Canon EOS R body with more than 75mp on the horizon [CR2]

Sep 10, 2018
92
40
maybe we can hope to see it sooner than later?

Photokina 2019 chancelled.

Imaging industry and Koelnmesse decide on new starting point for the new annual cycle

Following a successful photokina 2018, the German Photo Industry Association (PIV), as conceptual sponsor of the trade fair, and the event's organiser Koelnmesse have agreed not to organise the next leading global trade fair in May 2019, as initially planned, but in May 2020. From Wednesday 27 May 2020 to Saturday 30 May 2020, all the market leaders in the imaging industry are expected once again in Cologne. The decision to postpone the start of the announced annual cycle by one year is intended to give all participants the opportunity to further develop the new concept for photokina and to tap into new target groups among exhibitors and visitors in order to heighten the status of the trade fair as a global platform for the photography and imaging industry.
 
Nov 16, 2018
14
3
They can easily make a >75MP full frame sensor which isn’t BSI.

This APS-H sensor is not BSI, is smaller than full frame, and has 60% more than 75MP.
Of course Canon can pack a ton of pixels on a small area already, but that makes it practically useless except some very special scenarios in which we hardly ever shoot. Just look at their own video, and notice the heavily blown-out highlights on the right capture that are not blown out on the left one. The dynamic range they got out of the sensor downright sucks, although on the very shot it looks contrasty and appealing. Real world shooting is an entirely different story, though!

Keep in mind that the pixel pitch (the area of a single pixel) is a key determinant to the SNR of the sensor and thus the dynamic range you get out of it. Using a given technology, sensor sensitivity goes hand in hand with the pixel pitch. I don't need 75 MP or 122 MP if it means that I only can record tonal detail in a range of 9 stops of light or less! That means that most of my landscape shots would either have a badly blown out sky, or heavily noisy foreground (or whatever would be in a bit of shadow); and I would have to resort to doing HDR for pretty much every single shot. Imagine the image quality of some prehistoric entry-level digital Canon DSLR... except just resolution that I often can get by stiching multiple shots into a panoramatic image anyway. That's probably what a non-BSI 75 or 122 MP CMOS sensor would give you, even full frame.

The newest Phase One IQ4 has gone up to 150 MP, but that is a huge full frame medium format sensor with 16-bit ADC (that costs $50k - just the digital back). Although it roughly has the pixel pitch of the 5DsR, unlike the 5DsR it is a BSI CMOS sensor and also has 2 bits more ADC (=> whole 4 times the amount of tonal resolution captured), which allows for way cleaner input.

Moreover, (as surely some older post already reported), Sony is about to roll out a 60 MP 16-bit BSI CMOS sensor. My guess is that it is going to beat the output of the GFX in pretty much every way except the pixel sharpness: check this (Sony Alpha Rumors). No wonder in the end, all high-end cameras except perhaps just Canon is buying sensors from Sony. Of course they have a strategic advantage when it comes to technology. I'd love to see Canon competing with them and I really hope that Canon is going to deliver something similar in their high-MP EOS R, although I see just too many reasons to doubt it...
 
Nov 16, 2018
14
3
Of course Canon can pack a ton of pixels on a small area already, but that makes it practically useless except some very special scenarios in which we hardly ever shoot. Just look at their own video, and notice the heavily blown-out highlights on the right capture that are not blown out on the left one. The dynamic range they got out of the sensor downright sucks, although on the very shot it looks contrasty and appealing. Real world shooting is an entirely different story, though!

Keep in mind that the pixel pitch (the area of a single pixel) is a key determinant to the SNR of the sensor and thus the dynamic range you get out of it. Using a given technology, sensor sensitivity goes hand in hand with the pixel pitch. I don't need 75 MP or 122 MP if it means that I only can record tonal detail in a range of 9 stops of light or less! That means that most of my landscape shots would either have a badly blown out sky, or heavily noisy foreground (or whatever would be in a bit of shadow); and I would have to resort to doing HDR for pretty much every single shot. Imagine the image quality of some prehistoric entry-level digital Canon DSLR... except just resolution that I often can get by stiching multiple shots into a panoramatic image anyway. That's probably what a non-BSI 75 or 122 MP CMOS sensor would give you, even full frame.

