G7 killer

[Ray]

The ccd shif is the OPTICAL stabilization, the ISO is the second non-optical stabilization, so there are 2 ie "dual" stabilizations happening, one is optical and one is not.
[a href=\"index.php?act=findpost&pid=98630\"][{POST_SNAPBACK}][/a]

As I said, this term is misleading. By 'optical image stabilisation' I understand that the lens has some shifting elements. In the context of that definition, CCD shift should not be described as 'optical image stabilisation'.

In fact it's still not clear to me what CCD shift means. This is presumably different from anti-shake where the whole sensor moves. Is it another name for 'pixel shift' which video cameras have used for years?

[BJL]

As I said, this term is misleading. By 'optical image stabilisation' I understand that the lens has some shifting elements.
[a href=\"index.php?act=findpost&pid=98661\"][{POST_SNAPBACK}][/a]

Ray, you may understand "optical image stabilization" to mean doing it the way that you arwe familiar with from your Canon SLR system, but there is nothing in the combination of words that implies movement of one part of the optical system (lens elements) rather than another (sensor).

Practically, the result is about the same: the image of a particular part of the scene continues to fall on the same part of the sensor throughout the exposure time despite camera movements, reducing blurring due to camera motion at a given shutter speed. And this is essentially different from other approaches to reducing camera motion blur, which require higher shutter speeds, higher ISO, fixing the camera in place with a tripod etc.

Perhaps it would be useful for us to concentrate on the facts (like how the pictures look in the end)  and stop worrying or debating so much about what is the right and wrong way to use words like "optical stabilization", "enlargement", "lens speed" or "sensor speed". Or even "depth of field".

[Ray]

Ray, you may understand "optical image stabilization" to mean doing it the way that you arwe familiar with from your Canon SLR system, but there is nothing in the combination of words that implies movement of one part of the optical system (lens elements) rather than another (sensor).

[a href=\"index.php?act=findpost&pid=98715\"][{POST_SNAPBACK}][/a]

Quite true, and I said as much on the previous page. I don't think anyone could take Olympus to court for using a phrase outside of any generally more accepted meaning, but still within the fundamental meaning of the words.

Nevertheless, when companies do use expressions in this manner, outside of their usual meaning, some confusion inevitably results. Doing a quick Google search on Optical Image Stabilization I came across the following distinction made with respect to the use of these terms with video cameras.

What's EIS/DIS?

Electronic Image Stabilization and Digital Image Stabilization are completely electronic means for correcting image shake. As the shaky image hits the CCD chip, these systems reposition the active area of the chip (the location on the chip that the image is read from) to compensate for it, by re-addressing the area of the chip that they're reading from. If you've seen Rocky & Bullwinkle (a US cartoon involving a moose and a squirrel), think of Bullwinkle running back and forth with the bucket of water to catch Rocky after Rocky jumps from the high diving board (of course, Bullwinkle winds up in the water, but that's another story).
The EIS/DIS controllers look for motion vectors in the image (typically a widespread displacement of the entire image) and then decide how to "reposition" the image area of the chip under the image to catch it in the same place. The actual repositioning is done in one of two ways: one is to enlarge (zoom) the image digitally, so that the full raster of the chip isn't used. The controller can then "pan and scan" within the full chip raster to catch the image as it moves about. The other is to use an oversize CCD, so that there are unused borders that the active area can be moved around in without first zooming the image.
The zoom-style pan 'n' scanner can be detected quite simply: if the image zooms in a bit when EIS/DIS is turned on, then a zoom-style pan 'n' scanner is being used. Unfortunately, such methods reduce resolution, often unacceptably.
All EIS/DIS systems suffer from several problems. One is that, because the actual image is moving across the face of the chip, image shakes induce motion blur. Even though the position of an image may be perfectly stabilized, you can often notice a transient blurring of the image along the direction of the shake. Sometimes it's quite noticeable. To get around this, many EIS/DIS systems close down the shutter a bit to reduce blur. This reduces light gathering capability. You can't have everything, you know.
Another problem is that the motion-vector approach to stabilization can be easily fooled. If the area of the image being scanned doesn't have any contrasty detail that the processor can lock onto, the stabilization can hunt, oscillate, or bounce. This looks like a mini-earthquake on the tape, and it can occur at the most annoying times.
Also, the stabilization can work too well. Often when one starts a slow pan or tilt with EIS/DIS engaged, the system will see the start of the move as a shake, and compensate for it! Eventually, of course, the stabilizer "runs out of chip" and resets, and the image abruptly recenters itself.
The big advantage of EIS/DIS is that it's cheap.

