Foveon again

There is a new 10.2 Megapixel (3.4 megapixels x 3 colors)

Foveon X3 PRO 10M Image Sensor, used in the new Sigma SD 10 Single-

Lens

Reflex digital camera.

 

To read or download the Foveon News Release about this new Foveon

Image Sensor

please visit <a

href="http://www.sironline.com/pressrelease_foveon.htm">

http://www.sironline.com/pressrelease_foveon.htm</a>

Why don't they call it a 81.6 Megapixel sensor (3.4 megapixels x 3 colors x 8 bits)? Sounds even better on paper!

<p>

Someone should tell the marketeers at Foveon what a "pixel" is.

Someone should tell the poor consumers, what a pixel is. The marketeers know it very well, imho.

Now maybe I'm completely wrong about all this, but I'm going to have to agree with Michael. It's important to know the "quality of the pixel" and sensor when you differentiate between cameras but in the end, it is # pixels that counts the most to me. If you don't think it counts, then ask yourself, "which can be enlarged more...11 mp from a 1ds or this "10.2"mp from the foveon?"

 

From my understanding, the Foveon sensor's advantage is that there is less "processing" required by the camera post-capture. However, I'll take a Canon 1ds' true 11 megapixels over this "marketing hyped" 10.2 any day.

 

Please correct me if I'm wrong, but the resolution of the resulting "10.2mp" file is 2268x1512. That's only good enough for a 5x7 image at ~300dpi. How is that better than the image you can pull out of a 1ds?

 

Who knows, maybe I'm missing something here?

Who cares how many pixels they say it has? The people this camera is marketed to are not stupid enough to buy simply based on a MP number. If they are, then they deserve to have their money emptied from their wallets.

 

What comes down to it is the quality of the final output. Does the Foveon perform better or worse than other sensors? Are you satisfied with the final output?

 

The rest is useless posturing.

yes, you missed something. You missed my point:) What I mean is that marketing folks can fool lots of consumers with numbers like "10.2 megapixel" just because the half of the digital market is based on magic "megapixel" numbers. Everyday people don't know anything about sensor size, about the widest angle of the zoom lens on their digital, and <i>they don't care about</i>, they just feel happier and prouder with a bigger number of megapixels. This is already starting to be abused by the people mentioned by Bob Atkins, it seems.

conceptually this Carver mead sensor invention should blow everything else out of the water. the comparisons with film and other sensors on the older sensor in the sd9 seem really impressive in this article in Discovery Magazine http://www.discover.com/issues/dec-02/features/featphoto/

Just what is the apprehension or problem with this system/

jerry

I agree with you Jerry. The irrational widespread posturing against Sigma and Foveon makes no sense to me. I see in the various forums quite a lot. As consumers, we should all be welcoming any change in the status quo, we should applaud more choice. It is a "good" thing if Sigma becomes a success and tries to topple Canon, because that will force Canon to make better products.

 

Sigma product releases are sometimes met with disdain, as if people are threatened by them. I don't get it. There is supposed to be competition out there. Companies are supposed to come along and try to topple the current market leaders. That is what is supposed to happen.

 

I am not a marketing type but I wonder if this is part of what is called "branding". That is, people become so wrapped up in a brand name that they adopt it as part of their persona, thus resent newcomers on the block. Having virtually no brand loyalty, I confess I don't understand. (For the record, I have mainly owned Pentax manual focus gear but now shoot often with my wife's film Rebel.)

 

I think that what Sigma is doing with the Foveon is really interesting. On the surface the technology behind the Foveon chip appears quite elegant whereas Bayer interpolation seems like a kludge, to me anyway. Does it mean it's better, or that it can be better, I don't know. But it at least means that it's worth taking a look at and Sigma have decided to do just that.

 

What exactly is the problem with that?

