you can't use it when you don't have an internet connection. Why doesn't anyone think about this?
Because people prefer to think more about how to have an Internet connection all the time. Which is why iPods, cellphones, Palms and even my bloody DVR have Internet jacks, and why I can connect walking around or in my car, even in my rural area of America. It's also why I have wifi access in my home and business, why the local coffee shops have wifi, and in fact, why the local *pizza hut* has free wifi.
'Tis simply all-around better to be connected. And most of us don't need control of the software, as most of us aren't programmers anyway, or programmers interested in fussing with any particular software-as-service. Those programmers that do need access, already have it -- the program authors. RMS is missing the boat. Or perhaps has simply fallen off it.
In response to the announcement of Microsoft's innovative 3-application limit, Apple corporation has said it will release a version of OS X that will allow only one application to run at a time, but in a more friendly and artistically enhanced environment than Windows Reduced Vista(tm.) Apple announced the special version late Sunday evening, at a special event entitled "You're the One." Steve Jobs emerged from his semi-retirement to explain how Apple's invention of this one-to-one relationship between users and applications would "revolutionize computing." Jobs stated that the new OS would also herald a return to the one-button mouse, single monitors, and Apple's new "One-at-a-time" network stream technologies.
Overnight, the Linux community, leveraging its well known security advantages and high speed development based upon open source and developers active in all time zones at once, has released a beta of "Linux Zero", which they claim is the most secure operating system in the world, and the least confusing, by virtue of its enforcement of zero applications running. Linux authority Linus Torvalds said "if an application can't run, it can't bring worms or viruses into the system. In addition, user interaction is now limited to pressing the power button." Waxing optimistic, he went on to say that "We think even Windows users can learn to do this." He told this reporter "In fact, the price is zero, too!"
An unconfirmed rumor also developed this weekend of an OS that is so carefully and explicitly restricted that consumers interaction with it is limited to attempting to install it; as the rumor goes, completing the installation requires permissions that users simply do not have available to them. Such an operating system would provide the ultimate consumer safety net. When asked to comment, both Jobs and Torvalds derided the rumor as being propaganda. Both OS mavens insisted that technology wasn't up to such a challenge yet. The rumor, however, persists.
When contacted by the press for comments on these new developments, Intel explained that multi-core processors were designed specifically for reduced application counts. It is only now that the leading OS manufacturers are revealing their deep strategies for the decade of 2010 that Intel is able to comment on the real rationale for multiple cores. Technical Leader Sanji Ramahasmiran" laid out several reasons why systems with few- or single-application loads would benefit directly from multiple cores. He said "Our new 8-core dies will allow switching the same single task cyclically from one core to another, thus reducing the activity levels to 1/8th that of single-core designs and operating in a greener fashion, contributing less to global warming, and simplifying programmer APIs in any properly designed operating system."
Simply as a personal observation, I always enjoy seeing how competition ensures that corporations compete for the marketplace by leveraging their core competencies and working to out-do one another. The end users always benefit. No matter who your favorite OS manufacturer is, the industry finds a way to work to bring you the latest developments. Isn't technology wonderful?
Ah, but there are other factors as well, such as the fact that a large generator system is much more efficient at turning petroleum into energy than a car's ICE is, and that once the cars are taking energy from the grid, the means of generation can be weaned away from petroleum into any number of other sources without any change in the design of the car.
Pure electric cars make the most sense, period, of any system. Hybrids are interim solutions at best, and things like hydrogen based engines make almost no sense at all - no infrastructure, low efficiencies, etc.
The only technology that is lagging here is batteries / ultracaps. There's already a battery company with a 10-minute recharge time for medium performance batteries; they're ramping up production right now (yes, they're actually making batteries.) Here's a car based on them. Should EEstor or one of the other Ultracap companies come through with fast charging systems, we're off and running. There are no other problems remaining to solve that don't fall to engineering we already know how to do (for instance, a local charging station that builds it's stored charge slowly, overnight, when rates are low and utilities are looking to load balance, and dumps to your car in minutes.)
It is totally about the energy storage system in the vehicle.
So... in your view, products should never be researched and manufactured until your average bear can afford them?
