I shot about 1200 images last week with a Canon EOS 40D and a Canon EF 600 f/4L IS. On several of the images, I get a very strange effect on the out of focus areas. In the attached image, it is visible in the bushes in the background. It is hard to describe, but appears almost like double images.

If anyone has used the EF 600 f/4L IS that has experienced this phenomenon and has thoughts about how to eliminate it, it would be greatly appreciated. Note that I have also had similar issues with the Canon 1D and EF 400 f/2.8L IS combination, so it's likely not a technical problem with the lens or the body, just a peculiarity of the lens.Click to view attachment

I think it just happens sometimes with OOF specular highlights with the BWL's. You should see some of the weird stuff I get with my 400 f/2.8 with sun coming off cars in the background through chainlink fences. Saw similar effects with the 300 f/2.8 also.

Nill
~~
www.toulme.net

If you could repost a higher quality jpg, it might be easier to comment. The current image is so full of jpg compression artifacts that it is plain impossible to tell anything IMHO.

Regards,
Bernard

Nill, going through some of my images taken many years ago with the EF 600, I have found it to be more common than I originally thought. I wish I knew how to avoid it as it doesn't seem to be something that is controlled with the aperture setting.

Thanks, Bernard. Here is another with one of the questionable areas circled.

That looks different from what I'm describing... that's almost like a funny sort of doubling effect (and not limited to specular highlights, of which I don't think I see any in that shot).

Nill
~~
www.toulme.net

Well, the whole image is very soft. Assuming it's properly focussed, I'm guessing you have some lens shake that is smudging the detail, in addition to the out-of-focus blurring. In other words, you've got several effects going on with the blurred sage.

Here's a better example I took in 2003 with a D60 / EF 600 f/4L IS. The subject is in pretty good focus, and the animal in the background is smoothly out of focus, yet the bushes all have this wierd effect, even though some of them are even further away than the out of focus animal.

That's actually painful to look at. Maybe it's time to try Nikon. ;-)

That is indeed rather strange. Interesting, but strange.

Nill
~~
www.toulme.net

Some lenses just have ugly bokeh. I guess the 600 L is one of them. Nice bokeh is usually associated with lenses that have a larger number of diaphram blades, and where the diaphram blades are rounded.

Hi!

I can see something similar on my Minolta 400/4.5 when used with 1.4X extender. I have no explanation.

Best regards

Erik

It seems to be a factor of large aperture long lenses. A version of mirror lenses "donuts' if you will. I have the 600f4 IS and had the non IS, (in addition to a 300f2.8L). Both 600's have/had this as well as a previous FD400f2.8L. The EF 600's have a 9 blade aperture -- so it does not get much better than that. Heat waves may be a factor in magnifying the effect.

You will not hear about this as much in Nikon long lenses as not very many people use them. (You have all seen the side lines of sporting events the last 10 years).

What type of filter did you use, if any? I've read somewhere (cant remember where) about filters messing up OOF areas - or even the whole image - horribly. This guy took his brand-new mega-telelens on an African safari without testing it first and came back with what you describe. It turned out to be the filter he used.

I didn't use any filter, though there is what I assume to be a clear filter in the filter holder already (has a slide-in filter holder in the rear of the lens). I did check to make sure it was not smugged or dusty.

Why don't you ask the folks who made the lens?

http://www.usa.canon.com/consumer/controll...16&modelid=7320

Oh, yes, I have already tried every source I could think of. I started with Canon, went up to Canon CPS, asked a few regular sports photogs, asked the writers of photo blogs (including Luminous Landscape), all with the reply of what it possibly might be, but no way to pin it down and avoid it.

Since I seem to only really have this problem when the background is highly detailed (like bushes), I'm hoping the solution may still be around the corner with inquiries to the people who read this forum.

Hi there,
I noticed this on my 70-200 IS version, and this effect definitely wasn't there in the previous non-IS version..
Two possibilities at work here:
1) IS creates this effect in the out of focus areas in certain lighting conditions, or more likely in certain types of backgrounds
2) the optical formulas of all these IS lenses with their additional elements have something to do with this..

Try and find some non-IS lenses to check against and see if it's there.. more likely not.

