What would the FFF have to do in order to be competative?

xpatUSA wrote:

tarmov wrote:

xpatUSA wrote:

tarmov wrote:

abera wrote:

What I think would be an interesting version would be a chip with four layers with the bottom one being a very thick capturing mainly IR. It could be made perform quite a bit better in IR than current conventional cameras even when IR converted (as the photodiodes don't go deep enough to the silicon).

Yes, that might perhaps even improve the differentiation of 3 primary colors, wouldn't it?

Not if the layer depths remain unchanged, as they have been since the very beginning ...

I thought the color separation diagram and the 3x3 layer data transformation matrix suggest that additional layers would also improve the accuracy of 3 primary colors, due to the probabilistic nature of the process.

I'm thinking 4x3 if we're talking about the accuracy of 3 primary colors ...

A 4x4 transformation matrix would improve all 4 "colors" (the 4th being infrared).

Hmmm ...

Maybe I have misunderstood something.

Maybe. What is the source or basis for that statement ?

-Tarmo

The spectral diagram and the color (transformation) matrix show that the absorption of primary colors is spread out and that all 3 layers contribute to the values of all 3 primary colors. Without that there would be nonzero values only on the main diagonal cells, but the Foveon color matrix shows large nonzero values outside of the main diagonal.

The additional infrared layer would not just add a column and row to the 3x3 color transformation matrix, but it would also change the transformation values within the initial set of 3x3 (sub-)matrix. The values recorded in the infrared layer aid to separate red from green and red from blue. And part of that better separation is also due to the fact that infrared light itself also absorbs in all 4 layers (at different probabilities) and with 4 layers the share of infrared light in all those layers can be better constrained, which also helps to better constrain the values of 3 primary colors.

And one could also add an ultraviolet layer, to better separate blue from green and blue from red. One could even add 2 infrared layers: near infrared and far infrared. And 2 ultraviolet layers.

One could also add intermediate layers to the intersection of the blue and green spectral density region and to the intersection of green and red spectral density region. Perhaps those intersection layers do not really contribute to the separation of colors, thus the absorption there could be disregarded (not even read out, just let it get absorbed there).

-Tarmo