December 18, 2014, 02:55:58 PM

Author Topic: EOS 7D Mark II Talk [CR1]  (Read 29055 times)

Don Haines

• Canon EF 300mm f/2.8L IS II
• Posts: 3576
• Posting cat pictures on the internet since 1986
Re: EOS 7D Mark II Talk [CR1]
« Reply #75 on: October 10, 2013, 09:11:10 AM »
My guess - greater than 22.1 megapixels. That gives you 3x1920 pixels across, allowing for easy bining of a 3x3 pixel block from the sensor into a single pixel of 1920x1080 video... add a few pixels to each side for video stabilization and the 24MP sensor looks very likely.

How is binning at 3x3 on a grid that repeats at 2x2 easy?

Photosensors are in a 2x2 grid, but you're not binning a quad of photosensors; just the reds together, greens together (x2), and blues together...  So yes, it is easy.

In the attached image you bin the following:
r1+r1+r1+r1+r1+r1+r1+r1+r1=r1'
g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1=g₁1'
g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1=g₂1'
b1+b1+b1+b1+b1+b1+b1+b1=b1'
And you bin down right back into a bayer pattern which you can than save as-is, or demosaic, or whatever.

That method throws away half the light collected.

The 2x2 approach bins the 2 greens, 1 red and 1 blue in the block into 1 RGB pixel.  Simple and no light lost.

Rather than do the above, why not demosaic and interpolate with a simple interpolation method like bilinear?  That way, you'd at least keep all the light.

Where is there light being thrown away?

A normal pixel is 1 red, 1 blue, and 2 greens divided by 2... Let's say you have an 8 bit a/d converter.... that gives you a 24 bit pixel and you have 8 stops of dynamic range..

Bin 9 pixels together... (for ease of math I will just use 8 of the 9 pixels) with an 8 bit a/d converter you get an 11 bit number, or a 33 bit pixel, or 11 stops of dynamic range.... no light is lost, it is all summed up in the final number... you are trading resolution for dynamic range.
The best camera is the one in your hands

canon rumors FORUM

Re: EOS 7D Mark II Talk [CR1]
« Reply #75 on: October 10, 2013, 09:11:10 AM »

pedro

• 1D Mark IV
• Posts: 788
Re: EOS 7D Mark II Talk [CR1]
« Reply #76 on: October 10, 2013, 09:23:53 AM »
My guess - greater than 22.1 megapixels. That gives you 3x1920 pixels across, allowing for easy bining of a 3x3 pixel block from the sensor into a single pixel of 1920x1080 video... add a few pixels to each side for video stabilization and the 24MP sensor looks very likely.

How is binning at 3x3 on a grid that repeats at 2x2 easy?

Photosensors are in a 2x2 grid, but you're not binning a quad of photosensors; just the reds together, greens together (x2), and blues together...  So yes, it is easy.

In the attached image you bin the following:
r1+r1+r1+r1+r1+r1+r1+r1+r1=r1'
g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1=g₁1'
g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1=g₂1'
b1+b1+b1+b1+b1+b1+b1+b1=b1'
And you bin down right back into a bayer pattern which you can than save as-is, or demosaic, or whatever.

That method throws away half the light collected.

The 2x2 approach bins the 2 greens, 1 red and 1 blue in the block into 1 RGB pixel.  Simple and no light lost.

Rather than do the above, why not demosaic and interpolate with a simple interpolation method like bilinear?  That way, you'd at least keep all the light.

Where is there light being thrown away?

A normal pixel is 1 red, 1 blue, and 2 greens divided by 2... Let's say you have an 8 bit a/d converter.... that gives you a 24 bit pixel and you have 8 stops of dynamic range..

Bin 9 pixels together... (for ease of math I will just use 8 of the 9 pixels) with an 8 bit a/d converter you get an 11 bit number, or a 33 bit pixel, or 11 stops of dynamic range.... no light is lost, it is all summed up in the final number... you are trading resolution for dynamic range.

you are trading resolution for dynamic range.
don't understand any tech but a trade of res for dr would be the deal for many folks, I guess. anyway, I am all for high ISOs. But I can imagine, this might have its impact on high ISOs as well. Or is it like this: either better high ISOs or better DR? I don't know nothing about tech acutally, so thank you for any help.

I like every improvement on every body level as it will pay off in the next 5Ds as well
30D, EF-S 10-22/ 5DIII, 16-35 F/2.8 L USM II, 28 F/2.8, 50 F/1.4, 85 F/1.8, 70-200 F/2.8 classic,
join me at http://www.flickr.com/groups/insane_isos/

AvTvM

• 1D X
• Posts: 1183
Re: EOS 7D Mark II Talk [CR1]
« Reply #77 on: October 10, 2013, 10:03:48 AM »
you are trading resolution for dynamic range.
don't understand any tech but a trade of res for dr would be the deal for many folks, I guess.

