Sigma-enigma?

[Jorge]

It might be premature to start this discussion, but I just laid eyes on some sample photos taken with the soon to be released Sigma SD9. In the terms of sharpness this camera seems to outperform the canon D30/D60. This is due to the Faveon sensor where every pixel can in fact see all tree colours instead of the CMOS sensor where each pixel handles separate colours. The price of the sigma-enigma is supposed to be about just one third the price of a D60 (approximately 700 US$).

Actually I’m not sure that I like all that sharpness as it kind of draws the attention towards details that are not normally observed and thereby moving focus away from the overall composition. E.g. tiny details of skin structures in a close-up portrait are so highly visible that you almost feel like one of those invisible mites that walks around the human skin and that you can find blown up to dinosaur-size in textbooks of biology.

But surely this kind of sharpness could be put into good effect in all sorts of situations where a high amount of detail is wanted and the accustomed smooth style of canon digital pioneer cameras could soon go out of fashion???

Anyway the price alone would seem to show canon the back heel especially for those people who don’t drag a large pile of expensive and incompatible glass around – time will show. At least one can hope that canon and other manufacturers will lower the exorbitantly high prices and gear up for the new competition!

For those who haven't allready been there take a look at Erhardt’s photo-page:

http://www.foto-erhardt.de/modules.php?name=Testbilder&d_op=viewdownload&cid=67

[Roger_Cavanagh]

Jorge,

The Sigma SD7 is certainly not going to be one-third the price of a D60. The MSRP is USD1800.

I downloaded some of the sample images and I certainly wasn't hugely impressed with them - may be because all the hype raised my expectations. Some people have written (I know you didn't in this post) about the 3.43mp Foveon being equivalent to a 10.5mp, which is nonsense. There is more colour data collected, but certainly not 3 times as much image information.

I'm sure this technology will improve, but right now my experience with Sigma lenses doesn't encourage me to rush out and buy a camera that will only take Sigma lenses.

Regards,

[Jorge]

Ooops! You're right there. I just checked the expected price one more time: 1800 US$.

There's not much of a difference from an economical piont of wiev then. As for the quality of Sigma lenses I haven't got a clue.

[bigdave]

Just a question Roger, why do you find it so absurd that an image sensor could gather 3 times the amount of image information? I know there are very adequate descriptions of how foveon technology works, but maybe I'm missing something. I mean here's my reasoning:

Say you have 3 dots, they can only pick up red, green, or blue, respectively. Now say you modify each dot so that it can pick up all three colors by itself. The light gathered just went from 3 (1 per dot) pixels, to 9 (3 per dot). Isn't that three times the color information? In adition, the technology removes the need for interpolation, meaning even truer colors and sharper lines. If you have the time please explain what I've failed to consider.

[Jorge]

Bigdave might have a point there. I’m not familiar with the technical details concerning the Faveon processor, but consider this:

A full colour RGB file normally contains 24 bits of information per pixel; 8 bits for each of the three colours which add up to some 16.777.216 possible colour combinations.

If each pixel of the CMOS sensor stores only 8 bits of information (one colour) and the Faveon sensor stores 24 bits of information (3 colours) per pixel logically the Faveon must store three times as much information as the CMOS.

This does however not necessarily imply that a 3mp Faveon equals a 9mp CMOS since it’s not the same kind of information that is stored. Three pixels in a CMOS will theoretically measure different rays of light (or bundles of rays) while one pixel in the Faveon will measure the same ray (or bundles of rays). In theory the 3 pixels of the CMOS measure three times as many rays as one pixel in the Faveon, but each with 1/3 of the detail. In the end the amount of information from the 3 CMOS pixels will be the same as for the 1 Faveon pixel, but the values of the informations will be different! I’m not sure of the practical impact of this in terms of photo quality, but I recon the calculation of the data will have a great say here.

[Roger_Cavanagh]

BigDave wrote:
here's my reasoning:

Say you have 3 dots, they can only pick up red, green, or blue, respectively. Now say you modify each dot so that it can pick up all three colors by itself. The light gathered just went from 3 (1 per dot) pixels, to 9 (3 per dot). Isn't that three times the color information? In adition, the technology removes the need for interpolation, meaning even truer colors and sharper lines. If you have the time please explain what I've failed to consider.

My objection to saying this is the equivalent of 10.5mp Bayer-type chip...

I am making a distinction between data and information; more data does not always yield more information.

Yes, each pixel does capture separate values for RGB, that is 3 times the DATA, but the interpolation of the Bayer data does give us a full colour image. Yes, there will sometimes be errors, but the information value is to coin a phrase "greater than the sum of the parts". So the advantage of the Foveon chip is reduced to the errors caused by interpolation. I have no idea how to calculate that. :)

That's for chips with the same number of pixels. If we did have a 10.5mp chip, there would be a lot more resolution, so there would be more detail information in the captured image.

Jorge was not saying that the a 3.5 mp Foveon was as good as a 10.5 mp CMOS, or even suggesting it, but I _have_ seen that statement and his post just gave me the opportunity to vent a little. :)

Does that make sense?

Regards,