The newest Phase One IQ4 has gone up to 150 MP, but that is a huge full frame medium format sensor with 16-bit ADC (that costs $50k - just the digital back). Although it roughly has the pixel pitch of the 5DsR, unlike the 5DsR it is a BSI CMOS sensor and also has 2 bits more ADC (=> whole 4 times the amount of tonal resolution captured), which allows for way cleaner input.

Moreover, (as surely some older post already reported), Sony is about to roll out a 60 MP 16-bit BSI CMOS sensor. My guess is that it is going to beat the output of the GFX in pretty much every way except the pixel sharpness: check this (Sony Alpha Rumors). No wonder in the end, all high-end cameras except perhaps just Canon is buying sensors from Sony. Of course they have a strategic advantage when it comes to technology. I'd love to see Canon competing with them and I really hope that Canon is going to deliver something similar in their high-MP EOS R, although I see just too many reasons to doubt it...
There is actually another big issue with that many megapixels on a small area such as a 35mm sensor - diffraction. On the 5DsR, diffraction can be noticed as early as from f/7.1. If you max the megapixels to 75 and beyond on a 35mm sensor, you'll start to see it even earlier. Then at so often highly practical f-stops as f/9 and f/11, the advantage of those megapixels will start to fade away; and your high-MP camera will perform similarly to a lower MP camera, except that the lower MP camera will have greater tonal range. At that point, I'd rather have the latter. Just ask yourself based on your own shooting - what f-stops do you most often use? What is the distribution of f-stops across your real-world shooting? That should give you some clue as to whether the megapixels are worth it for you.
 

3kramd5

EOS 5D Mark IV
Mar 2, 2012
2,655
165
Canon may eventually go to BSI, but I imagine it would be in pursuit of bandwith from stacked designs rather than a marginal increase in per pixel DR.

An interesting compromise without changing to BSI would be binning. In high brightness range scenes, bin 4 pixels before the digital quantization for an according increase in effective well capacity and therefore dynamic range at the expense of resolution. In lower brightness range scenes, e.g. controlled light or low-sun landscapes, use the full resolution.
 
Likes: mk0x55
Nov 16, 2018
14
3
Canon may eventually go to BSI, but I imagine it would be in pursuit of bandwith from stacked designs rather than a marginal increase in per pixel DR.

An interesting compromise without changing to BSI would be binning. In high brightness range scenes, bin 4 pixels before the digital quantization for an according increase in effective well capacity and therefore dynamic range at the expense of resolution. In lower brightness range scenes, e.g. controlled light or low-sun landscapes, use the full resolution.
I can imagine that, too - in the end, it really depends on the majority of Canon's customer base, and they are mostly not technically-oriented pixel peepers that would demand ultimate image quality and complain if they don't get it. Binning sounds like a nice trick, but I generally find it hard to justify not upgrading the technology when virtually all competitors have done so. Their cameras are going to produce output that is technically way superior. BSI gives around 1 stop of advantage; 16-bit ADCs 2 stops. 3 stops of DR advantage is a non-trivial amount, especially when it makes a difference in my shooting, and Canon would really have to deliver amazing value in some other area for me to favor it in a buying decision. Canon has great ergonomics, build quality, colors and lenses; but other manufacturers are closing in pretty well. Fujifilm is a great example; but looking at Sony and the trends... it looks like they will get there.
 
