Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: feppe on August 30, 2009, 09:20:20 am
-
For years I've been dismayed at Canon and Nikon being stuck in the mindset of late 90s film cameras with their features. Their biggest innovations in digital have been a direct print button (Canon) and AF system with 14,725 focus points (Nikon). Sarcasm aside, of course there are new technologies, such as live view - which we all laughed at until we tried it -, and parametric auto-exposure. But by and large, dSLRs see only improvements to existing features which usually take generations to implement properly, while compact cameras from other companies pack in the truly new features.
Perhaps not surprisingly it seems to be Ricoh, Sigma, Samsung, Panasonic and Sony who innovate most, with pixel binning, expanded dynamic range by combining exposures in-camera, non-Bayer sensors, multiple aspect ratios, etc. Perhaps it is because they don't have a decades-long legacy with engineers stuck in another century, and aren't afraid to try something new.
But dSLR shooters are stuck with cameras with features from the late 90s, albeit much improved. Why aren't the features we see in compacts trickling down (up?) to dSLRs? Where are the new features? Where is automatic ETTR exposure? Where is truly smart parametric auto exposure control? Where is live view AF fast enough for handheld work? Dedicated sensor pixels to expand dynamic range (think older Fuji sensors)? Moire-less pictures without the softness inherent in high-pass filters? Automatic calibration of lenses?
While I'm afraid Sony S850's very low price point will force Canikon to accelerate the race to the bottom, perhaps they will instead race with new game-changing features, instead of a yet another generation of incremental improvements. Maybe G11 and S90 are a sign that at least Canon has finally "got it." Or the lower megapixel count might be just to protect their low-end dSLR market; their crippled movie-mode supports this notion.
I'm utterly disgusted by my Canon experience, with my soft 24-105mm f/4L, cryptic flash system, and the lack of innovation. I would sell all my lenses, the powerful but tedious 580EXII, and my bodies in a heartbeat if some other company showed innovation and dedication. Perhaps it will be Leica M9. Perhaps it will be Sony with the future A950; they are known for sticking to a long-term strategy - witness how they had pro-grade lenses long before they had a pro-grade body. Perhaps it will be Olympus with a truly astonishing follow-up to the EP-1.
-
Hi Harri:
I concour mostly with you, but I'm afraid that as long as they meet their sales goals , they won't innovate much.
Best
Eduardo
For years I've been dismayed at Canon and Nikon being stuck in the mindset of late 90s film cameras with their features. Their biggest innovations in digital have been a direct print button (Canon) and AF system with 14,725 focus points (Nikon). Sarcasm aside, of course there are new technologies, such as live view - which we all laughed at until we tried it -, and parametric auto-exposure. But by and large, dSLRs see only improvements to existing features which usually take generations to implement properly, while compact cameras from other companies pack in the truly new features.
Perhaps not surprisingly it seems to be Ricoh, Sigma, Samsung, Panasonic and Sony who innovate most, with pixel binning, expanded dynamic range by combining exposures in-camera, non-Bayer sensors, multiple aspect ratios, etc. Perhaps it is because they don't have a decades-long legacy with engineers stuck in another century, and aren't afraid to try something new.
But dSLR shooters are stuck with cameras with features from the late 90s, albeit much improved. Why aren't the features we see in compacts trickling down (up?) to dSLRs? Where are the new features? Where is automatic ETTR exposure? Where is truly smart parametric auto exposure control? Where is live view AF fast enough for handheld work? Dedicated sensor pixels to expand dynamic range (think older Fuji sensors)? Moire-less pictures without the softness inherent in high-pass filters? Automatic calibration of lenses?
While I'm afraid Sony S850's very low price point will force Canikon to accelerate the race to the bottom, perhaps they will instead race with new game-changing features, instead of a yet another generation of incremental improvements. Maybe G11 and S90 are a sign that at least Canon has finally "got it." Or the lower megapixel count might be just to protect their low-end dSLR market; their crippled movie-mode supports this notion.
