no title
June 20, 2006 2:56 AM Subscribe
Is it possible for digital camera sensors to absorb light faster than silver halides on film ?
I'm interested in this - i'm not very up on science but i cant believe digital camera's can only ape film at best - surely someones going to design some super fast sensor somewhere ?
I could be wrong but i'm interested if it's possible and wouldnt mind learning a bit more about the subject.
I'm interested in this - i'm not very up on science but i cant believe digital camera's can only ape film at best - surely someones going to design some super fast sensor somewhere ?
I could be wrong but i'm interested if it's possible and wouldnt mind learning a bit more about the subject.
Digital is different. It's wrong to think of it as "electronic film"; it isn't film at all.
CCDs have a wide dynamic range and a more linear response to light than film does, for one thing. As to whether they are "faster" than film, this is a case where the devil is in the details. It depends on the CCD and on the design of the camera.
No major astronomical observatory in the world today uses film; they've all switched to using CCDs. When they used film and wanted to take an image of a dim object, a single piece of film would be placed into a camera on the scope and exposed for minutes or even hours, on successive nights. But with CCDs, what they do is to dump the developing image about every fifteen minutes, and add them together digitally later to create the effect of a longer exposure. One advantage of that is that if one particular 15-minute exposure got ruined somehow (passing airplane, meteor) they can leave it out.
But the reason they can do that is because CCDs are so linear and because their dynamic range is so wide.
posted by Steven C. Den Beste at 4:01 AM on June 20, 2006
CCDs have a wide dynamic range and a more linear response to light than film does, for one thing. As to whether they are "faster" than film, this is a case where the devil is in the details. It depends on the CCD and on the design of the camera.
No major astronomical observatory in the world today uses film; they've all switched to using CCDs. When they used film and wanted to take an image of a dim object, a single piece of film would be placed into a camera on the scope and exposed for minutes or even hours, on successive nights. But with CCDs, what they do is to dump the developing image about every fifteen minutes, and add them together digitally later to create the effect of a longer exposure. One advantage of that is that if one particular 15-minute exposure got ruined somehow (passing airplane, meteor) they can leave it out.
But the reason they can do that is because CCDs are so linear and because their dynamic range is so wide.
posted by Steven C. Den Beste at 4:01 AM on June 20, 2006
I'm not sure what you mean by "absorb light faster". If digital sensors absorbed light faster, wouldn't every photographer in the world have to re-tune their brains in order to use digital? Most photographers know how a 1/60" shutter speed is going to affect their shot. If you modify digital sensors so that they absorb light faster, a picture taken with 1/60" shutter speed would then be different, wouldn't it? I think?
posted by antifuse at 4:10 AM on June 20, 2006
posted by antifuse at 4:10 AM on June 20, 2006
If you modify digital sensors so that they absorb light faster, a picture taken with 1/60" shutter speed would then be different, wouldn't it?
Not necessarily, because we have a measure of how fast the film is (ISO numbers). And formulas we can apply - the tradeoff between shutter speed, F-stop, and film speed. If digital cameras were faster (say, ISO 1000, ISO 2000..onwards and upwards), we could just adjust our F-stops and exposures accordingly. sgt.serenity wants to know why digital CCDs appear to so closely emulate the "speed" of normal, boring 400 ISO film, when surely we could be improving the technology and creating digital cameras that could take 1/2000th second shots inside without a flash.
I don't know the answer, but I second the question. But I have a feeling it has something to do with the actual amount of energy carried by the individual photons that are hitting the sensor. There has got to be some limit in how much can be achieved by the energy of a photon, whether that is in driving a chemical reaction on a film, or in generating an electric current in a CCD. I imagine someone will come along shortly and do the appropriate maths.
posted by Jimbob at 4:34 AM on June 20, 2006
Not necessarily, because we have a measure of how fast the film is (ISO numbers). And formulas we can apply - the tradeoff between shutter speed, F-stop, and film speed. If digital cameras were faster (say, ISO 1000, ISO 2000..onwards and upwards), we could just adjust our F-stops and exposures accordingly. sgt.serenity wants to know why digital CCDs appear to so closely emulate the "speed" of normal, boring 400 ISO film, when surely we could be improving the technology and creating digital cameras that could take 1/2000th second shots inside without a flash.
