Dynamic range testing of the Canon EOS R3 is complete

People who are unfamiliar with the longstanding tradition of Bill Claff providing excellent, repeatable dynamic range data for new cameras may bridle at some of the implications of the observations. But they should know that Bill has developed his methodologies openly over years with input from this and other communities. The data he produces, and his interpretations - which are made cautiously, which doesn't imply "speculation" or "guessing" - are useful to photographers who are looking to judge the capabilities and direction of various camera systems.

One of the hardest factors to suss out in this endeavor is the degree to which image data is fundamentally better/purer due to hardware advancements, and image data which appears better because of software interpretation - that may or may not present downsides as well. The dynamic range data Bill produces is rather immutable. Whether you think it appropriate to call it "cooking" or not is a semantic question the science won't answer, but it is indeed happening.

Whether a software-induced increase in perceived dynamic range is desirable, or has disadvantages - such as Sony's "star eater" issue - is often up to the individual photographer. Some appreciate a sophisticated algorithm improving the perceived data in a fashion that I likely couldn't quite do as well in post. Others will bridle at the implication that they don't have full control over the "pure" data, and will often suffer from the conceit that they'd do better doing the manipulation themselves.

Ever since I've been fooling around with Canon 1 series cameras, Canon has pursued a policy of "strategic ambiguity." They provide the quality perception boost, but they consider the "RAW" data to be that which comes out the end of the processing chain, so they need not address the nature of the processing that was done. This explains the manual talking about the RAW files being the raw data. Canon has decided semantically to consider the data that comes out the sausage machine to be, by definition, the raw data. This means they're true to their word in writing the manual as they have.

-tig

[Side note:
I'm perhaps halfway through the process of creating an R3 review for CanonRumors and Camnostic. I'm scheduled to talk to some CPS people in bit more than a week to ask some remaining questions. I will be certain to ask about this, but I do not expect a direct answer due to my experience in the past with this. The first time I asked a Canon rep about this was back when the 1DX II was new. With that camera and the 6D Mark II, as well as the SL2 and a few others, they concentrated in their marketing (more than with other cameras) on processor upgrades. I asked a rep why it is that after noting the increased frames per second, autofocus points, other new features, etc., a photographer should care about the processor version? What sort of benefit was it that an upgraded processor could provide that wasn't redundant to these other features? He responded that it marked the class of image quality improvements that could be made when taking shots. He indicated that this is one of the two reasons that the 1 series had a superior look and quality - the other reason being the sensor hardware. He did not tell me that Canon "cooks" the files, but he did indicate that the capacity of the processor was a direct factor in image quality. My sense has been ever since - corroborated with multiple conversations with Canon USA employees - that the firm wants to market the processor improvements as a proxy for image quality improvements in software, without implying that the pros aren't getting their "raw data." Which, in some interpretations, is a contradiction.]
 
  • Like
Reactions: 5 users
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
"How dare they cook the RAW files" is often indignantly shouted! Constant arguments about compressed or uncompressed RAW. Even over-exaggeration about the slightest differences in DR, that have been demonstrated to be undetectable by the human eye are constantly discussed.
Were did you see any shouting? I have not perceived any outrage over the noise reduction Canon employs of late.

It's been there in the R5 all along, and that body has been praised at every possible point. Now Canon seems to have expanded the range over which they apply it, but as it seems to have no meaningfull downsides, there's no point in complaining. Nobody's stars are being eaten and yet - likely the most important facet of them all - Canon finally domimates the DR rankings with both the R5 and the R3. Look who's d00med now! :p

It is simply interesting what Canon does with their sensors behind the scenes, and since they don't market official numbers for DR, noise, read speed and so on, testing on third party sites and discussions about their findings are the only way to give tickle this interest.
 
  • Like
Reactions: 1 users
Upvote 0
Jul 21, 2010
31,100
12,864
The question is not whether the file is "raw" but rather whether signal processing consulting neighboring pixels amounting to noise reduction has been performed. Such processing correlates pixels statistically with their neighbors.
Sorry, I was asking an unrelated question concerning Canon’s lossy CRAW format based on the point being made (or attempted, at any rate) about terminology. Personally, I don’t consider the output of a lossy compression to remain a RAW image (regardless of whether or not the removal of those data is discernible in my output).
 
  • Like
Reactions: 1 user
Upvote 0
Jul 21, 2010
31,100
12,864
One of the hardest factors to suss out in this endeavor is the degree to which image data is fundamentally better/purer due to hardware advancements, and image data which appears better because of software interpretation…
Lenses are an interesting parallel here. Canon now has three RF lenses, one of them an L-series lens, where they chose to sacrifice geometric distortion correction in the optical design in favor of forced correction in software (although obviously they can only force correction in their own software).

Is that ‘better’? I don’t know, but I’m happy with the output of my RF 14-35/4L and I don’t think we’d see a 16/2.8 lens with full optical corrections for $300.
 