That would be really upsetting news to people who made the first... oh, computers... televisions... CD players... PMPs... DSLRs... video cameras... multichannel stereos... general coverage receivers... GPS units... satellite television receivers...
You know. Things that come under the heading of "stuff that we've never had before because it requires high technology."
So I have a question. If these things aren't initially developed for people with deeper pockets, then where is the money to develop them supposed to come from? Over to you.
...Buggy whip manufacturers say that loans to "internal combustion engine" companies inappropriately drives technology that is "not ready for prime time"; they also state that "horses have a long way to go yet" when making their case that they should get the loan, rather than these newfangled folks. "Horses are cheaper", claims CEO of International Whipping Boys, Inc. Valley professor and writer Randall Stross wrote that was a risky waste of taxpayer money that would only benefit the wealthy and bailout VCs who'd sunk money into the money-losing internal combustion company.
Both metamoderation and moderation are a 100% waste of time on Slashdot.
Great posts are often lost at low ratings, and terrible posts get modded up. Slashdot editors pursue vendettas against various posters, and anonymous posts, regardless of content, are rarely modded at all.
This is a great site, with great content, but the only way to really experience that is to read at -1 and completely ignore the moderation, which simply does not work.
Palmer suggests that one of the two measurement events itself (either position or velocity) is not part of the real universe.
I don't read the article that way at all. The article says (emphasis mine):
According to Palmer's hypothesis, the invariant set contains all the physically realistic states of the universe. So any state that isn't part of the invariant set cannot physically exist.
Suppose you perform the Kochen-Specker thought experiment and measure the position of an electron. Then you ask what you would have found if you repeated the experiment, only this time measuring the electron's velocity instead.
According to Palmer, when you repeat the experiment you are testing a hypothetical universe that is identical to the real one except that the position-measuring equipment is replaced with velocity-measuring equipment.
This is where the fractal nature of the invariant set matters. Consider a place of interest you want to visit along a coastline. If you get the coordinates even slightly wrong you could end up in the sea rather than where you want to be. In the same way, if the hypothetical universe does not lie on the fractal, then that universe is not in the invariant set and so it cannot physically exist.
I read that as saying that the non-fractal math which can resolve velocity, is not, because it is not fractal, also resolving the position because the coarse, non-fractal math falls off the edge if you try and intercept the fractal in two non-fractal dimensions. The implication is that if you use fractal dimensions for the math, you'll get both answers instead of falling into holes in the foam. That's why he talks about "getting the co-ordinates wrong."
The idea here is that the world is deterministic, just not in the three dimensions we'd perhaps like to think it is. It never made sense that a particle with a known position had no velocity; if it's moving, it's moving. So since it is moving, the fault is in how we're measuring it. This may show why such a fault in our measuring approach exists, and how to measure without the fault. That's what makes it so interesting (in this specific case... of course, if the fractal idea is correct, it means a lot more than that as well!)
Well, the point of the article is that if the underlying structure of the universe is fractal, then it shows why, for instance, you can measure the position or the velocity of an electron, but not both; the general idea is that instead of a linear reality, the universe exists along a fractal edge, and answers derived using current quantum methods are literally falling off the edge because they're not finely enough resolved - they don't take the foaminess of the edge into account, so they miss the answer and land in a space that literally isn't part of the real universe - they're undefined. This is an illuminating and interesting idea, and it may point directly to how we could measure both at the same time, which would make a lot more sense to some of us. Me included.
He's not incorporated all of quantum theory into his fractal idea, so this is far from certain, but it is a lovely idea.
In my experience - which is considerable, I'm oooold - at 35, he won't have an age problem. That's not old enough to trigger the insurance companies to really mess with the company's expense of keeping him around under the current insurance setup. And who knows, by then, health care may look somewhat different.
A lot of ageism in tech companies is not being willing to pay for the experience an older employee usually brings to the table; but he's fresh out of school, so that doesn't apply to him. It seems to me that the odds are he'll do ok. He'll also have to accept starting wages, of course.
I went that way, basically have a collection of primes. 30mm, 50mm, 85mm, 100mm(macro), 200mm(L). Don't have the 135mm(L), but that's next on my list. All but the 30mm are EF; the 30mm would have to be replaced if I go FF.
