Are two cameras going to replace the Canon EOS R5? [CR]

[entoman]

By the same arguments, AI will remove the need for telephoto lenses as well. So, in future an Android or iPhone with AI will cover all photographic needs.

Yes, as I've said before, AI will eventually remove the need for telephoto lenses, as it will be possible to crop and "gigapixel" the resolution, and also it will be possible with depth data to produce realistic subject isolation/depth of field with a fixed aperture lens. All of our other "issues" such as noise, DR, camera shake, unwanted subject blur and much more will also eventually be tamed by AI. It's already happening. The form factor of cameras will change - smartphones are pocketable but have atrocious ergonomics for photography, so changes will happen in that field.

It's all a question of how quickly each development will find its way into the consumer marketplace. Personally I'd expect ILCs, pretty much as we know them, to exist for at least another 10 years, although a shift back to smaller sensors such as M43 is possible, as it is easier to implement AI tech with smaller sensors, and the public will favour smaller lighter lenses. In the meantime, I think sensor resolution increases will become far less frequent, and that AI will be used to increase the resolution of lower MP sensors. At first, this will be need to be done in post (possibly an integral part of the download process), but eventually it will all be done in-camera. Assuming (or hoping) that I live another 10 years, it will happen in my lifetime.

Also relevant is the direction, and future of photography itself. Mass market cameras have only existed for 200 years, roll film cameras have only existed for 150 years, and the first fully digital camera the Fujix DS-1P only arrived in 1988. Things have moved very fast since then, video has taken over as the most popular medium among consumers, and further huge changes will happen relatively quickly. So none of us can really foresee more than a few years ahead, and we're kidding ourselves if we think we can.

 

[Ph0t0]

For just how much longer will it actually be necessary or desirable for sensors to have high resolution. Within a very short space of time, AI tech could make them redundant, as it will be possible to artificially generate high resolutions and incredible sharpness from lower resolution sensors. Take a look at Gigapixel etc. The tech doesn't even need to be in-camera - it could be in the manufacturer's proprietary desktop apps, and could be batch processed.

Well, I guess if you are only interested in creating images with perception of detail, then a lower MP sensor and some AI will suffice.

If you want images that display actual details (not the details that AI thinks should have been there) then you will allways need good glass and sensors with higher MP count.

I mean... If you have a photo of a photo of a car in a distance and the number plate is just a blur of a couple of grey pixels, and you want to know the real number, then you need a telephoto or a better sensor.
If you don't care about the real number... Then AI can allways just create a random number and paint it on the plate and it will look great.

 

[jam05]

Then the lower resolution should have a stacked sensor. Not purchasing anymore high end cameras without a stacked sensor. $4k just for firmware is a no go for me. My R5 and R5C will last me for awhile. Will wait for the tilt shift lenses with AF

 

[David - Sydney]

How about a cheaper R5 with like 32MP for those who cant afford the R5II and RS or whatever its called. Make it a photo only camera for many of us who do not want video? Price it around $3099. I know. Keep dreaming…

Buy a R5 at the moment (in the US) with USD500 price reduction and bonus grip... sell the grip and voila! you are close to your target price
Don't forget not to change to video modes and it is a stills-only camera.
Of course, if you really don't want any video then the EVF/rear screen (besides the Q menu) is not for you as that is video off the sensor.

 

[David - Sydney]

One model that we have not mentioned is a replacement for the Ra. Simple enough to remove the low pass filter and have 2 models.
It has been 4 years since the Ra was released. Clearly we still have the R though.

Canon releasing the Ra after the 60Da would give some credibility that there was sufficient volume to warrant the cost of a second model. The only other competitor was the Nikon D810a

 

[PBguy]

A higher resolution body makes sense although I don't know if that means a completely new model or a high resolution R5. For me, 45 megapixels is more than enough for just about everything I shoot. I'm hoping to upgrade to an R5 soon from the R.

 

[David - Sydney]

For me, 45mp is a sweet spot for both high mp for cropping and for high bit rate video with no horizontal crop for 8k and simple 4KHQ (divide by 2 each direction). A slightly higher sensor resolution for a spec sheet won't offset the negatives (speed/file size/cost of new sensor etc)

I get that there is a niche for very high resolution but it would need to be sufficiently higher than the R5 to warrant the jump

If the R5ii gets the newer processor from the R6ii with greater bandwidth ie faster frame rate and full 14bit images at 20fps/45mp, improves the heat management plus most of the other incremental improvements people have asked for then the R5ii will be the 5Div after the 5Diii.... incrementally better all round but no big bang features.

