- Status
It doesn't look like there is anything surprising there. The D800 has about 50% more resolution than either 5D II or III, so it stands to reason its going to produce more lp/ph. The results look spot on as well, with optical aberrations dominating before @f/4, and diffraction dominating after @f/4. I'd have liked to see a 7D thrown in the mix, as its sensor actually has a higher spatial resolution than the D800. I wouldn't expect it to get hugely better results @f/4, but assuming the lenses were still capable of more than the sensor, I would expect the 7D to eek out just a bit more performance from the same lenses.
Upvote
0
According to this test, if you're a portrait photographer, or someone who shoots at wide open apertures, the 14 megapixel advantage of the D800 over the 5DIII is negligible? I rarely shoot wide open, but I just want to make sure I'm interpreting the data correctly.
Too bad they didn't get to put the 5DIII through the same battery of tests. If I carried through with my D800 purchase, which I'm not at this point, the 70-200 VRII was going to be my primary workhorse lens, so it's interesting to see how poorly the performed compared to the primes. Plus, with long lenses I usually shoot stopped down to f/8 or f/11, negating the D800's resolution even more.
Too bad they didn't get to put the 5DIII through the same battery of tests. If I carried through with my D800 purchase, which I'm not at this point, the 70-200 VRII was going to be my primary workhorse lens, so it's interesting to see how poorly the performed compared to the primes. Plus, with long lenses I usually shoot stopped down to f/8 or f/11, negating the D800's resolution even more.
Upvote
0
V8Beast said:
Yes, that would be accurate. "Best" or "maximum" resolution only occurs at one point (the normalization point for optical aberrations and diffraction), which is usually around f/4 these days. Thats why I've argued so hard about resolution in the past, as its a rather misunderstood concept. For common apertures, which tend to be more wide open for portraiture, and more stopped down for things like landscape, macro, and architectural photography, your likely to be well below the lenses maximum potential. I would say the only photography where I might regularly be around f/4 of f/5.6 is wildlife and sometimes bird photography, however more frequently I'm at around f/8 there as well.
Upvote
0
Even if most people know about using sturdy tripods, mirror lockup and cable release, I am not sure they are aware of exactly how much they need to pay attention as the MPs increase.
This article : http://www.josephholmes.com/news-sharpmediumformat.html talks about 45 MPs on a medium format system so it details procedure that are a bit extreme (and some irrelevant) for 35mm but still it is an eye opener in some ways. For example, he says to let the mirror come to rest for 6 seconds after it flips up to avoid vibrations. I don't know many photographers who wait even 2 seconds.
This article : http://www.josephholmes.com/news-sharpmediumformat.html talks about 45 MPs on a medium format system so it details procedure that are a bit extreme (and some irrelevant) for 35mm but still it is an eye opener in some ways. For example, he says to let the mirror come to rest for 6 seconds after it flips up to avoid vibrations. I don't know many photographers who wait even 2 seconds.
Upvote
0
DavidRiesenberg said:
There will be some pain at first, but after they learn, it will be second nature. It is not more difficult than a 7D or other 16mp+ crop.
Upvote
0
V8Beast said:
Technically speaking, when the spatial res. of the sensor the same as or less than the spatial res. of the lens, the more likely you are to get what appears to be a "razor sharp" shot. That apparent sharpness comes at the cost of resolution, though. Sharpness and resolution are disjoint things...one refers to how fine the degree of details can be resolved, while the other refers to the clarity of those details (MTF refers to both of those things at a given contrast ratio, which is what makes MTF valuable.) You can get "sharp" images at any resolution (if you want to get extremely technical, a 1 pixel sensor could take a sharp shot of anything, it just wouldn't be a useful image.)
The real question is whether you can get sharp shots at the resolution of your sensor. Inside of the "band of maximum resolution", which is maybe a 1 2/3rds stop area around the f-stop a lenses resolution peaks at (usually f/4, 173lp/mm, lets call that the ideal aperture), the sensor should always be sharply recording whatever the lens is projecting. Assuming proper exposure and generally following the 1/focalLength rule for camera shake (including the use of IS/VR lenses):
- Softness below the ideal aperture will usually be due to the lens...optical aberrations like chromatic aberration, longitudinal fringing, spherical aberration, etc.
- Softness above the ideal aperture will usually be due to diffraction.
