Understanding lens resolution

[danielleetaylor]

<p>Stephen Cumblidge - <em>It would be interesting to run a double blind test where one shoots the same scene with a 5D II and a 7D and see who can tell which is which from 20x30 prints.</em></p>

<p>Been there, done that. They can't until the ISO goes above 800.</p>

[g dan mitchell]

<p>"The old rule of thumb is that for close inspection you want 300 dots per inch."</p>

<p>That rule of thumb turns out to be quite irrelevant for photographic prints. Most printers regard about 180 resolution as pushing the lower limits for acceptable high quality prints.</p>

<p>I also want to second the comment that things other than lens resolution have a far greater impact on creating photographs that will stand up to large print sizes. I'm not saying the lens resolution is not important, but that many lenses provide more than sufficient resolution for quite large prints, and a whole range of other factors play into the final image quality. </p>

<p>I differ with some who claim that results from a 18MP cropped sensor camera will be essentially the same as those from a 21MP full frame camera. The 7D can, indeed, produce very fine image quality. However, the fact that larger film/digital formats have the capacity to create greater system resolution is essentially not something worth arguing about. </p>

<p>On the other side of that statement, unless you work rather carefully, know what you are doing in the post-processing phase, shoot with excellent technique, pay a lot of attention to things like accurate focus and camera stability and then push the upper boundaries of print size and quality... you probably won't see a difference. For example the differences at 12" x 18" prints are very, very subtle and would only be seen by an experienced viewer doing careful side by side comparisons. But somewhere above this size, depending on a bunch of variables, the will be differences.</p>

<p>On the other, other hand... a camera like the 7D offers some important functional features that you might find more appealing and useful than the increased resolution potential of a full frame camera like the 5D2.</p>

<p>Dan</p>

[JDMvW]

<p>Many people, even though they are not my friends, praise my brick wall pictures as being my personal best.</p>

[danielleetaylor]

<p><em>However, the fact that larger film/digital formats have the capacity to create greater system resolution is essentially not something worth arguing about.</em></p>

<p>Theoretical capacity is not the same as realized capacity. If all things were equal a FF sensor would out resolve an APS-C sensor. But it's quite obvious that pixel density is not equal between the 5D mkII and 7D/60D. And pixel density = resolution.</p>

<p>To put the system difference between the 5D mkII and 7D into perspective, 7D RAW files in ACR yield a higher extinction resolution on a chart test than 5D mkII RAW files in DPP or 5D mkII JPEG files. The difference in total pixels at the sensor is less than the difference introduced by choice of RAW converter. It is also most certainly less than technique, post processing, and lens choices. If you get everything else in the photographic chain right, you are not going to observe a difference in large prints due to the small difference in total pixels on the sensors.</p>

[philip_wilson]

<p>Daniel - it depends on what we are interested in. The 7D has higher sensor resolution as it can resolve more line pairs per mm of sensor area (lens issues and diffraction not withstanding). So if I only use 38% of the 5DII sensor area then yes the 7D has higher resolution. But what I care about is the resolution of my image and that is in line pairs per image height and here the 5DII has higher resolution that the 7D. Assuming that you print the image from the entire sensor then the 5DII has higher resolution than the 7D. For more information see numerous sources including Popular photography, Amateur photography, Bob Atkins, dpreview, DXO mark, Photozone etc... In my earlier post I mentioned that on a 20x30 print with the 5DII and 7D image side by side (even at 100 ISO) you can see the difference. However, this really does not matter as the moment you separate the prints so you can no longer directly compare it is no longer obvious. By the way I believe the human eye can resolve around 550 dpi - so clearly there is still benefit in printing above 300 dpi</p>

[g dan mitchell]

<p>"Extinction resolution" or not, a larger format 21 MP camera is going to have the system capacity to outresolve a 18 MP (or even a 21MP) smaller sensor across the image width. With lenses of equal resolving power, the larger system can "resolve more line pairs."</p>

<p>To avoid getting into an impassioned argument about "which format is better," let me clarify a few things again:</p>

<ol>

<li>Both formats can produce finely detailed images when used by a competent photographer.</li>

<li>Unless one works carefully and critically <em>and</em> makes prints that push the upper boundaries of print size, the difference is likely to be quite small and probably not of consequence to most shooters.</li>

<li>There are other factors besides system resolution that would argue for or against either format as the best choice for a particular photographer. For example, the 7D provides some functionality that is not found on the 5D2 that might be more important to some photographers. And some photographers might prefer the wider range of usable apertures (as per limitations due to diffraction blur) of the full frame system.</li>

</ol>

<p>It is not necessary nor accurate to claim that a) one type of system is always better than the other or that b) they are equal. They are not equal and the differences can be important - but they can cut both ways.</p>

<p>Dan</p>

[ilkka_nissila]

Photozone's MTF50 figures are not line pairs per mm but per sensor dimension. The figures for full frame are higher

because it is a physically larger sensor and also the pixel count is greater in the case of the sensors mentioned. So

the files are more detailed, as they should be. Now, if you compare a full-frame lens on a crop camera to an ef-s lens

on the same camera, it is entirely possible that the lens designed for the smaller sensor has an advantage in the

center. Lenses that cover a larger image area are typically not as sharp as those optimized for the smaller area. Buy

lenses that work well on the cameras that you intend to use! It's as simple as that.

