Dynamic range testing of the Canon EOS R3 is complete

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
Well, to me and many others I know not into the left brain side of art and photography it comes across as mental masturbation and incredibly dull but hey, I'd never ask for everyone to be just like me so carryon.
People can take wonderful photos and be true artists with only the slightest, if any, knowledge or interest in how a camera works. But, if you know the limitations of your gear, its sweet and weak points and how it works, you can get the most of out of it. And believe it or not, some people actaully have a genuine interest in how things work and also enjoy the artistic as well as technical aspects, which is why we really appreciate what the types such as Bill Claff do by working hard, measuring and explaining things to us.
 
  • Like
  • Love
Reactions: 11 users
Upvote 0
This is a question for someone with far more technical knowledge than I have, but I believe from past things I've read, the reason for it is that without the mechanical shutter closing after each exposure, stray light can impact the image following the exposure, introducing slightly more read noise. That's why you see the option for "Electronic First Curtain shutter," which still has a secondary mechanical shutter, and does not impact DR at all. That way, with the mechanical shutter closed, there is no additional light to generate read noise/heat on the sensor circuits for the brief milliseconds it takes to process the exposure.

Obviously there are ways to manage this, since Canon has improved the ES dynamic range in every release of a camera, but it's still slightly visible in the EOS R3 charts. It really doesn't make a difference at this point, though. Previously, Canon was also limiting Electronic shutter on the R5/R6 to 12-bit readout, but I'm sure with all the marketing around the uncompressed raw at 30 fps, they aren't limiting that anymore.
Thank you! Very interesting.
 
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
Well, to me and many others I know not into the left brain side of art and photography it comes across as mental masturbation and incredibly dull but hey, I'd never ask for everyone to be just like me so carryon.
Thank goodness not everyone has to be like you in this regards.

Without humans who are interested in technology and engineering and appreciate observing, measuring and understanding the technical aspects of this world, we wouldn't even be in the stone age yet.

The quality of an image may be something that goes beyond objectively measurable numbers. But nonetheless these numbers have meaningful implications for the images produced by the measured device.

You yourself ask how people like their images, which may be a meaningful piece of information to you. But people perceive and say all kind of silly things, and without the methods of quantifying what ever we can and considering statements by people in the context of these numbers, it becomes really easy to believe things that are out of line with reality.

As Alan said, this is valuable information, and it is factual. You may not have a use for it, but that does not devalue it.

The efforts of the engineers at Canon have paid of greatly with the technical performance of the R5 and R3. If the artist gave no thought at all to these technicalities, the engineers' work goes unappreciated, while still making life easier for the artists if they happen to stumble into a scene that demands capable technology.
 
  • Like
Reactions: 8 users
Upvote 0

DBounce

Canon Eos R3
May 3, 2016
500
544
I believe super fast readout, stacked sensors like the R3 has are not quite as good as a slower readout sensor when it comes to DR and noise performance.

Compare Sony A9ii to Sony A7Riii or iv and I believe the slower reading, higher MP sensors also have more dynamic range.

It’s actually a great achievement that the sensor of the R3 can match the image quality of the R5 IMO.

I’m sure someone will correct me, if these assertions are not quite right?

cheers
I’m not surprised at all that the R3 has better dynamic range than the R5… that’s simply progress. I would be disappointed if it didn’t.
From recent articles, the R3 beats Sony and Nikon in dynamic range. The R3 also manages to become the new low-light king. This is outstanding!
 
Upvote 0
Jan 29, 2011
10,675
6,121
I’m not surprised at all that the R3 has better dynamic range than the R5… that’s simply progress. I would be disappointed if it didn’t.
From recent articles, the R3 beats Sony and Nikon in dynamic range. The R3 also manages to become the new low-light king. This is outstanding!
It depends how you read the results. It is equally valid to say the only reason the R3 DR is ‘better’ than the older and more pixel dense R5 is because the R3 cooks it’s RAW files. Any progress made in the extra time to develop went into readout speed (not a small thing for some people) and cooking lessons.
 
