Patent: Variable Curved Image Sensor Concept

[dak723]

I hope these curved sensors go no where, as it would make Canon's new lens design job easier, 3rd party lenses would have to be Canon-specific, and the general consumer would lose as a result.

Existing lenses would work the same and future lenses designed for this sensor should have much better performance on the edges and corners. Sounds like a win win situation for everyone.

 

[Mt Spokane Photography]

There have been number of Canon patents for a variable curved sensor, but apparently the issues arising from stresses in bending the sensor made them impractical. The latest one is at least the 5th one. I wish Keith would publish the patent number, the link to the Japanese patent site never seems to bring up the patent for me, and I've been unable to link it in my posts.

I believe this one is for a method of bending the sensor mechanically using a mesh reinforcement to distribute the stress. I doubt that it will happen, but Canon obviously sees a advantage for curved sensors and has been working actively on them. Smaller, lighter lenses to match a small mirrorless camera? I actually think a curved sensor is going to happen sooner than we might think, Canon usually does things in stages, so first we may see a fixed lens camera where its easier.

http://thenewcamera.com/canon-working-on-curve-sensor-new/ (2016-173496)

http://thenewcamera.com/canon-patent-hybrid-curve-sensor/ (2016-201425)

http://thenewcamera.com/canon-patent-curve-sensor-iii/ (2016-197663)

http://thenewcamera.com/canon-patent-hybrid-curve-sensor-with-piezoelectric-elements/ (2016-213571)

 

[Jopa]

I could be wrong, but I think that's one of those "protective" patents that unlikely will see the light.

 

[Antono Refa]

I could be wrong, but I think that's one of those "protective" patents that unlikely will see the light.

The patent was published, so it has seen the light of day ;-)

Question is whether it would see the back side of a lens.

 

[keithcooper]

... I wish Keith would publish the patent number, the link to the Japanese patent site never seems to bring up the patent for me, and I've been unable to link it in my posts.
...

Sorry for the lack of info with some - the links in the hi-lows blog fail with Google Translate, since the Japanese patent office site blocks 'automated access' or some such

I'll try and remember to put the direct patent number in from the Japanese text - the only translated bits I see are after copying blocks of text directly into the translation.

Update: Added a link and found that the Japanese patent site seems to dislike deep links, so I've put a URL in, in full

 

[Bahrd]

What I had in mind is the smaller microlenses will direct less light into the pixels than gapless microlenses.

That is a good point.

If there is 5% gap between lens due to flexing, the reduction in area will be 10%. Not a terribly large difference, about a 1/6th of a stop.

Cannot microlenses remain gapless but simply put upside down? That is, with a flat side on "top"?

 

[Khufu]

Well, we won't be getting any true sub-44mm single-lens-element lenses on a 44mm flange EF mount, so it's just going to necessitate differently calibrated witchcraftery for wide angle lenses to work on an EF mount, right? Multiple lens lines using the same mount system during the crossover phase would also be a bit messy...

I can only really see this being put to use in a new system; fixed lens to begin with would be my guess, then maybe on to a shorter flange interchangable-lens mirrorless gizmo? A whole Full Frame sensor could wrap itself around a pinhole-like element at around 10mm for close to 180 degrees... sort of, couldn't it?

The idea of 24mm lenses actually at 24mm is cool! There's a chance my basic ideas of physics are way off, I'm just having fun thinking about possibilities!

 

[deleteme]

While this is clever, it presumes a flexible sensor that AFAIK is not in existence.

My question is: Can one patent the idea of a mechanism to control the curve of a flexible sensor when that sensor inhabits the realm of fantasy? I suppose it could also be a general patent that describes the adjustment of a flexible material in general.

 

[keithcooper]

While this is clever, it presumes a flexible sensor that AFAIK is not in existence.

My question is: Can one patent the idea of a mechanism to control the curve of a flexible sensor when that sensor inhabits the realm of fantasy? I suppose it could also be a general patent that describes the adjustment of a flexible material in general.

There are plenty of working examples of curved sensors, so thin enough chips will bend, but this one adds adjustable bending. The actual patent discusses this in terms of a stacked sensor and looks at interchip communication.

Making a chip that bends repeatably is indeed a challenge...

