The future of digital SLRs
Currently DSLR design still concentrates on packing more pixels onto the sensor and adding features to the camera firmware. As the megapixel race continues, however, it becomes more and more pointless. Most lenses can not resolve enough detail to require tens of megapixels, and sloppy technique often prevents that both lenses and sensors are used to their full potential. So as there is a limit of information that reaches the sensor, there's no point to use more and more pixels to record it. Once the pixel number of the sensor is higher than the total resolving power of the lens-photographer system, there is no advantage of adding more pixels. Also, more and more people will realize that they don't need 20 megapixels even for full page magazine prints. The novelty value of ever larger numbers in sales brochures will soon fade out.
So efforts will shift towards other areas. Mid- and longterm there will be increased emphasis on quality, usability, and getting rid of the past. Let's start with my look into the crystal ball:
Mechanical shutters are on the way out
Current DSLRs use mechanical shutters just like film cameras. That's not because it's the best way to construct a camera. It's because the technology is available and mature, and current full-frame CCD technology requires the sensor to be covered while it's read out. More advanced sensor technology does no longer need a shutter (as seen from today's digicams), but image quality does not reach that of full-frame CCDs. So with the quality-aware buyers of DSLRs, mechanical shutters are currently the right compromise.
However, I believe this will change. Sensor technology will improve to the point that image quality can be retained even without a mechanical shutter. Then there's no longer a need to use shutters. This will all of a sudden improve the system significantly:
- Shutters are high-precision mechanical parts. Or in other words, they are expensive. The money you save on shutters can be used better elsewhere.
- Shutters are slow. That's why SLRs have a flash x-sync speed that's significantly slower than their shortest exposure time. When you want to use flash at shorter exposure times, you have to use tricks like FP-sync flash or HSS. Without a shutter, none of these tricks are needed. You can use regular flash at full power up to the shortest exposure time. Again, more money to spend on something else.
- Shutters create noise and vibrations.
- Shutters sometimes fail. Without shutters you have fewer mechanical parts that can fail.
In future DSLRs there may be a simpler, slower, mechanical shutter that is only used to cover the sensor while you change lenses. But even this may be obsoleted (see below).
Oops, it's not a SLR anymore!
DSLRs still carry a lot of technology from old film cameras, namely the mirror, matte screen, prism and optical viewfinder. Again, this is not because it's the best way to build a camera. It's because the technology is available and mature. Looking at digicams again, you can see that you can build a digital camera without all that, using electronic viewfinders. But again, current technology is not up to the quality requirements of typical SLR buyers, so carrying over old camera technology is the right compromise at this time.
Again, I think this will change. Sensors will be improved to allow readout at high frame rates and at a lower resolution, and display technology will be improved with regard to resolution, speed and dynamic range. This will lead to electronic viewfinders that are just as good as current optical viewfinders. You will no longer need a mirror, matte screen and prism.
Throwing out old stuff can be liberating. Using electronic viewfinders will have the following advantages:
- The viewfinder no longer needs to be tightly coupled to the camera. It can be tilted up, swiveled and maybe even detached from the camera. You will never need an angle finder again. Alternatively, you can shoot tethered in a studio, using a large computer screen as a viewfinder.
- Viewfinder image and final image will never be out of sync. They are both taken by the same image sensor. It'll be easy to provide a 100% crop of the image in the viewfinder instead of the current 90-something percent.
- You have more freedom to manipulate the viewfinder image. Current optical viewfinders offer only a LCD status line and a few highlighted spots in the image area. With electronic viewfinders you can overlay the image with whatever you want. You can see a live histogram, important status messages etc., and the viewfinder image can be automatically displayed brighter when you use DOF preview to compensate for the light loss caused by the closed aperture. You can zoom into parts of the image for precise manual focusing. No need for an extra viewfinder loupe! You will always have a bright viewfinder image even when there is little light.
- You can record movies. Even advanced amateurs and even professionals may want to do this from time to time.
- Again, less noise and vibrations from the flip-up mirror. No need for a special mirror lock-up function.
- Again, fewer mechanical parts that may fail.
What you get will then be no longer a SLR in the literal meaning of the acronym. It will, however, still be a system camera, and it will look and feel like a SLR.
