Could curved sensors be made for digital cameras?

[Pherdinand]

Sensor dies are made the same way other semiconductors are made. 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 sensor. Fabs the world over have the same Specs for production of wafers. So, a curved sensor is not likely. Film can be curved..

It's possible that all this will change.. but not quickly. Bigger, maybe, but not curved. We can thank Dr. Lam (Lam Research) for this advancement.
http://moeadham.ca/intels-upcoming-450mm-capable-wafer-fab/

http://www.lamresearch.com/

at the moment the usual max die size in semiconductor chipmakery is 26x33mm (the 33 can be stretched a bit...in principle up to 42mm)

of course special cases do exist🙂

anyway. Curved sensor is not possible - on silicon.
But we're not restricted to silicon wafers anymore...

 

[Range-rover]

Thanks guy's

 

[peterm1]

I had myself wondered if using the human eye as an analogue (ie having a curved retina / sensor) might not give a whole new scope for optical innovation. Sounds weird when you think and compare it with a silicon sensor. Sounds perfectly normal and natural when you think about the biological eye. Of course there would be technical challenges. There are with everything. But in principle it seems feasible as we know that this is what is used throughout the biological world.

 

[Jamie Pillers]

Not a stupid idea!

Not a stupid idea!

Not a stupid idea, at all. The technological development of new materials is racing along at incredible speeds these days. There are a number of technologies being studied that will likely lead to shaped sensors: quantum dot arrays, spray-on black carbon polymer, and nano tubes. A lot of this technology is getting pushed along because of the potential market in high quality camera phones. Putting a tiny lens very close to a flat sensor creates lots of design problems. But if the sensor could be curved, then design of the lenses becomes simpler/cheaper.

I'm always surprised a bit reading negative responses to "what if" future-thinking threads here on RFF. Its seems silly anymore to say something can't be done, when we're living in such a stew of fantastic technological innovation right now! And the speed of the innovative leaps is nearly unbelievable! When I was a kid, our phone was a wooden box on the wall, with a crank handle on the side. Its only been 50+ years since those days and now my daughter has video chats with her friends on a gizmo only imagined in Dick Tracy comic strips! Making shaped sensors seems like kids play compared to the changes I've already seen. 🙂

 

[sevo]

I'm always surprised a bit reading negative responses to "what if" future-thinking threads here on RFF. Its seems silly anymore to say something can't be done,

It more of a "what for". And as for that, the ability to give sensors an arbitrary curve other than "flat" isn't very useful, unless it is employed to simplify the design of one particular lens. That is, it is of no advantage for interchangeable lens (or even zoom lens) cameras. It could make its way into low budget devices - but for that to be cost efficient, curving the sensor would have to be cheaper than a aspheric element fulfilling the same function. But that is still far away, as three-dimensional structures in current sensor substrates are seriously time-consuming and hence expensive in production, while pressed acrylic aspheric lenses in phone lens size are a few cents the dozen.

Now, curved EVF displays are even more desirable (as they'd have a constant distance to the eye point), might sell for enough to warrant somewhat expensive fabrication methods, and a curved OLED is more feasible than a curved large size silicon sensor. Early applications of curved optics IC design would probably be there.

 

[sevo]

I had myself wondered if using the human eye as an analogue (ie having a curved retina / sensor) might not give a whole new scope for optical innovation.

You mean as in "horribly poor camera and lens with a world class processor to reconstruct useful images from that blurry incomplete mess"?

Even if you wanted to grow a eye-like camera, you'd model it on the much better designed octopus eye. The vertebrate eye is a classic example that evolution does not pick the best solution, but the first available work-around. Due to what our eye developed from, the retina is attached bottom up and accordingly has all the cabling (nerves and blood vessels) on top of the active elements, plus a gaping hole in the middle, where the cabling meets. And that is just one of the many flaws of our eyes.

