SIGMA releases an official update about their Foveon X3 sensor project

[melgross]

No, a 20mp Foveon sensor has the resolution of a 20mp Bayer array sensor with no AA filter, which looks like a 30mp Bayer array with an AA filter (depending on the strength of that AA filter) I never was overdoing it.

Not exactly. The Foveon doesn’t need the Bayer convolution to get the different color sensing sites to show the same area, because they’re beside each other. With the Foveon, each site is on top of the other, so that isn’t required. The Bayer convolution loses placement data from the individual sites as they are mixed together. The resolution in green is about one third higher than in the Foveon because Bayer sensors use two green sites, as we are more sensitive to green. But as a result it has about 40% more red and blu, resulting in an equal amount of detail across the colors, giving about a 30% higher overall sharpness.

most of the tiny amount of detail lose from the AA filter can be recovered by the initial sharpness adjustment made by the software.

the filter isn’t anywhere responsible for the difference, because, at most, for a medium to high rez sensor, no more than about 5% of detail is lost.

you misunderstand what’s going on here. We’re not talking about countable pixels. That remains the same. It’s the sharpness of the individual pixel that’s affected.

 

[melgross]

Not exactly. The Foveon doesn’t need the Bayer convolution to get the different color sensing sites to show the same area, because they’re beside each other. With the Foveon, each site is on top of the other, so that isn’t required. The Bayer convolution loses placement data from the individual sites as they are mixed together. The resolution in green is about one third higher than in the Foveon because Bayer sensors use two green sites, as we are more sensitive to green. But as a result it has about 40% more red and blu, resulting in an equal amount of detail across the colors, giving about a 30% higher overall sharpness.

most of the tiny amount of detail lose from the AA filter can be recovered by the initial sharpness adjustment made by the software.

the filter isn’t anywhere responsible for the difference, because, at most, for a medium to high rez sensor, no more than about 5% of detail is lost.

you misunderstand what’s going on here. We’re not talking about countable pixels. That remains the same. It’s the sharpness of the individual pixel that’s affected. The effect is called, apparent sharpness, or effective sharpness.

 

[melgross]

The fact that some of the 3 Foveon layers pick up light meant for the other layers doesn't have to be considered being "poisoned". If the layer sensors have low noise, then they should be able to have a software 3x3 matrix filter to get the proper R G B values out of them. To show an example of this, the human eye has a green cone and a greenish-yellow cone (instead of a green and red cone). These 2 cones have a QE curve that overlaps each other as much as it differs from each other, but the eye/brain has circuitry that separates the values into what we consider green & red. If there is low noise then circuitry can do the same to eliminate this issue. Now, I'm guessing that the Foveon pixel elements have a lower QE curve and higher noise levels than desired so that would be the reason (I assume) for their lack of success.

That’s exactly what poisoned means. It has nothing to do with noise, which is different.

please don’t compare a camera system with human sensoriums. They are very different.