« April Fools | Main | The Problem of Microstock in a Nutshell »

Thursday, 01 April 2010

I am a bit surprised by this. I have used KM and Sony cameras at 200mm in churches with very good results. At 1/30 I would say I get 90% usable, and most of those excellent.

"At shutter speeds above 1/250 of a second, I am half-convinced that image stabilization produced less sharp results than no stabilization at all."

Exactly my experience with the E-P1, which is why switched IS on and off depending on the shutter speed. Quite a hassle, actually.

Of course, as you know, Ctein didn't test a K-M or Sony camera...we can never assume that a test of one product extrapolates to all other similar products, or that the results of a test of a feature in one camera applies to that feature globally. His comments pertain just to the camera and lenses he tested under the conditions he describes.


I have the Olympus 620, which has the same IS mechanism as the Pens, and I soon noticed the same thing. I'm not sure if I see a degradation of image quality using IS at higher shutter speeds at longer focal lengths, but I sure don't see any help. My only guess is that it just can't adjust quickly enough to make a difference. So I go back to the classic shutter speed = focal length trick (x2 for fourthirds) and things work well.

I want to thank Ctein for the kind words about and link to our Micro 4/3 group.

Having used image stabilization systems from Canon, Nikon, Tamron, and Panasonic my impression is that what Ctein found for the Panasonic lens tested here is unique to Panasonic amongst lens stabilization systems. It's just an impression. I haven't done a controlled comparison.

Could it be that Panasonic OIS is less effective for still photography because it is optimized for video? I'd be interested to hear from anyone who understands how stabilization technology for still and video applications may differ.

I have several K-M and Sony DSLRs and a Panasonic G1. I've noticed a real difference in image stabilization between them.

The K-M/Sony cameras work really well with stabilization turned on all the time (the Sony a bit better than the K-M 7D, but then they are newer generations of the technology). I think it helps even with high shutter speeds.

The G1, on the other hand, seems to work about the same as the Sony at low shutter speeds, but actually seems to make things worse at moderately high shutter speeds. At really high shutters speeds it doesn't seem to matter any more. I wish it worked more like the Sony version. I have lost a few shots on the G1 due to image stabilization induced blur. I don't think this has ever happened on the Sony.

I was not impressed at the VR performance of my Nikkor 70-200/2.8 (now the "old version"). I was rarely able to get sharp shots hand-held with VR at even 1/40sec.

I need to repeat this with my Sigma 120-400 lens, and perhaps with the LX3 camera.

Or maybe my 70-200 really IS broken?

Wonder whether this might relate to the mysterious, and apparently erratic, behavior of some Pentax K-X's,which seem to produce bad images using shake reduction at around 1/125?


Did you do a series with the camera+lens mounted on a tripod? I imagine this would at least provide a benchmark for maximum resolution, assuming accurate focus, no camera motion and no shutter vibration. (And, needless to say, IS was turned off).

Dear Clayton,

What Mike said. Plus, if you reread my column, you'll see I'm talking about an IS problem at high shutter speeds: "At low shutter speeds there was some benefit to image stabilization, but the higher the shutter speed went the less this was true."

For all you know, you'd have the same problem... or not. Maybe you should run some tests.

pax / Ctein

Somebody should finally get around to testing this methodically, reproducibly and on a large number of shots; a comparison across different systems and camera/lens classes would be quite enlightening, I think.

One could "record" movement with a camera dummy with gyro sensors inside (or even an iPhone with custom software strapped to a camera body), and have a machine reproduce these motions as the test shots are fired. Software could then automatically analyze the images' sharpness.