I really think we need to compare the foveon sensor in a completely new way . it is not really comparable to anyting out there as far as I am concerned. AS I understand it, each pixel is bigger than any other sensor on the market. Foveon pixels are 9.5 microns compared to the closest competitors 6.5 and 7.5. . That is capturing a lot more light and a lot better picture . In addition other sensors on capture something like 60 % of greens and 70 % of bluyes etc while the foveon chip captures 100% of each color. Nikon doesn't have a proprietory chip while the others do so there is a strong interest in hanging on to their own chip. But when another camera maker makes the move to foveon chip I think there will be another revolution in photography

I think my Nikon coolpix 5000 will feel like a brownie hawkeye in a couple of years. I can't afford a sigma SD10 but I am searching for a way to invest some of my pension money in Foveon because I believe this radical new sensor which combines all three colors in a single pixel rather adjacent to each other will be the sensor of the future. An image from this sensor IS a 10mb image no matter how one would want to argue the opposite or I am really missing something.

Check out the review of the SD10 at www.dpreview.com as well as th imaging resources. I have been researching a way to invest in this co. I have absolutely no connection iwth the company up here in Canada, just a retired teacher marvelling at this invention and its potential . My research has dug up these links . Check out the comparisons at: http://www.foveon.com/news_pressroom.html

and an intersting release from:

http://www.national.com/news/item/0,,727,00.html

Although they are a private company I have found a way to invest fairly directly in the company. Wish me luck!

Jerry, you are missing the fact that it is a 3.5 megapixel sensor which produces images with 3.5 million pixels, not 10. The files may be sharp for 3.5MP images since no antialiasing filter is used but it is a joke to compare it to the 11 MP Canon.

Yes, Ilkka, but each of those pixels produces 3 colours in their native operating mode, that is, not interpolated. The problem is that the 2 methods of generating images that the Bayer and Foveon employ are apples and oranges. Attempting to compare them using a single phrase, "XXX megapixels" cannot work.

 

Is Sigma completely candid calling their camera a 10 megapixel? Well, yes and no. There are only 3.5 million photosites, as they appear to the lens, but each records 3 independent colours. Is the marketing hype a little misleading? A little, sure it is. But when was the last time anybody called the Bayer-cameras to task for interpolating 2/3's of their colour data? And why aren't there threads of complaint about Fuji not only interpolating Bayer colours but going further and interpolating entire photosites the way they do in their octogonally oriented sensors.

 

Marketing people do what marketing people have always done. Calling one vendor/manufacturer to task and not others, although not wrong, is not quite right either. Rather, it's incomplete.

 

As another contributor stated, it's the end result that matters, and it's the job of consumers and information sites lke this one to point out the differences between claims and reality.

My point is that if you think the Foveon has 3.5 x 3 photosensors, then the Canon 1Ds has 33 million of them. The only difference is that the Canon uses an antialiasing filter to smooth out the light so that you avoid getting colour artifacts from the Bayer pattern. This reduces the overall resolution a little. Filtration isn't necessary with the Foveon, and it is becoming increasingly unnecessary with higher resolution Bayer sensors (see Michael Reichmann on Kodak's 16 MP medium format digital back) except for fashion/product type shots.

Ups, meant 3.5 x 3 million.

Ilkka,

 

I don't disagree with you. In one sense, their claim is misleading. OTOH, they can assert that since the 3 layers of sensor material each independently records a colour, then they can be considered "photosites" or "pixels". But because the three layers are top of one another and so look like one photosite to the incoming photons, then we balk at the marketing claim. This was what I was getting at in my post above. Attempting to compare the details of sensor technology becomes almost beside the point since they are apples and oranges comarisons.

the foveon pixel is bigger, 9.5 microns to 7.5 and 6.7 for nearest competitors AND it records 100% of each color as opposed to something like 60% of some colors like other sensors. Surely this makes a big difference in addition to the fact that within each pixel it is recording three separate pixles as opposed to adjacent pixels Yes it is comparing apples to oranges and it seems to me that the foveon is the sweeter

The Foveon sensor distinguishes colors by the fact that photons of different wavelengths tend to penetrate to different depths in the silicon. The pixels do not collect "100% of each color" and, in fact, since this is not a deterministic process, a good portion of each color (considering the spectrum divided into 3 bands, that is) winds up counting towards the *wrong* color. In addition, the Foveon pixels are somewhat less space-efficient than conventional pixels. I'd say at present there isn't a world-beating advantage one way or the other, if faux Foveon pixels are compared to faux conventional interpolated color pixels, anyway.

The problem with the Foveon sensor is that so far it seems to be a solution in search of a problem. If you've actually seen large prints made from a camera like a Canon 1Ds (or even a Canon 10D), they look pretty damn good. In fact they are excellent.

