Pages: 1 [2] 3   Go Down

Author Topic: In camera image stabilisation  (Read 22165 times)

dilip

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 60
In camera image stabilisation
« Reply #20 on: January 03, 2008, 12:51:26 pm »

Quote
John - is there some reason for us laymen, why future technology could not address this in some way? One might think it is a frequency issue to the movement with increasing focal length as much as just the distance the sensor has to move? But that is from a layman's perspective...  With all the coding built in these days, any reason why the camera could not adjust its sensor movement to the specific requirements needed for a particular focal length, within the limits of a particular sensor...just wondering??
[a href=\"index.php?act=findpost&pid=164814\"][{POST_SNAPBACK}][/a]

The issue is one of geometry.   IS (body or lens based) is introduced to address what we can affectionately call shooter shake. Slight shake in your hands typically has a component that translates to wobble in the optical axis (straight out from the sensor through the lens). Thus, if you want to think of it, your lens sways with an arcuate motion (I'm simplifying a bit...)

At a short focal length, a shake of a couple of millimeters isn't really noticable since it might end up being 1 degree, from a 45 degree field of view. However, at a long focal length, the same shake might be 1 degree from an 8 degree field of view.

To perform IS, the sensor has to move so that the center of the image remains somewhat centered.  If it needs to move 1/45th of its size that isn't a big deal (especially since on a reduced size frame, a conventional lens has an image circle bigger than the sensor). But if you want it to work just as well at a long focal length, the sensor has to move 1/8th of its size, in each direction.  Not only is there not usually enough space in the body, but the image circle cast by the lens isn't usually big enough to accomodate that.  Thus, in body IS doesn't perform as well at long focal lengths. (though it is likely that the degree to which the sensor is moved is calculated using the focal length of the lens as one of the inputs)

If we're putting the IS in the lens, we're using gyroscopes to detect movement, and then we have floating elements that can adjust to move the placement of the image.  If you have a 600mm lens with IS, the IS system is tuned for that focal length.  If you have a 70-200mm lens, the IS system can be tuned to work in that range, and in theory you can feedback the focal length to the logic processor to change how the floating elements are moved.  You cannot get that sort of tuning with a body that has to work with lots of different lenses.

wow... that took more explaining than I thought it would.

--dilip
Logged

DonWeston

  • Full Member
  • ***
  • Offline Offline
  • Posts: 148
In camera image stabilisation
« Reply #21 on: January 03, 2008, 01:30:44 pm »

Thanks Dilip - you gave a great explanation of the current issue of why body IS is not effecticve to present. I guess my understanding is incomplete though, as this does not explain why some of these concerns might not be solved down the road i.e. in the near future. It was not so long ago, that IS or VR tech gave only 1-2 stops of motion compensation , now we have lenses where 3-4 stops are possible under certain circumstances. Why could not the focal length information be transmitted to the body and thus have the body IS or VR adjust in parameters to yield the same compensation that the in lens system would accomplish. I always thought on a separate issue that IS was less effective for wide angles? Is this incorrect? Just trying to gain a greater understanding of how things work but have no engineering credentials to assist....Thanks again, Don
Logged

dilip

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 60
In camera image stabilisation
« Reply #22 on: January 03, 2008, 05:32:49 pm »

Quote
Thanks Dilip - you gave a great explanation of the current issue of why body IS is not effecticve to present. I guess my understanding is incomplete though, as this does not explain why some of these concerns might not be solved down the road i.e. in the near future. <snip>

The following will ask you to take some assumptions, and is somewhat general.

In order to give you an enhanced IS-system, an in body system has to be move the sensor around both faster and farther.  Faster allows it to track small deviations that happen quickly.  This will happen.  Farther is needed to compensate for bigger swings in the image, which happen at the longer focal lengths. A 600 mm lens gives you a field of view of 4 degrees.  If your vibration changes the angle 1 degree, the sensor has to be moved 1/4 of it size away from the centred position. At that point, most lenses aren't going to be projecting the image correctly (if at all).

In theory a 600mm lens with IS can be designed to deal with this much better.

The use of in lens IS does not mean that in body IS cannot also be used. There is nothing, other than complexity, to prevent hybridized systems.  But remember that complexity is often underestimated.

And we should also remember that full frame sensors are unlikely to be shifted around in body due largely to the fact that something like the 1Ds Mark II or Mark III already push the optical quality of lots of lenses in the corners. Shifting a sensor like that around would result in bringing those bad edges in towards the centre of the sensor, and push the edges either into the dark or at least into really bad parts of the image.

