Is lens contrast a tradeoff vs. resolution?

<p>In a <a href="http://visualsciencelab.blogspot.com/2011/11/brief-review-of-lens-you-probably-would.html">recent blog post</a> Kirk Tuck writes, "You can design a lens for high resolution or high contrast but not necessarily both. Nearly every lens is a compromise between those parameters."</p>

<p>Is that true? He also seems to suggest that higher contrast is not necessarily a desirable characteristic of a lens, which I'm not sure I agree with.</p>

<p>I just happened to test a 55/2.8 Ai-S Micro vs. a 60/2.8 AF-S Micro last night on a D700. My quick findings are that at ƒ/8, there's little to choose between them in terms of resolution; they're both pretty amazing. But the modern lens is much more contrasty, and a good third of a stop brighter at the same aperture setting.</p>

<p>This wasn't a particularly rigorous test, but my one-sample experience doesn't support Kirk's assertion. Any other opinions?</p>

<p>I disagree, most of my lenses have both resolution and contrast. My Contax G lenses were the best, but any Nikkor professional lens is just about as good. </p>

<p>It makes sense to me, but I`m not qualified to answer. Anyway, I take resolution as the "opposite" to sharpness... and sharpness use to be relative to acutance, or the cutting capacity.</p>

<p>Also, I agree that higher contrast is not always desiderable. In fact, softer lenses has always been a choice for portraiture. And softer lenses use to have better bokeh, at least in my experience.</p>

<p>About your own test; I wonder if such lenses, or other high performing ones can show a difference on a "not demanding" D700. I wonder if the limit is on the camera sensor, not on the lens`capacity.</p>

<p>BTW, Voigtlander has a current lens made in two "coating" versions, coated and non-coated (or it was multicoated and single coated?), specifically designed for <em>connoiseurs</em>; those b&w film shooters supposedly will prefer the uncoated one for the better gradation. This preference looks to be common amongst b&w specialists.</p>

<p>Personally, I don`t find one better than the other. I like both worlds; vintage lenses are great for its smoothness (I`m thinking on my LTM Canon 50/1.2), and current Leicas for its -impressive- microcontrast.</p>

<p>resolution is more a sensor feature, no? IDK, every really sharp lens i have is also pretty contrasty. i could see maybe a macro lens being designed more for sharpness, but i cant say i have a lot of lenses which are contrasty but not sharp.</p>

<p>I understand resolution as the capacity of resolving fine details, on film, sensor or lens. The higher resolution the finer gradation.</p>

<p>There is no good reason why good contrast and good resolution can't exist together in a lens. The exception to this would be if you also wanted a huge image circle or angle of view, but for modest image angles there's really no incompatibility between resolution and contrast. You might make a tenuous link between the complexity of a lens's design and its contrast, but that would be about as far as it goes.</p>

<p>In any case, contrast is affected by a whole host of things quite separate from the basic optical design of the lens. Edge blackening of elements, internal baffling, coatings and camera body design will all affect image contrast, as will the focused distance and ratio and distribution of light and dark in the subject. Not to mention condition and cleanliness of the camera body and lens, interaction of lens reflective surfaces with the inevitable reflection from the digital sensor or film. The list goes on....</p>

<p>Contrast and resolution are not absolute competitors to each other in lens design, but they are rather linked. Take a photo for example, put it in your favorite editing program, and play with the contrast. You will see that as you dial up the contrast, you lose detail, because a small patch that was many different shades of almost-black has now become all-black. I was stuck in an endless battle with a previous photo professor: I enjoyed lower contrast images, because it brought out the subtle detail of everything in my shot, while my professor was always pushing me to use contrast filters to make a scene pop. I think he just wanted to show me the flipside of the coin, so that I knew what I was missing out on by choosing to make my images the way that I did.</p>

<p>Contrast and resolution are closely related. High resolution corresponds to high contrast at high spatial frequencies i.e. small scale. In fact online MTF tests measure contrast at a given spatial frequency as a measure of resolution.<br>