AFAIK, Sony (actually using nothing more than a tweaked Minolta technology) is about the only example of in-body IS that has no quirks and works almost as advertised, even with 500mm lenses. And that's valid for KM-7D/5D, up to the latest A550; at least two stops "better" than simply hand-holding. In recent years Sony tried to push 4+ stops, but that's valid only in their marketing office.
Pentax seems to be equally effective at long(-ish) exposure times, but almost useless in the tele region. Oly has mixed results, mirroring the Pentax results but slightly better for tele. Same for sensor-based stabilised Nikons (in-body VR, sensor moving, for their "compact" superzooms).
Amongst the lens-based there is only Canon that, even for lenses like 100-400L with only 2 stops IS, seem to "get it" not only for the final result but, quite important, for the viewfinder sensation. Nikon VR lenses work nice, but the old implementation (like in the first version of 70-200) is rather unpleasant to watch, even if it's effective enough. The same for the (very few) PanaLeica interchangeable OIS lenses.
Sigma seems to be the least effective, and usually the most hated for the seasickness :D it would give.
An up-and-comer seems to be Tamron, which is very effective and has a rock-solid image in the viewfinder, with one caveat: it needs the longest time from the half-press 'till the full stabilisation takes effect; clearly not for the action/twitch shooters.

As of why is that... I hope that Ctein will get to the bottom of it; all I know is what I used (and sometimes learned to overcome or at least got to accept as is).

"Somebody should finally get around to testing this methodically, reproducibly and on a large number of shots"

I believe imaging-resource.com now does so.


Wow, what a shame. I bought a EP1 not long after they came out. I shot with it in good light for a day or so then dumped the images on my computer. I was dissapointed that a very large percentage of images were blurry. Images shot at 1/125th and higher. I figured it couldn't be camera shake with that high of a shutter speed and I had the stabilization ON. I took the camera back to my dealer and had to spend an hour showing him what was wrong to get my money back. I wrote that wonderful camera off as junk. Goes to show that you can't be a right brainer in the digial world.

imaging-resource.com (more precise, thru their sister site slrgear.com) tested only two examples: in-body stabilization for Oly E520 and lens-based stabilisation for Canon EF70-200mm f/4L IS.
Two tests in exactly a whole year... While extremely useful, this only represents just an apetiser.
What's even more sad is that they said to come upon products which fared pretty worse, but they kept mum about which ones would that be.
Considering that imaging-resource.com is happy (more often than not) about even the lowly me-too-cameras, I'm afraid that they're afraid to publish the really troublesome findings.
Still, can't wait for their next test.

Dear Gordon,

There wasn't any need for tripod comparisons. One could easily see the difference between the frames as smearing or overall loss of resolution in the high-speed tests, as shown in the illustration. It's clear, for example, that frame five is noticeably sharper than any of the others and close to perfectly sharp, if not entirely there.

At low shutter speeds, it's hard to tell how close to perfect image stabilization is, because small blurs tend to come out more symmetric. But at low shutter speeds, the benefits of image stabilization are obvious and overwhelming, compared to photographs made without it.

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 

Dear DDB,

You're going to have to show me some of your handholding techniques and tricks again, when I see you (in a day!). They didn't improve things for me when we tried them two years ago, but that was two cameras and two different IS systems ago.

Clearly I need to find a better handholding technique if I'm going to use the long lens sans tripod.

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 

This is somewhat tangential but I loved Jeffrey Friedl's take on what to shoot when doing stability tests of any kind. He used a test pattern on his LCD display when testing for tripod stability and that struck me as an excellent choice.


He has several other articles on the procedure—they're linked to in the piece referenced above—and they attracted quite a few comments and I believe the writers generally came around to his way of thinking, despite some initial scepticism.

Oh, that's too bad. I believe I just read the announcement and the description of the testing method when they decided they were going to test and report on this feature.


Next time I try hummingbirds I will see if I can turn my IS off. I have some shots at 1/1500 with the IS on, and don't see anything wrong with them, but doubt if the IS helped any. On the Sony's I think you can just leave it on.
I agree that we can't assume that tests with one system are valid with another, but I would expect some similarities.
Mike, you still have Sony and 4/3 available?

Very interesting tests these. It does look like you've simply reached the frequency limit of the system, at about 250Hz.

When the first few sensor-moving IS systems came out, everyone was saying (without testing) that for long lenses they would be less effective, because they can only move so far... which is crazy because (say) two stops worth of blur reduction is always going to involve movements of about 4 pixels. But what I didn't hear anyone saying, and didn't think of until now, was that a potential advantage of the lens-bases systems is that all the sensors and shifters can be aimed at the frequencies which will be relevant.

