Curved sensors - good or a bad development ?

[TacticDesigns]

I've just read what has been posted on this site about curved sensors.

Haven't really read too much.

But 2 questions.

1. Would curved sensors work well with trying to capture 180 or 360 degree pictures?

[...]

You're thinking of those old panoramic cameras where the film was held on a semi-circular arc and the lens was rotated mechanically. These are very subtly curved sensors which compensate for the field curvature, rather than using optical elements in the lens to correct it. This Wikipedia page gives some examples:

https://en.wikipedia.org/wiki/Petzval_field_curvature

J.

+1

Those panoramic cameras are exactly what I was thinking.

With the film, you could only bend (roll) it in one direction (axis).

But, with a curved sensor, it would be like you were rolling the film (sensor) in all directions.

Would that be the case?

Get on curved sensor and you've got ~180 degrees. Have 2 sensors (one on each side of the camera, like the Ricoh Theta) and you've got 360 degrees, but better IQ?

Take care & Happy Shooting!


--
My Personal Flickr Favs . . .
http://www.flickr.com/photos/tacticdesigns/sets/72157631300869284/
[FL][RP][LS]

this is the Sony curved sensor discussed here :

+1

Yes. I guess not as wrapped as the film was in the old panoramic cameras.

But if this gives some sort of advantage to IQ with wide angle lenses, then would this be a progression for wide angle cameras, including fisheye or panoramic cameras?

Yes - in the panoramic film cameras the film remained as what the mathematicians would call a developable surface, i.e. you could roll it out flat without stretching or creasing it - they would say that the film had zero Gaussian curvature - whereas the surface of these sensors it is really a curved surface; though with the current technology you clearly can't bend it very much.

+1

What is the benefit of being a "developable surface"?

I can see that because the sensor is manufactured into a curve (somewhere along the process) and then it's shape becomes set, it can't be straightened.

But . . . since the sensor records data, it is possible to "flatten" the representation of that data . . . it can be projected digitally onto a flat surface.

You don't need to flatten the data.

So, it may not be physically possible to flatten the sensor into a developable surface, but digitally it is possible? Yes? So, sort of a digital version of developable surface?

Not what's going on I think. In this case, the lens and the sensor will be matched and the resulting image will be flat field. The point is that a lens already struggles to be flat and we do many corrections to compensate for this.

Things like chromatic aberration and distortion?

And spherical aberration and coma etc.

+1

Now, there will be fewer corrections required because we'll stop trying to make it project a ruler flat image. Instead, we'll let it have a curve that matches the curve of the sensor. The result will be a naturally flat image.

I guess I'm trying to understand what you mean by the result will be a naturally flat image?

Do you mean that it will be a cleaner image (less aberrations), but the image still needs to go through digital processing / projection in order to be printed as a flat image, like a print?

What I'm saying is that it will be flat and not require any flattening - at least this is my expectation. IE, a lens wants to project an image that isn't flat. But if we meet that with a curved sensor, then the image will be flat.

+1

So we would design lenses that are happily not projecting a flat image in such a way that the curvature of the image matches that of the image sensor. So now, where you would have had poor edge performance, we've moved the sensor surface a bit to cure that. (For instance.)

Ok. I think we are using "flat" at different parts of the workflow.

I think you are referring to "flat" when the lens projects the image onto the sensor. In which case, yes . . . a lens designed to project onto a curved sensor could project a cleaner image.

The flat I'm talking about is when you take the image captured from the curved sensor and try to print it on flat piece of paper.

I'm lost with that question. At the very edge, the image will be sharp and free from aberration in theory and so in the center - the sensor is just transmitting pulses for a given location. Are you thinking as if you would flatten out the sensor and therefore have gaps in the image that need filling in? I guess I'm assuming the sensor is curved after the photo sites have been placed. Maybe not?

It actually wasn't a question.

I understand that we are using the word "flat" in different ways.

I agree with everything you say.