--dilip
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #23 on: January 03, 2008, 09:34:24 pm »

Quote
I admire your fancy imagination, but the floating element only has very limited usible movement for its purpose. Don;t forget even when there is no vibration, the effective image circle is still limited to that 36mm x 24mm region. If the vibration is too large for the moving sensor, it would be too large for the floating lens to maintain the good image quality.

Your technical cluelessness is showing here. With long lenses, the image circle is constrained primarily by the shape and size of the lens mount, not the lens itself. The longer the lens, the larger the image circle can be while still maintaing a reasonably high MTF. A Canon 600mm prime could easily cover 4x5 film with only a small MTF falloff in the corners.

Let's look at a scenario where the camera shake is such that the subject motion is 30mm at the sensor plane. With lens-based IS, moving a small lens element to keep the subject steady on the sensor is not that hard. The lens has plenty of spare image circle to keep the quality level up. Now given the same scenario with in-body IS, you have a serious problem. Even if you can move the sensor far enough and fast enough to stabilize the image (which is not bloody likely), the lens mount is going to cause vignetting.
Logged

250swb

  • Full Member
  • ***
  • Offline Offline
  • Posts: 220
In camera image stabilisation
« Reply #24 on: January 04, 2008, 04:24:33 am »

I can't help but see this entire discussion is solely based on the simple premise that Nikon and Canon lenses are so gigantic that they really do need in-lens IS.

But for a successful in-body IS system, like Olympus use, it is perfectly feasible to get four stops usable IS on say a 600mm (equiv) lens simply because it is so short (tiny) compared to a Canikon lens of the same reach. The geometry works for Olympus. So 'if' Canon are thinking of in-body IS, are they just fumbling around looking for another marketing ploy to hook people with, or perhaps looking towards a fundamental change of direction for the future? Perhaps a dedicated range of digital lenses that are smaller and lighter, and that don't need IS built in. And perhaps designed around the smaller APS-C sensor given the full format sensor is out performing the current legacy lenses anyway? There may be a big shake up of Canon culture to come, and in-body IS may be a positive move if seen perhaps in context of what else may be around the corner.

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #25 on: January 04, 2008, 09:30:13 am »

Quote
But for a successful in-body IS system, like Olympus use, it is perfectly feasible to get four stops usable IS on say a 600mm (equiv) lens simply because it is so short (tiny) compared to a Canikon lens of the same reach. The geometry works for Olympus.

Did you fail math or something? If you're shooting an Olympus of equivalent focal length to the Canon 600mm, in an equal-shake comparison, the Olympus sensor is going to have to move the exact same proportion of its size as the Canon sensor to achieve the same degree of stabilization, and both would have exactly the same amount of lens-mount vignetting of both used body IS.
Logged

aaykay

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 359
In camera image stabilisation
« Reply #26 on: January 04, 2008, 05:15:44 pm »

One additional thing we have to keep in mind (and I rarely hear any discussion on when talking about IS systems) is that it is VERY challenging trying to spot-focus or spot-meter a view-finder image that is not steady.  

An in-body stabilization system does not stabilize the image in the viewfinder and only enables a stabilized image to be CAPTURED.  

This is okay, with short (equiv) focal lengths, where the viewfinder shake is not that visible.  But at longer focal lengths, spot-metering/focusing would present challenges that are not present in the In-lens stabilized arrangement, where the image that gets to the viewfinder is stabilized (along with the captured image, obviously).
Logged

250swb

  • Full Member
  • ***
  • Offline Offline
  • Posts: 220
In camera image stabilisation
« Reply #27 on: January 06, 2008, 03:58:31 am »

Quote
Did you fail math or something? If you're shooting an Olympus of equivalent focal length to the Canon 600mm, in an equal-shake comparison, the Olympus sensor is going to have to move the exact same proportion of its size as the Canon sensor to achieve the same degree of stabilization, and both would have exactly the same amount of lens-mount vignetting of both used body IS.
[a href=\"index.php?act=findpost&pid=164987\"][{POST_SNAPBACK}][/a]

I'd have thought it was geometry I had failed?

As a very rough example Canon user moves the camera body accidentally up one degree and end of the lens moves 'say' an exaggerated ten millimeters off target if it is moving around the axis of the body. One degree movement for the Oly user becomes a five millimeter movement at the end of the lens because the lens is 'say' half the length. Likewise a similar thing happens when you balance the lens axis in the centre or reverse the whole thing and rest the lens hood on a wall. The amount of movement is amplified by a lens of longer physical length. So whatever the focal length the Oly body IS has to move a smaller distance to compensate for the same amount of movement.

