Future direction of high-resolution bodies

[shadowblade]

With the recent release of the D850 and A7r3, it has now been demonstrated that resolution and frame rate can come in the same package, and that, with competent AF systems, they make for very credible action cameras as well as having the resolution to do a good job of landscapes, reproduction, studio/commercial work and other situations where resolution is required (in fact, it could be argued that the D850 and A7r3 are better for field sports and wildlife than the D5 and A9, given the greater cropability of the images). They likely represent the first of a new breed of 'do-it-all' cameras, made possible by the increased data bandwidth and processor power that wasn't available in small packages just a few years ago.

But where does this leave the dedicated high-resolution camera - the future equivalents of the 5D2, D800 and A7r/A7r2? How would a manufacturer distinguish a high-resolution camera from a general-purpose camera, to gear each camera for a different task,while enticing users to buy both? It is clearly possible to push resolution much higher, at the expense of speed. But resolution may not be enough of a distinguishing feature any more. Nor will intentionally crippling features work - there is enough competition to ensure that, if one of the big three intentionally cripple the AF on the high-resolution body, or the output quality on the general-purpose body (also reducing the two bodies' ability to cover for each other), for instance, users will simply look to one of the other two manufacturers instead of just buying both bodies.

The D850/A7r3 are in the 40-45MP range, shooting at 8-10fps. A high-resolution body would likely be in the 60-80MP range, shooting at around 5fps, with the same AF capabilities and other features (dual cards, battery life, etc.). But would that be enough of a distinguishing feature? After all, 5fps is enough for a lot of action photography (AF speed and accuracy being far more important than frame rate), while 40-50MP is enough resolution for most purposes.

In addition, 8k video means that 39MP/25fps cameras are arriving in the near future. This will likely push 'general purpose' cameras to 60MP/10fps, further into the zone of diminishing returns.

I doubt the high-resolution/low speed/low ISO camera is going to disappear, unlike the action stills camera (which will become functionally indistinguishable from the 8k video camera, the two separate classes merging into one type). But it will likely have to evolve in other ways than just sacrificing speed for resolution to retain a meaningful distinction from the general-purpose bodies.

I doubt it will be a move towards medium format. There is simply too much to gain from sharing a common set of lenses with the general-purpose and action bodies - lenses which are increasingly sharp anyway - and too much to sacrifice in terms of weight and portability, particularly for multi-subject shooters who also need to carry a second, more general-purpose system.

Rather, more specialised sensors, geared towards maximising image quality at low ISOs, may be the way forward for these cameras. An increase in full well capacity would allow for greater DR, with an increase to 16-bit or greater output to handle the increased data. Multilayer sensors would allow for true RGB values for each pixel, eliminating colour artifacts, although possibly at the expense of high-ISO performance. Stronger colour filters (or a multilayer sensor with more than three layers - there's no reason you couldn't build a tetrachromat sensor, for example) would allow for greater colour accuracy and captured gamut, again likely at the expense of high-ISO performance. These features would be more difficult to build into a general-purpose body, which needs to perform at all ISOs, but could be valuable features in a body geared towards shooting at ISO 1600 and below.

In three years time, when we have a 60MP/10fps general-purpose camera, a 100MP/5fps slow camera may not seem very interesting. But make it a 100MP/5fps (or even 80MP/5fps) camera with 17 stops of DR, 16- or 18-bit output (with 16 bits of actual data, not just noise in the lower bits) and true RGB values for every pixel and it suddenly looks a lot more interesting next to the Bayer-sensor general-purpose body. There would be a meaningful choice between the two - some would buy one or the other, depending on need, while many would buy both, gaining ultimate IQ at low ISO while retaining the action capabilities of the general-purpose body, all with the same set of lenses and accessories. And the two bodies could still cover for each other in a pinch, with neither body being completely incompetent at the other's role.

Any thoughts?