What's optical stabilization?

Optical stabilization such as "SteadyShot" is descended from Juan de la Cierva's 1962 Dynalens design, a servo-controlled fluid prism used to steer the image before it hits the CCDs (in the '60s, of course, it steered images onto film or onto tubes!). In the late '80's and early '90's, Canon and Sony updated this technology for use in consumer gear, and it worked so well that Canon now offers a SteadyShot attachment for some of their pro/broadcast lenses.
The fluid prism is constructed of a pair of glass plates surrounded by a bellows and filled with fluid so that the entire assembly has a refractive index comparable to a glass prism. The angle of the prism is changed by tilting the plates; one plate can be rotated vertically, moving the image up or down, and the other rotates horizontally, steering the picture right or left.
Rotation rate sensors detect shake frequencies and tilt the front and back plates appropriately. Position sensors are also used so that in the absence of motion the prism naturally centers. The position sensors also detect when the prism is about to hit its limit stops, and reduce the corrections applied so that shake gradually enters the image instead of banging in as the prism hits its limits.
Optical stabilization of this sort is expensive, tricky to manufacture and calibrate, and must be tuned to the lens. Adding a wide-angle or telephoto adapter to a SteadyShot lens screws up SteadyShot; the processor doesn't know about the changed angle of view (all it knows is the current zoom setting) and thus over- or under-compensates for shake.
But for all that it works brilliantly: because the image is stabilized on the face of the CCDs, there is no motion blur; because rate sensors are used, the system isn't fooled by motion in the scene or by lack of detail; because a physical system has to move to reposition the image, there are no instantaneous image bounces or resets as can happen with EIS/DIS.
[It's interesting to note that on the XL-1, Canon added image motion-vector detection to the rate gyros on their optical stabilizer. As a result, the system seems to "stick" on slow pans and tilts just like an EIS/DIS system, although the recovery is more fluid and less jarring. On the other hand, it really does a superb job on handheld lockdowns.]

[BJL]

I don't think anyone could take Olympus to court for using a phrase outside of any generally more accepted meaning, but still within the fundamental meaning of the words.
[a href=\"index.php?act=findpost&pid=98778\"][{POST_SNAPBACK}][/a]

Your evidence for lens-based as the "generally more accepted meaning" of optical stabilization is really just evidence that this is the oldest and more widely used version, for the obvious reason that it is the only version that was usable with film. Taking the oldest or more common example of something to be the best or only correct version is a fallacy (I am tempted to say, very conservative or even reactionary thinking!)  A bit like ideas that one particular aspect ratio or frame size common in the film era is inherently the best or correct one under all circumstances.

[Ray]

A bit like ideas that one particular aspect ratio or frame size common in the film era is inherently the best or correct one under all circumstances.
[a href=\"index.php?act=findpost&pid=98904\"][{POST_SNAPBACK}][/a]

Surely you wouldn't accuse me of that degree of narrow-mindedness, BJL, after all the discussions we've have on the topic of aspect ratios   .

The only issue I have with this apparent abuse of terminology (by Olympus and others) is that the negative aspects of one form of image stabilisation compared with another, might be glossed over.

I get the impression that the 'pixel shifting' approach can reduce over-all resolution. I certainly got this impression when I first used a Panasonic videocam, which I foolishly bought some years ago because it was on special and boasted broadcast quality. I was so disappointed in the resolution and dynamic range, I never used it again after the first few weeks.  

[Jonathan Wienke]

I get the impression that the 'pixel shifting' approach can reduce over-all resolution.

You're confusing electronic stabilization, where a smaller area of the sensor is used and the "crop area" is moved around via bit-fiddling, versus Olympus' optical stabilization, which physically moves the entire sensor chip around, and has no resolution penalty.

[PeteC]

Wonder if somebody could hack a RAW option?   

JC
[a href=\"index.php?act=findpost&pid=97599\"][{POST_SNAPBACK}][/a]

Exactly. Surely the RAW function is purely a matter of firmware? Presumably Canon could issue an update to add RAW if they wanted to? Any point in leaning on them?

