Upload & Sell: Off
| p.5 #16 · Sony A7R sensor gets high scores from DxOMark |
sure, i'm just making an educated guess that a giant corporation isn't going to spend money they don't need to. i believe you were recently making statements based on your intuitive feeling about lens design, not sure why i shouldn't be allowed to make such guesses.
i'm not claiming an AA filter has no effect on color, i'm claiming it isn't part of the color filtering array. removing materials between reality and the sensor measuring them (color in this case) should intuitively () improve the data going to the sensor in they do anything. have you ever looked through a glass...Show more →
I won't comment individually but some general comments.
1. About AA filter, if AA filter have some kind color characteristic, this has to be considered in design of CFA to compensate it. Simply remove it will not make it. So window example is not really applicable before known exact the effect.
2. From D800E and D800 score, 77 total, RGB 2.02 1 1.49 vs 78 total RGB 2.04 1 1.48. you can made a reasonable guess how repeatability/reliability of this test can be. Now, you can see without remove AA physically but cancel it optically seems doesn't impact the score. Obviously, the AA do has some impact on this test or color itself. I don't want make any claim out of it more than it looks like. but 2.02 vs 2.04, 1.49 vs 1.48, this is kind tolerence more than I expect on this test. so I question your sample variation claim, but I certain can't prove you wrong based on that.
3. as engineer background, if I want design something meaningful, the test window (guard band) has to be reasonable small to make test result valid. I assume the same apply here. hence my difference is difference comment. 6 out 80 is not small, if random error, repeatability is this big, apply industry standard 6 sigma analysis, this is not very meaningful test at all.
However, I never disagree how individually should interpolate this number. 6 point won't tell me anything that which has better or worse color, or pleasing color but before we trash this industry standard test validation(again, here I assumed), Can't I assume this imply I need put more effort in my profile to get color match human eyes?
We may not argue anything at all other than my initial comment about color difference of RX1 and RX1R. again this seems a dead end!
But discuss has been very health so far