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

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
So 90D will likely be the same as 90D in terms of the DR, which is worse than 5DIV.
If we extrapolate that to the prospective high-res RS that is assumed to use the same tech but larger sensor area, is it a plausible conclusion that RS will have the DR worse than 5DIV?
 

Sharlin

EOS 6D MK II
Dec 26, 2015
1,033
531
Turku, Finland
So 90D will likely be the same as 90D in terms of the DR, which is worse than 5DIV.
If we extrapolate that to the prospective high-res RS that is assumed to use the same tech but larger sensor area, is it a plausible conclusion that RS will have the DR worse than 5DIV?
No, unless you’re talking about individual pixels. Resolution normalized and all else being equal, an FF sensor will always have a bit more than 1EV more DR than an APS-C sensor because of its larger surface area. That’s the whole point of FF sensors.
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
No, unless you’re talking about individual pixels. Resolution normalized and all else being equal, an FF sensor will always have a bit more than 1EV more DR than an APS-C sensor because of its larger surface area. That’s the whole point of FF sensors.
It's only applicable to some extent to the tests where they normalise and downsample everything to a certain smaller size. Downsampling reduces visible noise and noise affects DR. So yes with downsampling, a crop version of the same sensor has a lower DR

But such measurements are flawed as the whole point of high-res FF camera is to not downsample and use all pixels. So yes I'm interested in per-pixel sensor performance.

UPD: probably this chart is more relevant

I have no hope it reaches Sony's read noise, but I'd have expected it to be somewhere in between the best Canon and best Sony, not worse that 5DIV. That effectively means Canon failed to improve on the read noise in their newest generation sensors.
 
Last edited:

justaCanonuser

Grab your camera, go out and shoot!
Feb 12, 2014
437
251
Frankfurt, Germany
I'm bored so I'll post an essentially irrelevant comment. I could care less about 24p. I could care less about slight differences in DR, or even more than slight differences. I'd rather have an AA filter than not. I have two Canon DSLRs (AA filters, of course) and one Fujifilm camera (no AA filter). I really don't see any advantage to no AA filter. Any slight differences in sharpness are irrelevant as far as I'm concerned. I might even say that pictures can be too sharp. Buy what you want and argue about the technicalities if you want.
"Sharpness is a bourgeois concept", Henry Cartier Bresson once told the younger photographer Helmut Newton. Still my favorite quote from HCB, and I love to quote it again and again. According to him no AA filter would be more bourgeois than AA filter :devilish:
 

justaCanonuser

Grab your camera, go out and shoot!
Feb 12, 2014
437
251
Frankfurt, Germany
Even if Canon would succeed in beating DR auf Sony sensors, the hate on the web, in particular on DPR, wouldn't stop. So what? I don't care, but I'd love to see a 7D III or something really comparable asap. My old 7D II still does well. But it is the only camera I'd love also to have useful 4k video to have the option for pulling out useable stills from fast action. Plus, Canon should really improve the 7D II's not so brillant AF system. This is the biggest downside of the 7D II for wildlife.
 

justaCanonuser

Grab your camera, go out and shoot!
Feb 12, 2014
437
251
Frankfurt, Germany
Please please please. Can we talk about dynamic range here. This topic is for DR war, not 24fps war. Can we please troll each other with Canon's DR vs Sony's DR, not the framerate.
This topic is NOT called "initial testing shows Canon can do 24p if the camera is heated and slows down in 30p mode".
Yeah, but if the DR talk is based on tests done by a questionable source I think there is room for other questionable topics in this thread ;)
 
Please God, don't make us read more 24p posts. There is already a thread that has beat this dead horse to smithereens.
You are freaking right, it's a complete mess. Even Canon themselves are totally confused. This is what was posted to DPR and is not really funny - what a bunch of amateurs. They should stop messing with ppl and train their sfaff properly.

1567586934845.png

1567587026319.png
 
  • Like
Reactions: unfocused

AlanF

Canon 5DSR II
Aug 16, 2012
5,650
2,885
It's only applicable to some extent to the tests where they normalise and downsample everything to a certain smaller size. Downsampling reduces visible noise and noise affects DR. So yes with downsampling, a crop version of the same sensor has a lower DR

But such measurements are flawed as the whole point of high-res FF camera is to not downsample and use all pixels. So yes I'm interested in per-pixel sensor performance.

UPD: probably this chart is more relevant

I have no hope it reaches Sony's read noise, but I'd have expected it to be somewhere in between the best Canon and best Sony, not worse that 5DIV. That effectively means Canon failed to improve on the read noise in their newest generation sensors.
Bill Claff writes clearly in a link under the chart you present ( http://www.photonstophotos.net/GeneralTopics/Sensors_&_Raw/Sensor_Analysis_Primer/Photographic_Dynamic_Range_Summary.htm ) that he calculates PDR for an 8"x10" image held at approximately at arms length and he normalises to the appropriate circle of confusion. He is not downsampling and he is using all the pixels in those charts. The per pixel performance would be relevant if you are comparing the DR of say a print from a 1DX2 at 10"x15" with one from a 5DSR at 15.47"x23.5", where a single pixel from each has been enlarged to the same output size.
 

