Jump to content

Pixel density and lens quality


michael_young3

Recommended Posts

<p>A recent thread here asked about IQ issues on a 7D, and some of the discussion linked back to Merklinger's article on the Luminous Landscape (http://www.luminous-landscape.com/reviews/cameras/50d.shtml). To paraphrase, the pixel density of a 50D equates to 39 MP if extended to full frame. He concludes that image defects in his early tests were due entirely to the higher spatial resolution, and disappeared once the image was downsampled to more normal size.</p>

<p>The 7D's pixel pitch of 232 px/mm works out even higher, 46 MP if full frame.</p>

<p>Can this be right? P&S cameras with plastic gum drops (presumably, figuratively, and comparatively) for lenses have pixel densities ranging from mid-30 to 43 MP/cm^2. By comparison, the 5D2, 7D, and 50D have densities of 2.4, 5.4, and 4.5 MP/cm^2, respectively. If Merklinger's 24-105/4L is not sufficiently sharp for even the 50D's sparsely spaced pixels, how are the $100 consumer cameras able to capture any image at all? Does it seem reasonable that a P&S houses the sharpest lenses made?</p>

<p>Also worthy of comment, HM goes on to say the 50D's pixel pitch makes the sensor diffraction limited at f/7.5, giving the relationship as N=1600/pixelpitch in millimeters. The 7D's 232 px/mm predicts diffraction softened images at smaller than f/6.9. Again in stark contrast, the P&S with 43 MP/cm^2 works out to 655 px/mm, making them essentially pinholes at smaller than f/2.4. (I guess that answers the first part!)</p>

<p>Comments or thoughts?</p>

 

Link to comment
Share on other sites

  • Replies 61
  • Created
  • Last Reply

Top Posters In This Topic

<blockquote>

<p>Does it seem reasonable that a P&S houses the sharpest lenses made?</p>

</blockquote>

<p>What makes you think that is happening? At 100%, P&S 10MP images look close to a 5MP image seen at 200%. There's a point beyond which no more detail is being captured and the extra MP are just a marketing ploy. I think this ends up biting the industry in their rear-end as they first advertise cameras based on one single criteria and then they have trouble selling higher-end cameras to beginners that think their money should be buying them more MP. I think the fact that 7D comes up with 17MP instead of 10/12/15MP is sort of confirming this - camera companies still try to sell more MP.<br>

I think that what you usually see clearly on high resolution DSLRs like 7D is that zoom lenses are inferior - their limitations become much more noticeable than before, because the sensor is good enough to capture that detail - on P&S it's just not capable of doing that, so if you can't see much detail, you won't see aberrations either.<br>

I wish photosite density would be kept constant across cameras, sort of like how it was during the film era across various film formats - film was the same, except you got less of it on 135 than on 120. This would put an end to many useless discussions we have around noise performance, FF vs APC, etc - you would just pick the camera given your resolution needs (I am ignoring DOF issues, as I've always found those quite irrelevant - they're an advanced topic that most camera buyers won't ever know about - but noise is easily noticeable).</p>

Link to comment
Share on other sites

<p>Of course I don't think P&S lenses are sharper. Gumdrops, remember?</p>

<p>Maybe we can focus on the diffraction limited aperture for the sensors.<br>

5D2: 155 px/mm, f/10.3 (also 1DsMk3)<br>

7D: 232 px/mm, f/6.9<br>

50D: 213 px/mm, f/7.5<br>

300D: 136 px/mm, f/11.6</p>

<p>This roughly matches my experience with the 7D and 300D, and also with scrutiny of online ISO lens charts shot with the relevant bodies: diffraction limited at f/11 on FF; f/8 on 1.6x crop. It was explained to me (apparently in error) as a measure of quality of the lens that diffraction limit is reached at smaller f/ stop. I see now that this is not so, and is merely a function of pixel pitch. Maybe it's old news to you, but it certainly is a light bulb moment for me.</p>

 