You can work it out for yourself with some graph paper (thats paper with lines on it).

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #28 on: January 06, 2008, 10:40:08 am »

Quote
The amount of movement is amplified by a lens of longer physical length. So whatever the focal length the Oly body IS has to move a smaller distance to compensate for the same amount of movement.

You can work it out for yourself with some graph paper (thats paper with lines on it).

Geometry is a branch of mathematics.

And you're ignoring the fact that the Olympus lens mount is smaller than that of Canon, which is one side effect of designing the entire system for a smaller sensor and image circle. Though the subject movement in millimeters on the sensor plane for a given shake level is smaller with the Olympus than Canon, the amount of camera shake necessary to cause lens-mount vignetting is similar with both systems. B for effort, F for validity of arguments.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11310
    • Echophoto
In camera image stabilisation
« Reply #29 on: January 06, 2008, 05:59:52 pm »

Hi,

I use an Minolta 400/4.5 frequently with an 1.4x converter on Sony Alpha and can say that image stabilization works. If it works as well as other solutions I can't tell. My feeling is that 1/125 is possible hand held with this combo.

I don't think that the sensor need to move a lot, because it just moving during exposure.

Best regards

Erik

Quote
And does Sony have any long telephoto lenses available (400mm or longer), and if so, how effective is in-camera IS compared to lens-based IS of Canon or Nikon?
[a href=\"index.php?act=findpost&pid=162343\"][{POST_SNAPBACK}][/a]
Logged
Erik Kaffehr
 

250swb

  • Full Member
  • ***
  • Offline Offline
  • Posts: 220
In camera image stabilisation
« Reply #30 on: January 07, 2008, 04:18:00 am »

Quote
Geometry is a branch of mathematics.

And you're ignoring the fact that the Olympus lens mount is smaller than that of Canon, which is one side effect of designing the entire system for a smaller sensor and image circle. Though the subject movement in millimeters on the sensor plane for a given shake level is smaller with the Olympus than Canon, the amount of camera shake necessary to cause lens-mount vignetting is similar with both systems. B for effort, F for validity of arguments.
[a href=\"index.php?act=findpost&pid=165416\"][{POST_SNAPBACK}][/a]

Hmmm, B for effort and F for validity?

I'm sure you will put me right on this, but here goes. So you say the lens mount vignetting is similar on both systems, which so you don't get confused can both be called 'full frame' (in the sense that both systems have lens and sensor matched by design), yes?

OK, off the top of my head the minimum image circle for a 35mm lens is 43mm diameter? The Canon lens mount is 54mm diameter? The minimum Olympus image circle is 22.5mm diameter and the lens mount is 47mm internal diameter? So the Olympus lens mount is over twice as big a diameter as the minimum image circle needed. The Canon mount would have to be 86mm to be comparable, no? Is this the 'similar lens mount vignetting' you are suggesting? I'm obviously missing something in the equation given I failed maths and geometry.
« Last Edit: January 07, 2008, 04:25:41 am by 250swb »
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #31 on: January 07, 2008, 08:44:37 am »



And the E-3 mirror box is about 27mm wide. The sensor can only move 4-5mm before the mirror box or other bits of mechanism start getting in the way. And then there's the issue of the light trap at the rear of the longer lenses; how large is the opening on the back of the lens? At best, the Olympus can compensate for camera shake of about 1/2 the frame width.

OTOH, lens-based IS systems are not constrained by these limits; moving a small lens element in the right place can easily move the image circle projected by the lens the entire frame width.

And lens-based IS stabilizes the viewfinder image; body-based IS does not, at least when you're using the optical viewfinder. This can be a significant issue when trying to compose an image.
Logged

250swb

  • Full Member
  • ***
  • Offline Offline
  • Posts: 220
In camera image stabilisation
« Reply #32 on: January 08, 2008, 09:35:54 am »

Quote


But going back to geometry, it could be that a Canon in-body IS sensor (using the legacy designed lenses we are talking about) would need to move a larger amount than a 4/3rds sensor. This is because the light rays as you know fan out (with 'some' correction) as they leave the rear element,  spreading wider and wider. The more the camera is shaken off its central axis (the direct line between subject and centre of the sensor), so the sensor would need to move exponentially further to 'catch up' with the image given the focal point has to remain fixed, and with the light hitting the sensor at an increasingly oblique angle. So an in-body IS system on a Canon would be a poor idea without some new lenses (and design philosophy), as I originally mooted. The soft edges sometimes produced by Canon lenses would also be further exaggerated.