  • Like
  • Haha
Reactions: 5 users
Upvote 0

DBounce

Canon Eos R3
May 3, 2016
500
544
It depends how you read the results. It is equally valid to say the only reason the R3 DR is ‘better’ than the older and more pixel dense R5 is because the R3 cooks it’s RAW files. Any progress made in the extra time to develop went into readout speed (not a small thing for some people) and cooking lessons.
’Cooks it’s raw files” … where are you getting this from? I’m not going to argue the point, instead I’ll merely refer you to the the following:

Canon Eos R3 Advanced User Guide:
Page 152
”RAW images are raw data from the image sensor that are recorded to the card digitally as RAW or CRAW.”


From your statement you either believe:
A. You know more than Canon about the Eos R3
or
B. Canon is lying about the RAW files

I would like to suggest a third option:
C. You are mistaken

Canon consistently makes the top 5 for most patents filed worldwide each year. Is it really that hard to believe they might come up with a few improvements?
 
Last edited:
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
From your statement you either believe:
A. You know more than Canon about the Eos R3
or
B. Canon is lying about the RAW files

I would like to suggest a third option:
C. You are mistaken

Canon consistently makes the top 5 for most patents filed worldwide each year. Is it really that hard to believe they might come up with a few improvements?
There is a fourth option:
D. @privatebydesign has read the website where the measurements were made and you either haven't or missed it there and in the OP.
This is written in the image in the OP (sorry to repeat the image): The triangles down indicate noise reduction.


Screenshot 2021-12-11 at 17.46.09.png
 
  • Like
Reactions: 2 users
Upvote 0
There is a fourth option:
D. @privatebydesign has read the website where the measurements were made and you either haven't or missed it there and in the OP.
This is written in the image in the OP (sorry to repeat the image): The triangles down indicate noise reduction.


...
Yes, and posting the results was delayed as I double-checked and agonized over this.
I don't have a good way to quantify noise reduction so the down triangle only serves as an indication that it is present.
 
  • Like
Reactions: 7 users
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
’Cooks it’s raw files” … where are you getting this from? I’m not going to argue the point, instead I’ll merely refer you to the the following:

Canon Eos R3 Advanced User Guide:
Page 152
”RAW images are raw data from the image sensor that are recorded to the card digitally as RAW or CRAW.”


From your statement you either believe:
A. You know more than Canon about the Eos R3
or
B. Canon is lying about the RAW files

I would like to suggest a third option:
C. You are mistaken

Canon consistently makes the top 5 for most patents filed worldwide each year. Is it really that hard to believe they might come up with a few improvements?
In the section you quote Canon explicitly mentions CRAW, which is absolutely cooked (It uses LOSSY compression).

In the past Canon also had no problem misleading people into thinking mRAW and sRAW had anything to do with proper RAW formats.

And dual pixel RAW hasn't made a return since the 5D IV, has it? So really, the processing step that reduces two physical photosites into one digital value is already baked in for all recent Canon cameras.

Apparently Canon thinks adding a sprinkle of noise reduction into the mix is a good idea.

Just as I'm not aware of anybody having found a meaningful impact on picture quality from using CRAW, I doubt anybody has noticed any artifacts from the noise reduction yet. But that's probably why it is there.
 
  • Like
Reactions: 4 users
Upvote 0
Jul 28, 2015
3,368
570
’Cooks it’s raw files” … where are you getting this from? I’m not going to argue the point, instead I’ll merely refer you to the the following:

Canon Eos R3 Advanced User Guide:
Page 152
”RAW images are raw data from the image sensor that are recorded to the card digitally as RAW or CRAW.”


