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

MayaTlab

EOS 80D
Oct 6, 2015
185
66
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.
That's unlikely to be simple clock speed change as DR increased with the A7RIII.

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.
Which one do you think will be felt the most incremental between :
- going from 4,3fps to 5fps in servo, with still no liveview feed in between frames in continuous drive,
- or going from 3fps with AF/AE/liveview feed with 12bit files to 8fps with AF/AE/liveview feed with 14bit files ?


You know this is the reason? Citation?
It's just logic. The only reason you drop the bit depth in continuous mode on a camera that can produce 14bit files otherwise is if you want to halve the readout speed. And the only reason you'd do this on a mirrorless camera is to allow it to do what it relies on its sensor to do (AF, AE, liveview feed).

The original M does not. I was surprised to find out that the M5 does.
You wouldn't have been had you known how readout speed is a problem for Canon.

It kinda is. But the point was the engineering complexity which suggests Canon is not as far behind as forum posts would suggest.
Gave you a good reason why it isn't (can't read the sensor for both image information and phase information at the same time). One reason why you aren't getting AF in 120fps @720p. It's just a mixed bag, right now, on stills cameras.

DPAF may be complex from an engineering point of view. But it doesn't seem to require techniques that Canon will eventually have to adopt if they want to increase the readout speed of their sensors. That Canon's sensors show design complexity in one area doesn't mean that they show complexity in the one area where they're currently lagging.

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...
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.
I disagree with you assessment. The reason I always bring up the issue of blue channel noise under warmer TCs is that not only is this is the sort of lighting where you're very likely to be actually taking pictures, it's also something that will affect how a picture looks regardless of its print size. The way Lightroom deals with this is by turning darker areas magenta. The way C1 and DPP deal with this is by reducing the blue channel's saturation - resulting in a loss of colours.

Canon's best DR sensor is the 5D IV sensor, and it's behind by one stop.
I thought you were talking about the 5DSR. Indeed the 5DIV is more competitive in measured terms. But it's got banding at extreme pushes. That said the 5DIV is enough for me.
 

AlanF

5DSR
Aug 16, 2012
5,035
1,820
Thank you.



Canon and Nikon shooters using cameras over the past two decades could get good keeper rates with wildlife including BiF. That doesn't prove the A9 isn't giving a higher keeper rate, or giving a similar rate with less practice. Maybe it is for some people. I certainly won't deny the possibility.

But that belief by an A9 user should not be blown into 'shots others can't get' no matter how good his portfolio. Unless, of course, his portfolio actually includes shots of a technical nature which just cannot be found in the portfolios of others. I saw great shots, but I didn't see truly new/unique/technically never before achieved shots. (I doubt that exists for BiF.)



You have the equipment (5D IV; 5Dsr; 400 DO II; 100-400 II) and there are people producing incredible portfolios with that equipment. I would give the generic action photography answer of opportunity, settings, technique. You have to be out shooting a lot. You have to have your settings dialed in (including any AF adjustments). And you have to practice until you have technique down. Same as for any other action genre. Even a short pause in shooting (say, a few months) will drop your keeper rate until you've spent a few days shooting and are back in practice. I've experienced that with sports shooting. And I've read that Nat Geo photographers went through this all the time. Their keeper rate at the end of an assignment was far higher than at the beginning.

At any rate, he's not that good automatically because of an A9. And buying an A9 will not magically put anyone at his level overnight. He likely shoots that subject matter far, far more than you realize.
Here is an example of what I am mean by no shot is impossible. The best images in our thread on dragonflies of them in flight use the technique of twirling manual focus while in a continuous burst. The success rate is in the low percents, but well worth the effort https://www.canonrumors.com/forum/index.php?threads/dragonflies-and-damselflies.35543/page-9#post-744454
As you haven't answered my direct question "Do you try to take shots like those - they are just so difficult?" and have just made general advice akin to telling your grandfather how to suck eggs, that capturing small birds in flight is not in your repertoire?
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,196
481
That's unlikely to be simple clock speed change as DR increased with the A7RIII.
The frame rate change could be a clock speed change. That does not preclude other changes.

Which one do you think will be felt the most incremental between :
That doesn't change the point that new is new and incremental is incremental, and Sony fans play fast and loose with those definitions whenever they want to bash brands.

It's just logic. The only reason you drop the bit depth in continuous mode on a camera that can produce 14bit files otherwise is if you want to halve the readout speed.
Or if a downstream component can't handle it.

