Update: The Canon EOS R3 will be officially announced on June 29th

Tirmite

I'm New Here
Aug 6, 2018
15
8
I doubt it , more likely the same or less mp : I predict 24mp for the R1 which is going to be their flagship professional sports camera with probably twin Digic X processors and twin CF Express slots and maybe 40 fps, but they will probably make a high resolution R5 variant of perhaps 80-90mp
I respectfully disagree. I'm going to bet they go back to the days of offering two high end cameras for two distinct market segments like when they had an "S" model 1D for studio shooters, except the numbering will be different. R3 sounds like it can replace and exceed what the 20MP 1Dx currently does. The R3 will be the pro sports camera. And hit 30fps (like an assigned custom function button) and your "stills" just became a short video without having to stop and switch to video mode. That sounds like a sport photographer's dream. I think the R1 will be the high-end megapixel version for studio users/architectural/etc. who want/need massive files. Sony has already shown it can be done with the A1 and its 50MP sensor. Canon is almost forced to meet or exceed that milestone. Doubt there will ever be an R1c because they learned the hard way how the 5D cannibalized their video camera sales. It's why they priced the 1Dc at $10K when it was introduced. At that price point it didn't threaten there video sales. They probably won't offer a "killer" hybrid that kills their Cinema line sales. That's my theory, anyway.
 
  • Like
Reactions: Bob Howland

Chig

Birds in Flight Nutter
Jul 26, 2020
420
505
Orewa , New Zealand
I respectfully disagree. I'm going to bet they go back to the days of offering two high end cameras for two distinct market segments like when they had an "S" model 1D for studio shooters, except the numbering will be different. R3 sounds like it can replace and exceed what the 20MP 1Dx currently does. The R3 will be the pro sports camera. And hit 30fps (like an assigned custom function button) and your "stills" just became a short video without having to stop and switch to video mode. That sounds like a sport photographer's dream. I think the R1 will be the high-end megapixel version for studio users/architectural/etc. who want/need massive files. Sony has already shown it can be done with the A1 and its 50MP sensor. Canon is almost forced to meet or exceed that milestone. Doubt there will ever be an R1c because they learned the hard way how the 5D cannibalized their video camera sales. It's why they priced the 1Dc at $10K when it was introduced. At that price point it didn't threaten there video sales. They probably won't offer a "killer" hybrid that kills their Cinema line sales. That's my theory, anyway.
Interesting ideas but the following seem to favour my predictions :

Canon have stated that the R3 doesn't replace the 1DXiii and isn't their new flagship model which is all fact

There have been rumours of a high mp version of the R5 and surely this makes more sense as very few users will prefer the integrated and very expensive body for architectural work and the few that want it for studio work would be fine with a vertical grip instead.

Canon seems to consider the 1 D series as both their flagship model line and mostly aimed at professional sports shooters

Exciting times for those of us that prefer Canon but also for all camera companies and it'll very interesting to see what new and amazing gear comes out over the next year or two.

Just imagine what will be possible in 10 years from now
 
  • Like
Reactions: john1970 and pzyber

Chig

Birds in Flight Nutter
Jul 26, 2020
420
505
Orewa , New Zealand
Why would Canon follow up a pro body sports camera with another one? Why would anyone need 45 fps? May as well use video.

The Nikon Z9 will be their flagship with likely 50 MP, and a pro body. Canon won't challenge it with an R5 body camera. I expect the R1 to have at least 60 MP, probably more.
Well the R1 probably won't come out until 2024 at the next olympics and they will be releasing several cameras before then

Professional sports shooters always want higher fps to catch the perfect moment when a bat hits a ball or a diver just touches the water , etc and they definitely don't want bigger file sizes which offer little benefit and just slow their work flow down. Many of them just shoot jpeg as the results are plenty good enough.
 
  • Like
Reactions: pzyber

SwissFrank

from EOS 1N to R
Dec 9, 2018
624
352
Can you clarify. If they each individually have more noise (your comment above), how do they average to less noise as a single pixel?

It's a property of math, relied upon in statistics.

Say we're imaging something half-reflective of light, and it's so dark that even a pure white object will only yield one photon in each pixel of a hi-MP sensor. So our gray object should yield a half-photon per pixel. NO single pixel will have the correct answer of half-reflective, because there's no such thing as a half-photon. Instead they'll either have twice the real value, or a zero value. Noise is +- 100%, basically! (This is like: we know the odds when flipping a coin is 50% either way, but if we then just flip a coin one time, we cannot get 50%, we only get 100% heads or 0% heads.)

