Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: New focus mechanism for Canon 70D  (Read 9691 times)
dreed
Sr. Member
****
Offline Offline

Posts: 1246


« on: July 02, 2013, 02:04:39 AM »
ReplyReply

Canon's latest DSLR (the 70D) comes with a new focus technology built into the sensor called "Dual Pixel CMOS AF".

dpreview talk about it here:
http://www.dpreview.com/previews/canon-eos-70d/3

... however the most interesting comment they make about it is this:

"Finally, we're also really interested to see what happens when Canon - as it inevitably must - puts its new AF technology into a mirrorless camera. An 'EOS M2' with Dual Pixel CMOS AF and more enthusiast-oriented controls could completely transform Canon's fortunes in this sector, after the lukewarm reception received by the EOS M. This could also substantially negate one of the biggest disadvantages of mirrorless cameras so far - their relatively poor focus tracking capabilities - raising further questions about why you'd still bother with the bulk of an SLR. But that's an argument for another day."
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 8007



WWW
« Reply #1 on: July 02, 2013, 03:02:36 AM »
ReplyReply

Canon's latest DSLR (the 70D) comes with a new focus technology built into the sensor called "Dual Pixel CMOS AF".

dpreview talk about it here:
http://www.dpreview.com/previews/canon-eos-70d/3

... however the most interesting comment they make about it is this:

"Finally, we're also really interested to see what happens when Canon - as it inevitably must - puts its new AF technology into a mirrorless camera. An 'EOS M2' with Dual Pixel CMOS AF and more enthusiast-oriented controls could completely transform Canon's fortunes in this sector, after the lukewarm reception received by the EOS M. This could also substantially negate one of the biggest disadvantages of mirrorless cameras so far - their relatively poor focus tracking capabilities - raising further questions about why you'd still bother with the bulk of an SLR. But that's an argument for another day."

Do we know how this differs from the solution implemented by Nikon in their 1 series mirrorless cameras? Those track very well.

Thanks.

cheers,
Bernard
Logged

A few images online here!
heinrichvoelkel
Sr. Member
****
Offline Offline

Posts: 368


« Reply #2 on: July 02, 2013, 03:08:03 AM »
ReplyReply

Do we know how this differs from the solution implemented by Nikon in their 1 series mirrorless cameras? Those track very well.

Thanks.

cheers,
Bernard


read it here: http://www.newsshooter.com/2013/07/01/canons-new-eos-70d-a-revolution-in-autofocus-for-video/
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5130


« Reply #3 on: July 02, 2013, 08:17:47 AM »
ReplyReply

Do we know how this differs from the solution implemented by Nikon in their 1 series mirrorless cameras? Those track very well.
One claim is working well in very low light: down to EV0 and at f/11 (with the ability to operate well even when stopped down important for video obviously.) From what I have read, all previous on-senosr PDAF systems perform rather poorly in low light, at least compared to traditional SLR PDAF.

The other obvious difference form previous approaches is that it has millions of AF sensors, one available at almost every photo-site, which is potentially a great advantage for smooth tracking in video.

The claims seem oriented to focus tracking video; I wonder how fast and accurate it is for still photography and single AF mode compared to other technologies?


Disclaimer: I do not know any of these performance advantages for sure; note well the words "claim" and "potentially".
Logged
fike
Sr. Member
****
Offline Offline

Posts: 1373


Hiker Photographer


WWW
« Reply #4 on: July 02, 2013, 08:34:46 AM »
ReplyReply

Tracking focus on an screen won't help much with sports, wildlife or other action photography, but it will help with video.  This seems to be another example of Canon chasing the video business more than the still photo business. 

On the other hand, all this live focus aside, if they don't close the image quality gap with Nikon, they are going to be further marginalized as a photography leader.

