A high-megapixel EOS R camera is still on the roadmap [CR2]

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,348
22,524
It's all true but as I said we were talking about different 'enlargement ratios'.
I was talking about digital conversions. When you view 1:1, there's no compression and no enlargement of the data from your raw file. Anything other than 1:1 requires additional digital conversion (resampling). Most of the time it's downsampling so that you can fit large size image into the screen or a window on the screen.
Because there's no resampling at 1:1, you see the sensor noise as it is and that was important for my particular test.
Can we just agree on that your test is relevant to the extent that we should not expect the same signal to noise or DR in say a 1000x1000 pixel crop from a 80Mpx as from a 20Mpx sensor but for most purposes, like viewing a full image on a screen or printing it at the same size form both sensors, that it is not relevant.
 
  • Like
Reactions: 1 user
Upvote 0
Feb 28, 2013
1,612
272
70
Can we just agree on that your test is relevant to the extent that we should not expect the same signal to noise or DR in say a 1000x1000 pixel crop from a 80Mpx as from a 20Mpx sensor but for most purposes, like viewing a full image on a screen or printing it at the same size form both sensors, that it is not relevant.
We dont shoot subjects in a lab. The majority of people never enlarge larger than an A3 print with the vast majority of shots these days being digitally viewed we actually oversampling our output. It amuses me these types of conversation for years in Digital video the colour space was REC.709 now we have REC.2020 with a far greater luminance & colour gamut. These parameters have done more for picture quality as the colour space is greatly expanded. Its colour space that brings the greatest benefits, its pointless creating a 120MP camera if affordable lenses cannot exploit the resolution or if refraction minimises the benefits. Pixel matching to line pairs per mm makes far more sense that ever greater MPs.

Let me explain another way. The average multi-screen theatre in Europe has a 56ft diagonal screen. To actually see true 4K you need to sit in the front three rows, if in the future that screen upgrades to 6K or even 8K then you would need to stand in the area between the front row and the screen to actually see 6K or 8K which is totally impractical. We make 6K and 8K movie cameras now but the images are output at 4K, we oversample to retain the colour space but sacrifice some resolution.

Most modern camera electronics do a good job of minimising photon shot noise so the S/N ratio has improved vastly and the old argument between small & large pixels is less of an issue than it was and is still improving. So where can it improve? Lenses still have room for improvement and are improving, by this I mean aberrations and vignetting rather than all out resolution improvements. Dynamic range is improving and will continue to do so and with it so will colour gamut.

In reality the test charts are a guide, cameras & lenses are tools and tools are only as good as the people that use them. Forget charts, forget pixel peeping and go and shoot as often as you can and experiment that way your level of photography grows and you can enjoy the output not technical specs.
 
  • Like
Reactions: 2 users
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
Because there's no resampling at 1:1, you see the sensor noise as it is and that was important for my particular test.
I think if you could just stress that bit, that it is your particular test and that it does not have to impact the opinions of others on high resolution sensors, it could have reduced the number of back and forth posts on the subject drastically.
 
  • Like
Reactions: 1 user
Upvote 0
Can we just agree on that your test is relevant to the extent that we should not expect the same signal to noise or DR in say a 1000x1000 pixel crop from a 80Mpx as from a 20Mpx sensor but for most purposes, like viewing a full image on a screen or printing it at the same size form both sensors, that it is not relevant.

It's relevant when your target is large sized file or print and when you pixel-peep (and you pixel-peep when your target is large). Also it's relevant to things like amount of postprocessing you can apply; say noise reduction amount and parameters will depend on the noise, and then the noise limits certain types of digital manipulations.
It's absolutely true that in many cases for sensor comparison you have to normalise to the same size.

I think if you could just stress that bit, that it is your particular test and that it does not have to impact the opinions of others on high resolution sensors, it could have reduced the number of back and forth posts on the subject drastically.

Yeah I had to, although I tried to emphasise that in the previous posts. :)

The majority of people never enlarge larger than an A3 print

5DIV with 30Mp sensor can be printed at 300ppi on up to A2. If cropped, it won't reach 300ppi. So if your target is a large high quality print, you'll likely be pixel-peeping.
I know though that 200-240ppi is also ok and in many cases indistinguishable from 300ppi.
 