Thing is, the 3x3 binning example would mean an 89% resolution loss (8/9) and much smaller gains in DR. A/D is not 8 but 14 or 16 bit these days.  A 24 MP sensor would only yield 2.5 MP resolution. Good enough for HD video, but not for stills.

RGF

• 1D X
• Posts: 1323
• How you relate to the issue, is the issue.
Re: EOS 7D Mark II Talk [CR1]
« Reply #78 on: October 10, 2013, 10:41:09 AM »
Hope we get a mini 1Dx.

TrabimanUK

• EOS M2
• Posts: 199
• In the words of Brian Johnson - "Shoot to thrill!"
Re: EOS 7D Mark II Talk [CR1]
« Reply #79 on: October 10, 2013, 10:51:18 AM »
+1 - Mini 1Dx
Canon 1D, 5D, 5DIII, 70D, 60D, SX150-IR/UV, 50 f1.8, 70-200 f2.8 IS II, 70-300 f4-5.6 IS USM,
Sigma 10 F2.8 EX DC DIAGONAL FISHEYE HSM,  12-24 DG HSM, 24-70 f2.8 IF EX DG HSM

Don Haines

• Canon EF 300mm f/2.8L IS II
• Posts: 3576
• Posting cat pictures on the internet since 1986
Re: EOS 7D Mark II Talk [CR1]
« Reply #80 on: October 10, 2013, 10:53:00 AM »
you are trading resolution for dynamic range.
don't understand any tech but a trade of res for dr would be the deal for many folks, I guess.

Thing is, the 3x3 binning example would mean an 89% resolution loss (8/9) and much smaller gains in DR. A/D is not 8 but 14 or 16 bit these days.  A 24 MP sensor would only yield 2.5 MP resolution. Good enough for HD video, but not for stills.
I used 8 bit for ease of math

Since bining came up concerning video (1920x3 pixels across on a new sensor???), a 3x3 bining would be great for video. Bining could also be used in still modes as a tradeoff between resolution and DR... the great thing about using bining is that is implimented in software... if you want it, turn it on, if you don't, turn it off.... sort of like having a 24Mp sensor, a 6Mp sensor (3000x2000) and 2 stops DR gain with 2x2 bining, a 2.67Mp sensor (2000x1333) and 3 stops DR gain with 3x3 bining, and a 1.5Mp sensor (1500x1000) and 4 stops gain with 4x4 bining.... enought resolution for publishing on the web...

Of course, any of this would hinge on image formats that could store and/or use any extra dynamic range....
The best camera is the one in your hands

bvukich

• Spam Assassin
• 5D Mark III
• Posts: 735
Re: EOS 7D Mark II Talk [CR1]
« Reply #81 on: October 10, 2013, 11:22:46 AM »
My guess - greater than 22.1 megapixels. That gives you 3x1920 pixels across, allowing for easy bining of a 3x3 pixel block from the sensor into a single pixel of 1920x1080 video... add a few pixels to each side for video stabilization and the 24MP sensor looks very likely.

How is binning at 3x3 on a grid that repeats at 2x2 easy?

Photosensors are in a 2x2 grid, but you're not binning a quad of photosensors; just the reds together, greens together (x2), and blues together...  So yes, it is easy.

In the attached image you bin the following:
r1+r1+r1+r1+r1+r1+r1+r1+r1=r1'
g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1=g₁1'
g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1=g₂1'
b1+b1+b1+b1+b1+b1+b1+b1=b1'
And you bin down right back into a bayer pattern which you can than save as-is, or demosaic, or whatever.

That method throws away half the light collected.

The 2x2 approach bins the 2 greens, 1 red and 1 blue in the block into 1 RGB pixel.  Simple and no light lost.

Rather than do the above, why not demosaic and interpolate with a simple interpolation method like bilinear?  That way, you'd at least keep all the light.

I'm sorry, but you apparently don't understand what binning is.

You're saying instead of demosaicing just use each 2x2 block of subpixels to create one rgb pixel.  That is not binning, and would really gain you nothing.  It should probably be better than line skipping, though.  However this is not binning.

