Hi Paul,
The way I see it, the main advantages with Foveon is that it maintans resolution. Doing the colour interpretation always looses some sharpness. The Foveon is also in less need of an AA filter as it will not show colour moiré.
The Foveon has no CFA, so colour separation is based on absorption depth in silicon. That makes the design quite tricky and it will need strong colour separation in math, which increases noise. That is probably the reason it doesn't work well at high ISOs.
I would think that the new CFA design on the Thrichromatic is more dense and may be more ortogonal. Human vision has a lot of overlap, especially between L and M. So brain does a lot of colour separation work. CFA design is a compromise, it may be that P1 gives up a bit of high ISO performance for better colour separation in some areas.
The IQ3-100MP sensor has excellent reseves for high ISO.
It is a bit surprising that P1 pulls 15 EV DR (per pixel) from the pixel size. That could be a result of some smart trick. Nikon seems to play some trick with the D810, it seems to have larger FWC (Full Well Capacity) than other contemporary sensors.
Colour science has two parts, one is the CFA design, the other one is the colour processing pipeline. A part that we should not underestimate is detection of white balance. WB plays a major role in getting good colour.
Another important factor is IR cut off filtering. Landscape colours contain a lot of IR, as clorophyll does reflect very strong in short infrared. ColorChecker foliage patches don’t have a lot of IR content though.
A stronger IR filter may be beneficial for colour reproduction, just as an example.
So, there are many things P1 could have improved, or at least made differently.
Best regards
Erik
Foveon, as I understand it, uses 3 separate photodiode layers, red, green, and blue thus there is no interpolation for color. The standard Bayer pattern captures 1 color per photodiode, then interpolation is done to figure out the other colors. So if Red is capture, the camera will interpolate for blue and green. Downsides are that the Foveon layers are stacked, thus light transmission is limited and so the sensor so far tend to work best in the base to lower ISO ranges. This as far as I know has never been upgraded or fixed. There is no doubt to me that the Foveon chips can take amazing images, however the raw support for Foveon/Sigma is still very very limited.
This new tech from P1, as I can understand it from the limited amount of literature, still uses only one photodiode layer which appears to be the same as the current IQ3100. The difference is the color filter array which seems to allow for more precise color captures and no overlapping which can cause cross talk.
The need for interpolation still would be there as far as I can see since the P1 design is still only using 1 layer of photodiodes not 3 as Foveon does. Only 1 of the 3 basic colors is being captured, but that color per the claims of the marketing literature is supposedly more accurate.
It will be interesting to see some side by side shots from the existing IQ3100 and this new back.
Since the base ISO has been lowered to 35, I am also curious if this back will be more like a CCD back, in that it will best perform at base ISO to maybe 2 stops higher due to the difference in light transmission.
Paul Caldwell