[BernardLanguillier]

There are 2 main axis of progress:
1. Achievable quality of images when things go well. This is basically the result of lenses + sensor
2. Success rate or rate of failure

1. About #1: IMHO, the most significant progress that remains to be done is in the look/quality of lenses.

None of what you write on the sensor side will have any significant impact on images. Reaolution is now mostly irrelevant for 99% of photographers, DR is sufficient,...

The long term winner among the major brands (Canon, Sony and Nikon) will be decided by the quality of the rendering of their mirrorless lenses. What is the equipment that helps create beautiful images. Sony has alreasy played their cards and they are good but not great. It remains to be seen what Canon and Nikon can do. Nikon has communicated a lot about the look of their DSLR lenses, Canon hasn’t said much. We’ll what comes out in the end.

If you add minor players, it seems obvious that Olympus and Fuji have also a strong focus on the look of lenses and this may help them gain market share moving forward.

2. About #2: the main factor here is AF in terms of accuracy, speed, tracking ability, ease of control... This is an area where things can still be improved IMHO.

Cheers,
Bernard

[shadowblade]

There are 2 main axis of progress:
1. Achievable quality of images when things go well. This is basically the result of lenses + sensor
2. Success rate or rate of failure

1. About #1: IMHO, the most significant progress that remains to be done is in the look/quality of lenses.

None of what you write on the sensor side will have any significant impact on images. Reaolution is now mostly irrelevant for 99% of photographers, DR is sufficient,...

Yet there is a clear difference between pixel shift and non-pixel-shift images from the A7r3, especially in any situation where there are fine, repeating details (fabrics, hair, architectural textures, foliage, etc.). These often produce artifacts in non-pixel-shift images, but not in the pixel shift images. Furthermore, the pixel shift images demonstrate noticeably greater detail than the non-pixel-shift images - the resolution is still 42MP, but it's a much better 42MP than the non-pixel-shift images.

If there was no benefit to it, no-one would bother with it, and Sony wouldn't have included it as a feature.

Gathering true RGB values for each pixel using a layered sensor, instead of relying on interpolation, would provide the same benefits, without the motion artifact problems associated with pixel shifting. It would come at a cost in ISO performance, though, making it less suitable for a general-purpose body and likely the purview of a dedicated low-ISO body. And it would likely provide much more benefit to image quality than simply increasing resolution from 45MP to 60MP in the current generation, or 60MP to 80MP in the next (although there is little reason they wouldn't do both).

The long term winner among the major brands (Canon, Sony and Nikon) will be decided by the quality of the rendering of their mirrorless lenses. What is the equipment that helps create beautiful images. Sony has alreasy played their cards and they are good but not great. It remains to be seen what Canon and Nikon can do. Nikon has communicated a lot about the look of their DSLR lenses, Canon hasn’t said much. We’ll what comes out in the end.

This is completely missing the point of the thread. The point is regarding the distinction between the 'general-purpose' bodies (D850, A7r3) which shoot everything and the high-resolution/high-IQ/low-ISO specialist bodies (previously 5Ds, D800, A7r, A7r2, D3x, but none in this generation yet). You can put the same lens onto either body, so the lens lineup is not a point of distinction between the two classes of camera.

2. About #2: the main factor here is AF in terms of accuracy, speed, tracking ability, ease of control... This is an area where things can still be improved IMHO.

This is also a non-factor for product differentiation between general-purpose and high-resolution/low-ISO-focused bodies. What goes in one can easily go in the other, without a performance tradeoff - they'd be dumb to cripple one or the other for the sole purpose of market segmentation.

It's only a differentiating feature if it's something that can easily go into one type of body, but not the other (at least not without a significant tradeoff in core functions). AF and lens lineup work equally for both types of bodies. To make specialist high IQ/low ISO/low speed bodies relevant in the face of general purpose bodies already shooting 40-45MP/8-10fps in this generation, and likely 60MP/10fps in the next generation (necessary for the general-purpose bodies to remain distinct from the 8k video cameras), they're going to have to provide something that the general-purpose bodies simply can't provide, not just a few more megapixels. Fortunately, there are a few things they can include to make them distinct and relevant.

