 |
 | 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 / April 2006Canon to address "blown highlights" in 2007?
According to Pop Photography, Canon is going to bring out something in 2007 to fix the inability of sensors to deal with wide dynamic ranges. They (Pop Photography) said they'd say what it is next month. I'm wondering if they've figured out a way to regulate the sensitivity of individual pixels in an array by varying something (signal amplification?). >> According to Pop Photography, Canon is going to bring out something >> in 2007 to fix the >> inability of sensors to deal with wide dynamic ranges. They (Pop >> Photography) said they'd say what it is next month. I'm wondering if >> they've figured out a way to regulate the sensitivity of individual >> pixels in an array by varying something (signal amplification?). You mean like the Sony sensor in Fuji DSLRs?  Signaturewww.photosbydouglas.com www.weprint2canvas.com If you really must write,use my name at an above domain. >>> According to Pop Photography, Canon is going to bring out something >>> in 2007 to fix the [quoted text clipped - 4 lines] > > You mean like the Sony sensor in Fuji DSLRs? fuji dslrs use fuji sensors. >>> RichA wrote: >>>>> According to Pop Photography, Canon is going to bring out [quoted text clipped - 8 lines] >> >> fuji dslrs use fuji sensors. Yes, like the one they get from Sony? Signaturewww.photosbydouglas.com www.weprint2canvas.com If you really must write,use my name at an above domain. >>>> RichA wrote: >>>>>> According to Pop Photography, Canon is going to bring out [quoted text clipped - 10 lines] > > Yes, like the one they get from Sony? only fuji produce the paired pixel sensor for extra dynamic range. It is patented. I suppose they may pay sony to produce them to spec but i doubt it. However KM and Nikon do use Sony sensors. >> "D-Mac" <none@these.groups> wrote in message >>>>>> [quoted text clipped - 7 lines] >> It is patented. I suppose they may pay sony to produce them to spec >> but i doubt it. However KM and Nikon do use Sony sensors. I thought ony Wedding Photographers were pedantic. I'll rephrase then: Like to one they get Sony to make? Satisfied now? Signaturewww.photosbydouglas.com www.weprint2canvas.com If you really must write,use my name at an above domain. >>> "D-Mac" <none@these.groups> wrote in message >>>>>>> [quoted text clipped - 11 lines] > I'll rephrase then: Like to one they get Sony to make? > Satisfied now? nope >>>>>RichA wrote: >>>>> [quoted text clipped - 15 lines] > patented. I suppose they may pay sony to produce them to spec but i doubt > it. However KM and Nikon do use Sony sensors. Canon DSLRs already have very large dynamic ranges, limited by 12-bit A-to_D converters. The 1D mark II, for example, has measured dynamic range (max signal/read noise) > 3000:1, or 11.6 stops. See: http://www.clarkvision.com/imagedetail/evaluation-1d2 This is already greater than film (even print film): http://www.clarkvision.com/imagedetail/dynamicrange2 I've not seen any data on the Fuji sensors. Can they actually do better than 11.6 stops? Roger >>>>>>RichA wrote: >>>>>> [quoted text clipped - 29 lines] > >Roger They had an example of film versus digital DSLRs. The negative film was exposed 3 stops over and was able to produce a good image, it simply required more light through a dense negative. The digital image was blown completely. Maybe it's time for them to move to 16 bit A to Ds? -Rich > The digital image was blown completely. Maybe it's time for > them to move to 16 bit A to Ds? Maybe it's time for you to grow a brain? If the pixel well is saturated, even a 1000 bit ADC isn't going to help. >> The digital image was blown completely. Maybe it's time for >> them to move to 16 bit A to Ds? > >Maybe it's time for you to grow a brain? If the pixel well is >saturated, even a 1000 bit ADC isn't going to help. True, but if the RAW data had more dynamic range, you could afford to expose with more headroom as a default. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >> The digital image was blown completely. Maybe it's time for >> them to move to 16 bit A to Ds? > >Maybe it's time for you to grow a brain? If the pixel well is >saturated, even a 1000 bit ADC isn't going to help. Astronomical CCD cameras constantly saturate pixels (owing to bright point sources being images) and yet use 16 bit D/As. -Rich http://www.sbig.com/sbwhtmls/online.htm Astrophotographers take note: A 35mm format CCD with 11 million pixels. Another revolution in astrophotography is about to take place. If you have been waiting for a wide field CCD with the same coverage as 35mm film, here it is. Only this one is cooled, linear, self-guiding, 16 bits and optimized for high sensitivity and low noise performance in low light astronomical applications. Rich still hasn't grown a brain: > >> The digital image was blown completely. Maybe it's time for > >> them to move to 16 bit A to Ds? [quoted text clipped - 4 lines] > Astronomical CCD cameras constantly saturate pixels (owing to bright > point sources being images) and yet use 16 bit D/As. ... and they still can't do a damn about the saturation. Rich, you are a negative intelligence: http://users.pandora.be/vdmoortel/dirk/Physics/Gems/NegInt.html The more you learn, the more stupid you become. >Rich still hasn't grown a brain: > [quoted text clipped - 13 lines] > >The more you learn, the more stupid you become. More bits = greater dynamic range. -Rich >More bits = greater dynamic range. Greater *potential* dynamic range. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>More bits = greater dynamic range. > >Greater *potential* dynamic range. So no gains would be obtained by improving on the A/D converters? I can't answer this for consumer cameras, but I do know that image quality overall has been helped vastly by shifting to higher resolution A/D converters in the scientific use of CCDs and many of them only work in the monochrome area, not colour. -Rich >>>More bits = greater dynamic range. >> >>Greater *potential* dynamic range. > >So no gains would be obtained by improving on the >A/D converters? That's a different thing. There is more to improving on the converters, perhaps, than increasing the bit depth. Readout of ISO 100 shadows is currently in a poor state, compared to what sensors are capable of. Part of the problem is bit depth; part is not. >I can't answer this for consumer cameras, >but I do know that image quality overall has been helped >vastly by shifting to higher resolution A/D converters in the >scientific use of CCDs and many of them only work in the monochrome >area, not colour. Sure, but if you don't have clean read-out, it is of more limited value. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > That's a different thing. There is more to improving on the converters, > perhaps, than increasing the bit depth. Readout of ISO 100 shadows is > currently in a poor state, compared to what sensors are capable of. Part > of the problem is bit depth; part is not. What is the problem, then? >> That's a different thing. There is more to improving on the converters, >> perhaps, than increasing the bit depth. Readout of ISO 100 shadows is >> currently in a poor state, compared to what sensors are capable of. Part >> of the problem is bit depth; part is not. > >What is the problem, then? Read noise SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) > Read noise But why would this be a worse problem at ISO 100 than at eg 1600? >> Read noise > >But why would this be a worse problem at ISO 100 than at eg 1600? In absolute terms it isn't, but compared to other noise sources it is. By read noise, I am referring to all noise in the analogue chain after detection of the signal. The lower gain of ISO100 reduces the effect of noise sources in front of the gain stage, making the noise after the gain stage more significant. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>Read noise > > But why would this be a worse problem at ISO 100 than at eg 1600? See table 1 at: http://www.clarkvision.com/imagedetail/evaluation-1d2 With a read noise of only 3.9 electrons at ISO 1600, the read noise begins to be posterized ISO 400 and lower on the Canon 1D Mark II. At ISO 100, the 1DII records a maximum signal of about 53,000 electrons. With 12-bit digitization, 1-bit = 53000/4095 = 13 electrons, which is exactly what is measured. There is noise in the A-D conversion so that scrambles the posterization, and there are likely some other small noise sources too. Roger > With a read noise of only 3.9 electrons at ISO 1600, the read > noise begins to be posterized ISO 400 and lower on the Canon [quoted text clipped - 4 lines] > posterization, and there are likely some other small noise > sources too. Ah, now I see. Thanks. >> Read noise > >But why would this be a worse problem at ISO 100 than at eg 1600? Relative to the maximum RAW signal strength at that particular ISO, the highest ISO usually has the most noise. The variation, however, does not match the amplification level, so the absolute_signal to noise ratio is highest for the higher ISOs. Photon noise is fixed, in this regard, and dark current noise is determined by temperature and exposure time, so the differences are all readout-related (and therefore, the lower the ISO, the larger percentage of the noise that is read-related). Here's a visual demonstration of relative shadow noise at various ISOs and exposure indices: http://www.pbase.com/image/58032350/original Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >> That's a different thing. There is more to improving on the converters, >> perhaps, than increasing the bit depth. Readout of ISO 100 shadows is >> currently in a poor state, compared to what sensors are capable of. Part >> of the problem is bit depth; part is not. >What is the problem, then? Sloppy readout. If the 8 most significant bits of an ISO 1600 readout were used to replace the 8 least significant bits of an ISO 100 readout (both cover the same range of sensor charges), the image would be much cleaner. The deepest shadows of ISO 100 are clearly posterized, but that is not their only problem, as equally quantizing an ISO 1600 image taken with the same absolute exposure looks much better. There is *tremendous* room for improvement in low-ISO performance, with current sensor wells. Reading them cleanly is either difficult, expensive, or on a list of reasons for future upgrades. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > Sloppy readout. If the 8 most significant bits of an ISO 1600 readout > were used to replace the 8 least significant bits of an ISO 100 readout [quoted text clipped - 6 lines] > on a list of reasons for future upgrades. > -- Well, that is what you also said earlier (ie that there is room for improvement), and I'll take your word for it. But my question isn't "what happens", but "why is X the problem" (where X=sloppy readout). You say that reading them is/may be difficult; why do you think so? I've thought a bit about this, and I have to admit I can't see why this should be so (given that it can be done for ISO 1600, say), except if the problem is posterization (which you say it isn't, and I'll buy it). So, what exactly is the problem? After all, at any ISO, the signal is initially the same, and is just amplified more before being digitized at higher sensitivities. If we disregard bit depth problems when unamplified signals are digitized, I can't see any other quality-degrading operation happening. >Well, that is what you also said earlier (ie that there is room for >improvement), and I'll take your word for it. But my question isn't >"what happens", but "why is X the problem" (where X=sloppy readout). >You say that reading them is/may be difficult; why do you think so? I can only guess. It may be cost-prohibitive, but by how much? Enough that the cameras would have to cost $500 more? Or just $50 more, and they need to hear a lot of complaints for future models to get cleaner and/or deeper readout. It may only appear in the > US $4000 cameras at first. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >Canon DSLRs already have very large dynamic ranges, limited by >12-bit A-to_D converters. The 1D mark II, for example, has [quoted text clipped - 4 lines] >This is already greater than film (even print film): >http://www.clarkvision.com/imagedetail/dynamicrange2 But its a "hard" clip with the digital sensor and a "soft" clip on film. Its a bit like the old comparison of transistor and valve audio amplifiers - transistors gave better linearity and distortion than valves when operating well within their range, but valves sounded better because they clipped softly compared to the awful distortion produced by hard clipping transistors when driven by the occasional transient beyond their nominal range. A non-linear top end response to a digital sensor would eliminate the problem of blown highlights due to unexpected specular reflections. Say, linear response to 90% of the existing capacity and logarithmic response in the top 10%. It wouldn't take any additional photodiodes, as Fuji use, just the response of the CMOS buffer circuit, and flexibility of the pixel design is one of the key benefits of CMOS over CCDs. In fact, can't this be done with just two additional diodes, one of which is a zener, in a feedback loop around the amplifier in the pixel? SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >In fact, can't this be done with just two additional diodes, one of >which is a zener, in a feedback loop around the amplifier in the pixel? I would think that the big issues that get in the way are ones of calibrating the pixels to all have the same response in the RAW data. It's hard enough getting a few thousand rows and columns of single linear pixel readouts to have the same gain and blackpoint. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > >Canon DSLRs already have very large dynamic ranges, limited by > >12-bit A-to_D converters. The 1D mark II, for example, has [quoted text clipped - 22 lines] > In fact, can't this be done with just two additional diodes, one of > which is a zener, in a feedback loop around the amplifier in the pixel? As I understand the problem, highlight clipping is due to one or other of two effects. One, full wells. Given the dynamic range of good sensors, only an exposure long or large enough to cause well saturation can cause clipping by this means; and it should only be possible if shooting at the camera's lowest ISO. Shooting at higher ISO numbers results in fewer photons being captured because of the faster exposure, so full-well clipping at higher ISO's is a remote possibility, only perhaps with extremely bright specular reflections or gross overexposure. Two, the amplifier gain employed by the processor can cause clipping in jpeg images with only slight overexposure, as all highlights that exceed 255,255,255 in any amplified channel will be clipped, even if the sensel well was not full. RAW data processed in good converter software allows the effective gain to be adjusted to place the histogram where the operator wants it, so fixed-gain clipping problems are avoided. I have wondered if jpeg processing in-camera would benefit from