I definitely don't like the effect.. BUT Image Stabilisation is such a great tool that I will just have to live with a wierd OOF area.. The alternative is more often than not an unsharp image..

ihave seen this type of thing, when shooting primarily from 10am to 5 or 6 pm,, in other words, mid day shooting, when heat waves from the ground and plants begin to shimmer, and judging the shadows on your bison shots, you are shooting at miday in may or june,, due the light color of the calf, i also have the same problen when shooting animals on sunny days in the snow, at least mid day shooting when the sun begins to reflect off the snow and,,,,, it can totally ruin a photograph and cause some focusing problems,,, especailly if you are wanting detail from a somewhat distant subject,, thats my opinion and experiecne from that sort of shot,, in addtion, the compression of the lens on the subject and bg compounds the problem,,

I used to get this effect when using an IS lens placed on something steady like a monopod or back of a seat... I hadn't realised that this messes with the image .. you get what looks like a cross between a double image and camera shake.

Ros

Is there any way you can try to replicate the conditions with the lens but shoot on film, to compare, and to see if it has anything at all to do with digital sensors rather than the lenses?

A strange effect that I have seen with very long lenses is a sort of distortion where in, say, a range of heads at an angle to the image plane, the one in focus appears to be smaller than an out of focus one further away from the camera. Very disconcerting, and often visible on tv news shots.

Rob C

I get EXACTLY that effect with my 100-400L at times. Too often actually and I don't use a filter.

The suggestion to compare against film is an interesting one. Or even the IS possibility.

But unfortunately I don't know what causes it and Canon was no help either.

It looks just like heat waves to me. In the bison photo, look just above the fallen tree, in front of the bison. There's a clear little blob there.

I've seen this in numerous photos made with this lens (and with the 100-400 L). It's a 'bad bokeh' lens. You can minimize the effect by choosing backgrounds carefully, avoiding bright highlights or backgrounds that are near enough to the plane of focus that objects are semi-recognizable. Or, use a different lens.

Doug Herr
Birdman of Sacramento
http://www.wildlightphoto.com

[that is heat waves, or atmoshere. I see it alot shooting hot day time college football. no way around it except shoot when its cooler outside! your lens is fine, just bad shooting conditions!
chris in Texas

Actually, to me that looks like a bad case of nisen bokeh. It could also be camera shake, but my guess is that is nisen bokeh, which is "double lines" where there should be one thin line (doesn't affect solids, just thin lines). The bison appear to be quite well focused, and do not display the doubling effect you are seeing, and the foreground blurred parts do not seem to display the effect narly as much as the background. It appears that your lens has been overcompensated for spherical aberration, which is largely responsible for this effect. For details, see my website, which should explain this effect:

http://bokehtests.com/Site/About_Bokeh.html

Klaus
--
http://www.bokehtests.com/

Klaus, thanks so much! Finally, an answer to this problem! I regularly shoot with both the Canon EF 600 f/4L IS and the EF 400 f/2.8L IS and have never, never had anyone, including the techs from Canon, tell me what the problem may be.

I'm definitely recommending your wonderfully explanatory website to my colleagues.

Thanks again...Scott

Wow, so you have to choose between chromatic ab and bokeh! Great explanation, thanks. I always thought it was heat haze, although there may still be an element of this.

Actually, I believe spherical aberration and chromatic aberration are caused by different things. Spherical aberration is fixed with aspheric lenses and chromatic aberration by adding anomalous dispersion lenses. You can have perfectly corrected spherical aberration and still have chromatic aberration in your pics.

Klaus
--
http://www.bokehtests.com/Site/About_Bokeh.html

That's a great article Klaus. Just read this thread and you website on bokeh, and out of curiosity shot some pics with my Canon EF 70-300 mm IS f4-5.6 (on a Canon 400D). Hadn't noticed the effect so far, look at the reeds in front of the cat. Do you think IS has anything to do with it?


Gerard Kingma
www.kingma.nu

Click to view attachment

Click to view attachment

Well, I can't completely say that IS has nothingt o do with it, but I doubt it. THe artifact you have on that particular lens is in the foreground, and the background looks better to my eye, although it would be nice to have a deeper background with either a very bright spot or some reeds prominent in the background to say for sure. I would THINK that if IS was the culprit, you would get the same effect on both sides of the cat. If spherical aberration is the culprit, then you might have a real NICE blur to the background given that foreground blur. How does it do on difficult backgrounds... smooth nice bokeh or harsh nisen bokeh?