Link to comment
Share on other sites

<p>Really cool info, Michael. In my own totally-unscientific tests with my 7D I was able to pretty clearly see the difference in sharpness between 5.6 and 8, where it got sharper and shaper as I went from 1.4 to 5.6, but then seemed to degrade as I went to 8 and beyond. It's interesting to see that the math backs it up with that figure of f/6.9.</p>

<p>Fortunately for me, I very rarely shoot beyond f/1.4 or f/2, so it's not really a concern for me 99.9% of the time.</p>

Link to comment
Share on other sites

<p>Further thoughts:<br>

Color film (est. 9MP): 125px/mm, f/12.8</p>

<p>What are the consequences? For one, I can stop studying lens charts. The f/4 lenses I might have been considering are diffraction limited less than 2 stops down from wide open on the 7D. And for sure, TC 2x already pushes them past diffraction limit. I feel somewhat liberated.</p>

<p>Last.... I don't know that I agree completely with stopping the MP wars. We've only just very recently reached levels where the returns are diminishingly small. At any given moment in the past 5 years, somebody somewhere had already been saying that. So my very sharpest shots will be taken at wider than f/8. I'm weeping a little inside, but I can live with it. At worst, it's still better than "smaller" (but not lesser) bodies. At it's best, within this newly recognized limit, I already have shots that are pleasingly sharp, surpassing my very high expectations.</p>

 

Link to comment
Share on other sites

<p>Michael</p>

<p>my comment or thought is that everything is a swing, what you gain in one thing means you loose in another.</p>

<p>IF a lens for full frame has slightly less resolution than one for EF-S then it will be also benefited by the lower magnification applied to get the sensing area of 36x24 scaled up to a print of 8x12 than if taken on a 50D. The benefits of more pixels becomes a detraction with respect to noise related to the therefore smaller pixels. RN Clark has an <a href="http://www.clarkvision.com/articles/does.pixel.size.matter2/index.html"><strong>excellent page on this</strong> </a> on this.</p>

<p>If you really desire to enlarge more I believe strongly there is no substitute for larger capture format.It brings more with it than you think.</p>

<p>Lastly everyone forgets a key issue of lens testing, that is the process. <strong>You see when you test a lens you place it at a distance from the test chart which is related to the focal length</strong> . So when you use for instance Norman Korens test chart you place the lens at 51 times the focal length from the chart. For a 21mm lens this is 1.2 meters ... do you often photograph subjects at such distance?</p>

<p>Place the same chart at a more typical photographing distance and examine again what you get. Then compare this with a much lesser lens (say an EF24 f2.8) and tell me you can see any difference? I am willing to bet you also don't.</p>

<p>There are many many aspects to a good image, some of these technical issues may even be significant in that.</p>

Link to comment
Share on other sites

<p>Michael,<br /> It is not helpful to isolate these figure,<br /> <em>"Maybe we can focus on the diffraction limited aperture for the sensors.<br /> 5D2: 155 px/mm, f/10.3 (also 1DsMk3)<br /> 7D: 232 px/mm, f/6.9<br /> 50D: 213 px/mm, f/7.5<br /> 300D: 136 px/mm, f/11.6"</em></p>

<p>They only mean something when referencing same sized enlargements. Diffraction is the same on all sensors at any given aperture. Think about it, the light doesn't know or care why it is being bent through a small opening, but it will react the same whenever it passes through a same sized opening regardless of the medium behind it. The reason the limiting figures are given differently is because the assumption of the refference is to arrive at the same sized COC at any given print size. You need to enlarge smaller sensors more to get to the same sized print. But if you have a tripod mounted 600mm f4 at f16 and take a picture of a bird with a 7D and then take the body off the lens and mount a 5D MkII and take a picture. Now enlarge those two images so the bird is the same size, not the print, just the bird, the images are identical, with identical diffraction and depth of field. Diffraction can only be referred to in the context of COC and that is only relevant when referring to reproduction size and viewing distance.</p>

