Luminous Landscape Forum
Site & Board Matters => About This Site => Topic started by: samirkharusi on November 13, 2006, 01:36:37 am
-
Perhaps a useful future article on LL. I have yet to find a satisfactory explanation of what the ISO setting does in a DSLR. I am therefore requesting Michael (or anyone else) to talk to his Dr Know or somebody who actually KNOWS what is going on so that we can all get an authoritative briefing on what that ISO setting actually does. Let me explain why I find the topic rather muddy:
Astro CCDs normally have only one speed or ISO. Many output 16 bits. If you have each photosite have a fill of 65,000 electrons (or less) then your analog to digital digitiser has ample quantification "resolution" with 16 bits, 1 to 65536. It is rumoured that DSLR sensors also have only one fixed sensitivity. I do not know whether this is true or false. Assuming it is true then it is further speculated that if your output is only 12 bits (1 to 4096) then obviously you want to spread the most important part of your image signal to within 12 bits. Presto! Low signal/high ISO, use the 12 bits to examine the region closest to nil electrons. High signal/low ISO, use the 12 bits to examine the upper end of the electron well fill. I.e. the ISO button changes the information flow window only, not "sensitivity". I am not totally convinced that this is true, because no matter how you post process, it seems impossible to take an image at, say, 1/1000sec ISO 100 (16x under-exposed) and make it even approach an image with the same exposure (this time the "correct" exposure) at ISO 1600. I would have expected that if it were only a matter of data flow then one can achieve a reasonably close result in post processing, with some extra noise naturally. Corollary to all this? When DSLRs get 16-bit A/D converters (like some MF backs already?) then there will be no need for any ISO setting at all (like for astro CCDs). And Michael will no longer have to quibble about the placing of the ISO dial/display on future DSLRs. All the discussions I have come across to date are conjecture and guessing with some reverse engineering. Somebody somewhere must know what is going on. Anybody?
-
I always assumed that at higher ISOs the camera put more current through the CCD making it more sensitive but at the trade-off of noise.
-
I think ISO is made up of two components - AtoD gain, and software gain.
Graeme
-
Adjusting the ISO simply adjusts a voltage gain applied to the captured analog signal before it is digitized (and therefore quantized) by the A-to-D converter. In other words, the gain is applied in the analog domain at a very early step of image processing. This is an important technical difference from applying an exposure boost later on in the RAW conversion software.
Of course, any unwanted parts of the signal (e.g., noise) will also get boosted by the voltage gain.
Eric
-
Adjusting the ISO simply adjusts a voltage gain applied to the captured analog signal before it is digitized (and therefore quantized) by the A-to-D converter. In other words, the gain is applied in the analog domain at a very early step of image processing. This is an important technical difference from applying an exposure boost later on in the RAW conversion software.
[a href=\"index.php?act=findpost&pid=84954\"][{POST_SNAPBACK}][/a]
Mostly right, but I believe that some of the higher ISO settings on DSLR's go beyond the limits of this analog gain, and are done in the digital stage, reproducing what can be done later in raw conversion or post-processing. This is perhaps the case with ISO speed settings distinguished as "boost" or "HI", the ones above about 800 or 1600 in many DSLR's.
Indeed, I have read that in raw files from the Olympus E-1, there is no difference between ISO 800, 1600 and 3200 when one uses equal shutter speed and aperture, indicating that the latter are "digital pushes" of ISO 800 sensor and A/D convertor output. (1600 and 3200 are the optional "boost" settings on the E-1, disabled by default.)
-
Mostly right, but I believe that some of the higher ISO settings on DSLR's go beyond the limits of this analog gain, and are done in the digital stage, reproducing what can be done later in raw conversion or post-processing. This is perhaps the case with ISO speed settings distinguished as "boost" or "HI", the ones above about 800 or 1600 in many DSLR's.
Yes, I forgot to mention that in most of the current DSLRs, there is different behavior for the extreme ISO modes that are usually turned off (or made inaccessible) by default. Examples include the "H" ISO mode for Canon DSLRs which is ~ISO 3200, and the "L" ISO mode which is ~ISO 50) behave differently.