Pete

[Ray]

You're confusing electronic stabilization, where a smaller area of the sensor is used and the "crop area" is moved around via bit-fiddling, versus Olympus' optical stabilization, which physically moves the entire sensor chip around, and has no resolution penalty.
[a href=\"index.php?act=findpost&pid=98984\"][{POST_SNAPBACK}][/a]

Exactly my point, Jonathan. Even I'm confused, and that can't be good.  .

I get the impression from the 'dcresource' preview that the SP-550 actually employs 3 types of image stabilisation; the sensor shift, the high ISO/big aperture approach and 'pixel shifting' for video mode.

[GregW]

Panasonic are now offering RAW on an established digicam which sport harware vibration reduction.

http://luminous-landscape.com/forum/index....showtopic=14577

[BJL]

The only issue I have with this apparent abuse of terminology (by Olympus and others) is that the negative aspects of one form of image stabilisation compared with another, might be glossed over.

I get the impression that the 'pixel shifting' approach can reduce over-all resolution.
[a href=\"index.php?act=findpost&pid=98951\"][{POST_SNAPBACK}][/a]

If people are indeed subject to confusions (like you confounding of pixel shift post-procxeesing so physically moving the sensor), the solution surely is to look at results, not buzz-words.

At least read reviews of how a particular camera handles low shutter speed hand-holding, or better yet test it oneself. I suggest a double espresso first, to intensify the challenge to the technology.

[Ray]

If people are indeed subject to confusions (like you confounding of pixel shift post-procxeesing so physically moving the sensor), the solution surely is to look at results, not buzz-words.
[a href=\"index.php?act=findpost&pid=99465\"][{POST_SNAPBACK}][/a]

Absolutely, and I shall look forward to viewing comparisons between the the SP-550 at 500mm equiv focal length and model XX-xxx at 500mm which employs true 'lens shift' optical image stabilisation.

[BJL]

... true 'lens shift' optical image stabilisation.
[a href=\"index.php?act=findpost&pid=99516\"][{POST_SNAPBACK}][/a]

where by "true", you mean little more than "the type that I am familiar with, and which used to be the only type since it was the only type that worked with film"? You have totally failed to show that physically moving the back end of the optical system (the sensor) is any more or less "truly optical" than moving parts further forward (lens elements), so why are you persist in declaring that any other approach is "false"?

[Ray]

You have totally failed to show that physically moving the back end of the optical system (the sensor) is any more or less "truly optical" than moving parts further forward (lens elements), so why are you persist in declaring that any other approach is "false"?
[a href=\"index.php?act=findpost&pid=99648\"][{POST_SNAPBACK}][/a]

I don't have the  means to 'show'. All I can do is draw inferences from the discussions I see on the net. With long telephoto lenses, lens shift appears to be more effective than sensor shift. Is this not the case?

[BJL]

I don't have the  means to 'show'. All I can do is draw inferences from the discussions I see on the net. With long telephoto lenses, lens shift appears to be more effective than sensor shift. Is this not the case?
[a href=\"index.php?act=findpost&pid=99654\"][{POST_SNAPBACK}][/a]

Ray, you have completely ducked the point. I was not discussing which works better, I am discussing you unjustified use of the wording "real optical stabilization" to apply only to lens wiggling and not sensor wiggling.


But as to those discussions about how well the different methods work with long lenses, I find them seriously lacking in quantitative basis. All they seem to show is that sensor wiggling might require grater sensor movements for longer focal lengths, but that gives no comparison to how well it compares to lens wiggling.

More important, it is the real focal length that matters, not the "35mm equivalent".


If you need enough sensor movement to keep up with camera movement before the exposure starts as well as during, the amount of movement needed (in mm) depends on the actual focal length, independent of the format. This is because a given angular twist of the cameras moves the image relative to the sensor by about (focal length)/(angle). In the Olympus digicam under discussion, the maximum true focal length is about 80mm, small relative to the focal lengths of interest in Sony and Pentax sensor stabilized cameras. (And I expect compact digicam lenses to stay under about 100mm.)

Sensor based stabilization might struggle with real 500mm lenses.


