Jim Brennan
VHS Member Old Skool
Re: CCD to Film
I'm on that same train. If it crashes, we both go together.
Thank God for the science club.
I'm on that same train. If it crashes, we both go together.
Thank God for the science club.
CCD to Film
- Thread starter braw
- Start date
10s
Veteran
Re: CCD to Film
CCD Pixels are often thought of as buckets. If you set a hundred buckets on a given piece of ground and it began to rain it would collect so much water. If the you increased the size of those 100 buckets to a wash tub size (with the same size land area) you'll collect even more water, this is a given. If you had one giant swimming pool covering the same geographic land as the 100 buckets you'd collect even more water...because obviously the collection vessels are increasing larger, which in turn collects more rain water. So it is with light.
The larger the Pixels in CCDs the larger the collection buckets for photons are to fall in and create an electrical signal. So it takes less rain/ photons to activate the buckets/ pixel discharge. We can think of these bucket/pixels as triggers that activate an electrical signal.
One big pixel would be the ultimate sensor if all one wanted was an off on switch activated by light. But we don't seem to be too thrilled with the results, we want more, we want both.
Many small buckets increase the resolution because these collection buckets differentiate between where the ground is being hit by the rain, but they also become less likely to be struck by a rain drop so we think they are less sensitive.
The goal is to cover the ground with buckets the size of the rain drops and minimize the non collective area in between the buckets. In other words, pixels the size of photons ranging from 700nm to 400nm as tightly packed as possible will yield the best results. This will yield optimum resolution though each pixel/ bucket will seem to be less sensitive to being struck. When photons hit the area between the pixel/buckets we loose resolution & sensitivity.
New cameras are being designed to achieve 4k resolution so it can compete with film.
CCD Pixels are often thought of as buckets. If you set a hundred buckets on a given piece of ground and it began to rain it would collect so much water. If the you increased the size of those 100 buckets to a wash tub size (with the same size land area) you'll collect even more water, this is a given. If you had one giant swimming pool covering the same geographic land as the 100 buckets you'd collect even more water...because obviously the collection vessels are increasing larger, which in turn collects more rain water. So it is with light.
The larger the Pixels in CCDs the larger the collection buckets for photons are to fall in and create an electrical signal. So it takes less rain/ photons to activate the buckets/ pixel discharge. We can think of these bucket/pixels as triggers that activate an electrical signal.
One big pixel would be the ultimate sensor if all one wanted was an off on switch activated by light. But we don't seem to be too thrilled with the results, we want more, we want both.
Many small buckets increase the resolution because these collection buckets differentiate between where the ground is being hit by the rain, but they also become less likely to be struck by a rain drop so we think they are less sensitive.
The goal is to cover the ground with buckets the size of the rain drops and minimize the non collective area in between the buckets. In other words, pixels the size of photons ranging from 700nm to 400nm as tightly packed as possible will yield the best results. This will yield optimum resolution though each pixel/ bucket will seem to be less sensitive to being struck. When photons hit the area between the pixel/buckets we loose resolution & sensitivity.
New cameras are being designed to achieve 4k resolution so it can compete with film.
J_Barnes
Veteran
Re: CCD to Film
Somebody went a little wikipedia crazy! Anyone want to toss some Bayer Masking into the discussion?
Also, is it better to use a metal bucket over a wooden one, or will you have issues with voltage when you pack them tight?
Electronic image sensing is what everything is going to be in the future (I'm guessing Foevon over CCD or CMOS), but it's still got a tiny bit of work in terms of cost and quality before it makes sense to choose a digital image over a filmed one at the budget level of 35mm.
Somebody went a little wikipedia crazy! Anyone want to toss some Bayer Masking into the discussion?
Also, is it better to use a metal bucket over a wooden one, or will you have issues with voltage when you pack them tight?
Electronic image sensing is what everything is going to be in the future (I'm guessing Foevon over CCD or CMOS), but it's still got a tiny bit of work in terms of cost and quality before it makes sense to choose a digital image over a filmed one at the budget level of 35mm.
fomoDVXpal
Active member
Re: CCD to Film
10s that makes a lot of sense.
again, i talk as an engineering novice...but...
why don't you just have the large size pixels you want and then make heaps of them?
i.e large CCD chips?
i'm sure they can customize the lenses so they can transfer light to the larger size.
i mean, why can still photo cameras do it and not video cameras??
10s that makes a lot of sense.
again, i talk as an engineering novice...but...
why don't you just have the large size pixels you want and then make heaps of them?
i.e large CCD chips?
i'm sure they can customize the lenses so they can transfer light to the larger size.
i mean, why can still photo cameras do it and not video cameras??
