At 24 MP, that is 22% increase in lineal resolution over 16 MP. Some lenses may begin to be the limiting factor. Furthermore, image stabilization may be more essential also. At 16 MP, I can already make prints larger than I normally need, so increasing resolution is of little value to me. I hope improvements will be in dynamic range & high ISO performance along with low noise A/D conversion. My dream would be a camera with at least 18 ev or larger dynamic range at base ISO and at least 14 ev at ISO 6400.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Any credible rumours on K-r replacement?
- Thread starter Brett St Pierre
- Start date
GordonBGood
Veteran Member
Twong, I agree with you that most of us don't really need more resolution, although it will likely come whether we really think we need it or not. However, do you know what you are asking for in your Dynamic Range (DR) requirement?
First, for a state-of-the-art camera as you are speculating here, the limit to DR is likely mainly to be the black read noise of the sensor and Analogue to Digital Converter (ADC) combination. Since ideally the black read noise changes linearly with ISO sensitivity gain, in order to deliver 14 EV of DR at ISO 6400, one would need 20 EV of DR at ISO 100 if that were the "base ISO" to meet this requirement!
Next, current silicon photo sensors with their maximum possible efficiency don't have more than about 64,000 electrons as their Full Electron Well Capacity (FEWC) at an ISO of about 100 for a reasonable photosite density. In order to have a sensor black read noise low enough to support a 20 EV DR at this ISO would require a black read noise of less than one sixteenth of an electron; even an 18 EV DR would require a black read noise of less than a quarter of an electron. As electrons aren't really divisible, this essentially means zero sensor black read noise, which may be possible but very unlikely over all environmental conditions including sensor heat.
Then, even if the sensor had zero black read noise, such a wide DR requires at least an ideal 16-bit ADC for 18 EV of DR and 18 bits for 20 EV DR, and by "ideal" I mean that it must act as if the Effective Number Of Bits (ENOB) rating of the ADC is good to the very Least Significant Bit (LSB), which usualy means it needs at least a couple of extra bits.
Yet another limit to ever being able to deliver such a wide DR using silicon photo sensors is the random statistical arrival of photons or the so-called "shot noise" and the definition of the lower range of DR as where the standard deviation of the noise drops so that the Signal to Noise Ratio (SNR) is 1 : 1, which is also about the practical limit where small details start to get "buried in the noise". For very low signal levels and zero black read noise there are two solutions to the equation of this limit and one of the solutions is that the SNR is 1 : 1 when the signal is one electron = photon (at 100% efficiency), which for a 65,536 electron photo detector happens at 16 EV below the bright clipping limit (of 65536 electrons).
I think that practically the real limit to DR, at least for current silicon photo dectectors, is about 16 stops at lowest ISO of say 100, and it will likely take at least another five to ten years to get there from the current about 14 stops. Then the limit will scale with ISO sensitivity for 12 EV at ISO 1600 and 10 EV for ISO 6400 (without noise reduction).
It might be possible to increase this by an EV or two if CMOS sensor technology were adapted to larger sensors such as Medium Format (MF) or Large Format (LF) digital cameras, but the cost would be huge as the market is quite limited.
I think that the drive for wider DR will likely ease off once digital cameras exceed the DR of colour negative film at about 12 EV at lowest ISO, which they likely do now with the best current digital cameras delivering about 14 EV at lowest ISO.
Regards, GordonBGood
kitsios_spyros
Veteran Member
What a reply!!
Thank you for your explanation! I would still be dreaming for a crazy DR -like twong wrote- to be available in the next couple of years.
But, I can always keep on dreaming...
Hope that engineering will suprice us with earlier announcements of such a technology
Regards
Spyros
Thank you for your explanation! I would still be dreaming for a crazy DR -like twong wrote- to be available in the next couple of years.
But, I can always keep on dreaming...
Hope that engineering will suprice us with earlier announcements of such a technology
Regards
Spyros
Brett St Pierre
Veteran Member
Thanks Gordon, I always love your technical replies. I cannot refute the science so I just accept them as knowledge that is otherwise difficult to come and enjoy pondering the real world application of that knowledge for which you usually provide enough leads to keep me thinking a little while. Much better than the loose statements the rest of us usually make.
