Patent: Canon Foveon Sensor

Status
Not open for further replies.

Canon Rumors

Who Dey
Canon Rumors Premium
Jul 20, 2010
12,753
5,575
279,596
Canada
www.canonrumors.com
HTML:
<div name="googleone_share_1" style="position:relative;z-index:5;float: right; /*margin: 70px 0 0 0;*/ top:70px; right:120px; width:0;"><g:plusone size="tall" count="1" href="http://www.canonrumors.com/?p=13562"></g:plusone></div><div style="float: right; margin:0 0 70px 70px;"><a href="https://twitter.com/share" class="twitter-share-button" data-count="vertical" data-url="http://www.canonrumors.com/?p=13562">Tweet</a></div>
<p><strong>Canon still working on Foveon Technology

</strong>Canon has been working on 3 layer Foveon sensor technology for a while now, it still hasn’t appeared in a camera and we haven’t heard anything to say it will. However, the fact they continue to spend resources on it, tells us there is definitely a good application for the technology.</p>
<p><strong>Patent Publication No. 2013-93553 (Google Translated)</strong></p>
<ul>
<li>Publication date 2013.5.16</li>
<li>Filing date 2011.10.4</li>
</ul>
<p><strong>Imaging element of three-layer structure</strong></p>
<ul>
<li>Photodiode deep layer (hereinafter, PD) in the charge is diffused into the substrate</li>
<li>Thus sensitivity is poor red</li>
</ul>
<p><strong>Canon patent</strong></p>
<ul>
<li>I as a photonic crystal of irregular shape, the boundary surface of the PD of red or green,</li>
<li>In the periodic structure of uneven spacing, less than twice the 1/4 times or more the wavelength of light, or I shall be two times or less 1/2 times or more</li>
<li>Irregular shape has the properties to be spatially-localized or resonant light</li>
<li>The incident light is diffracted, it combines with the light of the boundary plane, the resonant</li>
<li>Prolonged exists than normal, the light becomes easy to be absorbed by the PD Thus</li>
</ul>
<p><strong>Source: [<a href="http://egami.blog.so-net.ne.jp/2013-05-22" target="_blank">EG</a>]</strong></p>
<p><strong><span style="color: #ff0000;">c</span>r</strong></p>
 
Dianoda said:
A foveon-style FF sensor in a canon body would be sweet. Canon, dooo ittttttt.....

I'm in love with my Sigma merrill cameras, I would be incredibly interested in a Canon fovean camera. The fact that an APS-C sized fovean sensor can capture as much or more detail than most full frame bayer sensors makes a full frame fovean sensor look really interesting.
 
Upvote 0
andyjaggy said:
Dianoda said:
A foveon-style FF sensor in a canon body would be sweet. Canon, dooo ittttttt.....

I'm in love with my Sigma merrill cameras, I would be incredibly interested in a Canon fovean camera. The fact that an APS-C sized fovean sensor can capture as much or more detail than most full frame bayer sensors makes a full frame fovean sensor look really interesting.

The Foveon has great potential and great flaws.

To fix the latter in a reasonable time, imho, a great deal of resources are needed - the sort of resources maybe only Sony and Samsung have.

The good thing is that manufacturers are understanding that the Bayer sensors (and the mechanical mirrors/shutters) are the bottlenecks of modern cameras' performance. However the Foveon is not the only solution. There's another prototype of sensor (can't recall the name) that uses the energy of the photons to translate it into wavelength and then into color information.
 
Upvote 0
I'm most pleased to read this! The only other rumor I have heard, from a year or two ago, was that Sony was interested in developing the technology. If Canon can make strides, that might give them an edge.

What this should tell all the foveon-Sigma haters, is that the concept definitely is worth the resources spent to develop it. Why? Because the basic idea, is the closest a digital sensor has come, to what film emulsion achieved. That of delivering full color at every location, within the lateral plane. Not only that, but film emulsion also achieved full color in the longitudinal plane...and a very thin plane it was. However, its plane was also that of flimsy plastic film which did not always form a perfectly flat plane within the camera, or the enlarger.

That said, I honestly do not think the foveon approach, will ever achieve parity with bayer or similar CMOS array approaches, with regards to the higher ISO noise floor. I could be wrong, but it just seems like the physics are against it. But that doesn't mean the foveon technique, won't ever be at parity, or even surpass other techniques...at the lower ISO's.

