Come play with the Canon EOS R RAW files

RobbieHat

EOS M50
Feb 4, 2015
41
31
54
Rocklin, CA/The Sea Ranch, CA
I have measured the dynamic range from the RAWs. It seems that the EOS R exactly matches the 5D4 performance, except at ISO 100, where it has around 0.2 EV less dynamic range compared to 5D4. Maybe not that bad taking into account that the sensor is now permanently switched on. EDIT: that could merely be related to sample variation (I just have seen that the 5D4 analyzed also showed 0.2 EV less DR in some files).

Here it follows the results, in the same units used by DXO (normalized). Between parenthesis, for comparison, the values from the 5D4 (got using exactly the same measurement procedure from this forum post, with this data).

ISO 100
FRO_0007 - 13.46 EV (13.60)
FRO_0150 - 13.25 EV (13.61)
FRO_0181 - 13.42 EV (13.61)


ISO 250
FRO_0078 - 12.91 EV


ISO 400
FRO_0230 - 12.99 EV (12.99)


ISO 3200
FRO_0201 - 10.97 EV (10.98)


ISO 6400
FRO_0362 - 10.13 EV (10.24)


ISO 10000
FRO_0380 - 9.48 EV

Since likely both cameras also share the same physical base ISO (the real ISO related to the labeled ISO 100) the comparison is fair.

The ISO 100 and 6400 pictures were underexposed, but that has been accounted for (assumed always a max saturation of 16383). These results are from the green channel (the EOS R has different read noise in each channel, just as the 5D4 did, likely due to some white balance correction done in the RAWs, which Canon never did in the past and neither has done in the 1DX2 -at least to this degree-).

The EOS R has a "masked" left area (the photodetectors at the left not exposed to the light) of 144 pixels width, compared to 136 in the 5D4. Same for the top (46 compared to 42). So maybe this sensor could be a new design (with similar technology, though). For those that have not heard about them, these extra pixels are not used by all manufacturers, but are a good habit from Canon, because allow good raw converters to improve the noise characteristics.

The 6D2 achieves 11.90 EV of dynamic range at ISO 100 with this same test. So Canon continues 1.55 EV above the old tech, but still 1.45 EV under my A7R3 results (14.85 EV). Come on Canon... you have done only just half of the job!. Meanwhile Sony did upped their performance (the A7R2 topped at 13.90 EV, same sensor as A7R3 but with poorer electronics).

For those interested, these results can be obtanined by downloading this tool and running e.g.:

C:\> dcraw -E -4 -j -t 0 -s all FRO_0007.dng
C:\> hraw mskstats -i FRO_0007_0.pgm -m 144 46 -c G
ReadNoise=2.73998 DR@30=12.4628 DR@8=13.4246 file { FRO_0007_0.pgm }
image { mean=654.274 min=489 max=15979 } left mask { mean=511.006 min=491 max=531 crop=68x4492+2+50 }


(to know the width/height of the masked areas, a trick is to edit the pgm with gimp and boost the brightness to 125 and contrast to 124).
Helpful information. Thanks for the analysis.

Good to see that this was not a step backward but disappointing that there was no improvement. I guess that was to be expected based on the release of the Mark IV and leverage of that sensor in other recent bodies. That still is a significant improvement over my existing 5dMIII and 5DSR dynamic range. If the future pro body has similar improvements in DR or better I will likely replace one of both of those bodies.

I too would really like to see Canon take the final step and close the DR gap with Sony sensor while maintaining all the other elements that make Canon great.
 
  • Like
Reactions: LetTheRightLensIn
Aug 3, 2018
4
2
I have measured the dynamic range from the RAWs. It seems that the EOS R exactly matches the 5D4 performance, except at ISO 100, where it has around 0.2 EV less dynamic range compared to 5D4. Maybe not that bad taking into account that the sensor is now permanently switched on. EDIT: that could merely be related to sample variation (I just have seen that the 5D4 analyzed also showed 0.2 EV less DR in some files).

Here it follows the results, in the same units used by DXO (normalized). Between parenthesis, for comparison, the values from the 5D4 (got using exactly the same measurement procedure from this forum post, with this data).