[Two23]

But where does this leave the dedicated high-resolution camera - the future equivalents of the 5D2, D800 and A7r/A7r2? How would a manufacturer distinguish a high-resolution camera from a general-purpose camera, to gear each camera for a different task,while enticing users to buy both?

If it's Nikon, they will simply make one camera $10,000 and the other one $7,000, LOL.


Kent in SD

[BAB]

This is completely missing the point of the thread. The point is regarding the distinction between the 'general-purpose' bodies (D850, A7r3) which shoot everything and the high-resolution/high-IQ/low-ISO specialist bodies (previously 5Ds, D800, A7r, A7r2, D3x, but none in this generation yet). You can put the same lens onto either body, so the lens lineup is not a point of distinction between the two classes of camera.[/font][/size]

[/size]
Not to sure general purpose bodies shoot everything, they might shoot it but the ? Is how well?[/font][/size]
I kinda have similar thoughts to Bernard, plus the image acquiring products today for top top IQ are less important for advertising since low res homemade images and video is more accepted actually preferred by by internet shoppers.[/font][/size]

[BJL]

Is it too much to hope that, with fast enough sensor read-out and in-camera processing, the distinction between high resolution and high frame rate bodies in a given format could disappear? It did not exist with film did it?

But even putting aside cynicism about camera companies crippling high-end products in order to sell two bodies to some photographers when one could do it all, there might still be legitimate differentiations like:
- multi-shot mode, adding expense and fragility that a sports/action/PJ photographer does not want.
- integrated vertical grip, greater battery capacity and extreme ruggedization, whose bulk is not wanted by many less action oriented photographers.

[shadowblade]

This is completely missing the point of the thread. The point is regarding the distinction between the 'general-purpose' bodies (D850, A7r3) which shoot everything and the high-resolution/high-IQ/low-ISO specialist bodies (previously 5Ds, D800, A7r, A7r2, D3x, but none in this generation yet). You can put the same lens onto either body, so the lens lineup is not a point of distinction between the two classes of camera.[/font][/size]

[/size]
Not to sure general purpose bodies shoot everything, they might shoot it but the ? Is how well?[/font][/size]
I kinda have similar thoughts to Bernard, plus the image acquiring products today for top top IQ are less important for advertising since low res homemade images and video is more accepted actually preferred by by internet shoppers.[/font][/size]

Yet, medium-format bodies exist. Pixel shift exists. People shoot panoramas and blend exposures. There is clearly a demand for better image quality (which includes, but isn't limited to, higher resolution - there are many ways to increase the output data other than just increasing the size of the two-dimensional matrix it's stored in).

Obviously, this demand isn't as great as for more general-purpose bodies, just as the demand for 15/20fps bodies isn't as great as the demand for general-purpose bodies. The D5 and A9 are never going to sell as well as the D850 and A7r3. This is why the 5D3 and D750 did so well (and I'd expect the D850 and A7r3 to do the same, even more so than the D810 and A7r2). But they are still in demand, as were the D810 and A7r2 (when they were the slow, high-resolution bodies of their time).

Despite general-purpose cameras shooting at 10fps and being able to hold just as capable an AF system, the action camera isn't going away - the dedicated action stills camera is likely to disappear as a separate entity, but, rather than dying out, is really just merging with the camcorder as a single device able to fulfil both roles (as a hand-portable camera capable of at least 39MP and 25fps - already, we have 6k video cameras shooting 19MP/25fps, which isn't too different from the A9's 24MP/20fps). The 'middle ground' covered by general-purpose cameras is getting wider, becoming capable of filling more roles that would once have required a specialist action body.

But what about the high-resolution camera (probably now better described as a 'high detail' camera, since resolution is not the only way, and probably not even the main way, it would set itself apart from the general-purpose cameras)? Given the different ways people go to now to increase captured detail, even with 42/45/50MP cameras, they certainly seem unlikely to disappear either.

