New Sensor Tech in EOS 7D Mark II [CR2]

[Krob78]

Really exciting at this point anything is a bonus! I will be glad to see Canon innovating again!

+1 I'm having a good feeling that Canon won't let us down with the 7D MkII! ;D

 

[Lee Jay]

Putting a few things together into PURE speculation...

It was mentioned before that the viewfinder would be pretty big - as big as the 5DIII. I pointed out that would make it pretty dim.

What if the "revolutionary sensor technology" is quad-pixel, for a cross-type AF sensor under every pixel, and that it works so well with a new processor that there's no need for a separate PDAF module in the bottom of the mirror box? That would enable the main mirror to be "fully silvered" instead of "partially silvered" which would mean a brighter viewfinder even at the same size.

I kind of doubt it, but it's a bit fun to speculate.

Would that not require the mirror to be up to AF? gogo EVF

I guess it would, so it would still need to be partially silvered to retain AF and not have nothing but a stupid EVF.

 

[justawriter]

Could this mean the introduction of the long rumored multi-imp iconograph? ;D
http://wiki.lspace.org/mediawiki/index.php/Iconograph

 

[bdunbar79]

Part of me wouldn't be surprised if this is the high MP body.

Unfortunately I wouldn't either.

 

[NancyP]

This sounds as if they have completed a new fabrication process, which would be good news. Hey, after at least two years of speculation, why not?

 

[dolina]

Announcement for the 7D replacement should fall on August or September.

 

[x-vision]

Assuming it's a Bayer sensor with multiple pixels under each microlens (like the 70D), there's not a lot they can do to improve sensor performance that's outside the realm of read noise. There are several ways to attack that one, and some of them involve doing clever things with the multiple pixels per microlens, such as reading out each one at a different ISO and then combining them, sort of like what Magic Lantern has done to increase DR.

Either that - or, it might not be a Bayer sensor in the first place.

By the look of things, the so called dual-pixel tech is actually quad-pixel already.
See my previous post on the topic here.

With a quad-pixel design, rather than having a single color filter per pixel, it's theoretically possible to have individual color filters for each of the four sub-pixels.
These color filters don't need to be monochromatic R/G/B filters anymore.
Instead, these could be a combination of di/poly-chromatic filters, from which the full color of a pixel can be derived.
That's better than a Bayer sensor, where two of the pixel colors need to be interpolated from neighboring pixels.

So, you never know. The 7DII could have the first non-Bayer sensor in a DSLR.
If they use a combination of dichromatic filters for each sub-pixel, they could achieve maybe 1 stop of ISO improvement vs a Bayer sensor.
I think Canon will inevitably implement this sooner or later, given that they have gone the quad-pixel route already.
The question is, will the 7DII be the first camera to have it - or will we have to wait more for that.

 

[Marsu42]

This may be one of Canon’s best kept secrets as it’s apparently going to be more than an “evolutionary” technology.

Reminds me of the newly developed sensor for the 6d (inc. the newly-developed 11-point af system )... which is a certainly nice, but it's all the same general sensor generation for a long time.

My bet: They won't release a "revolutionary" iq technology in a crop camera, but would target the high-end ff market first. Much more likely it's in the direction of on-sensor af, evf/ovf hybrid and video-stills combination for ultra-high fps.

 

[Lee Jay]

Assuming it's a Bayer sensor with multiple pixels under each microlens (like the 70D), there's not a lot they can do to improve sensor performance that's outside the realm of read noise. There are several ways to attack that one, and some of them involve doing clever things with the multiple pixels per microlens, such as reading out each one at a different ISO and then combining them, sort of like what Magic Lantern has done to increase DR.

Either that - or, it might not be a Bayer sensor in the first place.

By the look of things, the so called dual-pixel tech is actually quad-pixel already.
See my previous post on the topic here.

With a quad-pixel design, rather than having a single color filter per pixel, it's theoretically possible to have individual color filters for each of the four sub-pixels.
These color filters don't need to be monochromatic R/G/B filters anymore.
Instead, these could be a combination of di/poly-chromatic filters, from which the full color of a pixel can be derived.
That's better than a Bayer sensor, where two of the pixel colors need to be interpolated from neighboring pixels.

Wouldn't you be worried about that approach messing up the phase detection?

 

[x-vision]

Wouldn't you be worried about that approach messing up the phase detection?

Yes ;D.

