Could curved sensors be made for digital cameras?

[hteasley]

My only complaint about the M9 is the chromatic aberration using wide lenses like the Super Angulon. It strikes me that corner sharpness and color shifts would be less of a problem if the sensor plane was not flat, but curved, either cylindrically or spherically, for less of a distance and angle difference between the center and the corners of the sensor and the lens.

I figure it would be impossible to correct 100% accurately for every lens, but that there should be some sort of sweet spot where some level of curve could help all lenses. Is that crazy?

And is a curved sensor feasible? Are there any film cameras that curved the film behind the lens to achieve better corner and edge results?

Just idle curiosity on my part: I have no idea if this is a stupid idea or not.

 

[Sparrow]

I think lenses are corrected to have a flat projection as it is ... a curved sensor, I fear, would bugger things up a bit.

 

[Spanik]

This is the reason that a FF sensor is made in two pieces. The Mfg. standard won't allow for a die the size of a FF

I don't know where you get this. FF sensors and even those medium format 53.8x40.3 sensors are made as a single piece of silicon. There is no other way. Large sensors for astronomical applications are sometimes made as assemblies, but even those are multi chip setups.

It isn't because they have to split the readout to different amplifiers and AD's that the silicon is made in 2 parts.

Why do you think large sensors are so expensive? It is just because they are a single piece. And you cannot get many of them on a single wafer. So any defect means a scrapped die and less yield. The 450mm wafers have nothing to see with optical production.

Take a look at people that do make serious optical sensors: http://www.teledynedalsa.com/semi/technology/ccd.aspx Single die CCD sensors up to 100x100 mm²

 

[Sparrow]

Oh ... and folk go to extreme lengths to keep film flat for good reason too I suspect

 

[Sparrow]

I don't know where you get this. FF sensors and even those medium format 53.8x40.3 sensors are made as a single piece of silicon. There is no other way. Large sensors for astronomical applications are sometimes made as assemblies, but even those are multi chip setups.

It isn't because they have to split the readout to different amplifiers and AD's that the silicon is made in 2 parts.

Why do you think large sensors are so expensive? It is just because they are a single piece. And you cannot get many of them on a single wafer. So any defect means a scrapped die and less yield. The 450mm wafers have nothing to see with optical production.

Take a look at people that do make serious optical sensors: http://www.teledynedalsa.com/semi/technology/ccd.aspx Single die CCD sensors up to 100x100 mm²

Wouldn't 100x100 mm² be a cube?

 

[sevo]

The dies are flat, and there is little to be done about that, at any rate for the sensor sizes we are accustomed to. But microlenses optically do quite the same (in a even better way, being curved relative to the angle of incidence without affecting the focal distance) - so yes, we already have that. However you can only standardize new lenses to match all the same single curve (in which case we might make it a infinite radius - i.e. flat - and do away with the microlenses), and will need a different microlens (respectively curve) radius for each legacy lens. So this is more of a stop-gap measure for camera systems that want to improve their backwards compatibility, and remains a compromise that cannot eliminate the need for electronic correction.

 

[BobYIL]

Curved? No.. The semiconductor disc (wafer) is initially "grown" as solid bars of mono-crystalline structure. Then it's cut into slices, delivered as slices. The plant to use them first lap them to four-triangle class flatness & surface finish; meaning they should be as flat as possible before any photographic process or masking starts. (I once worked at TI for chip production.)

 

[sevo]

Chip works once had a gallery of canon, nikon, kodak sensors showing the seams. The dies are stiched with firmware. I think nikon's newest fab is still stitching FF sensors.

Nope, that is a different use for that terminology - they are stitched on the stepper, that is, the area of one sensor is made up of a mosaic of several exposures on the die (chips are made in a photolithographic process), as it exceeds the area that can be exposed in one go. The sensor chip itself is a single piece of silicone.

 

[sevo]

http://www.thefreelibrary.com/Tower...ching+Technology+for+Ultra-High...-a053609172

I told you so...

 

[Range-rover]

Remember Contax made a camera that had a vacuum on the back of the
pressure plate, the idea was to keep the film real flat for those
Zeiss lenses they used.