All of these are low light capable; the least among them (100mm, 200mm) are f/2.8, the rest either f/1.4 or f/1.8. That 135mm(L) I want is f/2.
I have some zooms -- the 100-400mm(L) is a great lens, but it's also a daylight lens... and I have the surprisingly good 18-55mm EF-S IS, but again, that's a daylight lens. I don't use either one very often.
The two downsides to all this are the heavy camera bag and the need to swap, which comes up fairly often.
As for the economy... not a good time to be starting most businesses... I think its much, much worse than the media or the politicos are admitting.
Yeah... I almost bought a 5D, but that whole black dot as a consequence of sensel overflow (overexposure) really made me unhappy. As did the fix (filling them with white so that a rightward white bloom replaces the rightward black dot propagation.) That's really not a good thing for someone who shoots astro, as I do. The amount of dynamic range in an astro photo often far exceeds the ability of the camera to represent it, and so overexposure on point light sources is basically a given. I have enough trouble dealing with coma and chromatic aberration as it is; I don't need to add blooming to my list of challenges.
I'm hoping for something that suits my interests this fall (60D?) or perhaps I'll hold out until the 5DmkIII or whatever it turns out to be. I'll have to dump all my EF-S lenses if I go FF, though, and that's kind of depressing. The up side is they tend to retain their value... it's just the exercise of selling it isn't very appealing.
When I said I wouldn't go past 15 mp, I was speaking in the context of APS-C sensors. My fault for not being that little bit more specific.
Very true... but bigger sensors also carry a heavy price premium. 50D street is about $1000; 5DmkII street is about $2700. And medium format... ouch.:)
"Sensels"? There's no need to invent new words. The word "pixel" means "picture element" and is relevant whether discussing monitors or CCDs. Stick with accepted terminology.
Pixels emit data. Pixel is a neologism coined from picture + element.
Sensels collect data. Sensel is a neologism coined from sensor + element.
They're not doing even remotely the same task. Consequently there is a practical reason to discriminate between them. If you are under the impression that anything arranged in a grid that does anything with light, regardless of task, is a pixel, then I suggest you get over to Wham-a-lart and ask them for some linoleum pixels and see what that gets you.
If this aggravates you, I'm sorry. Language evolves despite the digging in of your heels.
If that is the case, why do large format cameras create superior images to even the best small formats?
The sensors themselves are larger; a 10mp APS-C sensor (like the 50D's) is about half the area of a piece of 35 mm film, while a medium format sensor is many times larger, so the same ten megapixels are each much, much larger. If you were to increase the number of sensels until the sensel size was the same as the 50D's, you'd have (if I recall) about a hundred million of them, at which point they'd act the same. Until then, being larger, they are better behaved for many reasons; they have larger wells, so they can handle more dynamic range; the collect more photons (more area) so that the same degree of shadow exposure on a medium format sensor represents many more photons, and therefore a better signal to noise ratio, given the same noise levels in the row and column amplifiers, shot noise in the well, and so on. Another benefit is that because the sensels are larger, the demands on the lenses are less strict; the airy disc is huge compared to where it cause trouble for the 50D's sensor, so one can go to very high f-stops (and get great DOF) without encountering diffraction problems. Medium format is, in a word, wonderful. As long as you can afford the gear. Making large sensors ramps up the cost in a nonlinear and very wallet-unfriendly fashion.
No, that's not the case. The 50D, at 15mp in APS-C format only begins to encounter diffraction limits due to the airy disc just under f/8 with a 4.6 um sensel. The 40D ran into this at f/9 with a ~5.6 um sensel; presuming an APS-C sensor again in the next generation, there's plenty of room to go smaller and still have practical f-stops available that do not suffer from diffraction, thereby allowing maximum resolution at the focal plane. And of course, that doesn't stop anyone from using higher f-stops, it just means that they get to trade DOF for ultimate sharpness.
What is happening, however, is that because these smaller grids are so demanding, we're finding out just which lenses are the sharpest. That's actually an enjoyable evolution, and the answers aren't always what you'd expect. For instance, Canon sells lenses from $100 to $6000 and above. One of the sharpest in the line is the ~$400 (street price) 85mm f/1.8 -- and we're also seeing what happens when people use the wrong lens to evaluate a really high resolution sensor; DPReview used a 50mm f/1.4 lens to try to evaluate the 50D's ability to resolve detail, and they roundly blew their evaluation as a result.