10s
Veteran
Re: CCD to Film
Pixels need to be as small as possible just like film grain tries to be small as possible. What you've read here about large size pixels is rubbish. A large target area does make it easy for objects to hit it but it also reduces resolution.
Think of it this way, if we had one big giant target, it would be easier to hit then one micro tiny size target. That we can agree on. The trick is to put many micro targets next to each other that equal the size of the large target and then when you shoot your bow & arrow it will hit a target, just which one? The target surface is equal in it's ability to be penitrated.
Another way to explore this idea, take your computer monitor and reduce the number of pixels on the screen, what happens? Soon you get giant blocky looking garbage on the screen, and if you went far enough the screen would be one pixel. What's good about that.
Pixels need to be as small as possible just like film grain tries to be small as possible. What you've read here about large size pixels is rubbish. A large target area does make it easy for objects to hit it but it also reduces resolution.
Think of it this way, if we had one big giant target, it would be easier to hit then one micro tiny size target. That we can agree on. The trick is to put many micro targets next to each other that equal the size of the large target and then when you shoot your bow & arrow it will hit a target, just which one? The target surface is equal in it's ability to be penitrated.
Another way to explore this idea, take your computer monitor and reduce the number of pixels on the screen, what happens? Soon you get giant blocky looking garbage on the screen, and if you went far enough the screen would be one pixel. What's good about that.
Barry_Green
Moderator
Re: CCD to Film
Rubbish? Hmmm...
The answer to pixel size is: you want as many as it takes to cover the frame, and you want them to be as large as possible. Broadcast cameras have as many pixels as it takes to give them a properly-resolved frame, but no more, because they want each pixel to be as large as possible while still providing enough to resolve the image.
So you may find a 1/6" camera that has about 410,000 pixels, and a 2/3" camera (with over 13x as much surface area) will also have 410,000 pixels. The 2/3" camera's pixels will be 13x as large as the 1/6" camera's. And what do you find with 2/3" cameras as compared to 1/6" cameras? Much better low light performance, and better dynamic range.
Because bigger pixels are better.
As long as you have enough to resolve the frame. Obviously too-few pixels will result in soft footage. And too many result in aliasing and poor dynamic range and smear and worse low-light performance.
You want the pixels as big as you can possibly get them, and as close to a 1:1 ratio to the sampled frame dimensions as you can get.
Rubbish? Hmmm...
The answer to pixel size is: you want as many as it takes to cover the frame, and you want them to be as large as possible. Broadcast cameras have as many pixels as it takes to give them a properly-resolved frame, but no more, because they want each pixel to be as large as possible while still providing enough to resolve the image.
So you may find a 1/6" camera that has about 410,000 pixels, and a 2/3" camera (with over 13x as much surface area) will also have 410,000 pixels. The 2/3" camera's pixels will be 13x as large as the 1/6" camera's. And what do you find with 2/3" cameras as compared to 1/6" cameras? Much better low light performance, and better dynamic range.
Because bigger pixels are better.
As long as you have enough to resolve the frame. Obviously too-few pixels will result in soft footage. And too many result in aliasing and poor dynamic range and smear and worse low-light performance.
You want the pixels as big as you can possibly get them, and as close to a 1:1 ratio to the sampled frame dimensions as you can get.
J_Barnes
Veteran
Re: CCD to Film
10s, what you're saying makes logical sense, but surface logic rarely has bearing in quantum mechanics.
Essentially, it's an issue of voltage. All this talk of pixel size boils down to voltage rather then resolution, thats where it gets easy to mix up the facts.
The larger a pixel is, the more photons it receives and the more voltage it produces. It's just that simple, and it's the translation of photons to voltage that determines the low light performance of the individual pixels and thus the imager as a whole.
There is a limited amount of light available to excite each pixel and once you start adding pixels beyond a certain point, you begin to subdivide the photons available to each pixel, and thus subdivide the voltage output of those pixels.
If you squeeze four pixels of the same sensitivity into the space of one, you don't get a comparative level of voltage output. You have four pixels outputing 1/4 the voltage. Sure you have more apparent resolution, but at 1/4 the sensitivity to light.
Furthermore, CCD pixels need to capture a certain amount of photons to resolve a signal above the noise floor. The smaller a pixel is, the more light it needs to gather to be distinguished as information rather then noise. This is seen in the final image as dynamic range. The smaller a pixel, the lower the dynamic range, the larger the pixel...the larger the signal to noise ratio.