That Canon's sensors show design complexity in one area doesn't mean that they show complexity in the one area where they're currently lagging.
That's a fair statement. The problem is that they are typically not described as "lagging in one area", but as lagging in every area.

If you're correct and readout speed is an issue then I would agree it's a problem from the video side, and will be a problem on the mirrorless side, in terms of functionality, if not resolved soon. But it no more justifies the statement "Canon sensors are behind", implying all their sensors in all respects, than DPAF would justify the statement "Sony sensors are behind."

I'm also highly critical of the idea that Canon can't produce a faster readout sensor. I'd be willing to bet they've prototyped a number of them, just like they've prototyped 120mp sensors. If they have patents related to it then it's a guarantee a prototype is out there. When they actually go into production hinges on when they think it's profitable to do so. Since their marketshare has remained steady so far...

I disagree with you assessment.
Disagree all you want. The images are quite clear.

I thought you were talking about the 5DSR.
Fair enough.
 

MayaTlab

EOS 80D
Oct 6, 2015
185
66
The frame rate change could be a clock speed change. That does not preclude other changes.
I believe that there is some degree of relationship between ADC clock speed and DR. Slow clock speed is preferred. The A7RIII's sensor reduced by half the readout speed while improving DR.

Or if a downstream component can't handle it.
Possibly but I don't think so. That said with Sony' A7 series the sensor isn't always the bottleneck (the A7III's sensor, for example, can pump out more than 10fps 14bit shots given its scan time in silent shutter mode).

Canon's cameras, on the other hand... that's the sensor. The M50 has a silent shutter mode and it's quite slow. We'll know soon enough about the R.

That's a fair statement. The problem is that they are typically not described as "lagging in one area", but as lagging in every area.
Well that's not how I would describe it. They're just leading in some areas, behind in others.

DPAF could be seen as "leading". Even if it's mixed blessing right now for stills cameras, it's the future anyway.
If Canon's colour rendering is partly done to Canon's CFA design, then there could be that as well (At best it's only partly since there are so many factors affecting colour rendition that the sensor is only a tiny part of it).
All Canon cameras seem to behave well with long exposures. Cameras using Sony manufactured sensors seem to be either hit or miss - although it doesn't seem to be related much to the sensor's specs (The Pentax K1 seems great while the A7R is rubbish).

But they're lagging in more than one area otherwise :
- DR
- banding patterns and uniformity at low ISOs
- readout speed
- dual gain architecture (noise performance, particularly for small pixels cameras)

IMO the most pressing thing right now for Canon is readout speed, as it affects mirrorless cameras' basic operational qualities. They're doing amazing things to hide these problems (for example, Jared Polin falling for Canon's trick of freezing the EVF image when there's a blackout or showing a slideshow of the previous shot taken, to mask the existence of the actual blackout, something they perfected with the M50), but at some point it will clearly start to look bad for them.

I'm also highly critical of the idea that Canon can't produce a faster readout sensor. I'd be willing to bet they've prototyped a number of them, just like they've prototyped 120mp sensors. If they have patents related to it then it's a guarantee a prototype is out there. When they actually go into production hinges on when they think it's profitable to do so. Since their marketshare has remained steady so far...
The existence of a patent or a prototype doesn't mean that Canon can produce a particular sensor for a specific price. BSI and stacked sensors, which are a requirement for fast readout speeds long term, seem to require to introduce in the sensor production process steps that used to be quite foreign to sensor manufacturing until now. It would probably require a significant investment from Canon to enable these production processes industrially, ie at cost. Not just as a research process. What's remarkable from Sony is that they've actually been able to produce these sensors at cost and put them in actual products.

I have little doubts that Canon could do it. But for them the investment would possibly be way more disproportionate with expected revenues that it's a lot less attractive to plunge for it now.

The silver lining for Canon users is Canon's decision to sell their sensors to third parties. This could help justify the investments a little bit more.

Disagree all you want. The images are quite clear.
Let me be even clearer for you. Below, DPreview's test scene in tungsten mode, 5DSR vs A7RIII. First, without chroma noise correction :

Screen Shot 2018-09-15 at 11.29.34.jpg


Then, with the minimum amount of noise correction possible for each file so that most color blotches are eliminated :

Screen Shot 2018-09-15 at 11.30.19.jpg


See what's happening to the blue pants on the 5DSR file ? Well that's because the blue channel's noise is so elevated that C1's colour noise reduction algorithm has to undersaturated it so that it doesn't modify the grey point. DPP behaves similarly. LR's approach is to let the grey point shift (in general towards magenta).