This camera also looks like it is replacing the 7D camera and its rumored replacement.  The specs for the 70D take over most of the 7D ground with weather sealing and 19-point autofocus (optical viewfinder phase detection).
Logged

Fike, Trailpixie, or Marc Shaffer
marcshaffer.net
TrailPixie.net

I carry an M43 ILC, a couple of good lenses, and a tripod.
BJL
Sr. Member
****
Offline Offline

Posts: 5130


« Reply #5 on: July 02, 2013, 10:21:10 AM »
ReplyReply

fike,

    I agree that the main emphasis here is on video and other features (tilt screen with expanded touch operation, WiFi and image sharing ...) rather than core aspects of still image quality. Maybe Canon thinks that for the people choosing mainstream formats like EF-S, the IQ is now good enough for most customers, and so IQ increments have far less influence on purchasing decisions than these other features. The apparent end of the "single digit" line of EF-S series (no 7D replacement in four years, the name 6D used for 35mm format instead) also suggests that Canon is steering people who demand higher IQ towards 35mm format bodies like the 6D.

That said, this is offering PD AF in live view mode, which should also improve continuous AF tracking for still photography --- even if this is just a side effect, of still photography getting crumbs from the table of efforts aimed at improving video performance.
Logged
dreed
Sr. Member
****
Offline Offline

Posts: 1246


« Reply #6 on: July 02, 2013, 10:55:55 AM »
ReplyReply

    I agree that the main emphasis here is on video and other features (tilt screen with expanded touch operation, WiFi and image sharing ...) rather than core aspects of still image quality. Maybe Canon thinks that for the people choosing mainstream formats like EF-S, the IQ is now good enough for most customers, and so IQ increments have far less influence on purchasing decisions than these other features.

As an example, a friend of mine went to buy a new camera with his wife. I'd pointed out that Nikon had better IQ and would give him better results but when they got to the store, sales guy says "... and this Canon model has Wifi.' Wife: "We're buying the one with WiFi." Case closed. Canon have been focusing their camera designs on "features and gimics" that are gobbled up by those for whom any of the 2013 cameras is good enough.

Whether or not the new autofocus thing works and works well remains to be seen. I'm kind of curious if Michael will take a look at it or review it.
Logged
NancyP
Sr. Member
****
Offline Offline

Posts: 875


« Reply #7 on: July 02, 2013, 11:55:25 AM »
ReplyReply

I am still hoping for an improved sensor (dynamic range / SNR), 5D3/1DX-like autofocus, and 10 fps with deep buffer for the rumored 7D2. APS-C is still relevant for focal length limited uses such as bird photography.
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1679


« Reply #8 on: July 02, 2013, 12:12:13 PM »
ReplyReply

I am still hoping for an improved sensor (dynamic range / SNR), 5D3/1DX-like autofocus, and 10 fps with deep buffer for the rumored 7D2. APS-C is still relevant for focal length limited uses such as bird photography.
Relevant as in "most reach per dollar", and "highest #fps per dollar". If you can afford it, I assume that a D800 with a decent 200mm will give a Canon 7D with a 200mm a run for its money (provided that AF and framerate suits your task)?

-h
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1679


« Reply #9 on: July 02, 2013, 12:19:07 PM »
ReplyReply

Tracking focus on an screen won't help much with sports, wildlife or other action photography, but it will help with video.  This seems to be another example of Canon chasing the video business more than the still photo business. 
If/when they can do on-sensor PDAF that is as fast, accurate and well-tracking as the dedicated sensor PDAF, what do we need the flipping mirror for?
Quote
On the other hand, all this live focus aside, if they don't close the image quality gap with Nikon, they are going to be further marginalized as a photography leader.
How can you be a leader and be marginalized at the same time? Either they are leaders (in many ways they are, although not in certain technical image sensor disiplines), or they are marginalized (Kodak would seem like an example of this)

I hope that they will improve DR@low ISO, but I have a feeling that the 70D is not the model that gets this treatment. I'd be pleasantly surprised, though.

I have been very negative of Canon for some years, but I think that they have done well with the 100D and the 70D seems like a potentially strong model as well. When you are the biggest player, you want to widen your market (miniature DSLR, video-DSLR) while keeping costs down (re-use old sensor tech as long as the market tolerates it but not any further).