 

I agree that lens tests do not tell the whole story. However I consider "sample variability" a part of the quality of the

lens and there should be no excuses if the manufacturer allows it in their output. If a review site gets a lemon the

manufacturer deserves all the bad rap they get as a result. More expensive and simpler lenses typically have less

variability for obvious reasons (mechanical stability).

[bob_estremera]

<p>Thanks for all the input here. Even you Brad. Hope your K&E returns soon. And I'm sure you got very excited when the topic turned to 'extinction resolution'.<br>

The more I've researched and the more of my own images I look at critically since all of your help, the more I'm leaning toward the 5D with a few good primes that offer even performance across the lens when stopped down. A used 24-70 2.8 can be had for a good price at KEH also.<br>

When I can finally spring for it, I think I'll have a setup that I can just forget about technical stuff and shoot instinctively with only the image in mind.<br>

Thanks again everybody. This has been a big help.<br>

Bob</p>

[g dan mitchell]

<p>Ikka, your theoretical notion about the optical superiority of the EFS breed of lenses over EF and similar lenses on cropped sensor cameras is not borne out in actual practice. The main reason for producing EFS lenses is not that they are going to be optically better than other lenses.</p>

<p>This is not to say that there aren't some very fine EFS lenses, including the 17-55 f/2.8, the 10-22, the macro, etc. Others I didn't mention can also be quite credible lenses. But that doesn't mean that the EFS design is going to produce higher resolution than a non-EFS design.</p>

<p>The "advantage in the center" theory is one that is tossed around a lot without sufficient critical thinking. It is true that resolution falls off, along with light transmission, as we move toward the corners of the frame on any format, and that the farther from the center we have to go the more the resolution might decline. But this ignores several important things:</p>

<ul>

<li>The FF system starts out with higher center resolution of the camera/lens system by comparison to the center resolution of the cropped sensor system.</li>

<li>Even at the edge of an APS-C sized section of the FF sensor, the overall resolution in lp/frame width terms is still higher.</li>

<li>It continues to be higher beyond this point. On some lenses, and at some apertures, it remains higher all the way to the corners, while on others the situation might be different.</li>

</ul>

<p>In a worst case - a pretty bad lens on a FF body - it might be possible for the resulting image to have noticeably diminished corner resolution at the largest aperture in a very large print. In more common cases, and especially when shooting at more typical apertures, the corner performance on the FF system will be as good and more likely better than that of the smaller sensor system.</p>

<p>Disclaimer: I'm not saying that full frame is necessarily "better" than crop - there are other factors that can make this stuff unimportant and other factors more important for individual photographers. But the arguments that smaller sensor systems will equal (or even exceed!) the resolution capabilities of larger sensor systems are specious.</p>

<p>Dan</p>

[pete_ferling]

<p>+1 with Ken and Brad.</p>

<p>It a workflow issue, and expectations. Whether your are deliberate from the onset, or find yourself in a moment that you would expect to hit a decent 20x30 print. Then good technique, movement in the scene, enough lighting to keep at lowest ISO to have motion stopping shutter speed, etc. That's only half of it. We now have software and printing tools that extract a great deal of data, and create more perceived sharpness, even from what was otherwise junk antique lenses from the film days.</p>

<p>I look at my old film shots, from decent FD lenses, and I now have a second opinion when compared to my digital. Yes, those old 35mm prints are sharp, but only sharp as can be from the tools we had to deal with back then.</p>

<p>Granted, it's hard for some folks to understand just exactly what is possible now, when they have nothing to compare to how it was even ten years ago. Fortunately, many of our clients are not so well versed. I've sold many prints at 24" to customers whom never considered it was shot with a 40d and a kit lens.</p>

<p>Life is so much easier when you simply let this stuff go, and judge an image for how cool it looks.</p>

[ilkka_nissila]

<p><em>Ikka, your theoretical notion about the optical superiority of the EFS breed of lenses over EF and similar lenses on cropped sensor cameras is not borne out in actual practice.</em></p>

<p>I didn't say that EF-S lenses are optically superior. But the designer has an advantage if they only need to cover a small area they <em>can</em> make the lens superior in that area, if they choose to. Whether this actually happens with actual commercial lenses is another matter and product positioning and marketing play a key role here. Canon doesn't want to market high end EF-S lenses because they want the more expensive lenses to be always better even if they have to fight physics to get there. It would be much easier to make slow lenses better quality than fast lenses, for example, if marketing didn't decide that the more expensive lens <em>has to </em>be better quality to avoid confusing customers. This is a great pity. (The 70-200/4 is one exception to the general practice followed by most manufacturers.)</p>

[danielleetaylor]

<p>Philip - Daniel - <em>it depends on what we are interested in. The 7D has higher sensor resolution as it can resolve more line pairs per mm of sensor area (lens issues and diffraction not withstanding). So if I only use 38% of the 5DII sensor area then yes the 7D has higher resolution. But what I care about is the resolution of my image and that is in line pairs per image height and here the 5DII has higher resolution that the 7D. </em></p>