10s, what you're saying makes logical sense, but surface logic rarely has bearing in quantum mechanics.
Essentially, it's an issue of voltage. All this talk of pixel size boils down to voltage rather then resolution, thats where it gets easy to mix up the facts.
The larger a pixel is, the more photons it receives and the more voltage it produces. It's just that simple, and it's the translation of photons to voltage that determines the low light performance of the individual pixels and thus the imager as a whole.
There is a limited amount of light available to excite each pixel and once you start adding pixels beyond a certain point, you begin to subdivide the photons available to each pixel, and thus subdivide the voltage output of those pixels.
If you squeeze four pixels of the same sensitivity into the space of one, you don't get a comparative level of voltage output. You have four pixels outputing 1/4 the voltage. Sure you have more apparent resolution, but at 1/4 the sensitivity to light.
Furthermore, CCD pixels need to capture a certain amount of photons to resolve a signal above the noise floor. The smaller a pixel is, the more light it needs to gather to be distinguished as information rather then noise. This is seen in the final image as dynamic range. The smaller a pixel, the lower the dynamic range, the larger the pixel...the larger the signal to noise ratio.
J_Barnes
Veteran
Re: CCD to Film
I must assume you're trying to make a point and not simply trying to be antagonistic to everyone offering thoughtful and knowledgable information here.
So perhaps you can tell me how your question relates to the discussion?
EDIT: I don't mean that to sound snooty, I'm just not sure what you're implying or where you're going with that thought.
I must assume you're trying to make a point and not simply trying to be antagonistic to everyone offering thoughtful and knowledgable information here.
So perhaps you can tell me how your question relates to the discussion?
EDIT: I don't mean that to sound snooty, I'm just not sure what you're implying or where you're going with that thought.
10s
Veteran
Re: CCD to Film
Pananvision has a 12.4 mega pixel chip camera. Why didn't they make it half the pixel count and double the size of the pixels? DALSA has a 4k camera too....why did they put in all that money in R&D for high pixel chips if pixel count doesn't matter that much? the answer: it does matter, so much so they use it to compete and up the ante.
What has been said here is good, but there's more to it, and we can see that when we consider why the lead edge companies are pursuing these new chip designs.
Pananvision has a 12.4 mega pixel chip camera. Why didn't they make it half the pixel count and double the size of the pixels? DALSA has a 4k camera too....why did they put in all that money in R&D for high pixel chips if pixel count doesn't matter that much? the answer: it does matter, so much so they use it to compete and up the ante.
What has been said here is good, but there's more to it, and we can see that when we consider why the lead edge companies are pursuing these new chip designs.
J_Barnes
Veteran
Re: CCD to Film
The point isn't that pixel count doesn't matter, no one has said that, and the fact that companies are pursuing high pixel-count systems has not been challenged by anyone in this thread.
The point is only that more pixels results in lower light response. It's exactly the same as with film (to use your analogy). Illford Delta 400 has a much larger grain structure then Illford Delta 100 does.
Knowing this you can accurately make two distinctions with absolute certainty:
1. the 400 speed film has better low light response then the 100 speed film. Larger silver halide crystals are proportionately more responsive when compared to smaller silver halide crystals. A larger target area responds more then numerous smaller target areas comprising the same space.
2. The 400 speed film is not as high a resolution as the 100 speed film. Because the silver halide crystals are larger, there is technically less information captured on the film when compared to 100 speed. The 100 speed has smaller crystals that result in a tighter grain structure when exposed, increasing line definition and resolving detail to a finer degree then a 400 speed film.
The same is true for a CCD. Given equal dimensions a higher-pixel count CCD will resolve more detail then a low pixel count CCD, but conversely...a higher pixel count CCD won't respond as well in low light conditions as a low pixel count CCD.
The point being that sure, you can make a CCD any size you want and cram any number of pixels you want in that size...but once you start increasing the pixel count, you need to make more Herculean improvements in circuit and capacitor design to maintain a certain level of low-light responsiveness. This is why lab-grade CCD’s require Peltier cooling systems to keep them from melting, there’s some serious engineering in the substructure of a CCD system.
So the point is that with all things being equal, more pixels results in lower light response. That’s just a fact, not an opinion, but it also does not allow or prevent any specific design, pixel count, or CCD size.
The interesting question is, why is Panavision producing a 12 megapixel CCD in a camera that doesn’t output 12 megapixel images? Perhaps it’s pixel redundancy? Perhaps an effort to reduce chroma undersampling? I don't know, but that's what I suspect.
