CCD/CMOS vs Super CCD vs FOVEON

[chema_perez]

For an image quality point of view only, what's the best to do large prints and

large hight quality images. For studio, portrait and landscape work.

 

After some Googeling :

 

Nikon/Pentax/...(sony CCD) 10MP 3872x2595 More pixels

 

Canon CMOS XTi (futur 40D?)10MP 3888x2592 More pixels

 

Fuji Super CCD S5 6,1MP 3024x2016 Best dynamic range

 

Sigma Foveon SD14 4,7MP 2640x1760 Best color, pixel per pixel

performance, good range (guesing the performance will be equal or beter to the

SD10 but with more pixels)

 

In practical use the CCD and CMOS are very near in performance so a choice

between those two depends on other factors like avaiable lenses, price...

 

The Foveon gives ferfect pixel per peixel definition performance compared to

the others, but only 4,7MP, so if we downsample a CCD/CMOS image from 10 to 7

MegaPix with apropriate software I supose we can get same sharp results. The

dynamic range is beter than the CCD/CMOS.

 

The Super CCD has a bit more of pixel per pixel definition than 10MP CCD/CMOS

(just like the old nikon 6,1MP sensor had a bit more pixel per pixel definition

than the new 10MP sensor) But have to be downsampled a bit to get the same very

sharp result of the Foveon. This one has the best dynamic range, this is

critical when you can't modify the avaiable light and contrast of the subject.

It's also the best in hight ISO .

 

The 10MP CMOS/CCD have the worst dynamic range, hight ISO performance and pixel

per pixel definition but give a lot of pixels.

 

I'd like to comment this as I didn't compared for my self any of these cameras,

it's just googeling and theory. I supose the best one would be the one that

permits the bigest hight quality prints, even if it demand some time after

processing.

[mike_earussi1]

Everything's a tradeoff. Canon has the best image at high ISO. Pentax/Sony have more features for the price. Fuji, as you pointed out, has the best DR. Sigma, from what I've seen, seems to have the best image quality but at the price of the fewest features, it can only really be shot at 100 or 200 ISO, no weatherizing, no IS type lenses, and can only use Sigma lenses.

 

So ultimately it depends on what's more important to you.

[doug_axford1]

CMOS uses much less battery power. That was why I switched from Fuji years ago. Too many shoots interupted with battery changes. Almost never with Canon.

[dave_powell2]

As Mike said, tradeoffs are everywhere, to make decisions difficult.

 

If you haunt the dpreview site, you'll see that image-quality often seems to suffer when more megapixels are squeexed into the same sized sensor. For example, some users prefer the image quality (in a lower-noise sense) from the 5MP Konica/Minolta A1 than from the subsequent 8MP A2 (both of which used a 2/3" CCD).

 

Perhaps a REALLY expensive 16MP or 30MP studio digital could claim to be "the best," but other than that, all cameras have both praisers and detractors.

 

Sincerely,

 

Dave

[geoffs1]

If all you're looking at is image-quality, don't forget: http://www.betterlight.com/index.html

 

Cheers,

 

Geoff S.

[ned1]

I own a D200 and. Fuji S5. I can sum up the differences as:

 

Fuji: 6 stops of latitude. Which is one stop more than what can be reproduced in print. This means you can overexpose and still get a good image. Better colors. Good pictures at ISO 1600. Downside: 6mp is good for an 8x10 but not for something bigger. The fuji is a terrific camera for situations where you have little control of the light.

 

D200: upside is the 10mp sensor which is goot for 11x14s.

Downside: only 4 stops latitude so even if you exposure is perfect there are situation where you can never get a good print. Useless above ISO 400. Also the colors are not as good. A good studio camera but no good for available light.

 

I don't own a Canon or Sigma but from what I understand:

 

Canon mark II: best dslr on the market. 17 mps means it can make fairly large prints. 5 stops of latitude which means is you get you your exposure perfect you can always get a good print. Good at ISO 3200 although the Fuji comes close.