The surprise here perhaps is that the Panasonic lens doesn't appear to be able to handle frequencies fitting for its focal length. That Olympus may have aimed the camera's system at the kit lens seems less surprising. As Amin said, video may influence the designs too; anything above 60Hz is pretty useless there but 4, 10Hz are important.

I wonder if anyone has taken one of these apart and looked up the specs for the accelerometers they use?

I used to have this problem with an old Canon 28-135 IS - from memory, the first affordable IS lens available to the common man. I'd get shaky pictures at hand holdable speeds (1/125 and faster) with the IS on, but pictures were fine at slower speeds. I used to turn IS off for the faster shutters.

Unfortunately I have no explanation whatsoever!

This is mildly off topic, but Ctein, wow! That's a beautiful photograph, and if I read my EV chart right, iso 800 @ 1/8 @ f/1.7 is about EV 1 or 2, right? How wonderful!


Dear Bahi,

That's a pretty cute trick of Jeffrey's. Amin (of mu-4/3) did him one better when he was checking my IS results. He just photographed a normal mostly-white LCD screen. The individual pixels provided a nice high-resolution grid for checking relative sharpness.

Wish I'd thought of that!


Dear Will,

Thank you very much. I am quite fond of the print. I referred to this photo in my column on what tests don't tell you, a few weeks back. The ISO 800 version was printable; the ISO 1600 version fell right off the cliff into unusable noise.

pax / Ctein

Dear Improbable,

Several surprises for me. The first was that there was even a problem. The second was that it was as big a problem for the in-lens stabilization as the in-body. Third was that having them both on made no difference, on average, at any shutter speed, compared to having just one system on. I'd have expected better... or worse... but not no-matter.

Long lenses are harder to stabilize because the same acceleration produces more image movement. The resolution of the accelerometer has to improve in proportion to the focal length, to get you equally good results.

pax / Ctein

"Long lenses are harder to stabilize because the same acceleration produces more image movement"

This is certainly true for in-body IS, but it is not evident to me that it is necessarily true for in-lens.


I only got one IS lense, the EF300mm F4L IS. "the manual" say that IS is effective at 1/30-1/125 shutter speeds. I belive this is an older version of Canon IS, dont know if the newer version act differently. Its not my most used lense, and I never really done any IS tests with it.


It is relatively well known that the frequency at which the Nikon Vibration Reduction (VR) system works is somewhere around 1/500s. As a result VR can be counterproductive above that shutter speed.

I see no reson why it would be different for the OLYMPUS IBIS or the PANASONIC lens stabilization system.

By the why, here is what Thom Hogan says about « When should I use VR? When should I turn it off? »

VR should always be off unless you explicitly require it. It should always be off for shutter speeds over 1/500. It should be off if you're on a stable tripod even if the VR system says it is tripod aware. Basically, VR should be off unless you can guarantee that without it, you'll get camera motion in your shots.

Sounds clear clear to me.



If it is a problem damping high frequency motion, you might be able to develop a technique in which you deliberately introduce low speed motion, to reduce the high-speed shake. I am envisioning a sort "whole body wave" thing, where I am bouncing slowly on the balls of my feet, and carrying that motion up to a slow bounce of the camera.

Let the IS deal with the introduced motion, in hopes of reducing the motion it can't deal with.


Have you experimented with manual focus and with continuous shooting? Autofocusing takes time and image stablization can interfere. Especially when your camera (like most) gives getting the shot priority over complete and accurate focus. The whole process of half pressing the button to activate focus and IS I would assume (I use Canon) is difficult to judge on in boby stabilized systems, since you cannot see what's happening (opposed to lens IS).
Shooting multiple shots (burst) gives the system more time, bypasses the shutter priority and also reduces the shutter pressing motion.