Sigh. No, it definitely isn't. What you're doing here is taking the argument to extremes where it doesn't apply to current DSLR technology, which is not what the discussion was about.
Inactive areas (edges or otherwise) of sensels don't intercept photons in current generation sensors. There are microlenses over the sensels that direct photons outside the actual detector region into it, so that they are not lost. These lenses, at approximately 4.6 um square in the 50D (which sports the highest density sensor that I am aware of), aren't even *close* to a physical limit where they are losing significant numbers edge photons and significantly reducing the sensitivity at the well. The sensel's actual detector is smaller than the area of the grid it requires, but that's not the determining factor in its ability to collect photons over the unit area of its grid consumption. The microlens is. How good are the microlenses at capturing photons all across the sensor and directing them into the sensels? Well, Canon's comfortable calling them "gapless"; something to ponder over.
Having disposed of your error in asserting that smaller sensels lose significant numbers of photons at the edges, sensitivity to photon count is entirely a separate issue from the sampling resolution of the array; sensors next to each other of necessity take samples that do not resolve data that is in between them, but instead average that information to the degree that any photons actually arrive. The information is subsumed in the average, and cannot be recovered in its original form. By increasing the sensel density, we are placing detectors along a tighter spacing, and the samples will differ, whereas they would not have before; more discretely resolved information is passed along by the system as a result.
All of this has to account for other things in the system, such as continuously (relative) lowering diffraction limits of the lens with the increase in sensel density. I'm well aware of this. However, given informed use of good glass, knowledge of onset of significant spread outside the single sensel, an f-stop may be chosen that resolves detail at the focal plane with sensels the size of the 50D's or even smaller. The implication being there is still some room to go.
Before someone jumps on the aliasing bandwagon, again, granted these are physical limits we must deal with, but within the scope of a properly filtered increase of the sample rate, detail still increases just as I said.
And before you got going, the conversation was about actual, practical, existing camera technology. That's what made it interesting and useful. Please, let's return to that scope; if you want to argue about.001 um square sensels and the incipient failure of microlens technology 25 generations down the road, fine, but I'm not your guy. I'm talking about current technology and speculating about the next generation or two -- and that's all I'm interested in talking about.
One actual consequence of shrinking the sensels given any particular technology is that fewer photons go into any one sensor; given that the technology is constant, several of the noise sources also remain constant, and the signal to noise ratio drops on a per-sensel basis. This is the real elephant in the room, and it is what is likely to actually determine the end of the megapixel wars, when that actually gets here. We will develop technologies that are as low noise as we can practically manage, we'll match those with the smallest sensels we can make work, with the most dynamic range we can arrange for (both total photon count and the sensel's ability to discriminate between low counts and photon to electron conversion noise and so forth go into this limit), and that'll be the end of it.
This is only true if you *print*, which is not a given. Any person can benefit from these resolutions if they simply look at the image on a monitor. Printing, while again, nothing wrong with it, imposes many limitations. Personally, I very rarely print -- no call for it, no wall space, no particular interest.
To say that "The average pocket point-n-shoot type will not see any benefit from more megapixels" is silly. I can think of many mundane examples. Shoot a crowd; desire to look at one face more closely. If you shot with a low res camera and/or lens, you're going to be disadvantaged as compared to a higher res one. Any situation where you can't control the distance; for instance, a few nervous deer; you've got five seconds to shoot and you've got a short lens on. If you have a *sharp* lens and a high res camera, you can get them. Otherwise, too bad. Again, nothing to do with printing, and the benefit accrues to any shooter. I could literally go on for pages.
The presumption that the only task of a DSLR is to make prints or "whole images" is flawed right out of the gate, and so is any absolute declaration that depends upon it. That's just *one* thing you can do.
No, I'm not confusing anything. The more resolvable sensels there are on a sensor, when combined with any given lens, the more detail — magnification — will end up in the resulting image. This is physics. You can argue over terminology, that is, you may not want to characterize increasing sensel density as "magnification", but the end result of magnification is more detail, and this is also a beneficial end result of higher sensel density.
for example, when you shoot a photo of you diatom, do you take a photo of the whole ocean with nanometer scale pixel resolution and then blow it up till the diatom is visible. No! you simply maginify till the camera is capturing the volume the diatom is in, then smap the photo using a small number of pixels.
