Does Image Stabilisation adversely effect image sharpness?

[Mads S]

Hi Grumpy,

I read Thom's article, and I read your post, and I think you guys are saying the same thing... So I don't see why 'it is unfortunate'?
My intuitive reservation re. IS/OS would be the actual moving of the lens (group) in countersync with vibrations - I think it is 'clear' that if you're moving lens glasses around while exposing then you may get some blur. I also concede that it will be less than the blur you are causing by moving the whole thing around (camera shake) but I think this discussion has concluded that if you can 'eliminate' camera shake, then you're better off not having lenses move around on their own, or even just being loose enough that they CAN move.

Does anyone have a Canon EF L 400mm f5.6 lens for sale? It doesn't have IS...

Mads S

[Manfred M]

Hi Grumpy,

I read Thom's article, and I read your post, and I think you guys are saying the same thing... So I don't see why 'it is unfortunate'?
My intuitive reservation re. IS/OS would be the actual moving of the lens (group) in countersync with vibrations - I think it is 'clear' that if you're moving lens glasses around while exposing then you may get some blur. I also concede that it will be less than the blur you are causing by moving the whole thing around (camera shake) but I think this discussion has concluded that if you can 'eliminate' camera shake, then you're better off not having lenses move around on their own, or even just being loose enough that they CAN move.

Does anyone have a Canon EF L 400mm f5.6 lens for sale? It doesn't have IS...

Mads S

Actually what he and I are saying are quite different, certainly in how IS technology works. Many of his conclusions are based on some erroneous assumptions on regarding the technology and he extends that to some of the conclusions he draws. It appears to me that some other issue is in play because what he is reporting because it cannot be caused the IS, unless it is malfunctioning. I personally have a hard time with anecdotal evidence, and that is unfortunately a lot of what I read in his article.

All that being said unless you have a super high-resolution camera and are pixel peeping, I don't think you will see any difference. If you want the cleanest sharpest, most distortion free picture, then go for a fixed lens and shoot from a heavy duty tripod.

[William W]

I also remain unconvinced by the general conclusions of the Thom Hogan article which is cited in this thread.

Although unconvinced by the arguments because of the assumptions and therefore inconclusive misinformation touted as bona fide conclusions and rules: I shall limit the following comments to practical experience.

Specifically – Hogan cites a 400mm lens with VR – (and quotes IS is the equivalent in the Canon Range).

Apropos shooting Field Sports – I have used extensively the following lenses so to do: EF400/2.8L IS USM; EF 300F/2.8L IS USM; the EF 70 to 200F/2.8L USM and the EF 70 to 200F/2.8L IS USM and whilst my staple short lens is the NON IS version 70 to 200, I DO USE the IS in the “ON” position for shutter speeds above 1/500s on both the 300 and 400 lenses: even if the 400 is mounted on a monopod.

I can only speculate that, as I now shoot a lot of Swimming and in the past I have shot many Field Sports . . . my cohort of Photographers are “Not Real Professionals” according to Thom – because using IS on both the 300 and 400 is standard practice and so too is using the 300 Hand Held and the 400 on a pod – and shutter speeds are always (typically) faster than 1/500s.

Certainly my experience with the 70 to 200 NON IS lens, set at FL = 200mm and hand held – 1/1000s is NOT that fast and IS certainly benefits the Photographer at that shutter speed if the Lens is Hand Held: and I note that I am no slouch when it comes to hand holding at slow shutter speeds.

But the reality is, IF one looks hard enough there will be appreciable camera movement noticeable at the Tv Range 1/200s ~ 1/1000s with an Hand Held 200mm lens and on the balance of it, IS will be better “ON” than off . . . ergo at FL =300mm and FL =400mm IS will be an obvious bonus for hand held or monopod shooting. (Comments are in respect of using the lenses on a "full frame" camera).