 

Downside: $$$

 

Don't know much about it but in theory colors on the Sigma should rival film. Other than that it is unclear what advantage it has.

 

Hope this helps.

[robert lee]

"what's the best to do large prints and large hight quality images. For studio, portrait and landscape work."

 

The type of sensor doesn't really matter all that much. If you're shopping for a camera, take into account the overall system and specfically the lens mount. I'd stay away from Sigma.

 

Given the same digital camera generation, overall technical image quality is affected much more by optics, lighting, capture technique, and digital darkroom work than any sensor differences.

[Ed_Ingold]

To make a few corrections...

 

The dynamic range of a D200 is at least 7 stops, per DPReview.com. A 6MP camera provides the same or better prints as 35mm film camera, owing to the limitations of film. A 10MP camera is closer to medium format (645) in quality. The major manufacturers, notably Nikon and Canon, remain committed to Bayer-filter sensors and rectangular arrays. They have the resources to use other technologies, but do not, presumably for business reasons. CMOS is replacing CCD technology. The quality is equal to or better than from CCDs, and the structure allows for faster, more flexible image processing, using far less power (with less heat).

 

You buy into a system, not a particular set of features. With a company like Canon or Nikon, you have a wide selection of lenses and bodies, with complete interchangeability within that company. You can start low and build with good assurance of future compatibility (and company stability).

[jcuknz]

To call the Foveon a 'low res' camera is ignoring the point that while the rest of us are using Dufaycolor, Foveon is using the Ektachrome system.[Or Kodacolor if you like] So the 4.7Mp is the equivalent of 14.1Mp with the advantage that there is only one pixel doing the job of three for the rest of us.

 

I'm sure there are other pros and cons to each system but the basic superiority, potential if not realised, of the Foveon system is obvious and really it is a pity they are sitting on their patents instead of sharing it with the world.

[chema_perez]

Thanks for the posts.

I've been looking for the dynamic range capabilities of each acording to dpreview tests:

 

10MP Canon : 8.4 EV

 

10MP Nikon : 8.0 EV

 

6,1MP Fuji : 10 EV

 

The tests are on gray scales so it doesn't mean one can have a lot of information in the extremes.

 

In real life, after looking at the samples the difference is minimal, even between the Fuji and Nikon. The Fuji captures more detail in hight lights so it alows to overexpose a litle to bring details to shadows and stil have detail in the high lights (overexpose of about 2/3 or 1 EV?). That's not bad.

 

Like others said I dont like the idea of using only Sigma lenses. I can say Sigma is out of the competition here. If the Foveon was on a D200 with a bit more of resolution than 4,7MP...

[dave_redmann]

<I> Foveon . . . the 4.7Mp is the equivalent of 14.1Mp</I><P>

 

No. A 4.7 MP Foveon sensor (like on the Signa SD14) is very likely to be the equivalent, resolution-wise, of a 7.5 MP Bayer-array sensor (which is what almost all non-Sigma DSLR's use, and is independent of CCD vs. CMOS). This has been explained at length elsewhere on photo.net. The Fuji S3 and S5 use 6 MP of big pixels plus 6 MP of small pixels to increase dynamic range. Compared to a regular Bayer-sensor camera, various reports suggest that the effective resolution is somewhere between 7 MP and 10 MP.<P>

 

I would not get too worked up about the technology behind any given camera. Study up on what numerical ratings of resolution, dynanic range, etc. mean. Then read reviews that contain actual instrumented testing (primarily, in the US at least, DPReview.com and <I>Popular Photography & Imaging</I>) and experienced users' comments.

[chema_perez]

Excuse my ignorance, how much can we enlarge a 4,7MP image before start seeng the pixels?

 

I'm guessing 2640x1760 pixels is not enough but I don't have any idea!

[mike_earussi1]

Because it takes four Bayer sensors (two green, one red, one blue) to equal one Sigma sensor.