  • Like
Reactions: 1 users
Upvote 0

Pierre Lagarde

Canon, Nikon and So on ...
Aug 4, 2020
123
147
France
www.deviantart.com
Good results indeed. Coherent with the rest : got my hands on the R3 last tuesday and what specifically impressed me is how easy and pleasant it is to use. The handling is probably the best I've ever experimented. The most impressive is the fact that it's so light while feeling so tough (10g more than a D850 with still a full pro shape and handling !).
I tried the eye focus control too, but I must say that with glasses, it seems a bit quirky, at least to calibrate.
Anyway, it's an impressive camera, and the sensor that goes with it looks promising too.
Very well done Canon !
 
  • Like
Reactions: 1 users
Upvote 0

Pierre Lagarde

Canon, Nikon and So on ...
Aug 4, 2020
123
147
France
www.deviantart.com
My one handling gripe so far is the omission of the lower lug for the E1 hand strap. I really hope RRS includes one in the design of their R3 L-plate.
Well, I've had no time to test that so far, but it seems E1 won't be the best solution anyway (and is discontinued by now). Maybe the Nikon AH-4 could do better for you, don't know. There are some other (and maybe better) solutions, I think, anyway.
 
Upvote 0
Jul 21, 2010
31,100
12,864
Well, I've had no time to test that so far, but it seems E1 won't be the best solution anyway (and is discontinued by now). Maybe the Nikon AH-4 could do better for you, don't know. There are some other (and maybe better) solutions, I think, anyway.
The E1 design works great for me with other bodies. Canon discontinued it in favor of the E2, which includes a 1/4”-20 attachment (it’s basically the E1 with the add-on screw-in attachment, if you have a gripped camera (except the R3) you take off the 1/4”-20 attachment and just attach the strap itself to the lower lug.

The Nikon AH-4 is like the E2, with a lower attachment that uses the camera’s tripod socket.

The problem with that attachment method is it precludes use of a camera plate (e.g., Arca Swiss type) for mounting on a tripod. Having to unscrew a strap attachment to use a tripod would be a PITA. Moreover, that Arca-type camera plate is where I attach my BlackRapid strap or Spider Holster Arca clamp. So for me, if I can’t attach the hand strap to a camera plate (RRS, Kirk, etc.), I’ll have to forego the hand strap that I would otherwise always have attached to the camera.
 
Upvote 0
Jan 29, 2011
10,675
6,121
The SmallRig designed R3 L-Plate has the ability to mount a hand strap. I’ve used a few SmallRig items now and whilst they are not the quality of some in my limited experience they have easily done the job well.

 
  • Like
Reactions: 1 users
Upvote 0

Pierre Lagarde

Canon, Nikon and So on ...
Aug 4, 2020
123
147
France
www.deviantart.com
The SmallRig designed R3 L-Plate has the ability to mount a hand strap. I’ve used a few SmallRig items now and whilst they are not the quality of some in my limited experience they have easily done the job well.

... or maybe simply replace the original screw with that kind of thing could be enough : https://www.amazon.com/Release-Tripod-Bracket-Adapter-Mounting/dp/B073YQNT9Y
 
Upvote 0
Jul 21, 2010
31,100
12,864
Darn it! I have come to rely on the Peak Design hand strap.
The PD Clutch is a good option. I use their wrist strap on my M6 and R (and I’ve attached an anchor to an R3 lug). But I’d prefer the RRS modular L-bracket.

The SmallRig designed R3 L-Plate has the ability to mount a hand strap. I’ve used a few SmallRig items now and whilst they are not the quality of some in my limited experience they have easily done the job well.
Thanks! Modular, too. I’ll wait to see what the RRS design looks like (they haven’t finalized it, or at least not posted it), and if there’s no lug I’ll try the SmallRig version.
 
  • Like
Reactions: 1 user
Upvote 0
Jan 27, 2020
826
1,796
The Fourier analysis is 100% reliable and is confirmed with the spectral analysis.
Bill, you should be thanked, not just for joining the discussion and explaining things more in depth, but also for being patient with those "internet experts" who read a few articles and then cherry-pick a few sentances to "prove" their point, when, of course, they are not experts and have no point other than to try and show they are smarter than the real expert. Must be frustrating, but then again, this is just photography and the same thing happens on much larger, more important issues. Nonetheless, thanks.
 
  • Like
Reactions: 9 users
Upvote 0
I've always found @bclaff's findings interesting and informative. Clearly many people have different definitions of what they consider raw, what is NR, and what is "cooked".

My definition of RAW is the representation of the response to the light hitting the sensels on the sensor with no transform of that data to turn it into full RGB that conforms to a given colorspace like ProPhoto or sRGB. You have color managed, and you have raw. Color managed is typically full RGB at each pixel spot and a given colorspace, and RAW is a bunch of sensel values that needs processing to turn it into full color managed RGB.

With respects to NR or "cooked", frankly, unless it's completely destroying useful image data, I just do not care. Pretty much every manufacturer does it in one form or another in the never ending quest for better image quality, and you can't turn it off, so get over it. If it really made the image data worst for most of the users, the manufacturer probably wouldn't do it, and I have way more other things to worry about than whether or not my CR3 files are being "cooked" or have NR applied before being written out, like oh, I don't know.... Booking that next job so I can pay rent.
 
