Initial testing shows Canon’s new 32.5mp APS-C sensor improves dynamic range over predecessor

LSXPhotog

EOS RP
Apr 2, 2015
381
204
www.diossiphotography.com
Can you substantiate that it's "all but confirmed" or is this yet another echo of that ambiguous statement DP Review claims to have heard?
"All but confirmed" means it's not substantiated. LOL However, Imaging Resource and Camera Labs stated the reason the EVF is included in the kit is because of this reason and that's what Canon told them. You'd think that all these "journalists" would have done their jobs and asked questions... journalism is dead. It's now just people who can write and want free stuff. LOL (I'm an actual journalist)
 
  • Like
Reactions: pj1974

AlanF

Canon 5DSR II
Aug 16, 2012
5,758
3,111
"All but confirmed" means it's not substantiated. LOL However, Imaging Resource and Camera Labs stated the reason the EVF is included in the kit is because of this reason and that's what Canon told them. You'd think that all these "journalists" would have done their jobs and asked questions... journalism is dead. It's now just people who can write and want free stuff. LOL (I'm an actual journalist)
Judging by the recent reviews of Sony and Canon offerings, camera journalists in the main just people who can cut-and-paste and want clicks and free stuff.
 

yeahright

EOS M50
Aug 28, 2014
43
14
I do not understand the claim that using DPRaw in the 5D4 can be used to improve DR by one stop. After all it's equal to doubling the resolution, which also doesn't improve DR of same-generation sensors. Or is it that the dual pixel raw storage implementation in the 5D4 is simply buggy?

In


it appears that even though the 'auxiliary' raw file is not at its maximum, the 'main' raw file is - which points to a storage (too few bits) and not a sensor issue.
 

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,912
1,054
119
I do not understand the claim that using DPRaw in the 5D4 can be used to improve DR by one stop. After all it's equal to doubling the resolution, which also doesn't improve DR of same-generation sensors. Or is it that the dual pixel raw storage implementation in the 5D4 is simply buggy?

In


it appears that even though the 'auxiliary' raw file is not at its maximum, the 'main' raw file is - which points to a storage (too few bits) and not a sensor issue.
Its just maths.

If any number over 10,000 is pure white, if you are adding two pixels/one dual pixel readout it is the sum of the two, if each records 6,000 the accumulated reading is 12,000, a blown highlight, if you also record the half pixels you get a reading of 6,000, which is between 5,000 and 10,000, an extra stop of information in the highlights.

So any time the half pixel reads between 5,000 and 10,000 you get additional information. It's got nothing yto do with resolution, it is just the maths from counting the elections and the fact that any number higher then the highest possible is white, it doesn't matter if it is 1 higher or 1,000,000 higher, it is just white.
 
Last edited:

yeahright

EOS M50
Aug 28, 2014
43
14
Its just maths.

If any number over 10,000 is pure white, if you are adding two pixels/one dual pixel readout it is the sum of the two, if each records 6,000 the accumulated reading is 12,000, a blown highlight, if you also record the half pixels you get a reading of 6,000, which is between 5,000 and 10,000, an extra stop of information in the highlights.

So any time the half pixel reads between 5,000 and 10,000 you get additional information. It's got nothing yto do with resolution, it is just the maths from counting the elections and the fact that any number higher then the highest possible is white, it doesn't matter if it is 1 higher or 1,000,000 higher, it is just white.
Of course you need an additional bit to store the information. When adding two 14 bit values you must provide storage for a 15 bit number for the result, otherwise you might get overflow. So in your example, if the two sub-pixels each record a value of 6000, you shouldn't cram the result in a data format that supports only 10000 as maximum value, but instead store the correct result as 12000. Failing to do so is a buggy implementation of storage, but not an increase in sensor DR.
And if you divide a pixel in half, the signal to noise ratio decreases by a factor of two, that is, when you divide a pixel in two pixels, you also reduce the usable DR by one stop in the sub-pixels. If you add them together you get that one stop back, but it is the same DR you would get if you hadn't split the pixel in half in the first place...!?
 