And these facts and my experience and the shooting techniques and experience of others in my cohort led me to make this comment in Post #7

"Apropos “fast shutter speeds” . . . I suggest do some tests hand holding a 200mm lens and using 1/1000s and look at 100% 150% and 200%. – 1/1000s is not that fast for hand holding a telephoto lens, IF the file / negative is to be under close scrutiny."

Pertaining to this point - this link contains an extract of some of the in field testing using the EF 70 to 200F/2.8L IS USM and the EF 70 to 200F/2.8L USM.

The three selected cropped images are shot at: 1/250s; 1/500s and 1/1600s and each is the typical indication of the Camera Shake recorded when shooting Hand Held at 200mm.

Obviously some shots were "better"; but also there were some shots also “worse”.

The in field tests I have conducted lead me to set a shutter speed of 1/1600s as my typically ‘Best Practice Safe Shutter Speed’ when Hand Holding a 200mm lens on a Full Frame Camera. Obviously, I do use slower shutter speeds when hand holding: but I use 1/1600s as my "Best Practice" if I so can do.

But the point is, if we think of the adage of setting a Shutter Speed equal to 1/FL (seconds), whist that being a Rule of Thumb, such a Rule is not at very convincing when one gets to the longer Focal Lengths of the larger, physically longer and heavier lenses, for example, think of hand holding the 600/4L and shooting at 1/640s and expecting great results without using Image Stabilization.

In my experience and to the point of common sense, even contemplating the scenario of using an EF 400/2.8; (or 500/4 or 600/4) HAND HELD at shutter speeds of 1/500s or a tad faster (let’s say to 1/2000s) and NOT using the Image Stabilization seems quite ridiculous . . . but Thom writes:

“Rule #2: VR should normally be off if your shutter speed is over 1/500.

Indeed, if you go down to the sidelines of a football game and check all those photographers to see how their lens is set, you can tell the ones that are really pros: VR is usually off (unless they're on a portion of the stadium that is vibrating from fan action). Those pros have all encountered the same thing you will some day: if you have a shutter speed faster than the sampling frequency, sometimes the system is running a correction that's not in sync with the shutter speed. The results look a bit like the lens being run with the wrong AF Fine Tune: slightly off.

The interesting thing is that pros demanded VR (IS in the case of Canon) in the long lenses, then it turns out that they very rarely use it! I'd say that less than 10% of the shooting I do with my 400mm f/2.8 has VR turned on (and by the way, I hate the rotating VR switch on some of these lenses--it's so easy to not notice what position it is in.”

[Extract: op cit]


***


Maybe VR (Nikon) is different to IS (Canon).

But the Thom Hogan Article makes no such consideration: but rather in fact he lumps the two systems together in his critique.

However, please note my comments above relate ONLY to Canon Images Stabilization and Canon Lenses.

WW

[Mads S]

Hi Bill,

Thanks for your input. I agree that in many conditions/circumstances the benefits will outweigh the drawbacks. However, should you manage to 'eliminate' shake then I believe you will get better results without VR/IS/OS. I will personally work on that :-)

Mads S

[William W]

I agree that in many conditions/circumstances the benefits will outweigh the drawbacks. . .[if we manage to] 'eliminate' shake then I believe you will get better results without VR/IS/OS.

Mads,
Thanks for your kind words.

***

There are four points I’d like to (re)iterate regarding the underlined section quoted:

Firstly, (extreme) ‘elimination’ of camera shake is commensurate with using a solid tripod; a centre pole counter weight; sandbags on the tripod legs; remote release; mirror up and a light weight sandbag on the top of the camera / lens assembly.

Secondly, apropos Canon, the most recent issues of Image Stabilization has a tripod detect mechanism (ref: CS Post # 16 this thread).

Thirdly, Thom Hogan uses Nikon Gear. Although he does to lump “VR” being equal to “IS” in all manner and fashion, (AND that is another of the many assumptions he makes in his article.)

Fourthly, your conclusion also follows point three, above: i.e. you assume “VR” = “IS” = “OS”.