At apertures above ideal, you need to get about 2-3 stops beyond the initial DLA of the sensor before that really becomes a problem. If a sensor's DLA is f/8, it wouldn't be until about f/16 or possibly even f/22 that diffraction visibly affects the image. (Even at f/22, diffraction is often likely to have LESS impact on IQ than optical aberrations at say f/2.8 or wider when the ideal aperture is f/4.)
Consistent softness at all apertures can occur when the sensor significantly outresolves the lens. When the sensor outresolves the lens, a single line pair resolved by the lens will usually cover more than just two rows of pixels (or four, in the case of bayer sensors). You can usually realize improved resolution up to the point where the sensor outresolves the lens by about two fold (and even farther, however the cost to produce such a sensor will generally outweigh the cost of a better lens, and a better lens will help you realize greater gain than a better sensor at that point.) You may not, however, continue to experience improved sharpness as you increase sensor resolution. If you keep improving sensor resolution without improving lens resolution, the sensor is going to consistently start resolving enough detail to show the attenuation (...falloff) between line pairs, so instead of high acutance (strong contrast along the edge between a light line and a dark line) you'll get lower acutance (kind of like using a gausian blur along the edge between a light line and a dark line, creating a smooth gray transition rather than a hard edge.) Technically speaking, you are still resolving more detail, its just that beyond a certain point, the detail you are resolving appears less perfect (less sharp) than at a lower resolution.
In the band of maximum resolution, there is still a lot of room to gain more resolution with higher resolution sensors. However at other apertures, and in the corners of lenses, sensors are already outresolving lenses to some degree. At f/22, lens spatial resolution is limited to 68lp/mm MTF50 (good contrast...i.e. decently sharp). The Canon 7D with 4.3 micron pixels (116lp/mm), and the Nikon D800 with 4.8 micron pixels (102lp/mm), are getting closer to outresolving f/22 resolution two fold (136lp/mm), and are likely already outresolving f/2.8 two fold. Possibly even more so, as 900lp/ph with FF in lp/mm would be 900lp/24mm, or 37.5lp/mm. The D800 is capable of 2.8 times that, so its nearly 3x outresolving f/2.8 from the LensRentals test. The reason images might look slightly soft (i.e. not tack sharp) with a high resolution sensor like the D800's or the 7D's is because the sensors are outresolving the lenses at most apertures.
If you expect tack sharp photos strait out of the camera at any aperture (and don't print large), you are actually better off with sensors that have lower resolution. Ideally, a sensor capable of 68lp/mm would be tack sharp all the way from ideal aperture to f/22, and would only doubly outresolve f/2.8 (so it might not be tack sharp, but it should still be sharp enough.) The 5D Classic sensor has a spatial resolution of about 61lp/mm. The 5D II has a spatial resolution of 78lp/mm, the 5D III has a spatial resolution of 80 lp/mm, and the D800 has a spatial resolution of 102lp/mm. I would expect all three 5D cameras to produce what appear to be sharp shots more often than the D800 in hand-held situations. The D800 is more demanding of lenses, and it will be outresolving more apertures to a greater degree than any version of the 5D. I'm not sure I would go so far as to say you need "perfect" technique, however having better skill at hand-held photography, or the use of a tripod, will generally benefit the D800 more so than the 5D III.
Upvote
0
jrista said:
That makes sense. That said, the guy commenting on the LensRental website was complaining about the measures that must be taken to avoid vibration-induced softness, whether it's due to a long shutter speed, camera shake, mirror vibration, long focal lengths, or all of the above. If you jump up to a high MP sensor as in the D800 and experience softness due to any of those issues, anyone who understands the most elementary of photography techniques should be able to figure out what went wrong. If you can't, why on earth would you buy a $3,000 camera? To criticize Nikon about this, as if they're responsible for someone who gets in over their head with equipment they obviously don't know how to use properly, is just silly.
Upvote
0
V8Beast said:
One would generally think its elementary, however a lot of people have a lot of money, and its not always pros who buy $3000 cameras. There is a good reason why there is so much misunderstanding about sensor resolution, lens resolution, and image softness on the net (even by some professionals, who may be fantastic photographers, but generally don't understand the underlying technical, physical, and mathematical behavior of the gear they use.)