A lot of the difference in the contrast at low spatial frequencies is due to the coatings, although stray light can also reduce contrast which will be more noticeable when the lens is wide open.<br>

You compared a modern lens to a rather old one, and I would not be surprised if differences in coating technology were the main explanation for what you see. Both are primes, and will have good optical designs. There will also be differences between manufacturers too.</p>

<p>I am wary of saying that Kirk Tuck is mistaken, but I believe there is no reason why you cannot design a lens with high resolution and high contrast, and a high quality optic will have both. I do recall that with microscope objectives, planar ones (having a flat field) are in general said to have lower resolution than non planar ones, and this I assume is because the designer aims for an overall improved performance at the expense of the centre of the field. Perhaps that is what Kirk Tuck means. Certainly I was surprised at the high contrast of an old x100 non planar microscope objective (nothing special, just an old achromatic Prior) I own compared to a very modern planar x100 objective. But such a small sample does not say much.</p>

<p>i have noticed that my least-sharp lenses dont have much contrast, either. i think Kirk is applying some physics principle here, which may be technically true, but gets lost in real-world relevance.</p>

<p>R.J., I assume one thing is <strong>micro</strong>contrast, and another different, <strong>macro</strong>contrast. </p>

<p>Like in Ariel`s example, I think you are refering to the macrocontrast, while the unsharp masking tool could refer to microcontrast.</p>

<p>Mark, I can't fathom exactly what he meant to say, either. </p>

<p>After a quick scan through his blog, my best guess about what he meant to say comes from looking at the statement in question in the context of his previous sentence:<br>

<em>"...camera makers started creating lenses that added snap and sparkle back in at the expense of longer tonal ranging and high resolution rendering. You can design a lens for high resolution or high contrast but not necessarily both. Nearly every lens is a compromise between those parameters..." </em></p>

<p>One of my guesses is that what he really meant to say was:<br>

<em>"...camera makers started creating optical systems that added snap and sparkle (ie, contrast) at the expense of a longer usable tonal range. You can design an imaging system for a wide tonal range or for high contrast, but not necessarily both. Nearly every imaging system is a compromise between those parameters..." </em></p>

<p>The only other possibility that I can think of is that he is somehow trying to get across the idea that for all lenses, if you ask a lens to resolve higher spatial frequencies, the contrast goes down. In other words, he may be sorta-kinda describing in simple terms the general shape of MTF curves, all of which start with high contrast at low resolutions (low spatial frequencies) and wind up with no contrast at high spatial frequencies. I don't think many people would use the term "tradeoff" to describe this particular inverse relationship, but perhaps it was just a poor choice of words.</p>

<p>If it isn't either of these two explanations, I don't have a clue what he is talking about, and it may be just a hastily-written, inaccurate comment. </p>

<p>Tom M</p>

<p>PS - BTW, there is a great little introductory article on MTF (with lots of examples of MTF curves) right here on photo.net ( http://www.photo.net/learn/optics/mtf/ ) by our own Bob Atkins.</p>

 

<p>I am wary of saying that Kirk Tuck is mistaken...</p>

<p>Why? He's a practitioner/vocational school teacher, not an optical engineer. He presents no theoretical basis for his contention. It's little more than anecdotal observations that are no different than the observations posted by other people in this thread, and no more grounded in fact or theory either.</p>

<p>If one were to actually try to make the case that "You can design a lens for high resolution or high contrast but not necessarily both. Nearly every lens is a compromise between those parameters.", then one would reasonably start with a Laplace transform of the optical performance of lenses. Kirk can't do this because he doesn't know how and if he did, he would prove his basic premise wrong. </p>

<p>Again, I think Kirk Tuck is right.</p>

<p>Think on minute details; we`re talking about <em><strong>micro</strong></em>contrast. A resolutive type lens will show all this fine details in a subtle gradation, providing a "flat" image, while the contrasty lens will show just clear and dark borders, even "eating" subtle details (turning them to "clear" or "dark").</p>