 

[BobYIL]

I had myself wondered if using the human eye as an analogue (ie having a curved retina / sensor) might not give a whole new scope for optical innovation. Sounds weird when you think and compare it with a silicon sensor. Sounds perfectly normal and natural when you think about the biological eye. Of course there would be technical challeto recnges. There are with everything. But in principle it seems feasible as we know that this is what is used throughout the biological world.

Human eye has a field of view wider than 90 degrees which is wider than what we can see with the 21mm Super Angulon. Curved focusing plane may be helpful in reducing the optical deficiencies of very wide angle lenses however it poses more issues for the optical designers to calculate all focal lengths not based on a plane but on the same radius of curved surface; quite challenging. Imagine designing 50, 85, 135 and 400mm lenses all with the same curvature of image field of a 21mm lens; awful!

Assume also that they were able to do, then how to produce that curved sensor? It's not a lens to be ground or molded as curved.. Until the required diffusion and etching processes are accomplished the wafer undergoes a multitude of photographic/masking steps, even with plane masking such minute dimensions down to fractions of 1um creating huge issues (yes, with plane masking great issues!..) The sensor size of your cellular is in 1um range, meaning everything is squeezed in that cell into 1um.. Make 1um error on the surface curvature then throw away the whole wafer to trash.

 

[peterm1]

Human eye has a field of view wider than 90 degrees which is wider than what we can see with the 21mm Super Angulon. Curved focusing plane may be helpful in reducing the optical deficiencies of very wide angle lenses however it poses more issues for the optical designers to calculate all focal lengths not based on a plane but on the same radius of curved surface; quite challenging. Imagine designing 50, 85, 135 and 400mm lenses all with the same curvature of image field of a 21mm lens; awful!

Assume also that they were able to do, then how to produce that curved sensor? It's not a lens to be ground or molded as curved.. Until the required diffusion and etching processes are accomplished the wafer undergoes a multitude of photographic/masking steps, even with plane masking such minute dimensions down to fractions of 1um creating huge issues (yes, with plane masking great issues!..) The sensor size of your cellular is in 1um range, meaning everything is squeezed in that cell into 1um.. Make 1um error on the surface curvature then throw away the whole wafer to trash.

My point was more that there is clearly nothing in the laws of optics / physics that would seem to preclude this. But I tend to agree that the technical challenges would be huge and in particular what I neglected to say was that I could not see how you would build a tele lens - let alone a zoom.

What I was thinking though, was could this be a way of building a large sensor miniature camera in a manner that would allow smaller lenses (a major constraint with current technology). Assuming of course that you can overcome the problem you have rightly pointed to - how to build a sensor on a spherical surface.

 

[Sparrow]

My point was more that there is clearly nothing in the laws of optics / physics that would seem to preclude this. But I tend to agree that the technical challenges would be huge and in particular what I neglected to say was that I could not see how you would build a tele lens - let alone a zoom.

What I was thinking though, was could this be a way of building a large sensor miniature camera in a manner that would allow smaller lenses (a major constraint with current technology). Assuming of course that you can overcome the problem you have rightly pointed to - how to build a sensor on a spherical surface.

... well apart from the mathematics of the Planar Projection, which is what we actually trying to get to in the print.

 

[rodt16s]

at the moment the usual max die size in semiconductor chipmakery is 26x33mm (the 33 can be stretched a bit...in principle up to 42mm)

of course special cases do exist🙂

I seem to remember e2v show sensors around 61.4 mm x 61.4 mm, I think they are monolithic but I could be wrong

 

[rxmd]

at the moment the usual max die size in semiconductor chipmakery is 26x33mm (the 33 can be stretched a bit...in principle up to 42mm)

of course special cases do exist🙂

anyway. Curved sensor is not possible - on silicon.
But we're not restricted to silicon wafers anymore...

I guess you mean that's the area that can be exposed in a single exposure - full frame and medium format sensors, for example, get exposed in several consecutive exposures (which is partly what makes them comparatively expensive). The maximum wafer size in common use is currently 300mm/12" AFAIK, but I'm not an expert.