Now, take 4 pixels of this hi-MP sensor, either getting 0 photons (black) or 1 (white), and sum them. Consider receipt of a photon as a coin flip. Giving 0 for no photon, and 1 for a photon, the equally likely possibilities are:

0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111

Now you'll see a total of:

0 -- 1 time in 16. We're reporting this larger pixel to be black, 100% off its true value.
1 -- 4 times in 16. We're reporting this larger pixel to be 25, 50% off its true value.
2 -- 6 times in 16. We're reporting this larger pixel to be 50, its true value.
3 -- 4 times in 16. We're reporting this larger pixel to be 75, 50% off its true value.
4 -- 1 time in 16. We're reporting this larger pixel to be 100, 100% off its true value.

Instead of being off the correct value of .5 by 100% as before, now we're off by: 100 * 1/16 + 50 * 4/16 + 0 * 6/16 + 50 * 4/16 + 100 * 1/16 = 37.5%.

Now, take a lo-MP sensor with 1/4 the resolution. Its pixels are big enough they'll get 4 photons from a white object. Our gray object should return 2 photons. The math works identically to the above five cases and their chances of happening, giving the same 37.5% noise.

So, back to sensors. An 80MP back-side sensor should capture as many photons total as a 20MP sensor, though that means only 1/4 the photos per pixel and thus far higher noise per pixel. But then average four neighboring pixels together and the noise level comes down to exactly the same as the 20MP sensor.
 

SwissFrank

from EOS 1N to R
Dec 9, 2018
624
352
While the downsized R5 files @ 12,800 were not bad, I felt the R6 files were "crisper" with less noise when shot in similar lighting situations.
First, and this is absolutely no insult, if you knew which was which, a subconscious assumption may have biased your analysis. That's why real science is based on blind or even double-blind testing: make people rate the photos NOT knowing which camera took which. And maybe even have the person making the test not know which is which, so they don't fall into some unspoken subconscious ordering.

Second, Canon has always had sensors with the wiring on the front of the sensor (front-side sensor). The double MP sensor has twice the wires interfering with light, so it is NOT true that a hi-MP front-side sensor will pick up as many TOTAL photons as a lo-MP sensor. In contrast the R3 and one assumes all future Canons are back-side. The entire front of the sensor is sensitive to photons, whether there's 1 big pixel or a billion. In this case, while every single pixel in the hi-MP sensor will have more noise, the scene as a whole won't, so down-sampling (especially, downsampling by an integer factor like exactly 4 pixels to 1) should give an identical image with identical noise characteristics as well as identical resolution.
 

Bahrd

Red herrings...
Jun 30, 2013
184
127
It's a property of math, relied upon in statistics*.
* Photon's statistics is governed by Poisson distribution. This means, in particular, that not only 0 or 1 photon can be captured but also 2, 3 or even more (yet much less likely).
 
  • Love
Reactions: SwissFrank

SwissFrank

from EOS 1N to R
Dec 9, 2018
624
352
* Photon's statistics is governed by Poisson distribution. This means, in particular, that not only 0 or 1 photon can be captured but also 2, 3 or even more (yet much less likely).
Oh, exactly, of course! I'm just trying to make it as simple as possible. You'll agree that the result I'm presenting is correct, even though the example is cutting corners, right? Summing four Poisson distributions yields a Poisson distribution, but then when you divide by four you have a much smaller standard deviation. That's what I'm trying to convey.
 
  • Love
Reactions: Bahrd

Bahrd

Red herrings...
Jun 30, 2013
184
127
Oh, exactly, of course! I'm just trying to make it as simple as possible. You'll agree that the result I'm presenting is correct, even though the example is cutting corners, right? Summing four Poisson distributions yields a Poisson distribution, but then when you divide by four you have a much smaller standard deviation. That's what I'm trying to convey.
Absolutely! It's truly amazing that both Poisson and Bernoulli laws are so effective models of photon's behavior.
 

HMC11

Travel
CR Pro
Sep 5, 2020
58
51
It's a property of math, relied upon in statistics.

Say we're imaging something half-reflective of light, and it's so dark that even a pure white object will only yield one photon in each pixel of a hi-MP sensor. So our gray object should yield a half-photon per pixel. NO single pixel will have the correct answer of half-reflective, because there's no such thing as a half-photon. Instead they'll either have twice the real value, or a zero value. Noise is +- 100%, basically! (This is like: we know the odds when flipping a coin is 50% either way, but if we then just flip a coin one time, we cannot get 50%, we only get 100% heads or 0% heads.)