<p>Everybody wants a lens that is sharp (contrasty) together with high resolution; the best of both worlds... he says; <em>"Nearly every lens is a compromise between those parameters." </em><br /> <em> </em>I understand a lens can be designed for absolute contrast and infinite resolution, but I wonder if it can be made, and if something close, at what price. This are consumer mass products.<em><br /></em><br /> I believe it could be similar to film resolution/sharpness developing. Higher resolution use to be achieved with the finest film grain, processed with solvent developers. Under the microscope it means an small dot with a large diffused disk. If we want sharpness, a non solvent developer is needed to keep the border of the silver particle as clean and "edgy" as possible... it looks sharper.</p>

<p>You have then to choose; or sharpness, or resolution, or a compromise between both, with more or less of them depending on what we`re looking for.<br /> I suspect is the same for lenses; cannot see what`s wrong with it.</p>

<p>Jose, I'm of the opinion that the entire concept of microcontrast is a bit of a fiction. Contrast is contrast; so where does it stop being "macro" contrast and start being microcontrast? There is no disjoint to be seen in any measured MTF lens curve to signify a changeover point.</p>

<p>I seem to remember the term microcontrast coming into being at about the same time as the word "acutance" and acutance enhancing developers came into fashion. In such a context it has some meaning, in that microcontrast often refers to the area of chemical edge enhancement provided by acutance developing techniques. The digital equivalent would be the blur radius of an unsharp masking filter. However, without any artificial image enhancement, either chemical or mathematical, the concept of microcontrast disappears like smoke.</p>

<p>And BTW, the "border of a silver particle" is <em>always </em>sharp. It's a metallic silver filament that's totally opaque and can't have any "graduated border". Whether developed in a solvent developer or in the grainiest developer imaginable (actually, that'd be Rodinal).</p>

<p>@Jose - Unfortunately, the analogy with film is flawed. Film, post processing algorithms, sensor hardware, etc. can increase the contrast (and microcontrast) essentially without limit, yielding threshold-like processing, exactly as you described. However, the contrast in the image formed by a lens can not exceed the contrast of the scene in front of it. The best that a perfect lens can do is make sure no light spills from one point onto adjacent areas/pixels. This is why the vertical axis of MTF curves only goes from zero to one. </p>

<p>This observation has profound consequences. For starters, it immediately negates your statements, (a) <em>"... while the contrasty lens will show just clear and dark borders, even "eating" subtle details (turning them to "clear" or "dark") ..."</em>, and (b) <em>"... I understand a lens can be designed for absolute contrast ..."</em>. </p>

<p>Lenses, even theoretically perfect lenses, just can't do this. The most contrasty lens in the world can only turn a 50% contrast change across an edge in the subject into a 50% contrast change in the image of that edge. Film / developer combinations, post-processing, hardware, and many other things could, in principle, even turn a faint 5% contrast edge into a 100% contrast edge. </p>

<p>HTH,</p>

<p>Tom M</p>

<p>Kirk Tuck is absolutely correct. Thank you. Start here:<br>

http://www.luminous-landscape.com/tutorials/understanding-series/understanding-mtf.shtml<br>

Well, this is interesting: http://www.luminous-landscape.com/tutorials/understanding-series/lens-contrast.shtml<br>

As is this: http://www.imx.nl/photo/optics/optics/page63.html<br>

This is pretty good: http://www.imx.nl/photo/optics/optics/page93.html<br>

http://imx.nl/photo/zeiss/page116/page116.html<br>

or this: https://sites.google.com/site/seevve/historical-perspective-on-minolta-lens-design-philosophy<br>

some good tidbits here: http://photo.imx.nl/optics/optics/page81.html<br>

http://www.dyxum.com/columns/articles/lenses/sal-85f14cz/carl-zeiss-sal-85f14cz_review.asp</p>

 

<p>Rodeo, as you point out, the concept of microcontrast (and how it's distinguished from "contrast") is more than a bit hazy ;-). If pressed to discuss the concept, I will usually opine that microcontrast deals with spatial frequencies on the right half of the usual MTF curves and "regular" contrast with spatial frequencies on the LHS of MTF curves, with no sharp border between them.</p>

<p>Tom M</p>

<p>I'm glad you joined the conversation, Kirk. We'll get to the bottom of this much more quickly this way.</p>

<p>While interesting, the links that you provided seem to be only very general, broad introductions to lens testing, MTFs, etc.. Such material constitutes "the basics" for anyone who has worked professionally in optics. Perhaps, somewhere in the many pages you cited, there are places where these documents speak specifically to the statement of yours that is under discussion, but I would prefer a much more focused (pun intended) set of citations. Also, to be honest, I got worried when I saw that 4 out of your 8 citations are to a blog with the word, "Tao" in the title. Perhaps the single author of all these articles is indeed a true authority on the subject, but I would much prefer to have seen you give very specific citations to referred articles in journals like JOSA, university level books on the subject, etc.</p>

<p>The specific statement under discussion is: <em>"You can design a lens for high resolution or high contrast but not necessarily both. Nearly every lens is a compromise between those parameters."</em></p>

<p>The easiest way to prove this statement is wrong is to take any decent quality real-world lens, call it "Design #1", and modify it in a thought-experiment in the following ways:</p>

<p>a) First, remove all the AR coatings, but change nothing else. The contrast will do down, but the spatial resolution will be unchanged. Consider this design #2.</p>

<p>b) Next, (mentally) put the coatings back on, but randomly tweak the spacings between the elements by small amounts, say, a bit more than the usual production variability. The low frequency part of the MTF curve (ie, the ordinary contrast of the lens) will be absolutely unchanged from the original design, but the modulation (ie, contrast) at higher spatial frequencies will drop like a rock. Put differently the resolution will stink but the contrast at the lowest spatial frequencies will be completely unchanged. Consider this version to be design #3.</p>

<p>Thus, in principle, I have shown how one can independently adjust contrast and resolution in any lens, contradicting your statement. </p>

<p>Most people don't like a physicist's traditional Gedanken experiment approach to disproving a statement by offering a single conceptual counter-example, but I can assure you that this approach has proved to work very well time and time again in much more complicated situations. ;-)</p>

<p>As I said in my earlier posts, my guess is that you simply were not being precise in your statement and so it confused people. Perhaps you would like to elaborate on exactly what you meant.</p>

<p>Cheers,</p>

<p>Tom M</p>

<p>PS - Unfortunately, I'll be in transit for the next several hours, so I won't be able to participate during that period, but I'll be sure to check back in ASAP.</p>

<p>Thanks for the feedback, everyone. Especially Kirk!</p>

<p>When I spoke of contrast in my original post, I wasn't speaking of microcontrast. I took two identical shots, no difference except for the lens, and when looking at the full frame downsized to fit on my monitor, the shot with the modern 60mm is <em>much</em> contrastier than the shot with the old 55mm. Could be internal flare raising the black level on the 55, maybe... I'd have to look much more carefully to see.</p>

<p>The frequent confluence of the terms sharpness, resolution, acutance, and contrast causes confusion. I accept that "microcontrast" is a tradeoff against resolution, and Ariel's example is a good illustration of this. But that's not what I was talking about. Is that what you were talking about, Kirk?</p>

<p>I'm still a little perplexed about why contrast (not sharpness/resolution/acutance/microcontrast) would be a bad thing in a lens. The lens isn't adding contrast that's not in the original scene -- it's not making the bright things brighter and the dark things darker. It's not like cranking up the contrast slider in post, which is trading off contrast in one part of the curve for another. Is it?</p>