  • Like
Reactions: AlanF

Sharlin

EOS 6D MK II
Dec 26, 2015
1,061
559
Turku, Finland
So any time the half pixel reads between 5,000 and 10,000 you get additional information. It's got nothing yto do with resolution, it is just the maths from counting the elections and the fact that any number higher then the highest possible is white, it doesn't matter if it is 1 higher to 1,000 higher, it is just white.
But only if the FWC of the (sub-)photosites has enough headroom, which is then retained in the A/D conversion, right? It seems to me that a priori the subphotosites can't have individual FWCs equal to that of a single larger pixel, otherwise that would basically mean that the native ISO of the sensor is basically 50 (ignoring the fact that the subimages are subtly different!) So if a large pixel saturates at 10000, two subpixels with half the area should saturate at 5000, all else being equal.
 

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,912
1,054
119
Of course you need an additional bit to store the information. When adding two 14 bit values you must provide storage for a 15 bit number for the result, otherwise you might get overflow. So in your example, if the two sub-pixels each record a value of 6000, you shouldn't cram the result in a data format that supports only 10000 as maximum value, but instead store the correct result as 12000. Failing to do so is a buggy implementation of storage, but not an increase in sensor DR.
And if you divide a pixel in half, the effective noise amplitude increases by a factor of two, that is, when you divide a pixel in two pixels, you also reduce the usable DR by one stop in the sub-pixels. If you add them together you get that one stop back, but it is the same DR you would get if you hadn't split the pixel in half in the first place...!?
Nevertheless that is how they arrive at the numbers for the two files in the dual RAW system. The main file is the sum of the two sub pixels, the secondary file is the reading from one set of sub pixels. That is why the benefit seen is not in the shadows, the traditional improvement of lowering the noise floor, but in the shadows as it is adding extra range of tonality at the top.

Of course this means the benefit is very scene and exposure specific, if the main RAW file has no blown pixels then there is zero advantage to having dual RAW files.
 

yeahright

EOS M50
Aug 28, 2014
43
14
Nevertheless that is how they arrive at the numbers for the two files in the dual RAW system. The main file is the sum of the two sub pixels, the secondary file is the reading from one set of sub pixels. That is why the benefit seen is not in the shadows, the traditional improvement of lowering the noise floor, but in the shadows as it is adding extra range of tonality at the top.

Of course this means the benefit is very scene and exposure specific, if the main RAW file has no blown pixels then there is zero advantage to having dual RAW files.
Ok, but we can agree that the one extra stop of DR in certain situations is due to the very specific implementation of DPRaw in the 5D4, but it is not inherent to the DPRaw technology itself.
 

Timedog

EOS R
Aug 31, 2018
49
28
Quick question since you guys seem to be knowledgeable, I *thought* that all the non medium format cameras had 14bit ADC's (maybe I'm wrong) but how are Sony sensors claiming 15 stop DR if the 14-bit ADC physically cannot produce this? Seems like maybe it would be possible with two separate files or by halving resolution a la dpRAW, But then wouldn't Sony's normal RAW files have to be dpRAW sized?

Nevertheless that is how they arrive at the numbers for the two files in the dual RAW system. The main file is the sum of the two sub pixels, the secondary file is the reading from one set of sub pixels. That is why the benefit seen is not in the shadows, the traditional improvement of lowering the noise floor, but in the shadows as it is adding extra range of tonality at the top.

Of course this means the benefit is very scene and exposure specific, if the main RAW file has no blown pixels then there is zero advantage to having dual RAW files.
You wrote shadows there twice and I think you meant to say highlights? Having trouble deciphering your comment.
 
Last edited:

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
520
373
Australia
www.michaelborisenko.com
For two reasons. One, the image with less noise has less to gain for any given NR setting, and NR is not an unlimited adjustment as there is a point where it starts to destroy detail (luminance NR) or color (color NR).