On the other hand, the image projected onto the 4/3rds sensor is far more linear, and while not exactly parrallel the light rays hit the sensor square on. And it is within a proportionally much larger and usable image circle compared to your Canon lens design. So the problem of an exponential growth of the angle due to sensor movement is largely avoided. Thus the sensor has to move a shorter distance to achieve the same thing, and all within a very large image circle devoid of the problems of vignetting.  And if as you say the the 4/3rds sensor could 'only' move half the frame width, then this would be a mighty difference in where the camera is pointing, never mind simply stabilising the image.

If the number of words written on the subject were soldiers you should have won.  But just like the Romans encountering a much smaller army at Cannae, few of them stood up in the end  

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #33 on: January 08, 2008, 10:51:07 am »

Quote
Well I think thats a comprehensive out flank and encirclement by me. Having been wrong about the lens mount you fudge a change of direction.

My first instinct would be to suggest that if the sensor needed to be moved 4-5mm by the image stabilisation, then a tripod should be used 

And when shooting from a car, boat, or airplane, a tripod would accomplish what? Or what about shooting in venues where tripods are not allowed? Or you are shooting from a tripod, and there is a strong wind that is still causing camera shake? I've encountered all of these situations when shooting.

You still have said nothing that refutes my original point--that body-based IS becomes less effective as focal length increases, because the mirror box/lens mount imposes a hard limit on how much shake body-based IS can correct. And longer focal lengths are where IS is needed most.

Yes, the outer diameter of the Olympus mount is somewhat larger in proportion to the sensor size compared to Canon's lens mount. But that still imposes a physical limit to the effectiveness of Olympus' IS system that does not apply to the Canon. Focal length has little effect on lens-based IS system's effectiveness. With lens-based IS, the corrections made by the IS mechanism are magnified in direct proportion to the focal length of the lens, so lens-based IS is equally effective at all focal lengths. But with body-based IS, the distance and velocity that the sensor must move to provide a consistent level of shake compensation increases in direct proportion to focal length, and the hard limits of how fast and how far the sensor can be moved impose a hard limit to the maximum focal length at which effective IS can be achieved.

With lens-based IS, if you have a 600mm lens and add on a 2x teleconverter, the lens-based IS will be just as effective; if IS was eliminating 80% of the shake before adding the TC, it will continue to eliminate 80% of the shake afterwards. But with body-based IS, the limitations on the distance the sensor can travel and how fast the sensor can be moved will reduce the percentage of shake that can be compensated for by the mechanism. If the 600mm lens is already causing the body-based IS to work at 100% of its capacity to reduce shake by 80%, then adding the 2x converter means that the body-based IS mechanism will only be able to reduce shake by 40%.

Since I don't have an E-3, I can't test to determine the focal length at which its IS mechanism is operating at 100% capacity in terms of sensor movement speed and travel distance. But the effectiveness of it's IS mechanism in reducing shake beyond that point will fall off dramatically, with the percentage of shake being compensated for decreasing at the same time overall shake levels are increasing.
« Last Edit: January 08, 2008, 10:55:32 am by Jonathan Wienke »
Logged

Farmer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2848
In camera image stabilisation
« Reply #34 on: January 08, 2008, 03:54:25 pm »

Until such time as it is determined at which focal length this fall off occurs, there's no valid data to determine if it matters.  If it falls off at 2000mm (for example) it's hardly going to be an issue for 99.9% of all photogs.  If it falls off at 200mm then it's a big issue.

There are also limits to the amount of movement that any system, even lens based, will be able to deal with based simply on the ability to received light reflected from the intended target.  By the time you place a normal hood on the lens, there is a limited window through which light from the target can pass to the lens.  Certainly this is a reasonably sized window, but it's not as if lens based IS has an infinite capacity to deal with lens movement.

Anecdotal evidence suggests that in lens IS has perhaps a 1 to 1.5 stop advantage at the longer focal lengths (perhaps more in some cases - different lenses and different in camera systems make any generalisation difficult and, in reality, poor).  That doesn't mean that in most situations the in camera system isn't going to provide sufficient compensation.  With the added benefit of working with all lenses (the capacity added to something like a 50mm f/1.4 makes for an amazing capacity to capture in low light for example) the in camera system has tremendous appeal.

