tilt via IBIS?

[incoherent1]

What are the possibilities of using ibis movement of the sensor to tilt the sensor to achieve/emulate the focus plane tilt of tilt shift lenses?

 

[kenw]

IBIS systems do not tilt the sensor so they can’t be used to do that. All IBIS motion is in the plane of focus.

 

[incoherent1]

IBIS systems do not tilt the sensor so they can’t be used to do that. All IBIS motion is in the plane of focus.

Yeah, but any reason the ibis systems cannot tilt the sensor?

 

[BobORama]

What are the possibilities of using ibis movement of the sensor to tilt the sensor to achieve/emulate the focus plane tilt of tilt shift lenses?

Available solutions move the sensor in the focal plane, and optionally rotate the sensor along the perpendicular to the focal plane. So it won't do that.

Camera makers could create such a system by putting the whole works on a 3 point motorized mount to tilt the sensor plane. ( Imagine a tripod with 1 fixed and 2 motorized micrometer screws, etc.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos

 

[kenw]

IBIS systems do not tilt the sensor so they can’t be used to do that. All IBIS motion is in the plane of focus.

Yeah, but any reason the ibis systems cannot tilt the sensor?

Well I’m sure you could design one to do that, but none of the existing systems do. Maintaining the plane of focus to a very high degree of precision is absolutely critical so designing another two axes of actuation and control into the IBIS system would seem a tall order with a lot of expense for what would be a rather niche feature.

 

[MCLV]

What are the possibilities of using ibis movement of the sensor to tilt the sensor to achieve/emulate the focus plane tilt of tilt shift lenses?

As far as I understand, required tilt can be sometimes quite significant, see https://en.wikipedia.org/wiki/Scheimpflug_principle and there might not be physically enough room to tilt the sensor sufficiently without hitting shutter and/or other components.

 

[incoherent1]

What are the possibilities of using ibis movement of the sensor to tilt the sensor to achieve/emulate the focus plane tilt of tilt shift lenses?

As far as I understand, required tilt can be sometimes quite significant, see https://en.wikipedia.org/wiki/Scheimpflug_principle and there might not be physically enough room to tilt the sensor sufficiently without hitting shutter and/or other components.

Thank you, this addresses the why.

 

[D Cox]

If you really need tilt, you can use a Cambo Actus.

I'm not convinced that the movements that are provided in a large format monorail camera are needed at all in 35mm or smaller formats.





A shot like this needs all the movements on a 5x4 camera, but is all within the depth of field on 24x36mm format.

Don Cox

 

[Enginel]

No. The article exaggerates tilts as illustrations are meant to explain principle, not give exact scales.

Imagine such use case. You are going to take a image of A4 paper document (210x297) on a FF camera. The paper is at whopping 45 degrees to optical axis (because you want to avoid specular reflections or some other cause). 45 degrees => tangent 1. Your lens is set to 1:8 reproduction ratio. Lateral magnification is β=-0.125. Longitudinal magnification is square of lateral magnification, i.e. 1/64. So image of the paper projected by your lens will have tangent to optical axis 64 times smaller than the object does. If tilt occurs on long (36 mm) side then the left edge will be shifted 36/64 = 0.56 mm relative to right edge.

That is, you need only 0.56 mm (or 0.28 mm) to adjust focus for A4 sheet slanted at whopping 45 degrees to optical axis.

Is 0.56 mm so big that it would make all of this unusable in practice? For view cameras of course tilts be more relative to image diagonal. For m43 cameras tilts be even more minuscule, as β closer to zero.

Also, focal plane shutter is very likely soon to be gone due to sensor progress, as it'd be slower than electronic rolling shutter.

 

[Enginel]

Necessity for tilts comes from entrance pupil size, not sensor diagonal. A 210/4.5 on view camera would be almost identical to 50/1.2 on 36x24 camera.


Is it possible to use Cambo Actus to use native Sony lenses or at least Canon AF lenses (with focus-by-wire control)? I guess no.
A person asks about a feature which requires 20 grams to implement in cameras, works with all lenses, including long zooms and ultrafast normals, and gets advise to use 1 kg thing which would require to buy another set of fully manual lenses to use on it. It, of course, solves some certain use cases, but in no way a replacement for what OP asked for.

 

[Enginel]

You're asking for a feature that does not exist.