I'm utterly disgusted by my Canon experience, with my soft 24-105mm f/4L, cryptic flash system, and the lack of innovation. I would sell all my lenses, the powerful but tedious 580EXII, and my bodies in a heartbeat if some other company showed innovation and dedication. Perhaps it will be Leica M9. Perhaps it will be Sony with the future A950; they are known for sticking to a long-term strategy - witness how they had pro-grade lenses long before they had a pro-grade body. Perhaps it will be Olympus with a truly astonishing follow-up to the EP-1.
-
I think we're going to start seeing more of this. The issue was that for a long time the manufacturers could get away with simply increasing the resolution and maybe a couple other basic features (better battery life, bigger LCD).
I think we've reached a point where the resolution offered today is more than adequate for most users and we're hitting the limits of what 35mm format lenses can offer at the high end of resolution (e.g. diffraction limits, but also just resolution limits for the lenses).
What this means is that manufacturers are going to have to start competing on other features, and this should mean more innovation.
-
For years I've been dismayed at Canon and Nikon being stuck in the mindset of late 90s film cameras with their features.
Maybe, but my Nikon camera has marvelous flash, and good lenses albeit at a price. I can do things with a DSLR that I could not do with film e.g. shoot in mixed lighting, and get good white balance, or shoot long exposures and not suffer reciprocity failure, or colour shifts. (Remember Kodachrome?) It would be nice to have in camera stabilisation and an Olympus style sensor cleaner.
-
I think the next major innovation will be the death of the DSLR - perhaps Red will lead the way to mirror-free pro cameras.
-
There are for sure many areas where significant innovations could bring real world value for photographers, so what is preventing these innovations from surfacing?
I can think of several reasons:
- Priorities driven by perceived customers expectations focus resources on some other domains (more pixels,...),
- Technological difficulties make it expensive or impossible to implement some of the features,
- Manufacturers have agreed to control the pace of innovation in order to maximize the integral of the profit over time (the concurrent release of live view/video by all Japanese manufacturers could be explained by this), they want to make us buy one or 2 more generations of bodies before we decide that we can use our perfect DSLR for the next 10 years,
- Technological choices made at platform level make it hard to implement some features cleanly (interface between lens and body, physical packaging of body,...),
- All manufacturers rely on using some common supplier parts (chips,...) whose developement roadmap they don't totally control (live view comes to mind also),
- Some key thinkers inside the camera companies don't understand the need,
- They are stupid.
I let you pick which of mix of reasons is predominent for each of the features they didn't implement yet.
Cheers,
Bernard
-
I guess I don't feel like I miss many of these features you are discussing - pixel binning, etc. I look around and see the kinds of things people are achieving with their cameras, and I think if I am not able to achieve what I want with today's cameras, then it's probably my fault.
Just one example: see: http://www.mfilomeno.com/ (http://www.mfilomeno.com/) and look at the gorgeous things this guy Javier Vallhonrat is achieving with whatever camera he is using! It's killer, IMO.
I mean really, if he is able to achieve those incredible tones, sharpness, d.r., etc. then I feel like I really have no excuses. I believe that If I can't achieve what he is achieving in terms of IQ, I am probably lacking in imagination, skills, or both.
Just my opinion, of course.
-
I guess I don't feel like I miss many of these features you are discussing - pixel binning, etc. I look around and see the kinds of things people are achieving with their cameras, and I think if I am not able to achieve what I want with today's cameras, then it's probably my fault.
I'm only partly talking about IQ. I agree that IQ is almost there already - but since low-light cityscape photography is one of my main interests, I am in serious need of much more dynamic range and lesser shadow noise. Outside of IQ, UI is lacking severely in innovation. For example, why can't I dial a bracketed 5 and 80 second bracketed exposure? Instead, until recently I had to use a remote and a stopwatch to get anything beyond 30 secs - now Canon at least provides a second counter.
And the ideas I discussed were not meant to be an exhaustive list or even a kickstart for brainstorming; that was not the intention of the post. Besides, I'm not even talking about the truly revolutionary prototype-level ideas, such as plenoptic camera (http://graphics.stanford.edu/papers/lfcamera/), as most of the ideas presented exist already.