This is fine to a point, however in practical applications, additional detail capability in the camera's sensor reduces the need for longer lenses and consequently is very useful indeed. If for no other reason than getting more detail on your diatom.
There is a wide range of practical sensor detail that can be engineered into modern DSLRs, and your argument doesn't serve to limit that range at all.
Slashdot Mirror
Because people prefer to think more about how to have an Internet connection all the time. Which is why iPods, cellphones, Palms and even my bloody DVR have Internet jacks, and why I can connect walking around or in my car, even in my rural area of America. It's also why I have wifi access in my home and business, why the local coffee shops have wifi, and in fact, why the local *pizza hut* has free wifi.
'Tis simply all-around better to be connected. And most of us don't need control of the software, as most of us aren't programmers anyway, or programmers interested in fussing with any particular software-as-service. Those programmers that do need access, already have it -- the program authors. RMS is missing the boat. Or perhaps has simply fallen off it.
In response to the announcement of Microsoft's innovative 3-application limit, Apple corporation has said it will release a version of OS X that will allow only one application to run at a time, but in a more friendly and artistically enhanced environment than Windows Reduced Vista(tm.) Apple announced the special version late Sunday evening, at a special event entitled "You're the One." Steve Jobs emerged from his semi-retirement to explain how Apple's invention of this one-to-one relationship between users and applications would "revolutionize computing." Jobs stated that the new OS would also herald a return to the one-button mouse, single monitors, and Apple's new "One-at-a-time" network stream technologies.
Overnight, the Linux community, leveraging its well known security advantages and high speed development based upon open source and developers active in all time zones at once, has released a beta of "Linux Zero", which they claim is the most secure operating system in the world, and the least confusing, by virtue of its enforcement of zero applications running. Linux authority Linus Torvalds said "if an application can't run, it can't bring worms or viruses into the system. In addition, user interaction is now limited to pressing the power button." Waxing optimistic, he went on to say that "We think even Windows users can learn to do this." He told this reporter "In fact, the price is zero, too!"
An unconfirmed rumor also developed this weekend of an OS that is so carefully and explicitly restricted that consumers interaction with it is limited to attempting to install it; as the rumor goes, completing the installation requires permissions that users simply do not have available to them. Such an operating system would provide the ultimate consumer safety net. When asked to comment, both Jobs and Torvalds derided the rumor as being propaganda. Both OS mavens insisted that technology wasn't up to such a challenge yet. The rumor, however, persists.
When contacted by the press for comments on these new developments, Intel explained that multi-core processors were designed specifically for reduced application counts. It is only now that the leading OS manufacturers are revealing their deep strategies for the decade of 2010 that Intel is able to comment on the real rationale for multiple cores. Technical Leader Sanji Ramahasmiran" laid out several reasons why systems with few- or single-application loads would benefit directly from multiple cores. He said "Our new 8-core dies will allow switching the same single task cyclically from one core to another, thus reducing the activity levels to 1/8th that of single-core designs and operating in a greener fashion, contributing less to global warming, and simplifying programmer APIs in any properly designed operating system."
Simply as a personal observation, I always enjoy seeing how competition ensures that corporations compete for the marketplace by leveraging their core competencies and working to out-do one another. The end users always benefit. No matter who your favorite OS manufacturer is, the industry finds a way to work to bring you the latest developments. Isn't technology wonderful?
Ah, but there are other factors as well, such as the fact that a large generator system is much more efficient at turning petroleum into energy than a car's ICE is, and that once the cars are taking energy from the grid, the means of generation can be weaned away from petroleum into any number of other sources without any change in the design of the car.
Pure electric cars make the most sense, period, of any system. Hybrids are interim solutions at best, and things like hydrogen based engines make almost no sense at all - no infrastructure, low efficiencies, etc.
