« on: September 11, 2014, 05:21:59 PM »
DPAF uses every pixel covering 80% of the image sensor, no image data are lost, no interpolation is required, and on that 20 MP sensor you have 16 million AF points.
That's actually not true, and I'm pretty sure you know that. You need multiple DPAF pixels to be able to properly detect a phase differential, just like the strips in a dedicated PDAF sensor are made up of multiple pixels in strait or diagonal lines.
You don't actually have 16 million AF points.
Bear with me for a second. On a standard PDAF unit, there are discreet lines of pixels making up individual AF points. On DPAF (lovely similarity between acronyms), there aren't discrete lines of pixels, there is a grid of pixels.
Say it takes 3 adjacent pixels (random number, I'm sure it's far more) to constitute a line. If I have 10 adjacent pixels, I have 8 lines. My number of lines is limited by the borders.
Scale that up.
Given the total 20-ish megapixels and however many are necessary to constitute functioning lines, is it impossible that there are 16 million locations in which phase can be detected?
canon is not about invention. these times are long gone.
To that point... out of curiosity, if Canon followed Nikon's lead and started purchasing sensors from a third party vendor and instantly made up the low-iso-DR and resolution gaps, would that constitute innovation/invention?