Likes: bokehmon22
Mar 2, 2012
2,655
165
I can imagine that, too - in the end, it really depends on the majority of Canon's customer base, and they are mostly not technically-oriented pixel peepers that would demand ultimate image quality and complain if they don't get it. Binning sounds like a nice trick, but I generally find it hard to justify not upgrading the technology when virtually all competitors have done so. Their cameras are going to produce output that is technically way superior. BSI gives around 1 stop of advantage; 16-bit ADCs 2 stops. 3 stops of DR advantage is a non-trivial amount, especially when it makes a difference in my shooting, and Canon would really have to deliver amazing value in some other area for me to favor it in a buying decision. Canon has great ergonomics, build quality, colors and lenses; but other manufacturers are closing in pretty well. Fujifilm is a great example; but looking at Sony and the trends... it looks like they will get there.
I’m not sure we’ve seen a one stop improvement materialize from BSI.

As to the ADC, what improvement are you measuring in stops? You can expect better gradation, but I don’t know how to express that in stops. Adding two bits would allow two additional stops of DR to be linearly represented in a digital string, but that doesn’t translate back to the pixel characteristics (i.e., well capacity and noise).
 
Last edited:
Feb 25, 2015
132
39
I’m not sure we’ve seen a one stop improvement materialize from BSI.

As to the ADC, what improvement are you measuring in stops? You can expect better gradation, but I don’t know how to express that in stops. Adding two bits would allow two additional stops of DR to be linearly represented in a digital string, but that doesn’t translate back to the pixel characteristics (i.e., well capacity and noise).
Exactly! Imagine what would happen if Canon is cheeky and says "we're using 32-bit ADCs!", will the same people that argue that bits==stops DR in Sony sensors keep arguing that and say Canon now has 32 stops of DR?
 
Exactly! Imagine what would happen if Canon is cheeky and says "we're using 32-bit ADCs!", will the same people that argue that bits==stops DR in Sony sensors keep arguing that and say Canon now has 32 stops of DR?
True, there is more complexity to it than just how many bits the ADC have. Something to compare is the PhaseOne IQ4 sensor with about the 5DsR's pixel pitch that has 15(+) stops of DR - and it employs both 16-bit ADC and BSI. With about the same pixel sizes both the shadow and highlight recovery is just amazing... there are sample IQ4 RAW files to download and play with in Capture One.
It can be other things than BSI and the ADC bit depth that account for most of the difference, but it would be nice to know what it is if not these.

Canon sensors' weak spot appears to be readout noise - a notable and ISO-invariant amount of noise that is added after every single exposure (which doesn't allow for the ISO-invariance of the sensor). Boosting ISO to 800 and above, it gets to a bit less than a stop worse than e.g., Nikon D850 or Sony A7R3 when normalized by resolution (I guess mostly due to BSI); but at lower ISOs, the noise introduced by sensor readout spoils the output badly:
https://www.dxomark.com/Cameras/Com...-versus-Canon-EOS-5D-Mark-IV___1187_1177_1106

Also, comparison of the Full SNR (logarithmic view) of the sensors reveals quite a difference. When it comes to BSI vs. non-BSI (e.g., Nikon D800E vs. D810), I see about half a stop advantage in SNR in favor of the BSI sensor, so although it's not a full stop (you're right on that), it is still quite a lot (clearly, it allows the photo sites to gather more light).

Studio comparisons (to include some actual data)...:
https://www.dpreview.com/reviews/im...r16_3=100&normalization=full&widget=1&x=0&y=0
 

3kramd5

EOS 5D Mark IV
Mar 2, 2012
2,655
165
It can be other things than BSI and the ADC bit depth that account for most of the difference, but it would be nice to know what it is if not these.
With gapless microlenses, perhaps BSI gives some slight improvement to how much light a sensor gathers per unit exposure, but since dynamic range is a function of full well capacity (i.e., full exposure), I don’t expect it helps much from that perspective. Maybe there is less noise due to some thermal improvements (the wiring is on the heat sink side). That couple marginally improve dynamic range.

I’d bet the biggest players at this point are downstream electronics (how much noise couples in the analog portion, and heat), and quantum efficiency.