If it IS related to IS, then it is an interesting phenomenon. Back on the days before CT scans in medicine, sometimes to get a good image of a structure that doesn't show up well on x-ray (like a kidney), they would sometimes rotate the x-ray or patient around the axis of a kidney while taking the picture. This meant the only thing minimally moving would be the kidney, which would be "in focus" and everything surrounding it would "blur" because it moved more. I just can't remember exactly what the radiological technique was called. ?tomogaphy? This certainly could be going on with the lens as well, if you are moving, but the camera is keeping focus on the cat. BUt you would expect the same both back and front. The easy way to check if hat is undeed what happens is to place an object in the same location as the cat was, set up your cmera and lens in the same place, but on a tripod, turn IS off, and shoot again. If you get the same pattern, it is because of your lens' bokeh characteristics, and if it is gone, then you know it is due to IS.

Just some thoughts.

Klaus
--
http://www.bokehtests.com/Site/About_Bokeh.html

For simple geometric reasons, an image stabilizer (no matter whether it's in-lens or in-body) can eliminate camera shake only at the plane of focus. For out-of-focus parts of the image, the same IS action that reduces shake at the plane of focus will amplify shake elsewhere. Out-of-focus areas are blurred anyway so adding some shake-induced blur will hardly hurt ... usually. But unfortunately, defocus-induced blur is not the same as shake-induced blur; the two kinds of blur have different "textures," so adding them can result in very odd-looking out-of-focus rendition, or bad bokeh. So an image stabilizer at work actually can affect bokeh (and not for the better). Long lenses seem to suffer from this effect more than shorter lenses.

Furthermore, due to residual spheric aberrations most lenses have their nicest bokeh not at full aperture but at, say, half or one f-stop down from the widest. This is particularly true for fast lenses.

So in order to reduce the unpleasant bokeh I'd suggest to switch off IS (use a tripod instead) and to stop the lens down by an f-stop. If a tripod is not an option then maybe reducing camera shake through faster shutter speed (higher ISO setting) or through a monopod can also help, to a degree at least, as it will reduce the stabilizer's workload. If that doesn't help then I'm afraid you'll have to switch to a different lens.

-- Olaf

Klaus, very nice and enlightning explanation. Good work! Only one issue: At the end of that page, you seem to confuse perfect correction of spherical aberration with perfect bokeh. However these are two different concepts.

Perfectly corrected lenses usually will produce neutral circles of confusion, i. e. those with equal brightness across the whole area, and fairly sharp edges. However, neutral COCs will render an image with a bokeh that's only so-so ... not bad but not really nice either.

Lenses with slightly under-corrected spherical aberration will produce circles of confusion for background objects that are brighter at their centers and rolling off smoothly to their edges. COCs of that shape will render an image with perfect bokeh. However under-corrected lenses usually are not the best performers.

So perfect bokeh and perfect correction for aberrations are two different things. It's not impossible to design high-performance lenses which still have nice bokeh. But that requires the lens designer to take bokeh into consideration ... as if lens design wasn't complex enough already. So often, much effort goes into making the lens sharp and then the bokeh falls just where it may.

A lens maker who always makes their lenses not only sharp but also provides them with nice bokeh is Leica. Minolta also used to do this. A company that always corrected the hell out of their lens designs but hardly ever gave a damn for good bokeh was Nikon ... but they changed for the better recently.

-- Olaf

Thanks Olaf-

I will reexamine the page you were referring to. What you explain is exactly what I was trying to get across... "ideal" lenses with perfect spherical aberration, while nice and sharp and high contrast, have at best neutral bokeh. I will see where the confusion lies-- most likely in the little image circles saying "ideal" lens, which I defined earlier in the page as not ideal for bokeh, but a lens without "flaws" like spherical aberration. I would be a leica photographer if I could afford it, but now I am firmly in the minolta camp-- use a maxxum 7d. Unfortunately, my sony lenses I own are NOT designed for bokeh. The 70-200 ssm is a FANTASTIC lens, but not completely optimized for bokeh. It is not bad by any stretch of the imagination, but is close to an "ideal lens." I just ordered a Zeiss 24-70 SSM, but what I have seen about it is also not excellent bokeh. Sharper than any other 24-70 out there, but not optimized for bokeh. Some day I hope to investv in the sony 135 f/2.8 STF. Now THERE is a bokeh machine!