That blue channel noise issue is far more important than nearly everything else for most people's low light photography. It wreaks havoc at all print sizes, even small ones, and it's most likely to occur under the sort of lights people tend to shoot at higher ISOs.

There's a whole stop of noise performance difference between the two in that regard.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,196
481
...various sensor details...
We're waaay off the point. The original point was that Sony fans and the press tend to treat any change to a Sony sensor as "brand new" while referring to literally brand new Canon sensors as "reused tech." I don't think you disagree with that given some of your other statements. But as for specific sensor details, you and I are way into speculation that neither of us can prove without EE jobs at Sony and Canon.

Well that's not how I would describe it. They're just leading in some areas, behind in others.
A fair and accurate statement.

But they're lagging in more than one area otherwise :
- banding patterns and uniformity at low ISOs
I haven't seen that in any real world post processing in a long time.

They're doing amazing things to hide these problems (for example, Jared Polin falling for Canon's trick of freezing the EVF image when there's a blackout or showing a slideshow of the previous shot taken, to mask the existence of the actual blackout, something they perfected with the M50), but at some point it will clearly start to look bad for them.
On the blackout: isn't that almost everyone except Sony's 3rd gen FF cameras? Who else has truly blackout/slideshow free EVFs? The (now discontinued) Samsung NX1 is the only one that comes to mind.

The existence of a patent or a prototype doesn't mean that Canon can produce a particular sensor for a specific price.
It's a good bet they can do so profitably. There simply aren't wide gaps in semiconductor fabrication. The only time you see such a gap is when there's a patent protecting it, until someone works around the patent.

That's one of the fundamental flaws in the thinking of the author of that German paper. Another being that the highest volume fab will have the latest tech. As a counter example, over the decades Intel has repeatedly had their lunch handed to them on tech by smaller companies with far lower volume. That never mattered in the face of x86 backwards compatibility until smartphones where x86 didn't matter.

Point is: none of this is out of reach for Canon. But management has to approve.

I have little doubts that Canon could do it. But for them the investment would possibly be way more disproportionate with expected revenues that it's a lot less attractive to plunge for it now.
You're looking at one side of the equation and assuming the investment is large/difficult. A small added cost would not interest Canon management now. They're not losing marketshare. Outside of forums the consuming public doesn't seem to care.

I do agree with you that they're going to start to care about video and MILC fps (readout speed if you are correct on that point). I myself skipped the R over severely cropped 4k. I had the cash in hand to preorder, spent it on a lens instead. But until Canon management sees an impact to sales the cost could be $0.25 per unit and they won't spend it. If I have one critique of Canon, it's that the bean counters have too much say in camera body design.

Let me be even clearer for you.
By turning to the shadows of an underexposed and poorly white balanced tungsten image?

Below, DPreview's test scene in tungsten mode, 5DSR vs A7RIII.
Did you grab the wrong file for Sony? Because the Sony pants aren't blue on their site.

Screen Shot 2018-09-15 at 12.20.52 PM.png


Side note: with no NR of any kind it looks like the A7r III may be 1ev better at 12800 with a grossly underexposed tungsten image. The color splotching on the Sony is roughly equal at 25600. If you leave default color NR on the difference drops to <1ev again.

But...I wouldn't use a grossly underexposed 12800 tungsten image for any purpose from either camera. And performance in grossly underexposed tungsten images at 12800 is not relevant to the high ISO needs of anyone I know including pro photographer friends and casual photography family members. If anyone came to me with a Canon 5Dsr 12800 print like this test and complained about blue channel noise, I would not point them to a Sony. I would point them to an online course on exposure.
 

3kramd5

EOS 5D MK IV
Mar 2, 2012
2,927
310
But they're lagging in more than one area otherwise :
- dual gain architecture (noise performance, particularly for small pixels cameras)
If the can license the same scheme from Aptina that Sony did, it’s a fairly simple addition (specifically a switch and a capacitor on the pixel’s floating diffusion node) not likely limited by canon’s processes.
 

tmroper

EOS 80D
Sep 22, 2016
144
45
It wasn't that long ago that Sony earned most of its profits from their insurance businesses, and people started wondering if they'd exit electronics altogether. Now, these days, it's all about sensors. And tomorrow? Who knows with a conglomerate like that.
 

scyrene

EOR R
Dec 4, 2013
2,399
250
UK
www.flickr.com
Looking at this image https://flic.kr/p/HBhGPT I am horrified by the noise at 3200ISO in really good light. Kieth Reeder might be a bit over the top sometimes but he has a lot of very good 3200ISO 7D/7D MkII shots that look much better. Certainly my 1DX MkII 3200ISO images have much less noise than that image. Maybe it is a function of post processing but I am shocked at how weak the IQ is.