-
There's more to it that just a body with silicon in place of emulsion. The entire system upgrade has been fairly significant in terms of what it takes to get a picture today vs. twenty years ago. (i.e. Entire darkrooms have been replaced with a laptop and lightroom (or equivalent) software)). To say OEMs are stuck in the 90's... well, that's the object model from which we've all came from and are familiar with, and it will continue until we change the entire approach to how we shoot images, and you might not like the end result, as it might be too different.
-
True innovation for me will be tossing out the 3::2 rectangle. Tossing out the mirror too. Why DSLR's still use it is beyond me.
-
To me, Bayer sensors are the biggest hurdle and leftover from the old original digital technology. It's like trying to make a better and better incandescent bulb. Eventually you need to move forward with modern technology. As it is, it's at the effective limits of the technology and yet it still looks sub-optimal. Because, while it's very very nice, it's inherently a kludge.
Moires, jaggies, color artifacts, AA filters... the list goes on and on with the hardware and software because of this. Now I fully understand why they did it almost twenty years ago. Because back then it was really the only viable option with the poor technology they had at the time. But that was a LONG time ago.
Yet the makers of the new technologies that are the eventual future are way beyond "dropped the ball". Sigh. It's also why it's my greatest wish for modern photography. Because the promise is there and we're so close - but still nobody's doing anything about it.
-
To me, Bayer sensors are the biggest hurdle and leftover from the old original digital technology. It's like trying to make a better and better incandescent bulb. Eventually you need to move forward with modern technology. As it is, it's at the effective limits of the technology and yet it still looks sub-optimal. Because, while it's very very nice, it's inherently a kludge.
Moires, jaggies, color artifacts, AA filters... the list goes on and on with the hardware and software because of this. Now I fully understand why they did it almost twenty years ago. Because back then it was really the only viable option with the poor technology they had at the time. But that was a LONG time ago.
Yet the makers of the new technologies that are the eventual future are way beyond "dropped the ball". Sigh. It's also why it's my greatest wish for modern photography. Because the promise is there and we're so close - but still nobody's doing anything about it.
Currently, 99.9% of cars use combustion engines that use an architecture that many theoricians consider to be far from optimum.
Why? Because:
1. Huge amounts have been invested in this architecture, including a supply chain,... that make it a better option to stick to this highly tuned sub-optimal architecture instead of embarking on something new and risky.
2. The results are pretty good and it is unclear by when a new architecture with more potential would reach the levels of performance availalble today.
I agree that true RGB would be better, but the best Bayer sensors with AA filters (at the risk of getting negative comments I will once more put forward the d3x sensor) are pretty damed amazing in terms of sharpness and lack of articfacts. As of today, the impact of other imperfections in the image capturing system are typically a lot more important than the tiny imperfection introduced by the AA filter.
Cheers,
Bernard
-
My first response is that sitting here and grizzling about what manufacturers do or do not offer is pointless and a waste of time. However on careful reflection on Canon's offerings I've become so annoyed and cross I'm going to toss in my 2 cents worth. Don't get me wrong, they are not bad cameras on the whole, but
1) why can't I save as a DNG
2) have a MLU button
3) have no stupid direct print button
4) have user programmable exposure compensation. The list goes on. It's the feeling that the "features" I can have are decided by marketing department people bereft of souls that have never held a camera in their god-forsaken hands. This may well be true of other brands, but there is surely no need to be so obvious about it Canon. Having seriously considered Sony, if they can get better high iso performance and live view, it will be time to jump ship. Or even Pentax. They are interesting at least.
In the mean time, I shan't think about this again as it takes the pleasure out of shooting.
David
-
"Moire-less pictures without the softness inherent in high-pass filters?" - if only....
Aliasing is an inherent problem to be avoided in any sampled system. It doesn't matter what type of sensor you use, bayer pattern, monochrome sensor * 3 + prism, Foveon, etc. etc. All have the same fundamental issue....
