Some explanations from da Boss :
CP+2014
This is a great video, someone translated and summarized the gist of it at dpreview.
I also took notice of the color responses of the layers, so the top layer can indeed be called a "type of blue" but it does register red and green luminance too. This means that the B&W won't be as pure neutral as I thought but on the other hand, color accuracy can increase a bit.
So more than reducing wholesale the whole chroma to 1/4th, it's more like Blue = 19.6 MP but red & green ( and I am using the color terms loosely, it's not exactly RGB) are 4.9 MP.
Because the top layer is 1/2 sensitive to green and 1/4th sensitive to red (approx.) there are some guesstimates that could be done to increase the resolution, in particular of red- but not in all situations, and is tricky. But it's better than nothing.
While B&W is not as good as I had thought if this was a pure or mostly pure luminance layer, (the top one), it should still be pretty good.
And given the noise the Merrill has in red in particular, vs the old design, this design may very well capture in many situations as much detail with that color.
The top layer is effectively (take out your favorite photoshop RGB color wheel) - 255 B , 128 G, 64 R = a type of cyan.
Also the Sigma CEO mentions they improved the sensitivity by 1 ISO stop but they are still working on it. A bit of a bummer I was counting on two, but maybe they can eek out 1.25-1.5.
Also in a refreshing act of candid honesty, he mentioned that "Foveon's strong suit is not low light, and probably will never be, so if you are shooting in low light you should probably be using a better tool (camera) for that" (paraphrased). I respect and applaud his candid honesty here.
Given my experiments with the Merril at ISO 1600, 3200 and 6400 in B&W, I am still hoping that the new sensor could do decent B&W at ISO 6400 and I still wonder how much gain it would be in color when using Low Raw (true X3 capture with a super sampled "blue layer").
- Ricardo