 

[David - Sydney]

If true, I suspect it also means those hoping for a high MP R1 are in for some disappointment.

Although I am not the target market for a R1, you have to ask yourself how what features would a R1 need to have to warrant the flagship status?

The A1 and Z9 are high mp and frame rate bodies that justify their elevated prices. Currently we have either R3 for speed and R5 for mp. The logical conclusion for me would need to have both speed and mp with some methodology eg binning to have speed and lower mp.

 

[David - Sydney]

But the R5 can *already* shoot 45MP at 20fps, and I'd already taken it as granted that after 3-4 years Canon would be *capable* of producing a successor with 60MP and 30fps, at a price not much higher than the upcoming Z8.

It is often lost in the spec sheets that eshutter 20fps is at 12 bit and not 14 bit for mechanical (13bit for ECFS) whereas the R3 is 30fps at 14bit. The banding issue also means that I can't use it for indoor sports.

I'm guessing that an upgraded Digic X processor could handle 20fps 14 bit files. It could be the A-D convertor on the R5 sensor that is the limitation but that isn't the case for the R3 which has approximately the same total bit rate.

 

[unfocused]

Yes, as I've said before, AI will eventually remove the need for telephoto lenses, as it will be possible to crop and "gigapixel" the resolution, and also it will be possible with depth data to produce realistic subject isolation/depth of field with a fixed aperture lens. All of our other "issues" such as noise, DR, camera shake, unwanted subject blur and much more will also eventually be tamed by AI. It's already happening…

Just because you can do something doesn’t mean you should. This becomes a matter of personal ethics and responsibility. The accepted standard for nature, wildlife, sports, journalism, etc., is not to manipulate an image beyond what was possible using traditional darkroom techniques.

It’s a matter of personal choice, but for me I have no interest in taking and displaying pictures that don’t reflect my own abilities. I’m in this to push my own limits. I want to have the personal satisfaction of being able to say I was there and this is what I saw.

I may be naïve but I think that will always be the appeal of photography for many people.

 

[dirtyvu]

It is often lost in the spec sheets that eshutter 20fps is at 12 bit and not 14 bit for mechanical (13bit for ECFS) whereas the R3 is 30fps at 14bit. The banding issue also means that I can't use it for indoor sports.

I'm guessing that an upgraded Digic X processor could handle 20fps 14 bit files. It could be the A-D convertor on the R5 sensor that is the limitation but that isn't the case for the R3 which has approximately the same total bit rate.

if that's the case, then maybe it is a bandwidth limit since the resolution of an R3 pic is half of the R5. 30 24mp pics at 14-bit versus 20 45mp pics at 12-bit. someone do the math and see if they're the same...

 

[David - Sydney]

Just because you can do something doesn’t mean you should. This becomes a matter of personal ethics and responsibility. The accepted standard for nature, wildlife, sports, journalism, etc., is not to manipulate an image beyond what was possible using traditional darkroom techniques.

It’s a matter of personal choice, but for me I have no interest in taking and displaying pictures that don’t reflect my own abilities. I’m in this to push my own limits. I want to have the personal satisfaction of being able to say I was there and this is what I saw.

I may be naïve but I think that will always be the appeal of photography for many people.

Photography has had many technological changes since Daguerreotypes and from film to digital etc. Yet a small number of people still take 8x10 large format, dark room print and then scan them to their favourite social media platform. Part of it is the hands-on process and nostalgia.

That said, the majority of photographers (including the billions of phone camera users) now use "portrait mode"/ filters / post processing of their images.
AI creates new images but machine learning is more useful for everyone from AF tracking to removing unwanted objects, noise reduction, upscaling etc. For me, post processing is as essential as the technical image capture in the creative process.
Of course, we would love to have SooC perfection but we know that it is not truly possible without some conversion and the art + the science is the combination for me.

Competition rules have had to be changed over the last few years with use of stock images, compositing, HDR, significant post processing and AI generated images causing organisers headaches. Their response is to create new categories and mandating that raw images must be provided (or film negatives etc) before awarding prizes.

I predict that AI will be used to enable telephoto images from phones that can't physically get around the optical laws. When that would happen is an excellent question. It would be a good justification for a new phone. Given the exponential change in AI implementation, I would expect it sooner rather than later.
Less likely for APS-C/FF users of course.

 

[EOS 4 Life]

How about a cheaper R5 with like 32MP

R6 III.
The R6 is already a cheaper R5.