-h
Logged
shadowblade
Sr. Member
****
Offline Offline

Posts: 689


« Reply #10 on: July 02, 2013, 01:49:06 PM »
ReplyReply

fike,

    I agree that the main emphasis here is on video and other features (tilt screen with expanded touch operation, WiFi and image sharing ...) rather than core aspects of still image quality. Maybe Canon thinks that for the people choosing mainstream formats like EF-S, the IQ is now good enough for most customers, and so IQ increments have far less influence on purchasing decisions than these other features. The apparent end of the "single digit" line of EF-S series (no 7D replacement in four years, the name 6D used for 35mm format instead) also suggests that Canon is steering people who demand higher IQ towards 35mm format bodies like the 6D.

That said, this is offering PD AF in live view mode, which should also improve continuous AF tracking for still photography --- even if this is just a side effect, of still photography getting crumbs from the table of efforts aimed at improving video performance.

On the other hand, there's been no real improvement in full-frame IQ either, especially at lower ISOs (i.e. the bulk of landscape, studio and fine art photography), since the 1Ds3 and 5D2. Which is very disappointing, given Canon's fantastic and unmatched lineup of tilt-shift lenses.

Improved low-ISO noise and image quality could be a side-effect of having two photosites in each pixel, though - since each pixel is sampled twice, read noise should be averaged out, resulting in less-noisy low-ISO images, where read noise is the main cause of noise. I also hope they've moved to on-chip A/D conversion like every other manufacturer out there, or at least some other way of mitigating the pattern noise which plagues every single Canon sensor out there at the moment.
Logged
shadowblade
Sr. Member
****
Offline Offline

Posts: 689


« Reply #11 on: July 02, 2013, 01:52:12 PM »
ReplyReply

How can you be a leader and be marginalized at the same time? Either they are leaders (in many ways they are, although not in certain technical image sensor disiplines), or they are marginalized (Kodak would seem like an example of this)

They make great cameras, but stick crummy sensors based on outdated technology in them.

If only digital sensors were interchangeable like film - you could stick Velvia 50 into any film camera, regardless of the manufacturer, and chose a body on features and lens selection alone, rather than image quality.
Logged
Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1801



WWW
« Reply #12 on: July 02, 2013, 02:20:52 PM »
ReplyReply

"I assume that a D800 with a decent 200mm will give a Canon 7D with a 200mm a run for its money"
I hoe it does not come across as rude  becasue I generally like what you have to say like you Nancy, but comparing a D800 to a 7D is like comparing an octopus to a cuttlefish.
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
shadowblade
Sr. Member
****
Offline Offline

Posts: 689


« Reply #13 on: July 02, 2013, 02:29:21 PM »
ReplyReply

I am still hoping for an improved sensor (dynamic range / SNR), 5D3/1DX-like autofocus, and 10 fps with deep buffer for the rumored 7D2. APS-C is still relevant for focal length limited uses such as bird photography.

Not really - it's the high pixel density that's relevant, not the small sensor size.

I'd love a high-resolution, high-pixel-density full-frame camera for wildlife photography with long telephotos - that way, I'd be able to just pick a focus point, aim it at the animal's eye and track it as it moves, then crop around it for the final image (ending up with an effective sensor area around the same as that of a crop sensor) instead of having to continually change focus points depending on which way the animal is facing at any particular moment, or whether it's bobbing its head up and down. FPS is nice to have, but not critical, unlike with sports photography.
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1679


« Reply #14 on: July 02, 2013, 03:32:57 PM »
ReplyReply

"I assume that a D800 with a decent 200mm will give a Canon 7D with a 200mm a run for its money"
I hoe it does not come across as rude  becasue I generally like what you have to say like you Nancy, but comparing a D800 to a 7D is like comparing an octopus to a cuttlefish.
My point was that a D800 has about the same "reach" as my 7D.

I get that there are other differences between them.