As a side note, if IS has anything to do with the double line effect at all, then doesn't that make the Sony and Pentax sensor shift techhnology BETTER in this regard? The same effect should NOT happen with sensor shift, I wouldn't think, unless there is severe distortion on the lens being used.

I would be thrilled if someone would try their IS lenses on a similar scene, the same pic with IS on, and then tripod mounted with IS OFF. That would really settle things once and for all if this is due to properties of the lens optics, or properties of the IS system.

Klaus
--
http://www.bokehtests.com/Site/About_Bokeh.html

PS: I have edited the site ever so slightly to put quotes around the word "ideal" and explained again that all I mean is lack of spherical aberration, and fixed a sentence in the preceding paragraph. Thanks for pointing out the confusion

That's right. The Minolta/Sony 135 mm STF is the lens with the world's best bokeh ever. It uses a built-in and cleverly designed apodization filter to force the circles of confusion into the shape that naturally leads to perfect bokeh. However it comes at a price: even though the geometric lens speed is 1:2.8, the effective speed is only 1:4.5 due to the transmission loss in the apodization filter. So the lens is f/2.8; T/4.5. And it is manual-focus only even though it's made for Minolta A mount/Sony Alpha mount.

The Nikon DC lenses ('De-focus Control') try another approach to optimize bokeh. They allow the user to adjust spherical aberration through the range from slight under-correction to slight over-correction. So depending on the setting, the DC lenses will get optimized either for background bokeh or for foreground bokeh---but not both, unlike the STF. Also unlike the STF, they don't suffer from transmission losses.



No. As I already said in my previous post, the problem is the same no matter whether the image stabilizier is built into the lens (moving lens elements) or into the body (moving sensor).

Update: I just fired a few frames, to test if the image stabilizer really affects bokeh to a perceptable degree. I used a Konica-Minolta Dynax 7D (APS-C format, 6 MP, in-body image stabilizer) and a 560 mm 1:6.3 lens (Minolta AF Apo 400 mm 1:4.5 G with 1.4× TC). The focus distance was approx. 35 m/100 ft; the background had many sharp edges and specular highlights (cars in a parking lot) and was approx. 100 - 120 m/330 - 400 ft away. I tried f/6.3 at 1/500 s and f/16 at 1/80 s. For the hand-held shots, I was wielding the camera up and down vigorously, to give the stabilizer a real work-out. Still nearly all shots are sharp, even those taken at 1/80 s. For the tripod shots, I switched the stabilizer off. Result: ummm. At first sight: no difference. At a closer look: the hand-held shots show a tendency for harsher bokeh and hints of double lines, and the tripod shots don't ... but that's at the borderline of perceptability; it takes a very close look to notice. So I guess it will take a few more tests to settle this issue, using different combinations of focal lengths, apertures, shutter speeds, focus distances, and background distances.

-- Olaf

Well, first off, I am exceedingly jealous of your lens collection. Would LOVE to have a huge lens like that! Second, it is good to know that even on such a massive lens, the image stabilization on a Maxxum 7d is so effective... less effective than VR or IS? Don't know, nor have I seen a very convincing TEST to show this. That is, a test that is relatively well controlled and done by someone who does not start out with the answer they are trying to find.

I have thought through the IS/VR and in body image stabilization, and after thinking about it a bit, yes, it should happen with both systems about equally (I may make some photoshop diagrams to illustrate this point). Never thought about it, but the double lining seems, at least in theory, to be able to be caused by an IS/SSS system itself, on top of that produced by spherical aberration. The being the case, maybe I should use that tripod more often than I do!

Thanks for setting me straight.

Klaus
--
http://www.bokehtests.com/Site/About_Bokeh.html

Update: I have added the beginnings of a discussion of this Image stabilization bokeh effect on my website. Will add more in a few days when I get a chance.
http://bokehtests.com/Site/Stabilization_and_Bokeh.html

Hello
I did contact Nikon support if they know about this kind of effect (using their super tele VR lenses). They replied (very fast !!) that VR in generally has nothing to do with this effect, if the right VR mode is set. They further said that the effect may be caused by the specific lens characteristic (combination of different glasses and materials, polish characteristics) and can happen with "oversharp" lenses (whatever that means...)

Obviously they didn't fully understand the question.