Essentially, the "H" mode applies the same voltage gain as with ISO 1600, but after the signal has been digitized, the digital values are then doubled, thereby increasing by 1 stop. This means that the effective dynamic range is smaller by 1 stop in the shadows. Since this "doubling step" can be performed during RAW conversion, I see no advantage to using "H" mode for people who shoot RAW except perhaps a shorter workflow.
Conversely, the "L" mode applies the same gain as with ISO 100, but after the signal has been digitized, the digital values are then halved, thereby decreasing by 1 stop. This means that the effective dynamic range is smaller by 1 stop in the highlights.
Eric
-
Corollary to all this? When DSLRs get 16-bit A/D converters (like some MF backs already?) then there will be no need for any ISO setting at all (like for astro CCDs). And Michael will no longer have to quibble about the placing of the ISO dial/display on future DSLRs.
Not having ISO settings means a big change in method. How do you get a review image of the correct brightness? Or JPEGs saved on the card?
Sure, it would be nice if there was exactly the same digitization for all "ISOs", but you still need to hold onto the concept for default rendering purposes, and also if you want to use something like Av-priority mode and instruct the camera to use a fast shutter speed (as in setting a higher ISO).
All the discussions I have come across to date are conjecture and guessing with some reverse engineering. Somebody somewhere must know what is going on. Anybody?
[a href=\"index.php?act=findpost&pid=84893\"][{POST_SNAPBACK}][/a]
You're not going to get the kind of response you want. Reverse engineering is the only way you can figure this stuff out. Companies do not disclose their methods in any kind of detail.
When ISO 400 has only 18% more readout noise than ISO 100, then you know that they received unique amplifications. When ISO 3200 is all even numbers in the RAW data, and the noise is exactly the same as an ISO 1600 image doubled, you know that some simple math is used for 3200, and it has 1600-level amplification. When every 5th value is missing in a RAW histogram of an ISO 125 image, and the readout noise is exactly 1.25x as high, you know that the camera metered for ISO 125, amplified for 100, and multiplied the RAW data by 5/4, clipping away 1/5 of the linear dynamic range.
Canon's spokesperson has already said that they don't disclose such details about their cameras (a good way to avoid critical discussion). I would doubt that any of the other companies are willing to share their "secrets", either. Even if they made a statement, it may be incorrect. Reverse engineering is the only way to know these things.
-
Astro CCDs normally have only one speed or ISO. Many output 16 bits.
... Corollary to all this? When DSLRs get 16-bit A/D converters (like some MF backs already?) then there will be no need for any ISO setting at all (like for astro CCDs).
[a href=\"index.php?act=findpost&pid=84893\"][{POST_SNAPBACK}][/a]
The pre-amplifcation gain helps to reduce the effect of any noise introduced after that stage up to and including in A/D conversion. With some CMOS sensors, I believe that the pre-amplification is done very early, at the photosite itself, and so there might be distinct advantages in terms of reducing the effects of noise arising between the photosite and A/D conversion.
Doe anyone know the relative magnitude of the various noise sources that contribute to "read-out" noise including analog noise in the A/D convertor?
At what frame rates are astro CCD's read out? Higher read-out rates, as needed for standard digital cameras, increase read-out noise. Asrto CCD's might have negligible read noise. In fact I have read of other scientific CCD setups where read noise is less than one electron RMS. In that scenario, variable pre-amp gain (variable ISO speed) is probably irrelevant.
-
In fact I have read of other scientific CCD setups where read noise is less than one electron RMS. In that scenario, variable pre-amp gain (variable ISO speed) is probably irrelevant.
[a href=\"index.php?act=findpost&pid=85022\"][{POST_SNAPBACK}][/a]
One electron's worth of read noise from 65K wells with 16-bit digitization? I'll take it on my next DSLR. Can I get 5fps?
-
One electron's worth of read noise from 65K wells with 16-bit digitization? I'll take it on my next DSLR.
[a href=\"index.php?act=findpost&pid=85040\"][{POST_SNAPBACK}][/a]
Just pack your camera in dry ice
- DL
-
Does anyone know the relative magnitude of the various noise sources that contribute to "read-out" noise including analog noise in the A/D convertor?