The only technology that is lagging here is batteries / ultracaps. There's already a battery company with a 10-minute recharge time for medium performance batteries; they're ramping up production right now (yes, they're actually making batteries.) Here's a car based on them. Should EEstor or one of the other Ultracap companies come through with fast charging systems, we're off and running. There are no other problems remaining to solve that don't fall to engineering we already know how to do (for instance, a local charging station that builds it's stored charge slowly, overnight, when rates are low and utilities are looking to load balance, and dumps to your car in minutes.)
It is totally about the energy storage system in the vehicle.
So... in your view, products should never be researched and manufactured until your average bear can afford them?
That would be really upsetting news to people who made the first... oh, computers... televisions... CD players... PMPs... DSLRs... video cameras... multichannel stereos... general coverage receivers... GPS units... satellite television receivers...
You know. Things that come under the heading of "stuff that we've never had before because it requires high technology."
So I have a question. If these things aren't initially developed for people with deeper pockets, then where is the money to develop them supposed to come from? Over to you.
Both metamoderation and moderation are a 100% waste of time on Slashdot.
Great posts are often lost at low ratings, and terrible posts get modded up. Slashdot editors pursue vendettas against various posters, and anonymous posts, regardless of content, are rarely modded at all.
This is a great site, with great content, but the only way to really experience that is to read at -1 and completely ignore the moderation, which simply does not work.
I don't read the article that way at all. The article says (emphasis mine):
I read that as saying that the non-fractal math which can resolve velocity, is not, because it is not fractal, also resolving the position because the coarse, non-fractal math falls off the edge if you try and intercept the fractal in two non-fractal dimensions. The implication is that if you use fractal dimensions for the math, you'll get both answers instead of falling into holes in the foam. That's why he talks about "getting the co-ordinates wrong."
The idea here is that the world is deterministic, just not in the three dimensions we'd perhaps like to think it is. It never made sense that a particle with a known position had no velocity; if it's moving, it's moving. So since it is moving, the fault is in how we're measuring it. This may show why such a fault in our measuring approach exists, and how to measure without the fault. That's what makes it so interesting (in this specific case... of course, if the fractal idea is correct, it means a lot more than that as well!)
"Where'd the damn turtle go?"
"Ah, it fell off the edge of the universe again." Start over from the flat spot on that atom, would you?
Well, the point of the article is that if the underlying structure of the universe is fractal, then it shows why, for instance, you can measure the position or the velocity of an electron, but not both; the general idea is that instead of a linear reality, the universe exists along a fractal edge, and answers derived using current quantum methods are literally falling off the edge because they're not finely enough resolved - they don't take the foaminess of the edge into account, so they miss the answer and land in a space that literally isn't part of the real universe - they're undefined. This is an illuminating and interesting idea, and it may point directly to how we could measure both at the same time, which would make a lot more sense to some of us. Me included.
He's not incorporated all of quantum theory into his fractal idea, so this is far from certain, but it is a lovely idea.
So, the problem wasn't that God was playing dice with the universe, rather, it's just a nice Julia set?
Einstein must be rolling in the dimensions of his grave. Fractionally, of course.
In my experience - which is considerable, I'm oooold - at 35, he won't have an age problem. That's not old enough to trigger the insurance companies to really mess with the company's expense of keeping him around under the current insurance setup. And who knows, by then, health care may look somewhat different.
A lot of ageism in tech companies is not being willing to pay for the experience an older employee usually brings to the table; but he's fresh out of school, so that doesn't apply to him. It seems to me that the odds are he'll do ok. He'll also have to accept starting wages, of course.
I went that way, basically have a collection of primes. 30mm, 50mm, 85mm, 100mm(macro), 200mm(L). Don't have the 135mm(L), but that's next on my list. All but the 30mm are EF; the 30mm would have to be replaced if I go FF.
All of these are low light capable; the least among them (100mm, 200mm) are f/2.8, the rest either f/1.4 or f/1.8. That 135mm(L) I want is f/2.
I have some zooms -- the 100-400mm(L) is a great lens, but it's also a daylight lens... and I have the surprisingly good 18-55mm EF-S IS, but again, that's a daylight lens. I don't use either one very often.
The two downsides to all this are the heavy camera bag and the need to swap, which comes up fairly often.
As for the economy... not a good time to be starting most businesses... I think its much, much worse than the media or the politicos are admitting.
Yes, you are. From TFS:
Medium format gear is out of reach for the vast majority, and has no more relevance to them than do the cameras in the Spitzer space telescope.