-h
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5130


« Reply #15 on: July 02, 2013, 03:49:56 PM »
ReplyReply

A few comments to various pasts above:

1) having two photo-diodes under the single microlens and color filter of each of 20 million photosites will not give any noise advantage over having a single, larger photodiode at each of the same number of photosites: adding the signal from the two photodiodes more or less reproduces the signal that a single photodiode would give, and the same total photon count means the same shot noise, and so on. Except that having two smaller photodiodes probably wastes some space on extra wiring and circuitry, so reduces well capacity, and likely adds a little "blind area" between the pair of photodiodes where light is not detected. So the overall effect is likely slightly worse DR and QE than an equal number of "normal" photosites. Then only "noise advantage" would be comparing to the per pixel noise measurements of a 40MP sensor of the same size and technology, and hopefully the fallacy of comparing IQ "per pixel" rather than "per image" when comparing sensors of different pixel counts has been thrashed out here often enough.

2) For birding and the like, carrying the bigger, heavier, more expensive, slower frame rate D800 and then cropping from the same focal length  to get a bit less reach (15MP in place of the 20MP) sounds like an imperfect substitute! Ignoring all those other factors and addressing only the reach (which is still a bit worse worse) misses the point. Of course, if you already have a D800, you can make do quite adequately for such tasks, but let us not pretend that there is no advantage to using a tool better suited to that task.

3) As to the question of whether we will still need the flipping mirror if this new technology can AF as well as the best SLR AF: I already feel no need for a flipping mirror, but for some, issues like EVF lag and problems in very low light are still reasons for keeping the OVF option. (Not for much longer, I am guessing: the latest Sony EVFs and the new Olympus/Epson external VF4 seem quite impressive.)


P. S. There still seem to be people who expect 35mm format SLR's to end up with roughly the same pixel size as smaller SLR formats, so that they can counter any claimed advantage of the smaller formats through cropping. And yet the recent trend is almost the opposite: Canon and Sony now offer about the same maximum pixel count in APS-C as in their 35mm format cameras, and the D800 is the only exception --- and it still only matches only 16MP in DX crop, so significantly lower resolution than 24MP DX format sensors when used with equal focal lengths and cropping. If anything, there has been a trend of increasing the gap in lp/mm sensor resolution ("pixel density") in the eleven years since the 14MP Kodak 14N and 11MP Canon 1D arrived alongside 6MP APS-C alternatives.
« Last Edit: July 02, 2013, 04:00:48 PM by BJL » Logged
Bernard ODonovan
Full Member
***
Offline Offline

Posts: 114


« Reply #16 on: July 02, 2013, 03:55:15 PM »
ReplyReply

Some fun with numbers

Old Canon APS C’s are 22.3 x 14.9 = 1.61 Crop Factor

The new EOS 70D is 22.5 x 15 = 1.6 Crop Factor

The new sensor has 20.20 MP (Million Pixels) in the photo capture active area

The PD AF Photo Diodes are active 79.56% horizontal and 80% Vertical = 63.65% of the center of the active photo area.

That is 12.86 MP PD AF Pixels (active)

That is 25.71 MP PD AF Photo Diodes (active)

That is 7.34 MP Pixels not performing Sensor PD AF (inactive)

As we do not know if Canon kept all these pixels the same to avoid any frame banding, we can assume either they have a total of 33.05 MP photo diodes in the sensor in the photo capture area acting as the 20.20 MP sensor or if all the same type of pixel then 40.40 MP photo diodes in the capture area acting as the 20.20 MP sensor.

This also assumes Canon don’t sneak in a dynamic boost by having 40.40 MP photo diodes with half reading off hotter than the rest per pixel at capture point to enlarge dynamic range if that was possible on this camera. Given this camera is processing 20.20 Pixels up to 7 FPS, that is 155.4 MP/s (approx 8% more the D7100 at it’s full chat) it would need to be doing 310.8 MP/s for 40.40 Photo diodes although we know it has photo diode summing on the PD AF photo diodes, when they act as normal photo pixels, so that gets us back to 155.4 MP/s of actual sensor output ;-) Just the matter of reading off the possible enlarged dynamic range and alternating the hot and cold to avoid any banding or other optical artifacts.