<p>To question one area of the whole is impossible without a background understanding on all the other relevant aspects. So you need to understand COC, Airy discs, Nyquist limits, the concept of diffraction limited optics (good), optic resolution issues (imperfections, design limitations, manufacturing tolerances etc),sensor resolution, AA filter interference frequency etc .</p>

<p>The resolution limiting aspects are sensor density and how that is interfered with by the AA filter if present, diffraction and lens resolving power. If your lens is diffraction limited, ie it is not limited by its optics only by diffraction, then you are reduced to two factors, if your sensor is pixel dense enough to to resolve diffraction then you are limited to one, your image sharpness will be limited only by diffraction blur (your enlargement size and viewing distance). You normally end up with a trade off between the three.</p>

<p>If the optic is not limiting but the sensor is you are into Nyquist limit resolution until you stop down enough that the sensor can resolve the increase in COC. Most macro lenses will be in this example.</p>

<p>If the optic is limiting but the sensor isn't then you are done, once the diffraction gets worse than the optic limitations you are doubly done :-), most P&S's fall into this category.</p>

<p>Now with pixel dense sensors either the optics resolution or diffraction are normally the limits. Kit zooms will normally be optic limited initially but then move to diffraction blur. So stopping down past f10 or so the diffraction blur is worse than the lens blur.</p>

<p>Hope this helps, Scott.</p>

Link to comment
Share on other sites

<p>A few thoughts:</p>

<ul>

<li>The density of the P&S cameras means little when it comes to determining what does and doesn't make sense. The very high photosite densities there are more about marketing than about photography.</li>

<li>Watch out for that term "diffraction limited." It is easy to misconstrue what that is telling you. When a higher photosite density camera is said to have be diffraction-limited at a larger aperture, this does not in any way imply that a) that is the best aperture to use, or b) that the camera performs less well at smaller apertures when compared to cameras with smaller diffraction-limited apertures.</li>

<li>Since it is impossible to imagine a photosite density that will provide resolution <em>equal</em> to that of lenses (for a range of reasons I won't describe), there are only two realistic choices: a) the sensor is capable of out-resolving lenses, or b) the lenses are capable of out-resolving sensors. I think most would prefer the former over the latter.</li>

<li>Resolution is not the only issue. Even in situations in which the sensor out-resolves the lens, there is some thought that there can be advantages, for example you might obtain smoother gradients. If the cost of high photosite density sensors drops enough, cameras might do away with AA filters.</li>

</ul>

<p>Finally, all of this is largely moot except in the most extreme situations: very careful and skillful photographer almost certainly using a tripod and making huge prints.</p>

<p>Dan</p>

Link to comment
Share on other sites

<p>Photography doesn't begin or end with DSLRs and P&S cameras... Digital MF is up to 60 mpixel (on a 40x53mm sensor) and I've been using a 39 mpixel back (on a 37x50 mm sensor) for a long while. Now, with most MF lenses being crap in terms of raw resolution compared to most pro-level 35 mm format optics, I should have ran into all kinds of "my back outresolves my lenses" issues but this is simply not the case. In fact, I can hardly think of a lens (both Mamiya MF and Canon EF L and even some non-L optics) that is not up to the PhaseOne 45+ back or an 1Ds3/1D3/1D4. And I'd welcome well documented, real-life - no charts and brick walls, samples of good quality optics not being up to the 7D or 50D sensor in terms of resolving power.</p>

<p>I can't say that I don't enjoy talking about how many photons will fit on a pinhead, but most of this stuff has absolutely no practical bearing on photography.</p>

<p>Edit:<br>

Merklinger. Is that the guy who on the LL site proclaimed in 2003 (?) that the output from Canon D60 is on par with, or better than, MF film photography..?</p>

Link to comment
Share on other sites

<p>IMO we are getting to the point where we the market will demand a revolution in lens design. Which I think is great news for us all. It will take photography where it could not go before. I can't say where it will go but I'm eager to see it unfold. Before that we will see a surge in applications that will try and correct the defects of lenses, somewhat like we have now but more effective and better work flow. Additionally, sensor design will also evolve at some point to make better use of the limits of lens design.</p>