My use of the word "flat" is when you take that image that has been captured and you try to print it on a flat piece of paper. That's all.

I was just trying to understand how you were using the word "flat".

No, we're using the word the same way, but I'm not sure why the question about printing on a piece of paper. I not seeing how it's a problem. It's possible that I'm overlooking something and over simplifying.

Sorry.

My mind keeps jumping all over the place.

Originally I was thinking that this curved sensor would work nicely with a product like the Ricoh Theta. A camera that captures ~180 degree on each side, and stitches it together.

So my original thought was . . . when shooting RAW, you get the RAW image out of the curved sensor.

My first question was, would this be a good file to pull into something like Hugin to use in stitching? Because it would be a cleaner file than a flat sensor? Cleaner in that there is less aberrations? And wouldn't you want to preserve the curved aspect and not edit the file and make it flat before sending it to software like Hugin?

But . . . is that how the file would be saved? Or would it be "flattened"?

Sort of like what this person was asking . . .

Loading…

photo.stackexchange.com

And then I started thinking . . . ok, let's say you shoot RAW and save the picture.

Now you want to post the picture to Facebook, or print it to a piece of paper. Are you going to require a computer to process the picture, or will the camera itself be able to kick out a flat projection to share?

I am assuming the camera will be able to do that, and have built-in wifi to transfer it to your phone.

But then again . . . instead of sharing a flat image, why not upload a 360 degree view of where you are. So, what format would that be?

So, it was more . . . ok, you've got the capture. It's cleaner with less aberrations. But, then . . . now what? Where / how does that file go?

LOL.

Take care & Happy Shooting!


--
My Personal Flickr Favs . . .

My Favs

Some of my personal favourites from over the years . . .

www.flickr.com

[FL][RP][LS]

 

[Mark S Abeln]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

 

[stevo23]

I've just read what has been posted on this site about curved sensors.

Haven't really read too much.

But 2 questions.

1. Would curved sensors work well with trying to capture 180 or 360 degree pictures?

[...]

You're thinking of those old panoramic cameras where the film was held on a semi-circular arc and the lens was rotated mechanically. These are very subtly curved sensors which compensate for the field curvature, rather than using optical elements in the lens to correct it. This Wikipedia page gives some examples:

https://en.wikipedia.org/wiki/Petzval_field_curvature

J.

+1

Those panoramic cameras are exactly what I was thinking.

With the film, you could only bend (roll) it in one direction (axis).

But, with a curved sensor, it would be like you were rolling the film (sensor) in all directions.

Would that be the case?

Get on curved sensor and you've got ~180 degrees. Have 2 sensors (one on each side of the camera, like the Ricoh Theta) and you've got 360 degrees, but better IQ?

Take care & Happy Shooting!


--
My Personal Flickr Favs . . .
http://www.flickr.com/photos/tacticdesigns/sets/72157631300869284/
[FL][RP][LS]

this is the Sony curved sensor discussed here :

+1

Yes. I guess not as wrapped as the film was in the old panoramic cameras.

But if this gives some sort of advantage to IQ with wide angle lenses, then would this be a progression for wide angle cameras, including fisheye or panoramic cameras?

Yes - in the panoramic film cameras the film remained as what the mathematicians would call a developable surface, i.e. you could roll it out flat without stretching or creasing it - they would say that the film had zero Gaussian curvature - whereas the surface of these sensors it is really a curved surface; though with the current technology you clearly can't bend it very much.

+1

What is the benefit of being a "developable surface"?

I can see that because the sensor is manufactured into a curve (somewhere along the process) and then it's shape becomes set, it can't be straightened.

But . . . since the sensor records data, it is possible to "flatten" the representation of that data . . . it can be projected digitally onto a flat surface.

You don't need to flatten the data.

So, it may not be physically possible to flatten the sensor into a developable surface, but digitally it is possible? Yes? So, sort of a digital version of developable surface?

