I'm not an expert but a quick look, the patent says that the green conversion layer is an organic film sandwiched between a couple of ITO layers (Indium Tin Oxide which is a conductive transparent material). This is on top of pixel. The R and B are embedded in the silicon the way foveon is. The other difference is that this is back illuminated which I don't think the Foveon is.
It also looks like the internal doping scheme is quite a bit different than what I remember for the Foveon.
As Mike mentions the Foveon isn't back illuminated either. All in all, a much different design.
I find it interesting that they placed the green layer above, which definitely helps with color fidelity/separation of R and B layers, but requires some serious engineering to make sure the organic dye layer doesn't suck up all the blues before they hit the chip.
There's actually a HUGE difference between this, and the Foveon principle. The Foveon relies fully on the absorption properties of the substrate to differentiate wave-lengths (i.e - get colour). This makes the Foveon construction very dependant on after-capture processing, since the colour measurement straight from the sensor is pretty whacked up.
The main reason for this is the totally insufficient band-separation of R-G-B, or rather a total imcompatibility with the human S-M-L way of assessing colour. It also necessitates the use of light-intensity way out into IR and UV to get even close to a saturated result, and that screws up metamerisms in a a very noticeable way.
By adding a separate green-absorbtion layer on TOP of the substrate, you get a very different filter effect. You subtract the green bandwidth (centered on ~540nm) before you let light through to the depth-differentiated Foveon principle. This also makes the penetration depth differentiation a lot more strongly attenuated, you separate blue and red by a larger amount.
The negative with the Sony patent is that you still cannot make the green/orange separation overlap as the human eye does (and most good Bayer-based sensors do), so you still have a lot of metamerism failure her - smack in the middle of the very important skin-tone band. I'd expect very flat colour between warm red and warm green hues - flat as in "lack of hue discrimination", not as in "lack of saturation".
The positives about the patent is that Sony bypass the heavy post-processing by making the green band more than three times as narrow as the Foveon, and thereby get a much improved noise-characteristic. Foveon typically MEASURE a very good signal, but to get any usable colour you have to do a linear transform that amplifies colour by 3-5x, and this also amplifies noise by 3-5x. Sony also specify a totally different readout scheme, that minimizes the amount of circuitry that has to be present on Foveon-type sensor (that typically have less than 50% effective active area coverage - compared to 85-90% for a modern CMOS Bayer)
10X for the explanations, though I must admit I did not understand them fully (too thick I guess... ). I do however, wonder how the pics from such a sensor look like.
). I do however, wonder how the pics from such a sensor look like.