Discovering the SD15...

mike earussi wrote:

They're different nonetheless: http://www.dpreview.com/reviews/sigmasd10

and if you ever get a chance to play with one you might find you like the color.

OK - the main problem here is that it is so many variables. Sigma have experimented with almost every generation of their cameras. So - it is almost impossible to test it all and remember exactly how it was.

I remember the old and funky colors. My memory tells me it was SD9 and SD10. It was blue and yellow and also sometimes cyan. But, maybe you are right, and it was mainly SD9. I also remember the discussions about 7 lines that become 5 in a test chart, and if smooth or textured lawns were best when grass leaves were to small to be resolved and the pine needles.

Then I remember SD14. That was the prime time for magenta and green splotches. I also remember it as a time for desaturation of shadows and high ISO. Was that not so? I am rather sure. OK - but maybe the desaturated green when photographing brightly colored flowers was later. Maybe. Although I saw one in this very thread. But - maybe it was mainly the TRUE engine.

Then came the series of DP1 and DP2 variants. Green hue anyone? And green corners? And red grid? Evolutionary steps made them better and better, except when it sometimes got worse

And now, lets see ... those red cars that became pink ... was that not SD14? Yes, I think so. Or was it DP cameras? No ... I think it was SD14. And also red flowers that become yellow and/or orange and/or white.

Then came SD15 and SD1. SD15 was said to have the best colors so far. But ... not many images were posted on the forum really. So, I am not sure how good or bad it is. And the SD1 did cost $10,000 or so and only a hand full enthusiasts in the forum got one.

Then, after some agony ... came the SD1M and the DPnM cameras. The best colors so far (except maybe for SD10), but a bit harsh regarding fine texture.

The same thing happened with SD1M as for SD15, Not many images were posted in the forum. Some of the images posted rendered red as orange. That was never investigated. Moreover, it was said that SD1M did not handle legacy glass all that well with colored corners. Nothing that was demonstrated with examplas as far as I remember.

The DPnM cameras got popular though, and we have lots of nice images taken with those. Their exaggeration of micro contrast is often nice but sometimes makes it look too harsh IMHO. The Merrill cameras are superior monochrome cameras! The best so far. The enhanced micro contrast is a great plus then.

Then came the Quattro as a surprise to us all. There Sigma did abandon the idea with three detectors per pixel. Still three layers per pixel. The color is as good as the Merrill and the rendering is not as harsh. Some do not like that. Personally I do, even though I can see the use of Merrill. The resolution is slightly better than Merrill, but some claim that there is slight fuzziness at pixel level. There is also a noise, that some call sand.

The green and magenta splotches still plague Merrill and Quattro sometimes. The reason for this is quite simple - the green channel (after conversion) is very noisy. Strong noise reduction (before converting) is needed. And noise reduction might add a bias. If it does, then it will be green if the bias is positive and magenta if it is negative. And if it varies you get magenta/green splotches.

So - why is the green channel noisy? That is also quite simple. The top layer is desaturated blue. The bottom layer is desaturated red. The middle layer is very stronly desaturated green, nearly grey. So - to extract green you take middle layer minus top layer and also minus bottom layer. Can you visualise that the green channel almost only is noise?

BTW - another interesting feature is the extra clipping pixels in Quattro. A sparse pattern of pixels with lower sensitivity can be found in the Quattro sensor. Those are (probably?) used for compensating for over exposed areas. At least Foveon has such a patent.

I think I have seen the effect of those detectors in an image. Among the trees the sky is shining through - and all is out of focus - and therefore soft. In certain areas the signal clips. All thre channels clips eventually. Still the clipped areas are of a reasonable color and luminosity. Then, it clips to white with a narrow grey edge around the white patches.

Is that good?

Maybe. But the grey edges looked weird. And I assume it would not have worked so well if the area was sharp.