Two: human beings view photographs and human beings are not scientific instruments. The point where shadow detail becomes unusable in a print is the point where noise stands out to the viewer. As noise moves below that point it rapidly becomes invisible to the viewer. It's the difference between pixel peeping an underexposed black sock and admiring a print of Yellowstone or Yosemite.
Basically you've provided a long explanation on why noise reduction works for Canon but doesn't work for Sony! But in fact an image with 0.8 stop better DR will also have noisy shadows, just 0.8 stops deeper. Why aren't we able to apply noise reduction to those shadows??

Fine. Show me your real world image that illustrates how noticeable it is. Underexposed black cats and black socks do not count.
My suggestion to play with exposure slider was just to illustrate, for those who's interested, what is this 0.8 stop difference so that they can see it with their own eyes. It can be done on any image with deep enough shadows. It's just to get the feeling what 0.8 stop difference is, not convince you that 0.8 stops is significant.
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
520
373
Australia
www.michaelborisenko.com
Quick question since you guys seem to be knowledgeable, I *thought* that all the non medium format cameras had 14bit ADC's (maybe I'm wrong) but how are Sony sensors claiming 15 stop DR if the 14-bit ADC physically cannot produce this? Seems like maybe it would be possible with two separate files or by halving resolution a la dpRAW, But then wouldn't Sony's normal RAW files have to be dpRAW sized?
With a linear ADC, it's impossible to get more than 14 stops of DR from a 14-bit ADC.
The only way Sony can do it is to use a non-linear ADC. It will require additional digital processing to convert a non-linear 14-bit value to a linear 15-bit. Roughly speaking, it'll have to stretch it out. 14 bits can hold up to 16384 different values, 15 bits - 32768. So the range [0..16383] will need to be stretched to [0..32767], but not just by multiplying by 2 - it has to be non-linear and stretch the higher values more than lower values.

The result will lose some smoothness but that won't be noticeable I guess. That's the case if it does per-pixel conversion independently, but it may also involve using data from adjacent pixels.

I don't know if Sony really does that but that's perhaps the only way, if their claim about 15 stops of DR is true.
 
  • Like
Reactions: Timedog

Lurker

EOS 80D
Dec 8, 2012
159
20
Failing to do so is a buggy implementation of storage
No, it's just putting a limit on an open ended system. As long as the sensor is exposed to light it will count. The final result is from 0-infinity. No camera supports infinity, every camera can have blown highlights.
 

canonnews

EOS RP
Dec 27, 2017
340
295
Canada
www.canonnews.com
Ok, but we can agree that the one extra stop of DR in certain situations is due to the very specific implementation of DPRaw in the 5D4, but it is not inherent to the DPRaw technology itself.
no, it's due to the technology and how it has to be implemented. From what i gather;

the two halfs of the pixel may not receive the same amount of light because the sub pixels are seperated by 180 degrees of phase.
all sorts of bad things happen when 1/2 of the pixel oversaturates, and the other half does not, yet the total sum is less than the total pixel saturation level. Ie: total pixel is 10,000, A side is saturated at 5000, and B side has 3000. The color value is thus completely random. So they made it so that A could be 5750 (actual), and B could be 3000 and still be 8750 representing a true value under 10,000.

thus Canon built in some slop to ensure that this doesn't happen in most cases.
 

SteveC

M50 & T6i
Sep 3, 2019
352
204
"All but confirmed" means it's not substantiated. LOL However, Imaging Resource and Camera Labs stated the reason the EVF is included in the kit is because of this reason and that's what Canon told them. You'd think that all these "journalists" would have done their jobs and asked questions... journalism is dead. It's now just people who can write and want free stuff. LOL (I'm an actual journalist)
You did at least give a source that wasn't DP review--which is what I was looking for. We might be working off two different senses of the word "substantiated." I don't use the word to mean "certainty" but rather more like "there is some real evidence for this." You at least gave some reason (beyond DP Sony Shill Reviews) to believe it's true, and that gives it some substance. It at least somewhat backs up "all but confirmed" which came across to me as pretty confident and a phrase that a reasonable person wouldn't use without solid reason to believe it.