Of course, when you no longer have a mirror, you also no longer need to provide the space for it. Reducing lens registration (distance between image plane and lens mount) makes lens design easier, especially for wide-angle lenses. However, changing the lens mount would also mean that you have to buy new lenses. If you have thousands or even tens of thousands of dollars invested in your lenses, this is likely something that you don't want to do. So camera makers may keep the lens mount unchanged and use the available space for something else. For example, they may insert a protective glass into the space between lens and sensor. Dust would settle on the glass instead of the sensor, far away from the image plane and thus practically invisible. This would also make a protective shutter unnecessary.
Alternatively, makers may offer new lenses for a shorter registration distance, and a special extension tube to be able to use also old lenses that are designed for the longer registration distance.
There's one disadvantage. Without the flip-up mirror, it's no longer possible to use phase detection AF sensors as we know them today in SLRs. Camera makers will have to work hard to improve video AF to a level equal to today's phase detection AF.
Today's storage media are the bottleneck of the recording system. The write speed of today's NAND flash can not keep up with the increasing data rate from the sensor. So today's DSLRs need megabytes of image buffers so that at least a few images can be recorded at full speed before the camera has to slow down to write these images to the storage medium. These megabytes of RAM cost money and require power to operate, something that can be used better for other things (like electronic viewfinders). Future storage media will still use common formats like CompactFlash, SD card, Memory Stick, etc., but the technology inside will be something faster.
CPUs help lenses
Future lenses will provide additional information to the camera, for example their level of distortion and vignetting. While it's still best to optimize sharpness and chromatic aberrations in the lens, other things can more easily (and cheaper!) be corrected by the camera. So lens design compromises may shift towards sharpness and chromatic aberrations and away from distortions and vignetting, still yielding better overall results. For old lenses, the camera may either apply no corrections (gaining nothing, but also losing nothing), or cameras may be eqipped with a database containing these parameters for older lenses.
Sensor technology will concentrate more on improving dynamic range, color rendition and general image quality than on cramming ever more pixels on the same space. Photographers will probably profit most from increased dynamic range, which is one of the more problematic areas of current sensors. The Fuji sensors hint towards what we'll see in the future.
Color accuracy will also be improved. There are already sensors recording four colors instead of three, and this is something that we may see more often.
There's also still room to improve efficiency of current sensors, in terms of sensitivity, S/N ratio, and chip area usage. These improvements can all be done without making chips bigger, i.e. without making them more expensive.
Update from August 5th, 2008
Obviously Olympus and Panasonic have started making the changes described above. They have announced their Micro Four Thirds lens mount, which basically has all the above characteristics:
- No mirror
- Half the registration distance
- Contrast detection AF
- Electronic viewfinder
- Movie recording
- Can use old lenses with adapter
It remains to be seen how the new system will be realized in actual products, especially with regard to the quality of the AF system and the viewfinder. Also, the advantages for lens design will only be seen with new native lenses, not with adapted legacy lenses.
For details see the Four Thirds home page.
© 2008 Michael Hohner; This page was last changed on 2008-08-05
|#1: Comment posted by Torsten Crull on December 29, 2007 - 11:54:36 PM:|
|zu "Camera makers will have to work hard to improve video AF ..."
Der "phase detection AF" wird ja auch ständig verbessert. Ich mache Konzertfotografie und kann mir ehrlich gesagt nicht vorstellen, dass das mit "video AF" jemals annähernd so gut gehen wird, wie mit einem dann aktuellen "phase detection AF".
Gibt es Indikatoren dafür, dass die eine Technik die andere überholen wird?
|Michael Hohner answers:|
|So riesig kommen mir die Fortschritte beim Phase Detection AF in den letzten Jahren nicht vor. Wenn die Hersteller die gleiche Arbeit und das gleiche Geld in die Weiterentwicklung des Video AF stecken, dann könnte dieser sicher noch deutlich aufholen. Der Punkt im Artikel ist vor allem, dass diese Weiterentwicklung dringend nötig sein wird, wenn Phase Detection AF nicht mehr verwendet werden kann.|
|#2: Comment posted by Hans de Korte on February 18, 2008 - 04:54:29 PM:|
|How long will it take before we see a 'DSLR' camera on the market as you have described above?|
|#3: Comment posted by augustofretes on July 06, 2008 - 05:20:45 AM:|
|Unless EVF (or, well, technically, the sensor) can match the real state speed of an OVF (look at any P&S or Live view implementation they lag) and can match with resolution that one find in a OFV (which is not going to happen), I don't see how this will happen.