With general-purpose bodies taking up more of the middle ground, I can see these bodies evolving to become more specialised, geared towards maximal image quality rather than compromising aspects of their design for general use. They would likely go after the medium-format market, encroaching further and further on medium-format IQ (with a 100MP sensor, a sufficiently-sharp lens, and microlenses, filters and electronics geared towards maximal image quality, there's no physical reason a future 35mm camera can't match an IQ3 100MP digital back in terms of output quality), while retaining the 35mm advantages of lens selection, portability, price and commonality of AF systems with their more action-oriented counterparts. They would be able to do this because the general-purpose bodies already exist to cover situations where such sensors wouldn't be suitable (higher ISOs or frame rates). Essentially, they'd have one foot in each area, specialising in low-ISO image quality, while having the AF and other systems needed to back up a general-purpose system (and take the occasional 100MP action shot) when necessary, albeit with a lower frame rate.

Lenses aren't a distinguishing factor between fast, medium and slow cameras - you can put a super-sharp lens onto any of them, and they all benefit. They play a very important role in ensuring that camera systems continue to improve, but don't, by themselves, distinguish between various classes of camera body.

Medium format isn't sacrosanct. Remember large-format film cameras? We don't see too many of those any more, since medium-format digital now outresolves 8x10" colour film. Nor do we see too much demand for large-format digital, likely for a variety of reasons - they'd probably cost hundreds of thousands of dollars (more for commercial reasons than due to manufacturing costs), would be heavy and cumbersome to carry around, and, likely most of all, once you can match 8x10 film, you probably don't need all that much more detail (when you do, you're probably shooting something you can stitch, or are designing a billion-dollar spy satellite). We also don't see too many medium-format film cameras around, apart from the panoramic formats. These were once the staple press cameras, but full-frame digital now outresolves medium-format colour film and is far more portable. (Current resolution-focused full-frame cameras definitely beat 6x9-format colour film and probably come close to 4x5 colour film in terms of detail resolved, and obviously do much better in terms of ISO and DR, when using lenses that can keep up). With sufficient development of 35mm-format, detail-oriented cameras and sufficiently-sharp lenses (easier to make a sharp 35mm-format lens than an equally-sharp medium-format lens), there's no reason that medium-format digital won't itself largely become a relic, used by few other than well-heeled landscape photographers wanting something even bigger and better.

As for web output, the definition of 'web-sized' is changing fast. 4k monitors are now common. You can easily get 6k monitors. 8k monitors have now started to come out. Even the Samsung Galaxy S8 uses 2960x1440 resolution. For ideal screen output, the input needs to either come from a source with significantly higher resolution, or with true RGB values for each pixel (e.g. 6k input for 4k output) - and that's not even counting the cropping and panning of images commonly used for electronic output. The drone cameras and GoPros of the future won't be the present-day, low-resolution cameras using cheap, fixed plastic lenses. They will be 8k-capable video cameras, likely using deformable lenses with a wide zoom range. And, with production and use of imaging sensors exploding globally (they're used in far more things than just cameras), they won't be any more expensive than current drones and wearable action cameras. Yes, they'll produce 'web output'. But the 'web output' they need to produce will be very different from current-day, 700x400px Jpegs.

[shadowblade]

Is it too much to hope that, with fast enough sensor read-out and in-camera processing, the distinction between high resolution and high frame rate bodies in a given format could disappear? It did not exist with film did it?

That's because film never gave the opportunity to make that tradeoff. You couldn't use a lower-resolution film just to shoot faster.

But, now, data bandwidth is a resource that can be allocated, just like battery life. You can use it to shoot faster, or to shoot higher resolution/bit depth/colour sampling. I don't see the distinction between detail and frame rate disappearing until cameras can shoot 150MP video in 24-bit colour, with true RGB values for each pixel, when the whole question becomes moot anyway. At present, there is still demand for more detail/more resolution, so the distinction remains.