Upvote
0
jrista said:
First of all, I would like to thank you for your insightful input here. The level of your knowledge is great and makes me want to qote senator Clay in the Wire (don't know if you have seen it) 'Shiiiiiiiiiiiiiiiiiiiit'. That said. How is the 1D X in this equation? In my mind, it is at least partly a successor of the 1Ds III, and the went down on the megapixels.
Given that we are lucky enough to be able to pick your and Neuros brain in here, I would love for you to give some comments on what Canon lenses (both L and non-L) are most compatible with the 5D III and the 1D X (the latter one on what is communicated), as far as resolution capabilities of both body and lens.
Upvote
0
Quasimodo said:
First, thank you for the kind comments. Much appreciated. I like to educate, as I think it helps the community at large.
When it comes to Canon lenses, all of them perform very well, and their current L-series lenses from the last several years (lets say since the 5D II was introduced), and particularly Mark II lenses or new entrants (i.e. 8-15mm L fisheye) all perform excellently. Much older L-series lenses are still very good, and usually perform better than any non-L Canon lens (with the exception of a very few...such as the 100mm f/2.8 Macro and maybe the 10-22mm and 17-85mm EF-S lenses).
The main difference between older Canon lenses and newer ones are the types of special lens elements (UD glass, fluorite, aspheric, etc.), the quality of each lens element and the type of multicoating. Newer lenses tend to prefer fluorite elements over UD glass, as fluorite does a better job at reducing dispersion, and allows lenses to use fewer elements overall (reducing weight...swapping two UD elements for one fluorite element in the new 600mm f/4 L is how Canon reduced weight from nearly 12 pounds to just over 8 pounds.) One of the most important improvements of Canon's newest lenses is the multicoating. For decades, canon used Super Spectra multicoating, which was good and pretty standard, but not the best. Newer lenses use Subwavelength Structure Coating, or SWC. This type of coating is a form of nano coating, similar to Nikon's Nano Crystal Coat. The general idea was garnered from the design of moth eyes, which reflect a barely measurable amount of light due to their nano-scale structure. SWC is essentially the same thing, and reduces flare, ghosting and reflections to a bare minimum (in most cases, even if flare presents, its nearly invisible, and can usually be ignored in final images/prints since most viewers won't even notice it.)
If I had to recommend Canon lenses for use with their latest cameras, it would depend on the camera. I think just about any Canon L-series lens will perform superbly with both the 1D X and the 5D III. The 1D X offers a raw spatial resolution of 64lp/mm, and the 5D III offers a raw spatial resolution of 80lp/mm (factors such as microlens quality, low-pass filter, etc. will reduce those numbers slightly). Neither are high enough to worry about the sensor significantly outresolving the lens in most cases. From f/4 to around f/16, its likely the lens will outresolve the sensor to a degree (falling off the more you stop down.) You should get razor-level sharpness from both cameras.
When it comes to the likes of Canon's APS-C cameras, the story is a bit different. All of the current ones use the same 18mp APS-C sensor, which outside of f/4 and f/5.6, is going to be outresolving the lens. The more you open up or stop down, the more those sensors will outresolve. Optical aberrations are a different beast than diffraction, and even though the sensor will outresolve the lens at wide apertures, you'll encounter softness from a whole host of causes. Sometimes, you may actually want that softness...such as spherical aberration in portraiture, as it adds a nice aesthetic aspect (soft focus.) Diffraction presents more readily as classic "softness" or "lack of sharpness", however it is far more regular and mathematical, so its easier to counteract. Basic sharpening is often enough to combat one or two stops of diffraction softening. You won't really gain resolution (you'll gain a smidge), but sharpening will definitely increase acutance (contrast and hardness along the edges/boundaries of detail areas.)
If you are in the market for a 1D X or 5D III, and already have a significant investment in lenses, you should feel satisfied that your new camera will exhibit the best properties of those lenses. If you are not invested in lenses yet, I highly recommend getting only newer lenses, and the newer the better. Mark II versions of lenses, such as the 70-200mm Ls, 16-35mm L, 24-70mm L, the 8-15mm L fisheye, pretty much any one of Canon's newer telephoto L-series lenses, should all have their SWC anti-flare coating. Loss of contrast due to flare or near-nonincident light can significantly impact IQ at any aperture, and reducing flare to a minimum will improve IQ more than pretty much any other lens feature today. Better contrast, improved sharpness, and better color fidelity are the benefits of better anti-flare coating.