Now, take 4 pixels of this hi-MP sensor, either getting 0 photons (black) or 1 (white), and sum them. Consider receipt of a photon as a coin flip. Giving 0 for no photon, and 1 for a photon, the equally likely possibilities are:

0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111

Now you'll see a total of:

0 -- 1 time in 16. We're reporting this larger pixel to be black, 100% off its true value.
1 -- 4 times in 16. We're reporting this larger pixel to be 25, 50% off its true value.
2 -- 6 times in 16. We're reporting this larger pixel to be 50, its true value.
3 -- 4 times in 16. We're reporting this larger pixel to be 75, 50% off its true value.
4 -- 1 time in 16. We're reporting this larger pixel to be 100, 100% off its true value.

Instead of being off the correct value of .5 by 100% as before, now we're off by: 100 * 1/16 + 50 * 4/16 + 0 * 6/16 + 50 * 4/16 + 100 * 1/16 = 37.5%.

Now, take a lo-MP sensor with 1/4 the resolution. Its pixels are big enough they'll get 4 photons from a white object. Our gray object should return 2 photons. The math works identically to the above five cases and their chances of happening, giving the same 37.5% noise.

So, back to sensors. An 80MP back-side sensor should capture as many photons total as a 20MP sensor, though that means only 1/4 the photos per pixel and thus far higher noise per pixel. But then average four neighboring pixels together and the noise level comes down to exactly the same as the 20MP sensor.
That was well explained. Let me see if I get it right. Your 'model' assumes a uniform flux of photons, ie. it is equally likely to fall on any of the pixels. If, instead of that, we have a have a light (photon) distribution that have varying areas of light intensity (as in a real-world image), then are the following conclusions valid?

(a) in areas where there is sufficient light (well exposed), there will not be noticeable difference between an image taken by the higher MP sensor vs the low MP sensor when viewed at the pixel level (ie. at a viewing distance/enlargement of the image such that our eye can just about able to see each individual pixel) of the lower MP camera, i.e. say we compare 80MP vs 20MP, then when viewed at the pixel level of the 20MP image, the 80MP image would be essentially the same assuming everything else being equal (processing sensor tech etc). If we zoomed in further, then the 80MP image would appear sharper as it has more resolution.
(b) In low light areas, we can detect the difference between the 20MP & 80MP image when viewed at the pixel level of the 20MP sensor, i.e. the 80MP would appear noisy. Essentially, this is due to the 20MP pixel averaging the photons it receives and producing a single response for that pixel, whereas the 80MP would produce 4 signals that are not averaged, and hence would 'look' more noisy. I assume that if we view it from a greater distance such that our eyes essentially see an average of the 4 smaller pixels of the 80MP sensor, then the images from both would appear to be the 'same'.

Things are probably a lot more complicated than the above, however, does it mean that a higher MP sensor would have worst low light performance unless its technology, processing etc can sufficiently compensate?
 
  • Like
Reactions: SwissFrank

Sporgon

5% of gear used 95% of the time
CR Pro
You can be 100% sure the RP replacement won't have the R3's sensor or anything similar. In best case scenario it will get the R6's sensor or something similar.
In this situation, when looking at a ‘consumer’ targeted camera, I doubt the RPII (or whatever) would drop in mp. But I am usually wrong when guessing what Canon will do next ;)
The current RP is remarkably capable and Canon let the hand brake off for an “entry level” camera as long as you can live with the off chip ADC design.
 

EOS 4 Life

EOS R
Sep 20, 2020
1,069
841
Things are probably a lot more complicated than the above, however, does it mean that a higher MP sensor would have worst low light performance unless its technology, processing etc can sufficiently compensate?
I generally hate image tests for this reason.
They tend to measure the amount of noise.
Not all noise is displeasing, some of it can be easily corrected in post, and there is a lot of noise correction inside the camera.
Most of what they are testing is the built-in noise correction inside the camera.
In reality, all that matters is the final image and the time, effort, cost, and expertise required to achieve it.
The easiest thing to do is try cameras out and buy the ones that get you the best result for what you do.
To me, the biggest issue is the signal-to-noise ratio of camera information.
 

EOS 4 Life

EOS R
Sep 20, 2020
1,069
841
In this situation, when looking at a ‘consumer’ targeted camera, I doubt the RPII (or whatever) would drop in mp. But I am usually wrong when guessing what Canon will do next ;)
The current RP is remarkably capable and Canon let the hand brake off for an “entry level” camera as long as you can live with the off chip ADC design.
There is not supposed to be a camera named RP Mark II.
There may be an RP2 but that is not really an indication of whether it will be spec'd above or below the RP.
I am not sure that Canon will even start future camera names with RP but it makes sense to me to use that for any cameras with that body type.
It would also not surprise me to see a camera with an RP body with higher specs than a camera with an R body but it might be like the Rebel DSLRs where the smaller cameras are all cheaper.

