Well, a factor of 3 would be even more by-the-book, but let's not push it.

An obvious question arises: how well does dual IS work on an older camera? To check that, I've run a comparison at 100 mm for the same lens mounted on the original E-M1 (firmware 4.3):

• The IS benefit at 50% is 3.6 EV (compared with 4.3 EV for Mk.II)
• At 100% the effect is 3.8 EV (compared with 4.4 EV)

  Actually, I believe this value may be anywhere between 3.2 and 3.8 EV, due to a large spacing between shutter speeds used. Estimates at 50% are more robust.)

You can have a look at the related success rate graph or the numeric coefficients.

Other, non-IS lenses

Over the coming weeks I want to show here the IS data for some other lenses I have access to. Let me start from a beautiful piece of glass, the 12-40/2.8 MZD Pro.

Surprise! Even without in-lens IS, the results are very close to those of the 12-100 mm lens at 35 mm. Without IS, we get 50% success rate at 1/9.2 s (vs. 1/11 s), with — at 1/1.5 s (vs. 1/1.4 s). Stabilization effect measured at 50% is 2.65 EV (6.28×), while for the 12-100 lens — 2.88 (7.37×). The results for the IS PRO lens are uniformly a tad better, but nothing worth writing back home about.

Now, another favorite of mine: the MZD 75/1.8 ED prime, as close to a perfect portrait lens as they get (REMnds me of the 75/1.5 Biotar of 1952 from my Exakta collection).

No surprises here, exactly what I would expect from a non-IS lens on the E-M1: a factor of 5 gained in handholdable exposures.

Now, let us see how the Mk.II works with the 75-300/4.8-6.7 (which I consider perhaps the most underrated μFT lens) at the long end. I was a little disappointed with how the original E-M1 performed here, so it will be interesting to see if there are any changes.

Yes, there seems to be a difference between both camera bodies. The IS gives Mark II an extra EV of performance.

For the original E-M1, the values of v₅₀ were 5.72 and 3.88 EV without/with IS, respectively. (These values describe the horizontal position of the climbing lines.) For Mark II, the non-IS value remains, more or less, unmoved at 5.50, but the one with IS moved more to the left, to 2.43!

Looks like the sensor-shaking IS system has been seriously reworked by Olympus in E-M1 Mk.II. Being able to use a 300 mm (600 mm EFL) lens handheld at 1/30 s is definitely a luxury very few, if any, other cameras' users can enjoy.

Appendix: The v coefficients and shutter speeds

The values of v₀ and v₁₀₀, and v₅₀ are as described above and, in more detail, in the separate article on the method I'm using.

The last two columns show the improvement (IS versus no IS) in v₅₀ and v₁₀₀, which can be used as a measure of the benefits of image stabilization.