 

So what's the advantage of a Foveon sensor? While you may think Bayer interpolation is a bit of a cludge, in practice it works very, very well (when done properly). So unless the Foveon is way cheaper there's not much incentive for major manufacturers to use it since the results (i.e. the final print) won't be very different.

 

It's neat technology, but I've yet to see any convincing evidence it's actually better than existing sensors when it comes to performance in the real world. Maybe that will happen one day. It just hasn't happened yet.

Bob, I don't disagree with anything you wrote. Bayer sensors do work well, no doubt about it. And I don't know what advantage the Foveon brings or might bring in the future. There may not be any advantage to the Foveon design at all. Personally I was wishing for a cost advantage but there doesn't seem to be one so far, judging by the selling price of the Sigma bodies.

The main advantages are that with the Foveon sensor you don't have the chromatic aberration problems of Bayer sensors and that you don't need an antialiasing filter which reduces the sharpness of the image. Of course you can use more unsharp masking but by using the antialias filter you make the higher spatial frequency components of the image closer to the noise level, and applying the sharpening operation makes the noise more visible.

 

But they still need to increase the pixel count to comparable levels before we can make a meaningful comparison. Kodak uses a removeable antialias filter in their medium format digital back, it's removeable because so doing you get sharper images - with the risk of getting colour artifacts.

I think it's more accurate to say that Sigma haven't used anti-aliasing filters with Foveon sensors than Foveon sensors don't need anti-aliasing filters.

 

You don't get the chroma-based moire patterns from a Foveon sensor without an AA filter, but you do get spurious resolution effects and luminance aliasing artifacts as you would expect from sub sampling high frequency information.

 

I think Foveon chose not to use an AA filter because it can make the image look sharper while the aliasing effects aren't quite as obvious as those you'd get from a Bayer sensor without an AA filter. However those effects are still there.

Yeah, I thought of that too but I haven't seen it nor heard anyone talk about it so I figured it's not a serious problem. Anyway, this can be avoided by increasing the sensor pixel count or reducing the quality of the lenses. :-)

Can anyone explain what the actual background behind the Bayer pattern is? It doesn't really make sense to me, to organize the colours in the sensor in the way they seem to do. The human eye has one type of sensor for each colour, and they're distributed differently. Of the colour sensitive cones, only a few percent are blue senstive. The red and green are used for detection of colour edges.

 

I misunderstood the way they count pixels in the Bayer grid, sorry about that. Now I'm more puzzled than ever.

Does this camera fix the problems with the SD-9? Namely, the extremely cold color

rendition, the 'saturated pixel' color errors, and the obscene tendency to screw up

night photos? My buddy is looking at digital SLRs and I would love to recommend this

camera to him, but the Digital Rebel is here, now, and doesn't have any glaring

problems (image-wise, anyway). BTW I am a 10D owner, and have been about 100%

satisfied with my camera.

Would you also recommend he buys a camera that can only use Sigma lenses, since that's the case with the SD 10. Not that all Sigma lenses are bad, in fact some are quite good, but it is a limitation. With a 300D you can buy Canon, Sigma, Tamron and Tokina lenses, not to mention lesser names like Vivitar (who do a decent, cheap 100mm macro). With and SD 10 you get to pick from Sigma, Sigma and Sigma. That may be enough, but it's a point to keep in mind.

Since I´m in the process of finding my next digital camera I read now all the

explanations and interpretations to the "correct" numbers for resolution and pixels.

The Foveon sensor "sees" 10.2 million "pieces" of information without interpolation

and producess a picture with 3.4 million pixels ( using true measured RGB values and

not interpolated) whereas the others throw out > 6 million pixels with interpolated

RGB values and massive in-camera processing. The quality of the pictures is without

any doubt very good for both.

Now to be fair should we not allow at least a little out-of-camera processing to end

the resolution discussion?

I can not see a problem when someone wants a greater number of pixels to do such a

interpolation step later which by the way is one of the options of the software that

comes with the camera.

The result should not be worse than the pictures which everybody else gets right out

of the camera since the information used to do the job is "unmodified" yet.

Obviously there is more than one way to catch the light and shoot a decent picture.

I have also seen alot of SD-10 pictures today and I´m rather speechless about their

quality.