Binning is adding the signals from all the same colored subpixels a square block (the way phase one does it is slightly more complex, each color is in a square, but the squares for each color that are combined are offset).  Binning reduces noise (main reason for doing it), and should improve low light performance (since the "metapixel" collects photons like an equivalent size pixel would. The tradeoff is reduced resolution (which is not an issue for video) and increased moire/aliasing because the AA filter is effectively nullified.

canon rumors FORUM

Re: EOS 7D Mark II Talk [CR1]
« Reply #81 on: October 10, 2013, 11:22:46 AM »

• EOS M2
• Posts: 157
• Art Director, Visual Artist, Freelance Photography
Re: EOS 7D Mark II Talk [CR1]
« Reply #82 on: October 10, 2013, 11:35:00 AM »
I, for one, am a bit tired of my bodies being "video oriented". My 5DII, 7D and 60D have not shot a second of video. I guess I'll just have to hang on to them until the pendulum swings a bit in the other direction.

This.

I also shoot no video. I want a killer stills camera, preferably apsc because of prices. Features for improved stills shooting have come to a grinding halt, except for some gradual iso improvements- since 2009.

Be the best you, screw everything else.

mkabi

• Canon 6D
• Posts: 366
Re: EOS 7D Mark II Talk [CR1]
« Reply #83 on: October 10, 2013, 02:18:43 PM »
I, for one, am a bit tired of my bodies being "video oriented". My 5DII, 7D and 60D have not shot a second of video. I guess I'll just have to hang on to them until the pendulum swings a bit in the other direction.

I hate people that own multiple bodies, complain about the next iteration of canons not having enough megapixels and complain about video capabilities. (Sounds like a spoiled brat). In my opinion, save your money and invest in a medium format camera!!!

It gives you all the MP you want.... which in turn has amazing IQ.... and guess what? No video!
« Last Edit: October 10, 2013, 02:20:56 PM by mkabi »
Canon 7D - EF 16-35mm f/2.8 L, EF - 20mm f/2.8, EF 50mm f/1.4, EF 100mm f/2.0, Custom Cine 50mm & Custom Cine 35mm

Chosenbydestiny

• EOS M2
• Posts: 244
Re: EOS 7D Mark II Talk [CR1]
« Reply #84 on: October 10, 2013, 03:02:10 PM »
I, for one, am a bit tired of my bodies being "video oriented". My 5DII, 7D and 60D have not shot a second of video. I guess I'll just have to hang on to them until the pendulum swings a bit in the other direction.

I hate people that own multiple bodies, complain about the next iteration of canons not having enough megapixels and complain about video capabilities. (Sounds like a spoiled brat). In my opinion, save your money and invest in a medium format camera!!!

It gives you all the MP you want.... which in turn has amazing IQ.... and guess what? No video!

Now now, no need to resort to name calling. The video nay sayers can cry foolishly all by themselves in their little corner. I mean, it's obvious that video changed the price of the bodies, and even the lenses. For example, It's definitely not the dramatic performance upgrade to the 5D3 that supposedly made it so unaffordable. And the bayer sensors they've been using this entire time even before video was added isn't holding development back at all, no, supposedly it's....dun dun dunnnn. Video. Thanks to video, we couldn't have 62 AF points instead of 61 on the 1D-X. Boo frikkity hoo. How can I press the shutter button now knowing my camera has features I don't use, oh no. This is worse than the print button. Save me.
Nikon electric fan, gas stove, and slippers. Canon Elan 7, 1D Mark III, 5D mark III, and 2x6D. Canon 24mm Samyang/Rokinon, 85L, 135L, and many other lenses. 2x Canon 580ex II, third party speedlites, studio strobes

Lee Jay

• 1D X
• Posts: 1335
Re: EOS 7D Mark II Talk [CR1]
« Reply #85 on: October 10, 2013, 03:35:01 PM »
I'm sorry, but you apparently don't understand what binning is.

You're saying instead of demosaicing just use each 2x2 block of subpixels to create one rgb pixel.  That is not binning, and would really gain you nothing.

It's binning the two green pixels and it's what's used in the C300:

« Last Edit: October 10, 2013, 03:37:43 PM by Lee Jay »

Lee Jay

• 1D X
• Posts: 1335
Re: EOS 7D Mark II Talk [CR1]
« Reply #86 on: October 10, 2013, 03:37:05 PM »
My guess - greater than 22.1 megapixels. That gives you 3x1920 pixels across, allowing for easy bining of a 3x3 pixel block from the sensor into a single pixel of 1920x1080 video... add a few pixels to each side for video stabilization and the 24MP sensor looks very likely.

How is binning at 3x3 on a grid that repeats at 2x2 easy?

Photosensors are in a 2x2 grid, but you're not binning a quad of photosensors; just the reds together, greens together (x2), and blues together...  So yes, it is easy.