Not what's going on I think. In this case, the lens and the sensor will be matched and the resulting image will be flat field. The point is that a lens already struggles to be flat and we do many corrections to compensate for this.

Things like chromatic aberration and distortion?

And spherical aberration and coma etc.

+1

Now, there will be fewer corrections required because we'll stop trying to make it project a ruler flat image. Instead, we'll let it have a curve that matches the curve of the sensor. The result will be a naturally flat image.

I guess I'm trying to understand what you mean by the result will be a naturally flat image?

Do you mean that it will be a cleaner image (less aberrations), but the image still needs to go through digital processing / projection in order to be printed as a flat image, like a print?

What I'm saying is that it will be flat and not require any flattening - at least this is my expectation. IE, a lens wants to project an image that isn't flat. But if we meet that with a curved sensor, then the image will be flat.

+1

So we would design lenses that are happily not projecting a flat image in such a way that the curvature of the image matches that of the image sensor. So now, where you would have had poor edge performance, we've moved the sensor surface a bit to cure that. (For instance.)

Ok. I think we are using "flat" at different parts of the workflow.

I think you are referring to "flat" when the lens projects the image onto the sensor. In which case, yes . . . a lens designed to project onto a curved sensor could project a cleaner image.

The flat I'm talking about is when you take the image captured from the curved sensor and try to print it on flat piece of paper.

I'm lost with that question. At the very edge, the image will be sharp and free from aberration in theory and so in the center - the sensor is just transmitting pulses for a given location. Are you thinking as if you would flatten out the sensor and therefore have gaps in the image that need filling in? I guess I'm assuming the sensor is curved after the photo sites have been placed. Maybe not?

It actually wasn't a question.

I understand that we are using the word "flat" in different ways.

I agree with everything you say.

My use of the word "flat" is when you take that image that has been captured and you try to print it on a flat piece of paper. That's all.

I was just trying to understand how you were using the word "flat".

No, we're using the word the same way, but I'm not sure why the question about printing on a piece of paper. I not seeing how it's a problem. It's possible that I'm overlooking something and over simplifying.

Sorry.

My mind keeps jumping all over the place.

Originally I was thinking that this curved sensor would work nicely with a product like the Ricoh Theta. A camera that captures ~180 degree on each side, and stitches it together.

So my original thought was . . . when shooting RAW, you get the RAW image out of the curved sensor.

My first question was, would this be a good file to pull into something like Hugin to use in stitching? Because it would be a cleaner file than a flat sensor? Cleaner in that there is less aberrations? And wouldn't you want to preserve the curved aspect and not edit the file and make it flat before sending it to software like Hugin?

But . . . is that how the file would be saved? Or would it be "flattened"?

Shooting it with the matched lens, you're going to get flat. Optically, it's going to be flat when you get the image from the sensor and not require any "flattening" correction via processing. That's the idea anyhow. So to actually use it for panoramics, I think you have to use a standard lens and that would mess up your corrections. And I don't think it has a tremendous amount of curvature in the first place.

Sort of like what this person was asking . . .

https://photo.stackexchange.com/questions/79023/what-kind-of-files-can-i-get-out-of-a-ricoh-theta-s

And then I started thinking . . . ok, let's say you shoot RAW and save the picture.

Now you want to post the picture to Facebook, or print it to a piece of paper. Are you going to require a computer to process the picture, or will the camera itself be able to kick out a flat projection to share?

If you're still talking about the curved sensor/lens combination only and not the Theta, no. It's already going to be a flat image without special processing.

I am assuming the camera will be able to do that, and have built-in wifi to transfer it to your phone.

But then again . . . instead of sharing a flat image, why not upload a 360 degree view of where you are. So, what format would that be?

I don't think this gives any advantage for panoramics over any other sensor by virtue of it's curvature.

So, it was more . . . ok, you've got the capture. It's cleaner with less aberrations. But, then . . . now what? Where / how does that file go?

It's just like anything else at this point.

