05-15-2015, 07:15 PM
I think the short answer is... it is complicated! I had to look it up as I've not looked at it before in detail. So there are two sets of 4 pixel shifts. Each set of 4 overcomes the biggest flaw of bayer pattern and you get individual RGB sampling at each location.
The second set of 4 is where it gets more complicated. These are diagonally centred between the first 4, so the question then is how much benefit does that give? It wont quite give a complete doubling on X and Y axis, but if you rotate 45 degrees, you could argue it doubles the sampling rate on diagonals.
This is likely further complicated by the usable area of each detection site. We imagine these as being perfect squares, but chances are they're not. We generally like them as big as possible for maximum light collection potential, but this half-shift would result in significant overlaps and you wont get the full benefit. If they were smaller it could help with resolving, at the cost of less sensitivity.
I think we're in a similar situation to when Foveon first came out. Just how do you compare? In reality there probably isn't a single number improvement that represents all situations, but there is some improvement regardless.
The second set of 4 is where it gets more complicated. These are diagonally centred between the first 4, so the question then is how much benefit does that give? It wont quite give a complete doubling on X and Y axis, but if you rotate 45 degrees, you could argue it doubles the sampling rate on diagonals.
This is likely further complicated by the usable area of each detection site. We imagine these as being perfect squares, but chances are they're not. We generally like them as big as possible for maximum light collection potential, but this half-shift would result in significant overlaps and you wont get the full benefit. If they were smaller it could help with resolving, at the cost of less sensitivity.
I think we're in a similar situation to when Foveon first came out. Just how do you compare? In reality there probably isn't a single number improvement that represents all situations, but there is some improvement regardless.
<a class="bbc_url" href="http://snowporing.deviantart.com/">dA</a> Canon 7D2, 7D, 5D2, 600D, 450D, 300D IR modified, 1D, EF-S 10-18, 15-85, EF 35/2, 85/1.8, 135/2, 70-300L, 100-400L, MP-E65, Zeiss 2/50, Sigma 150 macro, 120-300/2.8, Samyang 8mm fisheye, Olympus E-P1, Panasonic 20/1.7, Sony HX9V, Fuji X100.