I owned a first generation DP2 for about 2 years, and shot over 5000 images with it (alongside my Canon DSLR and other gear). I loved it a lot, and sold it for a great price. Its images easily scaled to 25 megapixels. The ability to achieve natural color saturation in post editing, was better than anything else I've experienced (with minor caveats such as the weird tint toward the corners of the frame...easily correctable...and obviously nothing above ISO 200). I am now frankly craving a DP3 Merrill, for use in landscape stitched panoramas. Talk about cramming a ton of resolution into just a few shots...! The camera is said to take 8 or more seconds to write the RAW files, so I guess as long as the clouds or other objects aren't moving too fast, panoramas should be achievable. HDR panoramas might be near impossible though, depending on how wide the overall angle of view is...of course if there is no cloud or other movement at all, it might not be so bad.
 
Upvote 0
Albi86 said:
[...]

The Foveon has great potential and great flaws.

To fix the latter in a reasonable time, imho, a great deal of resources are needed - the sort of resources maybe only Sony and Samsung have.

The good thing is that manufacturers are understanding that the Bayer sensors (and the mechanical mirrors/shutters) are the bottlenecks of modern cameras' performance. However the Foveon is not the only solution. There's another prototype of sensor (can't recall the name) that uses the energy of the photons to translate it into wavelength and then into color information.

Why not Canon? AFAIK Canon develops sensors in their own company. O.k., they are behind others in some specs + IQ under different circumstances. But in 2005 when I decided to buy a 20D Canon had a great advantage: It's sensor reproduced great color and detail - much better than other cameras from other brands and it was a Canon developed +produced sensor.
They have to do something revolutionary - evolutionary development of existing 18MPix sensors will not satisfy us and I think we will see some revolutionary sensor within 2 or 3 years.

@paul13walnut5:
Good remark about pixel quality instead of quantity. A 12 or 15 MPix FF sensor three layer sensor without antialiasing and high ISO sensitivity would be outstanding!
 
Upvote 0
CarlTN said:
That said, I honestly do not think the foveon approach, will ever achieve parity with bayer or similar CMOS array approaches, with regards to the higher ISO noise floor. I could be wrong, but it just seems like the physics are against it. But that doesn't mean the foveon technique, won't ever be at parity, or even surpass other techniques...at the lower ISO's.

With lower resolution & larger photosites it could.

Again, from a video perspective this is very exciting, 3ccd's used to be the norm, makes for a very bulky camera, especially with much larger sensors in a block. So foveon might be the way forward, and of course for video you don't need huge MP counts.

A 16:9 8MP could give you 4k video.
 
Upvote 0
bchernicoff said:
Albi86 said:
There's another prototype of sensor (can't recall the name) that uses the energy of the photons to translate it into wavelength and then into color information.

Are you sure you aren't thinking of Roger's April Fool's Day post? http://www.lensrentals.com/blog/2013/04/wavelength-detecting-sensor-eliminates-bayer-filter-triples-resolution

Lol, yes. Maybe this is why I didn't recall it :P
My head moved the memory into the garbage can but forgot to clear it.


mb66energy said:
Why not Canon? AFAIK Canon develops sensors in their own company. O.k., they are behind others in some specs + IQ under different circumstances. But in 2005 when I decided to buy a 20D Canon had a great advantage: It's sensor reproduced great color and detail - much better than other cameras from other brands and it was a Canon developed +produced sensor.
They have to do something revolutionary - evolutionary development of existing 18MPix sensors will not satisfy us and I think we will see some revolutionary sensor within 2 or 3 years.

I was talking about money. Canon is not as big a company as Sony or Samsung. And imho before the Foveon actually gets to a point in which it can replace the Bayer sensors with no drawback, a lot of money has to go in R&D.
 
Upvote 0
"continue to spend resources on it, tells us there is definitely a good application for the technology"

But is a dSLR what they are thinking of?

Its also 2 years since it was filed, why does publishing it make it anything more than visible for IPR use?


Yes i love the idea, but its just an idea.
 
Upvote 0
The patent was filed on2011.10.4 and, as happens most of the time, it is published about 18 months later. This "grace period" buys time to the companies for the filing in the various countries before the content of the patent is made public. Usual procedure. What it just means is that by October 2011 the R&D of Canon completed a particular research project and thought appropriate to file for a patent. It does not mean that they are still working on this or that there is something in the pipe with this. How many patent did they already file on lenses which applications have not seen the light of day yet. Still waiting for the 50 f/1.2 II though :P
 
Upvote 0
The patent sounds like research more than like something they would produce. Canon works on a lot of different technologies trying for a breakthrough. This is a way to improve on the serious issues with Foveon type sensors, but I doubt if its enough to make them better than Bayer technology.