AlanF

Canon 5DSR II
Aug 16, 2012
5,650
2,885
"Sharpness is a bourgeois concept", Henry Cartier Bresson once told the younger photographer Helmut Newton. Still my favorite quote from HCB, and I love to quote it again and again. According to him no AA filter would be more bourgeois than AA filter :devilish:
Presumably, therefore, if you live up to the spirit of your quotes, all your images are soft?
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
Bill Claff writes clearly in a link under the chart you present ( http://www.photonstophotos.net/GeneralTopics/Sensors_&_Raw/Sensor_Analysis_Primer/Photographic_Dynamic_Range_Summary.htm ) that he calculates PDR for an 8"x10" image held at approximately at arms length and he normalises to the appropriate circle of confusion. He is not downsampling and he is using all the pixels in those charts. The per pixel performance would be relevant if you are comparing the DR of say a print from a 1DX2 at 10"x15" with one from a 5DSR at 15.47"x23.5", where a single pixel from each has been enlarged to the same output size.
Maybe I misunderstood their method, but if images from different sensors are printed on 8"x10" sheets, it's the same as downsampling. Or maybe they don't actually print them, just calculate - still it should be the same as downsampling.

If there's no downsamling at all, then the above mentioned Sony sensor in crop and FF modes would show the same DR. Otherwise I just don't understand how simple cropping, same sensor and same lens, just cropped, can reduce DR by 1 stop.
 

AlanF

Canon 5DSR II
Aug 16, 2012
5,650
2,885
Maybe I misunderstood their method, but if images from different sensors are printed on 8"x10" sheets, it's the same as downsampling. Or maybe they don't actually print them, just calculate - still it should be the same as downsampling.

If there's no downsamling at all, then the above mentioned Sony sensor in crop and FF modes would show the same DR. Otherwise I just don't understand how simple cropping, same sensor and same lens, just cropped, can reduce DR by 1 stop.
It's not downsampling - it's upsizing to the same size! The reason why the crop mode has less DR than the FF is because the crop needs 1.6x1.6 more upsizing than the FF to get to the same output size and so is amplifying the noise by 1.6x. Downsampling does give the same effect, but that's not how he calculates it.
 

Sharlin

EOS 6D MK II
Dec 26, 2015
1,033
531
Turku, Finland

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
It's not downsampling - it's upsizing to the same size! The reason why the crop mode has less DR than the FF is because the crop needs 1.6x1.6 more upsizing than the FF to get to the same output size and so is amplifying the noise by 1.6x. Downsampling does give the same effect, but that's not how he calculates it.
Hmmm ok... An actual print at 600ppi and 10" will take 6000 pixels on the longest side which corresponds to a 24 mp sensor. Still not upscaling. At 300ppi it'll definitely be downscaling.

But I don't know if they convert sensor pixels to printer's pixels. Maybe their model is just such as if an image was on the physical sensor pixels and then enlarged to a 8x10 print.
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
Note the disclaimer:


The 5D4 and 90D values are not comparable. The actual DR difference between the 5D4 and the 80D/90D is almost exactly what can be explained by the sensor size difference alone.
ok... best case that means the prospective RS would have the same DR as 5DIV, using the same measurements. Given the higher pixel count it wouldn't be that bad. But not exceptionally good either.
 

edoorn

EOS RP
Apr 1, 2016
206
107
it wouldn't be very good to be honest, since the A7R IV with 60mpix did manage to keep a high DR. So Canon would still be behind in this department by - for some, depending on the use - a significant margin
 

Bob Howland

EOS 7D MK II
Mar 25, 2012
441
46
Bill Claff writes clearly in a link under the chart you present ( http://www.photonstophotos.net/GeneralTopics/Sensors_&_Raw/Sensor_Analysis_Primer/Photographic_Dynamic_Range_Summary.htm ) that he calculates PDR for an 8"x10" image held at approximately at arms length and he normalises to the appropriate circle of confusion. He is not downsampling and he is using all the pixels in those charts. The per pixel performance would be relevant if you are comparing the DR of say a print from a 1DX2 at 10"x15" with one from a 5DSR at 15.47"x23.5", where a single pixel from each has been enlarged to the same output size.
If Claff is using a print for comparison, then the image had to go through a printer, hopefully the same printer for all images, and its image processor. How is he certain that the printer has no effect on dynamic range?
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
488
321
Australia
www.michaelborisenko.com
If Claff is using a print for comparison, then the image had to go through a printer, hopefully the same printer for all images, and its image processor. How is he certain that the printer has no effect on dynamic range?
Most likely he's not using an actual printer, but the DR is calculated as if the image were printed on an ideal printer. Effectively enlarging the sensor area to a 8x10 size print. In this case it's not downsampling but a smooth resizing, but it's not exactly clear from the explanations on the site.