Upvote
0
jrista said:Quasimodo said:
First, thank you for the kind comments. Much appreciated. I like to educate, as I think it helps the community at large.
When it comes to Canon lenses, all of them perform very well, and their current L-series lenses from the last several years (lets say since the 5D II was introduced), and particularly Mark II lenses or new entrants (i.e. 8-15mm L fisheye) all perform excellently. Much older L-series lenses are still very good, and usually perform better than any non-L Canon lens (with the exception of a very few...such as the 100mm f/2.8 Macro and maybe the 10-22mm and 17-85mm EF-S lenses).
The main difference between older Canon lenses and newer ones are the types of special lens elements (UD glass, fluorite, aspheric, etc.), the quality of each lens element and the type of multicoating. Newer lenses tend to prefer fluorite elements over UD glass, as fluorite does a better job at reducing dispersion, and allows lenses to use fewer elements overall (reducing weight...swapping two UD elements for one fluorite element in the new 600mm f/4 L is how Canon reduced weight from nearly 12 pounds to just over 8 pounds.) One of the most important improvements of Canon's newest lenses is the multicoating. For decades, canon used Super Spectra multicoating, which was good and pretty standard, but not the best. Newer lenses use Subwavelength Structure Coating, or SWC. This type of coating is a form of nano coating, similar to Nikon's Nano Crystal Coat. The general idea was garnered from the design of moth eyes, which reflect a barely measurable amount of light due to their nano-scale structure. SWC is essentially the same thing, and reduces flare, ghosting and reflections to a bare minimum (in most cases, even if flare presents, its nearly invisible, and can usually be ignored in final images/prints since most viewers won't even notice it.)
If I had to recommend Canon lenses for use with their latest cameras, it would depend on the camera. I think just about any Canon L-series lens will perform superbly with both the 1D X and the 5D III. The 1D X offers a raw spatial resolution of 64lp/mm, and the 5D III offers a raw spatial resolution of 80lp/mm (factors such as microlens quality, low-pass filter, etc. will reduce those numbers slightly). Neither are high enough to worry about the sensor significantly outresolving the lens in most cases. From f/4 to around f/16, its likely the lens will outresolve the sensor to a degree (falling off the more you stop down.) You should get razor-level sharpness from both cameras.
When it comes to the likes of Canon's APS-C cameras, the story is a bit different. All of the current ones use the same 18mp APS-C sensor, which outside of f/4 and f/5.6, is going to be outresolving the lens. The more you open up or stop down, the more those sensors will outresolve. Optical aberrations are a different beast than diffraction, and even though the sensor will outresolve the lens at wide apertures, you'll encounter softness from a whole host of causes. Sometimes, you may actually want that softness...such as spherical aberration in portraiture, as it adds a nice aesthetic aspect (soft focus.) Diffraction presents more readily as classic "softness" or "lack of sharpness", however it is far more regular and mathematical, so its easier to counteract. Basic sharpening is often enough to combat one or two stops of diffraction softening. You won't really gain resolution (you'll gain a smidge), but sharpening will definitely increase acutance (contrast and hardness along the edges/boundaries of detail areas.)
If you are in the market for a 1D X or 5D III, and already have a significant investment in lenses, you should feel satisfied that your new camera will exhibit the best properties of those lenses. If you are not invested in lenses yet, I highly recommend getting only newer lenses, and the newer the better. Mark II versions of lenses, such as the 70-200mm Ls, 16-35mm L, 24-70mm L, the 8-15mm L fisheye, pretty much any one of Canon's newer telephoto L-series lenses, should all have their SWC anti-flare coating. Loss of contrast due to flare or near-nonincident light can significantly impact IQ at any aperture, and reducing flare to a minimum will improve IQ more than pretty much any other lens feature today. Better contrast, improved sharpness, and better color fidelity are the benefits of better anti-flare coating.
Thank you
I have a couple of 'old' lenses; 50 1.4, and the 135L (which happens to be my favorite). Other I have are the 70-200 F2.8 IS II, 17L TS, 16-35 II, 100 F2.8 Macro HIS, and the 24-105 F4.0L. So its a mix, but I am very happy about it. Thinking about the new 24-70 II, and the MP-E65, and at this very moment I have a friend who is in NY buying the 2x teleconverter at B&H for me, which I will use with the 70-200 and the 135.
