14
\$\begingroup\$

I've read that lenses are designed to project an image onto a flat surface and that as image sensors grow larger, lenses will grow correspondingly larger and by some additional factor. Flat imaging was clearly important in the early days of photography where an image was projected onto the actual display medium as with Daguerreotypes.

Additionally, flat image projection certainly simplified film photography, where images could be conveniently stored on a rolled-up, long flat strip.

But why are we still projecting images onto flat surfaces if neither the storage medium, or the processing requires it?

Wouldn't lens design be vastly simplified if the image were projected onto a curved sensor? Shutters faster and more efficient if they were located at a fixed point? Is the problem with CCD design and sensor manufacturing? Are manufacturers pursuing this at all? Or are we destined to luggage-sized cameras.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ If you remove a constraint from engineers they should be able to make other things better. If not, it really wasn't a constraint. I would think this would happen in fixed lens cameras first because you don't need a full suite of lenses designed to the (same) curved plane and don't risk customers buying the wrong sort of lens and being disappointed. The question is whether it helps enough to create a path through the sensor industry that grinds the blank to a curve before making a sensor. \$\endgroup\$ Commented Feb 22, 2022 at 3:36
  • \$\begingroup\$ Are you asking about sensors curved in a single plane (as in cylindrical curvature) or in 2 planes (spherical curvature)? \$\endgroup\$
    – scottbb
    Commented Feb 24, 2022 at 0:58
  • \$\begingroup\$ Like the eye's retina....curved around the x and y axes, lightpath down the z axis. \$\endgroup\$ Commented Feb 24, 2022 at 4:44

4 Answers 4

12
\$\begingroup\$

Digital image sensors used in modern cameras are essentially silicon chips:

Sensor dies are produced in large batches on silicon wafers. The wafers are cut into many pieces with each piece housing a single sensor die. The larger the sensor die size, the lower number of sensors per wafer.

The wafers are sawn with diamond saws from silicon ingots, which renders them flat. I can only imagine that trying to saw or otherwise cut a surface curved in two dimensions would be very hard or impossible. Consequently, the "natural state" of a digital sensor is flat.

Image of two such wafers

a big wafer and a little wafer

(Image by Hebbe at German Wikipedia, via Wikimedia Commons.)

However, the raw wafers undergo considerable "post-processing" like polishing which conceivably could be used to grind one surface into a concave shape; but of course the succeeding processes currently are expecting a plane surface, so one must expect all kinds of costly adaptations and problems down the road.

That said, Canon has filed a patent for a curved sensor. I suppose that if the chip is thin enough it is flexible enough for the purpose.

\$\endgroup\$
17
  • 1
    \$\begingroup\$ You can easily bend Si wafers if you want to. By a little bit when they are 100 um+ and by quite a lot when it is down to about 40 um thickness. Unfortunately, it is very fragile plus there are issues fabricating high quality thin wafers in anything resembling cost effective way. So, what you save by lens in weight and cost would be lost by extra vibration dampening in the body and costly lens repair would be replaced by even more costly sensor replacement. \$\endgroup\$ Commented Feb 22, 2022 at 9:34
  • 1
    \$\begingroup\$ @Zizy What you see sometimes is cylindrical bending (e.g. researchgate.net/figure/…) Spherical/hyperbolic shapes will have a much larger bend radius. \$\endgroup\$ Commented Feb 22, 2022 at 10:09
  • 4
    \$\begingroup\$ Silicon chips being flat is also an artifact of the flat focal plane used in the photolitography processes. \$\endgroup\$
    – fraxinus
    Commented Feb 22, 2022 at 14:28
  • 2
    \$\begingroup\$ @Vikki it wouldn't be a very good crystal with a curve and without lattice matching. Far easier would be to cut a thick wafer and polish it as we do lenses. In fact this is done for infrared optics where silicon is transparent. As Peter says, deposition onto the curved surface would be problematic, but again, the optics industry does that for things like anti-reflection coatings. All this would make for very expensive sensors. \$\endgroup\$
    – Chris H
    Commented Feb 23, 2022 at 9:32
  • 2
    \$\begingroup\$ @ChrisH That's pretty cool. But they used a flat sensor and bent it afterwards. I didn't mean that Si can't be bent, just that it doesn't want to grow bent. Now, organic semiconductors might do the trick ;) But I doubt they'll replace Si-CMOS anytime soon... \$\endgroup\$
    – Max
    Commented Feb 23, 2022 at 10:13
10
\$\begingroup\$

Curved film planes were very common in low-end roll film cameras and even 35 mm simple cameras.