And... no more shutters, this may happen, but, for some reason it scares me, it's like feeling some basic things of photography are getting lost.
|Michael Hohner answers:|
|Well, never say never when it comes to technology, unless it's physically impossible. I don't see how a 50 fps 1-2 MP EVF is physically impossible.
Also don't think that the resolution of current optical viewfinders is that high. The matting structure of the matte screen limits the resolution.
|#4: Comment posted by Sandy Wilson (UK) on October 09, 2008 - 05:15:43 PM:|
|Watch out for RED Cinema (an American company)in early 2009 for their new revolutionary camera. It is going to shake up Canon and Nikon and the rest of the DSLR manufacturers.
There is nothing in their web site as yet,but there has been a mention of it by one of their directors. There has also been rumours on other blog sites, but no techinical details.
|Michael Hohner answers:|
|Once they have out more than just hot air, I will have a look. Until then I will not hold my breath. But my experience is that when a company itself declares their product "revolutionary", most likely it isn't.|
|#5: Comment posted by Detlef Bodart on December 02, 2008 - 07:09:03 PM:|
|One innovation no-one has mentioned yet is a binocular eye-level viewfinder. Binocular vision permits a better picture composition than the present SLR single-window type, and is one of the reasons why with all their limitations, Rolleiflexes and Hasselblads without prisms were so popular with professionals in spite of just showing mirror images. With some of the viewfinder screens of present budget DSLRs being simply pitiful regarding size, brightness and ability to focus, there seems to be more room and need for improvement here than with any other feature, under the aspect of finding selling points in a competitive market. It is time that people judge cameras by looking into them to see how visual images come over bright, large and clear, rather than by judging them just on account of pixels and outside ergonomics.|
|Michael Hohner answers:|
|If with “binocular eye-level viewfinder” you refer to TLR designs, then I have to disagree. TLRs have so many systematic problems (parallax, metering, double effort for lens, fixed lens, etc.) that they will surely not enter the mainstream again.|
|#6: Comment posted by Greg L. on May 13, 2009 - 01:52:31 AM:|
|With the demise of mechanical shutters and viewfinder mirrors, could we not also see the possibility of tilt and shift movements in the sensor plane, to allow digital cameras the same perspective-altering capabilities as large format view cameras?|
|Michael Hohner answers:|
|At least with shift lenses you need an image circle that is significantly larger than that of normal 35 mm lenses. It may work when using 35 mm lenses on APS-C cameras, but not when using 35 mm lenses on full-frame cameras. So this can not be done in the camera alone.|
|#7: Comment posted by Luis Augusto on September 21, 2009 - 04:39:17 AM:|
|You know, I think you're right. I wonder how long will it pass before we see a Pro Micro Fourth Thirds (EVIL) camera.|
|#8: Comment posted by Peter B on February 26, 2010 - 12:07:52 AM:|
|The major problem in using non retrofocus wide angles with digital cameras is not just the mirror but the sensors being unable to accept off axis light rays , hence for the Leica M9 a heavily modified sensor chip had to be designed to enable moderate wide angle lenses to be used , more extreme wide angles need redesigning with a degree of retrofocus to reduce the subtended angle ( the same consideration applies to using shift movements ).
Currently the electronic view finders are too poor to make for comfortable use hence DSLRs popularity , once the finders have instant response and good enough resolution to manually focus lenses these cameras will take over ( for low light portraits autofocus always seems to choose the wrong subject , manual focus tends to be easier than trying to get focus to lock in the right place eg. eye in focus not nose end! )
|#9: Comment posted by Chester on July 01, 2010 - 03:43:38 AM:|
|Resolution and update speed/AF speed issues aside, EVFs have another disadvantage: they consume much more battery power than optical finders. Powering a mini LCD screen takes more power than lighting up a line of LEDs. Moreover, the camera's processor has to work constantly to read data off the sensor and generate an image for the screen. This not only drains even more power, but the continuous sensor use might also heat the sensor up, increasing noise. Even on my compact, I often use the (terrible) optical tunnel finder to catch action and save battery power.
The EVF view would also be restricted by the camera's sensor's capabilities. The preview would be noisy or completely dark in low light, and could be hard to see in bright light.