Right now, you can shoot 24MP at 20fps, with 14-bit output and Bayer interpolation. With the same amount of data bandwidth, you could also shoot a 72MP, three-layer (RGB values for every pixel) sensor with 16-bit output at 2fps (1.94fps, to be exact). There's still a fair way to go until we get to a stage where there's no value in pushing IQ higher and higher and we can just push frame rates.

But even putting aside cynicism about camera companies crippling high-end products in order to sell two bodies to some photographers when one could do it all, there might still be legitimate differentiations like:
- multi-shot mode, adding expense and fragility that a sports/action/PJ photographer does not want.

Multishot mode doesn't add any expense and fragility if you already have in-body stabilisation. It's purely a software thing.

Given that action bodies and video cameras are likely to merge in the near future, and video cameras (and video shooting in general) practically demands stabilisation (even for short focal lengths) this doesn't really seem like a point of differentiation.

- integrated vertical grip, greater battery capacity and extreme ruggedization, whose bulk is not wanted by many less action oriented photographers.

Future grips are likely to be optional anyway, since it makes bodies more flexible and usable by more people. Think A9 rather than D5. Remember, the action-oriented film cameras didn't have integrated grips either - they only became necessary with early-generation digital cameras due to battery requirements. But battery life is improving.

Many people struggle to use the 1Dx or D5, simply because they are too big to easily handhold for many people. I won't use one without a monopod. As for 'lens balance' issues, that's really a non-issue when you're holding the lens-camera setup by the lens anyway, only using your right hand to manipulate the camera and press buttons rather than to support the weight of the camera. If it's really such a big issue, you can always gaffer tape a brick to it... there's no reason the weight needs to come from a non-removable grip.

As for the vertical shooting buttons, there's also no reason they can't be integrated into a 'mini-grip', which adds the buttons with minimal bulk and weight (but contains no batteries), or even integrated onto the base of the camera body itself.

[BJL]

Shadowblade,you might very well be right on all points, but since none of us know for sure, I will speculate:

- Maybe future 36x24mm format cameras will reach the resolution limits set by lenses and fundamental optics whole operating at video frame rates. Or a single body will be able switch between extreme resolution and higher frame rates by on-chip binning or such.  (Aside: It is strange to say that film lacks the _opportunity_ ty to make that trade-off, as if needing two bodies for two different priorities is an advantage.  Instead film has the advantage of never being forced to make that frame-rate vs resolution trade-off, because its image recording is a totally "parallel" process, not hampered by the time required to sequentially read-out and process the photosite data of an electronic sensor.)

- I have also long though that the high-end "action cameras" would be better off with modular, removable vertical grip/battery packs, yet the camera makes persist in disagreeing: they keep making those big, heavy "square" bodies. So maybe there are advantages to that integration (maybe better weather sealing or greater robustness or less bulk that body + add-on grip).

- I agree that once a camera has sensor-shift stabilization, it can easily have multi-shot, but so far the super-rugged flag-ship DSLRs do not have movable sensors, and maybe a niche of extremely rugged bodies will always favor lens-based stabilization and "fixed" sensors.

[ErikKaffehr]

Hi,

Eric Fossum, the inventor of the CMOS sensor, mentioned over at DPReview that Forza Silicon has developed a 133 MP sensor for NHK (Japanese Television) it is intended for 8K, but sampling each 8K pixel for RGBG. So all colours are sampled at 8K while the device is smaller than a three sensor with beam splitters solution.

https://www.forzasilicon.com/2015/03/nhk-and-forza-silicon-present-133-mpixel-60-fps-cmos-image-sensor-at-global-ic-design-conference-isscc/

Eric Fossum predicts that we will shoot 133 MP in a few years, that would be 2.3 micron pixels, about what is needed to make the Otus justice, I would think.