 

[Don Haines]

I have wondered for a long time why the bayer filter has survived. Instead of breaking up a pixel into 4 squares, G-G-R-B, why can't it be made into 3 rectangles of R-G-B where the colours no go from 25 percent of the area to 33 percent of the area and you gain about a third of a stop?

Now insert DPAF and imagine each pixel made up of 6 subpixels, 3DPAF pairs, and alternate the orientation of adjacent pixels so you can do DPAF in both the vertical and the horizontal plane.

that would certainly be a logical growth from what is in the 70D....

 

[Lee Jay]

I have wondered for a long time why the bayer filter has survived. Instead of breaking up a pixel into 4 squares, G-G-R-B, why can't it be made into 3 rectangles of R-G-B where the colours no go from 25 percent of the area to 33 percent of the area and you gain about a third of a stop?

Now imagine each pixel made up of 6 subpixels, 3DPAF pairs, and alternate the orientation of adjacent pixels so you can do DPAF in both the vertical and the horizontal plane.

that would certainly be a logical growth from what is in the 70D....

I think if you have quad, each pixel can do horizontal, vertical and diagonal in each direction phase measurements. So, no need for six.

 

[Loswr]

Either that - or, it might not be a Bayer sensor in the first place.

By the look of things, the so called dual-pixel tech is actually quad-pixel already.
See my previous post on the topic here.

You mean your previous post that was bogus and immediately discredited, because your conclusion was based on erroneous interpretation? :

 

[Don Haines]

I have wondered for a long time why the bayer filter has survived. Instead of breaking up a pixel into 4 squares, G-G-R-B, why can't it be made into 3 rectangles of R-G-B where the colours no go from 25 percent of the area to 33 percent of the area and you gain about a third of a stop?

Now imagine each pixel made up of 6 subpixels, 3DPAF pairs, and alternate the orientation of adjacent pixels so you can do DPAF in both the vertical and the horizontal plane.

that would certainly be a logical growth from what is in the 70D....

I think if you have quad, each pixel can do horizontal, vertical and diagonal in each direction phase measurements. So, no need for six.

The idea of six subpixels is to have pairs for DPAF, and the three pairs gets rid of the current bayer sensor with half the real estate devoted to green. This improves the sensitivity of red and blue by a third of a stop.

I did a quick drawing of the idea....

 

[x-vision]

I did a quick drawing of the idea....

The thing about this arrangement is that it uses single-color (monochromatic) filters.
So, you are still 'throwing away' 2/3rds of the incident light.
The trick would be to use more transmissive filters (say R+G, R+B, G+B) and thus throw away less than 2/3rds of light.
Your arrangement does improve resolution, though.

 

[x-vision]

You mean your previous post that was bogus and immediately discredited, because your conclusion was based on erroneous interpretation? :

It's all a matter of interpretation, I guess 8).
The arguments against my previous post were extremely weak.

 

[gsealy]

It seems to me that we have some general trends and concepts going on:
1.) Dual pixel technology (and the like) is going to be spread throughout the product line, it is a significant differentiator compared to competitors.
2.) Canon is adding value to each product through firmware updates, thus giving them a longer life cycle.
3.) Video is becoming more pervasive, I would expect the 7DII to have at least a clean HDMI out for external recording.
4.) 4K is in the future and while the 7DII might not have it now, perhaps the sensor will support it for a future upgrade done at a service center.
5.) Low noise at low light situations is a must.

I am not an expert on these things - but Canon has to be more than competitive. This is a major release in their product line, so they have to make a statement.

 

[rrcphoto]

Putting a few things together into PURE speculation...

It was mentioned before that the viewfinder would be pretty big - as big as the 5DIII. I pointed out that would make it pretty dim.

Not really . the difference would be negligible over the current 7D viewfinder. I didn't see people stating that the 50D viewfinder was far brighter than than 7D - which would be about the same ratio.

 

[Don Haines]

I did a quick drawing of the idea....

The thing about this arrangement is that it uses single-color (monochromatic) filters.
So, you are still 'throwing away' 2/3rds of the incident light.
The trick would be to use more transmissive filters (say R+G, R+B, G+B) and thus throw away less than 2/3rds of light.
Your arrangement does improve resolution, though.

interesting.... so you think that it could be done with the more transmissive filters? the combination (if it works) could get to only throwing away 1/3 of the light, as opposed to the current bayer which throws away 3/4 of the light.....

 

[pierlux]

I think CR guy knows more than he's reporting. C'mon Craig, tell us something more!