Upvote
0
I know using the 7d, which has a higher pixel density of the D800 and closer to the 45MP MF camera mentioned earlier, that to leverage the best out of the camera, I need to use the best lenses I can afford... crap in crap out... So comparing pixel densities, can it be said Canon lenses, especially the L lenses, compared to shooting on the 7D, are better equipped for higher resolution cameras than Nikon's lenses, or am I reading too much into this?
Upvote
0
awinphoto said:
Most issues will be solved by better technique. For sure, a higher mp body will see lens flaws better, but I expect that in most cases, its just a matter of user training and experience to get sharp images. Its normal to stop down with many lenses to get the sharpest image, there tend to be more flaws at widest aperture settings.
Very few lenses produce really sharp images wide open compared to being stopped down.
Upvote
0
Mt Spokane Photography said:
Fair enough... on my 7d my sweetspot i liked to keep my exposures if i could help it was F4-5.6, which on most my lenses was wide open, but overall I had little to no issues with the resolution I was getting my set-up...
Upvote
0
Excellent write-up jrista. And a good thread. Really helps explain some of the impact certain lenses have had on my 7D.
Upvote
0
awinphoto said:
I don't know a lot of specifics about Nikon glass, however what I do know is that its on par with Canon glass. Nikon lenses coated with Nano Crystal Coat would be very similar to Canon lenses with SWC.
When it comes to the 7D, there are probably fewer lenses than not that will help the 18mp sensor (which offers 115.97lp/mm) to resolve as much as possible. The number of apertures wherein you could resolve close to the theoretical maximum sensor resolution would be very few as well. I don't know if any lenses really actually achieve "perfection" at apertures less than around f/6.3 (where spatial resolution is 117.23lp/mm, about the same as the 7D's 18mp APS-C sensor). I think many get very close, but perfection is difficult to achieve, and some degree of optical aberrations will exist until you are really pretty much just capturing light from the center of the lens.
Resolution is a rather tricky thing, too. Sensors are capable of recording a fixed spatial resolution. Lenses achieve different spatial resolutions depending on their aperture, and their best resolution falls around one specific aperture, and falls off as you open wider or stop down narrower. The actual resolution you are recording, however, is a combination of all of the elements in the system that may have an effect on spatial frequencies. You can be as general or as accurate as you want in those calculations, and if you want to get very accurate you'll need to account for usually unseen factors...like the vertical and horizontal low-pass filters, the IR filter, teleconverters, even the quality of the microlenses above each photodiode in each sensor pixel. Anything that behaves as an optical lens or filter affects spatial frequencies, and will have an impact on your final resolution. To keep things more practical, I usually just account for lenses, teleconverters, and the sensor itself (its theoretical maximum resolution, ignoring the effects of the filters in front of it.)
To get much more technical, and explain everything with math (don't worry, the math is pretty simple, and it makes a LOT of sense in the end.)
The system resolution is really what your camera as a whole is capable of recording into photographs, and its always less than the lowest common denominator. System resolution can be derived by calculating the total "system blur", or the amount that each physical component of a system affects the reproduction of spatial frequencies in combination with each other. System blur is the square root of the sum of the squares of each independent factor of the system (i.e. the square of the sensor's blur + the square of the lens blur + the square of a TC blur). In other words:
Code:
systemBlur = sqrt(blur1^2 + blur2^2 + ... + blurN^2)Its tough to figure out the blur of a lens when optical aberrations dominate. More than one type of aberration usually affects spatial resolution at that point, they are more mathematically complex, and on top of all the optical aberrations, there is STILL diffraction to account for (which does affect the overall blur, however in and of itself it usually only contributes enough to cover a fraction of a pixel, so it doesn't matter.) When a lens becomes diffraction limited, optical aberrations are minimal (and possibly completely dominated by diffraction itself), and diffraction is easier to computer mathematically. Diffraction, the bending of light around edges (in this case, the edges of the diaphragm that border the aperture), produces a specific waveform, or airy pattern. The airy pattern is dominated by the airy disk in the center, which is primarily what your resolving a point of light to. Around the airy disc are concentric rings of light with far lower intensity. As the size of the physical aperture is reduced, the size of the airy pattern increases, with the central disk increasing in size, and the concentric rings reaching farther and becoming more intense themselves. Its the growth of the airy disk that softens resolution at smaller apertures.