Best regards
Erik

[shadowblade]

Shadowblade,you might very well be right on all points, but since none of us know for sure, I will speculate:

- Maybe future 36x24mm format cameras will reach the resolution limits set by lenses and fundamental optics whole operating at video frame rates. Or a single body will be able switch between extreme resolution and higher frame rates by on-chip binning or such.

That won't be for a while. You're talking hundreds of megapixels, and that's not even talking about superlenses, which use diffractive optics to resolve detail beyond the diffraction limit. Then there are still other ways in which image quality can be improved, without increasing the resolution - increased colour accuracy, true RGB values at each photosite, increased bit depth (meaningful bit depth due to sensors able to distinguish it, not just meaningless extra bits below the noise threshold), all of which require data bandwidth. Some of these things adversely affect the high-ISO side of things, at least for now; hence, they would be useful in a dedicated low-ISO, high-image quality body (e.g. Foveon-type sensors) but of limited use in general-purpose bodies.

But there is no need to reach the physical limits - there are also practical limits, where there isn't much point in increasing one particular aspect of performance further (at least for the immediate future) because increased performance in that aspect just wouldn't be particularly useful, outside of certain very specialised applications. We aren't near most of those limits yet, but are getting closer. For instance, if you have a 150MP sensor recording true RGB values for each pixel, there's probably not much benefit in increasing the resolution further. As you approach these 'soft' limits, you're probably better off improving other aspects of performance (aspects relevant to the envisioned use of the camera, obviously) rather than trying to push that one particular aspect further.

(Aside: It is strange to say that film lacks the _opportunity_ ty to make that trade-off, as if needing two bodies for two different priorities is an advantage.  Instead film has the advantage of never being forced to make that frame-rate vs resolution trade-off, because its image recording is a totally "parallel" process, not hampered by the time required to sequentially read-out and process the photosite data of an electronic sensor.)

Because film can't make that tradeoff. You can make a film camera shoot as fast as mechanically possible. But, want better image quality? Bad luck - you're just going to have to make it bigger, with bigger lenses, heavier parts, etc. You can't get better image quality in the same size of film without changing the film itself, and that's limited by chemistry.

- I have also long though that the high-end "action cameras" would be better off with modular, removable vertical grip/battery packs, yet the camera makes persist in disagreeing: they keep making those big, heavy "square" bodies. So maybe there are advantages to that integration (maybe better weather sealing or greater robustness or less bulk that body + add-on grip).

Battery life and supplied voltage, initially. And, later on, when these things were improved (if you can't deal with changing batteries every 2000 or more shots, you need better planning, not more battery life), public perception and macho attitudes by some photographers about 'real cameras' probably had a role to play. But it's clearly not mandatory - look at the A9, and look at the Canon 1V film camera.

- I agree that once a camera has sensor-shift stabilization, it can easily have multi-shot, but so far the super-rugged flag-ship DSLRs do not have movable sensors, and maybe a niche of extremely rugged bodies will always favor lens-based stabilization and "fixed" sensors.

That's because only Canon and Nikon build such cameras, and they built their systems around lens-based stabilisation.

There are lots of equally-rugged video cameras out there using sensor-based stabilisation. Sensor movement systems are already very rugged, since they don't have to move the sensor very far, nor support a particularly heavy weight. Mirror and shutter systems are far more vulnerable in that regard, being subject to far greater acceleration and far greater forces, and are far more likely points of failure.

Moreover, lens-based stabilisation is also a weak point, and far more likely to fail than sensor-based systems. A set of floating glass elements which need to be kept centred is far more susceptible to damage than a moving sensor. It is better for correcting movement in certain lenses, though, particularly long telephotos, which is why even those using movable sensors do it with some lenses.

In any case, what are you planning to do with the camera? Use it to hammer in nails? I've subjected all sorts of cameras to abuse and have never had one fail on me due to physical damage that wouldn't have also destroyed any other camera (melted by lava). If you're constantly dropping your cameras onto concrete, you have bigger problems than the camera not being tough enough. And the lens is unlikely to survive the same impact, either.

