Sometimes patents discloses quite weird things. Egami spotted that double new sensor design with hexagonal and triangle pixel design. The aim of the design is to increase the color fidelity by using a RGBCMY instead of the classic RGB pixel design. I doubt this kind of sensor design will ever make it into a real camera…
33 Responses
@Admin Fuji uses octagonal pixels in their S3/S5 cameras – they call them “R-Pixels”. It’s at least conceivable that Sony would give a chance a unique sensor designs as well since they do have such a vested interest in maintaining their majority of the sensor market.
As owner of Fuji S5 Pro, I might say that I would love..
..if Sony never tried same thing. 😀 Only if they managed to implement it a lot better. Fuji didnt.
Sony proofs that in lab adding white pixel RGBW significantly increase the ISO sensitivity under Exmor RS development, but due to some production issues, they drop the design.
I believe adding CMY pixel will improve the color depth, but may be under very limited medical/industrial usage
This is neither senseful nor necessary. Some RAW converters assume Beyer pattern, so you limit your choices. You still have the moire issue. And while using RGBCMY would widen the gamut of *displays*, it doesn’t do that for *sensors*. Colour fidelity is also increased by bringing the existing RGB pixels closer to the standard observer, google for “Luther-Ives condition”.
To sum it up: A lot of effort and trouble for a mild advantage. This effort is better invested elsewhere …
So to sum it up you’d rather stay with the current technology than explore what benefits would bring new technologies. Really don’t understand your way of thinking.
I’d like to see some advancements in sensor technology and move from square pixels to different shapes – why not? The only sensor tech available in commercial products are currently Foveon sensors and benefits are clear at lower sensitivities.
Huh? Every idea is good and should be evaluated by investing a lot of money for the working time of expensive engineers, filing patents, possibly producing prototypes? Aha.
Man you should be working for sony, it’s what they really need to open their eyes. They should fire their workforce of engineers and hire u instead
I’m curious about this. If adding C, M and Y pixels to monitors widens the range of colours they can display, therefore widening their colour gamuts, why wouldn’t adding those pixels to a sensor do the same by allowing it to capture those colours to begin with?
You are suggesting that sensor technology should not change because there is no software RAW converter that will support it initially? New software can be written, and old can be re-written; Sony can provide it’s own RAW converter (like all other manufacturers do) and 3rd Party software will adapt. Besides, the hardware and software algorithms typically get developed jointly.
Consider this:
The sensor does not have the normal square pixel array, so, the software does not map the sensor pixels to screen/jpg/png pixels, instead, the software calculates linear gradients between a hexagonal block to its neighboring six hexagonal blocks. Then it calculates virtual points in a square arrangement and maps those to be output. As far as the rest of the hardware (And the raw converter) is concerned, the sensor has square pixels. Is this process computationally expensive? Yes. Does current portable hardware have the juice to make these complex computations? Yes, at the cost of battery longevity. Should this be made? Why not try?
If the pixels are small enough and you overlap the large hex groupings, then you could almost eliminate moire patterns, because the overall effect of the grouping overlap is the same as that used in the X-trans Fuji sensor.
Anything that increases the colour gamut is fine to me. We’ll need displays with a colour space wider than sRGB/REC709 to display it, though.
Not a problem really. Majority of screens dedicated for photographers and graphic artists offer much wider gamut than sRGB.
I still have an old DSC-F828 which used extra colours in it’s sensor. RGB and Emerald. For usability, it is still my favourite camera/
I also!!
Interesting ideas. With more than three colors being sampled, the color rendition and reconstruction will be more reliable as they are today. Many people moan about problems with skin colors, plant’s green colors and the like. This is because the sensitivity and colors of eyes, sensor, display, and color prints differs. WIth 6 colors (and optional white), sensitivity and color reliability will increase massively. And I believe that RAW converters will be available soon after first products become available.
Perhaps 1 step closer to variable curved sensors.
Just like God designed eyeballs
😉
Insects have compound eyes in a hexagonal array.
So can we get 16 bit RAW out of it and JPEG with much improved colour fidelity & DR ?
Direct CMYK output?
And, what is the maximum useable ISO?
I guess it, as always, depends on the area size that doesn’t collect light. Not the color filter itself.