WW

[Ady]

I saw someone refer to these devices as “electronic gyroscopes” in an earlier posting. This is actually a bit backwards; 3-axis accelerometers can be used to build electronic gyroscopes, but not the other way around.

I'm no engineer so can't add anything sensible with regard to this but I was also under the impression that the guts of IS utilised gyroscopes on the basis that Canon says:

"The Image Stabilizer in a Canon lens works by shifting a stabilizer group of elements to compensate for motion detected by a gyroscope in the lens."

Have I misunderstood the references to gyroscopes or have Canon have over-simplfied to the point of misrepresentation?

Cheers,
Adrian

[Manfred M]

I'm no engineer so can't add anything sensible with regard to this but I was also under the impression that the guts of IS utilised gyroscopes on the basis that Canon says:

"The Image Stabilizer in a Canon lens works by shifting a stabilizer group of elements to compensate for motion detected by a gyroscope in the lens."

Have I misunderstood the references to gyroscopes or have Canon have over-simplfied to the point of misrepresentation?

Cheers,
Adrian

I am an engineer and have done some work in integrating control systems into equipment. I suspect Canon is using language that they feel a layman is more likely to understand. People know what a gyroscope is, but not that many are likely to have heard of accelerometers.

When we hand-hold a camera, the correctable image motion involved can be measured in millimeters or fractions of a millimeter. When it comes to camera shake, there are three axes of motion, relative to the camera sensor:

1. Horizontal, x-axis, i.e. side to side;
2. Vertical, y-axis i.e. up and down; and
3. Back and forth, z-axis; i.e. straight through the centre of the lens.

The first two will affect image quality if they are not corrected.

The third one is a non-issue and is really related to focus, not camera shake, and at the distances involved is negligible. These are subject to camera distance errors would show up as focus, rather than motion issues. The errors introduced by manual or auto focus would be greater than anything that we could effectively correct through stabilization; remember we are looking at distances measured in millimeters or fractions of a millimeter. Think of it as a depth of field issue, that is below the threshold where it can be detected.

This is why I say that no gyroscope is used. No manufacturer would waste money on components or processing power to measure for three axis movement, when only two will ever be used.

I hope I've clarified, rather than confused the issue.

[William W]

I suspect Canon is using language that they feel a layman is more likely to understand. People know what a gyroscope is, but not that many are likely to have heard of accelerometers.

I also suspect the language is watered down from technical to layman.
Also it is important to note where and for what purposes this "gyroscope" languge is used.


(I was writing this below, as you were writing your response above . . .)



Canon advertising uses this sentence in many descriptions of their lenses and usually under the heading “Key Technologies”:

"The Image Stabilizer in a Canon lens works by shifting a stabilizer group of elements to compensate for motion detected by a gyroscope in the lens."



This is usually the full paragraph used by Canon:

“The Image Stabilizer in a Canon lens works by shifting a stabilizer group of elements to compensate for motion detected by a gyroscope in the lens. There are two IS modes – Mode 1 for general shooting and Mode 2 for panning. This lens has an Image Stabilizer that gives up the equivalent of four stops of shutter speed so you can hand hold and capture acceptably sharp images.”


It occurs to me that the above paragraph is a simplistic and rather non technical encapsulation of the functionality and uses of Image Stabilization and mainly for advertising and or marketing purposes.


***

Perhaps a better description of IS is to be found here.


** *

Also, a detailed outline of Image Stabilization can be found in “Lens Works III - The Eyes of EOS” Fifth Edition, Sept 2004; Publishers and Editors Canon Inc, Lens Products Group;

Reference - Chapter, ‘Lens Technology’, Section 15, pp203~205.


WW

[Ady]

Manfred and Bill, I wouldn't for a moment suggest I am now an expert but I am at least pointing in a more appropriate direction, thanks to you both

Cheers,
Adrian

[Dave Humphries]

I don't disagree with where Manfred and Bill (mainly) have taken this so far.