So...I really think SLR viewfinders should stay. Good EVFs are already present in superzoom bridge cameras (Minolta DiMAGE A2 - http://www.dpreview.com/reviews/konicaminoltaa2/page3.asp <= I don't actually own this camera but it seems to have a good EVF), but they haven't made their way into DSLRs for some good reasons.
|Michael Hohner answers:|
|I think you're underestimating technical progress. The A2 is already 6 years old. With today's technology you could build a much better system. And while the power consumption of a EVF will always be higher than that of an optical viewfinder, there will be a line where the advantages of EVFs outweigh the disadvantages. Live View is already commonplace with DSLRs today, and the higher power consumption does not seem to be a major issue for their buyers.|
|#10: Comment posted by Chester on July 19, 2010 - 05:52:10 AM:|
|Live view (with the main sensor) seems to be used more for studio work, where composition and focus can accurately checked, and macro work, where articulated LCD screens let you take pictures without being a contortionist. People aren't using live view to compose every photo, which is why their batteries are lasting.
The contrast detection autofocus method used by cameras offering main sensor live view is just too slow for some other applications, like sports and wildlife. No matter what happens, contrast detection AF will be slower than phase detection, because it has to iteratively hunt for focus. Moreover, contrast detect AF will have a tough time following moving subjects because it can't tell if the lens is front or back focused.
Also, one of my favorite ways to take sports photos is to keep both eyes open (using the optical finder) and see the frame superimposed on my vision. Even crappy, inaccurate point and shoot finders work for this - I just crop later to change composition. I've tried this with camcorder EVFs, and they've really been a pain.
By the way, I have a Minolta 7000. The viewfinder is big, bright, and simply awesome. It's easy to see all of the viewfinder info in bright sunlight, and there's no noise in the finder when it gets dark. The phase detection AF is also quite good with the right lens. The screw drive makes enough noise to scare anything away, but hey, it's fast and it works! I'm not giving up on OVFs and phase detect AF anytime soon.
|#11: Comment posted by Serge on July 19, 2011 - 10:58:16 AM:|
|Your point or vue is pretty interesting and I do agree on most of it but you could maybe update it since new models have come up...., Sony slt 55 is a transluscent mirror dslr with electronic viewfinder combining all thé advantages you mentionned and effectvely allowing video shooting with phase detection autofocus and high speed photography...
As you also mentioned, the evf compensates light loss when viewing deep of field with the real aperture and its global quality is quite excellent, high définition and extented dynamic, plus a lot of features like C15 enlargement for manual focus and many useful informations and tools.
Anyway, thanks for your great work on this website, home you'll keep it growing.
Cheers, Serge, Belgium
|#12: Comment posted by rishio on August 20, 2011 - 11:37:56 AM:|
|Very well written and ahead of it's time. A couple thoughts:
"Sensor technology will concentrate more on improving dynamic range, color rendition and general image quality than on cramming ever more pixels on the same space."
Unfortunately this doesn't seem to be happening. The next generation NEX APS-C system is reaching 24 megapixels. Do you think there is still room to improve efficiency in sensors? Seems awfully cramped. Is full frame the only option to get that "space" back?
What's your viewpoint on sensor format. Do you think the future is full frame or APS?
"There's one disadvantage. Without the flip-up mirror, it's no longer possible to use phase detection AF sensors as we know them today in SLRs."
This may not be true. As stated on mirror less rumors, "The upcoming new LA-EA2 adapter comes with a built-in translucent mirror. This will allow you to use all A-mount lenses with fast phase detection autofocus on your NEX camera."
I imagine this allows sony to bring out a new, electronic (full frame?) A-mount when they remove the mirror and have an adapter that allows the camera with new mount to make use of phase detection for the older lenses. Do you think Sony will abandon the current A-mount for a new, full frame, electronic mount (like the emount) or will they keep the current A-mount alive and live with the longer flange distance it inherits? Or perhaps the days of the full frame are over...
|Michael Hohner answers:|
|Regarding sensor formats, I don't the future is any one size. There will probably always be all kinds of formats, from tiny to large, from cheap to expensive.
The LA-EA2 does not really invalidate what I've written, because it enables phase detection AF by using a semi-transparent mirror, like SLT cameras do. It does not flip up and down, but the basic principle is not much different from “traditional” SLRs.
|#13: Comment posted by Jarno Pelkonen on September 12, 2013 - 06:13:02 PM:|
Thank you for this fantastic and insightful article. Five years have passed since you wrote it and Mirrorless is clearly gaining ground. Do you have time and energy to write an update?
I am not a camera technology expert, but I have few questions / thoughts of the development:
1) Auto-focus performance without separate phase detection sensor?