Of course that doesn't speak to AF or Daniels comments.
The bird has been oversharpened (or at least the unsharp threshold was set too low) and the background has been subject to a blur filter and/or heavier noise reduction. I can't speak to the original points being made, but that looks like heavy-handed/clumsy postprocessing as much as anything.
 
Reactions: Keith_Reeder

MayaTlab

EOS 80D
Oct 6, 2015
185
66
I haven't seen that in any real world post processing in a long time.
I've recently had a big issue after Canon serviced my 6D. Apparently it seems that the sensor's' natural vertical banding pattern is corrected by a calibration process at the factory, because when Canon gave me my 6D back, I've had very noticeable banding. They had to exchange the camera as they couldn't solve it.

The 5DIV exhibits horizontal banding when pushed too far. Admittedly beyond a point where the image looks rubbish anyway (too much noise), so it's no big deal.

Basically, banding still is something that sometimes rears its ugly head in my experience with Canon, even if it's much less so than it used to be. Meanwhile I've never had any problem of that kind for years with my Fuji cameras (except the original X100).

On the blackout: isn't that almost everyone except Sony's 3rd gen FF cameras? Who else has truly blackout/slideshow free EVFs? The (now discontinued) Samsung NX1 is the only one that comes to mind.
It isn't about a blackout free camera. It's about providing a real time liveview feed in continuous burst, even if it's just between exposures - just like what an OVF looks like when the mirror is down, basically. M43 cameras were the first to provide such a feature, which is critical for framing reasonably well during bursts, and properly tracking a subject. Nowadays I think you would have a hard time finding a mirrorless camera that doesn't provide that feature, even if it's sometimes at a lower fps setting than the maximum (for example : Z6 at 5.5fps, A7III at 8). Except the Canons. None of them are able to do this. As a result what you see in the EVF of the M50 or probably the R, during burst shooting, is what happened more than 150ms ago.

It's a good bet they can do so profitably. There simply aren't wide gaps in semiconductor fabrication. The only time you see such a gap is when there's a patent protecting it, until someone works around the patent.

That's one of the fundamental flaws in the thinking of the author of that German paper. Another being that the highest volume fab will have the latest tech. As a counter example, over the decades Intel has repeatedly had their lunch handed to them on tech by smaller companies with far lower volume. That never mattered in the face of x86 backwards compatibility until smartphones where x86 didn't matter.

Point is: none of this is out of reach for Canon. But management has to approve.

You're looking at one side of the equation and assuming the investment is large/difficult. A small added cost would not interest Canon management now. They're not losing marketshare. Outside of forums the consuming public doesn't seem to care.

I do agree with you that they're going to start to care about video and MILC fps (readout speed if you are correct on that point). I myself skipped the R over severely cropped 4k. I had the cash in hand to preorder, spent it on a lens instead. But until Canon management sees an impact to sales the cost could be $0.25 per unit and they won't spend it. If I have one critique of Canon, it's that the bean counters have too much say in camera body design.
The unknown Towerjazz is reported to be gearing up for production of BSI sensors (granted, with Panasonic's involvement as well, and Panasonic now is a bigger player than Canon when it comes to sensors - Canon's sensors global market share has been dropping like a stone), so it's probably something Canon could do as well. But so far ? Zilch. Nada. Clearly they're not willing to spend the money. Sony was, though. To them it made sense. Probably, as the article suggest, because they're producing sensors for applications where these technological advances are more easily felt. And I think that's what the article is trying to convey : Canon is unwilling to spend the money because they're only looking at their camera / imaging products. Sony is, because, even if they possibly wouldn't invest just because of their camera business, they would because of their other clients (smartphones, automotive, etc.). Basically, Sony cameras benefit from Sony's wider sensor business.

We can't deny Canon's public declaration of intent to sell their sensors to third parties in the future. They're clearly seeing something interesting for them beyond their own imaging business.

By turning to the shadows of an underexposed and poorly white balanced tungsten image?
You know very well that the DPreview test scene isn't underexposed overall (clipping indicator in C1) :
Screen Shot 2018-09-16 at 00.43.49.jpg
And why shouldn't I look at what the shadows look like in a picture ? Aren't they part of the bloody picture too ?
Besides I neutralised white balance in the example above. By "poorly white balanced", are you actually suggesting that one can have the luxury to change the light bulbs at a venue when they fancy ?
BTW, the problem actually gets worse if the white balance is neutralised. That's because under tungsten lighting, it means boosting the blue channel relative to the others. If its noise floor is high, you start to run into the phenomenon that you can see in the picture above.