However, to muddy the waters - sorry guys

I'm sure I saw something, somewhere with a manufacturer boasting five axes of stabilisation, but I can't remember where now.

To add to Bill's three above, I can only think of;

4. Rotation about the lens centre (which could happen)

I recall the fifth was another kind of rotation, but I don't see how that works without being detectable/correctable by the other four - in the practical world.

If you go to extremes - you start to needing six accelerometer sensors, placed one at each end, each side and top & bottom of the lens/camera in order to differentiate rotational components from the linear ones in the three axes. Then you'll need five or six, not two or three, 'motors' to move the element group to correct these most accurately - but the law of diminishing returns probably kicked in sometime ago though

Pretty soon, you'll have designed the orientation system of a space craft

[carloshpvp]

I see that VR or IS or OS are more likely to appear on mid-long lens. Is it dispensable in a lens in 17-50mm range?

I ask this question because I'm trying to decide about getting a Tamron 17-50mm f/2.8 for a D7000 (Nikon DX, crop 1.5x) and there are VC (Tamron acronym for Vibration Compensation) and non-VC version of the lens. I've seen some people claiming that the non-VC has better optics then the VC version. The equivalent lens from Nikon, Nikkor 17-55mm f/2.8 doesn't have VR,

[Sponge]

I'm sure I saw something, somewhere with a manufacturer boasting five axes of stabilisation, but I can't remember where now.

Could you be thinking of the Olympus OM-D E-M5? It has 5-axis IS but it's in-body so not sure how it compares to a lens IS/VR/OS system

[Dave Humphries]

Could you be thinking of the Olympus OM-D E-M5? It has 5-axis IS but it's in-body so not sure how it compares to a lens IS/VR/OS system

Yes!

Thanks Patrick

... and here's the review with the diagram I was trying remember/describe in words; http://www.dpreview.com/reviews/olympusem5

Cheers,

[William W]

However, to muddy the waters - sorry guys
. . . Pretty soon, you'll have designed the orientation system of a space craft

haha! - yes the question becomes 'is it: "technology for technology's sake" ?' - that is always an interesting conversation.

I am not technically au fait with the Olympus OS System. But, as already mentioned, it is worthy to note the Olympus stabilization is in the camera and not in the lens: i.e. the system is ‘stabilizing’ the camera body for light rays which have already passed through the lens.

Canon also has Hybrid Image Stabilization – which accounts for camera shift (parallel movements to the Image Plane). Camera Shift Movement is shown by the green arrows in the Olympus Diagram Dave supplied in the dpreview link in Post #33.

Hybrid IS is used in the EF100F/2.8L Macro Lens.

WW

[William W]

I see that VR or IS or OS are more likely to appear on mid-long lens. Is it dispensable in a lens in 17-50mm range?
I ask this question because I'm trying to decide about getting a Tamron 17-50mm f/2.8 for a D7000 (Nikon DX, crop 1.5x) and there are VC (Tamron acronym for Vibration Compensation) and non-VC version of the lens. I've seen some people claiming that the non-VC has better optics then the VC version. The equivalent lens from Nikon, Nikkor 17-55mm f/2.8 doesn't have VR,

A stabilization System combined with a fast lens (Large Aperture) is a very useful tool for some shots and stabilization is very useful for some shots, at the wide to normal focal lengths –
Yes.

The specific answer regarding “dispensability’ is: it depends on the intended uses of the lens.

My advice is for you to consider very carefully how you will use the lens –i.e. what pictures will you make with it.
Stabilization in a Standard Range Zoom Lens is very useful for the following general shooting scenarios:
Travel Photography – where a tripod or monopod carriage is difficult.
Indoor Photography – where a tripod or monopod and Flash is prohibited (museums for example).
Night Scapes, generally – as a walk ‘about lens’ i.e. when one is not carrying a camera support.