ISO 100
FRO_0007 - 13.46 EV (13.60)
FRO_0150 - 13.25 EV (13.61)
FRO_0181 - 13.42 EV (13.61)


ISO 250
FRO_0078 - 12.91 EV


ISO 400
FRO_0230 - 12.99 EV (12.99)


ISO 3200
FRO_0201 - 10.97 EV (10.98)


ISO 6400
FRO_0362 - 10.13 EV (10.24)


ISO 10000
FRO_0380 - 9.48 EV

Since likely both cameras also share the same physical base ISO (the real ISO related to the labeled ISO 100) the comparison is fair.

The ISO 100 and 6400 pictures were underexposed, but that has been accounted for (assumed always a max saturation of 16383). These results are from the green channel (the EOS R has different read noise in each channel, just as the 5D4 did, likely due to some white balance correction done in the RAWs, which Canon never did in the past and neither has done in the 1DX2 -at least to this degree-).

The EOS R has a "masked" left area (the photodetectors at the left not exposed to the light) of 144 pixels width, compared to 136 in the 5D4. Same for the top (46 compared to 42). So maybe this sensor could be a new design (with similar technology, though). For those that have not heard about them, these extra pixels are not used by all manufacturers, but are a good habit from Canon, because allow good raw converters to improve the noise characteristics.

The 6D2 achieves 11.90 EV of dynamic range at ISO 100 with this same test. So Canon continues 1.55 EV above the old tech, but still 1.45 EV under my A7R3 results (14.85 EV). Come on Canon... you have done only just half of the job!. Meanwhile Sony did upped their performance (the A7R2 topped at 13.90 EV, same sensor as A7R3 but with poorer electronics).

For those interested, these results can be obtanined by downloading this tool and running e.g.:

C:\> dcraw -E -4 -j -t 0 -s all FRO_0007.dng
C:\> hraw mskstats -i FRO_0007_0.pgm -m 144 46 -c G
ReadNoise=2.73998 DR@30=12.4628 DR@8=13.4246 file { FRO_0007_0.pgm }
image { mean=654.274 min=489 max=15979 } left mask { mean=511.006 min=491 max=531 crop=68x4492+2+50 }


(to know the width/height of the masked areas, a trick is to edit the pgm with gimp and boost the brightness to 125 and contrast to 124).
Nice work! Thank you for the test!
 

preppyak

EOS 6D MK II
Oct 18, 2011
991
61
Still using default CA correction in ACR, with no other fussing, makes this pretty decent.

I don't think it's "serious purple fringing". An EF 50/1.4 would show much, much worse CA.
Ha, anyone who calls that serious purple fringing has never used an 85mm f/1.8 in bright sunlight at f/1.8. As you showed, thats easily corrected
 

Viggo

EOS 5D SR
Dec 13, 2010
4,234
795
The 50 1.2 is the most impressive of the lot to me. Almost makes me want to get an R camera just so I can actually take advantage of a decent 50mm lens for once...

The 28-70 appeared a bit softer to my eye at 70mm f2 than the 50 did at 1.2 (!). 28mm f2 on the other hand was fairly crisp. I'd be curious to see how the performance at 70mm f2.8 stacks up with the 24-70 II at that setting...
That’s my idea also, put an add up for my 1dx2 and placed and order on an R and 50 f1.2, I’ve been dreaming about that 50 since the first day I bought the EF version and was seriously disappointed ...
 

Viggo

EOS 5D SR
Dec 13, 2010
4,234
795
So... is he allergic to place focus correctly? Basically all of them are off... And what's up with the girl in the swimsuit running? it looks like focus actually hit, but it looks terrible...
 
Last edited:
  • Like
Reactions: KenRockwell

ykn123

EOS T7i
Oct 1, 2013
81
33
52
Germany
www.fotoflucke.de
The R - i really wish it had 8/10fps in Servo and a joystick (but need to see whether Tocu&Drag AF and the Mn Touch bar are good alternatives) - all other things that people complain about like IBIS and single slot and 4K crp are fine with me. The lenses alone are very tempting.
 