From your statement you either believe:
A. You know more than Canon about the Eos R3
or
B. Canon is lying about the RAW files

I would like to suggest a third option:
C. You are mistaken

Canon consistently makes the top 5 for most patents filed worldwide each year. Is it really that hard to believe they might come up with a few improvements?
The camera converts analogue signal (light hitting the sensor) into digital signals of 1's and 0's. And when the light hits the sensor to create an electronic signal the manufacturer needs to program in-camera the algorithms that define what is a valid signal and what is not. There is no such thing as 'raw' data if you define 'raw data' as in 'that is the data that came into the camera through the lens and that is what I have now recorded' - simply because (unlike film) you are not actually recording light. Sony provide an excellent example of that with their 'star eater' software which some astro photographers are very frustrated by where it interprets faint starlight as 'noise'.

You can hear the same thing in audio systems - different DACs will give a different sound to the system because of the way it is programmed to 'interpret' a digital system to analogue output. Or even how different ADCs will convert the sound from a vinyl record to digital output for the speakers.

So yes, the internal program any manufacturer uses is 'cooking' the data. There was a suspicion a few years ago (can't remember which model) that Sony's 'improved' dynamic range was also more a software advance than a real technological (sensor based) advance.
Cynically you could say that (historically, anyway) that Canon tried to give the user data that was as interfered with as little as possible so the user had maximum ability to get what they wanted, while Sony tried to give something that was most pleasing to the user with less work to do. It looks like Canon may not be edging towards that more cynical end of things.
All 'maybe', of course....
 
  • Like
Reactions: 2 users
Upvote 0
The camera converts analogue signal (light hitting the sensor) into digital signals ... There is no such thing as 'raw' data if you define 'raw data' as in 'that is the data that came into the camera through the lens and that is what I have now recorded' ...
The raw data is the value as it was read from the Analog to Digital Converter (ADC) probably with some offset (Black Level).
Plenty of cameras appear to report this raw data in their raw files.

If the value is manipulated in the firmware before it is written to the then it is "cooked".
Occasional manipulation of isolated pixels such as bad photosite repair isn't considered "cooking".
Changing every (or most) pixels based on inspecting their neighbors is "cooking".

Nearest neighbor algorithms are detectable using a variety of mathematical tools to inspect "raw" files.
My tools can indicate whether an algorithm was applied.
 
  • Like
Reactions: 3 users
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
The raw data is the value as it was read from the Analog to Digital Converter (ADC) probably with some offset (Black Level).
Plenty of cameras appear to report this raw data in their raw files.

If the value is manipulated in the firmware before it is written to the then it is "cooked".
Occasional manipulation of isolated pixels such as bad photosite repair isn't considered "cooking".
Changing every (or most) pixels based on inspecting their neighbors is "cooking".

Nearest neighbor algorithms are detectable using a variety of mathematical tools to inspect "raw" files.
My tools can indicate whether an algorithm was applied.
Does that near neighbour cooking lower detail as usually found with noise suppression methods?
 
Upvote 0
Does that near neighbour cooking lower detail as usually found with noise suppression methods?
Yes, but this is where the inability to quantify the amount of noise reduction hampers our ability to estimate the smoothing (loss of detail).
People seem happy with the results so it's probably academic.
 
  • Like
Reactions: 5 users
Upvote 0
Jan 29, 2011
10,675
6,121
Does that near neighbour cooking lower detail as usually found with noise suppression methods?
Now that is an interesting question! User feedback I have seen suggest the 24mp R3 is giving resolution output more comparable to earlier 30mp Canon sensors. So how much do the AA filter stack and NR cooked RAW files actually add up to...
 
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
Now that is an interesting question! User feedback I have seen suggest the 24mp R3 is giving resolution output more comparable to earlier 30mp Canon sensors. So how much do the AA filter stack and NR cooked RAW files actually add up to...
I've seen the same reports - they compare it with the 5DIV. But, the 5DIV doesn't stack up well against 24 Mpx sensors from its rivals. Canon has subsequently from the 5DIV weakened its AA-filters, as well as going over to a new type for the 1DXIII and R5. Both the R5 and R6 have asymmetric filters, with hardly any filtering in the vertical direction and weakened in the horizontal. As a result, the R5 45 Mpx sensor is very similar in measured resolution to the Nikon Z7 and D850 45.7 Mpx sensors that are without a filter. I found the R5 to be very similar to the D850 and if anything better than the 50 Mpx 5DSR (with its self-cancelling AA-filter), which Canon had claimed. Canon has really leapt forward with its sensors.
 