[dave_redmann]

<I>Excuse my ignorance, how much can we enlarge a 4,7MP image before start seeng the pixels? I'm guessing 2640x1760 pixels is not enough but I don't have any idea!</I><P>

 

IMOPO, for most images, 4.7 MP <I>from a Foveon sensor</I> will look pretty good at up to 11 x 14 inches (or roughly A3 to the metric world). But I've gotten some 24 x 30 inch (roughly A1) prints from my 6 MP DSLR that looked surprisingly good, especially considering the typical viewing distance.<P>

 

<I>Because it takes four Bayer sensors (two green, one red, one blue) to equal one Sigma sensor.</I><P>

 

No. Go read the actual resolution tests--DPReview found that the Sigma SD10's vertical resolution, from its 2268 x 1512 pixels, was 1550 lines, and its horizontal resolution was 1550 lines per picture height--in other words, a little bit less than most of the 6 MP Bayer-sensor DSLR's (say, roughly 5.5 MP). Based on the technology involved, the SD-14's resolution from its 2640 x 1760 pixels, <B>which is the file size you will actually get from the camera</B>, should prove to have about the same ratio, which is why I'm saying it's the equivalent of a Bayer sensor with 7.5 MP. I will not again explain the technical reasons--I've done so before. But please don't spout some theoretical resolution when (1) you don't know the test data and (2) you don't understand the theory. It just confuses people, and perpetuates serious misconceptions.

[robert lee]

"it takes four Bayer sensors ... to equal one Sigma sensor."

 

No, this isn't quite the case. The Bayer filter does reduce spatial resolution, but by no more than maybe few 10's of percent relative to the baseline sensor array. Foveon's architecture is an argument that raises the importance of chroma resolution. I think they're optimizing the wrong thing.

 

Try taking an image into Lab color space. See what you get when you throw away the L* channel versus a*, or b*.

[mike_earussi1]

To form one one color patch in a Bayer sensor they take the output from four sensors. But in B & W that is not necessary so the absolute resolution of a Bayer sensor will always be much higher in B & W than the Sigma. But I would expect the resolution to be much less for color which I have yet to see anyone do a test in.

[troyammons]

Dont let the Fovenon 4.7 mp output size fool you.

 

Go shoot a 5d next to a SD14, then upsize the SD14 and you will see that they are very close in resolving power.

 

The positives are, brilliant color and ultra high mtf at the chip, no AA filter and no color interpolation.

 

Its a pro chip in a consumer body although the new sd14 body is a big step up from the old sa cameras.

 

The downsides are lesser body functions, only sigma lenses but that does not bother me, and not as good high iso, and a bit expensive.

[roger krueger]

The Foveon doesn't have ANY inherent advantage in resolution over a Bayer chip.

 

What it does have is no AA filter--possible because you don't have color aliasing issues

like you do on a Bayer system. Look at the dpreview test pattern--this thing resolves

spurious detail well past its Nyquist limit, whereas properly AA'd Bayer cameras fade nicely

to gray as they run out of resolution.

[carl smith]

I've always found the Fuji SuperCCD could look a bit awkward with some subject matter, but as I haven't shot one lately I can't elaborate so I won't really count that against it.

 

It has a slight dynamic range advantage over my 1D MK II, but overall noise performance on the 1D MKII is better and resolution between the two is close.

 

I haven't got much experience with the X3 sensors but that Sigma DP1 camera is really tempting for me so that might be my first real extended use of a camera incorporating that technology. The Foveon can't physically match the spatial resolution of the standard CFA sensors, but color resolution is better. 4.7mp on a Foveon won't match the resolving power of a 6 or 8 or 10mp camera with the typical Bayer array for example, it simply isn't sampling the scene enough to do so, but from what I've seen the results are very smooth and appealing so the lack of spatial resolution isn't a big hit against it.

