On a DSLR, it only takes that much because the other 70% is needed for the OVF. And then the mirror flips up and the sensor gets all of the light.
Eh, so you're assuming that 30% directed to PDAF module is an afterthought, leftover? I would say, it is by design. And don't forget yet another path in DSLRs: metering module.
The SLT steals the 30% all the time which is a disadvantage of it relative to the SLR although I suppose one could also design the mirror to flip up in an SLT.
What has this got to do with anything? There are compromises with EVERY design, one chooses that gives up something to gain something else, INCLUDING a mirrorless camera which is really the point of this discussion. There is no free lunch.
The on sensor PDAF is fundamentally different because, well, it's on the sensor.
Fundamentally different in the way light is being received for AF. You're assuming that it uses less light to do the same thing that DSLRs/SLTs need 30% of the light for. Perhaps you're assuming it is significantly less (less than 20% or a third stop?).
That's right. It only blocks some (all?) of the light for a small number of pixels. For SLT, you are forced to divert 30% of the light over the entire area.
In terms of conservation of information, yeah it makes sense you'd need the same amount of light to get the same performance. In the case of CDAF, you need all the light for AF, but it turns out you can re-use it for the image, so there's no harm done. It's possible that for on-sensor PDAF at least some of that light can be re-used for the image as well. I think that's the case for the Canon version of on-sensor PDAF.
So the real thing I'm trying to get at isn't how much the AF needs to do its work. The real question is, how much is available to make the image.
You can't have one or the other. You've to deal with both (there would be no creation of light either way).
The PDAF sensors could possibly be able to register light intensity as an aside.
But worst-case, even if they completely cover some of the pixels, they block significantly less than having to block/cover the entire sensor for the small number of points.
--
Gary W.
My take is that less light for PDAF would equal ineffective system in even marginally low light conditions. This is why 30% (half stop) of light is diverted in DSLRs and SLTs to PDAF module. A rumor has been that Sony is going to reduce it to 20% (a third stop) with new generation of SLTs but I surely hope not at the expense of AF speed and effectiveness in low light.
For a PDAF to work effectively, I would assume a reasonable amount of light is necessary, which lilely has affected previous hybrid sensors a bit. The a6000 sensor is not only better than those, but it has a lot more pixels in use which is also effective in smaller apertures (by 2+ stops) compared to NEX-5R/6 (which lost OSPDAF at f/6.3).
You're missing the point here. The point isn't how much the PDAF needs to do its job. The point is how much do you need to steal from the generation of the image for the sake of AF.
SLR's and SLT's divert 30% of the light, but they utilize only a tiny fraction of that total for AF. Most of it is wasted. In the case of the SLT, this waste occurs even when the picture is taken, so the image quality suffers. For a DSLR, the AF gets nothing when the picture is taken, so the image sensor gets 100% of the light for making the image.
When you put the PDAF on the image sensor, now you have the opportunity to not waste anything. It's also possible the AF sensors can be repurposed during the image exposure to capture image data.
I'm not saying this is how it works. I'm saying this is how it could work. But one thing I am sure of is decades old data on SLR AF systems is not useful for this discussion.
In a separate discussion, TrojMacReady says he actually read Sony patents and believes their PDAF does use 30% of the light in the vicinity of the AF sensor. But I still don't know if that 30% is also not available for taking a picture.