Even given this, though, I don't believe it; I am hoping a real journalist can ask a more direct question and push for a clear answer.

Especially since the Canon rep's (mis?)quoted logic proceeds from a false assumption: The EVF is NOT included with the camera body-only (you only get it if you get a kit lens with the camera). If it truly were intended to replace/succeed the M5, they would have included it with all options, since without a VF the camera body-only can't replace the M5 body-only.

(As I already have one of those crappy 15-45 lenses (and never use it; I should probably just take it out of the camera bag) and have a Tamron 18-200, i have no need of the 18-150 kit either, so I'm SOL for getting the EVF with my M6 II. Which I will buy--once I am reasonably confident there won't be an M5 II. Maybe I can buy one of the kits at my local brick-and-mortar shop, and get enough money back selling the lens back to them as top condition unused to reduce the effective price of the EVF to below its stand-alone retail.)
 

koenkooi

EOS 7D MK II
Feb 25, 2015
493
291
(As I already have one of those crappy 15-45 lenses (and never use it; I should probably just take it out of the camera bag) and have a Tamron 18-200, i have no need of the 18-150 kit either, so I'm SOL for getting the EVF with my M6 II. Which I will buy--once I am reasonably confident there won't be an M5 II. Maybe I can buy one of the kits at my local brick-and-mortar shop, and get enough money back selling the lens back to them as top condition unused to reduce the effective price of the EVF to below its stand-alone retail.)
Same here, I already have an 15-45 and the deal for the 18-150 isn't that good since it's basically full price body + full price lens + "free" EVF. The white box version of that lens is already a lot cheaper and looking at the grey market it's even less. So for a savvy shopper it's more "half price EVF" and "free EVF".
So I preordered the M6II body only and will see after a few weeks if I actually want the EVF. Hopefully there well be white box versions of it by then :)

(The M6II release date is also on day 3 of the family vacation abroad, so I'm slightly miffed about that)
 

Don Haines

Beware of cats with laser eyes!
Jun 4, 2012
8,153
1,687
Canada
I call that hamburger. Great with cheese and bacon. ;)
But I need a camera that can film my hamburger in 24p, and that has enough DR to simultaneously show the details on the white cheddar in the sunlight while the blackened bacon in the shadows still shows it texture! OH THE HORROR!!!!!

BTW, you can not film the dog eating that hamburger in 24p because all the action occurs in less than one frame :)
 
Last edited:

Don Haines

Beware of cats with laser eyes!
Jun 4, 2012
8,153
1,687
Canada
With a linear ADC, it's impossible to get more than 14 stops of DR from a 14-bit ADC.
The only way Sony can do it is to use a non-linear ADC. It will require additional digital processing to convert a non-linear 14-bit value to a linear 15-bit. Roughly speaking, it'll have to stretch it out. 14 bits can hold up to 16384 different values, 15 bits - 32768. So the range [0..16383] will need to be stretched to [0..32767], but not just by multiplying by 2 - it has to be non-linear and stretch the higher values more than lower values.

The result will lose some smoothness but that won't be noticeable I guess. That's the case if it does per-pixel conversion independently, but it may also involve using data from adjacent pixels.

I don't know if Sony really does that but that's perhaps the only way, if their claim about 15 stops of DR is true.
It's been done before, and quite often in audio.

Remember Companders? Audio lines would have a frequency range of 300hz to 3000hz, so you would compress a wider range into that line, and at the other end you would expand it back to the original. It would give you a better frequency range, but at the expense of quality. This is a similar concept.
 
  • Like
Reactions: Quarkcharmed