[troyammons]

> Carl Smith, Mar 15, 2007; 10:21 p.m. 4.7mp on a Foveon won't match the resolving power of a 6 or 8 or 10mp camera with the typical Bayer array for example, it simply isn't sampling the scene enough to do so, but from what I've seen the results are very smooth and appealing so the lack of spatial resolution isn't a big hit against it.

 

Before you say that you should shoot one next to a D200 or a 5d. Its very close.

[dave_redmann]

<I>To form one one [sic] color patch in a Bayer sensor they take the output from four sensors.</I><P>

 

No. A Bayer-sensor camera uses a de-mosaicing algorithm to turn four sensor pixels (R-G-B-G) into four full-color pixels. To grossly oversimplify: The camera has a blue pixel. Now it needs to figure out how to turn that pixel into full color, i.e., it needs to figure out the red and green values for that pixel. The data available includes: the pixel's blue value; the red values from four of the eight adjacent pixels, which are red pixels; and the green values from four of the eight adjacent pixels, which are green. Obviously, from all of this information, the camera can make a pretty good estimate of what are the red and green values for the blue pixel. It's exactly the same for red pixels. Green is a tad different: only two of the adjacent eight pixels are red and two are blue, so the camera's estimate of the true red and blue values of the pixel at issue are not as good. But the net effect of all of this estimating (de-mosaicing) and the other technical aspects of the sensor design is that a Bayer sensor's linear resolution is about 80% of what it's pixel count suggests. That's why a 6 MP Bayer-sensor DSLR with 2000 vertical pixels is typically tested out at 1600 vertical lines of resolution. Total resolution is more-or-less vertical linear resolution times horizontal linear resolution. As 80% x 80% is 64%, a Bayer sensor delivers a real resolution about 64% of what its pixel count might imply.<P>

 

Now a Foveon sensor absolutely does use 3 'photo detectors' or whatever Sigma is calling them now to make one pixel. The R, G, and B are right on top of each other. That's why a Foveon sensor with 14,000,000 photo detector produces an image with about 4,700,000 pixels.<P>

 

So a Foveon with 4.7 MP gives you about the same resolution as a Bayer sensor with about 7.5 MP. Don't believe me--look up the instrumented test reports.

[troyammons]

I shot a SD9 next to numorous cameras S2, 10d, 20d, 30d, 5d, 14n, slr/c, and a bunch of film cameras etc and you numbers are off. The sd9/10 is about equal to a 20d or 7-8mp. The SD14 more depending.

 

Things are picking up though in the bayer department, and I have seen some very sharp bayer images lately so the gap may be giving way.

 

Still it does not matter so much they are all just tools, with built in obsolescence. Personally I still like film a lot.

[ned1]

OK, this is becoming a bit of a p**sing contest, but if DP review says the D200 has a 7 stop latitude they are simply out of their minds. 4 stops is the limit for ordinary CCD sensors. CMOS is 5. I've actually tested it. Not hard if you use a light meter, a white wall and a gray card. Yes, the S5 really is 6 stops and the D200 came out about 4 1/2.

[chema_perez]

Edward - I supose the light temperature is also important for those tests. All of us can't see the same number of greys. Besides this dpreview tests doesn't mean one can have lots of information on the extremes of that latitude, BUT if one can see the difference between each grey then the latitude is there. If you take a quick look to the test hen you could say the laitude is about 5 stops but if I take the time you can see more than that.

[ned1]

Latitude isn't about perception. Each stop means twice the amount of light as the previous one. When you reach the point where there is no longer any data then you have maxed out on the high end. When you have mostly noise then you have maxed out on the low end. Back in the film days the results could be calibrated by using a light meter on the subjects and a densitometer on the printed results. Now you can skip to second step by simply inspection of the data file.

 

Does anyone have a link to the DPReview article? From the discussion here it sounds like the didn't do a true latitude test, which involves taking many different calibrated photographs at exactly controlled light levels. Although used for a different purpose you can find a film-centric description of how it's done in Ansel Adam's "The Negative". They're a bit tedious to do, but once I see the article I'm going to do a full test on my D200 and S5.