Regards, Brett
Brett
http://www.pbase.com/shreder
The Journey is the Thing
Regards, Brett
--
Brett
http://www.pbase.com/shreder
The Journey is the Thing
Brett St Pierre
Veteran Member
Agreed Dean, especially at the price he has been offered... cheaper than a high end compact.
--
Brett
http://www.pbase.com/shreder
The Journey is the Thing
Brett St Pierre
Veteran Member
Fair comments. As much as I'd like the K-5 reality is my K-7 still takes great images (and is an great DSLR that suits me in every way - that body is a dream), has a lot of life in it and just like I skipped the K20D I'd prefer to wait for an even more of a quantum leap than just the next model. The only thing that would change that for me would be if Pentax looses it's way again and starts making big "lumps" for DSLR bodies. The current K5/7 body is about the best I have seen in 25 years of using SLRs and it warmed me that Pentax found their way back to making smallish, nice handling bodies.
--
Brett
http://www.pbase.com/shreder
The Journey is the Thing
Thanks for the reply. Your insight is always worth reading. I know I am asking a lot.
I think most of the camera buyers are new to photography and have little to no background in film. Their point of reference is their eyes and not film. (I work in a R&D organization, so this is normal for me to ask for more.) Even 16 ev or 17 ev DR is still 2 to 3 ev larger than K-5. Now that is a good reason for me upgrade from K-5 ...
It is difficult to reduce read noise, but what about well capacity? Increasing the surface area (FF or MF) is one way, but what about well depth? Or ... hold on to your seat ... split beam to 3 separate primary color light sensors? It would be difficult to implement this idea in the current DSLR, but it is quite feasible in a mirrorless camera. There is of course the usual alignment problem. Elimination of the Bayer pattern may also improve color accuracy.
Unless there is strong consumer demand, manufacturers will not voluntarily offer such solutions. Otherwise, we are more-or-less like mindless sheep being slaughtered for corporate profit.
I think most of the camera buyers are new to photography and have little to no background in film. Their point of reference is their eyes and not film. (I work in a R&D organization, so this is normal for me to ask for more.) Even 16 ev or 17 ev DR is still 2 to 3 ev larger than K-5. Now that is a good reason for me upgrade from K-5 ...
It is difficult to reduce read noise, but what about well capacity? Increasing the surface area (FF or MF) is one way, but what about well depth? Or ... hold on to your seat ... split beam to 3 separate primary color light sensors? It would be difficult to implement this idea in the current DSLR, but it is quite feasible in a mirrorless camera. There is of course the usual alignment problem. Elimination of the Bayer pattern may also improve color accuracy.
Unless there is strong consumer demand, manufacturers will not voluntarily offer such solutions. Otherwise, we are more-or-less like mindless sheep being slaughtered for corporate profit.
GordonBGood
Veteran Member
Twong, but the question remains, how would you use such a wide Dynamic Range (DR) considering most high contrast scenes don't have more than about 10 to 12 EV? Can you post some examples showing that you are limited by about 14 EV DR?
To partially answer my own question above, I don't think we are so much limited by 14 stops of DR at low ISO's but it would be nice to have more than eight EV/stops of DR at ISO 6400 (about nine EV/stops with the Pentax K-5 due to the compulsory Noise Reduction used for ISO 3200 and above).
What you are talking about is means of increasing effective light collecting efficiency, and I took some of that into account in the estimate of a Full Well Electron Capacity (FWEC) of 65,000. Increasing well depth of silicon photo detectors is difficult other than by increasing sensor area, but I anticipated that some gains might be made. Manufacturing of cameras and perhaps even very expensive video cameras seems to be moving away from the three sensor/split beam complexity and complications as with higher sensor photosite densities it is possible to get better image quality without it, and there will still be losses in the prism splitter. One of the main reasons that these three sensor designs exist for video was that the technology of the time wasn't able to produce fast enough and high enough resolution by other means, but that isn't true any more. Look up articles here by Joseph S. Wisniewski on this subject for more detailed analysis.