IS is not ONLY in Telephoto Lenses – but it IS perhaps a more noticeable ‘feature’ of longer lenses; Canon has IS in many of their Standard Zoom lenses – EF-S range: 17 to 55, 15 to 85, 18 to 55 etc, and in the EF Range 24 to 105.
Many Canon users were disappointed in the release of the EF24 to 70/2.8MkII when that lens did NOT have Image Stabilization and there is much discussion and contemplation about the recently released Tamron ‘equivalent’ which does have stabilization.

As one example – This shot would have been much easier to make with a stabilized 24mm Prime Lens – (at this venue Tripods and Monopods were NOT allowed):


“That Side”
Tech: EOS5D + EF24/1.4
Shooting: F/1.4 @ 1/8s @ ISO1600 HH

WW

[Manfred M]

Dave – just because a manufacturer’s literature talks about something doesn’t make it true. The literature and brochures we see are written by the Marketing Departments, rather than the Engineering groups. Their goal is to impress the prospective purchasers so that they buy the gear. Hyperbole and oversimplification is used, at the expense of technical accuracy. Let’s remember that a camera sensor is essentially a two dimensional object; and everything the lens captures is reduced to a position along the height and width of the sensor.

The diagram of the Olympus camera in the dpReview.com you are referring to is an excellent example of what I was referring to: http://www.dpreview.com/reviews/olympusem5

It looks good in sales literature, but is totally meaningless from a technical standpoint. They are claiming 5-axes; horizontal displacement, vertical displacement, yaw (rotation around the vertical axis), pitch (rotation around the horizontal axis) and roll (rotation around the axis through the centre of the lens). But if you understand the math, they are showing you exactly the same thing in two different ways of representing it. The green arrows represent motion using a Cartesian coordinate system and the purple arrows are using the polar coordinate system. The sensor doesn’t care; it is a two-dimensional entity. If anyone is interested; there is a famous novella from the late 1800’s; Abbot A. Edwin’s “Flatland” that describes how a three dimensional object would look to a two-dimensional being, which is very much what this is all about. Look at it for free (legally as the copyright has long expired) at: http://www.gutenberg.org/files/201/201-h/201-h.htm

The Olympus optical stabilization system uses a moving sensor design (in two directions), rather than the floating lens element approach used by Canon and Nikon. The implementations may be different, but the end results should be the same.

Again, as I said in my previous postings; two accelerometers; at right angles to each other on the same plane as the sensor are all that is required. We do need to remember that all of these companies install commodity components from third party suppliers in their lenses and cameras. I rather suspect that the reason that three-axis accelerometer assemblies might be used is that the 2 axis ones could be down to them being less expensive or even not available; hence the switch to a different design. So in a true marketing approach, this is offered up as a “new and improved” design, even though in reality nothing has changed. Sorry for the cynicism, but I have spent too many years working on product changes that the marketing departments would spin into “new and improved”, and the reality was “cheaper and better profit margins” or “we can’t get the original parts any more so we had to substitute something else”. These are the daily realities of any Engineering group involved in either product development or moving things from design to production.

As you can see from the catalogue from one of the major accelerometer suppliers, three-dimensional models are make up the bulk of the lineup: http://www.st.com/internet/analog/subclass/444.jsp

Bill: The Canon article (at least the first page) was a very good read.
http://cpn.canon-europe.com/content/...abilisation.do

The little Flash animation is not quite correct, as it shows the floating element handling the optical stabilization function as moving into a fixed position and then having the exposure occur. This does not match what is written in the article, but is fun to watch. The article confirms what I had previously written. I found the information of the design points running from 0.5 Hz the 20Hz quite interesting. While I can’t be certain, it does rather look like the floating elements rotate on gimbals rather than being displaced in axially. I come to that conclusion because the hole in the IS mechanism is round, not oblong.

I think your comment about the Olympus stabilizing the camera body is not quite right. My understanding of the Olympus and Sony system is that they use a floating sensor design; i.e. the sensor, including the AA filter and electronics built into the sensor, are repositioned during the exposure to compensate for camera movement. The camera body itself is rigid.