  • Like
Reactions: 1 user
Upvote 0

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,348
22,524
We dont shoot subjects in a lab. The majority of people never enlarge larger than an A3 print with the vast majority of shots these days being digitally viewed we actually oversampling our output. It amuses me these types of conversation for years in Digital video the colour space was REC.709 now we have REC.2020 with a far greater luminance & colour gamut. These parameters have done more for picture quality as the colour space is greatly expanded. Its colour space that brings the greatest benefits, its pointless creating a 120MP camera if affordable lenses cannot exploit the resolution or if refraction minimises the benefits. Pixel matching to line pairs per mm makes far more sense that ever greater MPs.

Let me explain another way. The average multi-screen theatre in Europe has a 56ft diagonal screen. To actually see true 4K you need to sit in the front three rows, if in the future that screen upgrades to 6K or even 8K then you would need to stand in the area between the front row and the screen to actually see 6K or 8K which is totally impractical. We make 6K and 8K movie cameras now but the images are output at 4K, we oversample to retain the colour space but sacrifice some resolution.

Most modern camera electronics do a good job of minimising photon shot noise so the S/N ratio has improved vastly and the old argument between small & large pixels is less of an issue than it was and is still improving. So where can it improve? Lenses still have room for improvement and are improving, by this I mean aberrations and vignetting rather than all out resolution improvements. Dynamic range is improving and will continue to do so and with it so will colour gamut.

In reality the test charts are a guide, cameras & lenses are tools and tools are only as good as the people that use them. Forget charts, forget pixel peeping and go and shoot as often as you can and experiment that way your level of photography grows and you can enjoy the output not technical specs.
You are looking from just one perspective and missing that there is a very large community of nature shooters, especially bird photographers, who frequently have to crop and view or print a tiny portion of the image! Just look at the Bird Portraits and BIF threads, they are nearly all crops. What you consider as pixel peaking is an actual practical reality in that world. Pixel-level sharpness does matter there. As they say YMMV as each type of photography has its own needs.
 
  • Like
  • Love
Reactions: 3 users
Upvote 0
May 11, 2017
1,365
635
5DIV with 30Mp sensor can be printed at 300ppi on up to A2. If cropped, it won't reach 300ppi. So if your target is a large high quality print, you'll likely be pixel-peeping.
I know though that 200-240ppi is also ok and in many cases indistinguishable from 300ppi.
If you had said this earlier this sub thread might have been shorter. It is often hard for a reader to guess the implicit qualifications in the mind of the poster.
 
  • Like
Reactions: 1 user
Upvote 0
Feb 28, 2013
1,612
272
70
You are looking from just one perspective and missing that there is a very large community of nature shooters, especially bird photographers, who frequently have to crop and view or print a tiny portion of the image! Just look at the Bird Portraits and BIF threads, they are nearly all crops. What you consider as pixel peaking is an actual practical reality in that world. Pixel-level sharpness does matter there. As they say YMMV as each type of photography has its own needs.
Im not missing that at all. Optimisation is the name of the game its pointless having a 120MP camera if your not using lenses that match the pixel pitch of the sensor. Equally defraction becomes an issue the more you raise the MP as does camera shake particularly if you cropping for birding.
Its no mistake that Canon fast RF primes are at least a 1/3rd more expensive than their EF cousins, quality costs. I would also image they have raised the line pairs per MM for the RF mount especially as they know their forward road map. No such thing as a free lunch.
 
Upvote 0

tron

CR Pro
Nov 8, 2011
5,222
1,616
Im not missing that at all. Optimisation is the name of the game its pointless having a 120MP camera if your not using lenses that match the pixel pitch of the sensor. Equally defraction becomes an issue the more you raise the MP as does camera shake particularly if you cropping for birding.
Its no mistake that Canon fast RF primes are at least a 1/3rd more expensive than their EF cousins, quality costs. I would also image they have raised the line pairs per MM for the RF mount especially as they know their forward road map. No such thing as a free lunch.
1. I agree about DLA. But the drop of IQ is gradual both for DLA and for lenses that do not cope 100% with very high Mpixel cameras.
2. The higher price can be mainly due to the fact that the lenses are new and not many years old like the EF versions.
3. We do not know if the RF versions are better. TDP shows the top 2.8L zooms 15-35 and 24-70 as equal with the RF versions having even more vignetting. The only clear advantage is the use of IS.
 