Hard to imagine the data rate of the PD AF section, as that is reading up to 25.71 MP Photodiodes at AF rates (fractions of a second), so my guess is they are either not reading all of them at the same time, or that the AF data is so much smaller than 14 bit pixel read for image data to allow a usable feed rate to the AF system to process in real time.

Let’s quote Canon’s old Marketing info:

‘’ The Canon DIGIC 5+ Image Processor brings phenomenal increases in processing speed and power. Data processing performance is 17x faster than the Canon DIGIC 4 Image Processor and features new algorithms that promote greater noise reduction at higher ISOs. Improved noise reduction provides greater image quality and a faster processing speed makes it possible to obtain high-quality shots during continuous shooting.’’

So compared to the ‘’Dual DIGIC 4’’ in the old EOS 7D, this new Camera has 8.5 Times the processing power (give or take the rest of the chips on board). Quite a big jump. It also needs to handle approx 12.2% more Sensor Pixel Data at image capture point (single frame). At 8 FPS the EOS 7D is cranking out 144 MP/s so that difference drops to 8% more data for the new EOS 70D to handle at full chat at it’s 7FPS max.

Note how Nikon equaled Canon’s data rate at max FPS (old Canon 7D compared to their new D7100). As above the new EOS 70D has gone 8% higher ;-)

The new EOS 70D would have 22.98 MP at 1.5 Crop factor so the new D7100 still has smaller pixels.

The new EOS 70D would have 51.71 MP at Full Frame. Now, if they had the extra photo diode on every pixel for that, we get 103.42 MP Photo Diodes… Phew…!!!

Just a bit of fun with numbers ;-)
Logged
Bernard ODonovan
Full Member
***
Offline Offline

Posts: 114


« Reply #17 on: July 02, 2013, 05:04:01 PM »
ReplyReply


Just found this, says approx 40.3 MP for the Photo Diodes  Smiley

http://www.youtube.com/watch?v=dGxLERJi5_Y

As for the AF, another guess whilst I search, is they group them some how to act as small AF points rather than just AF pixels to keep data rates for AF manageable. Guess we will learn more as the tech is explained
Logged
NancyP
Sr. Member
****
Offline Offline

Posts: 875


« Reply #18 on: July 02, 2013, 06:33:35 PM »
ReplyReply

I am not comparing the D800/E to 7D, they are different cameras for different purposes, and vive la difference.

 The beauty of APS-C format for birders is the ability to put "more pixels on the bird" (smaller pixels, greater pixel density) with a given focal length lens. Because the MINIMUM desirable focal length for shooting small songbirds is 400mm, bird photography can become very expensive. One needs either a closer approach or a longer lens to get an equivalent number of pixels covered on FF. I currently have the "budget" kit for high image quality bird photography with autofocus, an APS-C camera (Canon 60D - admittedly the 7D would have been better) and the Canon 400mm f/5.6L, allowing me to get high quality images for approximately $2,000.00 in total hardware cost and 2kg weight. That's approximately one-tenth the cost of a "professional" kit consisting of a Canon 1DX, 600mm f/4, TC 1.4, TC 2.0, and the tripod most people need for this 5kg lens and camera set. (Glenn Bartley, frequent LuLa poster, is one of the pros who uses the 7D; Art Morris is in the 1DX camp).
Logged
dreed
Sr. Member
****
Offline Offline

Posts: 1246


« Reply #19 on: July 02, 2013, 06:48:47 PM »
ReplyReply

I am not comparing the D800/E to 7D, they are different cameras for different purposes, and vive la difference.

 The beauty of APS-C format for birders is the ability to put "more pixels on the bird" (smaller pixels, greater pixel density) with a given focal length lens.

Lets put this another way: if both a FF camera and a crop (APS-C) camera both had sensors with equal pixel density then APS-C would offer no advantage to the birder.

Anyway, all of that is really irrelevant to the new focus mechanism in the Canon 70D.

With so many active photo diodes being used for autofocus, I wonder how well the AF will be able to track birds - well really, that'll still come down to tracking it with the lens for BIF.
Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad