Sony's curved sensors - this could be the near future or even a present reality?

[Aglet]

it's most likely for small camera systems to produce good low light results using cheap, simple, fast lenses. First use will probably be phone- cameras. Getting the required amount of curvature in larger sensor systems is likely to be much more challenging as slight (focus) errors of a few microns could occur with temperature shifts and if you're looking to make a compact, large sensor camera with this then you're really gonna have to decrease the radius of that sensor's curvature. So, more likely practical for 1/2.5 and smaller sensors.
Still, will be very interesting if they can effectively achieve this for larger sensor formats.

 

[dgatwood]

Microlenses were invented and put on your DSLR sensors because they couldn't make curved sensors.

AFAIK, the purpose of microlenses is actually to work around the fact that the sensors aren't flat, but rather recessed down into pockets in the silicon. Light rays near the edge of the sensor would otherwise fall way off unless you use a huge image circle. Changing the angle of those pockets might provide an advantage over the microlenses, but doing so doesn't inherently require curving the actual face of the sensor.

Where do you read such garbage? The purpose of the microlens is to increase the light gathering surface of the pixel, so that the space on the sensor top where there is circuitry and other stuff between pixels can be used to gather light to send into the pixel. This problem becomes more acute towards the edge of the sensor because rays of light are striking the sensor at and angle.

That's certainly an added benefit, but what I said is quite correct; unless you're using a back-illuminated sensor, the actual light-sensitive part of the chip is inherently recessed below the wiring. This results in light fall-off near the edges if you don't have microlenses.

Don't believe me? Read Leica's description of their microlenses array:

"This optimized micro lens design, based on many years of precision optical engineering experience, captures and concentrates even the most oblique rays on the sensor and reliably prevents image brightness fall-off at the edges and corners of the image."

Their words, not mine.

On the other hand, using a curved sensor brings with it a lot of other problems. For starters, the screen you show the image on isn't curved, and neither is the photo paper. So the image would have to be warped to fit the medium. Even ignoring all the aliasing problems that will likely cause (which I'd expect to be considerable), unless I'm mis-thinking this, the resolution on the photo will probably be lower near the edges of the sensor—built-in corner softness, if you will—unless the pixel density changes as you get closer to the edge of the sensor, in which case the SNR will be worse near the center of the photo. Neither approach seems particularly desirable. Or maybe I'm missing something.

Sounds like you're over thinking this in order to find reasons that it isn't a good idea as you're wrong about everything.

Build a pin hole camera with your lens or just some big pieces of cardboard and see what happens out the back.

Yes, you can use a flat focal plane, which produces vignetting, or a curved focal plane, which results in distortion. I'm not seeing your point here.

 

[expatinasia]

These sensors could come out sooner than we think. SAR is reporting this:

The image of this Sony curved sensor created quite a buzz but some people argued that this tech may not coming any time soon. Well this is wrong. Sony already made a first mass production run and realized 100 sensors. Sony device manager Itonaga said it clearly: “We are ready“.

http://www.sonyalpharumors.com/addendum-sony-says-we-are-ready-to-produce-curved-sensors-on-mass-scale/

If I have understood the physics of this properly, this could really be a game changer.

 

[GaryJ]

Some folk on CR must take an angry pill before logging on, as we say in Aus 'we are not playing for sheep stations' and also manners don't hurt either,still it is fun to read

 

[Don Haines]

it's most likely for small camera systems to produce good low light results using cheap, simple, fast lenses. First use will probably be phone- cameras. Getting the required amount of curvature in larger sensor systems is likely to be much more challenging as slight (focus) errors of a few microns could occur with temperature shifts and if you're looking to make a compact, large sensor camera with this then you're really gonna have to decrease the radius of that sensor's curvature. So, more likely practical for 1/2.5 and smaller sensors.
Still, will be very interesting if they can effectively achieve this for larger sensor formats.

This is what I was thinking....

Also, a curved sensor would work better with wide angle lenses, sort of like why a flat sensor works best with long lenses.... and since phones and p/s cameras tend towards the wide end of the spectrum, this could make the lens designs simpler.

And yes, there will be distortion, but look at the distortion you already have in those lenses... take an ipad or iphone and move it around and see how the image redraws as you move it... my bet is that the curved sensor will result in less distortion than the current sensor/lens combinations have now.

 

[aj1575]

My retina is curved. My eyeballs do not contain multiple elements. IF flat sensors were "simpler" to design for, I am quite disappointed that we have not evolved that superior level yet!

+1

I guess it is time to begin R&D on printing silicon wafers

There is reason to be dissapointed, not because your retina isn't flat, because it cant zoom.
There is the BIG problem with a curved sensor, the curvature only fits one focal lenght. If you like to go the middle way then and choose one curvature that fits everything, then you end up with a flat sensor, because it still ist the easiest shape, and producable with smaller tolerances (fitting two curved shapes, the sensor and the focal plane would be rather troublesome).

 

[privatebydesign]

These sensors could come out sooner than we think. SAR is reporting this:

The image of this Sony curved sensor created quite a buzz but some people argued that this tech may not coming any time soon. Well this is wrong. Sony already made a first mass production run and realized 100 sensors. Sony device manager Itonaga said it clearly: “We are ready“.

http://www.sonyalpharumors.com/addendum-sony-says-we-are-ready-to-produce-curved-sensors-on-mass-scale/

If I have understood the physics of this properly, this could really be a game changer.

And again, it will render the few lenses Sony already released for their FF mirrorless cameras obsolete. Time for another lens/system road map..........

 

[Steve]

And again, it will render the few lenses Sony already released for their FF mirrorless cameras obsolete. Time for another lens/system road map..........

