h a l f b a k e r yPoof of concept
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,
|
|
|
Lenticular lens displays are well known to create moving
or 3D visual effects when the device is viewed or tilted in
one axis of rotation.
Microlens arrays extend the principle of lenticular lenses.
The arrays of approximately hemispherical lenses provide
moving or 3D visual effects when the
device is viewed or
tilted in 2 axes of rotation.
Both lenticular and microlens arrays work at a limited
range of angles of tilt. This is because the focal length of
a lens is approximately the same at all angles (i.e. follows
a circular arc) but the image substrate is flat. Also lens
have more optical aberrations the further away from
perpendicar.
So my idea is to make the microlenses as Luneburg
lenses. Luneburg lenses work equally well at all angles,
thus overcoming the aforementioned problems.
Luneburg lenses are spheres which are made by layering
progressively lower refractive index layers. The Luneburg
lenses would be arranged in a regular array over the image
substrate which is made of a deformable material. Each
Luneburg lens placed directly above a corresponding
microimage. The Lunberg lenses would then be pressed
into the substrate (such that they are half submerged)
which is then solidified (e.g. UV curable
polymer).
The result would be more effective displays that would be
essentially colour holograms.
This could be extended to colour holographic video
displays if a deformable high resolution electronic display
could made (e.g. OLED).
Luneburg lens
https://en.m.wikipe.../wiki/Luneburg_lens
[xaviergisz, Jan 02 2020]
Fibre-optic image conduit
https://gfycat.com/...emandingHarrierhawk
[xaviergisz, Jan 04 2020]
Storage requirement for smaller frames of full-motion video
https://www.cl.cam....m/book/node111.html
[Voice, Jan 07 2020]
2.5µm pixel width
https://hardware.sl...lead-to-flawless-vr
Samsung and Stanford University have developed OLED technology that supports resolutions up to 10,000 pixels per inch. Perfect for this idea.
[xaviergisz, Oct 27 2020]
[link]
|
| |
I like the concept. Glad to see a use for even
higher pixel density displays. This seems feasible with
existing technology for large displays e.g. a billboard.
For smaller screens, this design seems to require very
small
pixels...each sphere appears as one pixel to the viewer,
but
each sphere must be illuminated by a grid of pixels with
one
pixel per ~degree of solid angle. For a color display,
presumably there would need to be 3 sets of spheres,
each
sized to avoid chromatic aberration at one specific
wavelength. |
|
| |
//For a color display, presumably there would need to be
3 sets of spheres, each sized to avoid chromatic
aberration at one specific wavelength.// |
|
| |
My guess would be that chromatic aberration would not
be an issue or could easily be avoided with proper
design. |
|
| |
//This seems feasible with existing technology for large
displays e.g. a billboard.
For smaller screens, this design seems to require very
small pixels...each sphere appears as one pixel to the
viewer, but each sphere must be illuminated by a grid of
pixels with one pixel per ~degree of solid angle.// |
|
| |
Yep, easier for big displays, but plausible for TV sized
displays also. Each Luneburg lens could be 500µm with
a 100x100 array of 5µm pixels behind it. Also could be
optimised for human binocular vision, with less vertical
pixels than horizontal pixels for each Luneburg lens. |
|
| |
I think I've pointed this out before, but "lenticular lens" just
means "lens-shaped lens" or "lens-related lens", which are
tautological. |
|
| |
Anyway, I like it. Maybe 3D TV could finally get and stay
popular if it didn't need glasses. But I'm more excited about
ubiquitous glassesless 3D displays in computers, tablets,
phones, etc. |
|
| |
Not necessarily. The majority of lenses are circular in plan; some specialised ones, like the anamorphic ones used for projection technologies like Panavision (which converts 3:4 35mm frames to widescreen) have a rectangular footprint. |
|
| |
So a lens can have a "lenticular" elevation (indeed it couldn't function as a lens if it didn't, and therefore wouldn't be a lens at all) but can also be "lens-shaped" in plan view. |
|
| |
The existence of rectangular lenses means that a rectangle
is a lens shape. |
|
| |
//in what way would these be superior to existing full
color holograms?// |
|
| |
Well, this would be relatively simple to make, not
requiring lasers and exotic materials. |
|
| |
The main difficulty would be printing the
microimages at
a high enough resolution. A typical printer prints
dots
with a dimension of approximately 100µm whereas
this
idea requires printing dots with dimension of 5µm.