Are you aware of fundamental limitations of Canon's Dual-pixel approach vs. a dedicated PDAF sensor (like in DSLR)? DPreview writes that the Dual-pixel sensor is as fast as dedicated sensor in 70D so it seems the technology is very promising. Are there other limitations than what the mirrorless system has vs. DSLR approach? E.g. are DSLRs using a separate sensor for metering and is there any benefits of it vs. using the main sensor?
2) Computational imaging
As you wrote, mirrorless design offers many cost and size advantages vs. DSLR, but maybe the biggest opportunity space for mirrorless design is in my view "computational imaging". Smartphones are leading the pack here and introducing plenty of imaging enhancements relying on extensive computation power and "always-on imaging sensor". These include zero-lag shutter, pre-emptive photo capture, best-shot-capture, automatic HDR and improved low-light imaging (longer shutter times and combination of several images).
Due to the opportunities of computational imaging, I believe Mirrorless design is the way to go in any sensor size. Any comments? Any thoughts on computational imaging and its applications in "DSLR" / Mirrorless camera domain? What are the trade-offs for not having mechanical shutter?
3) Lens design for shorter Registration/flange distances in Mirrorless systems vs. DSLR
It seems Both Canon and Sony have arrived to very similar flange distances with their mirrorless mounts (18mm). That is quite a different vs. DSLR designs (Canon EF mount has 44mmm flange distance). What I do not understand that why Canon made their mounts such that it cannot accommodate full frame sensor. Sony seems to be able to accommodate a full frame sensor, but they have raised concern over challenges of creating suitable full frame lenses for such a short flange distance (http://photorumors.com/2012/10/16/sony-full-frame-e-mount-interchangeable-lens-camera-is-possible-making-lenses-is-another-matter/).
What is your take on lens design challenges for digital full frame (35mm) mirrorless cameras with such a short flange distance (18mm)? As you mentioned digital sensors cannot register light arriving at steep angle (vs. film). Are the challenges more in telephoto or wide-angle end? What kind of issues there are and which of the challenges be compensated with algorithms and computation power?
4) Lens' capability to resolve
I do not know does this question have any point, it is more about curiosity.
You correctly wrote how megapixel race would cool off. First, there seems not to be many applications for ultra high resolution pictures beyond 20Mpx in consumer/prosumer markets (D800 owners may disagree, but Canon 5DmkII with its 21MPx is enough for 22" 300PPI prints). TVs are also just about starting to move to 4K (only 9MP) and that is plentiful of resolution even for 150" screens (PPI ~31, but with 2m viewing angle it is roughly similar perceived
resolution than looking at retina iPhone at 20cm). So it is high ISO performance, color accuracy, etc. that drive sensor size vs. just Megapixels. Of course extra megapixels help sloppy photographers like me to re-frame the images later in Lightroom. I would still set max sensible DSLR resolution limit to 40Mpx. I just cannot figure out what I would do with all the extra pixels (except digital "zoom" / cropping). Increased resolution goes also against Computational imaging trend since extra pixels increase processing requirements.
But out of curiosity, where lies the limit for lens' capability resolve? i.e. Where is the limit of "full-frame megapixels" (i.e. pixel density) where even the best lenses reach their resolution limits and further megapixel improvement will not yield any noticeable image resolution improvement? One lens reference could be latest Canon/Nikon 70-200mm/2.8 pro zoom and another reference could be good quality pro prime (like Sigma Art 35mm/1.4). The upper boundary for lens resolution of currently available technology is likely set by Coastal Optics 60mm UV-VIS-IR APO Macro (at any even remotely feasible manufacturing cost).
I find myself investing into especially sharp lenses even my camera is "only" 21Mpx (5DmkII). Just wondering how many of my current sharp primes I would need to replace with even higher quality optics in case DSLRs would go to e.g. 40MPx. Btw. The highest sharpness (Perceptual MPx) DxO Mark gives on any prime on Nikon D800 is only 28Mpx vs. D800 sensor's 36Mpx. The lens is btw all but affordable Nikon AF-S Nikkor 200mm f/2G ED VR II.
5) Concluding remarks
- It seems DSLR is about to be replaced by professional mirrorless designs.
- computational imaging opens up huge opportunities also in real-camera space. This is technology disruption in progress and camera houses need to catch the wave or get disrupted by other vendors.
- Optics seem to be already on critical path in resolution / megapixel race.
Any comments Michael or others?
|Michael Hohner answers:|