Photography folks are conservative and will invent bogus reasons to explain why it doesn't exist.

Imagine SLRs did not exist, and you were trying to explain to people that you're thinking of a camera where a flappy mirror used for framing, and also another flappy mirror mounted on the flappy mirror diverts light to relay system which projects to AF module, and all of this bounces 10 times per second, These problems with flappy-mirror-on-a-flappy-mirror have been solved.

You know the real reason why the feature you asked doesn't exist?
This feature is contra DSLR design, as OVF would not display any changes produced by tilted sensor.
A mirrorless camera is defined by what it does not have. Otherwise, it is supposed to look, act and feel as a DSLR. It is not supposed to add new features over a DSLR. Except, maybe Eye AF and histogram in viewfinder. That is okay. And a little saving in costs (which are not passed to customers). Despite some talking about 'mirrorless revolution' we are stuck by constraints.

 

[incoherent1]

No. The article exaggerates tilts as illustrations are meant to explain principle, not give exact scales.

Imagine such use case. You are going to take a image of A4 paper document (210x297) on a FF camera. The paper is at whopping 45 degrees to optical axis (because you want to avoid specular reflections or some other cause). 45 degrees => tangent 1. Your lens is set to 1:8 reproduction ratio. Lateral magnification is β=-0.125. Longitudinal magnification is square of lateral magnification, i.e. 1/64. So image of the paper projected by your lens will have tangent to optical axis 64 times smaller than the object does. If tilt occurs on long (36 mm) side then the left edge will be shifted 36/64 = 0.56 mm relative to right edge.

That is, you need only 0.56 mm (or 0.28 mm) to adjust focus for A4 sheet slanted at whopping 45 degrees to optical axis.

Is 0.56 mm so big that it would make all of this unusable in practice? For view cameras of course tilts be more relative to image diagonal. For m43 cameras tilts be even more minuscule, as β closer to zero.

Also, focal plane shutter is very likely soon to be gone due to sensor progress, as it'd be slower than electronic rolling shutter.

interesting....

 

[hjulenissen]

I guess that a tilting sensor would typically be quite slow (e.g. 0.5 seconds to go from one extreme to the other), and absolute tilt would be critical (you want 0 degrees of tilt to really be 0 degrees).

That is unlike IBIS that needs to react with little latency and many times a second to camera movements, and where a tiny "DC offset" might not be disaterous?

I think it is an interesting idea, but it seams that there will be little overlap between the IBIS mechanics and the tilt mechanics.

I for one would love the ability to have a "2-dimensional autofocus" where the system would sweep through different focus distance settings and sensor tilt in order to maximize actual sharpness of a scene region.

-h

 

[MCLV]

I guess that a tilting sensor would typically be quite slow (e.g. 0.5 seconds to go from one extreme to the other), and absolute tilt would be critical (you want 0 degrees of tilt to really be 0 degrees).

That is unlike IBIS that needs to react with little latency and many times a second to camera movements, and where a tiny "DC offset" might not be disaterous?

I think it is an interesting idea, but it seams that there will be little overlap between the IBIS mechanics and the tilt mechanics.

There are two options how to combine IBIS stage with tilting stage:

1. Tilting stage is put on the IBIS stage. This would require more robust and powerful IBIS stage in order to maintain its performance because larger mass has to be moved around.
2. IBIS stage is mounted on tilt stage. This would require more physical space to accommodate tilting movement of whole IBIS stage.

Neither solution looks as easy from engineering point of view. And it also seems that both options would consume more physical space in camera than IBIS alone.

I for one would love the ability to have a "2-dimensional autofocus" where the system would sweep through different focus distance settings and sensor tilt in order to maximize actual sharpness of a scene region.

-h

 

[MCLV]

No. The article exaggerates tilts as illustrations are meant to explain principle, not give exact scales.

Imagine such use case. You are going to take a image of A4 paper document (210x297) on a FF camera. The paper is at whopping 45 degrees to optical axis (because you want to avoid specular reflections or some other cause). 45 degrees => tangent 1. Your lens is set to 1:8 reproduction ratio. Lateral magnification is β=-0.125. Longitudinal magnification is square of lateral magnification, i.e. 1/64. So image of the paper projected by your lens will have tangent to optical axis 64 times smaller than the object does. If tilt occurs on long (36 mm) side then the left edge will be shifted 36/64 = 0.56 mm relative to right edge.