  • Like
Reactions: 4 users
Upvote 0

DBounce

Canon Eos R3
May 3, 2016
500
544
There is a fourth option:
D. @privatebydesign has read the website where the measurements were made and you either haven't or missed it there and in the OP.
This is written in the image in the OP (sorry to repeat the image): The triangles down indicate noise reduction.


View attachment 201614
This noise reduction ”cooked raw” is nonsense. Case in point, what happens when you add the Sony A7S3 to the chart? A camera that is well established to apply rather aggressive noise reduction at high base ISO 12800. Well from their chart it certainly doesn’t appear to show any noise reduction being applied until a lofty 102,000 ISO. Please explain?… I’ll save you the time… their speculation on noise reduction being applied is just that… speculation.

This might sound radical, but I think I’ll choose to trust Canon over… some website I’ve never heard of before?
 
  • Haha
Reactions: 1 user
Upvote 0
Jul 21, 2010
31,088
12,851
This might sound radical, but I think I’ll choose to trust Canon over… some website I’ve never heard of before?
Really? Then why did you suggest that Canon’s 30 fps spec for the R3 was just marketing hype, based on some videographer’s YouTube post where he was probably just using a shutter speed too slow for 30 fps because he didn’t RTFM? I now notice that he subsequently edited his post, "Update: the claimed 30 fps is only available with a shutter speed greater than 1/125 sec. This means that the 180° rule would not support this speed." That's exactly what I suggested was going on with his 'big reveal'. In other words, he confirmed that he posted erroneous information...which you swallowed hook, line and sinker.

Bill Claff has been reporting sensor analyses for many years and is well-respected in this field. He’s reporting real data from RAW file analyses, not just speculating.

Regarding the a7S III being, "...well established to apply rather aggressive noise reduction at high base ISO 12800," is that for still photography or for videography? Given your post linked above where the YouTuber who provided 'all the details' (according to you, that is...) was talking about 'open gate' and 'anamorphic', it seems clear that you have a focus on video. Earlier in this thread, @bclaff made it clear that his data apply to still images, not video. I'd suggest you stop using 'evidence' from the videography world to make conclusions about still photography.

Feel free to slink away from this thread as you did the other one.
 
  • Like
Reactions: 4 users
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
This might sound radical, but I think I’ll choose to trust Canon over… some website I’ve never heard of before?
Photonstophotos.net is the gold standard for DR measurements. And as you don’t appear to have taken it in, the whole of this thread started by the CR Guy is about his measuring the DR of the R3.
 
Upvote 0
This noise reduction ”cooked raw” is nonsense. Case in point, what happens when you add the Sony A7S3 to the chart? A camera that is well established to apply rather aggressive noise reduction at high base ISO 12800. Well from their chart it certainly doesn’t appear to show any noise reduction being applied until a lofty 102,000 ISO. Please explain?...
I have several objective tools at my disposal to make these decisions but as I indicated earlier it's hard to quantify the amount of signal processing (noise reduction) that is being applied. Ultimately the assignment of the down triangle symbol is a judgement call and I try not us use the symbol which is why the R3 results were delayed as I double-checked.

Regarding the Sony ILCE-7SM3 it's useful to look at the attached Read Noise as a function of ISO setting.
I decided that the drop above ISO 102400 warranted being labeled as noise reduction.
I see no evidence of anything at ISO 12800. There's a slight drop at ISO 4000 that I didn't deem significant.
Naturally you have the right not to accept these results.ILCE-7SM3_RN.png
 
Upvote 0