-- Olaf

I think they did, the question was very clear. So just for you Olaf the question in original German as sent to Nikon :-)

Frage:
als Natur- und Tierfotograf interessiere ich mich vor allem für lichtstarke Telebrennweiten und hier besonders für die Nikon Optiken mit VR Technik (AF-S VR NIKKOR 600 mm 1:4G ED und AF-S VR NIKKOR 600 mm 1:4G ED).
In dem bekannten Web-Forum von Luminous Landscape habe ich nun einen Thread gelesen, dass die Bildstabiliserung einen schlechten Effekt auf die Abbildungscharakteristik haben kann, der teilwiese so ausgeprägt ist, dass der Hintergrund ein extrem schlechtes Bokeh aufweist. Im Forum geht es allerdings um Canon Objektive mit IS Technik. Da nun Nikon genau wie Canon bewegliche Linsen verwendet, frage ich mich, ob bei Nikon auch so ein Effekt zu befürchten wäre. bzw. ist er Ihnen überhaupt bekannt oder können Sie sagen, dass der im folgenden Thread diskutierte Effekt überhaupt nichts mit der Bildstabiliserung zu tun hat:
Link auf diesen Thread:


Answer:
Bei richtiger Nutzung des VRs des Objektives (Stellung Tripod für Fotografie mit Stativ) sollte keine Verdopplung der Hintergrundstruktur auftreten. Der genannte Effekt tritt auch bei Objektiven mit überscharfer Abbildungsqualität auf, auch bei Objektiven ohne VR und ist unter Anderem abhängig von den verwendeten Gläsern, Schliffen und Kombinationen..

Obviously, they did not understand the question at all. They have no idea what you're talking about. They think you're talking about shake-induced blur and whether the stabilizer itself can create (rather than defeat) blur. They also refer to over-corrected spherical aberrations which is a different kettle of fish entirely.

Und noch einmal auf deutsch: Sie haben deine Frage ganz offensichtlich überhaupt nicht verstanden. Sie wissen gar nicht, wovon du sprichst. Sie glauben, es ginge um Verwacklungen und ob der Bildstabilisator selber welche verursachen könne. Außerdem beziehen sie sich auf überkorrigierte sphärische Aberrationen, was eine ganz andere Baustelle ist.

-- Olaf

Yes you are right. Indeed Nikon (you have to be a registerd Nikon user to send emails to their tech support) replied very fast (within one hour) so I do not think they took the time and followed this entire thread. Furhtermore they mixed the terms "tripod mode" from the D3/D300 Live View AF mode and the option for panning with VR lenses (only vertical blur correction). But as long as nobody has reproduced this bokeh effect (Stabilizer On/Off) it is not clear if IS/VR/AS can have such an effect as shown on some of the sample pictures here (especially the deer photo)

I have thought it over and I think VR can not have such an effect on bokeh. In theory (geometrical) there is a perspective shift due to camera shake (you do not shake with the entrance pupil location as the center of the shake). But the shift of background plane (defocused area) during exposure is neglibly small, too small at all to cause such an effect.
The graphics on http://bokehtests.com/Site/Stabilization_and_Bokeh.html are extremly exaggerated in scale. Keep in mind the sample picture of the deer : 600mm lens; shooting distance was maybe 70 meter (+ background additional 10 metres).
Furhtermore if VR has this shift effect to the defocused areas the effect would cause some kind of motion blur and not "double lines". The effect that was shown here in sample pictures occured with super tele lenses and tele zooms. Most telezooms show some kind of over corrected spherical correction in the tele range and together with the high amount of defocusing effects (which the human eye does not notice in normal vision) the "bad bokeh" appears. So in the end Nikon answer was OK.

Hi Mikael-

As the author of the exaggerated graphics that you referred to, I have to say, you are correct... I massively exaggerated the picture to illustrate the point, not to show EXACTLY what happens in the lens. Anybody who can't hold their lens steady to the point that it shifts 45 degrees each way is going to have some blurry pictures! BUt the point is still valid. Just as you wouldn't shake as much as that, you also would not have the image shifting from the left side of the picture all the way to the right. BUt the question is, what if there is a 1 degree shake, or a 0.1 degree shake. How much image shift would there be due to parallax? I intended to figure this out but then went on to bigger and better things, and have still not made it back there. ONe thing is sure-- the distance a pixel will travel due to parallax and a 0.1 degree shake will be much more on 600 mm lens than a 24mm lens.