Did you grab the wrong file for Sony? Because the Sony pants aren't blue on their site.
Are you colour blind ? No really, I sincerely ask the question. In the example that I showed, the guy on the right is clearly wearing blue pants. They turn to grey on the Canon because that's how C1 or DPP handle that blue channel noise (by desaturating it, to keep greys reasonably neutral).
 

MayaTlab

EOS 80D
Oct 6, 2015
185
66
If the can license the same scheme from Aptina that Sony did, it’s a fairly simple addition (specifically a switch and a capacitor on the pixel’s floating diffusion node) not likely limited by canon’s processes.
Well, then, note to Canon : I'm waiting :D.
 

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,422
444
119
The bird has been oversharpened (or at least the unsharp threshold was set too low) and the background has been subject to a blur filter and/or heavier noise reduction. I can't speak to the original points being made, but that looks like heavy-handed/clumsy postprocessing as much as anything.
I agree, I downloaded the 3200 iso RAW files for both the 1DX MkII and the A9 from DPReview (so it was a fair comparison), and they looked very similar to each other.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,196
481
...so it's probably something Canon could do as well. But so far ? Zilch. Nada. Clearly they're not willing to spend the money.
Why would they be when they hold majority marketshare by a wide margin?

Sony was, though.
What else were they going to do to try and gain marketshare against the big two?

You know very well that the DPreview test scene isn't underexposed overall (clipping indicator in C1) :
You know very well that the part of the DPReview tungsten scene which you choose IS underexposed. They basically light one side of the tungsten scene. You pick the unlit side to complain about noise, then pick the lit side to say it's 'not underexposed.' If you need to play games to support your claim, your claim is not true, or not true outside very narrow margins.

Screen Shot 2018-09-15 at 11.39.44 PM.png


And why shouldn't I look at what the shadows look like in a picture ?
If pixel peeping deep shadows of poorly lit tungsten shots at 12800 is your thing, knock yourself out. Just don't expect that to be meaningful to actual photography.

Are you colour blind ? No really, I sincerely ask the question.
I had the wrong section under the magnifier when I wrote that. Should be obvious from the screenshot but rather than be an adult about it you decide to be a child.

They turn to grey on the Canon because that's how C1 or DPP handle that blue channel noise.
ACR doesn't.
 

MayaTlab

EOS 80D
Oct 6, 2015
185
66
What else were they going to do to try and gain marketshare against the big two?
What you don't get and what this article suggests is that Sony being able to produce BSI or stacked sensors at cost isn't because they're scrambling to gain market share or whatever. It's because they have 40% of the sensor market share and have had to develop these technologies anyways to make sensors for smartphones, automotive applications, robots, etc. Basically, the investment to produce such sensors would have been made anyway, regardless of whether Sony makes cameras or not. They could drop tomorrow their Alpha business and they'd still invest in sensor manufacturing processes.

Canon's motivation to invest, on the other hand, is 100% reliant on their own products.

You know very well that the part of the DPReview tungsten scene which you choose IS underexposed. They basically light one side of the tungsten scene. You pick the unlit side to complain about noise, then pick the lit side to say it's 'not underexposed.' If you need to play games to support your claim, your claim is not true, or not true outside very narrow margins.

View attachment 180461

If pixel peeping deep shadows of poorly lit tungsten shots at 12800 is your thing, knock yourself out. Just don't expect that to be meaningful to actual photography.

ACR doesn't.
What ACR does, and what you can already see in the very picture you posted on your own, above, and without even pixel peeping, is that the areas that should be dark and black / grey, are in fact dark blue / purple / magenta. Don't you see that ?

Screen Shot 2018-09-16 at 10.31.16.jpg


This is why DPP and C1 desaturate the blue channel in these conditions. To keep greys neutral.

And this is why I consider elevated blue channel noise the n°1 problem at higher ISOs : you can still see its effect in a thumbnail.
 

dtaylor

Canon 5Ds
Jul 26, 2011
1,196
481
What you don't get...
I get it and I don't fully agree.

What ACR does, and what you can already see in the very picture you posted on your own, above, and without even pixel peeping,
I never posted a picture developed in ACR. I've posted screenshots from dpreview. And I'm tired of nitpicking this point. The A7r III's 25600 RAWs are not as good as the 5Dsr's 12800 RAWs which means it's <1ev ahead at high ISO.