But you knew that. Right?
Oh, that? That poster is simply wrong.
We simply aren't at a point where the lens is the limiting factor, even for APS-C sensors. For medium format... not even in the ballpark.
Yeah... I almost bought a 5D, but that whole black dot as a consequence of sensel overflow (overexposure) really made me unhappy. As did the fix (filling them with white so that a rightward white bloom replaces the rightward black dot propagation.) That's really not a good thing for someone who shoots astro, as I do. The amount of dynamic range in an astro photo often far exceeds the ability of the camera to represent it, and so overexposure on point light sources is basically a given. I have enough trouble dealing with coma and chromatic aberration as it is; I don't need to add blooming to my list of challenges.
I'm hoping for something that suits my interests this fall (60D?) or perhaps I'll hold out until the 5DmkIII or whatever it turns out to be. I'll have to dump all my EF-S lenses if I go FF, though, and that's kind of depressing. The up side is they tend to retain their value... it's just the exercise of selling it isn't very appealing.
When I said I wouldn't go past 15 mp, I was speaking in the context of APS-C sensors. My fault for not being that little bit more specific.
Very true... but bigger sensors also carry a heavy price premium. 50D street is about $1000; 5DmkII street is about $2700. And medium format... ouch. :)
Excellent post; kudos.
Not in the context of DSLRs, they haven't, and that's what the discussion, as per TFS, is about.
sensel. Not my invention, I assure you.
Pixels emit data. Pixel is a neologism coined from picture + element.
Sensels collect data. Sensel is a neologism coined from sensor + element.
They're not doing even remotely the same task. Consequently there is a practical reason to discriminate between them. If you are under the impression that anything arranged in a grid that does anything with light, regardless of task, is a pixel, then I suggest you get over to Wham-a-lart and ask them for some linoleum pixels and see what that gets you.
If this aggravates you, I'm sorry. Language evolves despite the digging in of your heels.
The sensors themselves are larger; a 10mp APS-C sensor (like the 50D's) is about half the area of a piece of 35 mm film, while a medium format sensor is many times larger, so the same ten megapixels are each much, much larger. If you were to increase the number of sensels until the sensel size was the same as the 50D's, you'd have (if I recall) about a hundred million of them, at which point they'd act the same. Until then, being larger, they are better behaved for many reasons; they have larger wells, so they can handle more dynamic range; the collect more photons (more area) so that the same degree of shadow exposure on a medium format sensor represents many more photons, and therefore a better signal to noise ratio, given the same noise levels in the row and column amplifiers, shot noise in the well, and so on. Another benefit is that because the sensels are larger, the demands on the lenses are less strict; the airy disc is huge compared to where it cause trouble for the 50D's sensor, so one can go to very high f-stops (and get great DOF) without encountering diffraction problems. Medium format is, in a word, wonderful. As long as you can afford the gear. Making large sensors ramps up the cost in a nonlinear and very wallet-unfriendly fashion.
No, that's not the case. The 50D, at 15mp in APS-C format only begins to encounter diffraction limits due to the airy disc just under f/8 with a 4.6 um sensel. The 40D ran into this at f/9 with a ~5.6 um sensel; presuming an APS-C sensor again in the next generation, there's plenty of room to go smaller and still have practical f-stops available that do not suffer from diffraction, thereby allowing maximum resolution at the focal plane. And of course, that doesn't stop anyone from using higher f-stops, it just means that they get to trade DOF for ultimate sharpness.
What is happening, however, is that because these smaller grids are so demanding, we're finding out just which lenses are the sharpest. That's actually an enjoyable evolution, and the answers aren't always what you'd expect. For instance, Canon sells lenses from $100 to $6000 and above. One of the sharpest in the line is the ~$400 (street price) 85mm f/1.8 -- and we're also seeing what happens when people use the wrong lens to evaluate a really high resolution sensor; DPReview used a 50mm f/1.4 lens to try to evaluate the 50D's ability to resolve detail, and they roundly blew their evaluation as a result.
Sigh. No, it definitely isn't. What you're doing here is taking the argument to extremes where it doesn't apply to current DSLR technology, which is not what the discussion was about.