As for your other point, I also agree. A unidirectional shake will produce a motion blur. But this will only occur if there is a single direction shake... if you shake from left to right ONLY during a photo. The minute the lens moves from one direction to the other, you will get a longer exposure for the object right at the point when directions change. and if you change multiple times, you will get two or more bright lines each time you change directions during your shake. So it is less important, or completely negligible, if your shutter speed is very fast. But if your shutter speed is slow enough to involve at least one change in direction you end up with a motion blur and bright spot. If you have time to change directions in your shake twice, you get the double line effect as I described on my website. Think of a very fast strobe light doing multiple exposures on a pendulum.... when changing directions at the edges of the trajectory the images will get closer and closer together. The faster and faster the strobe becomes, the closer you get to a real long exposure, and you will see images bunch up on each other at the edges. THis is the same as I am talking about, only the LENS is the pendulum.

The effect of VR is just that instead of having a big motion blur evenly throughout the image, the "in focus spot" is stabilized to a certain part of the picture. The effect described above will be more amplified the further away from the "in focus" spot is.

Now, as I have said many times, I am not a lens designer, a physicist, an optical engineer, or even a professional photographer. I am just an amateur photographer who finds this stuff very interesting, who has been burned by bad bokeh before and wants to understand it. So I may be wrong here. But I doubt it. Parallax is definitely a REAL phenomenon, and should affect ANY image with ANY lens that is rotating around any point other than the entry pupil, and should also be there with non-rotational movements as well. The reason it affects BOKEH in image stabilized lenses is that motion blur in the in focus area is stabilized. BUt again, it depends on how fast you shake, how long the lens focal length is, how long your shutter speed is, etc. And while I have very little doubt this phenomenon exists, the question is-- is it big enough to be relevant? Perhaps the shake when translated to the image only accounts for a motion of 0.0002mm at the sensor plane.... well, you won't be seeing this effect then.

If anyone can do the math that would be great. I can do it, but it will take time because I will have to figure out how to do it. If anyone knows and could tell me or do the calculation themselves, let me know!

Sorry for yet another very wordy reply.

KLaus
--
http://www.bokehtests.com/Site/About_Bokeh.html

Hello Klaus
I did not intend to critisize your graphics as to be exaggerated; that is what they were intended to be, to illustrate the theory of the VR effect.
Mathematically there is an effect, but I think other defocussing effects and lens specific bokeh characteristics are far more predominant. I really tried to reproduce the effect using a 400 mm lens; I took a lot of photos and tried to get as close as possible to the shutter speed where the IS just could compensate the shake (1/60 s) and compared that to the photos taken from tripod. Both were sharp on the plane I focused at and had exactly same bokeh.

So if someone could do the same, maybe we will have proof that the effect exists; I doubt it.

Exactly that is the question indeed.

I tried to compute the image shift due to parallax, and I came up with a very complex formula which is hard to compute due to numerical issues, as numbers of vastly different orders of magnitude are involved.

Anyway, my results so far indicate that image shift due to parallax generally is pretty small generally but under certain circumstances can be the same order of magnitude as the pixel pitch of a typical DSLR camera. So it can be up to a few pixels which definitely has the potential to affect bokeh.

For example if you're shooting a subject at a distance of 20 m/65 ft with a 600 mm lens, and if the background is at a distance of 25 m/80 ft, and if your lens revolves by 0.5° during the exposure around a pivot point 10 cm/4" in front of the entrance pupil then the background's image will get shifted by approx. 0.005 mm more than the subject's image. With a 12 MP APS-C camera this corresponds to approx. one pixel. The shift will increase (approx. proportionally) with the distance between subject and background, with the distance between entrance pupil and pivot point of the shake (this can be pretty long with telephoto lenses), and of course with the shake itself; it will decrease with the distance to the subject (i. e. more shift difference at close range).

-- Olaf

Maybe on lens based IS systems the shift of the lenses itself has general negative impact on the lens performance, which also cause bokeh to get worse than when turned OFF or when compared to sensor based IS cameras.

Hi Olaf-

So I have been thinking about this for the last couple hours.... I am starting to change my idea of how this all works. I figured you seem pretty knowledgeable, so I thought I might pose this here and ask for some help.