You can compute the size of the airy disk using the following formula:
Code:
1.22 * lightWavelength * fNumberAs you can see from that formula, different wavelengths of light produce different amounts of diffraction. It is easiest to just use the wavelength of green light to compute an average airy disk size that roughly accounts for middle-ground resolution. Bayer sensors are twice as sensitive to green light as they are to blue and red light, so it works out well in the end. If we take our f/6.3 aperture and the wavelength of green light in millimeters (@555nm), you can see that:
Code:
1.22 * 0.000555mm * 6.3 = 0.004265mmTo convert that to microns (which are easier to compare to sensor pixel pitch), you simply multiply by 1000um/mm:
Code:
blurCircle um = blurCircle mm * 1000um/mm
0.0042653mm * 1000um/mm = 4.265umHere is where things get very interesting. The 7D's 18mp APS-C sensor has 4.3 micron pixels...roughly the same size as the lenses blur circle at f/6.3. It should be easy to understand why the 7D is diffraction limited at about f/6.3...because that is exactly the point where the airy disk (the central bright point of light in the airy pattern resolved by a lens) is the same size as a sensor pixel. I believe officially the DLA (diffraction limited aperture) is f/6.9, which would allow for the airy disk to be just slightly larger than a single pixel, capable of affecting other pixels, thereby reducing resolution. (Its a bit more technical than that in reality, but roughly speaking thats the deal.)
Finally, we can determine the theoretical maximum spatial resolution of a lens in line pairs per millimeter as so:
Code:
spatialResolution lp/mm = 1000um/mm / (blurCircle um * 2)So plugging our numbers from above into that formula, we get:
Code:
1000um/mm / (4.265um * 2) = 1000um/mm / 8.53um = 117.23 lp/mmA diffraction-limited lens at f/6.3 is capable of a maximum of 117.23 line pairs per millimeter. Similarly, the 7D is capable of about 116lp/mm. The spatial resolution of a sensor can be easily computed as so:
Code:
sensor_resolution lp/mm = pixelRows l / sensorHeight mm / 2If we plug in the numbers for the 7D into that formula, we get:
Code:
3456 l / 14.9mm / 2 = 231.946 l/mm / 2 = 115.973 lp/mmAt this point, your probably figuring that you only need a diffraction limited lens at f/6.3 to resolve enough resolution to make the most out of the 7D. That would be great, but its still not quite that simple. Remember the concept of system blur from above:
Code:
systemBlur = sqrt(blur1^2 + blur2^2 + ... + blurN^2)Both the lens and the sensor are capable of resolving the same amount of detail, however the final resolution of our photograph is actually going to be much less. Our total system blur boils down to:
Code:
squrt((4.3um)^2 + (4.265um)^2) = sqrt(18.49um^2 + 18.19um^2) = sqrt(36.68um^2) = 6.056umWhile the maximum theoretical blur circle for our sensor is 4.3um, allowing for 115.97lp/mm, and the maximum theoretical blur circle of our lens is 4.265um, allowing for 117.23lp/mm, the theoretical blur circle of the lens+camera is a full 41% larger. If we plug our system blur into the necessary formulas to get spatial resolution in line pairs / mm, we get:
Code:
1000um/mm / (6.056um * 2) = 1000um/mm / 12.113um = 82.56 lp/mmOur system resolution is a meager 82.56 line pairs per millimeter!! Rather frustrating, however it might give some light to why the 7D appears "soft" much of the time. The sensor itself is producing about 40% more pixels than it needs to even at f/6.3, and the problem only gets worse the more you stop down. Similar things happen if you open the lens up too wide. If we increase our aperture to f/4, the picture improves, but its still not perfect (or even ideal):
Code:
F4Blur = 1.22 * 0.555um * 4 = 2.7um blur
F4Resolution = 1000um/mm / 185.185 lp/mm
systemBlurF4 = sqrt(18.49um^2 + (2.7um)^2) = sqrt(18.49um^2 + 7.335um^2) = 5.08um
systemResolution = 1000um/mm / (5.08um * 2) = 1000um/mm / 10.163um = 98.389 lp/mmSo what would it really take to make the BEST use of our 7D's 18mp APS-C sensor? Well, we could rearrange our blur formula a bit to compute a target blur:
Code:
targetBur = sqrt(lensBlur^2 + sensorBlur^2)
targetBlur^2 = lensBlur^2 + sensorBlur^2