I have an Olympus E510 with 14-54, and 50-200mm lenses. I have also tried legacy lenses (takumars) with adaptors. What I have noticed is that with the oly matched lenses I seem to be able to set the camera to tell it I have a 100m lens onboard and I do not get any blurring with any lens. However it was quickly apparent with the legacy lenses that I had to tell the camera the focal length it had on board or the IS worked against me. If I had 100mm set and then put a 35mm takumar on the camera the images were definitely blurred but as soon as I set the camera to 35mm everything was fine.

I haven't finished investigating, but I suspect that Oly camera's may be able to read the lens that is attached' if it's an oly, but otherwise may need to be manually told what is mounted on it.


You don't say whether you have in-body or in-lens IS, but the big advantage with a 500 mm lens (or even my 300mm on an APS-C sensor) is that it *really* locks the image in place in the viewfinder.

If there really are shutter speeds that are worse with IS, it would be nice if you could use IS for framing and turn it off to take the image. 8^)

All of the stabilization solutions have limits, some worse than others.

For what it is worth, the frequent comments from Nikon users that I've read suggest that VR (Vibration Reduction - a lens based stabilization Nikon offers) should generally be turned off at 1/500 second and faster as it reduces the sharpness at that speed and faster on many lenses.

The improvement that VR provides is very visible even in the viewfinder, in my experience with a 70-300 lens I use. That's one advantage of lens based stabilization, btw. It improves the auto-focus and your framing, not just the capture. By the way, that improvement is visible to me in the viewfinder, even with a monopod in use at longer focal lengths.

The m4/3 thread included a link to a very detailed Imaging-Resource study of this EP-1 problem which they call anomalous blur. The study focused on the 14-45, and found the problem at the same shutter speeds.

The study was the result of many days of work by Shawn Barnett, an EP-1 devotee.

You can find the study in the Imaging-Resource review of the EP-1 in the Optics tab.

Dear Roberto,

Doesn't matter where the correction is occurring, I'm talking about sensing the error. Suppose some particular acceleration is going to produce a 1/2 pixel smear with your 20mm lens. The IS system doesn't need to resolve that acceleration; it's too small to materially affect sharpness. Replace that 20mm lens with a 100m lens, and the same acceleration will produce a 2.5 pixel smear, which you *will* notice.

In lens may (or may not) be able to correct better, once the acceleration is resolved, but it needs to be resolved in the first place. The sensors are not infinitely sensitive.


Dear Tregix,

I think Thom is likely correct, but you can't simply generalize from one maker to another, because each company implements IS differently. What's true for Nikon (Thom's particular system) will not necessarily be true for Canon, Pentax or Olympus. You may see no reason why it should be different, but there is also no inherent reason it should be the same.

pax / Ctein

Dear Jan,

Yeah, I wondered about that. Did try bursts, got no difference, dammit. (But then I don't really trust AF, so I always press halfway and confirm that I've got a lock on what I want to photograph before I complete the exposure.)


Dear Bill,

On the Olympus, all mu-4/3 lenses are automatically keyed in as to focal length. I tested that to confirm it. There is manual FL setting for 'legacy' and non mu lenses; it has no effect with the mu lenses.


Dear Bruce,

The anomalous shake I-R found produces a very specific and characteristic blur pattern-- vertical and progressive from top to bottom of the frame. If you re-examine the test photos I posted here, you'll see the blurs are randomly oriented. You can't see the whole frame, but my blurs are not progressive down frames, either.

The problem may not have been entirely fixed by the v1.1 software, but it's clearly reduced it enough to allow other sources of blur to dominate.

The I-R report is useful because it's one of several that help to establish that IS is actually on at high shutter speeds (a common hypothesis folks toss out is that IS is getting automatically disabled above some shutter speed-- the data I've seen so far doesn't support that).

pax / Ctein
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com

I read somewhare about when the image stabilization is realy effective. And that help me alot. They say that the shutter speed below which the image stabilization takes effect is relative to the focal lenght. If we have the focal lenght 20mm (for example) then we have good chance to take good pictures at 1/20 shutter speed. Below that speed (but only for 2 or 3 f stops) image stabilization help. In practice I noticed that this calculation work.

The comments to this entry are closed.