Photoclubalpha

I heard reports somewhere (might have been the dpr review) that the AF in the 70D works good with some Canon AF motors but not so well with another, I think the USM was one that wasn’t so good and the STM type was good, I could have remembered that wrong though because there are two types of USM, ‘ring’ and ‘micromotor’ so it might have been just one of the USM types that was a bit slower than the other.
DXO gives the 70D sensor a lower overall score than the A77 sensor. (for what's it's worth)
Greg

Only tried the 18-135mm STM and the 50mm f1.4 USM
Both worked well, the STM is basically silent, don't like the focus by wire though..optics are decent though the Sony equivalent has more bite to it IMO.
I think it's one of their better "more rounded" cameras for stills and sure to grab some video users too (the video aspect works really well being dead honest, they trumped Sony here big time)

Not saying I'd rush out and get one (price is a tad high too) But I can see it doing pretty well for Canon
Sensor wise it's basically the same as the other sensors just a few more mp which is pretty irrelevant real world.

Still they beefed it up more than I expected.

Greg Beetham wrote:Yes Agorabasta I remember, but the divided pixels in the Canon are different to your design, the result AF wise might be similar but the Canon design wouldn’t have alternate one colour column layout, but even with that new method the jury is still out on how much light loss there actually is.

The particular design is irrelevant to the principle of on-sensor PDAF function where every sensel is divided in two halves, one looking more to one side, another - to the opposite. And such a design principle does improve the light-gathering ability at at the most oblique angles.

Greg Beetham wrote:The 70D doesn’t even score as well as the A77 overall and that camera is no world beater these days and wasn’t back in its day either.

Comparing like to the likes, that Canon sensor is better than all the earlier ones while sharing the same century-old manufacturing process.

Greg Beetham wrote:I remember seeing a diagram at one time of the Sony proposed on sensor PDAF and every second row was taken up with focus diodes (probably not the full height admittedly) which would amount to a large % image loss. The image is then re-constructed by data value association of adjacent pixels, for stills, but probably not for video. (or so I read)
I have no idea if that is the actual current or future Sony layout of course, they could have ended up doing something different.
Greg

That's obviously not the case, but the Sony variant of implementation is actually the worst of all possible as they take rather long stripes of sensels in the row right where it matters the most. Yet it takes a lot of special arrangements to even notice the negative effects by the beat freqs in the image.

Greg Beetham wrote:Ps As far as the 100 thousand AF pixels go for AF on the Oly, do you have a link? 100k is 100k in my book, it’s a lot more than none. My three DSLR’s don’t have any on sensor AF or exposure diodes.

About every earlier posted preview, including the dpr's, mentioned that the PDAF area on the sensor uses every 16th of the sensels for PDAF (or every 8th of green pixels), which makes for 1Mp for the full sensor or about 250-300kp for the area used.
So it's not that 100kp I so recklessly put up there, but more like an equivalent of 1Mp in the centre area.

But that's only 0.085 of a stop total or 0.17 stop in the green (luma) channel, so colour info/noise is not affected.