<p>That said, I like reading luminous landscape and think they do a good job overall and I respect a lot of their opinions. However, they are not engineers and their speculation on this subject is more entertainment than fact. They may be some what correct in their estimation of current and past implementations of technology, they certainly can't say exactly what principles are in play or say precisely what is causing the IQ issues they are seeing. Could be that the 24-105 they used was a little off? I think it is good they are questioning the cameras on the market, it makes the manufactures try harder to provide better products.</p>

<p>IMO after about 15 MP technique and lens choice make major impacts to IQ when viewed at 100%. And, at 21 MP a lot of peoples technique can't keep up. Does that mean that we should settle for that level? Of course not, asking for more is going to push the engineering and technology.</p>

Link to comment
Share on other sites

<p>Matt,</p>

<p>I disagree, taking the stereo analogy, when CD's came out they were all sampled at huge rates. When people moved that music over to computers they realised there was little difference between native sampling rates at over 1,000bps and 192bps, indeed all but the keenest ear and on good equipment can tell the difference between full sized CD's and resampled 128bps audio. After the hype and when given a choice people are pragmatic and have common sense. Very few people print out images that are resolution limited at 21MP, and those that are have higher spec'ed equipment to buy anyway.</p>

<p>21-25MP is a practical overkill limit for most users, the lenses we have available, as Michael says, are generally more than up to the job and as you say, most people are not sufficiently proficient or interested in their photography to warrant or utelise even more MP. The next push will be ever higher iso, broader DR, better video etc, none of which will be limited by current lenses.</p>

<p>I agree with Michael 100% too. It is fun understanding the academic physics behind the process but until you actually run into image destroying lens/resolution/diffraction problems (which I have never seen in normal use, apart from a few very crappy lenses) then it is just a thought process and we shouldn't worry about it much.</p>

<p>Take care, Scott.</p>

Link to comment
Share on other sites

<p>Allow me to further refine the thought. These are all good and worthy thoughts, but the point I was making is quite a bit more simple. Given no other reason to select one aperture in preference to another, we select one that fits the lighting while giving us good sharpness. It's a very common thought (isn't it?). Until yesterday evening, because I had been reading lens charts as though they held an answer, I would have set f/8, and maybe stop it down a third or two further for just a bit more DoF. Two months ago, I would have set it even smaller, f/11 to f/16. A lifetime of shooting sharp pictures says that's a good target if the light allows. The 7D is my first to fall far from this range. A very casual browse through my LR catalog bears it out. f/5.6 and f/8 good; smaller not so good. If we define "good" as eye pleasingly sharp at 1:1.</p>

<p>Who should care? I care. I was unknowingly making an ill-informed choice, based on assumptions that don't apply to the current situation. Readers outraged or perplexed by Darwin Wiggett's blog should care. His proof of the 7D's inconsequence is based on f/16 shots. You might care, if you've ever set it on a tripod in bright daylight and locked the mirror to shoot the sharpest image possible. If you carry more than one lens in your bag, you probably care. For sure, if you can distinguish the quality of a prime versus a zoom, you care deeply. I think that covers everyone not carrying a P&S for its acceptable IQ.</p>

<p>So, is this a universal relationship, diffraction limited N=1600/pixelpitch? I believe it is, insofar as the common definition of DoF is universal and acceptable. The 1600 comes from the same CoC that defines DoF. Nobody to my knowledge argues its validity strenuously.</p>

<p>This is news. The effect is easily observable in casually shot images. Diffraction degrades images shot on the 7D at apertures smaller than the relatively large f/6.9. This differs considerably from prior experience, where f/11 was a reasonable and safe estimate. Whether or not it shows up in 20x30 prints is a separate question. It's readily evident at 1:1, a very important magnification because that's how I like to view some images. Surely, if mediocre was good enough, I wouldn't have bothered to have this thought.</p>