 

[stevo23]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

Well said in a single paragraph what I've been struggling to put in words.

 

[Digitalis32]

Like curved monitors, it is just a fad. Regarding optical performance, it will improve vignetting performance with wide angle lenses, it is possible it will have the reverse effect on macro and long telephoto lenses which tend to have very focus fields and inherently low vignetting.

Also camera manufacturers might just as well change lensmounts while they are at it, it would keep people from complaining their shiny new toy doesn't work properly with their walkaround lens.

 

[Digitalis32]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

Well said in a single paragraph what I've been struggling to put in words.

*facepalm* there would have to be something very, very wrong with a lenses distortion profile if a curved sensor was needed to rectify it.

Spherical aberration and coma will still be problematic on a curved sensor, they could even get worse. Chromatic aberration will also be just as problematic on curved sensors as it is on flat ones.

--
A camera is just a camera. Who is behind, it matters far more.

 

[Nigel S]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

 

[FrancoD]

I think that the idea is to have smaller lenses with fewer elements.

So probably for phones first and compact single lens large sensor cameras.

 

[stevo23]

Like curved monitors, it is just a fad. Regarding optical performance, it will improve vignetting performance with wide angle lenses, it is possible it will have the reverse effect on macro and long telephoto lenses which tend to have very focus fields and inherently low vignetting.

I think it will improve more than vignetting.

 

[stevo23]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

Well said in a single paragraph what I've been struggling to put in words.

*facepalm* there would have to be something very, very wrong with a lenses distortion profile if a curved sensor was needed to rectify it.

Spherical aberration and coma will still be problematic on a curved sensor, they could even get worse. Chromatic aberration will also be just as problematic on curved sensors as it is on flat ones.

*facepalm* you're wrong.

 

[GodSpeaks]

I can see that a lens designed for a specific curved sensor might out perform a "normal" lens and a flat sensor. Mainly I would expect improved corner sharpness and the reduction of some aberrations.

But the key point would be that it would have to be a matched pair (lens and sensor) and so in practice would only work with a fixed lens camera. I really cannot see it being practical for interchangeable lens cameras at all.

I can see it being quite useful for some very specialized applications.

That said, I will NOT be rushing out to buy any curved sensor camera any time soon.

I consider this as about as practical a development, for normal photography, as curved screen or 3D HD/UHD TVs.

--
The greatest of mankind's criminals are those who delude themselves into thinking they have done 'the right thing.'
- Rayna Butler

 

[stevo23]

I'm going to assume that the raw file will be flat without any hint of curvature. The curvature as delivered by the lens will be precisely canceled by the curvature of the sensor. You'll get a well-corrected rectilinear image.

 

[stevo23]

I'd assume that in most cases, the image would be corrected to rectilinear

Animal and human eyes have curved sensors and low element-count lenses, yet my view appears to be rectilinear. Lines that are straight in real life appear to be straight to my eye.

+1

Good example.

Is there an example of what image gets projected onto the back of an eye? . . . like a simulation?

That would be pretty cool to see.

And then see how that is translated into what we see in our minds eye, so to speak.

Does anyone know of a sample of that?

It's a flat image - just like what you see. The rods and cones transmit the signals and they are mapped to a flat field.

 

[Turbguy1]

Actually (for most normally sighted humans) it is NOT flat, but a geometrically wide field corrected 3D representation of the world visible to us.

When you stop and ponder it, just what IS going on inside our minds, that practically instantaneously projects the 3D visual world around us??

It's really quite a feat!

That said, there are plenty of examples (aka, optical illusions), that can "fool" whatever IS going on.



--
Wayne
See more at:
http://www.flickr.com/photos/turbguy/

 

[stevo23]

Actually (for most normally sighted humans) it is NOT flat, but a geometrically wide field corrected 3D representation of the world visible to us.

But I think his question related to what one eye is doing as an analogy to a curved sensor. At least that's what I think he's after.