some sort of automatic gain control, based on how much lift is needed to place the highlights just short of clipping, rather than fixed gain controlled by ISO. Maybe the manufacturers do this already since it seems obvious, but there could be drawbacks like a processing penalty and space for the additional firmware, and a fixed safety margin might be preferable - in which case the safety margin should be controllable by the photog. Come to think of it, it is already controllable by exposure compensation. However, a dynamic control based on the individual exposure would be better. Colin D. >As I understand the problem, highlight clipping is due to one or other >of two effects. One, full wells. Given the dynamic range of good [quoted text clipped - 5 lines] >perhaps with extremely bright specular reflections or gross >overexposure. I don't understand something here, likely because of a stupidity surge of the sort that often affects me. But why would shooting at higher ISO numbers result in fewer photons being captured because of faster exposure? If I shoot outdoors at ISO 100 at say 1/100 and f/5.6 and then go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why would there be fewer photons? [N.B. The numbers are made up and not meant to reflect any particular shooting situation.] ---- Paul J. Gans >I don't understand something here, likely because of a stupidity >surge of the sort that often affects me. But why would shooting [quoted text clipped - 4 lines] >go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why >would there be fewer photons? Doh! Because you have gone indoors there *are* less photons! Unless you compensate by changing the exposure, which in your example you haven't, the you get less photons reaching the sensor. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>I don't understand something here, likely because of a stupidity >>surge of the sort that often affects me. But why would shooting [quoted text clipped - 6 lines] >> >Doh! Because you have gone indoors there *are* less photons! >Unless you compensate by changing the exposure, which in your example >you haven't, the you get less photons reaching the sensor. Ok. I think I get you. You are saying that because the light level is lower I will get fewer photons at 1/100 and f/5.6 indoors than outdoors. And I make up for it by upping the amplifier gain (which is what the ISO setting does, I guess.) So the statement up above is backwards. It isn't the higher ISO numbers that give fewer photons, it is the available light that gives fewer photons. The increased ISO setting simply compensates for the fewer photons. By the way, the backwards saying above was quoted by me but written by the OP. Reading it confused me. ----- Paul J. Gans >>As I understand the problem, highlight clipping is due to one or other >>of two effects. One, full wells. Given the dynamic range of good [quoted text clipped - 14 lines] > go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why > would there be fewer photons? In your hypothetical, the shutter was open the same amount of time but in lower light so fewer photons but with manual settings in the same lighting you'd get more photons by increasing ISO. High ISO highlight clipping is usually due to the interpolation (A/D converter?) of non-full wells but it's certainly possible to fill the wells *and* clip early due to high ISO interpolation. Here's an idea, have an option for bracketing ISO: pull the charge off the sensor once but convert it twice at high & low ISO then manually merge those exposures later. There seems to be more potential in the sensor than a single conversion can extract. >In your hypothetical, the shutter was open the same amount of time but >in lower light so fewer photons but with manual settings in the same >lighting you'd get more photons by increasing ISO. Changing ISO does *not* change the number of photons. Changing ISO merely changes the *responsivity* of the camera - slope of the digital output versus photon input. There is simply no way to "get more photons by increasing ISO", all that does is give a bigger digital number for the same number of photons. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >> In your hypothetical, the shutter was open the same amount of time but >> in lower light so fewer photons but with manual settings in the same >> lighting you'd get more photons by increasing ISO. > > Changing ISO does *not* change the number of photons. Not in itself but combined with 'normal' adjustments in shutter speed raising ISO results in the same number of photons... or more if it was impossible to handhold & you were underexposing at low ISO to reduce shake. If there aren't highlights at risk of blowing, then raising ISO can let you get more photons *and* read the shadows more cleanly. That is JPS's contradictory and confusing approach. Unfortunately much of my shooting is outdoors with some sky (or at night with bright lamps) so I can't overexpose unless it was two bracketed reads. I suspect it may be physically impossible to read the data twice though. Perhaps there is a way to meter each pixel as it's read and apply a curve at that time? > Changing ISO > merely changes the *responsivity* of the camera - slope of the digital > output versus photon input. There is simply no way to "get more photons > by increasing ISO", all that does is give a bigger digital number for > the same number of photons. > then raising ISO > can let you get more photons *and* read the shadows more cleanly. That > is JPS's contradictory and confusing approach. Raising ISO does not get more photons. The quantum efficiency of the detectors in a digital camera is fixed. The lens aperture, focal length, pixel size, and exposure time fixes the number of photons you collect for a given scene intensity, regardless of ISO. Raising ISO simply digitizes a different (lower) part of the signal stored by the sensor. Roger >> then raising ISO can let you get more photons *and* read the shadows >> more cleanly. That is JPS's contradictory and confusing approach. > > Raising ISO does not get more photons. Oops, you are correct... I was running in circles... > The quantum efficiency of the detectors > in a digital camera is fixed. The lens aperture, focal length, pixel size, > and exposure time fixes the number of photons you collect for a given scene > intensity, regardless of ISO. Raising ISO simply digitizes a different > (lower) part of the signal stored by the sensor. >>> In your hypothetical, the shutter was open the same amount of time >>>but in lower light so fewer photons but with manual settings in the [quoted text clipped - 5 lines] >impossible to handhold & you were underexposing at low ISO to reduce >shake. No, changing ISO does *not* change the number of photons. Ever!! The number of photons is determined by the scene, your lens aperture and the time that you expose the sensor for. ISO is completely independent of this, and is a measure of the "gain" of the camera in a photon to digital number conversion. >> Changing ISO merely changes the *responsivity* of the camera - slope >>of the digital output versus photon input. There is simply no way to >>"get more photons by increasing ISO", all that does is give a bigger >>digital number for the same number of photons. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>In your hypothetical, the shutter was open the same amount of time but >>in lower light so fewer photons but with manual settings in the same >>lighting you'd get more photons by increasing ISO. >Changing ISO does *not* change the number of photons. Changing ISO >merely changes the *responsivity* of the camera - slope of the digital >output versus photon input. There is simply no way to "get more photons >by increasing ISO", all that does is give a bigger digital number for >the same number of photons. Thanks. I've got that figured out now. I was confused by an early poster who said that increasing the ISO number decreased the number of photons. That's said backwards as I now understand it. With the same shutter speed and f/stop you get fewer photons in a darker environment. *THEN* upping the ISo number will result in greater amplification of what you've got or, as you say, a bigger digital number for the same number of photons. ---- Paul J. Gans > But why would shooting > at higher ISO numbers result in fewer photons being captured because [quoted text clipped - 3 lines] > go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why > would there be fewer photons? Usually, indoors there is less light, so 1/100 gets fewer photons than 1/100 sunny outdoors when there is more light. ISO gain means fewer photons are required to record an exposure. For example, a Canon 20D needs about 50,000 photons at ISO 100 for maximum signal (4095 in the 12-bit A-D converter). At ISO 1600, it needs only about 3100 photons to reach the same 4095 level. Roger >If I shoot outdoors at ISO 100 at say 1/100 and f/5.6 and then >go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why >would there be fewer photons? Assuming equally exposed results, they both colect about the same amount of photons. The difference is, the range of sensor charges digitized into meaningful numbers is smaller for the higher ISO. Sensor saturation is only a potential problem at the lowest ISO (or two, if the lowest is a "fake" low ISO with reduced highlights). At higher ISOs, the RAW data will saturate at a much lower level than the sensor would. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>If I shoot outdoors at ISO 100 at say 1/100 and f/5.6 and then >>go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why >>would there be fewer photons? >Assuming equally exposed results, they both colect about the same amount >of photons. I must have edited half that sentence and not the other half. I replied right after waking up from an unproductive nap. What I meant to say is that if the scenes were of equal brightness, the same aperture and shutter speed would collect the same approximate number of photons. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>>If I shoot outdoors at ISO 100 at say 1/100 and f/5.6 and then >>>go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why >>>would there be fewer photons? >>Assuming equally exposed results, they both colect about the same amount >>of photons. >I must have edited half that sentence and not the other half. I replied >right after waking up from an unproductive nap. What I meant to say is >that if the scenes were of equal brightness, the same aperture and >shutter speed would collect the same approximate number of photons. Don't worry. I understood that something was pied in your reply. I've been around here long enough to know who is mostly sensible and who is often not. ---- Paul J. Gans > I don't understand something here, likely because of a stupidity > surge of the sort that often affects me. But why would shooting [quoted text clipped - 4 lines] > go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why > would there be fewer photons? Presumably what he meant was that if I shoot at ISO 200 and 1/200s, f /5.6, then switch to ISO 100, same aperture and 1/100s (in exactly the same situation), then obviously twice as many photons will be detected in the second case. This is what higher ISOs do: give the "same" signal for a lower input (of course in reality the signal isn't the same due to various kinds of noise etc). >> I don't understand something here, likely because of a stupidity >> surge of the sort that often affects me. But why would shooting [quoted text clipped - 4 lines] >> go indoors and shoot at ISO 1600 at say 1/100 and f/5.6, why >> would there be fewer photons? >Presumably what he meant was that if I shoot at ISO 200 and 1/200s, f >/5.6, then switch to ISO 100, same aperture and 1/100s (in exactly the >same situation), then obviously twice as many photons will be detected >in the second case. This is what higher ISOs do: give the "same" signal >for a lower input (of course in reality the signal isn't the same due to >various kinds of noise etc). Yup. Folks have straightened me out on this. In truth, that's what I would have thought. As you say, "Presumably what he meant..." was just this. ---- Paul J. Gans > I have wondered if jpeg processing in-camera would benefit from some > sort of automatic gain control, based on how much lift is needed to [quoted text clipped - 6 lines] > exposure compensation. However, a dynamic control based on the > individual exposure would be better. Yeah that would be a neat firmware hack to do for geeks! Metering would expose for brightest possible without clipping - then you *have* to go back & darken things in post-processing but it gives better quality images. It would work for raw too, EC during raw conversion is better than levels on a jpeg but it's not as good as setting the appropriate ISO. When I adjust EC in CS ACR and hold down ctl to show blown areas, there are few images where a real improvement in clipping can be achieved, normally sliding down won't clean up blown highlights & sliding up won't remove many blocked shadows. >When I adjust EC in CS ACR and hold down ctl to show blown areas, >there are few images where a real improvement in clipping can be >achieved, ACR is not honest about clipping. It nails RAW levels that are in the the vicinity of the RAW maximum to 255 in the output, and stretches the extreme highlights with extreme contrast with aggressive -EC. You need another converter to get natural highlights that are already clipped, or back them down with the "brightness" control (which shouldn't really be necessary). >normally sliding down won't clean up blown highlights & >sliding up won't remove many blocked shadows. You're never going to get your deepest shadows from ACR unless you set "shadows" to 0, instead of 5. With shadows set to 0, +4 EC exposure is actually about +6. ACR's "numbers" are not really meaningful in an exact sense. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>When I adjust EC in CS ACR and hold down ctl to show blown areas, >>there are few images where a real improvement in clipping can be [quoted text clipped - 14 lines] > actually about +6. ACR's "numbers" are not really meaningful in an > exact sense. I didn't explain well. All I meant to say is that adjustments in the RAW conversion are not as effective as adjusting ISO or exposure in the field. Sometimes you can extract decent lost detail, most of the time it is not as good as it seems and that's confirmed by looking at the clipping view with alt held down. Yes the shadows slider is the other one where you can hold down alt & see clipping. For a big stretch of blown sky, in ACR if I hold down alt & drag EC down, the edge of the blown area might reduce slightly and I might even get some rough tones in the middle if I crank it all the way down but then the rest of the image is very dark and those sky tones are badly posterized & very noisy. Likewise for the shadows slider. In most cases the optimal point for those sliders is at zero though, as you watch the clipping, the best exposure is still the original numbers... occasionally I can make a real improvement, more often not. I don't doubt that other