In any event, here is a translation of part of the description. Its hard to read, as all such machine translations are.




DETAILED DESCRIPTION


[Detailed Description of the Invention]
[Field of the Invention]
[0001]
The present invention relates to a photoelectric conversion device, a manufacturing method for the same, and a photoelectric conversion system.
[Background of the Invention]
[0002]
A Patent document 1 has disclosed the photoelectric conversion device which detects two or more colors by a unit pixel by including two or more photoelectric conversion parts in a unit pixel along a depth direction. In this photoelectric conversion device, it is used that the rate of the absorption of light of each color by a semiconductor substrate changes with depth from a semiconductor substrate surface.
[Citation list]
[Patent literature]
[0003]
[Patent document 1] JP,2008-500768,A
[Summary of Invention]
[Problem to be solved by the invention]
[0004]
since a substrate is deep and light with long wavelength, for example, red light, can carry out until entering, the generated electric charge can diffuse to a substrate. This may bring about the loss of photosensitivity. In the photoelectric conversion part which was arranged at the deep position among two or more photoelectric conversion parts, the loss of photosensitivity takes place easily. This influence can become remarkable with micronization of a pixel.
[0005]
In the photoelectric conversion device with which two or more photoelectric conversion parts were arranged along the depth direction, there is the object of this invention in providing advantageous technology, in order to make high photosensitivity of the photoelectric conversion part which was arranged at the deep position.
[Means for solving problem]
[0006]
One side surface of the present invention starts a photoelectric conversion device, and the aforementioned photoelectric conversion device, The component in which light has the 2nd page that was arranged at the incident opposite side of the 1st page and the 1st above-mentioned page, Two or more photoelectric conversion parts which were arranged at the inside of the aforementioned component along the depth direction from the 1st above-mentioned page are included, Any at least one except the aforementioned photoelectric conversion part which was arranged at the nearest position from the 1st above-mentioned page among two or more aforementioned photoelectric conversion parts. The difference of elevation has large uneven shape at the interface with the aforementioned component from the 1st above-mentioned page rather than the aforementioned photoelectric conversion part which was arranged at the nearest position, and the interface of the aforementioned uneven shape is characterized for the light which entered from the above-mentioned 1st page side, and arrived at the interface of the aforementioned uneven shape by localization or making it resonate.
[Effect of the Invention]
[0007]
According to the present invention, in the photoelectric conversion device with which two or more photoelectric conversion parts were arranged along the depth direction, it is advantageous in order to make high photosensitivity of the photoelectric conversion part which was arranged at the deep position.
 
Upvote 0
I enjoy using my Sigma DP1x but the body design and controls are quite clunky. I'd like to see the merger of Sigma and Fuji and have Foveon tech put in a Fuji X100s style body. It would be both a beauty and a beast!
 
Upvote 0
A layered, Foveon-style sensor is definitely intriguing. It has a lot to offer, but at the same time, it also has it's limitations. From a raw spatial resolution standpoint, Foveon will probably never compete with Bayer...at least, not with the same low levels of noise. Foveon can be upscaled, but the results are not quite the same (remember, bayer has limited color resolution, but it still has full luminance resolution.)

The most intriguing technology I've seen to date that maximizes resolution and light gathering/quantum efficiency is Panasonic's MCS, or Micro Color Splitting. Instead of FILTERING colors, red light is split from the incoming light at each pixel, and directed to the appropriate neighboring pixels. The result is White + Red and White - Red pixels that, when interpolated, contain all of the color and luminance information that reached the sensor (there can still be losses due to IR Cut and transmission losses, as well as due to Q.E. losses).

Unlike either Foveon or Bayer, an MCS sensor achieves high spatial resolution as well as low noise. I'd really like to see Canon develop something like this, instead of a layered sensor...I think the end result is even better high ISO performance as well as more color-rich low ISO performance. (And, theoretically, by gathering more light overall, one would suspect dynamic range to improve as well...but I don't really know how an MCS design might affect read noise...which is really the fundamental issue Canon faces these days...noisy high frequency downstream components adding gobs of read noise.)
 
Upvote 0
You will need a new type of monitor to really use it the way it could possibly be made. Imagine no pixels, no 'resolution', just a standard size that everything is scaled to. It might be vector-based with infinite resolution if the picture is digitally made. Or if it captured with this type of Foveon sensor, it will have to use some math processing to figure out what color to put where.
 
Upvote 0
Status
Not open for further replies.