When it comes to BSI vs. non-BSI (e.g., Nikon D800E vs. D810), I see about half a stop advantage in SNR in favor of the BSI sensor, so although it's not a full stop (you're right on that), it is still quite a lot (clearly, it allows the photo sites to gather more light).
I believe that is a comparison of cameras which both use FSI sensors. D850 was the first Nikon SLR to use BSI, unless I’m mistaken.

BSI is an important architectural step, but I don’t think it has been demonstrated as a panacea for visible image quality. Rather, it’s necessary for straightforward stacked sensor design.
 
Likes: mk0x55
With gapless microlenses, perhaps BSI gives some slight improvement to how much light a sensor gathers per unit exposure, but since dynamic range is a function of full well capacity (i.e., full exposure), I don’t expect it helps much from that perspective. Maybe there is less noise due to some thermal improvements (the wiring is on the heat sink side). That couple marginally improve dynamic range.

I’d bet the biggest players at this point are downstream electronics (how much noise couples in the analog portion, and heat), and quantum efficiency.



I believe that is a comparison of cameras which both use FSI sensors. D850 was the first Nikon SLR to use BSI, unless I’m mistaken.

BSI is an important architectural step, but I don’t think it has been demonstrated as a panacea for visible image quality. Rather, it’s necessary for straightforward stacked sensor design.
Thanks for sharing your insights!

Your are right, D810 has indeed a non-BSI CMOS sensor. Then the more correct comparison from me would then be the D810 vs. D850, which shows the same SNR figures, normalized to resolution. Interesting. I learned something new.

Well, I'm curious to see what Canon comes up with in the not-so-distant future.
 
... (see above) ...
Another hot question to me would be Canon's off-chip ADC architecture vs. a more modern on-chip one that Sony uses.
This might actually be the greatest source of impairment to image quality in Canon cameras... Canon's sensors themselves might actually be just as capable as Sony's.

In 2015, a senior executive of Canon said that they decided to concentrate on on-chip ADC architectures in the future:
https://www.imaging-resource.com/ne...-m-enthusiasts-more-aps-c-lenses-new-printers
Dave Etchells & Masaya Maeda said:
DE: This is actually a very technical question. I’m not sure if it’s one that you would be free to answer or not, but with sensor technology some have pointed to the analog-to-digital conversion implementations being very critical for image quality and dynamic range. Can you tell us whether Canon currently uses on-chip or off-chip A/D converters?

MM: Right now, we use both on-chip and off-chip, but recently I made the decision going forward to concentrate on the on-chip.
 

3kramd5

EOS 5D Mark IV
Mar 2, 2012
2,655
165
Yes, notionally doing the quantization on sensor should be beneficial (less opportunity for analog noise to be inserted; digital coupling is easier to mitigate). There are mixed reports about whether canon currently has on sensor ADC architecture, but I think it’s all speculative (unless chipworks has done a tear down).
 
Jun 6, 2016
269
13
Yes, notionally doing the quantization on sensor should be beneficial (less opportunity for analog noise to be inserted; digital coupling is easier to mitigate). There are mixed reports about whether canon currently has on sensor ADC architecture, but I think it’s all speculative (unless chipworks has done a tear down).
----

If Canon is going to do on-sensor-chip Digital Signal Processing (DSP) and ADC (Analog to Digital Conversion), then they might as well look to TI (Texas Instruments) for their incredibly fantastic DSP and ADC intellectual property. Especially, get the 32-bits downsampled to 16-bits per colour channel ADC so they can take best advantage of the "Natural Anti-Aliaising" that happens within the Nyquist sampling theorem.

Canon just need to stack the ADC and DSP circuits on the rear of the Canon Image Sensors. If possible, do synchronous parallel readouts for each line of pixels so that a form of global shutter can be effected at a reasonable price. Simultaneous multi-line readout isn't TRUE Global Shutter but it IS a lot better than current offerings on DSLRs!