When you stop and ponder it, just what IS going on inside our minds, that practically instantaneously projects the 3D visual world around us??

It's really quite a feat!

That said, there are plenty of examples (aka, optical illusions), that can "fool" whatever IS going on.

Articles anywhere you can share?

 

[stevo23]

I can see that a lens designed for a specific curved sensor might out perform a "normal" lens and a flat sensor. Mainly I would expect improved corner sharpness and the reduction of some aberrations.

But the key point would be that it would have to be a matched pair (lens and sensor) and so in practice would only work with a fixed lens camera. I really cannot see it being practical for interchangeable lens cameras at all.

I can see it being quite useful for some very specialized applications.

That said, I will NOT be rushing out to buy any curved sensor camera any time soon.

I consider this as about as practical a development, for normal photography, as curved screen or 3D HD/UHD TVs.

 

[Turbguy1]

No articles, just ponder it...

Even with ONE eye, you still have 3D "clues" that our "image processing" uses to generate a 3D world around us. Binocular vision is not an absolute necessity, but is a GREAT enhancement!

--
Wayne
See more at:
http://www.flickr.com/photos/turbguy/

 

[dynaxx]

Me, me, me !

Don't we care more about the development of photography than the personal, petty losses that we experience when technology moves forward ?

Curved ( flexible with adjustable curvature, possibly ) sensors have the potential to give us better IQ and smaller/cheaper/lighter/longer lasting lenses. I may be mistaken but I would guess that the width/height of a curved sensor would be less than a flat sensor with the same area so smaller medium format or FF bodies may be on the cards. Is that good or bad ?

This comment is very similar to one I heard at work, bemoaning the phasing out of deisel engined vehicles.

 

[antoineb]

- curved sensors seem like a smart and logical development.

- this being said they have yet to find their way in mainstream cameras.

- when they do, it is likely that they first find their way into Smartphones. Because smartphone companies have way more money for R&D and have way more products and faster product cycles to play with.

- when curved sensors finally arrive to dedicated cameras, then yes lenses optimised for flat sensors will lose some performance at the edges. But it's always the same thing: for most photography, what happens at the edges is not relevant - except perhaps on some landscapes. And, many lenses still do not do such a great job at the edges because it's just very difficult to achieve: either you leave the light alone and you lose some focus precision, or you tr y to get focus precision but you have to compromise on other factors.

BOTTOM-LINE: I think you can safely continue to buy any good lens you feel you need.

If worse come to worst, say in 6 months Canon and Nikon and Sony announce enthusiast 24x36 sensor cameras with curved sensors, then you current equipment will still function just fine. But it's more like 6 years, at best, and perhaps more like 10 years. If that.

So I see some of the big names are lodging patents on new curved sensors and lenses to work with them. I'm pondering what this means for those of us with established and extensive lens collections - will these become obsolete and we have to buy a whole new set of lenses if the future = curved sensors ? I know lens development happens and in time replacement of lenses might need to happen but usually you envisage this happening gradually and selectively since some lenses can last us 30+ years of great service and results. It would be a bitter pill if we had to ditch all our existing lenses to use with shiny new bodies that featue curved sensors? Do you think such bodies would exist in parallel with bodies that used existing sensor tech?

I odn't mean this to be a doom and gloom post but I think it raises a valid point - is this a good time to stop buying expensive lenses if they might not work on new cameras featuring these new curved sensors ?

--
"When words become unclear, I shall focus with photographs. When images become inadequate, I shall be content with silence." Ansel Adams.

 

[stevo23]

No articles, just ponder it...

Even with ONE eye, you still have 3D "clues" that our "image processing" uses to generate a 3D world around us. Binocular vision is not an absolute necessity, but is a GREAT enhancement!

Great, I get it. But the image in your eye that falls on your retina - flat. No need to "decode" it into flat because it's already flat. And to go from curved to flat, you just map the rods and cones to a flat surface.