<p>Contrast, according to both Leica and Zeiss, comes at a cost of resolution. Neither is necessarily bad and both are design choices. Contrast leads to better edge acutance but lower overall resolution. Until camera makers hit 16 or so megapixels the trade off could be made to increase contrast and still deliver adequate resolution for the combination of sensor and AA filter but when they started to go above that the reviewers noticed that their was no more "real" detail only interpolated detail as the sensor resolution increased which pointed to a fixed limit for resolution at a certain (popular) contrast level. Better glass formulations get you both but consumers always prefer snap over reality so the lenses that are designed for consumer cameras tend to get designed that way. <br>

As sensors increase in resolution they'll need lenses optimized for that parameter. The lenses WILL deliver a flatter image rendition. That's not a bad thing for us pros and hobbyists that are used to doing their own post production but we represent only a tiny subset of the overall market and the people who want to take images (JPEG) directly out of camera and submit to Walgreens, Costco and other Kiosk print suppliers or look at uncorrected images on computers will complain. Their technique (and I'm generalizing) is handheld and doesn't scrape the ceiling of resolution requirements as would your good technique. They'll insist on snappy glass. We want glass that's flatter but much more detailed because we can increase contrast and saturation in PS but we can't effectively go in the other direction.<br>

That's why it's an important topic. Good contrast is not a bad thing in a lens but it's the same as the paradigm of strength and flexibility in a swimmer. Too much muscle impedes fluid and articulated movement thru the water while not enough strength makes one too slow. There's always a balance. <br>

As you optimize a lens for resolution the optical choices made DO limit contrast and vice versa. In science there is really no free ride. So each maker strives to change the things that limit lower contrast while maintaining resolution. <br>

Think of it like this: an sRGB file has a limited gamut. A limited range of colors and contrasts. A pro RGB file has a much wider gamut BUT when you print the wider gamut on a machine printer you'll get a very, very flat print. The file has much more color and contrast information in it but the print can't utilize it. Lens makers look at the entire market for their glass and design for the sweet spot of the market. The majority. And the masses love snap, crackle pop. We might need specialty optics that go beyond that. One of my favorite lenses is the Zeiss 100mm Makro. It's very sharp and has very high resolution but it's not nearly as contrasty as the Tamron 90 micro, which I also own. But if you analyze a very high res ( Canon 1ds3 ) file from each lens you'll find more discernable detail in the less snappy Zeiss lens and more impression of sharpness in the Tamron lens because the contrast is optimized to create the highest impression of acutance.<br>

The Zeiss is not inferior but might require a nudge in post to look the same, but with more detailed resolution. That's my point. It's nice to be able to choose.<br>

It's well known amongst industry experts who've been around for a while that lenses are computed for different tasks. You might compromise contrast for flat field or flat field for center sharpness or geometric correction for lower vignetting. To believe that you can have it all is folly. You might be able to have it all but we mere mortals will not be able to afford that optic.<br>

And I'm not really addressing microcontrast though Erwin Puts covers that in some of the essays he wrote about lens design philosophy in 1999, 2001 and 2002.<br>

It's true that I'm not an optical expert but I was trained as an electrical engineer and do have a rudimentary grasp of optics and physics. And I've been away from it long enough to understand that it's all a compromise. Heat chases size chases bleed chases heat......You can never have it all. That's why, in electronics the pursuit is for the right compromise between size, power and speed.....<br>

That's all the nerd talk I have left for the day. I've just finished the final edits on my book about LED's for Amherst Media and I'm too exhausted to argue about stuff you could all quickly research on the web...It's not my opinion, it's published industrial science.<br>

As to the authority I cite, Erwin Puts has been a leading expert in photographic optics for years, has worked with Leica and just published a very technical 600+ page (densely packed) book about Leica and optics. Read it and we'll talk. His credentials are very good. But there are ample white papers from Zeiss and Leica. I am not paid to do other people's research. Nor is my blog audience composed of physicist who are deeply invested in technical arguments.</p>

