Foveon sensor?

[slackercruster]

Anyone try one? More MP than a D800. Why are they not more mainstream? I never heard about one till today.

 

[usayit]

A megapixel in a CMOS or CCD chip is a bit different than a megapixel in a Foveon sensor. I personally think its a wonderful design and wished it would be taken forward. Unfortunately, the only way to get a Foveon sensor is built within a mediocre body that is priced very high.

 

[Solarflare]

The best Foveon Sensor currently has about 16 Megapixel or some such.

Foveon multiplies this by three with the argument that they have all 3 colors in every spot, but thats not how it works. I'll grant Sigma that Bayer sensors indeed have a lot less resolution than they officially have, maybe 30% less or some such. But not 70% less.

So if Sigma would manage a 25 Megapixel sensor for full frame, they would be about in the area the Nikon D800E is right now. Minus any antialiasing problems. However, with their current way of computing their pixel size, they would claim they had a 75 Megapixel sensor.

The Foveon X3 is a very nice idea. But right now, in practice, it currently leads to serious problems with bad high ISO tolerance and thus bad performance in low light, as well as some problems with proper color rendition.

These issues might or might not vanish in future. You know how it works. Since the 1960s, everyone says tomorrow we'll use gallium arsenide for computers. We are still on silicone.

So maybe in 10 years we all use Foveon X3 fotosensors. Or maybe not. However, in the first case, most likely not in Sigma cameras. As usayit already mentioned, all currently available Foveon X3 cameras kind of suck in respect to their features. Unless Sigma significantly improves in this respect, they will remain in a niche.

Nikon has, by the way, a compareable patent. Their idea is having dichroic prisms on the sensor, for each pixel, instead of for 3 different sensors, one for each color, like its usually done in many video cameras. The problem with this idea is the reduced size of the sensor - basically again introducing low performance with low light situations. It should work nicely, though, if Nikon manages to STACK the sensors on top of each other, going 3D with their sensor design.

I guess Nikon so far hasnt managed this. If they would, they would indeed gain a huge advantage.

 

[EDL]

Different sensor technology that's still more or less in its infancy. Whether it'll grow and get better is anyone's guess. Even if it does there's still no guarantee it'll become mainstream...might end up similar to the beta max/VHS outcome....time will tell.

 

[Trever1t]

Solorflare pretty well summed it up...interesting technology but perhaps we will never see its fruition.

 

[KmH]

The Foveon image sensor taps into the fact that light of different wavelengths penetrates silicon to different depths. File:Absorption-X3.png - Wikipedia, the free encyclopedia
Pixels that can see red, can also see green, and blue. Pixels that can see green, can see blue, but can't see red. Pixels that can see blue, can only see blue.

In 2008 Sigma bought Foveon, Inc., the California company started in 1997 that developed the image sensor.

Foveon was initially known for their high-end digital portrait camera systems built around a color-separation beam-splitter prism assembly.

 

[Garbz]

Different sensor technology that's still more or less in its infancy. Whether it'll grow and get better is anyone's guess. Even if it does there's still no guarantee it'll become mainstream...might end up similar to the beta max/VHS outcome....time will tell.

This technology is not in its infancy. It has existed for more than 15 years and been well researched over that period. The problem is cost and specs vs performance. So far every new foveon sensor has been bested by beyer counterparts in nearly all tests. The AA / debeyering is bested by the far higher native resolution of the typical sensors, and despite the foveon sensor having the advantage of photosites which are physically larger, and in theory a 3x advantage in quantum efficiency due to no beyer filter they are still beaten in noise / high ISO performance.

Then there's the other problem of what's pleasant to see. You see 100% awesome visual acuity due to a lack of anti-aliasing filter sounds awesome on paper, however it's not awesome for the same reason that computer games running even on the maximum supported screen resolution still benefit from anti-aliasing. By having perfect visual acuity tiny details may skip over pixels, i.e. a power cable that is rendered as less than 1 pixel wide may actually skip pixels in the picture and look discontinuous. However when AA is applied the cable would become less sharp and slightly faint but also continuous which is more inline with how our eyes see.

Personally it's a nice technology for scientific purposes, but it has drawbacks for conventional photography.

 

[Solarflare]

Personally I think the comparison with anti-aliasing of computers is more confusing than helpful, because its a completely different thing.

The anti-aliasing of fotosensors is because fotosensors see the world through colored glasses (Bayer or Fuji X-Trans or whatever). Anti-aliasing helps to preserve color information which could otherwise be lost. So the question of anti-aliasing for fotosensors is finding the right amount of it, and too much is definitely bad, and you can leave it out like with the D800E and still get some kind of result.

Anti-aliasing on the computer is because the computer will compute the color of the middle point of the pixel. Not the sum of colors in each pixel, as our eyes or a camera would do. Anti-aliasing means the computer will compute the color of multiple points in the same pixel and color the pixel to the average color of these multiple points, thus getting closer, but obviously still not perfect, to what that color would look like would the scene be real and not just triangles forming surfaces painted with textures in a computer. In short, the more anti-aliasing on the computer, the better.

 

[unpopular]

The anti-aliasing of fotosensors is because fotosensors see the world through colored glasses (Bayer or Fuji X-Trans or whatever). Anti-aliasing helps to preserve color information which could otherwise be lost. So the question of anti-aliasing for fotosensors is finding the right amount of it, and too much is definitely bad, and you can leave it out like with the D800E and still get some kind of result.

An optical antialiasing filter takes a single bundle of light and splits it into four bundles of light, allowing any given point of the projected image to be averaged out over the area of the bayer mask. Too much anti-aliasing and you'll end up with more of those separated bundles ending up in region where they don't belong, too little and you'll end up with an region receiving information from another point of light.

There are ways around this. You could skip the AA filter and simply move each of the recorded pixels digitally. My RAW processor has a mode which permits this called "Half", it is extremely fast and the color and noise levels is very good, but at the cost of resolution. You could also move the sensor by one pixel, sampling each color component in the same region, this technique is called Color Co-Site Sampling. However, exposure time would have to be compensated and currently only Sony would be remotely ready to implement such a technique and only as an option - though it would be an interesting option for since resolution would effectively be doubled.

 

[sovietdoc]

Nikon is currently playing around with idea of moving the AA filter inside the lens..

 

[unpopular]

what?! that is a terrible idea.

 

[dxqcanada2 1]

Sony has been trying to figure out a foveon type sensor: One more foveon like Sony patent! | sonyalpharumors

 

[unpopular]

The foveon approach is interesting, one could implement phase detection into one layer of the chip. The concept is provable, Fuji incorporated Phase detect into one half of six imaging pixels in one of their P&S cameras.

 

[Solarflare]

That means we can have Phase AF on the sensor ?

That would indeed be a massive advance that would remove the main advantage of DSLRs over other types of cameras.

 

[unpopular]

See:

Autofocus - Wikipedia, the free encyclopedia