  • Like
Reactions: 1 user
Upvote 0
Feb 13, 2018
209
178
Very good points and lenses are important, but I was checking just one specific aspect of sensor performance. I'm a pixel-peeper and I spend quite some time in 1:1 view mode when editing. More noise means less room for certain things such as shadow lifting, sharpening, noise reduction, heavy or not so heavy crop etc.
Then it is simple. The 1DXiii (and rumored R6) with their "low" 20mp sensors are the cameras for you.
More pixels, more noise. At an extreme a theoretical 1 pixel sensor has the highest accuracy of luminosity. Another theretical sensor with infinite pixels will have infinite noise (it is essentially a binary on/off /pixel). Real sensors are in the middle and you get to choose your preference. As tech improves, noise is reduced (sensitivity, adc, software etc.)
 
Upvote 0

cayenne

CR Pro
Mar 28, 2012
2,868
796
I have a question. Do you think next very high res FF bodies will extinguish the MF market? Or is there a future for it? With all the new techs coming in mirrorless bodies, the prices and the flexibility it provides in real life, how could MF shooters resist to switch?.. :unsure:

Well, there still is something to be said for MF's larger physical sensors and >= PM counts for resolution, etc....

I don't think that market is going to go away very soon....but who knows.

Interesting times we live in.
 
  • Like
Reactions: 1 user
Upvote 0

Sporgon

5% of gear used 95% of the time
CR Pro
Nov 11, 2012
4,719
1,537
Yorkshire, England
Well, there still is something to be said for MF's larger physical sensors and >= PM counts for resolution, etc....

I don't think that market is going to go away very soon....but who knows.

Interesting times we live in.

Indeed it's quite easy to get 200mp equivalent resolution out of a 6x7 MF piece of film and without the diffraction limit that's going to plaque those seeking maximum resolution from 100 mp + in FF format.
 
Upvote 0
Then it is simple. The 1DXiii (and rumored R6) with their "low" 20mp sensors are the cameras for you.
So if I complain about noise in a car when it speeds above 60, your suggestion would be to buy a car that has a hard limit of 40?

More pixels, more noise.

It simply is not true, had you actually read my message. It shows examples of sensors with smaller pixels and less noise.
 
Upvote 0
Feb 13, 2018
209
178
So if I complain about noise in a car when it speeds above 60, your suggestion would be to buy a car that has a hard limit of 40?

It simply is not true, had you actually read my message. It shows examples of sensors with smaller pixels and less noise.

No, I'm merely saying that if you spend a lot of time in 1:1, you are going to have a better time with the 20mp sensor compared to a higher resolution sensor (given same generation of sensor tech).

What I and some others say is physics. An 80mp sensor receives 1/4th of photons/pixel compared to a 20mp one, while maintaining a similar noise level (if its same tech). However it also enables a less intrusive noise canceling at same correction levels.
Down-converting to the same resolution should yield a similar noise performance (this is why DXO normalizes resolution).

Your comparison is mood, because it varies sensor tech and resolution. It simply shows that not all sensors are created equal. I'd even go so far as to say that random noise is less of an issue in general (easily corrected) but other sensor properties can be a lot harder to deal with e.g. banding

Hope this helps.
 
Last edited:
  • Like
Reactions: 1 user
Upvote 0
No, I'm merely saying that if you spend a lot of time in 1:1, you are going to have a better time with the 20mp sensor compared to a higher reaolution sensor (given same generation of sensor tech).
If I buy a high-res camera, I intend to use this high resolution somehow. Print large or crop heavy etc. If I'm always to downsample, you're right, I can buy a lower-res camera.

But I'm interested in high resolution and noise performance at native resolution. I'm not interested in sensor performance at an arbitrary downsampling ratio.

It simply shows that not all sensors are created equal.

Exactly. It shows what I get at different native resolutions and different sensor designs from different manufacturers.

Downconverting to the same resolution should yield a similar noise performance (this is why DXO normalizes resolution).

First of all, it doesn't yield similar performance even after normalisation, and that allows to compare sensors using a different metric that eliminates sensor resolution. Normalisation is good if you want to compare sensors with different resolutions and rank them. It's also a useful metric, but can only be used within DxO ranking system. But you lose information on how sensors perform at their native resolution.

You can use different metrics/rankings for different aspects of camera performance, DxO, Photostophotos, pixel noise at 1:1, whatever, why do you see a problem there?
 
