Read this Article about Sensor development and why canon has a hard time

3kramd5

EOS R6
Mar 2, 2012
3,084
405
sony could also just start to stack images.
not the same as reading both halves and on half of the dual pixel and combining them.
i know... but there are ways to increase DR with sonys new quad pixel sensor design.
Please elaborate. I was under the impression the quad pixel architecture is merely an altered CFA pattern. How does one increase the dynamic range of the capture without stacking?

What would be neat is if Sony brought its smooth reflections app (in camera stacking written out in a raw file) to the A9; with its high frame rate you could get some big noise advantages. Edit: unless there is a trade with the electronic shutter.

Nobody but Canon has DPAF.
... and Samsung.
 
Last edited:

sdz

EOS RP
Sep 13, 2016
253
193
Pittsburgh, PA
Can't you take excellent images with Canon's current sensors?

I can. And that won't change in the future...
These improvements which distinguish Sony from its competitors mostly occur at the margins, a place where the benefits gained by innovative technology pale when compared to the costs of developing and producing that technology. Sony has the cost advantage because of its sensor business. The key area where these technical and cost advantages are important are found in the sensor readout speed. Dynamic range is less of an issue for Canon than its inability to produce sensors that can produce high frame rates, with high dynamic range and pixel count. Canon has patents applications in hand that, were they to become sensors, might address this problem by increasing the output speed of these new sensors.

The competition is not yet over.

But, Sony has an economic advantage that will enable it to keep the technical advantage It has over Canon.

What makes me wonder about all of this is Red Cinema cameras. Red is the true market leader when it comes to high performance sensor technology. Red went from a startup to a market leader in the space of a few years. How could a rich man's fever dream turn into a sensor design powerhouse? And, does knowing this about Red give us hope for Canon?
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,628
1,179
Too long a post to answer each part of it. So I'll focus on this one. If you knew what you were talking about, you'd know that the A7RII and A7RIII have a completely different sensor. In less than two years of development, Sony managed to cut by half the readout speed of their 42mp sensor while improving DR slightly. The A7RIII's sensor can be read in 14bits as fast as the A7RII could in 12bits, and twice faster in 12bits.
If this means "completely new" then the EOS R sensor is completely new. So were several of Canon's 18mp sensors.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,628
1,179
The key area where these technical and cost advantages are important are found in the sensor readout speed. Dynamic range is less of an issue for Canon than its inability to produce sensors that can produce high frame rates, with high dynamic range and pixel count.
Is there any real confirmation of this? It would neatly explain the cropped 4k issue. But at first glance it flies in the face of 50mp x 5 and 20mp x 16, all at 14-bit. Not to mention the readout for DPAF to work as well as it does.

I always guessed that Canon's cropped 4k was a processing issue downstream, not a readout speed issue.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,628
1,179
Too long a post to answer each part of it. So I'll focus on this one. If you knew what you were talking about, you'd know that the A7RII and A7RIII have a completely different sensor. In less than two years of development, Sony managed to cut by half the readout speed of their 42mp sensor while improving DR slightly. The A7RIII's sensor can be read in 14bits as fast as the A7RII could in 12bits, and twice faster in 12bits.
Speaking of sensor readout speeds...didn't the 42mp A7r II switch to 12-bit RAWs in continuous shooting mode?

The 5Ds / 5Dsr were released 5 months prior to the A7r II and could do 50mp x 5 fps x 14-bit. Am I correct in understanding that it took Sony "less than two years of development" to finally catch up to Canon on this spec? It sure looks that way from the A7r II manual.
 

sdz

EOS RP
Sep 13, 2016
253
193
Pittsburgh, PA
Is there any real confirmation of this? It would neatly explain the cropped 4k issue. But at first glance it flies in the face of 50mp x 5 and 20mp x 16, all at 14-bit. Not to mention the readout for DPAF to work as well as it does.

I always guessed that Canon's cropped 4k was a processing issue downstream, not a readout speed issue.
You may be right about this, but I suspect processing power is easier to gain than sensor capacity. There certainly is a bottleneck somewhere. I doubt Canon would produce crop 4K if it could provide 4K from a full sensor read.
 

sdz

EOS RP
Sep 13, 2016
253
193
Pittsburgh, PA
Sometimes I think we split hairs too much over things that aren't that big a deal. There is a whole lot of hair splitting going on and not many photos being posted.
This. Ford v. Chevy/Apple v. Windows once again.

