h a l f b a k e r yYeah, I wish it made more sense too.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Modern camera lenses require several glass elements in order to properly project an image onto a planar surface. This makes for a heavy and expensive lens, particularly when a large aperture is desired.
The complexity of lenses could be significantly reduced by projecting onto a curved surface instead.
A CCD manufactured as a sector of a sphere allows a lens to create a sharp image without having to account for spherical aberration.
Obviously this presents some manufacturing challenges, the most difficult of which is forming the silicon. Once a wafer is obtained however, the rest of the process isn't too difficult. Photolithography also benefits from the design, as manufacturing equipment needn't account for spherical aberrations either.
Circular CCD chip for digital cameras
See [cowtamer]'s Oct 31 2006 annotation, and subsequent annotations [hippo, Feb 12 2013]
Spherical CCD sensor by Andanta
http://www.andanta....w_products_2012.pdf
Note link is a PDF.
16 mega pixels. [xaviergisz, Feb 12 2013]
Excellent site about optics
http://toothwalker.org/optics.html
[mitxela, Feb 12 2013]
Sony breaks new ground with camera sensors curved like the human eye
http://www.phoneare...e-human-eye_id57124
[xaviergisz, Jun 14 2014]
Single Crystal Turbine Blade
http://www.rothbiz....rbine-blade-as.html
[bs0u0155, Jun 17 2014]
Sony have made this now
http://www.dpreview...s-and-better-images
[hippo, Feb 25 2015]
...and Apple have patented it
http://connect.dpre...small-camera-module
[hippo, Feb 01 2016]
"The challenge of highly curved monolithic imaging detectors"
https://www.eso.org...PIE2010/7742-27.pdf
2010 article, cited by the Apple patent. Covers the basic idea and a manufacturing technique. [notexactly, Feb 01 2016]
"The challenge of highly curved monolithic imaging detectors"
https://www.eso.org...ing%20detectors.pdf
Redundant link [notexactly, Feb 01 2016]
[link]
|
| |
I like it. There aren't many new ideas in image capturing other than "add more pixels, and make the pixels better" so [+] |
|
| |
//A CCD manufactured as a sector of a circle// |
|
| |
Indeed I do. Here I researched spherical geometry to ensure sector was the proper term, and I carelessly ended up eliminating an entire dimension. |
|
| |
Good idea - this came up before in an annotation (see link) |
|
| |
Agreed, a good idea. It may be technically difficult,
though - the wafers on which the CCDs are built are
sliced from a cylindrical ingot and then polished to
present a single crystal plane on the surface. Not
sure how you'd do this as a spherical surface. |
|
| |
Either carve out a curved surface from a wafer several milimetres thick, or get a flat wafer and use molecular beam epitaxy to build up the surface? |
|
| |
If you carve it, the silicon crystal plane will be all
over the place. |
|
| |
Adding a 4th dimension projection would be one
obvious solution. |
|
| |
For watching it means that the viewer has to maintain a fixed distance from the spherical-section screen. |
|
| |
Do the optics maintain their relative simplicity if you want to +/- zoom ? |
|
| |
Would the surface be concave or convex? For some
reason I originally thought convex, but I'm guessing
it
would actually be concave, matching the function
of
the eye. |
|
| |
Also, given a high enough resolution, could this not
be done in software? In essence, by “offsetting”
each pixel based on the difference between where
it is in space and where it would be if the sensor
were curved. |
|
| |
You could do that, but the granularity of the photo
will be slightly higher at the edges than the middle
simply due to angle of incidence. |
|
| |
I had this idea a few years ago -- simply from the fact that the retina doesn't suffer from blue/red shifts at the edges -- and got the blunt reply of "too expensive to manufacture". |
|
| |
Even if you could manufacture it, the curvature would have to be matched to a fixed focal length. This isn't an enormous problem; cameras like the Fuji x100 have shown there's still a demand for fixed lens cameras. |
|
| |
Another problem though is that it doesn't actually solve, for instance, chromatic aberration. The edge-of-frame effects are gone (transverse chrom abs) but the loss of sharpness and blooming would remain (longitudinal chrom abs). There's an excellent website I found <link> which explains lens aberrations really well. Thoroughly worth reading the whole site if you have an interest in optics. |
|
| |
//could this not be done in software?// That is a
very good point. |
|
| |
//the granularity of the photo will be slightly