I don't know the answer, but I second the question. But I have a feeling it has something to do with the actual amount of energy carried by the individual photons that are hitting the sensor. There has got to be some limit in how much can be achieved by the energy of a photon, whether that is in driving a chemical reaction on a film, or in generating an electric current in a CCD. I imagine someone will come along shortly and do the appropriate maths.
posted by Jimbob at 4:34 AM on June 20, 2006
If I'm following Sarge correctly, he's referring to the amount of "fill" in the image. The digital images can produce a more vibrant result under the right conditions (deeper darks, appearently more livid colors), but as Steven explained, it's due to the dymanic range of the Charged Coupled Devices.
That particular range is also known as the Gamma curve. As digital images are encoded in binary, they can be more readily tweaked for certain results after capturing the image; a number of cameras today, for example, can process the image to reduce red-eye, or motion blur if the photo was taken without a tripod mount. In more elaborate instances, such as with Photoshop, HDR, or High Dynamic Range techniques can come into play.
posted by Smart Dalek at 4:41 AM on June 20, 2006
That particular range is also known as the Gamma curve. As digital images are encoded in binary, they can be more readily tweaked for certain results after capturing the image; a number of cameras today, for example, can process the image to reduce red-eye, or motion blur if the photo was taken without a tripod mount. In more elaborate instances, such as with Photoshop, HDR, or High Dynamic Range techniques can come into play.
posted by Smart Dalek at 4:41 AM on June 20, 2006
I believe film has a greater dynamic range, the range between the darkest darks and the lightest lights. Digital is overtaking film even here, but is not there yet, at least on commercially available cameras. I am not sure about those supercooled sensors used in astronomy.
posted by caddis at 4:51 AM on June 20, 2006
posted by caddis at 4:51 AM on June 20, 2006
At one extreme end of the electronic spectrum, you have devices capable of detecting a single photon. But they are too expensive and bulky to use in regular photography.
sgt.serenity wants to know why digital CCDs appear to so closely emulate the "speed" of normal, boring 400 ISO film, when surely we could be improving the technology and creating digital cameras that could take 1/2000th second shots inside without a flash.
Well, this depends a lot on the camera. The basic problem faced with almost all systems has to do with signal vs. noise. If you amplify the signal, you also amplify the noise with the signal. Filtering doesn't help that much because it's almost always a destructive process. It's very difficult to design a filter that does not chop off some of your signal in ways that may not be desirable.
You can also tackle the problem by reducing the noise in the system, but this gets expensive. Another strategy used by both film and digital cameras is increasing the sensor size, but this also becomes expensive.
I'd argue that most digital and film cameras are built around market competition with a strong bias for something that produces images similar to what we see with the naked eye, and everyday human interactions as the subject. An ISO range of 25-800, a roughly normal lens, and shutter speeds between 1/250th and 1/4th of a second will allow you to capture most of the events you are likely to see in your life. Beyond that, you are getting into more and more specialized markets.
posted by KirkJobSluder at 5:13 AM on June 20, 2006
sgt.serenity wants to know why digital CCDs appear to so closely emulate the "speed" of normal, boring 400 ISO film, when surely we could be improving the technology and creating digital cameras that could take 1/2000th second shots inside without a flash.
Well, this depends a lot on the camera. The basic problem faced with almost all systems has to do with signal vs. noise. If you amplify the signal, you also amplify the noise with the signal. Filtering doesn't help that much because it's almost always a destructive process. It's very difficult to design a filter that does not chop off some of your signal in ways that may not be desirable.
You can also tackle the problem by reducing the noise in the system, but this gets expensive. Another strategy used by both film and digital cameras is increasing the sensor size, but this also becomes expensive.
I'd argue that most digital and film cameras are built around market competition with a strong bias for something that produces images similar to what we see with the naked eye, and everyday human interactions as the subject. An ISO range of 25-800, a roughly normal lens, and shutter speeds between 1/250th and 1/4th of a second will allow you to capture most of the events you are likely to see in your life. Beyond that, you are getting into more and more specialized markets.
posted by KirkJobSluder at 5:13 AM on June 20, 2006
I think by "fast" you mean signal-to-noise ratio. For a given a mount of light (signal), faster film has less visible grain (noise). Dynamic range is also (effectively) proportional to signal-to-noise ratio, so we don't need to discuss that separately.
Signal-to-noise ratio is something you can easily discuss about CCDs, especially noise. As I understand it, the vast amount of noise is introduced by electrical noise the op-amp that turns the tiny charge that builds up in the CCD's cells into a voltage that can be measure by an analog-to-digital convertor. This is eminently solvable by improved shielding, better circuit design, etc.