This is because they improve the image quality of a lens with a spherical focal surface, as is the case with the simplest lenses used in consumer cameras back in the day -- the meniscus. Meniscus lenses were common as far back as wet plate days (1850s on), and were popular because they were cheap. Kodak sold the Vest Pocket Kodak with an achromatic meniscus, but did what was necessary to use a flat film plane: installed the lens concave to the world, behind the aperture and shutter.

Later, for marketing reasons, meniscus lenses were often installed the other way around, convex to the world, which allows placing the lens in front of shutter and aperture and avoids "the lens has fallen off!" conversations at the camera store. Unfortunately, that way 'round a meniscus has a curved focal surface; this was compensated, especially in rectangular frame formats, by constraining the film surface to a cylinder which minimized the error, mostly confining defocus visible on the contact prints of the day to the extreme corners of the frame.

This "optimization" for lens cost is very unlikely to appear in digital cameras, however, because it would be offset by the costs associated with fabricating a non-planar sensor surface. Designing lenses with flat focal plane is old news; the Cooke triplet, Periskop, Rapid Rectilinear, and Tessar all have flat focal plane, and predate the introduction of roll film as we know it in 1901.

If you want to look it over, however, try to find one of the "plastic fantastic" 35 mm cameras that have been sold over the past fifty or so years. Ilford and Lomography sell them now as "reusable" (as opposed to disposable) cameras, at prices around US$30 to US$50, and cheaper examples can be had direct from China. All of these have simple meniscus lenses and curved film planes.

\$\endgroup\$
7
\$\begingroup\$

It's not like there isn't any research being done in this direction:

Origami silicon optoelectronics for hemispherical electronic eye systems

...reshaping conventional planar sensor systems into hemispherical formats would empower visual recordings with features that are beyond what state-of-the-art cameras can see, such as infinite depth of field, wider view angle, and lower aberrations

These guys have taken rudimentary silicon imaging technology and folded them into curved shapes, mimicking insect eyes:

enter image description here

Admittedly, their results are fairly primitive but the paper is already over 4 years old. Modern CCDs also had humble beginnings!

\$\endgroup\$
1
  • \$\begingroup\$ Those incorporate the lens and sensor together -- each sensor fragment has its own lens, like an insect's compound eye. Not at all the same. \$\endgroup\$
    – Zeiss Ikon
    Commented Feb 22, 2022 at 18:30
1
\$\begingroup\$

Or are we destined to luggage-sized cameras. If you consider what can be done with mobile phone cameras, we're really not.

Using that sort of technology and a small sensor we can already build a fixed focal length, fixed aperture, variable focus camera only about 1cm thick. Applying this to a dedicated camera, with other advances from the last couple of decades, we should be able to get the image quality of an SLR from say 20 years ago in a small compact.

But then you hit an ergonomic limit. Cameras have to be held steady, and the controls have to be operable with human hands. Compacts without viewfinders, just like phones, have to be held tens of cm in front of your face, which makes them prone to shake. Viewfinders take up space, and good ones as on SLRs take up a fair bit.

So there's not much demand to make cameras even smaller than we already can. If you want small, just use the phone you're probably already carrying. If you want good quality, you can do pretty well with something that will fit in a pocket. Bulky cameras get you versatility, including better performance over more metrics in a wider range of situations.

Curved sensors are of interest in astronomy. In space-based telescopes they could improve the quality in a given payload.

\$\endgroup\$
3
  • \$\begingroup\$ Understood. But keeping things from getting bigger is different from making things smaller. Demand for higher resolution and resulting larger sensors is forcing lenses to grow huge. Let's not even look at fast lenses for full frame cameras....OMG! Not to mention how impractically heavy some of these zoom lenses have become. As I understand it, that's driven by the physics of planar image projection. \$\endgroup\$ Commented Feb 23, 2022 at 16:36
  • \$\begingroup\$ Most of it isn't driven by the need to project an image onto a plane. A little bit of bulk and a little bit of weight are, but that's all. Sensor sizes haven't changed in decades (full frame matches 35mm film) and higher resolution comes from smaller pixels instead. There is some demand for faster lenses, and those are bigger \$\endgroup\$
    – Chris H
    Commented Feb 23, 2022 at 16:55
  • \$\begingroup\$ Admittedly, that's a good point. I also get that modern lenses also contain image stabilization and autofocus technology. But modern CCD resolution is far superior to film; might that require far better (and bigger?) lenses. The fact that unlike film, CCDs require a very specific light path also forces a specific lens design. \$\endgroup\$ Commented Feb 23, 2022 at 17:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.