Canon DOES NOT HAVE MUCH TIME anymore to do the above. Other players are coming online with DISRUPTIVE Audio/Video/Stills images and lens technology that gives truly Professional DSLR and Cinema-level features at Consumer/Prosumer-level prices! Canon is about to GET SQUEEZED BIG TIME so they better not hold back on the NEWEST technology and much better prices!
 
Likes: mk0x55

dtaylor

EOS Rebel T7i
Jul 26, 2011
941
84
There is actually another big issue with that many megapixels on a small area such as a 35mm sensor - diffraction. On the 5DsR, diffraction can be noticed as early as from f/7.1.
People treat diffraction like a brick wall or a boogeyman. It is neither. You are never at an IQ disadvantage for using a higher sampling frequency (more MP). The IQ improvement, versus a lower resolution sensor of the same format, does not drop off at the so called 'diffraction limit'. Diffraction just becomes apparent if you were to compare your image to the same resolution sensor in a larger format.

Well beyond that 'limit' you are not resolving the detail you otherwise could, but you're also not doing any worse than a lower resolution sensor.
 
Nov 16, 2018
14
3
I would agree. But it's not even 1 full stop better than the 5Dsr. Maybe 2/3rds stop. And that's with 3 years newer tech and the much praised Sony patents/fab.
Well, from what I could see, the D850 is quite a bit cleaner in the shadows. It might be 1 stop on the average across the tonal range though. The shadow SNR of the D810 is ~12 dB higher than the 5DsR, which equals to about 16 times cleaner than the 5DsR (i.e., 4 stops cleaner). As you go toward the highlights, it becomes better:
https://www.dpreview.com/reviews/nikon-d810/15

I own the 5DsR, shoot it and sort of love it, but I know it can be pretty horribly noisy. If I need to get rid of the noise, I need to do mean stacking, which is only usable for static scenes like landscapes if there isn't much motion that would mean hurting the image. Of course, you could argue that this noise perfectionism is just for pixel peeping... it often does not matter for practical use... well, it might, based on what you shoot and how much you need to lift the shadows to make the image you wanted. Also, to get the shadow detail to the same level as the D810, I would have to take 16 shots of the same scene and mean stack, which often is prohibitive, especially if you shoot at night with 2 minute exposures on tripod... + other two minutes to cut systematic noise and hot pixels.
I have to say though that without having this camera, I very likely wouldn't have learned tricks to get around its image quality limitations. So feel grateful in a way... although I often wish that my camera too had the image quality of the Nikon D810/D850.

People treat diffraction like a brick wall or a boogeyman. It is neither. You are never at an IQ disadvantage for using a higher sampling frequency (more MP). The IQ improvement, versus a lower resolution sensor of the same format, does not drop off at the so called 'diffraction limit'. Diffraction just becomes apparent if you were to compare your image to the same resolution sensor in a larger format.

Well beyond that 'limit' you are not resolving the detail you otherwise could, but you're also not doing any worse than a lower resolution sensor.
I fully agree and meant that - it just makes the high resolution useless if the diffraction limits you to the extent that you don't get more resolution using the higher resolution sensor than you get from the lower resolution one. In such cases, you could trade the resolution for other desirable parameters - such as the dynamic range. Thinking about 51, 75, 122 and 250 MP on 35 mm sensor, diffraction may become visible at f/8, f/6.7, f/5 and f/3.5, respectively. The Nikon D810 has it from f/9.5 and D850 from f/9. Just for the fun comparison, the $50k Phase One IQ4 back with 150 MP may suffer from diffraction from f/7.1.
https://www.photopills.com/calculators/diffraction
 
Last edited:
Jun 6, 2016
269
13
I would agree. But it's not even 1 full stop better than the 5Dsr. Maybe 2/3rds stop. And that's with 3 years newer tech and the much praised Sony patents/fab.
---

In MOST situations, this SHOULD NOT be a problem, because one SHOULD use extra light (a flash or even the cheapest 250 lumens solid diffused light) to make up that extra stop!