Inactive areas (edges or otherwise) of sensels don't intercept photons in current generation sensors. There are microlenses over the sensels that direct photons outside the actual detector region into it, so that they are not lost. These lenses, at approximately 4.6 um square in the 50D (which sports the highest density sensor that I am aware of), aren't even *close* to a physical limit where they are losing significant numbers edge photons and significantly reducing the sensitivity at the well. The sensel's actual detector is smaller than the area of the grid it requires, but that's not the determining factor in its ability to collect photons over the unit area of its grid consumption. The microlens is. How good are the microlenses at capturing photons all across the sensor and directing them into the sensels? Well, Canon's comfortable calling them "gapless"; something to ponder over.
Having disposed of your error in asserting that smaller sensels lose significant numbers of photons at the edges, sensitivity to photon count is entirely a separate issue from the sampling resolution of the array; sensors next to each other of necessity take samples that do not resolve data that is in between them, but instead average that information to the degree that any photons actually arrive. The information is subsumed in the average, and cannot be recovered in its original form. By increasing the sensel density, we are placing detectors along a tighter spacing, and the samples will differ, whereas they would not have before; more discretely resolved information is passed along by the system as a result.
All of this has to account for other things in the system, such as continuously (relative) lowering diffraction limits of the lens with the increase in sensel density. I'm well aware of this. However, given informed use of good glass, knowledge of onset of significant spread outside the single sensel, an f-stop may be chosen that resolves detail at the focal plane with sensels the size of the 50D's or even smaller. The implication being there is still some room to go.
Before someone jumps on the aliasing bandwagon, again, granted these are physical limits we must deal with, but within the scope of a properly filtered increase of the sample rate, detail still increases just as I said.
And before you got going, the conversation was about actual, practical, existing camera technology. That's what made it interesting and useful. Please, let's return to that scope; if you want to argue about .001 um square sensels and the incipient failure of microlens technology 25 generations down the road, fine, but I'm not your guy. I'm talking about current technology and speculating about the next generation or two -- and that's all I'm interested in talking about.
One actual consequence of shrinking the sensels given any particular technology is that fewer photons go into any one sensor; given that the technology is constant, several of the noise sources also remain constant, and the signal to noise ratio drops on a per-sensel basis. This is the real elephant in the room, and it is what is likely to actually determine the end of the megapixel wars, when that actually gets here. We will develop technologies that are as low noise as we can practically manage, we'll match those with the smallest sensels we can make work, with the most dynamic range we can arrange for (both total photon count and the sensel's ability to discriminate between low counts and photon to electron conversion noise and so forth go into this limit), and that'll be the end of it.
This is only true if you *print*, which is not a given. Any person can benefit from these resolutions if they simply look at the image on a monitor. Printing, while again, nothing wrong with it, imposes many limitations. Personally, I very rarely print -- no call for it, no wall space, no particular interest.
To say that "The average pocket point-n-shoot type will not see any benefit from more megapixels" is silly. I can think of many mundane examples. Shoot a crowd; desire to look at one face more closely. If you shot with a low res camera and/or lens, you're going to be disadvantaged as compared to a higher res one. Any situation where you can't control the distance; for instance, a few nervous deer; you've got five seconds to shoot and you've got a short lens on. If you have a *sharp* lens and a high res camera, you can get them. Otherwise, too bad. Again, nothing to do with printing, and the benefit accrues to any shooter. I could literally go on for pages.
The presumption that the only task of a DSLR is to make prints or "whole images" is flawed right out of the gate, and so is any absolute declaration that depends upon it. That's just *one* thing you can do.
From TFS (emphasis mine):
No, I'm not confusing anything. The more resolvable sensels there are on a sensor, when combined with any given lens, the more detail — magnification — will end up in the resulting image. This is physics. You can argue over terminology, that is, you may not want to characterize increasing sensel density as "magnification", but the end result of magnification is more detail, and this is also a beneficial end result of higher sensel density.
This is fine to a point, however in practical applications, additional detail capability in the camera's sensor reduces the need for longer lenses and consequently is very useful indeed. If for no other reason than getting more detail on your diatom.
There is a wide range of practical sensor detail that can be engineered into modern DSLRs, and your argument doesn't serve to limit that range at all.