Years ago I chose Canon. I went with Canon because of the color. And, thanks to a good camera and high quality lenses, an unschooled bungler like me can produce nice photos. The images I've seen from the EOS R were gorgeous. This has barely mattered as the complaints come by the bucket. For me, switching to Sony or Nikon or Fuji, etc. is a nonstarter. Besides, would a switch to Sony matter much for the images I produce? I doubt it. Pixel peeping might find more noise in the Canon image. But only someone looking for noise would notice it. Everyone else would consider them worthy.

Sometimes we should enjoy what we have because what we have is good enough.
 

Keith_Reeder

I really don't mind offending trolls.
Feb 8, 2014
957
469
60
Blyth, NE England
Cameras have got so good that nowadays it is not about the quality of the images but the keeper rate and likelihood of getting them in the first place.
And if Canon cameras were incapable of delivering extremely well on those criteria, jumping ship would make sense.

I've done so myself in the past - from Nikon - and I would again without a second thought if I couldn't routinely achieve my photographic aims with my Canon gear (I'm as demanding as any other photographer, and - shooting wildlife, birds and sport - probably more demanding than most).

But Canon just delivers, and anyone who suggests otherwise is almost certainly blaming the wrong "problem"...
 
  • Like
Reactions: pj1974 and ethanz

Keith_Reeder

I really don't mind offending trolls.
Feb 8, 2014
957
469
60
Blyth, NE England
...its inability to produce sensors that can produce high frame rates, with high dynamic range and pixel count. Canon has patents applications in hand that, were they to become sensors, might address this problem by increasing the output speed of these new sensors.
"Inability" is not the same as "disinclination" - Canon has so far been pretty clearly just not that interested in getting into this game, presumably because they believe their customer base doesn't see such things as a priority.

I suspect they're right, if market share is anything to go by (yes, that again - you can't ignore this, one of the few available facts in these discussions).

I'm aware that some of Canon's recent public pronouncements might imply otherwise, but personally I put more store in what someone does than what he says...
 

MayaTlab

EOS 90D
Oct 6, 2015
193
81
If this means "completely new" then the EOS R sensor is completely new. So were several of Canon's 18mp sensors.
It takes probably quite a bit more effort to halve the readout speed of a sensor than to go from 4,3fps with AF (5DIV) to 5 (R).

Is there any real confirmation of this? It would neatly explain the cropped 4k issue. But at first glance it flies in the face of 50mp x 5 and 20mp x 16, all at 14-bit. Not to mention the readout for DPAF to work as well as it does.

I always guessed that Canon's cropped 4k was a processing issue downstream, not a readout speed issue.
Speaking of sensor readout speeds...didn't the 42mp A7r II switch to 12-bit RAWs in continuous shooting mode?

The 5Ds / 5Dsr were released 5 months prior to the A7r II and could do 50mp x 5 fps x 14-bit. Am I correct in understanding that it took Sony "less than two years of development" to finally catch up to Canon on this spec? It sure looks that way from the A7r II manual.
In the case of the 5DS' 50mp x 5fps and the 1DXII's 20mp x 16fps, the only thing these cameras' sensors are tasked to do when performing these bursts is to simply pumps out exposures. The reason the A7RII dropped to 12bits is that, being a mirrorless, it's also performing AF / AE, and at 3fps, provides a liveview feed in burst. Now we're at 8fps in 14bits with AF and liveview feed in the case of the A7RIII with a 42mp camera.
BTW, all of Canon's M cameras drop to 12bits in burst as well, and none of them manages to provide a liveview feed in between frames in continuous drive mode. That's a key area of performance that is yet to be known with the R.
It's fairly easy to know that Canon's sensors struggle with readout speed : the M50's or the 5DIV's significant rolling shutter in cropped 4K is all we need to know. If Canon wanted to provide FF 4K, they'd need to skip so many lines that IQ would be hideous.

BSI was supposed to improve high ISO. It made no practical difference in FF sensors. Reading forum posts you would think Canon FF sensors must be behind on high ISO. Looking at actual RAW files they're right at the top.

Nikon and Sony's best DR sensors beat Canon's best DR sensor by 1 stop on DR. At base ISO. If you ignore dual pixel RAWs. That 1 stop amounts to a couple ticks on a noise slider when pushing shadows.

Nobody but Canon has DPAF. Which is a rather complicated sensor engineering feat. While I have video spec related complaints about the EOS R, early reviews indicate it's one of the best focusing MILCs out there. Meanwhile Nikon Z models are getting hammered for AF issues in early reviews. Do they not use Sony sensors? I guess you can't blame them. It's only the recently released gen 3 Sony A7 bodies which offer decent FF AF in the Sony mirrorless line.