raw converters might be able to squeeze a little more out but I prefer the simpler workflow of doing everything in photoshop. Roger shows some pretty significant improvements with 64 bit ImagesPlus as the raw converter, that might be nice for special cases but sounds like a pain for every image. http://www.mlunsold.com/ILNewFeatures.html It seems a much more tedious process. > Roger shows some pretty significant improvements with 64 bit > ImagesPlus as the raw converter, that might be nice for special cases > but sounds like a pain for every image. > http://www.mlunsold.com/ILNewFeatures.html > It seems a much more tedious process. Actually, IP is a much simpler process. Because of IP's 64-bit math, I don't have to tune each raw conversion. Converting from 12-bit raw to 16-bit tiff should be trivial with essentially no loss of information. That is what I see with IP. But with photoshop and rawshooter essentials (two that I've tried) there are serious compromises. You must fiddle with the sliders to tune each image to get good results. With IP, I simply batch convert with no tuning. I then do a levels in IP by checking max signal, and adjusting the high end to maximize signal. I am doing more and more in IP and as a result seeing more and more artifacts with photoshop's 15-bit math. I've just spent about 8 hours prepping a 4-foot panoramic in photoshop, mostly fixing artifacts, on an image done entirely in photoshop several years ago. http://www.clarkvision.com/galleries/gallery.bird/web/c12.19.2002.crane.takeoff. sequence.b-1100.html Contrast that to my latest, mostly IP processing, that was the fastest processing I've ever achieved due to simplified work with fewer processing artifacts in IP: http://www.clarkvision.com/galleries/gallery.bird/web/great.blue.herons.the.kiss .JZ3F8149.f-700.html I also had to do essentially no noise suppression after upsizing and Richardson-Lucy image restoration for 16x22 inch prints. I credit this to a good initial signal-to-noise ratio, but fewer crappy artifacts from 15-bit math. I'm planning to do a study on photoshop's 15-bit math and the processing artifacts it produces. This stuff was not an issue in the old days of a few years ago, but digital S/N is so high now, 15-bit math is becoming limiting. Or maybe it's photoshop's funny way it does the math to speed operations. Some things it does uses additions when it should be doing multiplies. Roger >> Canon DSLRs already have very large dynamic ranges, limited by >> 12-bit A-to_D converters. The 1D mark II, for example, has [quoted text clipped - 6 lines] >> > But its a "hard" clip with the digital sensor and a "soft" clip on film. If you examine the dynamicrange2 web site above, you would see for the 1D mark II that the hard clip of the digital sensor is only 2/3 stop below the print film, and 1/3 stop above slide film. This is for getting any information off the film with a 16-bit scanner. But then the digital goes way below the print or slide film in the shadows. > A non-linear top end response to a digital sensor would eliminate the > problem of blown highlights due to unexpected specular reflections. Say, [quoted text clipped - 5 lines] > In fact, can't this be done with just two additional diodes, one of > which is a zener, in a feedback loop around the amplifier in the pixel? The problem is the full well. The photons convert to electrons in a linear process, so even if you want to encode the output non-linearly, the sensor is still inherently linear and has a full well. All canon, or any other manufacturer would have to do is add more bits, and include some in headroom. For example, the 1D Mark II sensor full well is 80,000 electrons, but the ISO 100 maximum signal is about 53,000 electrons. With some extra bits, you could have about 2/3 of a stop more headroom. Many canon cameras already have a bit shift for jpeg data, which auto shifts the data if the jpeg will saturate. It is a custom function to turn on. I use it all the time. Roger >>> Canon DSLRs already have very large dynamic ranges, limited by >>> 12-bit A-to_D converters. The 1D mark II, for example, has [quoted text clipped - 41 lines] > >Roger With regard to the shadow detail in digital. What does it matter if its there if it consists of near black and white "colour" and next to no tonal gradations? If you bring up the level enough to see it, it looks so odd and so ugly you might rather have it as black or near black. -Rich > With regard to the shadow detail in digital. What does it matter if > its there if it consists of near black and white "colour" and next to > no tonal gradations? If you bring up the level enough to see it, it > looks so odd and so ugly you might rather have it as black or near > black. > -Rich Well, perhaps you have a bad experience with raw converters. Shadows have detail and color, which you can see well if you expose longer. Digital cameras have linear sensors and record the shadow information but with increased noise. Some raw converters do a poor job of getting shadow detail and color. See: Digital Camera Raw Converter Shadow Detail and Image Editor Limitations: Factors in Getting Shadow Detail in Images http://www.clarkvision.com/imagedetail/raw.converter.shadow.detail The larger the full well, and the larger the pixels, the more electrons that can be collected and the better the shadow information will be recorded. The problem I am seeing is that the 15-bit math of photoshop is not adequate to properly render and manipulate high dynamic range images. Photoshop's high dynamic range mode has very few tools. So working with high dynamic range data one might need other image processing programs. I use a system that does 64-bit math. Roger > The problem is the full well. The photons convert to electrons in > a linear process, so even if you want to encode the output [quoted text clipped - 6 lines] > signal is about 53,000 electrons. With some extra bits, > you could have about 2/3 of a stop more headroom. How exactly do the photodiodes work? Do the 'wells' really have depth? Thicker silicon chips? > Many canon cameras already have a bit shift for jpeg data, > which auto shifts the data if the jpeg will saturate. > It is a custom function to turn on. I use it all the time. > How exactly do the photodiodes work? Do the 'wells' really have depth? > Thicker silicon chips? The "well" is a potential well where electrons get trapped. Yes, it does have physical size, much as a capacitor has physical size. Larger pixels have larger potential wells allowing more electrons to be collected. Roger >> How exactly do the photodiodes work? Do the 'wells' really have >>depth? Thicker silicon chips? [quoted text clipped - 3 lines] >physical size. Larger pixels have larger potential wells >allowing more electrons to be collected. Roger, in a CMOS chip that "potential well" *is* a capacitor. It is a potential well in a CCD, the difference there is that it is dynamic, the trapping region of the well being produced as a depletion layer underneath an insulated electrode when a (relatively high) voltage is applied to it. Thus in a CCD the size of the well (or virtual capacitor) can be changed by the voltage applied to the electrode. As clock cycles sequence in different phases on the CCD adjacent electrodes have their voltages modified to create and destroy adjacent wells, thus moving the charge through the silicon to the readout in the same way as a fluid is moved through a tube by a peristaltic pump. In a CMOS sensor this creation and destruction of dynamic virtual capacitors to move the charge does not occur. The capacitor on a CMOS sensor is fixed and the charge integrates on that to produce the signal voltage directly, requiring only buffering and multiplexing for readout - although some CMOS devices can switch additional capacitors in parallel to increase or decrease the storage capacity of the pixel. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>> How exactly do the photodiodes work? Do the 'wells' really have >>> depth? Thicker silicon chips? [quoted text clipped - 11 lines] > is applied to it. Thus in a CCD the size of the well (or virtual > capacitor) can be changed by the voltage applied to the electrode. OK, a capacitor as I understand is a device that holds a charge of photons/electrons (I may be way off?) so the silicon beneath the photodiode opening is like a catcher's mitt or sponge for electrons, then later the camera calls up all the balls that were caught & counts them (sometimes miscounting due to analog/digital-quantization problems) so then couldn't you make the catcher's mitt deeper, the sponge/silicon thicker? > As clock cycles sequence in different phases on the CCD adjacent > electrodes have their voltages modified to create and destroy adjacent [quoted text clipped - 6 lines] > although some CMOS devices can switch additional capacitors in parallel > to increase or decrease the storage capacity of the pixel. SignaturePaul Furman http://www.edgehill.net/1 Bay Natives http://www.baynatives.com >OK, a capacitor as I understand is a device that holds a charge of >photons/electrons (I may be way off?) so the silicon beneath the [quoted text clipped - 3 lines] >problems) so then couldn't you make the catcher's mitt deeper, the >sponge/silicon thicker? You could, but it isn't trivial. A capacitor is effectively two plate electrodes separated by an insulating dielectric. The size of the capacitor is proportional to the area of overlap of the electrodes, inversely proportional to their separation and proportional to the electrical permitivity of the dielectric. On the chip, one of the electrodes is the silicon substrate and the other is a metal layer and the dielectric is a layer of silicon oxide (glass). You can't make the electrodes any larger because the capacitor has to fit inside a pixel - it is already the biggest component in the pixel. You can't make the dielectric any thinner or it will break down at the operating voltage. So the only realistic option is to use a different dielectric with a higher permitivity, such as a ferroelectric material like lead zirconium titanate (PZT) or something similar. That is a significant departure from standard silicon chip processing and very few plants have the capability to do that and fewer still prepared to let such materials anywhere near their plant. The potential for contamination knocking out a few billion dollars of investment is significant. Samsung in Korea have done some work making very dense memory chips with ferroelectric layers as the dielectric, and I was involved in a development using thin film ferroelectric layers for thermal imager detectors (ferroelectrics are also pyroelectric). The biggest problem is the process temperature of the ferroelectric required to obtain a decent dielectric constant, which at >900degC is well above the melting temperature of the metals used in the semiconductor process. The mismatch is shown in this graph http://www.ams.mod.uk/ams/content/docs/ti/presentn/uncooled/sld022.htm You have to resolve that problem before you get a bigger catchers mitt. ;-) SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) [] > http://www.ams.mod.uk/ams/content/docs/ti/presentn/uncooled/sld022.htm A most interesting presentation (from someone involved in thermal imaging and cooled detectors 25 years ago!). David >[] >> http://www.ams.mod.uk/ams/content/docs/ti/presentn/uncooled/sld022.htm > >A most interesting presentation (from someone involved in thermal imaging >and cooled detectors 25 years ago!). Sadly, completely out of date now, only 4 years after it was given - but MOD still use it for internal awareness, hence its general availability. It should be getting updated sometime soon. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) > In article <yN9Vf.41761$wl.5954@text.news.blueyonder.co.uk>, David J > Taylor <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk> [quoted text clipped - 8 lines] > but MOD still use it for internal awareness, hence its general > availability. It should be getting updated sometime soon. I would appreciate knowing about any updates, thanks. David >> But its a "hard" clip with the digital sensor and a "soft" clip on >>film. [quoted text clipped - 3 lines] >below the print film, and 1/3 stop above slide film. This is for >getting any information off the film with a 16-bit scanner. Maybe I am missing it, but the only charts on that page that appear relevant to the discussion are 8a/b, which are the same data represented on different scales. Neither seem to show the 2/3 stop difference in limit that you are talking about. If they are the charts you are referring to then I don't think they address the issue I was referring to. These charts show a representation of the digital sensor response *after* processing a raw output with a standard curve. The raw output image is, as you note on the page, very linear - consequently the generation of the response curve is after the digital quantising ADC stage. It is saturation of that ADC stage that causes the effect of blown highlights. By comparison, the curved film response is real and entirely in analogue, prior to quantisation at the ADC, irrespective of the number of bits in the scanner. The tapered response to highlights of the film clips overloads softly, whilst the digital linear output is hard clipped and *then* response shaped. >But then the digital goes way below the print or slide film in the >shadows. Shadows are a different matter entirely. >> A non-linear top end response to a digital sensor would eliminate the >>problem of blown highlights due to unexpected specular reflections. [quoted text clipped - 10 lines] >non-linearly, the sensor is still inherently linear and has >a full well. You seem to be overlooking something significant - these are CMOS sensors, which means that there is not actual potential well as such. The photocharge is integrated on a capacitor - some of which is actually the parasitic capacitance of the photodiode itself. The integration of that charge produces a voltage on the capacitor directly, which is read by the output circuits. Simple V = Q/C. It is a little more complicated than this, but that is the gist of it. However, the voltage on the capacitor only forms as a linear function of integrated charge because the capacitor has a very low leakage, otherwise the voltage would integrate exponentially. Thus, creating a circuit which increased the leakage path above a certain threshold voltage would significantly modify the linear conversion in the highlights. That is relatively easy to achieve with two diodes, since the I-V curve of a forward biased diode is strongly exponential. This would produce a linear response up to the voltage at which the diode becomes forward biased and a logarithmic response from there on. The result is a linear response with a soft cut-off within the saturation limit of the "well" and, most significantly, it is implemented before any ADC. >All canon, or any other manufacturer would have to do is add more >bits, and include some in headroom. For example, the 1D Mark II >sensor full well is 80,000 electrons, but the ISO 100 maximum >signal is about 53,000 electrons. With some extra bits, >you could have