[shadowblade]

Hi,

Eric Fossum, the inventor of the CMOS sensor, mentioned over at DPReview that Forza Silicon has developed a 133 MP sensor for NHK (Japanese Television) it is intended for 8K, but sampling each 8K pixel for RGBG. So all colours are sampled at 8K while the device is smaller than a three sensor with beam splitters solution.

It's also much more robust than a beam splitter/three sensor solution.

With a three-sensor system, if even one of them jolts or misaligns slightly, the images no longer match.

But, using a Bayer array with standard microlenses comes with its own problems. Essentially, you're throwing away two-thirds of the light hitting the sensor - red pixels throw away the green and blue parts of the spectrum, green pixels throw away the red and blue, while blue pixels throw away the red and green. The three-sensor system avoids this, but introduces other problems.

Aside from multilayer sensor solutions, this can also be overcome using a different microlens array in front of a normal sensor. A normal microlens focuses incident light onto the photosite directly behind it; a colour filter then filters out the wavelengths that the photosite isn't supposed to record (the photosite itself being essentially colour-blind - it's just seeing the world through a coloured filter). But a different microlens array would use both a lens to focus light, but also a microprism to split the incident light by wavelength, directing each part of the visible spectrum onto different photosites to represent red, green and blue. Each microlens-microprism complex would, therefore, direct light to three different photosites - there would be no need for colour filters, and no light wasted by being absorbed by the filters. It is, in effect, a form of pixel binning, with three times as many photosites as there are pixels in the final image. But there's no reason you couldn't use, say, a sensor with 300 million photosites to produce a 100MP image, with true RGB values for each pixel - after all, Canon already uses dual-pixel technology, with each pixel consisting of two different photosites (and the sensor as a whole containing twice the number of photosites as the stated resolution of the sensor).

This would have the effect of increasing the ISO performance of the camera by about one-and-a-half stops (since three times as much light is being collected by the photosites during any given exposure). If the FWC of the sensor can be increased by a similar amount, that would lead to an increase in DR by the same one-and-a-half stops; otherwise, it would lead to a sensor with a base ISO of 250 or so (but with the same DR and noise characteristics at ISO 250 as the same sensor with normal microlenses would have at ISO 100).

Eric Fossum predicts that we will shoot 133 MP in a few years, that would be 2.3 micron pixels, about what is needed to make the Otus justice, I would think.

Sounds about right.

I'd expect the next-generation action bodies - due in 2020 - to shoot 8k video, which puts them at 39MP/25-30fps. Any less and they wouldn't really have any reason to exist, since there will be 8k video cameras around which can do this; it would be easy enough to tweak the settings slightly to make them shoot stills at 1/1000s, 25fps, rather than video footage, rendering any lower resolution/slower action stills camera pointless. Essentially, action stills and video cameras would merge into the same camera - they are currently on a convergent path, with stills cameras getting faster and faster while video cameras gain more and more resolution; a 6k video camera shooting 19MP/25fps really isn't too different functionally from an A9 shooting 24MP/20fps.

This would likely put the D850/A7r3 successors (whether they have mirrors or not) in the 60-70MP range, in order to maintain a meaningful distinction between them and the action bodies/video cameras. And, to maintain a gap between themselves and the general-purpose bodies, this would put the slow-shooting, IQ-focused bodies in the 100-150MP range (although, by that point, it is increasingly likely that such IQ-focused cameras will have to distinguish themselves by means other than just resolution, since it would be starting to hit the point of diminishing returns - higher bit depths and true RGB values for each pixel being the most obvious distinguishing features).

[BJL]

That won't be for a while. You're talking hundreds of megapixels . . .

As Erik K points out, over 100 MP at video frame rates is already implemented, so for the needed bandwidth, "a while" might only be one or two generations of new cameras. I do agree with this:

But there is no need to reach the physical limits - there are also practical limits . . . For instance, if you have a 150MP sensor recording true RGB values for each pixel, there's probably not much benefit in increasing the resolution further. As you approach these 'soft' limits, you're probably better off improving other aspects of performance . . .