Well, companies need to turn profits in order to fund more research, and producing cameras with features that many photographers never use would appear to to contrary to that requirement. In fact, the majority of everyday photographers rarely use a scene capture DR of more than about nine EV. You'll notice that many of the laments about limited DR are often related to shooting JPEG's or JPEG processing which intentionally has the displayed DR compressed to about this due to a contrast boost in digital image development even though the captured range is much wider or lamenting limited "highlight headroom", which can be compensated with today's cameras by just slightly underexposing and modifying the development.
Regards, GordonbGood
paulkienitz
Veteran Member
Of all the numbers in that post, the one most easily changed is the 65000 electron ceiling for the well capacity. I'm sure there are all kinds of technical tricks that could be played to stretch the upper end of the response curve beyond where today's sensors start clipping. Even if such techniques become noisy and nonlinear and inaccurate, they could still go a long way towards producing nice-looking highlights in a picture that's otherwise exposed much the same as today. That could give you another 2 EV or so of DR on the top end.
--
keepers: K10D, Sig 17-70, DA 55-300, FA 50/1.4 "Billy Bass"
discards: DA 50-200 "zipper", F 100-300, Sigma 135-400 "piglet", M 400/5.6 "the Great Truncheon"
--
keepers: K10D, Sig 17-70, DA 55-300, FA 50/1.4 "Billy Bass"
discards: DA 50-200 "zipper", F 100-300, Sigma 135-400 "piglet", M 400/5.6 "the Great Truncheon"
Film responses are quite non-linear at both ends of the exposure curve .... and many photographers actually like it that way. And non-linearity is measurable and correctable by software. You can have it either way with an adjustable parameter. Wow ... that is a marketable feature along with increased DR.
16MP Obsolete!?!
Gimme a break!
Only about three years back a friend of mine was at one of the Canon professional gatherings where the Canon rep was answering questions. He was quite candid in admitting that Canon had trouble developing and even testing lenses (their very best LD stuff mind you) that could possibly resolve 16MP.
DO YOU THINK NORMAL CONSUMER OR MIDGRADE GLASS IS EVEN CLOSE TO RESOLVING 16MP?
This STUPID megapixel race is biting us all in the butt. For me, a 12 or 14MP image is all could want.
If they want to make top-end specialty cameras with 24MP sensors that require $3000 lenses to actually resolve that... then fine... it's a specialty product.
Unfortunately though the high pixel counts will filter into all categories and we will all be saddled with massively increasing volumes of image data we cannot put to use. So we'll have the choice of in-camera interpolation down to a lower resolution file or post processing all our images just so they're manageable! A friend recently bought one of the new Canon bodies and was really rather annoyed to find that the next step down from 18MP in-camera was 8! So he HAS to post process everything or deal with the larger files. No thanks. We may well find ourselves remembering the good-ole' days when we had 12 and 14MP bodies.
And... as was explained, very well, earlier in the thread, as MP count increases, sensor site sizes decreases impacting the DR and noise characteristics of the imager.
Those of you who can't get enough megapixels need to face it... just like the computer industry has had to give in to physics so will the camera industry. WE WILL NOT HAVE 50, 80 or 100 MP DSLR's in the future. We are already bumping up against the limits of silicon sensors and lens physics.
Sorry for the rant but I hate this megapixel marketing crap!
--
Happiness is a want... Contentment is a choice.
Gimme a break!
Only about three years back a friend of mine was at one of the Canon professional gatherings where the Canon rep was answering questions. He was quite candid in admitting that Canon had trouble developing and even testing lenses (their very best LD stuff mind you) that could possibly resolve 16MP.
DO YOU THINK NORMAL CONSUMER OR MIDGRADE GLASS IS EVEN CLOSE TO RESOLVING 16MP?
This STUPID megapixel race is biting us all in the butt. For me, a 12 or 14MP image is all could want.
If they want to make top-end specialty cameras with 24MP sensors that require $3000 lenses to actually resolve that... then fine... it's a specialty product.