Life is even easier with "just in time" stabilization that is only active during exposure, not during composition. My estimate is that
1. Without stabilization, decent sharpness needs the image motion across the sensor during exposure to be 1/1000th or less of the sensor width, and a smaller fraction for higher resolution needs: maybe no more than about two pixel widths.
2 This presumably happens at traditional hand-holdable shutter speeds.
3. To get four stops of stabilization requires handling 16 times that traditional hand-holdable shutter speed, a situation where without stabilization the image might move as much as 16/1000's of the way across the sensor, but not more than about that.
4. So sensor moving stabilization only needs that much sensor movement during the exposure. For a compact digicam sensor, this means less than 0.2mm of movement. For a 35mm format sensor, it would mean less than 0.5mm.

And for any format, the amount of sensor movement needed during exposure does not depend on focal length or angular field of view! It depends only on how many stops you want to go beyond the hand-holdable limit.

[Ray]

Ray, you have completely ducked the point. I was not discussing which works better, I am discussing you unjustified use of the wording "real optical stabilization" to apply only to lens wiggling and not sensor wiggling.

If lens shift as opposed to sensor shift does not have any inherent advantage, then it really does not make any difference which terminology is used. The only issue I see here is the possibility that consumers might be misled into believing they are getting a better image stabilisation system than they actually are getting in practice.

More important, it is the real focal length that matters, not the "35mm equivalent".
If you need enough sensor movement to keep up with camera movement before the exposure starts as well as during, the amount of movement needed (in mm) depends on the actual focal length, independent of the format. This is because a given angular twist of the cameras moves the image relative to the sensor by about (focal length)/(angle). In the Olympus digicam under discussion, the maximum true focal length is about 80mm, small relative to the focal lengths of interest in Sony and Pentax sensor stabilized cameras. (And I expect compact digicam lenses to stay under about 100mm.)

I find it difficult to reconcile this statement with the 1/35FL rule for a reasonably sharp hand-held shot enlarged to 8x10" print size. You are not suggesting, are you, that without image stabilisation, sharp images are possible using 1/7th sec exposure with the average P&S using a 'standard' focal length?

[BJL]

If lens shift as opposed to sensor shift does not have any inherent advantage, then it really does not make any difference which terminology is used.
[a href=\"index.php?act=findpost&pid=99740\"][{POST_SNAPBACK}][/a]

Perhaps we should leave it at that, instead of using prejudical wordings like "real optical image stabilization".

find it difficult to reconcile this statement with the 1/35FL rule for a reasonably sharp hand-held shot enlarged to 8x10" print size. You are not suggesting, are you, that without image stabilisation, sharp images are possible using 1/7th sec exposure with the average P&S using a 'standard' focal length?
[a href=\"index.php?act=findpost&pid=99740\"][{POST_SNAPBACK}][/a]

Be careful to distinguish absolute distance that the image moves (in mm) from angular distance or the number of pixels that the image moves across, which is more in line with how much blur one would get without stabilization. Angular movement, or pixels of movement, depends on the angular FOV of the lens, and so depends roughly on "35mm format equivalent focal lengths", but actual distance moved depends mainly on "true focal length". Of course, the same distance of image movement will cause different amount of image blurring in different formats due to the different degree of enlargement needed to get the same sized print.

The reason I estimated movement distance (in mm) was to estimate how much sensor movement is needed to compensate for it. It seems to be that the physical limitation on sensor stabilization (do you approve this wording?) is mostly how much sensor movement is needed. With smaller sensors perhaps having the advantage of lower weight, so that moving over equal distance requires less forces and energy, while large sensors and pixels having the advantage that the movements can be less precise.


By the way, several reviews of Panasonic's OIS lens stabilization indicate that it achieves somewhat better image quality in "mode 2", where stabilization is done only during the exposure, than in "mode 1", where stabilization operates also during composition (so also stabilizing the VF image.) Maybe because in mode 2, exposure starts with the stabilizing elements waiting in the optimal central position, rather than risking already being pushed to the limits of their movement during wobbly composition as could happen in mode 1. If so, the same could apply to leaving the sensor in central position until the shutter release starts.

An experiment I did: with my narrowest FOV lens, 200mm in 4/3" format, I see the VF image move around by about 1/10th of the frame width, which means about 2mm across the frame (which is 18mm wide), so about 1/100th of the focal length. I propose as a cautious rule that "mode 1" sensor stabilization that operates during composition probably needs to be be able to move the sensor about 1/50th of the focal length. That is only a couple of mm for "super-zoom" digicams with their sub-100mm maximum focal lengths, but maybe up to a cm or so with super-telephoto SLR lenses at 400mm and up.