My understanding of how IS works was the same as yours- you could only stabilize a single plane of focus, and the rest would not be as affected by image stabilization. Now I am not so sure this is correct. If you imagine a rotational motion blur on a non-image stabilized lens (say you took pictures of point sources at 10m, 10m, and 1000m), wouldn't you see a streak of light that would be EXACTLY the same distance for all three points of light? I think the motion blur will be exactly the same, as long as you are rotating about the entrance pupil (otherwise how does a panoramic tripod work-- you rotate it a bunch of degrees and the foreground and background objects move the same distance on the sensor plane). My guess is that the IS systems work by assuming that the rotational axis of your shake is AT THE ENTRANCE PUPIL. This means that the IS system (or SSS) can stabilize the ENTIRE IMAGE at ALL focal planes. This would be much easier to implement by a lens manufacturer. Of course, there will be very slight parallax shifts because you will not be rotating precisely around the entrance pupil, but the effect should not be that important, at least not on anything other than a very long focal length lens.

Is this incorrect? As for this topic, I had previously stated that if this effect exists at all, then it could be explained by parallax error. It turns out that after thinking about it, IF it truly can only stabilize a single focal plane, then there is a much bigger effect than parallax that would explain the phenomenon that people are reporting here. I still think the answer to this question is most likely that the changed optics of the IS versions of their lenses just overcorrected spherical aberration and led to a bad case of nisen bokeh.

Let me know what you think.

Klaus
--
http://www.bokehtests.com

That's right. The problem arises when rotating around some arbitrary pivot point which is not identical to the entrance point. This would introduce some error into the system ... and the point is: the error will vary for the different subject distances. You'd get a different error for every distance.



This is what I am thinking, too.

A camera has six degrees of freedom to move in three-dimensional space. There are three axes, and for each axis the camera can do two things: move along and rotate around. Current image stabilizers sense, and compensate for, only two of the six possible movements: rotations around the lateral and the vertical axes (i. e. pitch and yaw). Rotations around one pivot point can be expressed as a linear combination of rotations around some other pivot point plus shift movements along the axes. So neglecting shifts basically is the same as assuming the lens' optical center, i. e. the entrance point (which is where the entrance pupil and the optical axis intersect), as the pivot point indeed.



No, it doesn't. All parts of the image will get stabilized evenly only when the pivot point of the shake actually is at the entrance point.



Well ... in my previous post I explained that, when rotating around a pivot point which is not the entrance point, the difference between the parallaxes for different subject distances can exceed the pixel pitch of a modern DSLR camera and thus, actually might affect bokeh.



Of course, it is possible that other effects besides parallax errors also may affect bokeh. Optical image stabilizers work by deliberate decentering of the lens---which typically does not exactly improve the optical performance. It's possible that an IS lens in decentered state shows a bokeh that's different from the bokeh in centered state. It also is possible that this phenomenon might affect bokeh to a more significant degree than the parallax errors. The two phenomena also may add up and emphasize each other.

Just speculating ...

Perhaps I should point out that the different bokehs at centered and decentered states (if they actually are different in the first place) is a concern with in-lens image stabilizers only, while the effect of parallax errors concerns all types of image stabilizers, in-lens and in-body.

-- Olaf

hmmmm. Intersting. I have some stuff to think about and try to understand. Sorry if this rambles a bit. Thinking about the way that we both, I think, assume that optical stabilization works (I agree that only 2 degrees of freedom are likely compensated for), while it doesn't TECHNICALLY stabilize the whole image plane, in reality it pretty much does. If you were to mount a lens on a swivel mount attached directly to its entrance pupil, and do whatever gyrations you might want, the optical stabilization would stabilize the WHOLE image, no? It would seem then that ALL of the difference between how the in focus object is affected by the stabilization system and how out of focus objects are so affected would be due to PARALLAX.

If we assume that the IS system presumes rotation about the entrance pupil, there is no need for distance input from the camera, so it attempts to stabilize the whole image. And, it should be pretty successful at it, I would think. Since parallax will affect the whole image at EVERY distance (albeit greater for close objects and less for distant objects), this should pretty much stabilize the whole image, plus or minus a few pixels here or there that will be slightly different for close up and background images? We should not be talking about a single plane of focus that is completely rock stable, and then some areas outside of that plane that widely gyrate... merely a few pixels here or there, right? Similar to the amount you had calculated previously in one of the above posts.