Now, perhaps we can improve OLPFs so that they have a sharper cut-off slope. However, to achieve a sharp slope with filters, we need negative lobes on the filter, and hence negative photons - and that's not going to happen soon. Negative lobes are easy with math, but not so in a real physical filter that filters photons.
But.... Do we want sharp filters, because sharp filters ring, and ringing is a nasty artifact too. You can have aliasing, ringing, or soft image. Pick any one!
Solution of course, is to run the sensor resolution right up into the diffraction / lens limits and omit the OLPF and keep increasing the resolution of the sensor until no aliasing is visible. This is not efficient, but should work quite nicely. At this point you don't need fancy Foveon like RGB sensors (you'd not actually want to use a Foveon as the colorimetry isn't great, and the increase in noise to realize a colorimetry isn't great either - silicon is not a good color filter) and Bayer is more than adequate, cheaper, effective.
Graeme
-
Thanks Graeme,
I wanted to make the same points myself, but you expressed it much better than I could.
Erik
"Moire-less pictures without the softness inherent in high-pass filters?" - if only....
Aliasing is an inherent problem to be avoided in any sampled system. It doesn't matter what type of sensor you use, bayer pattern, monochrome sensor * 3 + prism, Foveon, etc. etc. All have the same fundamental issue....
Now, perhaps we can improve OLPFs so that they have a sharper cut-off slope. However, to achieve a sharp slope with filters, we need negative lobes on the filter, and hence negative photons - and that's not going to happen soon. Negative lobes are easy with math, but not so in a real physical filter that filters photons.
But.... Do we want sharp filters, because sharp filters ring, and ringing is a nasty artifact too. You can have aliasing, ringing, or soft image. Pick any one!
Solution of course, is to run the sensor resolution right up into the diffraction / lens limits and omit the OLPF and keep increasing the resolution of the sensor until no aliasing is visible. This is not efficient, but should work quite nicely. At this point you don't need fancy Foveon like RGB sensors (you'd not actually want to use a Foveon as the colorimetry isn't great, and the increase in noise to realize a colorimetry isn't great either - silicon is not a good color filter) and Bayer is more than adequate, cheaper, effective.
Graeme
-
I personally have a very interesting understanding of this issue, being outside, and now inside camera development, and what really got me was the "nuances". What may appear obvious, is seldom so when you get down into the actual implementation details. Hence my "if only..." - "if only sampling theory worked that way..."
The art of camera design is the art of intelligent compromise, and I don't mean deliberately limiting quality or functionality, but the tricky balance between mutually exclusive parameters like the triangle I mentioned earlier - {alias free, ringing free, sharp image} - you can have any one, or even some degree of combination, but you can't have a totally sharp image, with no aliasing and no ringing!
Graeme
-
The art of camera design is the art of intelligent compromise, and I don't mean deliberately limiting quality or functionality [emphasis mine], but the tricky balance between mutually exclusive parameters like the triangle I mentioned earlier - {alias free, ringing free, sharp image} - you can have any one, or even some degree of combination, but you can't have a totally sharp image, with no aliasing and no ringing!
I mean exactly that. Companies deliberately limit the quality or functionality of otherwise identical products all the time. That's why you don't see proper bracketing in entry-level dSLRs, although it's just a software switch. It is disabled to justify the higher cost of mid- and high-price offerings which do include it. Economic term for this is price discrimination, which I'm perfectly fine with as a finance pro and even as a customer.
But as I was lamenting, we don't get many of the most exciting compact camera features in dSLRs. So even if I wanted to pay a premium to get hassle-free expanded dynamic range from dual bracketed exposures in a dSLR, I can't. I'd have to get a Ricoh CX-2 - or do tedious non-creative contrast masking in post since I prefer dSLR glass.
-
Yup, I fully understand that type of compromize occurs. I wish it didn't. I much prefer trying to do the best you can!
Graeme
-
Some very valid and interesting points, enjoyed reading the thread. My take is that, a DSLR is just that, a Digital version of an SLR camera. Priority has been given to sensor developments to try and replicate film. The actual bodies have more or less been left alone, as have features wanted as expressed in other posts.