Isolate the blue channel shot under dim tungsten light with a full moon and Jupiter and Saturn aligned. It doesn't change a thing.
 

jeffa4444

EOS 6D MK II
Feb 28, 2013
1,415
67
65
JonSnow premise is about quantity to amortise R&D costs and this now hinders Canon. Red Camera have much smaller quantities yet offer at least three different sensors, all at the forefront of cinematography, with great dynamic range. Canon still represents the majority of professional photographers making a living from their tools because the camera is only half of the story the other being lenses. The new RF 50mm f1.2L lens have been compared to the EF 50mm f1.2L II with much better control of chromatic aberrations resulting in cleaner images. This is partially because the new mount allows modern concentric (parallel light) type designs to be employed with a closer back-focus. Similarly the technical challenges are not simply resolution or dynamic range but colour space & colour imagery and here Canon has a much better record than Sony. Add the benefits of dual pixel technology and Canon offer something different not a "me too" product and for that we should all be grateful because it retains choice.
 

jeffa4444

EOS 6D MK II
Feb 28, 2013
1,415
67
65
I get to play with Sony cameras we own for rental. Everything from the A9 down. They make great shots but frankly they dont move the game on anywhere like you would think looking at specifications. We also rent Nikons. Good equipment, solid performers but for every one Sony we rent, we rent four Canon cameras, for every Nikon we rent we rent three Canon cameras.
Canon service is second to none, reliability and endurance legend. The Sony cameras are not as robust and whilst the "G" lenses are great optics, they dont really outshine Canon in the day to day world away from MTF machines and CIPA test charts. Ive just returned from South Africa on business, the rental companies there dont really rent Sony, they do rent Nikon but its Canon that pays the rent whether shooting fashion or animals in the wild. Cant argue with profit.
 

3kramd5

EOS 5D MK IV
Mar 2, 2012
2,927
310
JonSnow premise is about quantity to amortise R&D costs and this now hinders Canon. Red Camera have much smaller quantities yet offer at least three different sensors, all at the forefront of cinematography, with great dynamic range.
Hasselblad and PhaseOne have great sensors too. Like Red, they don’t make them.
 

jeffa4444

EOS 6D MK II
Feb 28, 2013
1,415
67
65
That's unlikely to be simple clock speed change as DR increased with the A7RIII.



Which one do you think will be felt the most incremental between :
- going from 4,3fps to 5fps in servo, with still no liveview feed in between frames in continuous drive,
- or going from 3fps with AF/AE/liveview feed with 12bit files to 8fps with AF/AE/liveview feed with 14bit files ?




It's just logic. The only reason you drop the bit depth in continuous mode on a camera that can produce 14bit files otherwise is if you want to halve the readout speed. And the only reason you'd do this on a mirrorless camera is to allow it to do what it relies on its sensor to do (AF, AE, liveview feed).



You wouldn't have been had you known how readout speed is a problem for Canon.



Gave you a good reason why it isn't (can't read the sensor for both image information and phase information at the same time). One reason why you aren't getting AF in 120fps @720p. It's just a mixed bag, right now, on stills cameras.

DPAF may be complex from an engineering point of view. But it doesn't seem to require techniques that Canon will eventually have to adopt if they want to increase the readout speed of their sensors. That Canon's sensors show design complexity in one area doesn't mean that they show complexity in the one area where they're currently lagging.



I disagree with you assessment. The reason I always bring up the issue of blue channel noise under warmer TCs is that not only is this is the sort of lighting where you're very likely to be actually taking pictures, it's also something that will affect how a picture looks regardless of its print size. The way Lightroom deals with this is by turning darker areas magenta. The way C1 and DPP deal with this is by reducing the blue channel's saturation - resulting in a loss of colours.



I thought you were talking about the 5DSR. Indeed the 5DIV is more competitive in measured terms. But it's got banding at extreme pushes. That said the 5DIV is enough for me.
Show examples on the banding on the 5D MKIV? We rent tons of them and Ive never seen it or heard of it.
 

3kramd5

EOS 5D MK IV
Mar 2, 2012
2,927
310
PhaseOne and Hasselblad use Sony sensors, Red design their own sensors and independent foundries make them (not Sony).
Right*, hence they aren’t really applicable to the discussion. They don’t face the capital burden of continuously improving fabrication.


*Well, I don’t know much of the design Red does in house as opposed to flowing specifications to, reportedly but not confirmed, CMOSIS.