Upvote 0
Jul 21, 2010
31,100
12,864
With respects to NR or "cooked", frankly, unless it's completely destroying useful image data, I just do not care. Pretty much every manufacturer does it in one form or another in the never ending quest for better image quality, and you can't turn it off, so get over it. If it really made the image data worst for most of the users, the manufacturer probably wouldn't do it, and I have way more other things to worry about than whether or not my CR3 files are being "cooked" or have NR applied before being written out, like oh, I don't know.... Booking that next job so I can pay rent.
As discussed earlier, this is essentially an academic discussion. Any ‘cooking’ of the RAW file is baked in and cannot be undone. I like knowing what is happening ‘behind the scene’, and I’m grateful to Bill for his analyses…but I’m not going to return my R3 because he uses downward-pointing triangles instead of circles on a plot.
 
  • Like
Reactions: 4 users
Upvote 0

GoldWing

Canon EOS 1DXMKII
Oct 19, 2013
404
279
Los Angeles, CA
en.wikipedia.org
All the graphs are really "cool". But the "money" is in IQ.

The Z9 has superior IQ to the R3 and 1DXMKIII

Canon seems to be pushing this "ultra-high ISO" advantage.... I don't shoot in the dark but perhaps it's a new trend in sports photography.

Turn off the lights and take pictures of the NFL, NBA, MLB and the Olympics in darkness.... Could catch on ;)

Next time... We'll have to shut 80% of the lights in Yankee Stadium....

IQ ='s Money also known as revenue.... My clients don' pay for charts and graphs..... :)
 
Upvote 0
Jul 21, 2010
31,100
12,864
All the graphs are really "cool". But the "money" is in IQ.

The Z9 has superior IQ to the R3 and 1DXMKIII

Canon seems to be pushing this "ultra-high ISO" advantage.... I don't shoot in the dark but perhaps it's a new trend in sports photography.

Turn off the lights and take pictures of the NFL, NBA, MLB and the Olympics in darkness.... Could catch on ;)

Next time... We'll have to shut 80% of the lights in Yankee Stadium....

IQ ='s Money also known as revenue.... My clients don' pay for charts and graphs..... :)
Thanks for illustrating that not everyone has high IQ.



I’m talking about image quality, of course.
 
  • Like
  • Haha
Reactions: 7 users
Upvote 0
All the graphs are really "cool". But the "money" is in IQ.

The Z9 has superior IQ to the R3 and 1DXMKIII

Canon seems to be pushing this "ultra-high ISO" advantage.... I don't shoot in the dark but perhaps it's a new trend in sports photography.

Turn off the lights and take pictures of the NFL, NBA, MLB and the Olympics in darkness.... Could catch on ;)

Next time... We'll have to shut 80% of the lights in Yankee Stadium....

IQ ='s Money also known as revenue.... My clients don' pay for charts and graphs..... :)
As an engineer and owner of an engineering company the data from PP is of great interest.
As a photographer for 34 years the data from PP is of great interest.
As a photography company the data is of great interest.

Here in the UK we have not had an opportunity to use the R3 and our main bodies are the R5.
We regularly shoot at very high ISO so understanding how to get the best from the camera and our post processing flow is vital to the business and sanity.

I think it's unlikely we would move to the R3 but I look forward to having some time with one in the future.

As for the Nikon I've not used one and haven't read enough to have suitable knowledge to form an opinion.

I think the advantages of the R3 being advertised are wide and diverse and not just high ISO performance.
 
  • Like
Reactions: 1 users
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,356
22,535
All the graphs are really "cool". But the "money" is in IQ.

The Z9 has superior IQ to the R3 and 1DXMKIII

Canon seems to be pushing this "ultra-high ISO" advantage.... I don't shoot in the dark but perhaps it's a new trend in sports photography.

Turn off the lights and take pictures of the NFL, NBA, MLB and the Olympics in darkness.... Could catch on ;)

Next time... We'll have to shut 80% of the lights in Yankee Stadium....

IQ ='s Money also known as revenue.... My clients don' pay for charts and graphs..... :)
Your comments are usually shots in the dark.
 
  • Like
  • Haha
Reactions: 4 users
Upvote 0

DBounce

Canon Eos R3
May 3, 2016
500
544
All the graphs are really "cool". But the "money" is in IQ.

The Z9 has superior IQ to the R3 and 1DXMKIII

Canon seems to be pushing this "ultra-high ISO" advantage.... I don't shoot in the dark but perhaps it's a new trend in sports photography.

Turn off the lights and take pictures of the NFL, NBA, MLB and the Olympics in darkness.... Could catch on ;)

Next time... We'll have to shut 80% of the lights in Yankee Stadium....

IQ ='s Money also known as revenue.... My clients don' pay for charts and graphs..... :)
Did your clients pay for your work prior to the Z9? Only people on sites like this pixel peep. But that said, high ISO is not only about shooting at night. It’s also useful when you require a fast shutter speed and need to crank up the ISO to compensate. Or don’t Nikon shooters do that? And what about dynamic range? I hear the Z9 has worst dynamic range than the Z7… so it stands to reason it will have less than the R3. Resolution is only one aspect to image quality.
 
Upvote 0