Upvote 0
Jan 27, 2020
826
1,796
FWIW...recently bought an older EF 100-300 lens and took some shots on my R in crop mode. Then took the same shots - same composition, same ISO, same f-stop (wide open at 5.6 on this lens) - using the M5. Yes, I know not the same sensor, but similar generation. Images from the crop camera (and smaller pixels) definitely had more noise at the image (not pixel) level. Your results may differ.
 
Last edited:
Upvote 0
Feb 13, 2018
209
178
If I buy a high-res camera, I intend to use this high resolution somehow. Print large or crop heavy etc. If I'm always to downsample, you're right, I can buy a lower-res camera.
But I'm interested in high resolution and noise performance at native resolution. I'm not interested in sensor performance at an arbitrary downsampling ratio.

Exactly. It shows what I get at different native resolutions and different sensor designs from different manufacturers.
That all good and well. We all would like a high-res high-iso sensor but they have to be traded off against each other. So it looks like we get 3 different cameras with 3 vastly different resolutions. I'm trying to explain that high-res will have limitations at high iso (see 5Ds) and lower res will do antiproportionally do better in low light (see 5Div and 1DXiii). This is entirely based on the assumption that Canon will use similar sensor tech, otherwise this statement is mood (see 6Dii vs 5Div)

First of all, it doesn't yield similar performance even after normalisation, and that allows to compare sensors using a different metric that eliminates sensor resolution. Normalisation is good if you want to compare sensors with different resolutions and rank them. It's also a useful metric, but can only be used within DxO ranking system. But you lose information on how sensors perform at their native resolution.

You can use different metrics/rankings for different aspects of camera performance, DxO, Photostophotos, pixel noise at 1:1, whatever, why do you see a problem there?
I don't when the statemet refers to SNR of each pixel/signal (as you do in this post). The argument becomes a bit misleading if this is used as a basis to compare sensor performance (e.g. by plotting them side-by-side) of cameras build for largely different use cases. Higher res is just not universally better (or worse) and Canon's flagship uses the lowest resolution sensor of them all for a reason.
 
Upvote 0
The argument becomes a bit misleading if this is used as a basis to compare sensor performance (e.g. by plotting them side-by-side) of cameras build for largely different use cases.
But there's the same problem with DxO and normalisation. You can compare score for different cameras but the comparison becomes a bit misleading as you're probably not buying a hypothetical low-res but high-ranked camera when you absolutely need a 80Mp one which ranked lower on DxO. Also the absolute value of the 'landscape dynamic range' from DxO becomes useless if you want to print your 80Mp at max size, that dynamic range value wasn't calculated for your conditions.

After normalisation like they do on DxO and Photonstophotos, you deal with just one parameter (measured dynamic range or landscape score) and you can rank cameras/sensors. Their ranking is useful but still has limitations.

When you look at per-pixel performance, you can't easily rank sensors because you have to consider two parameters: noise and resolution. However you can (say in the future) look at per-pixel noise on R5 vs R5s and decide if you want to trade resolution for noise.

So both methods are useful within their scope and when you understand their limitations.
 
Upvote 0

SteveC

R5
CR Pro
Sep 3, 2019
2,678
2,592
FWIW...recently bought an older EF 100-300 lens and took some shots on my R in crop mode. Then took the same shots - same composition, same ISO, same f-stop (wide open at 5.6 on this lens) - using the M5. Yes, I know not the same sensor, but similar generation. Images from the crop camera (and smaller pixels) definitely had more noise at the image (not pixel) level. Your results may differ.

It would be interesting to see what you get when one compares those to the newer sensor in the 90D and M6 mk 2
 
  • Like
Reactions: 1 user
Upvote 0
Jan 27, 2020
826
1,796
It would be interesting to see what you get when one compares those to the newer sensor in the 90D and M6 mk 2
It looks like - based on photons to photos website - that the new cameras you mention have approximately 1/2 stop improvement in DR compared to the M5. Will be interesting to see if the new R models will have a similar improvement.

My guess is that 1/2 stop will not significantly alter my results. Again, others may have different results. Without a tripod, as I mentioned in a different thread, I also had no noticeable increase in resolution shooting at 1/500th of a second. Some pics with the FF crop were better, some with the more pixels of the M5 were. How still I could hold the camera seemed to be the deciding factor. Again, just my conclusion, but my guess is that many folks shooting very high MPs aren't actually getting much better resolution that if they were shooting with a much lower MP camera if shooting hand held.
 
  • Like
Reactions: 1 users
Upvote 0