If this is true, then optical stabilization should work much better (a few pixels less blur) on distant objects than close objects. Moreover, this may not only affect the bokeh of a shot, but if this is all correct, it may actually affect the sharpness of the image, and provide a maximum allowable sharpness for image stabilized lenses that will be different depending on how far from the camera the object is. Again, the effect should be the same if it is SSS, or IS/VR.

Also, one thing I wondered about that you had said above. You mentioned that the centering of the lens changing should only affect IS/VR type lenses and not sensor shift technologies. There is something different, however, that can affect sensor shift technologies, as well, and it is pretty much analagous, I would think. When the sensor shifts, the light path through the lens that is hitting a particular sensor pixel will be DIFFERENT than it was before the shift. Since different areas of lenses may have different bokeh characteristics, this should be able to affect bokeh too (an example would be the poor bokeh performance of the sony Zeiss 24-70 SSM lens that seems to have really quite acceptable bokeh when focused at 10 feet in the center, but extremely poor bokeh at the same distance on the sides of the lens). If you don't believe that lenses have different bokeh in different areas of the lens, see my tests of the Zeiss mentioned above at http://www.bokehtests.com/Site/Sony_Zeiss_24-70_Bokeh.html.

Thanks for the time and discussion. I am learning a lot.

Klaus

I think it would indeed, as long as you don't exceed the stabilizer's working range. However hand-held camera shake does not generally pivot around the entrance point exactly.



That's the question. I can hardly think of any other effects (besides varying lens bokeh) ... but that doesn't mean too much as I am no expert. I am only rambling, just like you do.



Good point. However as far as I know, image stabilizers do take distance information into consideration ... so I am a bit confused.

Distance information is good for two things. First: Geometry of the imaging. Second: Properties of the lens. Lenses do change their properties according to focus distance, namely effective focal length (and a few other parameters). Question is: Does the image stabilizer know which lens it is stabilizing? Well, in-lens stabilizers most likely do . Do in-body systems adapt to the particular lens model? Another question is: Does the stabilizer adapt to the focus distance as such? If so then there is something going on you and me are not aware of.



Right. In most cases, the difference between stabilization of in-focus and out-of-focus parts of the image are minuscule; in some cases they may affect the bokeh to a perceivable degree.



Actually, it does! Image stabilizers don't work well in the macro range. The main reason for this is the fact that the effects of the camera movement's other (uncompensated) four degrees of freedom become more significant at close range.



Yes, but still it was there in the image (albeit elsewhere). So it's not really analoguous I'd think.



This is true only at full aperture. Most lenses have not-so-good or even poor bokeh at their full apertures. Stop the lens down by one f-stop or even just half an f-stop, and in most cases bokeh will improve considerably.

-- Olaf

I would think that sensor shift technologies must also know to some extent what lens they are stabilizing... after all, the distance to shift the sensor would be very different for a 12 mm lens that moves 1 degree and a 1200 mm lens that moves one degree.

As for most lenses having better bokeh at a few stops past wide open aperture, I have read that from several sources, so it must have a solid basis. My testing of the Sony SAL 24-70 lens, however, does not show that so much, or rather, the story is much more complicated. It loks like the bokeh produced at the edge of the picture DOES improve, but the bokeh in the center of the lens actually gets slightly worse when going from f/2.8 to f/4 or even f/5.6. So it seems to be very lens dependent whether Bokeh gets better or worse when stopping down. but so far that is the only lens I have put through any kind of rigorous testing... but I wouldn't be surprised to see things like this in multiple lenses from any lens line.

Technically, also, if a lens has "good bokeh," that is, specular highlights having a less than sharp border and no ring artifact, one would expect the lens to have WORSE bokeh stopping down. Since spherical aberration takes place at the edge of the lens, cutting the edge off with an aperture blade will tend to make it more like an "ideal" lens with a complete lack of spherical aberration... in other words, bokeh will get more neutral. For lenses that are overcorrected for spherical aberration, stopping down should eliminate some of the ring artifact, and thus produce better bokeh, but for lenses with good bokeh stopping down should produce worse bokeh. Obviously, this is a large generalization, however. If my understading is wrong, I would love to be corrected!

Klaus
--
http://www.bokehtests.com