Just Add Extra Light = Image Noise And Light Gathering Problem Solved!

Sony currently has the best general market image sensors, but since Canon is no slouch in using advanced image processing to get cleanish images, I would say the newest Canon mirrorless offerings are about equal in terms of light gathering power versus image noise to Sony Sensor-based cameras BUT the Nikon D850 is STILL BETTER at having less noise at the same ISO-level and/or aperture! The newest Canon Full Frame mirrorless cameras are promising but NEWER players are coming very soon that WILL undercut on prices and out-deliver on technology and features.

Canon's time is short in terms of it's CURRENT price per performance and features metrics. It falls short and is ABOUT to get an tush-whooping from some pretty big electronics giants in the next little while! They need to bring out some NEW HUUUUGE-ON-FEATURES stills and video cameras lest they get dumped on their keester by some NEW biggest-boys-on-the-block company!

A large MF sensor of 8+ microns per photosite at 4:4:4/4:2:2 and 10-to-16 bits per channel interframe-encoded DCI 8k Video at 60 fps AND 50 megapixel 4:4:4 16-bits-per-channel JPEG-2000 and RAW combined Stills/Video camera WILL make that BIG SPLASH than Canon so desperately needs now! They BETTER GET HOPPING !!!!
 

dtaylor

EOS Rebel T7i
Jul 26, 2011
941
84
Well, from what I could see, the D850 is quite a bit cleaner in the shadows.
It has about 2.5ev greater DR at base ISO. By ISO 400 this drops to 1ev.

I own the 5DsR, shoot it and sort of love it, but I know it can be pretty horribly noisy.
This would be the 5th or 6th time on this website where someone has claimed this, and when I post test images the claim disappears. I'm tired of posting images and already have in this thread.

Shipping FF cameras, right now, are within 1ev of each other in terms of high ISO noise. The 5Dsr was never noisy. When it was released it was arguably the best high ISO body out with the caveat of being artificially capped to 12,800. (There was no reason for canon to cap it at 12,800 instead of 25,600. And there's really no reason for anyone to ship any ILC with a cap over 25,600 since they're all useless past that point unless your goal is a 3x5 that looks like it came from expired 110 film.)

The 5Dsr is, however, a terrible camera to pixel peep if you don't realize what's happening when you pixel peep. You can sort all the 5Dsr reviews into two categories: compared at same view/print size, and only compared while pixel peeping. The category will tell you the reviewers opinion on its high ISO performance. This is why nearly the entire world of online reviewers was saying the 5Dsr was "not a high ISO camera" while Imaging Resource was telling you how massive your prints could be even at high ISOs.

With the exception (today) of the D850, no other 35mm camera's image is magnified as much while pixel peeping.

Also, to get the shadow detail to the same level as the D810, I would have to take 16 shots of the same scene and mean stack,
That's a ridiculous claim. At base ISO you would need your normal shot and a shot at +2.5ev to match a D8x0's DR. By ISO 400 that drops to a 1ev difference, and drops a little bit more as ISOs progress. So you would need 2 exposures 1ev apart. Though, in practice, a 1ev DR difference is trivial and can basically be eliminated with NR.

I fully agree and meant that - it just makes the high resolution useless if the diffraction limits you to the extent that you don't get more resolution using the higher resolution sensor than you get from the lower resolution one. In such cases, you could trade the resolution for other desirable parameters - such as the dynamic range.
Again, we're not seeing higher DR from lower pixel density sensors in ILC cameras right now. I don't know why. In theory the full well capacity should be better the larger the pixel site. Yet the highest DR bodies for both off chip ADCs (5Dsr) and on chip ADCs (D80x0; A7r3) have relatively small pixels.
 