Range

 

[t6un]

And is a curved sensor feasible? Are there any film cameras that curved the film behind the lens to achieve better corner and edge results?

FYI: Schmidt cameras require a spherical sensor, and it has been a major PITA for astronomers since 1930's. http://en.wikipedia.org/wiki/Schmidt_camera

 

[literiter]

I've wondered if their may eventually be any advantage for each lens to have it's own sensor, fitted to the back of each lens and tailored specifically to fit each perfectly.

It may solve the dirty sensor problem that happens from time to time as well.

 

[rodt16s]

Interesting, I've held 6" Si wafers that have been background to 50um (2mils), quite flexible you can wobble it like paper, it has quite a metallic sound. (Std A4 paper is around 100um (4mils)). Whether grinding a wafer to this thickness for optical is practical.. Hmm not my field.
I would think the package assembly to give a curved sensor is going to quite a challenge. I'm not from optical packaging, so don't have much insight of the minute requirement for the die alignment Z tolerances etc..
Curved Si might have some fun with thermal expansion
But die placement and making the interconnects to the die would require an interesting set of equipment.. as most currently work in a flat and/or perpendicular plane... something to doodle later

 

[kdemas]

I've wondered if their may eventually be any advantage for each lens to have it's own sensor, fitted to the back of each lens and tailored specifically to fit each perfectly.

It may solve the dirty sensor problem that happens from time to time as well.

That would be called a "Ricoh GXR".

 

[Spanik]

The Nikon D3 has two independent sensors that were joined - on a single silicon substrate

That's another thing. Has indeed nothing to do with yield or die size but with timing issues. Once the pixel matrix gets too large it becomes slow to read out as a single unit. So they split it in 2 halves they can clock and read independently. The issue is not the seam as the pixel matrix is regular but the problem is that you have to use more ad convertors and have to correct for the differences between them.

I've wondered if their may eventually be any advantage for each lens to have it's own sensor, fitted to the back of each lens and tailored specifically to fit each perfectly.

You mean like a Ricoh GXR?

Are there any film cameras that curved the film behind the lens to achieve better corner and edge results?

Agfa Clack. Difference is they did this so they could use a very simple lens and still get acceptable results. Some panorama camera's with a rotating front lens also curve the film plane.

 

[t6un]

I've wondered if their may eventually be any advantage for each lens to have it's own sensor, fitted to the back of each lens and tailored specifically to fit each perfectly.

It may solve the dirty sensor problem that happens from time to time as well.

This, plus a curved sensor, plus custom software processing = our eyes.

Flat sensors are only special because they are universal: there is only one flat shape versus all the different spherical ones.

 

[NeeZee]

Wait! Does that mean I will need a curved monitor to watch my shots?

The swing lens panoramic cameras (widelux, noblex etc.) expose on curved film, by the way - for different reasons of course...

 

[Range-rover]

Wait didn't Leica angle the lenses of their sensor's to except more light for
better lens corner performance??

Range

 

[willie_901]

The X100 has micro lenses optimized to match the angles of light that exits the Len. The lenses for the corners are different HN those used in the center. But this is custom tuned for its fixed focl length lens.

Leica does something with the sensor and/or micro lenses to help with the corners, but I have no idea what it is.

 

[BobYIL]

The X100 has micro lenses optimized to match the angles of light that exits the Len. The lenses for the corners are different HN those used in the center. But this is custom tuned for its fixed focl length lens.

Leica does something with the sensor and/or micro lenses to help with the corners, but I have no idea what it is.

-Offset microlenses are employed to cope up with oblique rays coming out of the rear elements close to the sensor surface.

- Offset lenses can not be 100% solution for wide & extreme wide angle lenses unless supported by appropriate "correction" software features too.

- In spite of offset lenses + correction software some issues like corner "smearing" or color shift toward edges can not be "handled" 100%, like the case of 21/4.5 Biogon on dgital M's.

- Offset microlenses may affect the rays from longer focal lengths (like teles) adversely.

The best solution for sensor sizes of APS-C and larger is to design the lenses either as retrofocal or according to the sensor size they will be used for; like the highly modified formulas used on X-P1 lenses (or the new 35/2 Sonnar on the XR1 which I suspect of having a peculiar formula) combined with specific correction programs.