[Ray]

BJL,
Maybe this all boils down to one type of image stabilisation technology being more practical to implement in a particular type of camera design, perhaps sensor shift being more suitable for the smaller camera and lens shift more suitable for the larger camera.

I notice that Canon are claiming a 3 stop advantage with their new 70-200L F4 IS, as they do with the 24-105 IS. Building image stabilisation into the lens, allows for improvements without upgrading the whole camera. This wouldn't be a concern with a P&S camera.

Doing a bit of Googling on the subject, I notice that some photographers take multiple shots in continuous mode when trying to get a sharp shot with IS and slow shutter speeds. It seems to be often the case that, whilst most of the shots might be substandard with regard to sharpness, the odd one or two might be very acceptably sharp.

I'll try employing this technique. I have noticed that 1/13th exposure with my 24-105mm IS at 24mm (using the 5D) is a bit risky and 1/6th even more so, yet a 3 stop advantage would convert 1/6th sec into the equivalent of 1/50th, which should be fast enough for 24mm.

[howiesmith]

BJL,
Maybe this all boils down to one type of image stabilisation technology being more practical to implement in a particular type of camera design, perhaps sensor shift being more suitable for the smaller camera and lens shift more suitable for the larger camera.

I notice that Canon are claiming a 3 stop advantage with their new 70-200L F4 IS, as they do with the 24-105 IS. Building image stabilisation into the lens, allows for improvements without upgrading the whole camera. This wouldn't be a concern with a P&S camera.

Doing a bit of Googling on the subject, I notice that some photographers take multiple shots in continuous mode when trying to get a sharp shot with IS and slow shutter speeds. It seems to be often the case that, whilst most of the shots might be substandard with regard to sharpness, the odd one or two might be very acceptably sharp.

I'll try employing this technique. I have noticed that 1/13th exposure with my 24-105mm IS at 24mm (using the 5D) is a bit risky and 1/6th even more so, yet a 3 stop advantage would convert 1/6th sec into the equivalent of 1/50th, which should be fast enough for 24mm.
[a href=\"index.php?act=findpost&pid=99946\"][{POST_SNAPBACK}][/a]

OPINION

It may be that Canon uses IS in the lens, not as a practicle measure, but with a mind toward backfitting IS to older camera bodies.  Perhaps similar to autofocus motors in the lens instead of the body.

[Ray]

OPINION

It may be that Canon uses IS in the lens, not as a practicle measure, but with a mind toward backfitting IS to older camera bodies.  Perhaps similar to autofocus motors in the lens instead of the body.
[a href=\"index.php?act=findpost&pid=99949\"][{POST_SNAPBACK}][/a]

No, Howard. There's no compatibility between Canon's newer autofocus, IS lenses and older pre-autofocus bodies, just as there's no compatibility between the older Canon FD lenses and the newer bodies supporting autofocus.

This has always been considered a disadvantage of the Canon system in relation to a Nikon system which allows the use of older lenses, even though autofocus is disabled.

[BJL]

BJL,
Maybe this all boils down to one type of image stabilisation technology being more practical to implement in a particular type of camera design, perhaps sensor shift being more suitable for the smaller camera and lens shift more suitable for the larger camera.
[a href=\"index.php?act=findpost&pid=99946\"][{POST_SNAPBACK}][/a]

Maybe so.

Building image stabilisation into the lens, allows for improvements without upgrading the whole camera.
[a href=\"index.php?act=findpost&pid=99946\"][{POST_SNAPBACK}][/a]

That cuts both ways:
- to get the newest lens-based stabilization, replace each lens that you want stabilized, use existing bodies,
- to get the newest sensor-base stabilization, replace each body (one for most of us) and use all existing lenses with stabilization.

By the way, I am with Howard on the idea that for Canon and Nikon, lens-based stabilization has some advantage of backward compatibility with many SLR's already in use, film and digital. This compatibility goes back about twenty years for Canon, even if it fails with truly ancient pre-EOS bodies. But it could also be simply that they already have good lens-based stabilization technology, and so have less incentive to invest in developing a different one, like sensor-based. (A bit like designing DSLR's for at least partial and transitional compatibility with their substantial lens systems.)