<p>I have to agree with the contrast v. tonal range argument, at least from my own experiences. I have owned Zeiss 'C' and 'CF' lenses for my Hasselblad (the Cs are an older coating), and I find that even though both versions of a lens are theoretically the same, the C versions deliver a slightly longer tonal scale (particularly in the highlights), at the expense of less contrast and worse colour accuraccy. My C lenses also look sharper, but I suspect that this is only because the more contrasty coatings of the CF lenses may occasionally hide fine detail.</p>

<p>I'm also not ruling out the possibility that I had 'good copies' of the C lenses and 'bad copies' of the CFs either.</p>

<p>But as far as overall contrast vs. resolution in lens design, I can't say that I have any experience. I know that sharper lenses generally have more busy bokeh, but I think that the fact that lens design and coating design are both evolving means that you need much more knowledge than I have to talk about lens design only, without the coatings being a possible variable.</p>

There is a very good article about resolution and contrast in Leica Fotografie International 7/2011 written

by Michael J Hussman. It is a very technical article but worth the effort. In resume, good resolution is

the result of the correlation between lens and sensor, an harmonic balance between the resolution

capacities of this two elements will give you the best possible resolution and good contrast.

 

Also, I think that we need to consider the dynamic range that a sensor is capable to deliver. When we

use Nikon D-light function, we all experience a decrease in contrast but not a decrease in resolution.

 

My hiphotesis right now is that contrast is dependant on lens and sensor capabilities individualy, and

resolution, as the article points out, has to do more with the "symbiotic" relationship between lens and

sensor.

 

Everyone, escuse my english, I tried my best!

<p><em>"And I'm not really addressing microcontrast though Erwin Puts covers that in some of the essays he wrote about lens design philosophy in 1999, 2001 and 2002".</em><br /> Me too :) Erwin P. is a somewhat controversial kind of expert, but highly respectable in my opinion. I use to read everything coming from him.</p>

<p>It`s important to adress this concept, if not, we are talking about nothing. It is measurable and can be translated to "technical" charts (intensity and MTF). As everything in optics, it`s a complex topic, which I`m obviously not qualified to chat about.</p>

<p>(To R.J.: Well, I was trying to put a "film analogy" for easier understanding. I`m obviously refering to silver halide grains, not to the <em>silver molecule</em>... :) That`s the physical vs. chemical development. In the opposite to acutance developers, the dissolving action occurs in the border of the grain by <em>removing</em> them and making them softer looking. The non-solving action of e.g. the mentioned Rodinal make the grains to grow (actually, that silver filaments), and providing sharp looking edges).</p>

<p>KT: <em>"...I am not paid to do other people's research... </em></p>

<p>However, citations, to be useful, should be specific, and not require searching through tens of thousands of words to find something that may be relevant. For example, I would like to see one clear statement from an authoritative source (eg, journal article, university level text, etc.) that says what you claimed in your most recent post: <em>"Contrast, according to both Leica and Zeiss, comes at a cost of resolution.", </em> ie, which defines clearly and quantitatively the terms used. That shouldn't be too hard, should it?</p>

<p>KT: <em>"...Nor is my blog audience composed of physicist who are deeply invested in technical arguments...."</em></p>

<p>You got it wrong. Physicists are invested in getting to the truth, not arguments. This statement of yours says it all: you don't want a substantive discussion, just a voicing of opinions. You didn't respond to even one of the points I raised. The discussion is over as far as I am concerned.</p>

<p>Tom M</p>

 

<p>Not having read the other posts (sorry)...</p>

<p>Perhaps what the person meant is that engineers can add lens elements to a design to make more refined corrections that result in higher resolution, but by doing so, they've introduced additional reflecting surfaces, thereby lowering the contrast.</p>

<p>While this may be a tradeoff in theory, I don't think it's a tradeoff in practice. A poorly corrected lens will have low contrast BECAUSE it has low resolution -- because light is hitting the focal plane where it shouldn't. Also coatings are so good these days that lenses with numerous elements do not have contrast issues.</p>