That is, you need only 0.56 mm (or 0.28 mm) to adjust focus for A4 sheet slanted at whopping 45 degrees to optical axis.

Is 0.56 mm so big that it would make all of this unusable in practice? For view cameras of course tilts be more relative to image diagonal. For m43 cameras tilts be even more minuscule, as β closer to zero.

Also, focal plane shutter is very likely soon to be gone due to sensor progress, as it'd be slower than electronic rolling shutter.

I admit I didn't calculate it like this. I just looked up tilt range for FF lenses, which seems to be up to 8.5 degrees. That corresponds approximately to 3.6 mm over FF sensor height or 5.4 mm over FF sensor width.

 

[hjulenissen]

I largely count on the smartphone camera industry to do the heavy lifting when it comes to pouring PhD hours into investigating different means of improving image quality through novel ideas and the miracle of mass production.

When image sensor pixel density and lens improvements have reached the point where diffraction and fundamental physics (along with unit size) limits image quality, you need to think differently to improve IQ. Machine learning seems to be the answer to all problems these days, but also shooting long sequences of short exposures, to be assembled computationally into one, superior image.

Perhaps tilting sensor is a sensible way to increase sharpness in diffraction limited cameras for some scenes. But would not a smartphone manufacturer rather do focus stacking in order to bring out details at both near and far?

-h

 

[D Cox]

I largely count on the smartphone camera industry to do the heavy lifting when it comes to pouring PhD hours into investigating different means of improving image quality through novel ideas and the miracle of mass production.

When image sensor pixel density and lens improvements have reached the point where diffraction and fundamental physics (along with unit size) limits image quality, you need to think differently to improve IQ. Machine learning seems to be the answer to all problems these days, but also shooting long sequences of short exposures, to be assembled computationally into one, superior image.

Perhaps tilting sensor is a sensible way to increase sharpness in diffraction limited cameras for some scenes. But would not a smartphone manufacturer rather do focus stacking in order to bring out details at both near and far?

-h

I think Olympus m4/3 cameras also have built in stacking. Focus bracketing is also quite common and there is plenty of software to do the stacking.

But all this requires a still subject.

Don Cox

 

[hjulenissen]

I think Olympus m4/3 cameras also have built in stacking. Focus bracketing is also quite common and there is plenty of software to do the stacking.

But all this requires a still subject.

Don Cox

I think that stacking can be done in an intelligent manner where some movements can be tolerated. For tourists shooting a monument with other tourists in the way, it would be trivial to implement e.g. a median operation to remove moving subjects as long as the camera is on a tripod (or can be "stabilized" in some other way).

I think that a tilting sensor does not generally work well with scene movement, but some movements can be tolerated. If you are hand holding your camera, a slow-moving sensor tilt may have a hard time tracking a horizontal beach in front of you. On the other hand, a surfer on the waves at a distance is no problem.

My point is that (focus) stacking reduce the problem to one of general camera hw and compute/algoritms. Smartphone manufacturers would rather solve that first, rather than implementing novel hardware, unless there is real significant benefits to going the hw route.

-k

 

[MCLV]

I think Olympus m4/3 cameras also have built in stacking. Focus bracketing is also quite common and there is plenty of software to do the stacking.

But all this requires a still subject.

Don Cox

I think that stacking can be done in an intelligent manner where some movements can be tolerated. For tourists shooting a monument with other tourists in the way, it would be trivial to implement e.g. a median operation to remove moving subjects as long as the camera is on a tripod (or can be "stabilized" in some other way).

I think that a tilting sensor does not generally work well with scene movement, but some movements can be tolerated. If you are hand holding your camera, a slow-moving sensor tilt may have a hard time tracking a horizontal beach in front of you. On the other hand, a surfer on the waves at a distance is no problem.

Yes, I don't see any issues with moving subjects when using the tilt since you get everything in a single shot. However, the issue might be to adjust the tilt quickly enough when moving the camera around. also the question would to whether the tilt would be adjusted automatically or manually.

My point is that (focus) stacking reduce the problem to one of general camera hw and compute/algoritms. Smartphone manufacturers would rather solve that first, rather than implementing novel hardware, unless there is real significant benefits to going the hw route.

-k

I agree that focus stacking is a very attractive solution when increased DOF is needed without many of complications of tilting either the lens or sensor.