  • Like
Reactions: Viggo

rsdofny

EOS M50
Oct 18, 2013
37
15
I am blown away by how clear the photo captures the reflection of the beach and the photographer. Is it normal to see the tips of the 2 eyebrows to be out of focus and the eyelashes on the non-focused eye is in focus again?
 

MayaTlab

EOS 80D
Oct 6, 2015
191
77
The R - i really wish it had 8/10fps in Servo and a joystick (but need to see whether Tocu&Drag AF and the Mn Touch bar are good alternatives) - all other things that people complain about like IBIS and single slot and 4K crp are fine with me. The lenses alone are very tempting.
These two limitations have the same cause (sensor readout speed). So in fact you very much want FF 4K :D.
 

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,966
1,178
119
I am blown away by how clear the photo captures the reflection of the beach and the photographer. Is it normal to see the tips of the 2 eyebrows to be out of focus and the eyelashes on the non-focused eye is in focus again?
Yes.
 

BeenThere

EOS 6D MK II
Sep 4, 2012
858
187
The 50 1.2 is the most impressive of the lot to me. Almost makes me want to get an R camera just so I can actually take advantage of a decent 50mm lens for once...

The 28-70 appeared a bit softer to my eye at 70mm f2 than the 50 did at 1.2 (!). 28mm f2 on the other hand was fairly crisp. I'd be curious to see how the performance at 70mm f2.8 stacks up with the 24-70 II at that setting...
Great Zooms don’t quite have the IQ of great primes although they can get pretty close in the center of the frame. Also, it’s hard to draw conclusion from just a few uncontrolled images. Fun to speculate!
 

tmc784

I love photography
Apr 1, 2018
77
22
I have 6L lenses, why should I spend more $$$$ to buy a M camera and adapters for my L lenses ? I just wait for 5DV
 
Sep 10, 2018
1
0
Did anyone try to play with the shadows and recover them with the photo of the beach FRO_0007 and the computer FR0_0362 ? Playing in PS with the DNG seems not to recover them or work well. It looks like the noise is gorgeous after +3EV even at ISO 100 . There is not comparation with the 5D MkIV RAW where you can rise/recover the shadows a lot better.
 
Last edited:

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,966
1,178
119
Did anyone try to play with the shadows and recover them with the photo of the beach FRO_0007 and the computer FR0_0362 ? Playing in PS with the DNG seems not to recover them or work well.
I agree, I actually posted the linked images on another thread, that predated this one, and I said basically the same thing.

There are many things to like about this system, improvements in DR are not one.
 
  • Like
Reactions: LetTheRightLensIn

Act444

EOS 6D MK II
May 4, 2011
1,029
120
That’s my idea also, put an add up for my 1dx2 and placed and order on an R and 50 f1.2, I’ve been dreaming about that 50 since the first day I bought the EF version and was seriously disappointed ...
Not too pleased about not being able to use that 50 with my 5D4 but it is what it is.

Great Zooms don’t quite have the IQ of great primes although they can get pretty close in the center of the frame. Also, it’s hard to draw conclusion from just a few uncontrolled images. Fun to speculate!
True that. Highly awaiting more samples, resolution tests, and an update to DPP to actually see what the RAWs are made of...
 

Mikehit

EOS 5D MK IV
Jul 28, 2015
3,227
416
I agree, I actually posted the linked images on another thread, that predated this one, and I said basically the same thing.

There are many things to like about this system, improvements in DR are not one.
I wonder if Canon have made the same 'developments' to the 5DIV sensor that they made for the 6D2 sensor compared to 6D. That had a mixed bag of comments and it took some while for the positives (few that they were) to start emerging.
 
  • Like
Reactions: Keith_Reeder

privatebydesign

Would you take advice from a cartoons stuffed toy?
Jan 29, 2011
7,966
1,178
119
I wonder if Canon have made the same 'developments' to the 5DIV sensor that they made for the 6D2 sensor compared to 6D. That had a mixed bag of comments and it took some while for the positives (few that they were) to start emerging.
To be sure the 5D MkIV, for all the complaints from testers, is a very good sensor with a lot more potential, but I don't think the R developers had the brief to mess with the sensor, just everything else!
 
  • Like
Reactions: pj1974