<p>But all that aside, it still doesn't explain the phenomenon that started all this. I have a Tokina 80-200/2.8 ATX lens that displays ruinous CA everywhere in the frame on the 7D, but was not objectionable or even noticeable on a 300D. None of the above seems to explain this.</p>

Link to comment
Share on other sites

<p>Michael,</p>

<p>It seems to me you are missing your own point.</p>

<p>When you enlarge you 7D image to 1:1, or pixel for pixel, you are enlarging it a lot more than you were your 50D. If you print out your 7D and 50D the same size the diffraction is exactly the same, it is just as blurred in both images. The diffraction is no more or less with the two cameras, but now, due to the increased pixel density, your diffracted blur is better resolved by the 7D. The effects of diffraction on the COC are entirely dependent on the reproduction size and viewing distance. You can't enlarge your 7D image more and say it shows more diffraction than your 50D, it doesn't, they are identical.</p>

Link to comment
Share on other sites

<blockquote>

<p>IMO we are getting to the point where we the market will demand a revolution in lens design.</p>

</blockquote>

<p>Got any <strong>practical</strong> (i.e. based on know physics and not involving liquid nitrogen cooling...) ideas how to vastly improve transmission of visible light through a complex optical system..?</p>

Link to comment
Share on other sites

<p>That's all true, Scott. Nonetheless, two shots taken side by side at different apertures will show differences in the detail captured in the parts at critical focus. If I have no other reason to choose one over the other, meaning DoF and other considerations aside, I want the aperture capturing the most detail.</p>

<p>The difference isn't theoretical. It's quite pronounced in the tools I most often use to view pictures. Looking at the 2 shots side by side, one at f/7.1 and one at f/11, I won't say, "Oh, they're the same. It was certainly good enough when I had only 6 MP. Watch what happens when I downsample it and gain back all that wasted disk space. You'll see. They're the same. They have to be: the lens doesn't know or care. All that extra detail? You didn't need it. How big were you going to print anyway? It isn't all that good to begin with."</p>

<p>That doesn't sound like a conversation I would have. It actually went like this: "What the CR*P? I'd better go reshoot that."</p>

 

Link to comment
Share on other sites

<p>Michael,</p>

<p>You need a reality check on your viewing technique then. The diffraction and blur and DOF and CA are exactly the same on your 300D and your 7D, you are disappointed that your 7D doesn't look better than your 300D at 3x the magnification, that is crazy, it will look a damn sight better than the 300D at the same magnification!</p>

<p>It's analogous to saying, my old car went 60mph and did 30mpg, my new car is rubbish because although I go everywhere at 180mph it doesn't still do 30mpg. If you went 60 in your new one it would do more than 30mpg.</p>

<p>The 7D is way way better than the 300D, it suffers diffraction exactly the same as the older camera, you can't look at it 3x closer and not expect to see more blurring, you will see three times more, but it will give you more detail at the same size. 1:1 for the two cameras is not the same size, you have to understand that.</p>

Link to comment
Share on other sites

<blockquote>

<p>The difference isn't theoretical. It's quite pronounced in the tools I most often use to view pictures.</p>

</blockquote>

<p>Photography has always been about the intended output so if you shoot for people who only pixel-peep @100% (on a 72 DPI monitor...a joke in itself if you get my drift...) then all I feel is pity. The truth is that on real life output - regardless of its size - on something different than a 72 DPI monitor neither you nor the the most acutely sensitive person would be able to tell apart a well-executed picture taken - say at f/8 and f/11, even though (say, for 1Ds3) f/10 is "diffraction limit." Now, of course we (photographers) should be aware of the technical aspects of the craft (after all photography is a technical endavour) but we also should keep the whole technical aspect in perspective (i.e. does it matter for the intended application...) And the size of the intended output is not always the most important consideration: say, if you shoot for glossies (esp. higher quality mags) you'll soon realize that you really need to put an extra "oomph" in your pictures because, despite their smallish size, they are intended for "nose surfing" - close-up viewing from the "average" reading distance. And so forth... It's good to be aware of all this stuff but I draw a line at demonizing it..</p>

