From what I can tell based on Canon's current patent, they may not actually be producing a "layered" sensor as much as they are taking a new approach towards utilizing the photodiode area in each pixel...some of it is allocated to red, some to green, some to blue, utilizing silicon itself to filter the light (same as Foveon), allowing a "single pixel" to detect all three colors. I am not sure what the real pros/cons of that might be, or if thats actually what they will end up doing in the long run. It seems more like a modified CFA approach than a true layered design like Foveon, and it might resolve one of the key issues with Foveon's design...differentiating how many electrons in a given photodiode are to be allocated to blue, green, and red readouts.
Thanks! That's useful! Do you have any info on to what extent photosite size would affect the DR and ISO performance? with the gapless microlens tech, and some 15.3 MP on the sensor the photosites would be fairly large, I imagine, but I wonder how that would translate into DR and ISO performance on a layered sensor...
Well, we are assuming Canon would produce a sensor with that low of a resolution. They might...it might be their favorite size...18mp. It may be more dense. Either way....DR and noise will be key issues for any layered sensor, done Foveon-style or some other way. Foveon sensors don't fare well at high ISO, namely because green (the most prevalent band of light) and red are only registered deeper into the photodiode. Foveon senses blue at the top, green in the middle, and red at the bottom. This generally should balance things out, as silicon is more sensitive to red light than green, and more to green than blue. But there is also a lot more row/column activate and readout wiring (as well as some other control logic) per-pixel in a layered sensor than in a bayer sensor. That means there are more obstructions, so more light gets reflected or absorbed as heat at the green and red levels, reducing light further. Since this is INSIDE the photodiode, there isn't really any way you could use microlenses to mitigate the problem. A really advanced design might actually stack separate photodiodes, sandwiching microlens layers. I imagine that would be FAR more difficult to manufacture and very cost prohibitive.
However Canon does it (assuming they even ever do), you'll have to make some tradeoffs. DR for color fidelity and resolution (on a pixel-normal basis anyway.)