 

[Khufu]

These sensors just need micro-hinges between the pixels, but hinges need oiling; I recall Nikon testing out an early self-oiling sensor in the D600...

 

[darth]

Years ago Microsoft and Canon had a patent license agreement and collaborated on some projects [https://www.cnet.com/news/microsoft-and-canon-shake-on-a-patent-sharing-agreement/].

Years later Microsoft developed a small curved sensor that is sharper than a "50mm 1.2 on a 1ds3" on a camera phone. [https://mspoweruser.com/microsoft-researchers-create-new-curved-camera-sensor-sharper-commercial-camera]
It is possible that they are sharing patents and ideas with Canon.

Microsoft claimed that a curved sensor reduces the number of glass elements needed since it doesn't need to correct a curved image to a flat sensor. Less glass elements mean a sharper image. If canon makes a curved sensor and a new lens line that has fewer elements but is curved sensor only (so EF, EF-S, EF-M, EF-curved) that only works on these new bodies, they can make new lenses that are light, small, fewer glass elements, and sharp, with hopefully less variation from fewer elements. Could be a real breakthrough and allow much higher MP counts that can blow away anything we have now. Looking forward to it.

 

[Mikehit]

Of course, it will need new processing algorithms - you are turning a 3D image of a curved sensor into a flat 2D image so need to remove the distortions. And if the curvature changes depending on the lens used, or maybe even the focal length used on a zoom, that means you will need a different algorithm for each situation because the rectilinear distortion will be vary.
The more I think about this the more challenges there seem to be....

 

[Antono Refa]

Of course, it will need new processing algorithms - you are turning a 3D image of a curved sensor into a flat 2D image so need to remove the distortions. And if the curvature changes depending on the lens used, or maybe even the focal length used on a zoom, that means you will need a different algorithm for each situation because the rectilinear distortion will be vary.
The more I think about this the more challenges there seem to be....

I think it's the other way around.

Lenses focus to a curved plane, rather than a flat one. Therefore a flat sensor's pixels are in the wrong place, requiring algorithms to correct the problem, e.g. sharpen to correct for lost sharpness.

A curved sensor moves the pixels to where the light actually focuses to, alleviating the need for some of those algorithms.

If corrections are needed, I suspect only one algorithm would be required, with parameters varying with the amount of curvature, etc.

 

[Mikehit]

I think it's the other way around.

Lenses focus to a curved plane, rather than a flat one. Therefore a flat sensor's pixels are in the wrong place, requiring algorithms to correct the problem, e.g. sharpen to correct for lost sharpness.

A curved sensor moves the pixels to where the light actually focuses to, alleviating the need for some of those algorithms.

If corrections are needed, I suspect only one algorithm would be required, with parameters varying with the amount of curvature, etc.

In flat sensors, most of the conversion to a flat surface is done by the lenses, not software - expensive lenses are expensive because of all the different elements that are needed to convert a spherical image from a simple lens into a flat surface with minimal distortion. How do you think they created flat images in the days of film? No software there.


In mathematical terms a flat surface is the surface of a sphere of infinite radius and once you look at it like that the impact of distortions become apparent. The smaller the radius the greater the curvature and greater the distortions induced by forcing it to a flat surface.
Try the reverse - smooth a piece of paper over a football and see how much creasing there is, then try the same with a tennis ball.

 

[mb66energy]

Just thinking:
What about multi area DPAF to calibrate sensor curvature for your lens (e.g. using stars at night as objects close to infinity)?

@mikehit: Maybe they calculate the lens to project an image to the curved sensor which is finally rectilinear if you look at it in the final image? So the optimization of field curvature / distortions is done by lens AND sensor together? An additional degree of freedom without introducing more glass between object and sensor ...

 

[djack41]

Pie in the sky. Canon has its customers discussing fairy tales. Meanwhile. Canon takes 5 years to update cameras with 4 year old technology.

 

[mb66energy]

Pie in the sky. Canon has its customers discussing fairy tales. Meanwhile. Canon takes 5 years to update cameras with 4 year old technology.

Maybe.

I have bought the 200D and I am very satisfied with the DPAF in live view especially for closeups of flowers in nature with wide open aperture settings. DPAF is some years old, but it is still unique as far as I know in terms of a combination speed / non-hunting / precision.