higher at the edges// That is also true. However,
one could have different pixel sizes at the
edges...no, hang on, you can't tesselate squares
of sizes that increase smoothly in both directions. |
|
| |
Maybe the software solution is simplest, with
resampling to give a constant granularity. |
|
| |
On the other hand, is it really so difficult to make
lenses that correct for the planarity? The lens
already has multiple elements to provide focus
and achromaticity - how many extra lens elements
are actually needed to correct for planarity? |
|
| |
//could this not be done in software?// No. If the edges of the frame are out of focus for whatever reason, no amount of software will recover that. |
|
| |
However, the latest generation of digital SLRs match the profile of the lens and do correct what they can (distortion/edge of frame colour shifts) in-body. |
|
| |
//If the edges of the frame are out of focus for
whatever reason, no amount of software will recover
that.// |
|
| |
But would the edges be out of focus or just
distorted? |
|
| |
"Distortion" with regards to photographic lenses means straight lines being represented as curved on the image. |
|
| |
Chromatic and spherical aberration are a loss of sharpness because some of the light is out of focus. Curving the image plane does not stop this, it simply makes the effect uniform across the frame. Multi-element lens designs actually reduce these effects. The one thing a curved image plane would prevent would be astigmatism. |
|
| |
But I really do recommend the toothwalker site for better explanations. |
|
| |
One thing I should also mention - there are cameras that actually use single element lenses these days: disposable film cameras. Many of these do actually curve the film plane to try and improve the image, of course practical constraints mean it can only be curved in one plane. |
|
| |
Assuming space-time is curved, as allowed by GR theory, just make the camera really really big. |
|
| |
Or hand a tiny piece of black hole behind the lens? (might require a sturdy tripod to stand the camera on) |
|
| |
It's seems somewhat ironic complaining about "aberration" on here.. |
|
| |
//If you carve it, the silicon crystal plane will be all over the place.// |
|
| |
You only have to approximate a sphere, though, so within any individual pixel the crystal plane can be flawless. Imagine starting with a perfectly flat wafer and then somehow building up each pixel by the correct number of crystal lattice units, so that the pixels are individually flat but the overall shape of the surface is spherical. |
|
| |
Making the interconnects work could be tricky. |
|
| |
Actually, I've recently had cause to deal with silicon
wafers which are thin enough to be very flexible, so
perhaps this isn't as unfeasible as I'd first thought. |
|
| |
Flexible probably doesn't have the precision
required, and it almost definitely can't do a bowl
shape with the required accuracy. |
|
| |
In order to get the precision for an optical
application, you're almost definitely either going to
have to somehow force the crystal to grow in shape,
or be able to grind it to shape. |
|
| |
//force the crystal to grow in shape//
That I like. Is there a "catalytic" surface that silicon will condense on, that can also be removed/dissolved? Could you entice the entire CCD sensor structure to "grow" (self-assembling)? Maybe use custom-engineered enzymes? |
|
| |
////force the crystal to grow in shape//// |
|
| |
Ain't gonna work. You need a single crystal plane. |
|
| |
//a single crystal plane// |
|
| |
well, check the <link>. They're growing single crystal
turbine blades, so it's not a massive intellectual leap
to suppose they can make the whole 'plane in one
crystal. I think the furry dice hanging from the rear
view mirror might prove technically challenging,
however. |
|
| |
More seriously, collimated beams are a bit of a pain
to make, but after that, optics gets very easy. It's
tough to get everything lined up in x & y, adding in
another dimension with tolerances down in the
individual nanometres is not to be encouraged. |
|
| |
//Flexible probably doesn't have the precision
required, and it almost definitely can't do a bowl
shape with the required accuracy.// |
|
| |
You would mount the flexible wafer on a solid curved
support. You could at least get curvature in one
dimension. |
|
| |
Reading the claims of the Apple patent, it seems that it
doesn't claim the actual idea here (which is covered in the
2010 article it cites: [link]), but only a camera of this type
with various specific optical parameters. (IDK why they'd
patent that…) |
|
| |