Link to comment
Share on other sites

<p>Re. 300%: 173%, actually, or equivalently 58% going the other way. It's a squared relationship.</p>

<p>The analogy is a pretty good fit. You might not notice if you didn't have a reason to care. If you didn't understand the relationship or misremembered it in your haste, the disparity with your expectations will drive you nuts.</p>

Link to comment
Share on other sites

<p>This is crazy. If it means something to you that f/6.9 on the 7D is as sharp as you'll get, I'm glad I was able to help. It's contrary to long previous experience, not intuitively obvious, and explains a lot. Marginalizing the observation as "shooting only for peepers", or somehow turning it into a comparison with the 300D is just nuts.</p>

<p>I like these grapes; they're sweet and juicy.</p>

 

Link to comment
Share on other sites

<p>Here with my old 4x5 digital scan backs; the lens requirements are only modest. The 35 and 50 megapixel backs have a 7x10cm scan area. The small guy shoots a 5000x7000 pixel image; the big guy a 6000x8400 pixel image. 10cm is 100,000 microns; thus the smalll guy has a pitch of 14 microns; the big guy 12 microns. If the lens is *good enough* for the sensors pitch; one can get a sharp image.<br>

The poor assumption you all have led your self into a mental rut is assuming that a 100 buck P&S digital has a bad lens; since your assumption is gravely in error; you conclusions are faulty and in error. Optically it is easier to make a good lens for a smaller format camera that a larger one. P&S digitals are made by the millions too; thus the tooling costs per unit sold are way less than a low volume lens.</p>

Link to comment
Share on other sites

<p>Here with my old 4x5 digital scan backs; the lens requirements are only modest. The 35 and 50 megapixel backs have a 7x10cm scan area. The small guy shoots a 5000x7000 pixel image; the big guy a 6000x8400 pixel image. 10cm is 100,000 microns; thus the smalll guy has a pitch of 14 microns; the big guy 12 microns. If the lens is *good enough* for the sensors pitch; one can get a sharp image.<br>

The poor assumption you all have led your self into a mental rut is assuming that a 100 buck P&S digital has a bad lens; since your assumption is gravely in error; you conclusions are faulty and in error. Optically it is easier to make a good lens for a smaller format camera that a larger one. P&S digitals are made by the millions too; thus the tooling costs per unit sold are way less than a low volume lens.</p>

Link to comment
Share on other sites

<p>Here with my old 4x5 digital scan backs; the lens requirements are only modest. The 35 and 50 megapixel backs have a 7x10cm scan area. The small guy shoots a 5000x7000 pixel image; the big guy a 6000x8400 pixel image. 10cm is 100,000 microns; thus the smalll guy has a pitch of 14 microns; the big guy 12 microns. If the lens is *good enough* for the sensors pitch; one can get a sharp image.<br>

The poor assumption you all have led your self into a mental rut is assuming that a 100 buck P&S digital has a bad lens; since your assumption is gravely in error; you conclusions are faulty and in error. Optically it is easier to make a good lens for a smaller format camera that a larger one. P&S digitals are made by the millions too; thus the tooling costs per unit sold are way less than a low volume lens.</p>

Link to comment
Share on other sites

<p>Here with my old 4x5 digital scan backs; the lens requirements are only modest. The 35 and 50 megapixel backs have a 7x10cm scan area. The small guy shoots a 5000x7000 pixel image; the big guy a 6000x8400 pixel image. 10cm is 100,000 microns; thus the smalll guy has a pitch of 14 microns; the big guy 12 microns. If the lens is *good enough* for the sensors pitch; one can get a sharp image.<br>

The poor assumption you all have led your self into a mental rut is assuming that a 100 buck P&S digital has a bad lens; since your assumption is gravely in error; you conclusions are faulty and in error. Optically it is easier to make a good lens for a smaller format camera that a larger one. P&S digitals are made by the millions too; thus the tooling costs per unit sold are way less than a low volume lens.</p>

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...