lensBlur^2 = targetBlur^2 - sensorBlur^2If we plug in the numbers for our sensor and target blur when the target blur is the same as the sensor blur:
Code:
lenaBlur^2 = (4.3um)^2 - (4.3um)^2 = 0Well bugger. We would have to have infinite lens resolution (lens airy disks that are essentially infinitely small points of light) to actually achieve a total system resolution that is the same as our sensor resolution. A better, more realistic way to put it is that it is impossible for any combination of system components to actually produce a system resolution that equals the lowest resolution. System resolution has an asymptotic relationship with the lowest component resolution. Lets say we just want to get very close. Lets say we want to achieve 114 lp/mm resolution with our lens+camera:
Code:
lensBlur = sqrt((4.4um)^2 - (4.3um)^2) = sqrt(0.87um^2)
systemResolution = 1000um/mm / (sqrt(0.87um^2) =
systemResolution = 1000um/mm / 0.9327um * 2 =
systemResolution = 1000um/mm / 1.8655um = 536.05 lp/mmWe need an unbelievably stellar lens, capable of a whopping 536.05lp/mm, to improve system spatial resolution to 114lp/mm. That would be about 98% of the theoretical maximum. In terms of lens aperture, that would be:
Code:
1.22 * 0.000555um * fNumber = 536.05lp/mm * 2
fNumber = (1000um/mm / (536.05lp/mm * 2)) / (1.22 * 0.000555mm * 1000um/mm)
fNumber = (1000um/mm / 1072.1l/mm) / 0.6771um
fNumber = 0.9327um / 0.6771um
fNumber = f/1.377A perfect (i.e. diffraction limited rather than optical aberration limited) lens at an aperture of f/1.38 would be necessary to achieve 114lp/mm with the 7D sensor. To my knowledge, such a lens does not exist (or if it did, it would have to be an extreme supertelephoto lens, where most incident light is already collimated, producing very little optical aberrations to start with.) When it comes to system resolution (or system blur), you get the most benefit by improving the lowest common denominator. In this case, the sensor is the lowest common denominator when we are using f/4. As such, we could gain more system resolution by increasing sensor resolution relative to lens resolution. If we used a sensor capable of 173lp/mm, the same as a diffraction limited lens at f/4, we would have a 40mp APS-C sensor. Our total system resolution would be about 122lp/mm at f/4. As you have probably figured, a 40mp APS-C sensor would be quite a feat to manufacture, probably have some very undesirable noise and electronic characteristics, and would likely be extremely expensive. It would also likely exhibit similar softness, as ironically, the sensor is still producing images with 40% more pixels than are necessary to produce a sharp photo.
Personally, I prefer my lenses to outresolve my sensors a bit, which means the sensors never produce more pixels than necessary to create a sharp photo strait out of the camera.Outside of extremely fine detail (which for the most part full-frame cameras would be incapable of resolving in the first place), any comparison between such a camera and full-frame with the same megapixels would make it seem like the higher resolution APS-C was "soft". In reality, larger details appear soft relative to the full-frame, but your resolving finer detail overall. A real-world example (assuming all else being equal...i.e. our hypothetical 40mp sensors have the same noise characteristics, color fidelity, dynamic range, etc. despite being different physical sizes) might be shooting a portrait such that you wanted your subjects eye lashes to appear sharp, vs. shooting the same portrait where you wanted the fibers in your subjects iris to appear sharp. The eye lashes might appear a touch soft in the APS-C photo, compared to how they appeared in the FF photo...but the APS-C photo is resolving the iris itself in far more detail than the FF could ever aspire to. You could sharpen the APS-C photo a bit, and your eye lashes would be superb...although the iris may seem a bit over-sharpened now (and your exposing every single blemish of your lovely model's face in depressing detail as well!) In the end, neither 40mp camera, FF or APS-C, really matters, since your printing that amazing portrait photo on a full-page magazine spread that is a meer 10x12" in size...in which case, you could have probably done superbly well with a meager 10 or 12mp camera (possibly even less). :
Upvote
0
- Status