In the attached image you bin the following:
r1+r1+r1+r1+r1+r1+r1+r1+r1=r1'
g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1+g₁1=g₁1'
g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1+g₂1=g₂1'
b1+b1+b1+b1+b1+b1+b1+b1=b1'
And you bin down right back into a bayer pattern which you can than save as-is, or demosaic, or whatever.

That method throws away half the light collected.

The 2x2 approach bins the 2 greens, 1 red and 1 blue in the block into 1 RGB pixel.  Simple and no light lost.

Rather than do the above, why not demosaic and interpolate with a simple interpolation method like bilinear?  That way, you'd at least keep all the light.

Where is there light being thrown away?

I guess I misunderstood your figure.  I thought the top-left red pixel was made by keeping the four reds and throwing out the greens and blues in that 3x3 block, and the same for the other final blocks.

9VIII

• 5D Mark III
• Posts: 673
Re: EOS 7D Mark II Talk [CR1]
« Reply #87 on: October 10, 2013, 04:05:01 PM »
Improved high ISO video performance would be very welcome. 7d footage at 1600 and above just looks horrible. I understand how the smaller pixel size on APS-C sensors is a limitation, but it would be SOO nice to have a Canon crop at least equal to just the Mark II as far video ISO performance.

Would it be possible to create a ~24mp APS-C sensor (so lets say pixel size of: 4µm) but use larger pixels (something closer to full frame size: 6.4µm.....or even bigger like 8µm) for the pixels designated for video, (resulting in AWESOME low-light performance)? Or does having different pixel sizes on the same sensor simply not work for whatever reason?

And I can think of a ton of photo applications for such technology as well. I'm one of those weirdos who would love a 8-10MP camera with ridiculous pixel size and thus great high ISO performance. Plus getting 14fps+ should be a breeze with that image size.

Ugh, the "large pixel" myth rears its ugly head again. Lower resolution sensors gain you nothing but a smaller file size. Full frame sensors are better in low light because of increased total surface area.

Binning is adding the signals from all the same colored subpixels a square block (the way phase one does it is slightly more complex, each color is in a square, but the squares for each color that are combined are offset).  Binning reduces noise (main reason for doing it), and should improve low light performance (since the "metapixel" collects photons like an equivalent size pixel would. The tradeoff is reduced resolution (which is not an issue for video) and increased moire/aliasing because the AA filter is effectively nullified.

Binning should be standard practice for photographs as well as video, given that the display industry has always counted multiple sub pixels as a single pixel.

It's actually kind of disheartening to think that a 24MP bayer filter sensor is actually producing a 6MP image by any other standard. Yes they do a very good job of interpolating data, but you're still not getting true RGB color on a per pixel basis, and with an AA filter you're throwing out all the fine details anyway. With my 5D2 I can easily cut the resolution in half with minimal loss of detail.
If someone wants a "real" 8MP camera they should actually be asking for a 32MP bayer filter with binned output.

-100% RAW-

canon rumors FORUM

Re: EOS 7D Mark II Talk [CR1]
« Reply #87 on: October 10, 2013, 04:05:01 PM »

AvTvM

• 1D X
• Posts: 1183
Re: EOS 7D Mark II Talk [CR1]
« Reply #88 on: October 10, 2013, 05:01:39 PM »
This is worse than the print button. Save me.

good comparison

* Both, the ubiquitous marked-in-red red "record video" button on every DSLRS and the former
"direct print button" are utterly useless to the overwhelming majority of DSLR users.

* Both are not user-reassignable to something useful.

* direct print buttons were out of the way, somewhere on the left side of the camera body and did not interfere with stills capture ever

* The marked in red "capture stupid video"-button is directly under users thumbs in a prime ergobnomic property location on every camera today

* Direct print buttons are gone, the record video button unfortunately is still around

9VIII

• 5D Mark III
• Posts: 673
Re: EOS 7D Mark II Talk [CR1]
« Reply #89 on: October 10, 2013, 05:28:03 PM »
I was playing around with the diffraction calculator here (http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm)
It looks like 64MP is about the highest resolution I would want on a Canon APS-C bayer sensor, details would still be pristine at f4, and only a little degraded (from the insane 64MP native resolution) at f5.6. I have to wonder how binning would affect diffraction though, if a 64MP bayer image binned to 16MP would go back to being diffraction limited only above f8.
-100% RAW-

canon rumors FORUM

Re: EOS 7D Mark II Talk [CR1]
« Reply #89 on: October 10, 2013, 05:28:03 PM »