So I can't give you figures on how current CCD tech compares to film, but I can say that there's no technical reason for them not to be much much better. I'd imagine the current chips in cameras are no better than film because they're optimised for other things instead (cost, reliability, flexibility, etc).
On previews: What Kirk said.
posted by cillit bang at 5:20 AM on June 20, 2006
Signal-to-noise ratio is something you can easily discuss about CCDs, especially noise. As I understand it, the vast amount of noise is introduced by electrical noise the op-amp that turns the tiny charge that builds up in the CCD's cells into a voltage that can be measure by an analog-to-digital convertor. This is eminently solvable by improved shielding, better circuit design, etc.
So I can't give you figures on how current CCD tech compares to film, but I can say that there's no technical reason for them not to be much much better. I'd imagine the current chips in cameras are no better than film because they're optimised for other things instead (cost, reliability, flexibility, etc).
On previews: What Kirk said.
posted by cillit bang at 5:20 AM on June 20, 2006
CCDs are vastly more sensitive than halides -- in particular, some of the top of the line sensors can record less than 10 photons hitting the sensor a minute.
They have to be very cold to reach such sensitivity, but they are able to do so. They're also very noisy at that sensitivity, you really need to measure the noise of each pixel and subtract it out of the final exposures when you're working with sensitive CCDs. Thus, the "dark frame" -- a shot taken with a shutter closed, to read out all noise, no signal. Dark frames are a big part of astronomical observations, often taken before and after a series of images.
Consumer CCDs are less sensitive, because of the noise problem. By ignoring the weaker signals, you can increase the signal-to-noise ratio. This can make a big difference, most consumer cams set to ISO 400 mode are quite obviously noisier than the same cams running at ISO 100.
The reason they're so much more sensitive is quantum efficiency, or how much of the light impacting the sensor is captured. Film rarely goes above 2%. Modern consumer CCD easily hit 60%, and top of the line CCDs approach 75%.
Moderating this is fill factor. A strip of film has a fill of 100% -- the entire area captures light. CCDs don't, they're divided into discrete pixels. Basic CCD lose about half of their imaging area, so a efficiency of 60% and a fill of 50% means you're catching 30% of the light -- still much better than film. Top of the line CCDs with microlenses have an efficiency of 75%, and a fill factor approaching 90%, so you're looking at around 67% effective quantum efficiency -- a dramatic improvement.
Camera CCDs typical have a filter mask installed on them, so the resolution of each color channel is less, the typical ratio is 1:1:2 R:B:G.
Astronomical sensors use filter wheels, so each channel uses the entire sensor (and they use many more filters, for sensing purposes.)
Finally, the biggest win of CCDs -- halides, once they react with line, are finished. CCDs are readout and blanked, ready for the next exposure. So, over long time periods, CCDs are fantastically more sensitive than film, because you can use them repeatedly.
posted by eriko at 5:24 AM on June 20, 2006
They have to be very cold to reach such sensitivity, but they are able to do so. They're also very noisy at that sensitivity, you really need to measure the noise of each pixel and subtract it out of the final exposures when you're working with sensitive CCDs. Thus, the "dark frame" -- a shot taken with a shutter closed, to read out all noise, no signal. Dark frames are a big part of astronomical observations, often taken before and after a series of images.
Consumer CCDs are less sensitive, because of the noise problem. By ignoring the weaker signals, you can increase the signal-to-noise ratio. This can make a big difference, most consumer cams set to ISO 400 mode are quite obviously noisier than the same cams running at ISO 100.
The reason they're so much more sensitive is quantum efficiency, or how much of the light impacting the sensor is captured. Film rarely goes above 2%. Modern consumer CCD easily hit 60%, and top of the line CCDs approach 75%.
Moderating this is fill factor. A strip of film has a fill of 100% -- the entire area captures light. CCDs don't, they're divided into discrete pixels. Basic CCD lose about half of their imaging area, so a efficiency of 60% and a fill of 50% means you're catching 30% of the light -- still much better than film. Top of the line CCDs with microlenses have an efficiency of 75%, and a fill factor approaching 90%, so you're looking at around 67% effective quantum efficiency -- a dramatic improvement.
Camera CCDs typical have a filter mask installed on them, so the resolution of each color channel is less, the typical ratio is 1:1:2 R:B:G.