At what frame rates are astro CCD's read out? Higher read-out rates, as needed for standard digital cameras, increase read-out noise. Asrto CCD's might have negligible read noise. In fact I have read of other scientific CCD setups where read noise is less than one electron RMS. In that scenario, variable pre-amp gain (variable ISO speed) is probably irrelevant.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=85022\")
Yes, higher read-out rate increase noise. I do know that astronomical CCDs are read out at low rates and this is one reason why they must be cooled. Otherwise dark noise at the low readout rates would degrade the noise unacceptably during the long time interval of the read process.
To get faster frame rates I believe the Nikon D200 has 4 readout channels, each reading its own portion of the sensor. If these are not balanced properly this leads to banding and you have to send in the camera for adjustment.
This article explains some of the physics:
[a href=\"http://www.mssl.ucl.ac.uk/www_detector/optheory/darkcurrent.html#Readout%20noise]http://www.mssl.ucl.ac.uk/www_detector/opt...Readout%20noise[/url]
-
Not a reply to anybody in particular, but simply some extra info. A Canon 20D has been measured to have about half the Read Noise of many astroCCDs, despite slower reading in the astroCCDs. A 20D also seems to have its lowest Read Noise at ISO 1600, slightly lower compared to other ISO settings. Nobody is more obsessive regarding noise than astrophotographers, and most have come to a consensus that it does not matter much what ISO you use. An hour integration time, made of, say, 12x5minute exposures, will end up with the same Signal to Noise Ratio in the final processed image regardless as to whether it was shot at ISO 1600, 800, 400 or 200 (keeping away from the extreme ISO settings). Previewing an image when we have no ISO settings at all is not a major issue. Just an auto-stretch (contrast/brightness) setting in the camera's firmware will give a reasonable view of what was captured. All astroCCD software that I have used offers this facility.
We need Dr Know...
-
Not a reply to anybody in particular, but simply some extra info. A Canon 20D has been measured to have about half the Read Noise of many astroCCDs, despite slower reading in the astroCCDs. A 20D also seems to have its lowest Read Noise at ISO 1600, slightly lower compared to other ISO settings.
ISO 1600 has the lowest readout noise in units of electrons, or relative to absolute exposure. The difference is not small, however, with other ISOs. The absolute read noise is 7x as high at ISO 100, 3.7x at ISO 200, 2.1x at ISO 400, 1.4x at ISO 800.
Nobody is more obsessive regarding noise than astrophotographers, and most have come to a consensus that it does not matter much what ISO you use. An hour integration time, made of, say, 12x5minute exposures, will end up with the same Signal to Noise Ratio in the final processed image regardless as to whether it was shot at ISO 1600, 800, 400 or 200 (keeping away from the extreme ISO settings).
That contradicts what was already discussed, at least as far as the Canon is concerned. Your statement is kind of vague, though. Parts of the recipe are missing.
We need Dr Know...
[a href=\"index.php?act=findpost&pid=85077\"][{POST_SNAPBACK}][/a]
I still don't understand what you think is such a mystery.
-
ISO 1600 has the lowest readout noise in units of electrons, or relative to absolute exposure. The difference is not small, however, with other ISOs. The absolute read noise is 7x as high at ISO 100, 3.7x at ISO 200, 2.1x at ISO 400, 1.4x at ISO 800.
That contradicts what was already discussed, at least as far as the Canon is concerned. Your statement is kind of vague, though. Parts of the recipe are missing.
I still don't understand what you think is such a mystery.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=85133\")
In astrophototography one has to separate the noises one can do something about from those beyond the user's control. Read-Noise in a 20D is smallest at ISO 1600, but the noise in a final, fully processed image, usually made up of a large stack of frames, depends on many other factors, the major one being photon-statistical (or quantum statistical) noise in the skyfog (light pollution) itself. There is also thermal noise, etc. Current understanding is such that one can actually put up equations that desribe much of this stuff very satisfactorily, except that ISO setting (!), because most of our understanding as to what it does is guesswork. Hence we need Dr Know to say what the basic principles in setting ISO are, e.g. on-chip, off-chip, analog domain or after A/D conversion, prior to reduction to 12-bits or after, etc. I expect that all camera manufacturers use the same basic principles and then tweak here and there. The tweaks would presumably be the proprietary areas, the principles probably common knowledge in the industry. There is no ISO term we can confidently put into these equations to optimise the SNR in the final image. There is also no agreed test procedure/protocol (e.g. like for measuring Read-Noise) that leads to an answer like "use ISO 400 for this much skyfog and ISO 1600 for that much skyfog. Here is a simplistic note I have written up on the minimal length of sub-exposures in a stack:
[a href=\"http://www.samirkharusi.net/sub-exposures.html]http://www.samirkharusi.net/sub-exposures.html[/url]
and here are some of the equations I was alluding to (for astroCCDs and hence no mention of ISO):
http://www.starrywonders.com/snr.html (http://www.starrywonders.com/snr.html)
For the daytime shooter I think it would be of great interest to learn if ISO setting would be going the way of the dodo in a couple of years (when 16-bit A/D converters come on stream) or not. Personally I find it all very intriguing, much more than pixel peeping (which I also do occasionally).
-
For the daytime shooter I think it would be of great interest to learn if ISO setting would be going the way of the dodo in a couple of years (when 16-bit A/D converters come on stream) or not. Personally I find it all very intriguing, much more than pixel peeping (which I also do occasionally).
Considering that certain digital back manufacturers claim 16-bit A/D, and still offer several ISO settings, it doesn't appear to be going the way of the dodo because of that.
And the ISO settings are useful for other things related to exposure, such as achieving a longer or shorter shutter time without changing DoF.
But perhaps that setting will go away with the dodo, if ISO 50 through 6400 are equivalent in terms of final noise; then I could just set the exposure in terms of shutter speed and f-stop, and adjust the exposure compensation afterwards to get the "right" exposure.
It's a bit dizzying to think of what photography would be like without choosing a "film speed", so maybe I'm way off the mark here.
-
It's a bit dizzying to think of what photography would be like without choosing a "film speed", so maybe I'm way off the mark here.
[a href=\"index.php?act=findpost&pid=85391\"][{POST_SNAPBACK}][/a]
My vision of the Camera of the Future is one where you choose Av and Tv values in manual mode, and the feedback the camera gives is a logarithmic histogram based on high-res metering, with some kind of flickering or speckling of the lower ranges so that you recognize when S/N is being compromised, so you can let more light in if you don't need the high f-stop or fast shutter speed. I couldn't say "good-bye" fast enough to discreet one-stop ISOs with high read-out noise.
For auto-exposure, user-programmable EV to AV/Tv mappings based on the lens mounted would be used.
-
It's a bit dizzying to think of what photography would be like without choosing a "film speed", so maybe I'm way off the mark here.
[a href=\"index.php?act=findpost&pid=85391\"][{POST_SNAPBACK}][/a]
I've imagined it so many times, I'm ready *now*. All I need is a high-res metering that gives a logarithmic histogram in the edge of the viewfinder to give me an idea of S/N, and I'm set to go.
-
A 20D also seems to have its lowest Read Noise at ISO 1600, slightly lower compared to other ISO settings.
[a href=\"index.php?act=findpost&pid=85077\"][{POST_SNAPBACK}][/a]
Does this mean that by applying that extra amplification (to "ISO 1600") early, right at the photo-sites, the relative impact of subsequent noise in the read-out and A/D conversion is minimized?
Naively, this makes sense, and suggests something like Samir's strategy, or at least my understanding of it, which is as follows.
1) amplify a lot and early, say to "ISO 1600", to minimize the effect of noise introduced later in the process (read-out, A/D).
2) use an A/D converter than can handle the resulting high maximum signal strengths, for example in situations where one is using "exposure index ISO 100", and thus amplifying to a level of four stops of "over-exposure". That is, add four bits or so of highlight headroom to the A/D convertor. So yes, go to 16 bits A/D or even more, in that direction.
Then selected ISO speed affects shutter speed/aperture choice in AE modes, but goes into the raw file as a mere suggestion for default raw conversion, like WB, sharpening, and such. With manual shutter speed and aperture setting, the light meter reading could be used to record a "suggested" ISO speed in the raw file.
My question is how large a signal-to-noise ratio an A/D convertor for a portable camera can have these days, or in the foreseeable future. That limits its useful bit range, regardless of what the spec. sheets say. No point the A/D being able to handle very high input levels if the noise added by the A/D convertor is much more that 1/65536=1/2^16 of that maximum signal, because then the least significant of those 16 bits are all A/D noise, no signal, and "16-bit" becomes purely a paper spec.
Aside: with current SLR sensors, the only possible benefit of 16 is covering up to four stops of exposure level error, which is related to what I (and Samir?) propose as a deliberate strategy. With correct exposure, abut 12 bits is enough to record the S/N ratio of any current SLR sensor itself, and trends at the high end (Kodak and Dalsa) seem to be holding the line or even dropping S/N ratio for the sake of increased pixel count. (Leaving the options of binning or down-sampling when the user prefers extremes of high DR at less high resolution.)
-
Hence we need Dr Know to say what the basic principles in setting ISO are, e.g. on-chip, off-chip, analog domain or after A/D conversion, prior to reduction to 12-bits or after, etc. I expect that all camera manufacturers use the same basic principles and then tweak here and there
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=85355\")
I too wish that Dr. Know would enter into the discussion and enlighten us, but in the meantime, here are a couple of posts by Dr. Clark (an astrophysicist and avid photographer) that deal with read noise, dynamic range, and ISO.
He states that when unity gain is reached (the ISO where one photo-electron results in one increment in the ADU [analog to digital unit or raw pixel value], there is no reason to increase the ISO on the camera any further since it will only decrease the dynamic range without recording additional shadow detail. Unity ISO is shown for various cameras and ranges up to about 1600.
[a href=\"http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html]http://www.clarkvision.com/imagedetail/eva...-1d2/index.html[/url]
http://www.clarkvision.com/imagedetail/doe...tter/index.html (http://www.clarkvision.com/imagedetail/does.pixel.size.matter/index.html)
-
I too wish that Dr. Know would enter into the discussion and enlighten us, but in the meantime, here are a couple of posts by Dr. Clark (an astrophysicist and avid photographer) that deal with read noise, dynamic range, and ISO.
He states that when unity gain is reached (the ISO where one photo-electron results in one increment in the ADU [analog to digital unit or raw pixel value], there is no reason to increase the ISO on the camera any further since it will only decrease the dynamic range without recording additional shadow detail. Unity ISO is shown for various cameras and ranges up to about 1600.
[a href=\"index.php?act=findpost&pid=85473\"][{POST_SNAPBACK}][/a]
I've never agreed with that conclusion of his. He ignores the issue of 1-dimensional noise - line banding. This is *the* demon of Canon high ISO shadows; it is much more visibly potent than random noise alone, in many images. Banding is still greatly improving from ISO 800 to 1600 (relative to absolute signal) with recent Canons, even when random noise is starting to taper off, so just to have less banding, it seems worthwhile to go to ISO 3200.
He draws conclusions about ISO 3200 based on what the camera actually does; not on what real ISO 3200 amplification would be. His camera just doubles the data that is digitized for ISO 1600 but metered for 3200 - and then draws conclusions based on the noise of ISO 3200, which doesn't really exist on his camera, the way ISOs 100 through 1600 do.
The idea that digitization ends at unity (1 ADU = 1 electron) is not logical to me at all. I believe 3 or 4 ADUs per electron will have slightly less posterized data. Don't forget, the ADC is not counting electrons, it is digitizing a buffered, amplified fascimile of the charges in the wells. Oversampling the electrons allows more accurate counting, as the noise can't flip the result as far in either direction.
Here's the read noises of the 20D blackframe at various ISOs; the yellow line is the total noise (divided by 10 to fit on the graph with the other two lines). ISO 100 is straightforward, and the other ISOs are divided by 2, 4, 8, 16, and 32 to show their strength relative to absolute exposure (values proportional to electrons, IOW):
(http://www.pbase.com/jps_photo/image/65737967.jpg)
As you can see, the values for the artificial ISO 3200 are abruptly flattened, more than the trend suggested by the other ISOs. This is hardly true of the total read noise (which is the only thing he and most others measure), but very true of the blue line which traces horizontal banding noise, *the* demon of Canon high ISO shadows (despite being only 10% as strong as the total noise). I believe that a gain-based ISO 3200, and maybe even 6400, would improve the pattern noise situation, despite the likelyhood that no significant gains in total noise may occur.
-
I do not understand the specifics of the "banding" issue, but overall, it does make sense that as far as random noise is concerned, once the signal is amplified to the point that A/D quantization error is of order of one photo-electron or less, more amplification will not help. And if anything, I would expect progress in A/D convertors to reduce the increment in input voltage level corresponding to one step in output level, meaning that pre-amplification to rather less than ISO 800 is likely to be enough. Even counting to within two photo-electrons is probably completely adequate in practice, because photon shot noise will be significantly more than that at any pixel that got enough light to be worth printing as anything other than pure black.
More to the point, DSLR sensor well capacities seem unlikely to go much beyond the current upper limits, so should stay under 2^17=131,072, so a 17-bit A/D convertor could do "exact" counting of every photo-electron up to a well capacity that large, so one pre-amplification level would then cover all cases. Who cares if it is ISO 100 or 3200 or in between? If 16 bit RAW is needed, maybe the decision as to whether to delete one bit from the top or bottom could be done after examining such 17-bit raw data.
And A/D convertors of more than 17 bits exist, for audio recording at least.
-
And A/D convertors of more than 17 bits exist, for audio recording at least.
One of the great revolutions in A/D conversion is the 1-bit converter (using a D flip-flop frequency delta-sigma modulator) (http://heim.ifi.uio.no/~matsh/ds/DrMats.pdf), rather than the clunky and complex 24-bit (or what-have-you) Nyquist-rate converters. Bonus: it's simpler!
-
I've never agreed with that conclusion of his. He ignores the issue of 1-dimensional noise - line banding. This is *the* demon of Canon high ISO shadows; it is much more visibly potent than random noise alone, in many images.
[a href=\"index.php?act=findpost&pid=85542\"][{POST_SNAPBACK}][/a]
John,
Thanks for the information. Very well reasoned as usual. Both you and Roger are considerably more advanced than I in these matters, so I can't really judge. Roger does determine noise by subtracting two frames taken together under identical conditions, so any repeatable pattern noise is not accounted for. However, his approach does eliminate problems from nonuniform targets and illumination when one merely takes the standard deviation of pixel values for a uniform patch. How do you measure noise?
The idea that digitization ends at unity (1 ADU = 1 electron) is not logical to me at all. I believe 3 or 4 ADUs per electron will have slightly less posterized data. Don't forget, the ADC is not counting electrons, it is digitizing a buffered, amplified fascimile of the charges in the wells. Oversampling the electrons allows more accurate counting, as the noise can't flip the result as far in either direction.
[a href=\"index.php?act=findpost&pid=85542\"][{POST_SNAPBACK}][/a]
That makes sense, but there must be a point of diminishing returns when the ADU number is greater than the number of electrons captured. One is after all measuring voltage, which is related to the charge and capacitance of the pixel well together with the analog amplification. However, the voltage should increase in discrete steps with each electron. Some degree of oversampling makes sense, just like carrying an extra decimal place in calculations so as to avoid rounding errors. If you are using a ruler to measure an object to the nearest millimeter, it doesn't help to calculate with six decimal places.
Bill
-
The idea that digitization ends at unity (1 ADU = 1 electron) is not logical to me at all. I believe 3 or 4 ADUs per electron will have slightly less posterized data.
[a href=\"index.php?act=findpost&pid=85542\"][{POST_SNAPBACK}][/a]
But isn't it true that at any photosite receiving enough light to deserve being rendered as anything except pure black, you need enough photo-electrons that photon shot noise is well over one electron? (RMS photon shot noise is about the square root of the number of photo-electrons.) If so, then A/D quantization errors of one electron will be buried in that other noise, except at pixels that fall below the black-point.
(It is different in technical applications like astro-photography, where they care about counting every photon, even in very dark parts of the image where S/N ratios are horribly low.)
-
But isn't it true that at any photosite receiving enough light to deserve being rendered as anything except pure black, you need enough photo-electrons that photon shot noise is well over one electron?
No. The significance of unity is an illusion. It is a simple, appealing concept, with no basis in reality. It's just the crossing point of the input and output of the square root function. Hypothetically, if there were no noise other than shot noise, and you had a 1 gigapixel sensor, you could get perfectly usable images when only 20% of the sites received one single photon each. You could probably recognize a face in an image where only 1% of the pixels received photons. You could use a 1 gigadot printer, or downsample the image.
The real-world problem is readout noise, which varies RAW values by much more than one electron. This changes the 1:1 problem a bit, but even here, there is nothing significant about the 1:1 point. You can truncate the last few bits in an ISO 1600 image, and not notice the difference at all in a normally rendered print; counting individual photons is fuzzy well above the unity level; the more amplification or bit depth you use, the more accurate the signal and noise are rendered, and there is a small, but real, improvement in IQ. Nothing magical should happen at the 1:1 level. It really has no significance. Read noise does NOT occur in discrete units of sensor well electrons.
(RMS photon shot noise is about the square root of the number of photo-electrons.) If so, then A/D quantization errors of one electron will be buried in that other noise, except at pixels that fall below the black-point.
An image is a community of pixels. The better the noise is sampled, the closer you get to a true image. There is no benefit in ignoring things because they are too noisy. "Buried in noise" is a metaphor, but isn't really what is happening. Noise never masks anything by 100%. Readout noise is not a low wall behind which you can see no signal; it is an uneven ground which alters the height of signal stakes that are resting on it; the more accurately that the uneven ground is represented, the less falsehood is in the heights of the tops of the stakes.
(It is different in technical applications like astro-photography, where they care about counting every photon, even in very dark parts of the image where S/N ratios are horribly low.)
[a href=\"index.php?act=findpost&pid=86256\"][{POST_SNAPBACK}][/a]
Shooting birds in the forest with long, slow telephoto/TC stacks when the foliage is at peak growth is not much different.
Someone gave me two RAW files from a Konica-Minolta DSLR, a K7 I believe, with the same under-exposed image at both ISO 1600 and 3200, with the same manual exposure. Even with more noise, relative to the Canons, I could still see clearer shadows in the 3200, which, unlike the Canons, is analog in ISO nature. the 1600 had slightly higher chromatic noise, and more distinct line-banding.
-
The significance of unity is an illusion.
[a href=\"index.php?act=findpost&pid=86298\"][{POST_SNAPBACK}][/a]
I agree that unity is an illusion, or at least an extreme that is not a natural practical choice. At most, it limits the worst case quantization error to 1e, and no matter how much more one amplifies, the error will in fact not get any less, because the original error sources all come in quanta of one or more electrons, even if subsequent amplification and charge-to-voltage conversion renders these discrete errors into variations in a continuous voltage signal.
But I think, contrary to you, that counting accurate to one electron is more than enough precision, rather than not enough, by considering the minimum magnitude of the errors already present in the signal reaching the A/D convertor.
The real-world problem is readout noise, which varies RAW values by much more than one electron.[a href=\"index.php?act=findpost&pid=86298\"][{POST_SNAPBACK}][/a]
I agree that the noise levels arriving at the A/D converter are already "much more than one electron", and that is precisely why I believe that an A/D convertor that measures accurate to a single photo-electron ("unity conversion") is already _more_ precision than has any practical value.
The real world problem is the total of read-out noise plus other noise present in the signal reaching the A/D convertor, including photon shot noise. You might be right that in practice, read-out noise dominates, but I use photon shot noise only because fundamental physics sets a known, absolute lower limit on noise levels in the signal reaching the AD convertor. That is, photon noise gives an upper limit on the "precision" of the signal that arrives at the A/D convertor.
No matter how low other noise sources get, the minimum possible noise level (in RMS variation in the count of photo-electrons in the electron well) will be the square root of the number of photo-electrons of signal, the contribution of photon shot noise. This can somewhat arbitrarily be divided into two cases:
Case 1: photo-electron signal less than about 16e [higher, like 25e, might be a more reasonable cut-off?]
Case 2: photo-electron signal of about 16e or more [or 25e?]
Case 1: Under 16e is 12 or more stops below the maximum possible signal in larger current DSLR photosites, so 8 or 9 stops below mid-tone levels at ISO 100 and 4 or 5 stops below mid-tones at ISO 1600. Total black on a print comes about four stops below mid-tones, so these levels are well into total black on a straight print even from an ISO 1600 image. Moreover, the S/N ratio can be at most the square root of the electron count, so at most 4:1 with this few photo-electrons, and this is so miserably low that I cannot imagine any interest in rendering such extremely dark and noisy pixels as anything except pure black.
That is, the black point would almost surely be set above the 16e level, so that the signal from a pixel receiving such a low illumination level would be transformed to level zero, eliminating any visible print noise in any part of the image receiving such low illumination.
[25e would move minimum S/N ratio up to 5:1, still I suspect uselessly low for artistic photography, and so destined to be zeroed out by the black point. Kodak has suggested 10:1 as the minimum S/N ratio needed for "acceptable" image quality, along with 40:1 for "excellent".]
Case 2: Signal of 16e or more has photon shot noise of at least 4e RMS, and thus thus the total "analog domain noise" (photon noise , dark current noise , read-out noise, pre-amplifier noise, and any others I have missed) in the signal entering the A/D convertor already has this much "error". An experimentalist would probably tell you that there is no point in measuring a quantity down to less than 1/4 of the error already present: your already get more than enough precision, "two bits" to spare, if the A/D convertor counts accurate to the level of one photo-electron, as with "unity conversion".
To quantify this, adding 1e RMS of quantization noise to 16e RMS of "analog stage" noise give a total of sqrt(17)=4.12e RMS, an increase of 0.12e, or 3% in total noise. I doubt that this change, in the deep shadows, would be even slightly visible.
And as the signal increases, the effect of an additional 1e RMS quantization error becomes less. For example, setting the black point a bit higher by requiring at least 5:1 S/N ratio at "non-black pixels" increases the minimum electron count to 25e, minimum "analog phase noise" to 25e RMS, and then adding 1e RMS to 25e RMS only increases the total noise by 0.1e, less than 2%.
If any posterization in extremely dark parts of the image still arises at the transition from black level to lowest non-zero level (which I doubt), it can be eliminated by interpolation of additional finer levels of near black, a kind of "dithering". There is no need to use the illusory precision of even more "accurate" A/D conversion to produce such levels: that would simply be using random noise in the signal as a form of random dithering.
The weakest point in this argument is the somewhat arbitrary choice of S/N ratio thresholds like 4:1, so I am open to evidence and arguments that very dark pixels with S/N ratios less than 4:1 are worth printing as anything other than pure black. But for now I am skeptical, given expert opinions like the guideline of a 10:1 minimum S/N ratio stated in a Kodak technical paper.
-
John,
Thanks for the information. Very well reasoned as usual. Both you and Roger are considerably more advanced than I in these matters, so I can't really judge. Roger does determine noise by subtracting two frames taken together under identical conditions, so any repeatable pattern noise is not accounted for. However, his approach does eliminate problems from nonuniform targets and illumination when one merely takes the standard deviation of pixel values for a uniform patch. How do you measure noise?
For blackframes, I just look at the standard deviation of the blackframe, and the blackframe with averaged lines, horizontal and vertical, for banding noise. The values agree in IRIS and Photoshop, as long as you view the greyscale in PS as color (PS gets confused in greyscale mode, and does not report values properly). IRIS makes it very easy - you just load the RAW file, and type the stat command, or highlight a rectangle and right click on "Statisitics". For the banding, I measure in PS, after doing an image resizing to one pixel wide or high, and back to full size, with bicubic.
For actually exposed areas, I measure noise by shooting a color checker card out of focus, to get rid of any luminance detail in the card itself, then splitting the RAW into the 4 color channels (2 greens), and measuring the the mean and standard deviations of the flat areas within the squares (careful not to include bokeh from the black edges or their shadows, or dust spots if present).
That makes sense, but there must be a point of diminishing returns when the ADU number is greater than the number of electrons captured.
There are certainly diminishing returns, but there is no significance to the unity point per se.
One is after all measuring voltage, which is related to the charge and capacitance of the pixel well together with the analog amplification. However, the voltage should increase in discrete steps with each electron. Some degree of oversampling makes sense, just like carrying an extra decimal place in calculations so as to avoid rounding errors. If you are using a ruler to measure an object to the nearest millimeter, it doesn't help to calculate with six decimal places.
[a href=\"index.php?act=findpost&pid=85877\"][{POST_SNAPBACK}][/a]
Who says we're only interested in the nearest millimeter, though. We don't look at images to count photons. We look at images to see what is suppposed to be in them, and even "diminished returns" are better than not having those diminished returns, when you really need to see what's down there in those shadows, and as little of the artifacts as possible.