Yeah pretty sure its been pointed many times that these sensors are most likely going to be used for fixed lens cameras like the RX series

as we say in Aus 'we are not playing for sheep stations'

what does this even mean? I can't even begin to parse this weird proverb

 

[privatebydesign]

And again, it will render the few lenses Sony already released for their FF mirrorless cameras obsolete. Time for another lens/system road map..........

Yeah pretty sure its been pointed many times that these sensors are most likely going to be used for fixed lens cameras like the RX series

Which makes it even more of a niche. Are we really short on choice in the >$2,500 fixed lens ff market? And who cares? Are corners that much of a problem for everybody as to need new cameras and lens/lenses?

I suspect this will be a VHS vs Betamax type contest, sure curved sensors have the theoretical potential to offer "better" IQ, but at what cost? And seeing as how flat sensors/film and lenses have been around for over a hundred years and consumers seem more driven by size and convenience, or price, than IQ, I think Sony have yet anther cool product with little to no market.

 

[dgatwood]

Microlenses were invented and put on your DSLR sensors because they couldn't make curved sensors.

AFAIK, the purpose of microlenses is actually to work around the fact that the sensors aren't flat, but rather recessed down into pockets in the silicon. Light rays near the edge of the sensor would otherwise fall way off unless you use a huge image circle. Changing the angle of those pockets might provide an advantage over the microlenses, but doing so doesn't inherently require curving the actual face of the sensor.

Where do you read such garbage? The purpose of the microlens is to increase the light gathering surface of the pixel, so that the space on the sensor top where there is circuitry and other stuff between pixels can be used to gather light to send into the pixel. This problem becomes more acute towards the edge of the sensor because rays of light are striking the sensor at and angle.

That's certainly an added benefit, but what I said is quite correct; unless you're using a back-illuminated sensor, the actual light-sensitive part of the chip is inherently recessed below the wiring. This results in light fall-off near the edges if you don't have microlenses.

Don't believe me? Read Leica's description of their microlenses array:

"This optimized micro lens design, based on many years of precision optical engineering experience, captures and concentrates even the most oblique rays on the sensor and reliably prevents image brightness fall-off at the edges and corners of the image."

Their words, not mine.

Leica's words match mine. The sensors are flat. It is the pixels that are the holes in the surface of the sensor. Further, the light rays don't "fall off", they just don't enter.

That's what the term "fall off" means—a reduction in brightness at the edges compared with the middle. The brightness "falls off" (decreases) as you get closer to the edge of the sensor.

There is reason to be dissapointed, not because your retina isn't flat, because it cant zoom.
There is the BIG problem with a curved sensor, the curvature only fits one focal lenght. If you like to go the middle way then and choose one curvature that fits everything, then you end up with a flat sensor, because it still ist the easiest shape, and producable with smaller tolerances (fitting two curved shapes, the sensor and the focal plane would be rather troublesome).

I assume they would set the curvature based on a focal length right in the middle of the typical shooting range. That way, the most common length requires no focal plane correction, everything uses less correction on average, and some lenses correct the focal plane in the opposite direction from others.

 

[arcanej]

This link (supposedly) shows the first picture from the curved sensor:
http://www.mirrorlessrumors.com/world-s-first-image-taken-with-the-famous-new-sony-curved-sensor

 

[Max ☢]

What was Sony thinking when they submitted this image? it's blurry!! if this is supposed to show the advantage and performance of their new sensor, then it's pretty underwhelming... sure, there's no vignetting, but does this feature really called for a new sensor technology?

 

[arcanej]

A traditional planar sensor will only have part of a properly focused image out of focus. The advance of curved sensors lets a photographer have an entire properly focused image blurry.... Progress!

Or something.

 

[aj1575]

Do you like to buy Sony curved sensor; you can do it now, the KW1 will be sold shortly.

Take a look at this awesome piece of professional equipment...

http://www.gizmag.com/sony-kw1-perfume-selfie/33484/

 

[mb66energy]

My retina is curved. My eyeballs do not contain multiple elements. IF flat sensors were "simpler" to design for, I am quite disappointed that we have not evolved that superior level yet!

The eye has only very good center sharpness. But to the periphery sharpness/resolution declines and the colour sensitivity too; but the peripheral sensors are faster (some flourescent tubes flicker in the periphery but are stable in the center).
On the other hand you need a very wide field of view (roughly 180 degree horzontally) to be aware of potential threats. The "eye-brain" directs the sweet spot of our eye to the potential threat.

Our "eye-brain" scans the scenery and stores details - we think that our eyes are optically very good but in fact they are not compared to good primes e.g.

So photography has to store a lot more of detail than we might read in a glance.

Where our eye excells is the dynamic range which might be 20 stops which is
FROM full sunlight (1000 Watts per square meter)
TO micro LED torch illuminating a medium large room (0.001 Watts per square meter)
through (1) using an auto aperture, (2) high DR sensors and (3) relying on two types of sensors activated for its special purposes.

 

[mb66energy]

I really like that. But I guess we are lightyears away from such a device...Or Sony could take us by surprise by 2016?

http://www.northlight-images.co.uk/cameras/Canon_rumours.html

I think curved sensors are a "money-printing-machine" for companies if it comes for interchangeable lens cameras: It converts all other lenses to high tech waste.

A better approach should be to make the sensor surface matt black so they collect each photon idependent from the angle of incidence. If I knew how to do that I would do it but ... I have no idea yet.

 

[AcutancePhotography]

as we say in Aus 'we are not playing for sheep stations'

what does this even mean? I can't even begin to parse this weird proverb

Be careful what you ask for. Do you really want to know about Australians, sheep and the games they play?