This is apparently indeed the case with Sony’s implementation, but nothing would prevent a different implementation where the AF pixels are read at a higher speed than the imaging ones (a multiple obviously).
In fact I am wondering whether the Sony implementation is really the one you are describing for the PDAD based focusing phase. It may be the case if a final contrast based AF is needed. But this also is technological in nature.
In other words, I believe that a lot can still be invented here.
Cheers,
Bernard
Hi Benard,
What you said it could be very true... my speculation why Sony maybe does it this way, and this goes very very far away of what I could know about CMOS sensor technology and AF technology, so, take it with a very big grain of salt:
- For things like Eye AF or Face detection my speculation is that they are using information from more than the PDAF pixels. PDAF pixels give information about where to move the lens, rest of pixels information about what the camera is seeing.
- Not sure if the Sony tracking algorithms use this. But since they can read the whole image (or a big number of pixels), for what you can understand from Sony presentations, it will be interesting for Sony engineers to use all the info improve their algorithms.
I suspect that the ability or read quick an image in Sony sensor does not come from the need of better AF reading (it is a benefit to it...), I think their major objective was to reduce EVF lag and eliminate blackout (one of the major criticism for action photography that they had before the A9). Also, for video and action photography, this minimizes the effect of rolling shutter (curiously enough... Sony with the A9 did a Canon like move... not allowing to have Picture Profiles for video on it).
Now other thing, again, big grain of salt here, reading only the PDAF pixels will probably require specific control lines for those pixels in sensor design, this can complicate the design of the metal connections of the sensor chip. Although... after writing this, and thinking of the stacked sensor design of the A9 and A7 III chip, since the image is very quickly copied to a DRAM chip, access to the PDAF pixels only will be just addressing those memory address that it should be very quick...
As I said in this thread or in other one talking about mirrorless... I have a very big nerdy side
Regards,
David