For all the endless talk online about "superior" Sony sensors the 3 year old Canon 5Ds and 5Dsr still easily tie the D850 and A7r3 for image quality and large prints at low ISO, and are only very slightly behind at high ISO.
Let's address the rest. BSI was never supposed to improve high ISO for larger sensors. It's only for smaller ones that it does. However BSI does enable a more flexible approach to wiring the sensor, and that's on thing that helped enable the A7RII, A7RIII or D850's readout speed (as Imaging-Resource interviews suggest). More importantly it's also a prerequisite for stacked sensors. No BSI ? No stacking. So far we've seen no BSI from Canon. And a patent means zilch for consumers if it can't be produced at a specific cost.

DPAF undoubtedly is the future but it doesn't come for free or without compromises, particularly for stills cameras. One inconvenient of it is that DPAF sensors can't feed both phase information and actual image information at the same time to the camera (analysis of DP raw files suggest that DPAF sensors have only one ADC path and can't read the A and B channels at the same time). Given the M50's rolling shutter in 4K, I suspect that the reason there's no DPAF in 4K is that the sensor's readout speed is too slow to both perform a DPAF reading and an image reading. Masked pixels sensors are compromised in many ways, but they can provide both phase AF and image information at the same time. In 2018 DPAF isn't the bee's knees. It will increasingly be once readout speed increases.

DP raw is not a practical solution to increase DR in all cases as you can get parallax errors, and it doesn't play nice with the softwares most people use. That said as JonSnow suggested DP raw prefigures the sort of exciting things we'll be able to do with our cameras. Well actually, it's the iPhone that really prefigures that, but I digress.

And finally, the 5DS(R) isn't only slightly behind at high ISOs. It's a full stop behind (particularly under warmer TCs as its blue channel noise is elevated) :

Screen Shot 2018-09-13 at 08.22.14.jpg

That could be partly explained by another sensor trick that is yet to be seen in a Canon camera : current Sony sensors use Aptina's dual gain architecture to have, basically, two base ISO values (64/100 and one higher up, around 640/800) to improve noise performance.

And its DR deficit is a bit more than one stop. Compared to what is currently produced by Sony, it's rather 2 stops.

And before anyone thinks otherwise, let me just say that I'm a Canon user who's completely uninterested in Sony cameras, and who doesn't want to see a Sony sensor inside a Canon. But Canon's current deficit in terms of readout speed will have to be addressed at some point, as it's not just a question of IQ, but simple, basic operational qualities. As this article suggests, they may have quite a few difficulties ahead in that area.
 

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,346
7,150
Oh, I'm well aware of that.

Have you noticed that their images aren't actually any better?
As MikeH has written, it's not that the quality is any better but whether you are able to get the shot in the first place and how many you have to take for really good keepers. What many of the improvements to lenses and bodies do is to extend the range of what you can do. Where Nikon and Sony have done is to have much better tracking for Birds in Flight against fussy backgrounds. I am very happy with the shots I get from my Canon gear when I can get them, but there are times when the tracking fails.
 

3kramd5

EOS R6
Mar 2, 2012
3,084
405
My Nokia 7 Plus has DPAF by using dual Sony sensors (each with its own lens).

Not a good solution for an ILC, though.
Samsung however has IP for what is a DPAF solution very alike Canon’s, so much so I suspect it’s licensed or part of some technology sharing agreement.
 
Last edited:

Keith_Reeder

I really don't mind offending trolls.
Feb 8, 2014
957
469
60
Blyth, NE England
As MikeH has written, it's not that the quality is any better but whether you are able to get the shot in the first place and how many you have to take for really good keepers..
And as I suggested to Mike, nobody using Canon kit well will have any problems on that score. For every one that has jumped ship, far more haven't.

I tend strongly to the belief that when people argue that they've given up on Canon because the equipment is a de facto barrier to them achieving their aims, they're being rather economical with the truth: like Art Morris jumping to Nikon because - he said - only Nikon allowed him to achieve shots "like this" (referencing a pretty unchallenging shot of a Pelican taking off), on a website full of images at least as technically challenging, taken with Canon kit.

There's nothing wrong with wanting a change (which may or may not equate to needing to change), but the case for changing as being the "only way" to get results, is overstated, and deeply unlikely to bear up to critical scrutiny...
 
Last edited:

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,346
7,150
And as I suggested to Mike, nobody using Canon kit well will have any problems on that score. For every one that has jumped ship, far more haven't.