Other
problems would include aligning the microimages
with
the lenses, and accounting for the stretch when the
lenses are pressed in. |
|
| |
These problems could be overcome by using the
film
used in pre-digital cameras. Thus, the Luneburg
microlens array would be embedded into unexposed
film. A repositionable screen would illuminate the
film
from different angles; each angle would expose one
pixel behind each of the lenses. The process
repeated
until all pixels behind each lens is exposed (for a
100x100 array, this would be 10,000 exposures). |
|
| |
I'm not sure how difficult it would be to make the
Luneburg microlenses. Spheres are generally easy to
make because the symmetry puts them at the bottom of
an energy well. That is, droplets and bubbles will
naturally form spherically. |
|
| |
The layers could be made of uniform depth either
through equilibrium of surface tension or through self-
assembly molecules. |
|
| |
Polymers can be made with precisely controlled
refractive index. |
|
| |
Once the Luneburg microlenses are formed arranging
them in a regular array would be trivial, e.g. with an
apertured plate. |
|
| |
An alternative way of making the video version could be
made with a flat (super high resolution) display. |
|
| |
The flat display is covered in adjacent optic fibres
forming a layer. The optic fibre layer would be machined
with an array of hemispherical recesses in which the
Luneburg lenses would fit. |
|
| |
I feel like an entire display, with lenses, could be produced photolithographically. We already have the tech to produce nanometer-scale transistors and whatnot, surely layering up tiny lenses can't be all that difficult. |
|
| |
Re the photographic film implementation: I think you'll need a
projector rather than a screen, but I think it should work
otherwise. |
|
| |
Re the photolithographic implementation: I'd use
photolithography to make the pixels and subpixels (maybe
actually "angels" pronounced like "angles", I guess), with
recesses for the lenses to fit into, and then just pour the
lenses on. (They can be coated with UV-activated glue that's
only activated after they're vibrated into place and the excess
ones are shaken off.) |
|
| |
If you make the lenses photolithographically, I expect there'll
be roughness on their surfaces and also inside them (i.e. each
isosurface of refractive index will have spatial quantization
noise). That might not be a problem if you make them with
sufficiently shortwave light (or electrons) and then use them
at visible wavelengths, though, as long as the roughness is
much smaller than the visible light's wavelength. |
|
| |
//I feel like an entire display, with lenses, could be
produced photolithographically. We already have the
tech to produce nanometer-scale transistors and
whatnot, surely layering up tiny lenses can't be all that
difficult.// |
|
| |
Maybe for a prototype, but for mass production doing it
photolithographically seems very inefficient. |
|
| |
//Re the photographic film implementation: I think you'll
need a projector rather than a screen, but I think it
should work otherwise.// |
|
| |
Yep, a projector or screen that is super collimated so the
image is projected in a single direction so that each
screen/projector pixel exposes a single
corresponding pixel behind the Luneburg lens at each
angle. This collimation could be done with an array of
tubes (with black interior surface). |
|
| |
Instead of the screen/projector moving to different
angles, the screen/projector could be stationary while
the
Luneburg array device moved to different angles. |
|
| |
The exposure process could be done continuously, so
the screen/projector would play an animation of the
10,000 frames while the device was pivoted in a spiral
motion. |
|
| |
Umm..what happens to the other 9,999 frames,
that's a serious storage problem. |
|
| |
//Umm..what happens to the other 9,999 frames, that's a
serious storage problem// |
|
| |
So this device would squeeze 10,000 frames (let's say
each frame being a full HD image (1080×1920)) all into
one device. That is 10000x1080x1920 pixels
(20,736,000,000 pixels) in total. |
|
| |
At any instant a viewer with two eyes would only see
two slightly different full HD images (forming a single
stereoscopic image in their mind). |
|
| |
'spose how thick the wood be on the frames, maybe aluminum ? |
|
| |
It wouldn't be necessary to story all 10,000 frames. Compression would make it trivial to bring it down to 100 frames. |
|
| |
//It wouldn't be necessary to story all 10,000 frames.
Compression would make it trivial to bring it down to
100 frames.// |
|
| |
Not sure what you mean. Could you explain this a bit
more? |
|
| |
^ the computer clock rate of processing in between the needed viewing frame rate. |
|
| |
You don't need to store 10,000 frames at once, just enough voxels to paint the whole 3D picture. |
|
| |
The held voxel data model still has to be processed to turn out data the pixels, under the lenses, need to display for an image that the viewer can again process into thought. |
|
| |
// a projector or screen that is super collimated so the image is projected in a single
direction so that each screen/projector pixel exposes a single corresponding pixel
behind the Luneburg lens at each angle. // |
|
| |
I'd look into telecentric lenses. |
|
| |
Alternatively, you could expose the film by pressing it directly against a very
pixel-dense screen, before applying the lenses. But then you have the problem of
aligning the lenses with the printing again. |
|
| |
On the other hand, I just realized: with the lenses in the way, how do you develop the
film? |
|
| |
Presumably, unless you were mass-producing identical 3D images (and maybe even
then), you'd just store the 3D model and render each angle on the fly. Much less
storage required in that case, but more processing if you're making more than one. |
|
| |
// while the device was pivoted in a spiral motion // |
|
| |
Makes sense, but I think it would need to be carefully designed and calibrated such
that the step size between frames and between turns of the spiral was exactly the
same as the spot size of the Luneburg lenses. If the step size is bigger, then, at
certain angles, a viewer will get only a black image. If it's smaller, then the viewer
will get a blurry image. I think. |
|
| |
ETA: I've also just realized that either the display needs to be backlit or every viewer
needs a headlamp. |
|
| |
//On the other hand, I just realized: with the lenses in the
way, how do you develop the film?// |
|
| |
I suppose you could carefully removed the lenses,
develop the film, and the replace the lenses. |
|
| |
// 10,000 frames //
// [storage] // |
|
| |
Storing 10,000 frames at HD quality is trivial; it's 8
minutes of HD video. This is to imprint the static image
into the film. To make an active 3D display would have
to deliver 20 billion pixels per frame and I want 60
frames per second, so 1.2 trillion pixels per second.
Each pixel would be 3 bytes, so 3.6 terabytes per
second. A completely uncompressed 90 minute movie
would be |
|
| |
20 petabytes. I think compression of 100:1
would be feasible, so a mere
200 terabytes. A terabyte
of storage (HDD) costs $18, so $3,600 to store a
movie. |
|
| |
//I've also just realized that either the display needs to
be backlit or every viewer needs a headlamp.// |
|
| |