If Canon stays true to form then the follow-up would be RP10, entry-level would be RP100, and top-level would be RP1.
 
Last edited:

blackcoffee17

EOS RP
Sep 17, 2014
688
852
In this situation, when looking at a ‘consumer’ targeted camera, I doubt the RPII (or whatever) would drop in mp. But I am usually wrong when guessing what Canon will do next ;)
The current RP is remarkably capable and Canon let the hand brake off for an “entry level” camera as long as you can live with the off chip ADC design.

The RP II just needs a bit better sensor (can stay 24-26mp), maybe one with on-chip ADC and full frame 4K with dual pixel AF. Even if it's only at 30fps. And maybe make it a bit faster, like 5-6 fps with tracking AF.
 

Hector1970

EOS R
CR Pro
Mar 22, 2012
1,348
600
Well the R1 probably won't come out until 2024 at the next olympics and they will be releasing several cameras before then

Professional sports shooters always want higher fps to catch the perfect moment when a bat hits a ball or a diver just touches the water , etc and they definitely don't want bigger file sizes which offer little benefit and just slow their work flow down. Many of them just shoot jpeg as the results are plenty good enough.
The world cup in Nov 22 so this is plenty of time to have an R1 available. I can't see an R1 having less MP than an R3. The R1 will have to be better in all aspects to an R3 to justify its position at the top of the tree. File sizes are less relevant with communication speeds rapidly increasing, cropping allows for more editing options. You can always reduce file size - you can't increase it. FPS could would arguebly reach a point where its an obstacle to work flow. HIgher FPS means more sorting through practically identical images. You may as well take stills from video. For an R1 to justify its price it will have almost amazing autofocus ability. I wonder myself if the R3 was something planned from the beginning or whether it was an afterthought of an R1 developed so far and Canon either thought it didn't justify the R1 badge or some breakthrough that wouldn't be ready in time and decided to go with an intermediate solution of calling it an R3 to be competitive with the top Sony camera. I'd certainly take 50MP over 50FPS in terms of usefulness. I can't wait for the R3 reviews. If there is a major leap forward in focusing ability it will be of great interest to me.
 

EOS 4 Life

EOS R
Sep 20, 2020
1,069
841
I wonder myself if the R3 was something planned from the beginning or whether it was an afterthought of an R1 developed so far and Canon either thought it didn't justify the R1 badge or some breakthrough that wouldn't be ready in time and decided to go with an intermediate solution of calling it an R3 to be competitive with the top Sony camera.
If R3 was the planned successor to the 1DX then it would have 2 CF Express slots.
I get the feeling R3 was planned before the Z9 announcement and Canon is not as obsessed with what Sony does as social media is.
 
Aug 7, 2018
353
305
Isn't there a kind of noise that gets overproportionally worse when pixels get smaller? Otherwise it is strange that sensors optimized for low noise often have a low resolution. I am thinking about that sensor by Canon for example that can do images with ISO 3,276,800 or so and only has a very low resolution. Of course if noise is random at each pixel, you can average it out, but there also is noise with other frequencies that effects a hundred or so pixels in the same way, Averaging it out will not work there and it could get worse on small pixel, because on small pixels tha signal has to be amplified much more than on a large pixel.
 

EOS 4 Life

EOS R
Sep 20, 2020
1,069
841
Isn't there a kind of noise that gets overproportionally worse when pixels get smaller? Otherwise it is strange that sensors optimized for low noise often have a low resolution. I am thinking about that sensor by Canon for example that can do images with ISO 3,276,800 or so and only has a very low resolution. Of course if noise is random at each pixel, you can average it out, but there also is noise with other frequencies that effects a hundred or so pixels in the same way, Averaging it out will not work there and it could get worse on small pixel, because on small pixels tha signal has to be amplified much more than on a large pixel.
Cameras focused on high ISO have lower resolution.
I guess maybe they also have less noise at low ISO but they capture a lot less information.
 

lethiferous

EOS M50
Nov 3, 2020
30
15
Yes, at the current rate of market change that should only take a decade or so. Did you know that the best-selling ILC in Japan for the last two months was a Canon DSLR?
Pretty cure cost of entry and low price ef glass plays a huge role in that. A rebel with access to old cheap EF lenses is a lot cheaper.