<p>I'm not talking about cropping a 5D mkII image to give the same FoV as on a 7D. I'm talking about just the sensors. The 5D mkII has a 21 MP sensor and the 7D has an 18 MP sensor. And that 3 MP difference, at the end of the photographic chain, amounts to less than one's choice of RAW converter. (See DPReview's chart tests.)</p>

<p>I'm not going to get into the print debate again. Suffice it to say that I've never had anyone tell me which came from which, either with large prints or with 100% crops presented repeatedly in this very forum. If somebody could readily tell you which print came from which, then you probably did not sharpen the 7D file appropriately. (Amateur Photographer UK came to the same exact conclusion, btw.)</p>

<p><em>By the way I believe the human eye can resolve around 550 dpi - so clearly there is still benefit in printing above 300 dpi</em></p>

<p>As with everything else, the human eye's ability to resolve detail depends on detail contrast. Given B&W line art the human eye can resolve detail well past 1200 dpi. Given average color photographic detail, you've hit the limit at around 250 ppi. (Note I said average. A silhouette for example, having higher contrast, might benefit from higher printing resolution. I can recall one print test I performed with a few friends using MF drum scans. The color detail was done by 250 ppi, but the model's black hair against the white backdrop saw improvement to around 500-600 ppi.) </p>

[danielleetaylor]

<p><em>a larger format 21 MP camera is going to have the system capacity to outresolve a 18 MP (or even a 21MP) smaller sensor across the image width. With lenses of equal resolving power, the larger system can "resolve more line pairs."</em></p>

<p>Only at the extinction resolution of the lens (MTF10). But sensors are no where near that pixel density.</p>

<p>Out of camera a 21 MP FF sensor will yield a higher MTF50 lpmm result than a 21 MP crop sensor, all other things being equal. The average photographer would not describe this as "higher resolution" but as "greater sharpness" because the same detail is present in the crop file, just at a lower contrast (i.e. MTF40). The trick is that they are so close, because the formats are so close in size, that simply sharpening the crop output will yield the same MTF50 result. This is pretty much true for sensors within a certain MP range of each other as well.</p>

<p>Which brings us back to this: your choice between Canon's 21 MP FF sensor and their 18 MP crop sensor is less important than: technique; lens; RAW converter; post processing skill; printer; paper choice. At the end of the photographic chain it's non-observable. It's hardly observable when you isolate everything else and look for it in a chart test.</p>

<p><em>Unless one works carefully and critically and makes prints that push the upper boundaries of print size, the difference is likely to be quite small and probably not of consequence to most shooters.</em></p>

<p>They're quite small and not of consequence even when you work this way.</p>

[danielleetaylor]

<p>Ilkka is correct regarding lenses. It is easier to design a lens for a smaller format and maintain a given level of quality through the manufacturing chain. Also, there are examples of crop camera/lens combinations out performing FF camera/lens combinations where both lenses are high end, disproving the claim that FF "always has" a resolution advantage and would only be worse with a poor lens. My 17-40L is not a "poor lens", but my 7D+Tokina 11-16 out performs it on a 5D mkII.</p>

<p><em>But the arguments that smaller sensor systems will equal (or even exceed!) the resolution capabilities of larger sensor systems are specious.</em></p>

<p>There are now multiple crop bodies which out resolve the early generations of FF bodies.</p>

[g dan mitchell]

<p>So Daniel, then we can extrapolate from your logic that if cropped sensor DSLRs can produce the same system resolution as full frame sensor DSLRS that MF digital systems must not produce better resolution than full frame DSLRs and that four/thirds cameras must be able to resolve as well as cropped sensor cameras and, what the heck, point and shoot size sensors can resolve as well as four thirds, all of which logically proves that...</p>

<p>... point and shoot size sensor cameras can logically produce system resolution equal to that of MF digital systems with similar numbers of photosites.</p>

<p>When I come up with a logical chain that points in directions like that, I always go back and re-check my assumptions.</p>

<p>Dan</p>

[mark_pierlot]

<blockquote>

<p>There are now multiple crop bodies which out resolve the early generations of FF bodies.</p>

</blockquote>

<p>This is a red herring. The debate in this thread concerns the <em>current</em> generation of full frame and crop bodies.</p>

<p>Obviously the 7D out-resolves the 5D, for example; however, here we're comparing the resolution of, e.g., the 7D to that of the 5D<em>II</em>.</p>

[g dan mitchell]

<p>Mark, that's why I didn't even bother to respond to it. Pretty silly comment, wasn't it...</p>

[danielleetaylor]

<p><em>When I come up with a logical chain that points in directions like that, I always go back and re-check my assumptions.</em></p>

<p>As you should. It is false to assume that what we observe between 35mm and APS-C holds true for all larger differences between two formats. (Fallacy of multiplication.)</p>

[danielleetaylor]

<p><em>This is a red herring. The debate in this thread concerns the current generation of full frame and crop bodies.</em></p>

<p>Your reading comprehension is poor. I was not addressing the debate but a <strong>specific claim</strong> made by a <strong>specific person.</strong></p>