Canon will release four new full-frame cameras in 2020 [CR2]

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
Yep. But all four smaller sensels do not necessarily all receive the same number of photons. Poisson distribution and all, not to mention edges between objects in the scene with different brightnesses.
You are both right. Leaving aside edge effects, the sum total of photons hitting 4 1/4-sized pixels (sensels) will have the same Poisson distribution as 1 full-sized pixel but the smaller pixels will lead to more noticeable granular noise. It's a trade-off between detail and noise. But, it's easier to extract detail from a noisy background than it is from one where the detail has been smoothed.

It leads us on to what is the optimal size of pixels for the current sensors. I slightly prefer the heavily cropped files I get from my 50 Mpx 5DSR (20 Mpx APS-C equivalent) to those from my 32 MPx 90D, although when not heavily cropped both are pretty indistinguishable.
 
  • Like
Reactions: Joules

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
But he was distinctly NOT talking about using the full area of the FF sensor!
I was replying to the last part of his post to point out that future crop sensors, or cropped areas from a FF, don't have much room for improvement over current ones.
.... maybe future technology can improve the crop performance over what crop sensors currently achieve?
 

SteveC

M50 & T6i
Sep 3, 2019
475
341
I was replying to the last part of his post to point out that future crop sensors, or cropped areas from a FF, don't have much room for improvement over current ones.
Ah. Got you, and yes, that's certainly true.
 
  • Like
Reactions: AlanF

Michael Clark

Now we see through a glass, darkly...
Apr 5, 2016
1,716
938
You are both right. Leaving aside edge effects, the sum total of photons hitting 4 1/4-sized pixels (sensels) will have the same Poisson distribution as 1 full-sized pixel but the smaller pixels will lead to more noticeable granular noise. It's a trade-off between detail and noise. But, it's easier to extract detail from a noisy background than it is from one where the detail has been smoothed.

It leads us on to what is the optimal size of pixels for the current sensors. I slightly prefer the heavily cropped files I get from my 50 Mpx 5DSR (20 Mpx APS-C equivalent) to those from my 32 MPx 90D, although when not heavily cropped both are pretty indistinguishable.
Assuming none of the smaller sensels are at full well capacity, it will. But if some but not all of the smaller photosites are at FWC, then the end result will be different than if the same number of photons had struck a larger photosite that did not reach FWC. In the former case, some information (photons that struck the sensels already at FWC) is lost. In the later case every photon is counted, though at lower resolution.
 

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
Assuming none of the smaller sensels are at full well capacity, it will. But if some but not all of the smaller photosites are at FWC, then the end result will be different than if the same number of photons had struck a larger photosite that did not reach FWC. In the former case, some information (photons that struck the sensels already at FWC) is lost. In the later case every photon is counted, though at lower resolution.
From your previous posts, I know you what you are assuming so let me spell it out for everyone who hasn't been following, using as an example again two sensors, one of which has sensels (pixels) that are 1/4 the size of the other (say an 80 vs a 20 Mpx sensor). The full well potential of the large sensel is 4x larger than of the smaller sensel but there are 4 small ones occupying the same area as 1 large one so 4 small have effectively the same full well potential as 1 large sensel. Michael is implicitly and correctly assuming that the statistical noise in each small sensel (Poisson distribution) is greater than that for the large sensel because they receive on average only a 1/4 of the number of photons. Accordingly, there is a probability that some of the small sensels are saturated where others are not, and this will be more frequent than for the large sensel. Whereas this is perfectly correct, the probability of random saturation happening at low levels of light is very low. The probability of some being saturated and some not will have to be at high levels of light where they are all closer to being saturated. But, at high levels of light, fluctuations are statistically lower. So, in practice, it's not going to be very significant.
 
  • Like
Reactions: Michael Clark

Kit.

EOR R
Apr 25, 2011
1,626
967
Yes, but dithering only works well when none of the sensels/photosites combined into one pixel are at full well capacity/fully saturated...
When your luminous exposure is just slightly above the saturation level, it is quite possible that your full pixel will be saturated, but one of your quarter-pixels won't. Averaging over quarter-pixels in this case seems to give a slightly higher DR at the expense of linearity near the saturation values.
 

Quarkcharmed

EOS 5DMkIV
Feb 14, 2018
784
628
Australia
www.michaelborisenko.com
It has the same DR and low light performance as the 80D
and what appears to be the same AF module. I bought it for the Mpx, the 10fps, point focus and excellent mirrorless performance in LV. It's a very nice piece of kit that could be even better. Remember that the low light performance is the same as the 80D only when printed or viewed to the same size, but not per pixel. A 1000x1000px grid on the 90D would have worse noise than a 1000x1000 on the 80D, for example.
Ok, I found out why I thought 90D had a better DR than 80D


Not sure if it still holds or photonstophotos measurements are more accurate.
 

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
Ok, I found out why I thought 90D had a better DR than 80D


Not sure if it still holds or photonstophotos measurements are more accurate.
That was an initial estimate that caused a stir but the measurements were wrong. There was some discussion on the net until the photonstophotos was posted.
 

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
Yep. Now it downgrades my expectations on the prospective high-res R. It'll probably be lagging about a stop behind the Sony. Best case half a stop, if we check 90D against cropped A7RIV.
The DR of the 90D is pretty much the same as the cropped A7RIV over the whole range apart from a couple of points around iso100, which may be or not be real.
 

AlanF

Canon 5DSR II
Aug 16, 2012
6,199
4,058
But ISO 100 is crucial, this is where we shoot landscapes...
I'd be ok with 70-80Mp though and keeping the same DR as 5DIV. Time will tell.
Not crucial for me! I live in the iso600+ world. I would find 70-80 Mpx too large to handle as my standard format and would want a genuine RAW (or CRAW) crop mode for general usage and occasional use of FF for my type of nature photography. We all want something different and so have to live and let live.
 
Last edited:

mppix

EOS T7i
Feb 13, 2018
95
57
I think you're going to see the RF line revert to what it was with film, sans the IX bastard child.

If canon can do a sub $800 full frame camera, they'll do the entire line with full frame. and why not?

Canon has a HUGE advantage over the other companies, they can manufacturer in Japan - and it's almost entirely all automated. Especially for something like a mirrorless camera.
Maybe - IC/sensor cost came down considerably due to waver size. However, the real cost difference of APS-C vs. full frame is lens cost (see Fuji). We will have to see if canon considers APS-C lenses worth making.

PS. Just my 2 cents. I haven't shot an APS-C camera in years and if it were up to me, I'd rather have a high spec'd FF body.
 

tron

EOS 5D SR
Nov 8, 2011
4,389
627
Maybe - IC/sensor cost came down considerably due to waver size. However, the real cost difference of APS-C vs. full frame is lens cost (see Fuji). We will have to see if canon considers APS-C lenses worth making.

PS. Just my 2 cents. I haven't shot an APS-C camera in years and if it were up to me, I'd rather have a high spec'd FF body.
Even for birding having a FF camera helps with BIF photography (1.6*1.6 = 2.56 times the area).
 

stevelee

FT-QL
Jul 6, 2017
1,319
324
Davidson, NC
Well, I can make a good (not great) photo with a cardboard box, gaffers tape, tin foil, a tiny hole and some Ilford paper.On top of it it's a whole lot more enjoyable than clicking on a button full of electronics and a keyboard. Now, do I do it very often? No. I like to spread the love around though. Like Canon offering different bodies for different strokes.
When I was a kid, I made pictures with a Kodak Brownie and made contact prints. The prints were small, but the tones were gorgeous, better than I ever could get with enlarger paper.
 
  • Like
Reactions: Michael Clark