Astronomical sensors use filter wheels, so each channel uses the entire sensor (and they use many more filters, for sensing purposes.)
Finally, the biggest win of CCDs -- halides, once they react with line, are finished. CCDs are readout and blanked, ready for the next exposure. So, over long time periods, CCDs are fantastically more sensitive than film, because you can use them repeatedly.
posted by eriko at 5:24 AM on June 20, 2006
Absorb light faster? According to the laws of the universe, the speed of light is 186,000 miles/sec. If a digital sensor or a silver halide crystal records the light at all, it will record it when the light hits. Otherwise, after 1/186,000 of a second the light is already a mile away. Too late to record it then. So my bet is on both systems recording it instantaneously as the light hits.
Not sure what you have in mind? Perhaps a super high speed camera? You think maybe film wouldn't be fast enough to record light from one of those? The highest speed film camera recorded at 40k fps and the highest speed ccd camera at 1M fps. The limiting factor with film is obviously the extreme difficulty in transporting film that fast rather than any limitation in the film's ability to record light at such a high shutter speed. Irregardless, whether you have film or ccd, at an extremely high shutter speed, you need lots of light or high sensitivity.
posted by JJ86 at 5:45 AM on June 20, 2006
Not sure what you have in mind? Perhaps a super high speed camera? You think maybe film wouldn't be fast enough to record light from one of those? The highest speed film camera recorded at 40k fps and the highest speed ccd camera at 1M fps. The limiting factor with film is obviously the extreme difficulty in transporting film that fast rather than any limitation in the film's ability to record light at such a high shutter speed. Irregardless, whether you have film or ccd, at an extremely high shutter speed, you need lots of light or high sensitivity.
posted by JJ86 at 5:45 AM on June 20, 2006
I think this question can best be phrased as:
If CCDs are so much better, why can't we set our cameras to 6400 and 12600 ISO speeds and still get good quality results?
But the answer appears to be: because they get really noisy, just like film.
posted by bonaldi at 6:01 AM on June 20, 2006
If CCDs are so much better, why can't we set our cameras to 6400 and 12600 ISO speeds and still get good quality results?
But the answer appears to be: because they get really noisy, just like film.
posted by bonaldi at 6:01 AM on June 20, 2006
because they get really noisy, just like film
well, it's noise vs grain, and as a user of both b&w film and digital, interestingly enough, I still prefer actual film grain.
I'm currently trying to make some Leica C-LUX1 ISO 1600 shots presentable, and the noise is massive. I've recently worked with old Ilford 3200 negs, exposed at 6400, and they print beautifully -- even with massive grain. same for Tri-X, that I routinely expose at 800 -- grain looks better than digital noise to me.
posted by matteo at 7:03 AM on June 20, 2006
well, it's noise vs grain, and as a user of both b&w film and digital, interestingly enough, I still prefer actual film grain.
I'm currently trying to make some Leica C-LUX1 ISO 1600 shots presentable, and the noise is massive. I've recently worked with old Ilford 3200 negs, exposed at 6400, and they print beautifully -- even with massive grain. same for Tri-X, that I routinely expose at 800 -- grain looks better than digital noise to me.
posted by matteo at 7:03 AM on June 20, 2006
having said that, I'm planning to sell all my film cameras, except for an old family Leica M2, in the next few months, while they still have some value.
posted by matteo at 7:05 AM on June 20, 2006
posted by matteo at 7:05 AM on June 20, 2006
Response by poster: If CCDs are so much better, why can't we set our cameras to 6400 and 12600 ISO speeds and still get good quality results?
Aye thats pretty much the question , cheers B.
posted by sgt.serenity at 9:18 AM on June 20, 2006
Aye thats pretty much the question , cheers B.
posted by sgt.serenity at 9:18 AM on June 20, 2006
OK, now that I understand the question better, I think that the people who mentioned noise are right. The CMOS sensor in many Canons takes OK pictures at ISO 3200, but that is with in-camera noise reduction on and using something like Noise Ninja to remove more noise in post-processing. If the sensors in digital cameras were cooled as are the CCDs in telescopes, you could get more light sensitivity but with much more bulk, weight, and complexity for the cooling system.
posted by TedW at 12:44 PM on June 20, 2006
posted by TedW at 12:44 PM on June 20, 2006
cillit bang: As I understand it, the vast amount of noise is introduced by electrical noise the op-amp that turns the tiny charge that builds up in the CCD's cells into a voltage that can be measure by an analog-to-digital convertor.