Im not clear how this configuration would work.
It does however seem that these pixels would be huge explaining why sony intends to limit resolution to 30+ mp.
When Fuji did hex sensor array they onterpolated the image res to double the native resolution
I need to go back and read AnimalsBarry’s posts on CMYK sensor again….
Where was it?
That was in the post about the canon phase detect sensor.
It describes a different Sony ycg 2layer in 4partsthe plus a seprate link to the patent description.
My phone did not allow long comments that day so i broke it up.
I think it was about 2 or 3 days ago.
Note that was a differrent new patent i was talking about.
I actually drawed something like that last year (guess Im not only one who came to this). Tho just that right side variant, splitting pixels didnt occur me as good idea (if it should be possible to produce.).
Otherwise I think we will see it one day.
Tho I think that making simply hexagonal RGB is much better idea (it should lower moiré issues both with or without AA).
Adding “white” isnt bad, but theres issues with change of demosaicing. Fuji X learned that too that making unsupported CFA is nice, but not without consequences.
Actually found my “drawing”. 😀
http://www.ulozisko.sk/obrazky/623425/matrix.jpg
Made it 26.10.2012 (timestamp is there).
I think that RGB pattern could work rather well, issue would be lowered DR maybe (cause you need strong green channel). Tho with bit of thinking I would figure something out.. just dont find that particulary useful for me. 😀
The human eye proves that all colors that we matter about can be interpolated from RBG. Looks like a boring brute force solution (instead of clever interpolation), which does not have to be bad in practical terms however. Many things are solved well enough by brute force and if this does it, why not?
Just some basic color theory that might help anyone trying to figure this out.
The primary light colors are red green and blue. Add them together you get white light.
The primare absorption (pigment) colors are
Yellow absorbs only blue light.
Magenta absorbs only green light.
Cyan absorbs only red light.
Add them together you get black pigment.
Just a little usefull info on the topic. Not actually a comment
That WAS usefull information. DO you have more usefull information packed in your head you can share with us? Please share more usefull information, I want to listen to how usefull it is.
So if Yellow absorbs blue light, and blue absorbs red light, and red absorbs green light, what does green absorb? Does green absorb orange?
You criticize whatever I say because you dont like my comments about high resolution however you fail to notice that 85% of my comment have nothing to do with resolution.
While I have your attention compare the quality of your comments with the quality of the comments you are criticizing.
Light is NOT red/green/blue. Light is a continuous spectrum from infrared to ultraviolet. The human eye happens to use a particular way of sensing the colour. That doesn’t mean that it “the only way”. There are some advantages to using a sensor which mimics the way our eyes see things, but that doesn’t mean we have to blind ourselves to alternatives.
And the idea that we can’t use a CMY sensor because some RAW converters assume Bayer matrix – that’s insane and completely backward! If we followed that logic there could never be any advances in sensor design, because they would break the limitations of existing RAW converters.
Curious how anything other than emerald/cyan is of use color-wise. I’m not colorometry expert but I do know RGBE can give very true perceived color. Perhaps it’s more for laboratory/industrial application.
Good to know many options are being looked at – each allows something to be learned, even if it’s not used in this form.
One problematic element to this pixel arrangement is that current output devices utilize the traditional square pixel arrangement and therefore interpolation would ALWAYS BE REQUIRED to use these images.
I dont actually see how interpolation would improve the IQ just that it would be necessary.
This sensor would probably not be the best choice for highest possible resolution.
I have not yet figured out how or if the color profile works so I won’t comment on that.
True Barry, but consider that even for the square Bayer pattern, images are interpolated for each color. If the hexagon “groups” are the same size as a RGGB group, the resolution would really be the same.
You may be right but….
The pixel groups do not line up vertically like they do in the bayer array hence a little additional interpolation…..
And six or seven color pixels take up more space than the four of the bayer array therefore they would either be smaller or fewer in number.
Thats why it seemed to me that yhis would be geared towards improved color rather than higher resolution.
I am not at all certain if my guess is correct.
Sony’s CINEALTA F65 uses this sensor. That’s why it has a color gamit that surpasses film. It’s just 20.3 megapixel but Sony claims that it will do 8K RAW. So this sensor insome way also enables more pixel interpolation or something,