about 2/3 of a stop more headroom. In most cases where people are complaining about blown highlights I doubt that 2/3 of a stop would be at all significant. They could do that even at the moment if the diodes were uniform enough, but some margin will always be required (perhaps this is the 2/3 stop you are referring to) to prevent the saturation limits being non-uniform. However, doing that with only 12-bits on the ADC would seriously compromise the shadow response - even more than it already is. If Canon put more bits on their processing it will be to provide better shadow response so that HDR effects can be produced on a single frame exposed for the highlights, rather than the multiple exposure techniques required currently. >Many canon cameras already have a bit shift for jpeg data, >which auto shifts the data if the jpeg will saturate. >It is a custom function to turn on. I use it all the time. It doesn't stop blown highlights appearing in RAW. SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>> But its a "hard" clip with the digital sensor and a "soft" clip on film. >> [quoted text clipped - 7 lines] > on different scales. Neither seem to show the 2/3 stop difference in > limit that you are talking about. You do need to read the text too ;-). The relevant text is in the paragraph above figure 3: http://www.clarkvision.com/imagedetail/dynamicrange2 "The highlights are similarly recorded in all three tests (digital, print film, and slide film), and have intensities in the very bright regions in the digital file within 4% of each other. Some portions of the test images were driven to saturation. To do this, a series of 15 exposures were recorded as a function of exposure time, both above and below the metered exposure and the ones selected had the same brightness at the high end for the same locations on the test target. The exposure times were: Fujichrome velvia: 0.0 stop, 1D Mark II digital: +0.3 stop, and the Kodak Gold 200: +1.0 stop from standard metering. By selecting these exposures, the dynamic range from saturation to the noise floor could be measured for each medium, and the three data sets are registered at the bright end. This is why the curves in Figures 8 and 10 overlay each other at the high end (upper right in each Figure)." What this means is that to drive the image to saturation is the slide film got there at meter compensation 0.0 stops, the print film at +1 stop and the digital +2/3 stop. By saturation on the film, I mean I couldn't get any more detail from my 16-bit scanner. Metered exposure times agreed exactly accounting for film speed ISO differences, and agreed with the digital. For example, the digital metered 0.0 exposure at iso 100 was half the metered time of the iso 200 print film. How good is the scanner? I have reduced my use of split density filters when using film as I find I can recover highlights in film scans that I previously could not print (of get a professional lab to print) with traditional enlarging methods. I have had similar experience with both my scanner (sprintscan 4000) and drum scanners. While it may be possible to get a little more information from negative film, I don't think it would be much. So, for the 1D mark II, saturation is only 1/3 stop lower than print film in this test. It is also my experience in the field that the 1D mark II does very good metering, certainly better than any other camera I've owned, or used (including work). > In most cases where people are complaining about blown highlights I > doubt that 2/3 of a stop would be at all significant. For the 1D mark II, that 2/3 stop improvement would put the saturation point at 1/3 above print film for the same exposure. > They could do > that even at the moment if the diodes were uniform enough, but some [quoted text clipped - 6 lines] > for the highlights, rather than the multiple exposure techniques > required currently. Yes, I've said for a long time the DSLRs need more bits, It was disappointing that at PMA no 14 or 16-bit cameras were introduced. >> Many canon cameras already have a bit shift for jpeg data, >> which auto shifts the data if the jpeg will saturate. >> It is a custom function to turn on. I use it all the time. >> > It doesn't stop blown highlights appearing in RAW. Yes, that's true, but it does help when you are within range. Metering system on other DSLRs I've used, including the 10D and D60 were quite poor and blew the highlights often. The 10D jpeg top end is one stop below the raw saturation. The 1D II rarely blows the highlights in either raw or jpeg. Perhaps that is what Canon means that they will address the issue: better metering on lower end cameras. Roger >> Maybe I am missing it, but the only charts on that page that appear >>relevant to the discussion are 8a/b, which are the same data >>represented on different scales. Neither seem to show the 2/3 stop >>difference in limit that you are talking about. > >You do need to read the text too ;-). Of course and I had no problems with your explanation of how the curves were normalised. >What this means is that to drive the image to saturation is the >slide film got there at meter compensation 0.0 stops, the print [quoted text clipped - 4 lines] >For example, the digital metered 0.0 exposure at iso 100 >was half the metered time of the iso 200 print film. I am a little surprised that negative film essentially only provided 1-stop of over-exposure latitude, which seems to be what this concludes. Given that I often de-rated most negative films by between half and one stop in any case, I would have expected far more incidence of blown highlights than I have found. I still believe this is due to the naturally reducing responsivity of film in the highlights, which digital doesn't, at present, do. >While it may be possible to get a little more information from >negative film, I don't think it would be much. However, that little more information would be soft clipping so it doesn't appear as hard in the final result. I can't see any way of digital resolving this unless they introduce a soft clip technology to the pixel. Derating the ISO and adding more bits will only degrade the noise performance and clipped highlights will still occur. >Yes, that's true, but it does help when you are within range. >Metering system on other DSLRs I've used, including the 10D and >D60 were quite poor and blew the highlights often. The 10D jpeg >top end is one stop below the raw saturation. The 1D II >rarely blows the highlights in either raw or jpeg. Fundamentally though, the problem is that the camera meters for an average exposure, whatever metering option is actually used, whilst the sensor clips on highlights, which may only be a small (and thus unmetered) part of the image. The only real way to avoid this is to meter for highlights (which is generally what a scanner will do, with a prescan) and that requires a meter sensor which has similar resolution as the image sensor. I don't mean exactly the same resolution, perhaps 1/64th of the pixel count would be good enough, but covering the entire frame. One thing I do miss from the OM-4 spot meter was the ability to multi-spot meter and automatically expose for highlight or shadow, which was an attempt in the analogue film world to address this problem. The 5D doesn't seem to have a similar facility - I don't know if the 1D-II has. One of the advantages of multi-spot metering was that I could click on several objects in the frame that looked like they might be the dominant highlight and the camera would automatically sort them in order and then determine an exposure to ensure that