Thing is, would we loose sight of why we take photographs if a manufacturer produced a camera with every conceivable feature, thus, it might become a very expensive point and shoot, with the only thing left to the user being composition. Another point to consider, if more functions are included, the less thought is needed in taking a shot, would this make it slightly boring, I don't know.
Saying all that, I'm sure manufacturers have already created a feature laden model, but as others have said, whilst sales are still good, why would they release it.
-
Yup, I fully understand that type of compromize occurs. I wish it didn't. I much prefer trying to do the best you can!
It's not as bad as it sounds. The economic justification for it means that companies can extract money from consumers who would be otherwise unable to afford the product, or alternatively, willing to pay more for it. It's not all bad for us consumers: we get a product line which has something for every wallet, and features which otherwise would not be financially justified to put on the camera.
-
The art of camera design is the art of intelligent compromise, and I don't mean deliberately limiting quality or functionality, but...
Graeme
One theory of marketing is to try to avoid supplying anything that will do everything that anyone will want it to do for a decade... I think Red do not use that theory.
Another theory of marketing is to make something that will do everything that anyone will want it to do for a decade, but price it so that not everybody will buy it immediately.
I have recently bought my first "pro" digital camera, with the option to upgrade, as, until they invented 60 Mpx cameras that you could use as a point-and-shoot or on a view camera, I was not tempted.
Looking forward to the red 617.
-
Yup, I fully understand that type of compromize occurs. I wish it didn't. I much prefer trying to do the best you can!
Graeme
Graeme,
Since I come at this from a commercial photographer's point of view (when your a hammer, everything is a nail), I look at the still and motion cameras we have today vs. what was around 7 to 8 years ago and honestly I don't see much difference.
Maybe costs have plummeted on comparable systems like the 5d2 compared to a 1ds3 or a digital back, or the RED vs. what was available with a 35mm cinema camera, but as far as being able to offer up anything to a client that is really different or unique for advertising and marketing we're still pretty much producing, concepting, working and delivering imagery the same way we were 10 years ago.
Right now the economy has put everything into sharp focus as to what works, what doesn't what's needed, what's expendable and now more so today than anytime I can remember, I get this feeling that the advertising world is kind of sitting on their hands waiting for the next big thing, or the next big idea that will drive their clients into spending and clients are sitting on their hands waiting for that something that will drive their customers into opening their wallets.
when I look at real innovation, something like the new Iphone just blows away compared to what we have in professional equipment.
Shoot video, shoot stills, one button click to e-mail them to one person or the world, blog em, web em, comment on them get responses all while you can talk to your boyfriend, check the weather, hell it's such a different mindset compared to the professional equipment side of life.
With our equipment and software, we're still working traditional and nothing cross purposes that well. Sure a 5d2 on manual focus will shoot a video image, a RED with manual focus will produce a still image out of the motion clips good enough for reproduction and though both cameras are quite amazing, they don't really offer 1/2 of the usability of something like the iphone I mentioned and this somewhat stumps me.
Why nikon has a d3 that has amazing focus for stills, shoots 10fps almost for long periods but can't just ramp up and shoot 30 fps kind of confuses me. As somebody that works on the camera design side you probably know the answers to all of this, but how expensive and difficult is it to make take that Nikon or Canon, add another processor, a large buffer and let it rip?
The professional world is still living in this historic past. A 1ds2 nor Nikon D3 looks and works pretty much like a film camera, a RED pretty much like a film camera and even finding common ground on software is almost impossible.
Your much better versed and studied in this than I, but why we can't put a video clip in lightroom and use it's familiar functions and controls to work video as we post process the stills is kind of perplexing that Adobe isn't working day and night to get this out in a week. (maybe they are).
Maybe your next RED will have autofocus, high iso, lenses from 300mm to 10mm, familiar to use post processing software, or even better, easy to use in camera color and tone correction, I don't know, you do, but I do know that if you make the rounds from AD agency creatives, to corporate CEO's even the ones with advertising budget to spend, today aren't really sure where to put it, what will work, what medium will have the most play.