Well, there is an optimization.. The noise generated by the sensor is related to the area. So, if costs demanded it, you would match the noise level of your electronics to the area of your CCD. Since the CCD cost is still a significant part of the cost of a camera (I think), if electronics noise is dominating, a smaller CCD should be used to save some money.
Also, I've seen digicam sites tell people to buy the lower resolution camera in a model range. Presumably, in that situation the electronics don't change, and the overall size of the CCD doesn't change, but the individual pixels get bigger (there are fewer of them in the same area). If doing that results in a noticeable reduction in noise, then the electronics noise is not dominating.
eriko: By ignoring the weaker signals, you can increase the signal-to-noise ratio.
That sounds really strange, I think it is because we are treating dynamic range and SNR as the same thing (err...). Astronomical instruments trade dynamic range for sensitivity, and in the process the SNR goes down. Okay, but why..
I've heard that high sensitivity CCDs are easily damaged, but I'm thinking that it is actually the electronics that get damaged. Some kind of super low noise and high gain transistor that amplifies the CCD signal might get overvoltaged by a bright flash. If that's right, it is the electronics that limit the SNR in astronomical instruments..
I should probably go looking for some facts.. It is very interesting!
JJ86: Absorb light faster? According to the laws of the universe, the speed of light is 186,000 miles/sec.
Strange as it sounds, looking at eriko's numbers regarding quantum efficiency, 'absorb light faster' is actually a reasonable way to put it. A given area of CCD can absorb more light in a given period of time than the same area of film. That's faster.
matteo: well, it's noise vs grain, and as a user of both b&w film and digital, interestingly enough, I still prefer actual film grain.
In digital audio (much more my area of expertise) it is common to add dither (mimicking analogue noise) to mask the digital noise (not exactly noise, actually, which is the problem - more like distortion). Very interesting to see that the same problem is being noticed by digital photographers - makes sense.
I'll shut up now :P
posted by Chuckles at 9:06 PM on June 20, 2006
Well, there is an optimization.. The noise generated by the sensor is related to the area. So, if costs demanded it, you would match the noise level of your electronics to the area of your CCD. Since the CCD cost is still a significant part of the cost of a camera (I think), if electronics noise is dominating, a smaller CCD should be used to save some money.
Also, I've seen digicam sites tell people to buy the lower resolution camera in a model range. Presumably, in that situation the electronics don't change, and the overall size of the CCD doesn't change, but the individual pixels get bigger (there are fewer of them in the same area). If doing that results in a noticeable reduction in noise, then the electronics noise is not dominating.
eriko: By ignoring the weaker signals, you can increase the signal-to-noise ratio.
That sounds really strange, I think it is because we are treating dynamic range and SNR as the same thing (err...). Astronomical instruments trade dynamic range for sensitivity, and in the process the SNR goes down. Okay, but why..
I've heard that high sensitivity CCDs are easily damaged, but I'm thinking that it is actually the electronics that get damaged. Some kind of super low noise and high gain transistor that amplifies the CCD signal might get overvoltaged by a bright flash. If that's right, it is the electronics that limit the SNR in astronomical instruments..
I should probably go looking for some facts.. It is very interesting!
JJ86: Absorb light faster? According to the laws of the universe, the speed of light is 186,000 miles/sec.
Strange as it sounds, looking at eriko's numbers regarding quantum efficiency, 'absorb light faster' is actually a reasonable way to put it. A given area of CCD can absorb more light in a given period of time than the same area of film. That's faster.
matteo: well, it's noise vs grain, and as a user of both b&w film and digital, interestingly enough, I still prefer actual film grain.
In digital audio (much more my area of expertise) it is common to add dither (mimicking analogue noise) to mask the digital noise (not exactly noise, actually, which is the problem - more like distortion). Very interesting to see that the same problem is being noticed by digital photographers - makes sense.
I'll shut up now :P
posted by Chuckles at 9:06 PM on June 20, 2006
Response by poster: wow, this post is still open - i still think this is a pretty interesting question and we'll see dslr's moving away from merely copying film slr's.
posted by sgt.serenity at 4:32 PM on January 11, 2007
posted by sgt.serenity at 4:32 PM on January 11, 2007
This thread is closed to new comments.
posted by kaytwo at 3:31 AM on June 20, 2006