the brightest was within the film range - or the darkest with shadows. I can, of course, do the same thing manually with the Canon, but it requires a bit of memory and then manual intervention - much slower. However such a facility would, IMO, be much more useful on digital than it ever was on film, for exactly this reason. >Perhaps that is what Canon means that they will address >the issue: better metering on lower end cameras. Perhaps, but I suspect it is likely to be a more fundamental change, and one that will be applicable across the range. Either way, I hope it *isn't* just to derate the ISO and add more bits, which IMO would be a fudge with too much cost in noise. We will have to wait and see. ;-) SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>> Maybe I am missing it, but the only charts on that page that appear >>>relevant to the discussion are 8a/b, which are the same data [quoted text clipped - 22 lines] >naturally reducing responsivity of film in the highlights, which digital >doesn't, at present, do. Film negatives max out at a certain density, where there should be areas that although you could expose through them on the enlarger, they should be nearly devoid of detail, the film dye/grain being uniformly exposed. But given that a shadow area could be 9 stops darker than an adjacent light area (in a sunlit scene) it could take a massive overexposure to do this so its likely you would always be able to print some detail from a negative. Unlike digital. Popular Photography's test showed that a 3 stop overexposed negative easily produced a reasonable looking print, almost normal. But take digital three stops over: http://www.pbase.com/andersonrm/image/57723663 and its blown to bits. Yet, two stops over and there are no major problems with the image. Although, there is missing data compared with a correctly exposed image, to the tune of about 20%. http://www.pbase.com/andersonrm/image/57723660 That one stop makes all the difference! -Rich > Film negatives max out at a certain density, where there should be > areas that although you could expose through them on the enlarger, > they should be nearly devoid of detail, the film dye/grain being > uniformly exposed. Yes, that is what I observed in my scans. On negatives, the signal went into the scanner noise, and on slides, it went completely light (all the same max value). Interestingly, in both cases the signal became so uniform if I computed the signal/noise of a box area, it became extremely high. But the true signal was just a constant with no detail. > But given that a shadow area could be 9 stops > darker than an adjacent light area (in a sunlit scene) it could take > a massive overexposure to do this so its likely you would always be > able to print some detail from a negative. Unlike digital. I have the opposite experience. I can bring out detail from scanned film that I could never get printed, in both highlights and shadows. Roger >Film negatives max out at a certain density, where there should be >areas that although you could expose through them on the enlarger, >they should be nearly devoid of detail, the film dye/grain being >uniformly exposed. It's fairly well known in this forum that you have little experience of digital imaging, but have you ever shot any negative film? SignatureKennedy Yes, Socrates himself is particularly missed; A lovely little thinker, but a bugger when he's pissed. Python Philosophers (replace 'nospam' with 'kennedym' when replying) >>Film negatives max out at a certain density, where there should be >>areas that although you could expose through them on the enlarger, [quoted text clipped - 3 lines] >It's fairly well known in this forum that you have little experience of >digital imaging, but have you ever shot any negative film? About 3000 rolls over a few years, when I had a darkroom. -Rich >>>> But its a "hard" clip with the digital sensor and a "soft" clip on film. >>> [quoted text clipped - 48 lines] >that the 1D mark II does very good metering, certainly better than >any other camera I've owned, or used (including work). How does Canon's underrating of the sensitivity of their sensors effect your testing? Or does it? >Yes, I've said for a long time the DSLRs need more bits, It was disappointing >that at PMA no 14 or 16-bit cameras were introduced. What is the hold back in this regard? Cost? -Rich > How does Canon's underrating of the sensitivity of their sensors > effect your testing? Or does it? I don't believe canon is underrating the sensitivity. Exposures for a given ISO agree well with film and other digital cameras. In fact, I have used 3 different canon DSLRs as my light meter for my 4x5 images, and I get them spot on. I usually shoot 4 4x5 frames: two on the DSLR metered value, one under and one over a half stop. It is rare that the metered value is incorrect. On the 4x5, I shoot velvia. >>Yes, I've said for a long time the DSLRs need more bits, It was disappointing >>that at PMA no 14 or 16-bit cameras were introduced. > What is the hold back in this regard? Cost? I think cost, and power are both factors. Canon probably has their 12-bit AtoD in their digic II processor chip set. We'll probably have to wait for digic III. If other camera manufacturers come out with 14 or more bits, it would help force the competition. Roger In message <4425D3B2.8030105@qwest.net>, "Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote: >I don't believe canon is underrating the sensitivity. Exposures >for a given ISO agree well with film and other digital cameras. >In fact, I have used 3 different canon DSLRs as my light meter >for my 4x5 images, and I get them spot on. I find that odd, as you have stated that the Canon 10D was one of your DSLRs. At ISO 100, my 10D is clearly understated by about 2/3 stop; it meters for ISO 64, and even though the JPEGs from the camera are prone to blow-out, the RAW data has all the headroom to support it. Putting both my 20D and 10D on a tripod-mounted lens, and taking the same absolute exposure in manual mode, the RAW DNs are about 50% higher for the 20D. I can shoot a Gretag-Macbeth Color Checker with my Sekonic L558 set to ISO 32 and incident, and the white square comes out at about 3700 (IIRC; I know it didn't clip) in the 10D's RAW green channel. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > In message <4425D3B2.8030105@qwest.net>, > "Roger N. Clark (change username to rnclark)" <username@qwest.net> [quoted text clipped - 16 lines] > ISO 32 and incident, and the white square comes out at about 3700 (IIRC; > I know it didn't clip) in the 10D's RAW green channel. Hmm, I have shot film, 10D and 1D mark II side by side and get the same meter value for "easy" scenes. The 10D and 1D II start to differ in complex lighting, with the 10D blowing highlights. When I use my digital camera as a light meter for my 4x5, I do zoom in on subjects and "spot" meter, thus compensating for complex lighting. Roger >According to Pop Photography, Canon is going to bring out something in >2007 to fix the >inability of sensors to deal with wide dynamic ranges. They (Pop >Photography) said they'd say what it is next month. I'm wondering if >they've figured out a way to regulate the sensitivity of individual >pixels in an array by varying something (signal amplification?). No, its call more pixels just like MF backs do today. Signature"I have been a witness, and these pictures are my testimony. The events I have recorded should not be forgotten and must not be repeated." -James Nachtwey- http://www.jamesnachtwey.com/ |
Enable EMail Alerts
Start New Thread
Reply to this Message |
 |