There are a lot of technical hurdles that need to be overcome before a full frame CCD can directly compete with 35mm film in both quality and cost of operation. Bayer masking and chroma undersampling are both significant issues that will need to be addressed eventually. The answer won’t be prismatic 3-chip designs, so unless they can come up with some vast Peltier-cooled super-sampled chip…someone is going to have to implement something similar to foevon.
Either way, I’ll shoot with the camera someone pays me to shoot with.
The point isn't that pixel count doesn't matter, no one has said that, and the fact that companies are pursuing high pixel-count systems has not been challenged by anyone in this thread.
The point is only that more pixels results in lower light response. It's exactly the same as with film (to use your analogy). Illford Delta 400 has a much larger grain structure then Illford Delta 100 does.
Knowing this you can accurately make two distinctions with absolute certainty:
1. the 400 speed film has better low light response then the 100 speed film. Larger silver halide crystals are proportionately more responsive when compared to smaller silver halide crystals. A larger target area responds more then numerous smaller target areas comprising the same space.
2. The 400 speed film is not as high a resolution as the 100 speed film. Because the silver halide crystals are larger, there is technically less information captured on the film when compared to 100 speed. The 100 speed has smaller crystals that result in a tighter grain structure when exposed, increasing line definition and resolving detail to a finer degree then a 400 speed film.
The same is true for a CCD. Given equal dimensions a higher-pixel count CCD will resolve more detail then a low pixel count CCD, but conversely...a higher pixel count CCD won't respond as well in low light conditions as a low pixel count CCD.
The point being that sure, you can make a CCD any size you want and cram any number of pixels you want in that size...but once you start increasing the pixel count, you need to make more Herculean improvements in circuit and capacitor design to maintain a certain level of low-light responsiveness. This is why lab-grade CCD’s require Peltier cooling systems to keep them from melting, there’s some serious engineering in the substructure of a CCD system.
So the point is that with all things being equal, more pixels results in lower light response. That’s just a fact, not an opinion, but it also does not allow or prevent any specific design, pixel count, or CCD size.
The interesting question is, why is Panavision producing a 12 megapixel CCD in a camera that doesn’t output 12 megapixel images? Perhaps it’s pixel redundancy? Perhaps an effort to reduce chroma undersampling? I don't know, but that's what I suspect.
There are a lot of technical hurdles that need to be overcome before a full frame CCD can directly compete with 35mm film in both quality and cost of operation. Bayer masking and chroma undersampling are both significant issues that will need to be addressed eventually. The answer won’t be prismatic 3-chip designs, so unless they can come up with some vast Peltier-cooled super-sampled chip…someone is going to have to implement something similar to foevon.
Either way, I’ll shoot with the camera someone pays me to shoot with.
10s
Veteran
Re: CCD to Film
Good info, i think we are in agreement
Why are they making these high pixel chips?
Dalsa, Panavision, ARRI...are trying to equal 35mm film with 4k video cameras. HD has far less resolution than film. They aren't interested in wasted valuable money on R&D for silly goals. This is where it should get really interesting in the next 5 years.
Good info, i think we are in agreement
Why are they making these high pixel chips?
Dalsa, Panavision, ARRI...are trying to equal 35mm film with 4k video cameras. HD has far less resolution than film. They aren't interested in wasted valuable money on R&D for silly goals. This is where it should get really interesting in the next 5 years.
Re: CCD to Film
..i wouldnt say HD is a silly goal. theres an issue of "how much resolution or what quality is enough" . just as most people cant tell the differnece between a jpg at 60% quality and a full uncompressed file. most people cant tell you whether somthing is shot on HD or film and they dont really percieve any difference. espescially if the HD can replicate the gamma and color reproduction of film. its the same as the issue with SD vs HD. sure HD has alot more res. but the real question is .. do people actually care? IMHO there is a point of silliness at both ends of the stick. it may be silly to play with HD if you can simply develop a great chip at actual film res, but its just as silly to make the film res chip if nobody can tell the difference or cares about the difference between its output and the HD.
that said ... id personally take the 4k chips anyday.. perhaps it is overkill for what people actually percieve in the end, but .. i tend to like overkill
..i wouldnt say HD is a silly goal. theres an issue of "how much resolution or what quality is enough" . just as most people cant tell the differnece between a jpg at 60% quality and a full uncompressed file. most people cant tell you whether somthing is shot on HD or film and they dont really percieve any difference. espescially if the HD can replicate the gamma and color reproduction of film. its the same as the issue with SD vs HD. sure HD has alot more res. but the real question is .. do people actually care? IMHO there is a point of silliness at both ends of the stick. it may be silly to play with HD if you can simply develop a great chip at actual film res, but its just as silly to make the film res chip if nobody can tell the difference or cares about the difference between its output and the HD.