I also find it unfortunate that some of the upgrades to the IS functionality appear to be minor technical tweaks that could be easily applied backwards through lens firmware updates and make backwards compatible improvements. So far as I can tell, the only company that seems to update in-lens firmware seems to be Panasonic, but there may be others I am aware of.

[William W]

The Canon article . . . The little Flash animation is not quite correct . . . While I can’t be certain, it does rather look like the floating elements rotate on gimbals rather than being displaced in axially. I come to that conclusion because the hole in the IS mechanism is round, not oblong.

I think so -

FYI :




Ref: Lens Works III (op.cit.) pp204~205


***

I think your comment about the Olympus stabilizing the camera body is not quite right. My understanding of the Olympus and Sony system is that they use a floating sensor design; i.e. the sensor, including the AA filter and electronics built into the sensor, are repositioned during the exposure to compensate for camera movement. The camera body itself is rigid.

Yes. I concur: My understanding is that in the Olympus system the sensor is floating and it is repositioned.


Your comment was predicated by poor phraseology by me and /or misunderstanding of the exact meaning of my text by you.

The word stabilizing was written in inverted commas to indicate it was being used in a loose and non-technical meaning. It was not intended to mean the (whole) camera body was being stabilized, but rather that the stabilization system was contained within the camera body - and not the lens.

Just as we might use the term “lens stabilization” as opposed to "camera stabilization": the whole lens itself is rigid and NOT being stabilized – but rather some elements inside the lens are being repositioned – but the stabilization system is contained within the lens.

WW

[Manfred M]

Bill - You are a treasure trove of great information! Thank you very much for this.

This exploded diagram shows pretty well confirms everything I've written. One can see the two acclerometers, the two positioning sensors, etc. The only thing I am scratching my head about is why there are eight (electro) magnets, rather than the four I would have expected.

I was trying to figure out how they would dampen the oscillations from the positional control circuit, and that is likely what these might be doing. Four actuators (as a a balanced pair for each axis) to move the optical stabilization lens element and four damping magnets (again as as a balanced pair on each axis) to smooth damp the motion would be my best guess.

[Dave Humphries]

Hi Manfred,

I do 'get' what you mean about marketing departments taking liberties with the truth and spinning things to the extreme.

However, I also think that there might be a small additional performance improvement to IS/VR/VC if there were more micromotors, attached to the lens compensation group, or sensor corners, that could move the 'corners' fore and aft assymetrically to capture light rays that would have arrived at the sensor "off perpendicular" due to yaw or pitch movement, as distinct from lateral positioning in an X/Y direction shown in the diagrams Bill contributed.

The concept of roll movement correction is only correctable on the sensor based implementations, since rotating a lens group won't achieve the desired effect. A sensor could be rotated to compensate for roll

Now, I'm not saying there are these extra sensors and micromotors, but technically, it might be possible - however I also fully accept the improvement might be, to all intents and purposes, negligible. Also that it is possible/likely that you're 100% right and it's all a load of marketing 'tosh' by Olympus - but surely, they wouldn't tell fibs, would they?

This is turning into an interesting and educational thread - despite my efforts
(honestly, I am being serious)
I have therefore given it a couple of tags in the Tag Cloud to allow reference back to the useful links provided by Bill and yourself.

On the topic of the frequency range of oscillations Canon mention; 0.5 to 3 Hz and 3 - 20 Hz, Nikon obviously decided to separate these; which is why most of their lenses have the "Normal" and "Active" VR mode switch. I assume that "Active" switches out some of the circuit dampening to allow the correction of the faster 'vibration' frequencies.

Cheers,

[William W]

Manfred - et al, I just found this -

http://software.canon-europe.com/fil..._Book_9_EN.pdf

I did not know an electronic version was avaialable publically and for free, from Canon.

I still like the hard copy, hard cover book, though.

WW