As we would all know, it's often about the perceived general application, usability and cost rather than which system is technically superior (which is often debatable - take the old Beta vs VHS scenario).  At the highest end, it seems a hybrid system will possibly be available (partciularly for Canon and Nikon who already have lenses with IS - it's more of a challenge for Oly, Sony, etc to add lenses to their in camera system than vice versa for Canon and Nikon).
Logged
Phil Brown

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
In camera image stabilisation
« Reply #35 on: January 09, 2008, 08:36:43 am »

Quote
There are also limits to the amount of movement that any system, even lens based, will be able to deal with based simply on the ability to received light reflected from the intended target.

Nope. Canon's IS system is gyro-based; it will work just fine even with a lens cap on. It is equally effective no matter what the light level is. You can hear a faint whirring noise when the gyros spin up while IS is active.

Quote
Anecdotal evidence suggests that in lens IS has perhaps a 1 to 1.5 stop advantage at the longer focal lengths (perhaps more in some cases - different lenses and different in camera systems make any generalisation difficult and, in reality, poor).

Most tests of Canon's IS lenses indicate 2-3 stops of stabilization. My personal experience agrees with this.

Quote
Until such time as it is determined at which focal length this fall off occurs, there's no valid data to determine if it matters. If it falls off at 2000mm (for example) it's hardly going to be an issue for 99.9% of all photogs. If it falls off at 200mm then it's a big issue.

The anecdotal comments I've seen from Olympus users would seem to indicate that the fall-off point is a lot closer to 200mm than 2000mm. But I can't claim rigorous testing for this.
Logged

Farmer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2848
In camera image stabilisation
« Reply #36 on: January 09, 2008, 03:42:52 pm »

Quote
Nope. Canon's IS system is gyro-based; it will work just fine even with a lens cap on. It is equally effective no matter what the light level is. You can hear a faint whirring noise when the gyros spin up while IS is active.[a href=\"index.php?act=findpost&pid=166109\"][{POST_SNAPBACK}][/a]

Sorry, I was unclear.  I meant that if the movement is so much that there is no longer a path for the light of the subject to reach the lens then it doesn't matter how well stablised you are, you can't take the picture.  No system can make light bend around a hood and back into the lens :-)  In other words, there's a point at which it doesn't matter if the system could theoretically compensate because the path is physically blocked.

In camera systems functiong fine with the lens cap on too, of course.

Quote
Most tests of Canon's IS lenses indicate 2-3 stops of stabilization. My personal experience agrees with this.

This doesn't tally with my experience up to 600mm comparing Canon with Konica-Minolta where I would say it's 1.5 - 2 stops.  The newer Sony update to the system is roundly accepted as superior to the original KM implimentation.

If you've been using glass longer than 600mm then I will take your word for it as I have no experience of my own.

Quote
The anecdotal comments I've seen from Olympus users would seem to indicate that the fall-off point is a lot closer to 200mm than 2000mm. But I can't claim rigorous testing for this.

I've not used the Olly system.  The KM system (and from reports the Sony version) continues to function without noticable drop-off up to the 500-600 range, but that is not tested with any scientific rigour.

Again, I don't think there's any doubt that the in lens systems have an advantage at longer focal lengths, but for the majority of users, over 600mm is not something they're going to experience (remember, the DSLR market is again growing and it's heading down the track, away from high end users because that market is already saturated), so it probably doesn't matter because either system is going to function just fine for them.  Also, developing in camera technology will help to support any company's non-DSLR range.

The idea of a hybrid system for high end users could offer the absolute best at the appropriate price point.
Logged
Phil Brown

DarkPenguin

  • Guest
In camera image stabilisation
« Reply #37 on: January 09, 2008, 04:16:05 pm »

In camera stabilization doesn't get rid of the viewfinder bouncies.
Logged

Farmer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2848
In camera image stabilisation
« Reply #38 on: January 09, 2008, 04:44:39 pm »

Of course, but some people find that more or less useful depending on what they're shooting and their own personal preference.  That feature of in lens stablisation can also be annoying at times when you *want* to quickly adjust a small amount to see what's just out of frame and you have to wait.  That's a minor, minor issue and overall it's a useful feature, but it's not anywhere near in the same league of being useful as the actual stablisation and stop gains possible as a result.
Logged
Phil Brown

aaykay

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 359
In camera image stabilisation
« Reply #39 on: January 09, 2008, 05:27:01 pm »

Logged
Pages: 1 [2] 3   Go Up