that said ... id personally take the 4k chips anyday.. perhaps it is overkill for what people actually percieve in the end, but .. i tend to like overkill
J_Barnes
Veteran
Re: CCD to Film
True, they are high resolution chips, but the panavision camera, while having a 12 megapixel sensor, outputs only to HD resolution. The Dalsa, while having a 4k resolution sensor (reportedly), only 3000 of that horizontal field gets hit by a lens. Why? Not to mention that the effective resolution of these sensors winds up being so much smaller then their specs suggest because of the way they sample color. Even 4k chips, to the audience member they are probably indistinguishable from the resolution of a 35mm frame, but they're not quite enough resolution to truly equal 35mm. It's certainly a good step!
Honestly, I think the engineers working on these CCDs aren't designing them for serious use, I think they're designing them as inbetweens before they tackle the problems of color resolution, heat dissapation, throughput, data transport, etc.
I think these cameras are only R&D goals, not products. They have to show them off to keep their name in the mix, but I don't believe that they are the serious "second generation" offerings as so often ascribed. They're the steps between HD cams and 35mm emulation cams we will eventually have. (whatever final resolution they wind up at isn't that important)
True, they are high resolution chips, but the panavision camera, while having a 12 megapixel sensor, outputs only to HD resolution. The Dalsa, while having a 4k resolution sensor (reportedly), only 3000 of that horizontal field gets hit by a lens. Why? Not to mention that the effective resolution of these sensors winds up being so much smaller then their specs suggest because of the way they sample color. Even 4k chips, to the audience member they are probably indistinguishable from the resolution of a 35mm frame, but they're not quite enough resolution to truly equal 35mm. It's certainly a good step!
Honestly, I think the engineers working on these CCDs aren't designing them for serious use, I think they're designing them as inbetweens before they tackle the problems of color resolution, heat dissapation, throughput, data transport, etc.
I think these cameras are only R&D goals, not products. They have to show them off to keep their name in the mix, but I don't believe that they are the serious "second generation" offerings as so often ascribed. They're the steps between HD cams and 35mm emulation cams we will eventually have. (whatever final resolution they wind up at isn't that important)
Re: CCD to Film
i think that they need to work on developing a very fast laser scanner that can scan a 35mm area at speeds up to 60times per sec. that way they just have the lense throw onto a white pure flat screen type of surface, and the scanner just scans the projected image at whatever res , and at the given fps. it would capture all the detail from the deepest shadows to the brightest brights, because its a scanner it should capture all of the available light accurately representing the full spectrum of available dynamic range in the image and accuratley recreate the image the lense is throwing.
i know were talking about a really fast scanner there, but it would really kick butt.
.. anyway :
i think that they need to work on developing a very fast laser scanner that can scan a 35mm area at speeds up to 60times per sec. that way they just have the lense throw onto a white pure flat screen type of surface, and the scanner just scans the projected image at whatever res , and at the given fps. it would capture all the detail from the deepest shadows to the brightest brights, because its a scanner it should capture all of the available light accurately representing the full spectrum of available dynamic range in the image and accuratley recreate the image the lense is throwing.
i know were talking about a really fast scanner there, but it would really kick butt.
.. anyway :
10s
Veteran
Re: CCD to Film
"Honestly, I think the engineers working on these CCDs aren't designing them for serious use, I think they're designing them as inbetweens before they tackle the problems of color resolution, heat dissapation, throughput, data transport, etc."
Why has Dalsa, a new entrant in the high rez market, set up a large new service facility in LA for major motion picture production if they aren't serious?
They're spending millions to get this up off the ground. Are all those engineers & ASC cinematographers that silly? ....No, I don't think so, the must be on to something.
I like the scanning idea....you might be on to something.
"Honestly, I think the engineers working on these CCDs aren't designing them for serious use, I think they're designing them as inbetweens before they tackle the problems of color resolution, heat dissapation, throughput, data transport, etc."
Why has Dalsa, a new entrant in the high rez market, set up a large new service facility in LA for major motion picture production if they aren't serious?
They're spending millions to get this up off the ground. Are all those engineers & ASC cinematographers that silly? ....No, I don't think so, the must be on to something.
I like the scanning idea....you might be on to something.