Likes: AlanF
Aug 16, 2012
4,230
448
I really have to agree with what dtaylor writes about DR and noise. In the range in which I mainly work, iso640 - 6400, the 5DSR, 5DIV and D850 all have the same DR http://www.photonstophotos.net/Charts/PDR.htm#Canon EOS 5D Mark IV,Canon EOS 5DS R,Nikon D850 and I sometimes show pairs of images comparing the 5DIV at iso 6400 and the 400mm DO II f/4 with the 5DSR at iso 6400 and 100-400mm II at f/5.6 pushed to iso 18900 equivalent in post (+1.56ev) where the two are very similar when scaled to the same size. It helps to have the best RAW converter for reducing noise - DxO with PRIME noise reduction. Oh, here is the bellbird again, buried deep in the New Zealand forest green canopy to prove it. 5DIV on top, 5DSR below as it clearly hasn't been registered by some.
bellbird_male_2B4A8000_DxO_CRiso6400.jpg
bellbird_male_3Q7A8969DxO_bellbirdmale_CRiso6400+1.56ev.jpg
 
Nov 16, 2018
14
3
It has about 2.5ev greater DR at base ISO. By ISO 400 this drops to 1ev.



This would be the 5th or 6th time on this website where someone has claimed this, and when I post test images the claim disappears. I'm tired of posting images and already have in this thread.

Shipping FF cameras, right now, are within 1ev of each other in terms of high ISO noise. The 5Dsr was never noisy. When it was released it was arguably the best high ISO body out with the caveat of being artificially capped to 12,800. (There was no reason for canon to cap it at 12,800 instead of 25,600. And there's really no reason for anyone to ship any ILC with a cap over 25,600 since they're all useless past that point unless your goal is a 3x5 that looks like it came from expired 110 film.)
Perhaps I didn't make myself clear enough - I was not comparing the cameras being shot at ISO 640-6400 (shooting e.g. landscapes or architectures at that high ISOs is often unacceptable to me). I meant shooting at the base ISO and if needed, up to something like ISO 320. The limits I'm referring to show up when lifting shadows in pictures. I also claimed above in this thread that from ISO 800, the cameras become almost equal in terms of noise performance.

That's a ridiculous claim. At base ISO you would need your normal shot and a shot at +2.5ev to match a D8x0's DR. By ISO 400 that drops to a 1ev difference, and drops a little bit more as ISOs progress. So you would need 2 exposures 1ev apart. Though, in practice, a 1ev DR difference is trivial and can basically be eliminated with NR.
True, mean stacking can be avoided if you want to blend exposures instead. My mistake that I didn't mention this option. However, again, to achieve this (the same shadow noise level with the 5DsR as with the D810, normalized to its resolution) for a night shot in the dark, you need to get a 4 stops brighter exposure to use shadow areas from. If your base exposure is 6 minutes (and another 6 minutes to get rid of hot pixels and systematic noise), +4 EV from that is 192 minutes of having the camera busy for the second exposure (96 minutes of that is shooting the subject). That is no win compared to mean stacking 16 exposures - it takes the same amount of time. Not to mention the fact that the longer the exposure, the more noise you get (dark current...). So doing exposure blending for then cleaning up the shadows is only beneficial when you have much shorter exposures; and the benefit dwells in not having to mean-stack that many images in postprocessing. It might also be worth considering that doing exposure blending like that is a more laborious process in Photoshop (although not by much). Mean stacking is a brute force method compared to it, but you get much cleaner picture across the entire tonal range.

Since Canons get closer to Sony sensors the higher ISO we use; a shortcut to that long shooting procedure for later mean-stacking or exposure blending would be using a higher ISO instead, and mean-stacking and/or exposure-blending shots with that. This is due to that mean-stacking four ISO 800 shots from the 5DsR results in a picture cleaner than ISO 200 shot on the same camera, which would be the equivalent to it when using a completely ISO-invariant sensor. Note however, that the higher ISO you use, the less dynamic range gets captured and if there is a highly dynamic scene (even at night... like a nicely lit house on a darker, shadowy meadow), there will be penalties - you might be forced to go exposure blending and/or HDR, and blend more than two exposures to get what you want.
 
Last edited: