 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Digital Photography / DSLR Cameras / February 2007Canon high ISO rumour - article
Bottom of page 4 of the article. Citing new technology from South Korean company Planet 82, Canon's Westfall says camera sensors could conceivably someday deliver image quality at ISO 6400 that approaches today's ISO 100. http://www.digitalphotopro.com/articles/2007/janfeb/imagesensors.php > Bottom of page 4 of the article. > [quoted text clipped - 3 lines] > > http://www.digitalphotopro.com/articles/2007/janfeb/imagesensors.php While I generally have great respect for Westfall, the comment seems pretty out of place. In the beginning of the article, they state quantum efficiencies (QE) range from 25 to 70% (correct). So how does one get a factor of 16 (hint: you can't go greater than 100% QE). If they mean small pixel size P&S cameras of today could someday perform like todays DSLRs by a modest increase in QE in combination with larger sensor and pixel size, then I'll agree. But no one has found a way to reduce noise beyond photon counting noise, which is a physical limit. Note in the article right after the Westfall statement, the Foveon guy didn't believe it either. More info on pixel effects size as discussed in the story: http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary Roger >> Bottom of page 4 of the article. >> [quoted text clipped - 21 lines] > > Roger You might see a demonstration on Miami CSI. >> Bottom of page 4 of the article. >> [quoted text clipped - 21 lines] > > Roger You don't need more width you need more volume of active silicon. You can do this by making the cells deeper and still have the same physical size. This technique is used in DRAM to make the cells hold more electrons. Any such technique could be applied to larger sensors too, so newer DSLRs will still have an advantage over compacts even though current DSLRs may produce worse picture quality than newer compacts. There has to be some driving force to get people to upgrade now that digital exceeds the resolution of film (and the lenses). ;-) dennis@home skrev: > You don't need more width you need more volume of active silicon. > You can do this by making the cells deeper and still have the same physical > size. That would increase the dynamic range, which in and by itself would be a very good thing, actually better than increasing sensitivity further IMO. (Phtotogs are ungrateful. They have already forgotten that DSLR:s already have given them much higher ISO sensitivity for colour photography than they could get with film. They just want to get even more.) But how coulld it increase sensitivity? If essentially all incident photons already are trapped until well saturation becomes limiting, how could improving well capacity lead to higher sensitivity? Jan B?hme >That would increase the dynamic range, which in and by itself would be >a very good thing, actually better than increasing sensitivity further >IMO. (Phtotogs are ungrateful. They have already forgotten that DSLR:s >already have given them much higher ISO sensitivity for colour >photography than they could get with film. They just want to get even >more.) Most output media can handle at most about 6 stops. Increasing dynamic range without reducing noise levels simply means lower ISO values, which require longer exposures, which are (except for landscapes) a problem.  SignatureThat was it. Done. The faulty Monk was turned out into the desert where it could believe what it liked, including the idea that it had been hard done by. It was allowed to keep its horse, since horses were so cheap to make. -- Douglas Adams in Dirk Gently's Holistic Detective Agency Most Canons that I see don't do well much above 400 iso. I try to keep my XT at 200 most of the time. I get no grain, no noise, that way. The new XTi is probably the same way. I'd give my eye teeth to get my hands on one of those bodies. I might just do it when funds are available. I'm selling some old equiment on ebay now, in hopes of raising enough to get the XTi body. Harry > >That would increase the dynamic range, which in and by itself would be > >a very good thing, actually better than increasing sensitivity further [quoted text clipped - 13 lines] > by. It was allowed to keep its horse, since horses were so cheap to make. > -- Douglas Adams in Dirk Gently's Holistic Detective Agency Today, HarryO50 made these interesting comments ... > Most Canons that I see don't do well much above 400 iso. I > try to keep my XT at 200 most of the time. I get no grain, no [quoted text clipped - 5 lines] > > Harry Harry, I have a year-old XT which works well, although my experience matches yours on noise above ISO 200. Besides the extra 2 MP, what specific features/improvements do you see in the XTi? >> In article >> <1168679686.966722.160310@38g2000cwa.googlegroups.com>, [quoted text clipped - 19 lines] >> -- Douglas Adams in Dirk Gently's Holistic Detective >> Agency SignatureHP, aka Jerry > Today, HarryO50 made these interesting comments ... > [quoted text clipped - 12 lines] > extra 2 MP, what specific features/improvements do you see in the > XTi? Good question. I upgraded from a Rebel to a Rebel XT for several reasons but my next upgrade will be to at least the 30D level, maybe the 5D level. I don't see much reason to move from the XT to the XTi. Greg Signature"All my time I spent in heaven Revelries of dance and wine Waking to the sound of laughter Up I'd rise and kiss the sky" - The Mekons Today, G.T. made these interesting comments ... >> Today, HarryO50 made these interesting comments ... >> [quoted text clipped - 17 lines] > 30D level, maybe the 5D level. I don't see much reason to > move from the XT to the XTi. I thought about upgrading this summer to a 30D but I very much like the smaller size and 7 oz. less weight of the XT, so when it breaks, I'll probably buy the XTi or whatever has replaced it. SignatureHP, aka Jerry >> Today, HarryO50 made these interesting comments ... >> [quoted text clipped - 18 lines] > > Greg I'd like that nice big screen on the XTi, but I have young eyes so I'll let all the old farts out there snap up the XTi's. ;) SignaturePhil Today, Phil, Non-Squid made these interesting comments ... >>> Harry, I have a year-old XT which works well, although my >>> experience matches yours on noise above ISO 200. Besides the [quoted text clipped - 11 lines] > eyes so I'll let all the old farts out there snap up the > XTi's. ;) Forgot about that one! I have old fogey eyes and even with my cheaters on it is hard to get a decent idea if my just-shot pic is or is not OK. Basically, all I can do is decide on the rough composure and the rough exposure, can't tell much else. For composure issues, I try to include 20-25% more of the area surrounding the main subject (if I can) and crop for best composure whilst post-processing. For exposure, a real challenge for me when using flash in museums, I set the brightness of the LCD down which better approximates what I see in the actual image. I've got about 10,000 actuations with my XT so it probably has considerable life left in it. I usually shoot at the 4 mega pixel size and resize down to about 1.5 MP. I rarely print these days and even when I do print to 8.5 x 11 on my Canon 6600, results on glossy paper are more than acceptable to me with the approx 130 PPI available. But, if I ever needed a really large print, I can always go back to the unedited image (I keep them all in a separate folder) and re-edit). Occasionally, I use 8 MP as a form of digital zoom which works very well in situations where I didn't bring along a long enough zoom lens and when I'm trying to create a much larger final image to include both photos and text, e.g., on the signage often found in front of exhibits in museums, so I can read the text. As I said earlier, I toyed with the 30D very briefly last summer but decided against it because it is much larger and over 7 oz. heavier. Yes, both can be an advantage if you have big hands and/or you want to use its higher inertia to control shake, but my energy and stamina is quite limited due to some health issues and I am close to 5 pounds with my longer Canon zoom and their 430 EX external flash, so every ounce counts for me. You and others made some very useful comments in this thread which I appreciate and will store in my memory bank. I just was asked to edumacate (that's a techie word!) a friend who couldn't decide between the new Nikon D80 and an XTi. I told him the same thing that I'd been told when shopping for either a D70 or XT last year: pick up each one and see which feels better and which has controls that seem more natural for you, and take a gander on how much glass costs and what focal lengths are available. He decided on the larger D80 because, to him, the controls were much more intuitive. Makes sense to me! SignatureHP, aka Jerry > Most Canons that I see don't do well much above 400 iso. I try to keep > my XT at 200 most of the time. I get no grain, no noise, that way. [quoted text clipped - 4 lines] > > Harry If you say this, then no camera does well. Canon DSLRs are regarded by astrophotographers as the top performing digital cameras (only cooled CCDs are better). The noise statistics I presented on the page: http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary shows that. In particular look at Figure 3 (read noise), Figure 6 (ISO at unity gain), and Figure 7 (low signal sensitivity). If you don't like the noise from a DSLR, you'll hate the noise from a small pixel P&S camera ;-). For example, compare the iso 1600 images from a DSLR to ISO 100 images from a P&S: http://www.clarkvision.com/imagedetail/does.pixel.size.matter2 Roger > Most Canons that I see don't do well much above 400 iso. I try to keep > my XT at 200 most of the time. I get no grain, no noise, that way. [quoted text clipped - 4 lines] > > Harry Oh!! Reference to two test cases: http://www.pbase.com/nchen711/image/72995412 http://www.pbase.com/nchen711/image/72653897 >>> That would increase the dynamic range, which in and by itself would be >>> a very good thing, actually better than increasing sensitivity further [quoted text clipped - 12 lines] >> by. It was allowed to keep its horse, since horses were so cheap to make. >> -- Douglas Adams in Dirk Gently's Holistic Detective Agency Philip Homburg skrev: > >That would increase the dynamic range, which in and by itself would be > >a very good thing, actually better than increasing sensitivity further [quoted text clipped - 4 lines] > > Most output media can handle at most about 6 stops. True. But this isn't all that relevant for how many stops of sensitivity it would be advantageous for the image recording to have. >Increasing dynamic > range without reducing noise levels simply means lower ISO values, > which require longer exposures, which are (except for landscapes) a > problem. In practically all the cases where you really need higher dynamic range, it _isn't_ a problem, though. These situations typically involve images which in part are lit with bright sunlight. And higher dynamic range without reducing noise levels doesn't _simply_ mean lower ISO values. It means lower ISO values with better shadow resolution. So you expose for your highlight, which should take quite a low ISO at normal shutter speed and aperture if you really need the extra dynamic range, and get your desired resolution out of the shadows. If we talke today's DR as natural, the best way of describing a higher dynamic range in ISO terms is that you can use a lower ISO for the highlights and a higher one for the shades, in the same exposure. Jan B?hme > dennis@home skrev: > [quoted text clipped - 12 lines] > photons already are trapped until well saturation becomes limiting, how > could improving well capacity lead to higher sensitivity? Jan, Your are correct, increasing the well size only increases the photons collected and thus the dynamic range. It does nothing to change sensitivity. Pixel size also determines true sensitivity. Larger pixels collect more photons and photons/pixel/time is sensitivity. Camera manufacturers define ISO to be a fraction of full well capacity, thus a sliding scale. We see the result in noise: small cameras have more noise; it is a fundamental physical limit (photon counting statistics). More info on this subject: Digital Cameras: Does Pixel Size Matter? Factors in Choosing a Digital Camera http://www.clarkvision.com/imagedetail/does.pixel.size.matter Digital Cameras: Does Pixel Size Matter? Part 2: Example Images using Different Pixel Sizes http://www.clarkvision.com/imagedetail/does.pixel.size.matter2 The f/ratio Myth and Digital Cameras http://www.clarkvision.com/photoinfo/f-ratio_myth Roger >Pixel size also determines true sensitivity. Larger >pixels collect more photons and photons/pixel/time >is sensitivity. Camera manufacturers define ISO to be a fraction >of full well capacity, thus a sliding scale. I don't think ISO is determined by just the full well capacity. It is combination of the full well capacity and the sensitivity. Higher sensitivity results in a higher ISO, but a higher full well capacity results in a lower ISO. SignatureThat was it. Done. The faulty Monk was turned out into the desert where it could believe what it liked, including the idea that it had been hard done by. It was allowed to keep its horse, since horses were so cheap to make. -- Douglas Adams in Dirk Gently's Holistic Detective Agency >>Pixel size also determines true sensitivity. Larger >>pixels collect more photons and photons/pixel/time [quoted text clipped - 6 lines] > Higher sensitivity results in a higher ISO, but a higher full well capacity > results in a lower ISO. I said that full well capacity does not change sensitivity. I said true sensitivity is photons/pixel/time. Regardless of well capacity and relative response of the ISO definition, true sensitivity is a function of pixel size and quantum efficiency, with larger formats giving better sensitivity, assuming the same megapixel count. The larger pixels simply collect more photons much like a larger bucket collects more rain drops than a smaller bucket. The larger bucket may not fill up any faster, but more you do get more rains drops. Same with pixels and photons. But noise in digital camera images is dominated by photon counting statistics, so larger pixels collectong more photons give better images. Roger > I said that full well capacity does not change sensitivity. Of course a change in the full well capacity changes the ISO sensitivity. If you double the full well capacity then you need twice the photon count to get it full. So you have half the ISO. > I said true sensitivity is photons/pixel/time. True _quality_ is dictated by time*photons/CFA-pixel. > Regardless of well capacity and relative response of > the ISO definition, true sensitivity is a function of pixel > size and quantum efficiency, How do you define "true sensitivity"? When you have two sensors with the same qe, same CFA-mask and the same full well capacity but one has 2x the CFA-pixel active area than the other, then in your opinion does the 2x provide a higher "true sensitivity"? When all other aspects are held the same, the sensor with the 2x CFA-pixel area of course has 2x the ISO but it means that only half of the dynamic range that the 2x CFA-pixel could provide is being used. This does not mean "true sensitivity" but false sensitivity. > with larger formats giving better sensitivity, > assuming the same megapixel count. Better _quality_. > The larger pixels simply collect more photons much like > a larger bucket collects more rain drops than a smaller > bucket. That is correct, you simply need a larger bucket to collect them all. > noise in digital camera images is dominated by photon > counting statistics, so larger pixels collecting more > photons give better images. Yes, better _quality_ images. Then, when that quality is high enough, it is possible to provide false sensitivity by reducing that high quality, by using only a small portion of the full dynamic range that the camera is capable to provide. Just like it is being done in all the cameras even if the true quality of these cameras do not warrant such reduction. My understanding is that the true sensitivity of sensor is a function of qe, fill factor and the transmittance of the CFA-mask only. Timo Autiokari >> I said that full well capacity does not change sensitivity. > >Of course a change in the full well capacity changes the ISO >sensitivity. If you double the full well capacity then you need twice >the photon count to get it full. So you have half the ISO. The number of photons entering a well of unit area in unit time is independent of the depth of the well. Imaging that 1000 photons pass through a 3 micron diameter opening in to a well in 1ms; how does making the well deeper increase the number of photons? The only ways of making the well more sensitive are to either make the area larger or improve the efficiency. With current efficiencies of ~30% there is limited potential in that route. Making the well deeper will improve the dynamic range thus allowing you to capture more shadow detail without blowing highlights, it can NOT capture more photons per unit time. SignatureIan G8ILZ There are always two people in every picture: the photographer and the viewer. ~Ansel Adams > The number of photons entering a well of unit area in > unit time is independent of the depth of the well. That is correct. > Making the well deeper will improve the dynamic range thus allowing > you to capture more shadow detail without blowing highlights, No, making the well deeper (increasing the full well capacity) and doing nothing else affects to nothing. But doing so provides you the possibility to collect a longer time. So when you double the full well capacity _and_ you also double the collecting time then, and only then, you will have improved dynamic range. Doubling the collecting time in the above example (while keeping the concept of camera unchanged) can only be achieved by introducing an other ISO sensitivity setting that is half of that ISO setting that was effective before the doubling of the full well capacity. Timo Autiokari >> The number of photons entering a well of unit area in >> unit time is independent of the depth of the well. [quoted text clipped - 6 lines] >No, making the well deeper (increasing the full well capacity) and >doing nothing else affects to nothing. It does allow you to capture more shadow detail, any half competent photograph knows the techniques for doing this, i.e. using a longer exposure, but it can not be done in a single exposure if the sensor does not support it. >But doing so provides you the possibility to collect a longer time. So >when you double the full well capacity _and_ you also double the >collecting time then, and only then, you will have improved dynamic >range. No, doubling the full well capacity doubles the available dynamic range, you don't have to use it to make the dynamic range of the sensor double. >Doubling the collecting time in the above example (while keeping the >concept of camera unchanged) can only be achieved by introducing an >other ISO sensitivity setting that is half of that ISO setting that was >effective before the doubling of the full well capacity. Wrong! If each photon produces one electron and one photon arrives every second making the well twice as deep will not produce two electrons every second (improbable unit values and efficiency used for simplicity). SignatureIan G8ILZ There are always two people in every picture: the photographer and the viewer. ~Ansel Adams >Wrong! I try to make this more simple for you by using an example: We have a camera that is photon limited (this means that the dominant noise is the photon shot noise) and it has such a sensor that the intrinsic ISO sensitivity (the lowest setting on the ISO dial) is 100. Now we change the sensor of this camera, the new sensor has full well capacity that is two times larger than what it was in the original sensor. No other changes are made at this time. The full well capacity means a capacitance, it is just a storage for the collected electrons. There are other ways to achieve larger capacity than to make the "well deeper" as you put it. Since we did nothing else to the camera (we only increased the storage space) it is behaving just similarly as the unchanged camera, except that the modified camera has an unnaturally large overexposure margin (1 f/stop larger margin). If you do not use this overexposure margin the camera does not give any more dynamic range than the unmodified one. You are simply utilizing only half of the storage capacity. In case you do use this margin you need to know to systematically overexpose by one f/stop, only then you will get better dynamic range (better by square root of 2 since this is a photon limited camera). But by this you have changed the basic concept or meaning of what is meant by camera, they do not by default require such a systematic overexposure. So, we now need to make another modification to the camera, we add an ISO 50 setting to the ISO dial. It affect (in the first place) to the exposure metering, in effect requiring double the exposure time than what the ISO 100 setting requires. Now we have a camera that behaves like any other camera, it has intrinsic ISO sensitivity of only 50 (so half of that what the unmodified camera has) but it does utilize the full storage capacity and therefore has 1.41 times the dynamic range of the unmodified camera. Timo Autiokari >>Wrong! > [quoted text clipped - 11 lines] >the collected electrons. There are other ways to achieve larger >capacity than to make the "well deeper" as you put it. I did not originate the term "well deeper", although it is a useful analogy for collecting "photo rain". >Since we did nothing else to the camera (we only increased the storage >space) it is behaving just similarly as the unchanged camera, except >that the modified camera has an unnaturally large overexposure margin >(1 f/stop larger margin). In other words an improved dynamic range. >If you do not use this overexposure margin the camera does not give any >more dynamic range than the unmodified one. You are simply utilizing >only half of the storage capacity. It still has an improved dynamic range, just because you photograph a uniformly lit grey card does not mean the sensor used has no dynamic range, it only means that you are not using all that it can provide. >In case you do use this margin you need to know to systematically >overexpose by one f/stop, only then you will get better dynamic range [quoted text clipped - 11 lines] >unmodified camera has) but it does utilize the full storage capacity >and therefore has 1.41 times the dynamic range of the unmodified camera. Agreed, it does have more dynamic range, but it is not more sensitive; the number of photons (~3) which will produce one electron has not changed by making the 'well deeper', all that has changed is that you can accumulate more electrons through a combination of longer exposure and wider aperture (or stronger illumination) before the sensor is overloaded. SignatureIan G8ILZ There are always two people in every picture: the photographer and the viewer. ~Ansel Adams > > The number of photons entering a well of unit area in > > unit time is independent of the depth of the well. [quoted text clipped - 6 lines] >No, making the well deeper (increasing the full well capacity) and doing >nothing else affects to nothing. Are you sure? It seems to me (and I'm certainly not an expert) that DR is a ratio of the difference between blowout (overflowing well) and black (empty well). The morw DR available, the more shades of brightness possible in the recorded image between blowout and black. If you increase the well depth, you increase the number of photons that can enter the well before overflowing (blowout) occurs. Thus, more DR. No? SignatureSenate Democrats proposed ethics reform legislation on Tuesday. It calls for lawmakers to pay the real cost of corporate jet flights and the full cost of skybox tickets for sporting events. If you want to know ahead of time what's going to happen to this bill, you simply need to watch the last five minutes of Old Yeller. > <timo.autiokari@aim-dtp.net> wrote: > >No, making the well deeper (increasing the full well capacity) > >and doing nothing else affects to nothing. > Are you sure? Sure I am sure, I said "...and doing nothing else". > If you increase the well depth, you increase the number of photons that can > enter the well before overflowing (blowout) occurs. Thus, more DR. No? No. In this situation you must _in addition_ increase the exposure accordingly (by using a slower shutter or a wider aperture -setting) in order to get that extra dynamic range into meaningful use. In other words: Into that use what we refer to when we speak about dynamic range of a camera. In other words: Into comparable use than how all the other cameras utilize the available dynamic range that the sensor provides. Having only an un-naturally large margin against over-exposure is waste of the dynamic range. We would very soon declace that such a camera has exposure metering fault. The only way for truely high-quality high-ISO performance is to have CFA-pixels with larger active area. Timo Autiokari timo.autiokari@aim-dtp.net skrev: > > If you increase the well depth, you increase the number of photons that can > > enter the well before overflowing (blowout) occurs. Thus, more DR. No? > > No. In this situation you must _in addition_ increase the exposure > accordingly (by using a slower shutter or a wider aperture -setting) in > order to get that extra dynamic range into meaningful use. Depends on how you exposed before, though. Your reasoning assumes that one always exposes to the right. While this is generally a good strategy for digital cameras, (and better, the more noise one expects) it might sometimes, particularly when the dynamic range of the incident light is large, be preferable to let the highlights blow, and expose for midtones or shadows instead. If you exposed for shadows or midtones when your DR was inadequate, then increased well capacity will indeed result in more DR without you doing enything else. Jan B?hme >> <timo.autiokari@aim-dtp.net> wrote: >> >No, making the well deeper (increasing the full well capacity) [quoted text clipped - 13 lines] >of a camera. In other words: Into comparable use than how all the other >cameras utilize the available dynamic range that the sensor provides. That doesn't seem right. If the well is overflowing (exhibiting blown out highlights), deepening the well will allow more range between black and blow-out. Increasing the exposure will let more photons in, but the original problem was too many photons for the well depth to handle. Increasing the exposure now only replicates the problem. >Having only an un-naturally large margin against over-exposure is waste >of the dynamic range. We would very soon declace that such a camera has >exposure metering fault. Having the "un-naturally large margin against over-exposure" *IS* increased DR in this case. Or else I'm not understanding what you're saying (entirely possible). Can you explain *where* I'm wrong? >The only way for truely high-quality high-ISO performance is to have >CFA-pixels with larger active area. > >Timo Autiokari SignatureNew Jersey threw out a one-hundred-fifty-year-old law Thursday that prevents idiots from voting. It threw the presidential race into chaos. As if Hillary Clinton wasn't facing enough problems, William Shatner is now leading in New Jersey. > Having the "un-naturally large margin against over-exposure" *IS* > increased DR in this case. No it is not since the increased exposure has to be offered by the additional setting on the ISO dial, so when the camera was natively ISO 100 it now has double the storage capacity for electrons but is natively an ISO 50 camera. It has 1.41 times the dynamic range _at_ that ISO 50 setting only. Only at the ISO 50 setting there are more detected photons. At ISO 100 setting it will have the very same dynamic range as it had before the full well capacity was increased because when use that ISO 50 camera at ISO 100 setting then the camera will amplify the sensor signal by 2x (so it will clip at half of the full well) and the situation is exactly the same as it was before the storage capacity was increased. In case of a photon limited camera (a camera where the photon shot noise is the dominant noise source) the dynamic range can be estimated to be Sqrt(signal) where signal is the amount of detected photons. The amount of detected photons can be affected by: 1) the conversion factor (quantum efficiency) but there there is a limit, it can only be 1 and in current cameras it is maybe 0.25. So about two stops could be had here at extremely high price. 2) the use of more transparent CFA mask. There is much more tight limit here, they do need to filter at least a little and they are already passing a lot. 3) increasing the active photon collecting surface area of each CFA pixels. The larger this area is the more photon it will collect. A change to the full well capacity does change the dynamic range of the camera BUT only at the expense of the native ISO sensitivity. So, a change to the full well capacity _alone_ does nothing. Because the camera has to behave like all the other cameras behave. I try again with a little bit more manly example. Say we have a natively ISO 100 camera. We then give it 16 times the full well capacity. Now it requires 16 times the exposure until the well becomes full. So where do we get such an increase _in such a manner_ that the camera still behaves (measures) like all the other cameras. There is no other possibility but to declare that this camera is now an ISO 6.25 camera. At ISO 6.25 this camera has as much as 4x the dynamic range, but the ISO 6.25 is somewhat restricting. We then could use this natively ISO 6.25 camera at ISO 100 setting but it means that the signal amplifiers are set to 16x so it is using only 1/16th of the full well capacity (clips everything that goes above), and the image quality and the dynamic range is exactly the same as it was before the increase of the full-well capacity. Since exactly the same amount of photons are available, they simply reside in a different kind of storage. Timo Autiokari > I try again with a little bit more manly example. Say we have a natively > ISO 100 camera. We then give it 16 times the full well capacity. Now it > requires 16 times the exposure until the well becomes full. So where do > we get such an increase _in such a manner_ that the camera still behaves > (measures) like all the other cameras. There is no other possibility but > to declare that this camera is now an ISO 6.25 camera. Looking over your contributions in this thread, I have to conclude you don't know what you are talking about here. ISO is computed by measuring the amount of light needed to obtain some (fixed, specified) output response of some media. Be it film or CCD, doesn't matter. Some fairly simple algebra can be brought to bear that shows ISO is a simple function of the signal-to-noise ratio. Dynamic range has nothing to do with it, since the physical response remains the same whether the buckets are itty bitty tiny ones ones or behemoth, monster gigantic pixels the size of an Olympic pool. See ISO Standard 12232:2006 for the nitty gritty. > ISO is computed by measuring the amount of light needed to > obtain some (fixed, specified) output response of some media. Sure. So in your opinion for the digital cameras the ISO can be computed by measuring the above, and this can be performed no matter in what setting the ISO dial of the camera happens to be??? > Some fairly simple algebra can be brought to bear that shows > ISO is a simple function of the signal-to-noise ratio. It is, and the signal-to-noise ratio is very closely related to dynamic range. In case of a photon limited camera they are the same. > Dynamic range has nothing to do with it, since the physical response remains > the same whether the buckets are itty bitty tiny ones ones or behemoth, The dynamic range of a digital camera is _defined_ by the full well capacity but _at the same time_ is the native ISO sensitivity of the camera defined by it. How can this be so difficult to understand for you? When you multiply the exposure time by 16 then you will collect 16 times the electrons in a given situation. You only need to increase the full well capacity by 16x to make it possible. This makes the native ISO sensitivity of the camera to be 4 stops lesser, but _at that native ISO sensitivity_ you will get 4x the dynamic range than what you had before the increase of the full well capacity. Timo Autiokari > > ISO is computed by measuring the amount of light needed to > > obtain some (fixed, specified) output response of some media. > > Sure. So in your opinion for the digital cameras the ISO can be > computed by measuring the above, and this can be performed no matter in > what setting the ISO dial of the camera happens to be??? What opinion? I gave you the reference to the standard for measuring the ISO. Did you read it? You might also want to look into Stroebel (et al) "Basic Photographic Materials and Processes" for their discussion on this subject. > > Some fairly simple algebra can be brought to bear that shows > > ISO is a simple function of the signal-to-noise ratio. > > It is, and the signal-to-noise ratio is very closely related to dynamic > range. In case of a photon limited camera they are the same. In a photon noise limited camera, the noise cf. "signal to noise" arises from the incoming light itself, not any property of the sensor. For example, the color of the camera, the gender of the user, the day of the week, or even (oh no!) it's dynamic range. > > Dynamic range has nothing to do with it, since the physical response remains > > the same whether the buckets are itty bitty tiny ones ones or behemoth, [quoted text clipped - 3 lines] > camera defined by it. How can this be so difficult to understand for > you? The DR of a pixel is the power ratio between the "mininum discernable signal" on the one side (ie, when you can finally distinguish between read noise and actual light) and the signal level when the sensor response begins to compress (ie, when the well is full, when the read-out amplifier starts going wonky, etc). Why you conflate this with the sensitivity remains a mystery. If you have two sensors, one which does the job of obtaining a fixed response with 1000 photons and another which needs 10000 photons, the latter will have an ISO rating 10x lower, as per the standard. And this will remain true regardless of the DR of the sensor. > What opinion? Your opinion/interpretation on how the standard defines the measurement method. > In a photon noise limited camera, the noise cf. "signal to noise" > arises from the incoming light itself, not any property of the sensor. That is correct (in my examples I did wrote that the camera was assumed to be photon limited). Now, camera- and sensor manufacturers want to ignore the photon shot noise completely because it is, as you correctly pointed out, the property of the light itself. But this is totally misleading to the user because the camera operates by the light so photon shot noise noise will be seen there in our images among all the other noises. E.g. we would not be very happy with a camera even if it had sensor signal-to-noise ratio of 1000000:1 but with such a very very tiny CFA-pixel area that gives 10:1 signal-to-noise ratio due to the photon shot noise. So, to speak about the dynamic range and about signal-to-noise ratio of a camera while leaving out the photon shot noise is not meaningful at all. Unless you are in the business of selling sensors or cameras. > The DR of a pixel is the power ratio between the "mininum discernable > signal" on the one side (ie, when you can finally distinguish between > read noise and actual light) and the signal level when the sensor > response begins to compress (ie, when the well is full, when the > read-out amplifier starts going wonky, etc). That is correct. I have not said otherwise. > Why you conflate this with the sensitivity remains a mystery. Do you refer to the sensitivity of the sensor chip or do you refer to the ISO sensitivity setting of a camera? > If you have two sensors, one which does the job of obtaining > a fixed response with 1000 photons and another which needs > 10000 photons, the latter will have an ISO rating 10x lower, > as per the standard. That is correct. And the camera that has the latter sensor has sqr(10) times the dynamic range. But if you increase the full well capacity of the former by 10x then that camera will have the same dynamic range as the latter, but 100x lower ISO rating. (Assuming photon limited devices). >And this will remain true regardless of the DR of the sensor. OK, it is clear now. You are speaking about the properties of the sensor in the manner how sensor manufactures specify them. I have been speaking about the properties of digital cameras from the standpoint of the user, who have absolutely no possibility to avoid the photon shot noise, it will be there in the images with all the other noises. Timo Autiokari > > In a photon noise limited camera, the noise cf. "signal to noise" > > arises from the incoming light itself, not any property of the sensor. > > That is correct (in my examples I did wrote that the camera was assumed > to be photon limited). This is incomplete. The signal-to-noise ratio is the square root of the sum of the squares of the photon noise plus read noise plus thermal noise. Depending on signal level, exposure time, and temperature of the sensor, any one may dominate. For high signals, short exposures (seconds or less), and room to low temperatures, photon noise dominates. > Now, camera- and sensor manufacturers want to ignore the photon shot > noise completely because it is, as you correctly pointed out, the > property of the light itself. But this is totally misleading to the user > because the camera operates by the light so photon shot noise noise will > be seen there in our images among all the other noises. If you read the sensor data sheets, you would see photon noise is not ignored. > E.g. we would not be very happy with a camera even if it had sensor > signal-to-noise ratio of 1000000:1 but with such a very very tiny [quoted text clipped - 12 lines] > > That is correct. I have not said otherwise. In previous threads you have argued against the minimum discernible signal and used numbers higher than is possible to get with film. Have you changed your ideas? > > Why you conflate this with the sensitivity remains a mystery. > [quoted text clipped - 8 lines] > That is correct. And the camera that has the latter sensor has sqr(10) > times the dynamic range. ISO can and is defined in two different ways: 1) relative to maximum signal (this is done in digital cameras), and 2) signal-to-noise ratio relative to film (the astronomical CCD manufacturers do this, claiming iso's something like 30,000 for similar chips as used in digital cameras. Digital cameras, from small pixel P&S cameras collect fewer photons at ISO 100 than do a large pixel DSLRs at ISO 100 because the ISO is defined as a function of full well capacity. Regarding two sensors with full capacity of 1000 photons and one with 10000 photons, assuming the read noise was the same, the dynamic range is 10x difference not sqr(10). Some sensor data: http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary > But if you increase the full well capacity of the former by 10x then > that camera will have the same dynamic range as the latter, but 100x > lower ISO rating. (Assuming photon limited devices). No, if you assume the definition used for digital cameras, and no change in sensor active area, then the iso scales linearly with full well capacity, not squared. > >And this will remain true regardless of the DR of the sensor. > [quoted text clipped - 5 lines] > photon shot noise, it will be there in the images with all the other > noises. With regard to your original post to my post on Jan 13 (I've been in Africa), here are my responses: > Roger N. Clark wrote: > [quoted text clipped - 3 lines] > sensitivity. If you double the full well capacity then you need twice > the photon count to get it full. So you have half the ISO. No. Sensitivity is about detection. You are confusing maximum signal with being able to collect photons. > > I said true sensitivity is photons/pixel/time. > > True _quality_ is dictated by time*photons/CFA-pixel. Quality and sensitivity are two different things. > > Regardless of well capacity and relative response of > > the ISO definition, true sensitivity is a function of pixel [quoted text clipped - 8 lines] > could provide is being used. This does not mean "true sensitivity" but > false sensitivity. Sensitivity concerns the number of photons collected relative to the noise in a given time interval: photons/pixel/time. For example, let's assume two sensors have the same read noise (e.g. about 4 electrons), and that you want to take a picture of stars or city lights at night. Even though the quantum efficiency of the two sensors may be identical, the larger pixel sensor will collect more photons so has more effective sensitivity. This is illustrated at: http://www.clarkvision.com/imagedetail/does.pixel.size.matter2 > > with larger formats giving better sensitivity, > > assuming the same megapixel count. > > Better _quality_. While better quality, the quality is directly related to collecting more photons, thus greater sensitivity. > > The larger pixels simply collect more photons much like > > a larger bucket collects more rain drops than a smaller [quoted text clipped - 11 lines] > is capable to provide. Just like it is being done in all the cameras > even if the true quality of these cameras do not warrant such reduction. It is not false sensitivity. True sensitivity is the number of photons collected divided by the noise. This effect is well illustrated on the above does.pixel.size.matter2 web page. > My understanding is that the true sensitivity of sensor is a function of > qe, fill factor and the transmittance of the CFA-mask only. The fundamental sensitivity is number of photons you collect divided by the noise. While things like quantum efficiency are fundamental limits to the efficiency of conversion, sensitivity to detect a given signal is different, and includes everything needed to make a measurement, including the optical system. For example, a larger telescope detects fainter stars because it collects more photons. Digital cameras at a given f/ratio have sensitivities proportional to the active area of the pixel. Try a 10 second exposure at f/4 50mm (35mm) equivalent exposure at ISO 200 on a P&S camera and on a larger pixel DSLR. Which shows more stars? The camera that shows more stars is more sensitive because it collects more photons. Hint: the its the larger pixel camera. Roger >dennis@home skrev: >> You don't need more width you need more volume of active silicon. >> You can do this by making the cells deeper and still have the same >> physical >> size. > But how coulld it increase sensitivity? If essentially all incident > photons already are trapped until well saturation becomes limiting, how > could improving well capacity lead to higher sensitivity? Sorry. I answered the wrong question (again!, it must be age). It should increase the dynamic range. I suppose if you want to do something really radical you could have an individual photon counter per pixel. Then you would have sensitivity and dynamic range. It would bring digital cameras into the digital age too. > You don't need more width you need more volume of active silicon. > You can do this by making the cells deeper and still have the same physical > size. > This technique is used in DRAM to make the cells hold more electrons. Unfortunately, DRAMs don't record photons not coming in at a vertical angle very well ... > even though current DSLRs may > produce worse picture quality than newer compacts. Picture quality is not just (absense of) noise. Show me one 8MPix camera where the optics can really deliver the resolution the sensor records. -Wolfgang Wolfgang Weisselberg skrev: > Picture quality is not just (absense of) noise. Show me one > 8MPix camera where the optics can really deliver the resolution > the sensor records. OTOH, the resolution the sensor records is significantly finer than the one we get to use. Anti-aliasing and Bayer interpolation see to that. Jan B?hme > Wolfgang Weisselberg skrev: >> Picture quality is not just (absense of) noise. Show me one >> 8MPix camera where the optics can really deliver the resolution >> the sensor records. > OTOH, the resolution the sensor records is significantly finer than the > one we get to use. Anti-aliasing and Bayer interpolation see to that. On the gripping hand, even with AA and interpolation, even high quality SLR lenses don't manage to match 8 MPix resolution, except with near optimal apertures and in the center. Just look at http://www.photozone.de/8Reviews/index.html and compare the MTF at different apertures ... -Wolfgang On 4 Feb, 14:14, Wolfgang Weisselberg <ozcvgt...@sneakemail.com> wrote: > > Wolfgang Weisselberg skrev: > >> Picture quality is not just (absense of) noise. Show me one > >> 8MPix camera where the optics can really deliver the resolution > >> the sensor records. > > OTOH, the resolution the sensor records is significantly finer than the > > one we get to use. Anti-aliasing and Bayer interpolation see to that. > On the gripping hand, even with AA and interpolation, even > high quality SLR lenses don't manage to match 8 MPix > resolution, except with near optimal apertures and in the > center. True. But the exceptions you give are rather important, as we do tend to use lenses at optimal apertures - lighting permitting - and generally, even if we follow the rule of thirds, have the principal focus of interest far closer to the center of the image than to its edge. Jan Böhme >Show me one 8MPix camera where the optics >can really deliver the resolution the >sensor records. With all dSLR cameras the resolving power of the camera (that is the result of anti-aliasing filter + sensor + CFA interpolation) is _far_ lesser than the resolving power of any lens. What do you refer to with "the resolution the sensor records"? Timo Autiokari > >Show me one 8MPix camera where the optics > >can really deliver the resolution the > >sensor records. > With all dSLR cameras the resolving power of the camera (that is the > result of anti-aliasing filter + sensor + CFA interpolation) is _far_ > lesser than the resolving power of any lens. Interesting claim, can you provide some sort of proof? Why does the f/1.4 50mm (Canon) suffer from noticable softness when wide open, even in the center, compared to using it, at, say f/2.8? How comes that with almost all lenses the borders are weaker than the center when wide open or near wide open? Obviously, these lenses does *not* (always) resolve as much as the sensor can resolve, for the same sensor can give better results with the same lens with stopping down, and, yes, people do see differences in resolving power when using different lenses with the same sensors. Which clearly shows that the lenses do not outperform the sensor. > What do you refer to with "the resolution the sensor records"? Raw data from the sensor, with Bayer interpolation added as needed. -Wolfgang >Bottom of page 4 of the article. > [quoted text clipped - 3 lines] > >http://www.digitalphotopro.com/articles/2007/janfeb/imagesensors.php Note the word "conceivably." This means that he can think of a time when this might happen. He's not saying this is in the works, but that it might be sometime down the road. I can conceive a production, 4-passenger car that will get 200MPG on unleaded regular gas that will provide all of the protection of a full-size car, the trunk space of a 75 Coupe De Ville, with adaptive LCD panels to make it look like different cars on different days. No one has such a car in the works, but I can conceive it. SignatureSenate Democrats proposed ethics reform legislation on Tuesday. It calls for lawmakers to pay the real cost of corporate jet flights and the full cost of skybox tickets for sporting events. If you want to know ahead of time what's going to happen to this bill, you simply need to watch the last five minutes of Old Yeller. This ties in with Canon's comments. Planet82: Imaging Breakthrough? Planet82 are a Korean company based in Seoul, who are developing nanotechnology based imaging systems, building on research from the Korea Electronics Technology Institute. Last year they demonstrated the first 'Single Carrier Modulation Photo Detector' (SMPD) image sensor. A nanometer (nm) is one billionth of a meter, around the size of a few atoms or a small molecule. Nanotechnology is technology that works with materials at approaching this size - around 100nm. By designing structures on this scale, Planet82 claim they can make sensors that respond to just a few photons, with an amplification that makes them up to 2000 times more sensitive than the photodiodes used in current sensors. The SMPD sensors can be smaller than current sensors and use less power, but most importantly will be able to produce images at much lower light levels - perhaps down to 0.1 lux, making them rather better than human vision. Sensitive enough to take pictures by moonlight, or in a room lit by a single candle. (Typical domestic lighting levels are around 50-400 lux.) At next week's International Consumer Electronics Show at Las Vegas, Planet 82 will have a 2Mp SMPD color sensor. The first applications for this are likely to be in CCTV systems, but camera phones and digital cameras are likely to follow. However it may be some time before we can throw away our flashes and take photographs in near darkness. It may well take some years to scale up the SMPD sensor to a suitable size for high-resolution imaging, where current sensors already severely test available lenses and the wavelength of light imposes its own limits. Breakthroughs in imaging technology sometimes take quite a while to materialize in usable products. The Foveon 3 color chip which promised so much when I wrote about it in 2002 as a Digital Breakthrough has not yet quite lived up to its promise, although cameras such as the Sigma SD14 and Sigma DP1 expected shortly may put this right. Planet82's SMPD sensor, assuming it delivers as promised, may revolutionize the way we take pictures away from daylight in a few years time. Friday January 5, 2007 > This ties in with Canon's comments. > [quoted text clipped - 11 lines] > to 2000 times more sensitive than the photodiodes used in current > sensors. But the problem isn't to make small enough detectors, it's that if you make them smaller then the number of incident photons will be smaller (for a given flux). The point is that, even if you're able to detect every single photon (which is, in principle, ok), if the total number should have been eg 5, then, due to the random nature of their arrival, you'll get large random fluctuations in this number: noise. Really. This is ignoring issues of maximum possible charge stored per unit area (related to well depth) and other things, it assumes perfect detectors, wells that never overflow etc. It's hard to believe, I know, but all the people who keep going on about this here didn't just make it up to show off. Books have been written about the random fluctuations arising from the particulate nature of matter (and light). It's not a technological limitation, it really is there, and some thought will tell you that the reason "we" are the size we are and not 10^6 times smaller is basically this (ie the fact that stuff is made up of particles). Of course, all the problems can be overcome (assuming perfect everything, no overflow etc) by simply allowing more photons to fall on the detector (ie longer exposure times). But this isn't really what you're saying. According to RichA <rander3127@gmail.com>: [ ... ] > A nanometer (nm) is one billionth of a meter This depends on whose "billion" you are using. In the US, a billion is 10^9, but in the UK, a billion is 10^12. While I was born in the USA, and have lived here for most of my life, I feel that the UK system is more consistent. You obviously used the US billion (10^9). I'll skip the rest of this. Enjoy, DoN. SignatureEmail: <dnichols@d-and-d.com> | Voice (all times): (703) 938-4564 (too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html --- Black Holes are where God is dividing by zero --- |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.