Unfortunately though the high pixel counts will filter into all categories and we will all be saddled with massively increasing volumes of image data we cannot put to use. So we'll have the choice of in-camera interpolation down to a lower resolution file or post processing all our images just so they're manageable! A friend recently bought one of the new Canon bodies and was really rather annoyed to find that the next step down from 18MP in-camera was 8! So he HAS to post process everything or deal with the larger files. No thanks. We may well find ourselves remembering the good-ole' days when we had 12 and 14MP bodies.
And... as was explained, very well, earlier in the thread, as MP count increases, sensor site sizes decreases impacting the DR and noise characteristics of the imager.
Those of you who can't get enough megapixels need to face it... just like the computer industry has had to give in to physics so will the camera industry. WE WILL NOT HAVE 50, 80 or 100 MP DSLR's in the future. We are already bumping up against the limits of silicon sensors and lens physics.
Sorry for the rant but I hate this megapixel marketing crap!
--
Happiness is a want... Contentment is a choice.
Jack Simpson
Forum Pro
.... there will be a replacement within 12 months, Meester Brett But, of
course, that's true of (almost) digital cameras
Cheers,
Jack
--
STREET PHOTOGRAPHY DOCUMENTARY: http://www.youtube.com/watch?v=6kkHKP4Gnd0
( UPDATED NOV 16th )MY BLOG.... http://www.nakedmanonawire.blogspot.com
MY EMAIL ADDRESS IN IS MY 'VIEW PLAN'
It's amazing what one can do when one doesn't know what one is doing
But, of course, that's true of (almost) digital cameras
Cheers,
Jack
--
STREET PHOTOGRAPHY DOCUMENTARY: http://www.youtube.com/watch?v=6kkHKP4Gnd0
( UPDATED NOV 16th )MY BLOG.... http://www.nakedmanonawire.blogspot.com
MY EMAIL ADDRESS IN IS MY 'VIEW PLAN'
It's amazing what one can do when one doesn't know what one is doing
A great article also making the point that we are nearing the limits of imager technology.
http://www.clarkvision.com/articles/digital.sensor.performance.summary/
Just as we don't have 10Ghz computers on our desktop (which certainly should have arrived by now - barring physical limitations) we won't have 100MP DSLR's someday.
Even Nikon is basically admitting that 12MP is plenty! Their $2700 D3s is 12MP. Only their absurdly expensive D3x makes the jump to 24MP.
I sincerely wish Canon had not started sticking 18MP sensors in lower-end cameras and hope 24MP stays up in the realm of specialty pro cameras. I don't want a 24MP enthusiast camera.
--
Happiness is a want... Contentment is a choice.
http://www.clarkvision.com/articles/digital.sensor.performance.summary/
Just as we don't have 10Ghz computers on our desktop (which certainly should have arrived by now - barring physical limitations) we won't have 100MP DSLR's someday.
Even Nikon is basically admitting that 12MP is plenty! Their $2700 D3s is 12MP. Only their absurdly expensive D3x makes the jump to 24MP.
I sincerely wish Canon had not started sticking 18MP sensors in lower-end cameras and hope 24MP stays up in the realm of specialty pro cameras. I don't want a 24MP enthusiast camera.
--
Happiness is a want... Contentment is a choice.
paulkienitz
Veteran Member
Absolutely. That nonlinearity can totally be a feature, not a bug. I can think of two or three different ways it could be done, only one of which is to genuinely make the electron bucket bigger. That would avoid the nonlinearity but would probably also make the readout slower, which would cause problems for video.
And the problem could also be tackled with the Fujifilm approach of having small low-sensitivity photosites in between the main high-sensitivity ones. That might be the best way to go, though not the cheapest. You'd have to work around their patents, for starters.