Because film can't make that tradeoff.

No, film cameras do not need to make that trade-off: the same film camera can offer both the highest frame rate and the highest resolution etc. that the format itself allows: how is it a disadvantage relative to digital cameras that the latter — so far — have to compromise on one to max out the other?

On integrated vs modular vertical grips, I hope you are right:

But it's clearly not mandatory - look at the A9, and look at the Canon 1V film camera.

The Sony A9 might be an example of what is to come (I hope!), but it might just be that Sony does not yet put much emphasis on that sports/PJ market. The Canon EOS-1 series (and the Nikon F series) of film cameras only dropped their integrated vertical grips when that sports/PJ "machine-gunning" market was moving to digital cameras; their digital counterparts at the same time and since have continued the "big square brick"  integrated style.

And on stabilization and sensor-shift multi-shot, I again hope that you are right — I am a big fan of 5-axis IBIS, and look forward to hand-held multi-shot. I am just not totally confident, and so playing the devil's advocate. Let us see ...

[shadowblade]

As Erik K points out, over 100 MP at video frame rates is already implemented, so for the needed bandwidth, "a while" might only be one or two generations of new cameras. I do agree with this:

That's 133MP at 1 colour per pixel. It doesn't mention the bit depth.

If you're talking about maximal quality, where you're just pushing up the frame rate because image quality has already been maximised, you need 3 colours per pixel. That triples the bandwidth. You're also likely talking about somewhat greater than 14 bits of bit depth - let's say, 18 bits. A 133MP body with true RGB values in each pixel and 18 bits of bit depth would require 4 times as much data bandwidth as a 133MP normal sensor with 14 bit output shooting at the same frame rate. That's probably about 10 years away.

Moreover, there may be compromises to be made which involve more than just bandwidth. Some methods of increasing image quality - Foveon sensor, for example - can greatly increase low ISO image quality, at the expense of high-ISO sensitivity. Such a sensor would be well suited to an IQ-focused camera, but you probably wouldn't want to put it in a general-purpose camera. So you're still left with multiple bodies, even though you have the bandwidth needed to do the job with just one. There are ways around this (e.g. the microlens-microprism complex I mentioned earlier) but it just depends which one they decide to implement.

Also, no mention of what body this thing is going in - is it suitable for a still-camera-sized body and battery, or does it need a production camera with mains power?

No, film cameras do not need to make that trade-off: the same film camera can offer both the highest frame rate and the highest resolution etc. that the format itself allows: how is it a disadvantage relative to digital cameras that the latter — so far — have to compromise on one to max out the other?

The disadvantage is that you can't max out the image quality at all - you get what you're given. You can pump the 'bandwidth' as high as you like, by increasing the frame rate, but there isn't a single thing you can do to increase the image quality, barring bringing out a new type of film.

On integrated vs modular vertical grips, I hope you are right:The Sony A9 might be an example of what is to come (I hope!), but it might just be that Sony does not yet put much emphasis on that sports/PJ market. The Canon EOS-1 series (and the Nikon F series) of film cameras only dropped their integrated vertical grips when that sports/PJ "machine-gunning" market was moving to digital cameras; their digital counterparts at the same time and since have continued the "big square brick"  integrated style.

None of the Canon EOS-1 series bodies, dating from 1989-2000, had integrated grips. You could add on a power drive booster to get to 10fps, but that was strictly optional. And the 1V is perhaps still the toughest camera body around.

The Nikon F5 had an integrated grip, but the preceding F4 and the succeeding F6 didn't (using motorised grips for increased frame rate if needed). The F6 was released when digital cameras were around, but the F4 was released in 1988.

[the_luminous_french]

Color fidelity, DR for photo and Codec and Sound for video... a lot of things can be improved...

I think we are going to see two types of camera... it used to be frame rate vs resolution...

the future : Video camera that can take Photographs vs Photo camera that can record video ;-)