I tend strongly to the belief that when people argue that they've given up on Canon because the equipment is a de facto barrier to them achieving their aims, they're being rather economical with the truth: like Art Morris jumping to Nikon because - he said - only Nikon allowed him to achieve shots "like this" (and referencing a pretty unchallenging shot of a Pelican taking off), on a website full of images at least as technically challenging, taken with Canon kit.

There's nothing wrong with wanting a change (which may or may not equate to needing to change), but the case for changing as being the "only way" to get results, is overstated, and deeply unlikely to bear up to critical scrutiny...
Your stock-in-trade answer that the reason why people don't get results is because they are not using equipment "well" is not only condescending, but it also does not "bear up to critical scrutiny". Arbitrage, who is a stunning BIF photographer, has sold his Canon gear and has gone over to Sony and Nikon, and he is certainly not being "rather economical with the truth". He has done it to get those shots he hasn't been able to get with Canon. Now, I am not ready to give up on Canon, but I am competent enough to know its limitations in AF and how far I can go with it.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,628
1,179
It takes probably quite a bit more effort to halve the readout speed of a sensor than to go from 4,3fps with AF (5DIV) to 5 (R).
You want to place incremental Sony sensor improvements in the category of 'completely new' while placing incremental Canon sensor improvements in the category of 'reused and old.' I'm not going to play that game. We know the EOS R sensor has some changes to the circuitry based on the AF configuration. The Sony sensor might involve nothing more than a clock speed change (which can be enabled by manufacturing process improvements with the same circuit design) or even a change downstream.

New is new. The 5Ds 50mp and 5D IV 30mp sensors never existed before. Incremental is incremental. The 42mp and 24mp Sony sensors clearly existed before the third generation, and the third generation clearly built upon those previous designs. And it's hypocrisy for Sony fans to say a Canon incremental sensor is 'old and reused' while praising an incremental Sony sensor for being a revolution.

The reason the A7RII dropped to 12bits is that, being a mirrorless, it's also performing AF / AE,
You know this is the reason? Citation?

BTW, all of Canon's M cameras drop to 12bits in burst as well,
The original M does not. I was surprised to find out that the M5 does.

Let's address the rest. BSI was never supposed to improve high ISO for larger sensors.
Tell that to the ad departments and fanboys.

In 2018 DPAF isn't the bee's knees.
It kinda is. But the point was the engineering complexity which suggests Canon is not as far behind as forum posts would suggest.

And finally, the 5DS(R) isn't only slightly behind at high ISOs. It's a full stop behind
LOL! Your images show the exact opposite. All three 25600 images look worse.

But let's make the point clear with some screenshots of an area that has some real detail. Being a full stop behind means the D850 and A7r III should look just as good at 25600 as the 5Dsr at 12800. Let's see...

Screen Shot 2018-09-13 at 7.23.04 PM.png


Screen Shot 2018-09-13 at 7.23.36 PM.png


Nope and nope. They both look worse set 1ev higher. I bet if dpreview had underexposed the scene 1ev and pushed in post for an effective 25600, the differences would be minor.

And its DR deficit is a bit more than one stop.
Canon's best DR sensor is the 5D IV sensor, and it's behind by one stop.
 
  • Like
Reactions: pj1974

dtaylor

Canon 5Ds
Jul 26, 2011
1,628
1,179
Finally had time to read the article.

  1. It's a puff piece.
  2. The central thesis relies upon an unproven assumption.
  3. It perpetuates the "Sony is superior in every measure" mythology.
  4. According to the article the end of photography is near any way.
I kinda wish I had the time I spent reading it back.
 
  • Like
Reactions: Mark D5 TEAM II

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,346
7,150
What shots would those be?
He's a most phenomenal amateur bird photographer see - https://www.flickr.com/people/100907765@N08/
He has had the time and money to work through the best Canon, Nikon and Sony gear, and is now able to capture small birds in flight that he could not do with Canon. His comments are spread through nearly 20,000 posts on the FM site. I occasionally look at them and remember the gist: Canon is still the fastest to lock on but its AF-C is slightly jittery, leading to lower number of keepers and most importantly it is not in the same league for tracking as the A9 or D500. His BIFs speak for themselves. Now, I can get some good shots of BIFs, like this puffin travelling like a rocket on Farne, with my Canon gear (5DIV + 400mm DO II) , but I can't match that guy - but maybe I would never be able to with any gear. But, I believe him.

Puffin+Sandeels_2B4A4924-DxO_vvb.jpg