--
keepers: K10D, Sig 17-70, DA 55-300, FA 50/1.4 "Billy Bass"
discards: DA 50-200 "zipper", F 100-300, Sigma 135-400 "piglet", M 400/5.6 "the Great Truncheon"
paulkienitz
Veteran Member
Yes we will. But this does not mean we'll have 50, 80, or 100 MP finished images. It does mean we can get rid of antialiasing filters, have no worries about moire artifacts, and stop relying on guesswork to figure out the colors in a bayer interpolation.
paulkienitz
Veteran Member
It's not at all rare that the only way to really save the highlights is to underexpose drastically, not "slightly". And I don't think it's an abstruse theoretical or technical issue for the average unsophisticated snapshooter... the fact is that today's cheap digicams constantly turn clouds and waterfalls and so on into solid clipped white, and it's not uncommon to blow out patches of skin if the light isn't behind the shooter. In that area the average snapshooter might well still be better off with film.
paulkienitz
Veteran Member
Don't give him a K10d, start with a K200d.
I guess that MPixels will increase on cameras as screens increase size. For example, in the future when Sony sell wall paper full of leds for decoration where you can put on your photos or videos, there will be cameras FullMegaHD prepared for that.
Well correct me if I'm wrong, but...
from what I am aware of, the physics of imagers are starting to become a barrier. A photosite on an imager is going to be limited by the number of photons that are incident upon that area of the imager for a given aperture, illumination level and exposure time. As the photosite shrinks so shrinks the number of photons incident upon it under the same parameters. There's no getting around it. Thus the reason that tiny P&S imagers don't approach the quality of APS-C or full frame. In this 50 or 100MP imager scenario the 12MP imager many of us are using would be represented by only a small portion of the imager area - very similar to the size of a little P&S imager. Do you think that small portion would return the same image quality as what we're using now.
Will some super-technology take care of all the physics issues?
Will some new unobtanium laden glass be used to create lenses that can resolve 100 million pixels accurately?
We are getting close to the point of diminishing returns.
Since you suppose that if we're going to have 100MP imagers why not 200MP or 500MP?
At what point will physics or just plain pointlessness stop the pixel count?
--
Happiness is a want... Contentment is a choice.
from what I am aware of, the physics of imagers are starting to become a barrier. A photosite on an imager is going to be limited by the number of photons that are incident upon that area of the imager for a given aperture, illumination level and exposure time. As the photosite shrinks so shrinks the number of photons incident upon it under the same parameters. There's no getting around it. Thus the reason that tiny P&S imagers don't approach the quality of APS-C or full frame. In this 50 or 100MP imager scenario the 12MP imager many of us are using would be represented by only a small portion of the imager area - very similar to the size of a little P&S imager. Do you think that small portion would return the same image quality as what we're using now.
Will some super-technology take care of all the physics issues?
Will some new unobtanium laden glass be used to create lenses that can resolve 100 million pixels accurately?
We are getting close to the point of diminishing returns.
Since you suppose that if we're going to have 100MP imagers why not 200MP or 500MP?
At what point will physics or just plain pointlessness stop the pixel count?
--
Happiness is a want... Contentment is a choice.
paulkienitz
Veteran Member
Physics limits the tradeoff between sharpness and noise. It also limits the accuracy with which a photon can be positioned. But we are still a long ways short of that second limit, and the first limit has nothing to do with the number of pixels, only with the total noise across the picture at a given exposure level. Making a picture sharper also makes it noisier, and you can always reduce the noise by giving up some sharpness, but unless you have the raw megapixels, you can't push the tradeoff the other way.
With today's megapixel counts, our pictures are pretty much locked into a pixel size determined by the camera, or to dramatically reduced sizes such as web jpegs. Once the sensor megapixel count goes into overkill, we'll be able to tune noise vs. sharpness in a pretty nearly continuous way instead of being kind of locked into a predefined point on the tradeoff curve. You might set your camera so that at ISO 100 you're producing 24 MP images, but at ISO 1600 the images are coming out at only 6 MP. Or whatever turns out to be the appropriate balance. The raw image might be 96 MP, but if you don't want to deal with that directly, you don't have to.
The only limit of physics is that there's no point in making photosites smaller than photons, so they shouldn't be smaller than about one micron. Other than that, there is nothing to lose by pinning down the location of each photon with more exactitude.
Keyboard shortcuts
- f
- Forum