 

[BobYIL]

My point was more that there is clearly nothing in the laws of optics / physics that would seem to preclude this. But I tend to agree that the technical challenges would be huge and in particular what I neglected to say was that I could not see how you would build a tele lens - let alone a zoom.

What I was thinking though, was could this be a way of building a large sensor miniature camera in a manner that would allow smaller lenses (a major constraint with current technology). Assuming of course that you can overcome the problem you have rightly pointed to - how to build a sensor on a spherical surface.

Peter, so far nothing could be "invented" any further than the offset microlenses; that also have their own shortcomings. The best remedy is the retrofocus lens design, minimum optical problems but size issue. However Fujifilm engineers -they have extensive experience from miniature lenses up to LF or wide-range zooms for studio TV cameras- came up with revolutionary lens designs (extralarge rear elements) with their XP1 lenses and succeeded with no offset microlenses whatsoever. On their sensor even the 35/1.4 Summilux asph. show color fringing near edges and corners but the 35/1.4 Fuji does not. Remarkable.. (The 35/2 Sonnar too could be of this type, I suspect..)

If we are to see FF-mirrorless compact bodies then it will be possible after their engineers accomplished some unconventional designs for the FF lenses to be used on them.. (Don't mind much about the sensor..)

 

[sevo]

I seem to remember e2v show sensors around 61.4 mm x 61.4 mm, I think they are monolithic but I could be wrong

I've seen spec sheets listing monolithic sensors up to 125mm square, with 100mm being a established size for aerial mapping and recon cameras. And that is non-classified stuff in regular production, so there will be bigger ones in labs and classified spyware.

I suppose 26x33mm (probably some maker's near FF sensor?) is the biggest unstitched size, and past that, you enter the territory of stitched monolithic chips.

Multi-chip designs are uncommon and probably were only done while feasible chips were tiny - all references I can find date back to the seventies or eighties. Some (star mapping) telescope cameras use chip arrays - but these are one-of-a-kind constructions, it is cheaper to build a custom camera if you need to digitize a telescope originally built for photographic plates than to replace the telescope, and the sky being essentially static for astrophotographic purposes the superimposed lattice grid can be eliminated on a second scan pass over the same area. Besides, astronomers have unusual low noise requirements which may have them opt for a particular sensor regardless of whether they need several to fill the image area.

 

[rxmd]

I suppose 26x33mm (probably some maker's near FF sensor?) is the biggest unstitched size, and past that, you enter the territory of stitched monolithic chips.

I would surmise that full frame sensors will become a lot cheaper when photolithography advances the area of a single exposure beyond those 900-ish mm².

 

[Jamie Pillers]

Or maybe the curve changes with lens mounted. 🙂

 

[FrozenInTime]

Looks like Sony have a Dali inspired curved sensor !

http://egami.blog.so-net.ne.jp/2012-09-26

 

[citizen99]

... 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.

Here's another:
http://www.flickr.com/photos/43334883@N03/7893681388/in/set-72157631319992170
[actually with a (non-Super) Angulon 😀 ]

 

[Alfasud]

Curved Film Plane

Curved Film Plane

Minox used a curved pressure plate in they sub-mini cameras. As these were, of course, fixed lens. I assume that the curve was dictated by lens design.

 

[Spanik]

Of course if the sensor area was malleable, you could use air or fluid to shape the device (asphere) to capture the specific lens characteristic.

Like this? http://www.dpreview.com/news/2004/3/3/philipsfluidlens

 

[Pherdinand]

I guess you mean that's the area that can be exposed in a single exposure - full frame and medium format sensors, for example, get exposed in several consecutive exposures (which is partly what makes them comparatively expensive). The maximum wafer size in common use is currently 300mm/12" AFAIK, but I'm not an expert.

yep. die is the image that is created in one exposure. larger ones are stitched (which is def'ly not a regular process on a semiconductor chipmakers' litho machines. Not on the ones we are making at asml, at least 😉 )
300mm is the wafer size used for highest efficiency/speed in lithography, indeed. 450 is on the horizon...