I think all mediums from motion to still, blogging to web, web to print, print to broadcast all shot under one production will be the future, but as slow as most of the equipment makers work, it will be a long time until we have devices that will do this.
JR
-
Great questions - and I'm probably not the right guy to answer most of them. I can talk a little on the readout speed sensor and image processing stuff though, as that's an area I've had to investigate.
With a stills camera, we have a physical shutter, which means readout from the sensor can be comparatively slow, without causing visual artifacts. When the readout needs to go faster, you need to read and reset either the whole sensor or line very much faster. From what I understand, to reduce this read/reset time is a non-trivial task. To avoid visual artifacts, the read/reset time needs to be fast enough so that the whole frame can be read out much faster than the goal fps for the sensor. And that is why, for instance, the 5D2 appears to only read out a third of the lines per frame as it just doesn't have a fast enough read/reset time to do more.
Adding a buffer to extend record time would be great - but how big a buffer and what sustained data rate? For the RED One, uncompressed data rates at 4096x2304x12bits at even just 24fps are quite large - over 330MB/s, so for even just a 4minute load you'd need not just that very fast sustained rate but 80GB storage too. So that's why we went down the REDCode RAW compression route, which puts all that onto a 8GB CF card.
So to get stills quality at video frame rates means you need a much faster sensor, and either new compression technology or very fast storage. It's not actually too much a function of "processing" if you're going un-compressed, or very much a function of special processing if you use a compression technique.
Why not Lightroom type processing on traditional video? Video is very rarely recorded RAW, and hence Lightroom style processing just wouldn't work, as that needs the larger dynamic range an lack of burned in gamma and matrix to function as it does. The second issue is that many things we take for granted in stills processing are made tricker for movie processing due to that movement inherent in movies, tracking windows / masks / grads as the image moves, for instance. Finally, of course, it's not to bad to get real-time feedback on a single image, but to do so at 24fps (at a minimum) for moving video is quite a trickier computational problem.
The digital stills "problem" has traditionally been a far easier one to solve than video. For over 50 years, although video cameras got better, they were still fixed at "standard definition" resolutions. Look at how slow the adoption of HD has been compared to the leaps in resolution we've seen in digital stills, and how digital has had the benefits of raw for ages, where video is mostly still in the "JPEG" type of processing. However, computing power is catching up, and now we can see what a stills / movie convergence can do. And I think once that parity of image quality is achieved, we'll start to see the innovations that are possible.
Graeme
-
I agree that Canon/Nikon/Sony should bring the best and new features to the hi-end cameras, but we photographers need basically 2 things to obtain the best work:
a) Good lenses
Outstanding film, in the digital world a very good sensor
Today the above 3 companies have excellent lenses, but on the sensor side, still need to make improvements, such as Sony A900 with D-max noise.
The rest of the goodies help you take thousands of pictures to select from, but best of them will need a) and to be outstanding, and above all, your eyes.
Gonzalo
-
Hi,
Did you try to shoot in the night using sensor based Live View?
Erik
True innovation for me will be tossing out the 3::2 rectangle. Tossing out the mirror too. Why DSLR's still use it is beyond me.
-
Hi,
Did you try to shoot in the night using sensor based Live View?
Erik
Bring back the wire-frame viewfinder ... and does anyone know of one that zooms auto as the lens zooms? (or an auto-zooming lens shade?)
-
A few innovations that should be coming down the pike very soon (we are starting to see the first one in a few cameras):
* Genuine automatic dynamic range expansion. This one is actually fairly easy, but may require being on a tripod to get right, since multiple exposures are required in the most obvious method.
* Super-resolution - obtaining a higher genuine resolution than the sensor has. This one is also fairly well developed, and it could be implemented using the same mechanism as in-camera sensor-shift image stabilization.
What seems to be tough, but is needed pretty badly:
* Better automatic white balance.
And now a difficult one, but one which would truly be the next, er, dimension in photography:
* True distance encoding per pixel. This would be "true" 3-D. That is, not stereo, but 3-D in the sense that each pixel would be red, green, blue plus distance.
Cheers.