 

[EOS 4 Life]

I'm guessing that an upgraded Digic X processor could handle 20fps 14 bit files.

The 1D III and R3 already can with the current Digic X so that does not seem to be the limitation.

 

[EOS 4 Life]

I think we will get an R3II first if the planned R1 is outspecced by Sony/Nikon. They need to leave enough space for a R1 or R3II sticker.

Canon could also call it the R2.

 

[EOS 4 Life]

Definitely. I think almost all future cameras will be stills/video hybrids

Tell that to Hasselblad.

 

[EOS 4 Life]

5DII and 5DIII had very few extra pixels, so Canon could also do a 50 MPIX R5II and an 80 MPIX R5S.

50 MP would be a quite pointless upgrade from 45 MP but I guess some people felt that about going from 20 MP in the R6 to 24 MP in the R6 II.

 

[David - Sydney]

So I don't think 90MP at 20 fps is really possible, or the R5ii being able to shoot raw faster than 20 fps.
45 at 20 is pretty much the max a CF express type B card can do. 45 MBish at a transfer rate of 900 MB/s (more so a real world multi file write scenario) is 20 files per second.
60MP files ~60MB is maxed at 15 files per second
90MP files ~90MB is maxed at 10 files per second.

It makes sense why the A7Rv is capped at 10 fps on its CF Express type A cards with their slower write speeds.
IF Canon makes a R5s as a high resolution body, they would either have to use an internal or external SSD in order to shoot fast, and again that processing would have to be insanely fast to process the data on that sensor. I don't think realistically its something that could even be used for sports or anything fast motion and fast capture as it might have banding or other issues with that large of a pixel density.

Realistically if you are product or landscape or some sort of studio photographer with a need for a R5s to replace a 5Ds, I don't think anything more than 5-10 fps is really required.
Thats my 2 cents. Maybe I am wrong, who knows.

Let me know if my back-of-the-envelope calculations are off!

R3 on electronic shutter = 24.1mp * 30fps * 14bit = 10.1 Gb/s
R5 on mechanical shutter = 45mp * 12fps *14 bit = 7.6 Gb/s (probably limited by the shutter assembly)
R5 on electronic shutter 45mp * 20fps *12 bit = 10.8 Gb/s
R6ii = 24.3mp * 40fps * 12bit = 11.7 Gb/s => In theory the R6ii could have had 30fps @14bit but it would not have had the same ultimate speed impact on the spec sheet.

=> So the bandwidth appears to be a maximum of ~11gb/second with the Digic X in those bodies
Ultimately, Canon should given the user the choice... higher speed with lower bit depth or slower speed higher bit depth. At least the R3 can vary the eshutter rate and maintain 14 bit.

To your point and if existing bandwidth is the bottleneck:
- 60mp * 12fps (mechanical or eShutter) @ 12 bit or 60mp * 10fps * 14 bit would be possible (~8.4 Gb/s)
- 90mp * 8fps * 14 bit = ~10Gb/s
=> I reckon that either of these scenarios would be acceptable in a high mp body where outdoor light or indoor controlled lighting would not cause banding.

As Bryan from TDP says "Is dropping to 12-bit a problem? While there is a phsychological difference, the image quality difference will seldom be noticed in ideally exposed images. Reduced quality will show primarily in smooth gradients such as the sky when contrast is adjusted, and as illustrated earlier in this review, noise levels increase in brightened images, especially in the shadows. Most of the images in this review were electronic shutter captured, and I haven't noticed the 12-bit difference. "

Also, the CFe write speed is only relevant when clearing the buffer. The buffer depth will manage the max frame rate until full. Quality CFe Type B cards can handle sustained ~1.4Gb/s write speeds ie about 10x slower than the internal bandwidth.
If you are interested in the buffer depth for the R5/R6, then Bryan's testing is at
https://www.the-digital-picture.com...-R6-Buffer-Capacity-During-High-Speed-Capture

 

[EOS 4 Life]

Canon can bump the low-end to 32MP and cap the high-end at 50MP

Sony is already ay 61 MP.
Canon won't stay behind Sony in the megapixel race.

 

[EOS 4 Life]

45 at 20 is pretty much the max a CF express type B card can do

Sony is getting 50 MP at 30 FPS with the slower CF express type B*.
Canon RAW files are a lot smaller than the Sony ones.

*I do realize that there are a lot of caveats and it is more like 20 FPS in the real world.