Understanding lens resolution

<p>While working my way toward deciding on a 60D (love the flipout LCD) versus 5DMK2 as an upgrade to my 450D, I have been researching lens resolutions and getting a bit confused.<br />Making prints up to 20X30 is one objective. And I've already read every post of that 28 page thread comparing 7D to MK2 for large prints so no need to rehash that here.<br>

Thinking that print quality is highly dependent on the glass, I thought that if I decided on the 60D, I should at least pair it with 'L' lenses in the event I ever make the jump to FF. And if I don't, I still have Canon's presumably best lenses.<br>

Using photozone for test data, I checked out the 17-40 and found something I don't understand that I'm sure has to do with testing on FF versus cropped.<br>

The 17-40 resolves 2443 lines at 5.6 on a 50D. On the 5DMK2, it resolves 3422 lines.<br />Even my kit lens 18-55 resolves around 2400 lines at 5.6 as does the more expensive EF-S 15-85.<br>

Two questions: <br />1. Is it true that the 17-40 does not provide more resolution of sharpness than the EF-S lenses, including my kit?<br />2. Why are the resolution figures so different when tested on FF versus cropped?<br>

I'm hoping this thread might be invaluable information for lots of photogs besides me.<br>

Thanks, Bob</p>

 

<p>Print quality is highly dependent on many factors, the quality of the glass being only one of them (and certainly not the most important for <em>all</em> prints). </p>

<p>Forget resolution, get a camera and use it. Hand holding limits your resolution tremendously, contrast, exposure, subject matter, EV etc etc are all big factors. I don't understand your resolving lines figures either, lines per what? Inch, mm, sensor?</p>

<p>Detailed 20"x30" prints from a crop camera are, depending on your personal opinion of acceptable, unrealistic, unless you limit viewing distance. </p>

<p>Nose-to-glass on a 20x30 print is an unrealistic viewing distance. </p>

 

<p>True, as is expecting people to stay ten feet away.</p>

<p>What Rob and Scott say is absolutely to the point.</p>

<p>As for</p>

<blockquote>

<p>2. Why are the resolution figures so different when tested on FF versus cropped?</p>

</blockquote>

<p>the answer is that different parts of the total image cast by the lens are being measured. APS-C (it's its own thing, not a 'cropped' anything) images measure only the middle 22.3 x 14.9 mm of the image on Canon bodies, while the 35mm-sensor (apx. 24x36mm) covers more of the edges of the lens field.</p>

<p>The old idea was that the APS-C (aka DX) format used the better central portion of the field, the so-called "sweet spot"; but many tests of modern lenses suggest that this is not a universal principle.<br /> In any case, lenses are often not as good out to the edges as they are in more central portions of the image.</p>

<p>(A classic example of the differences can be found in the little EF 50mm f/1.8 lens, which is wonderful on a APS-C body, but more "how good for what it costs" on a 35mm-sensor.)</p>

<p>If you read the Photozone test you will see a little chart that shows the camera resolution limits - their 50D chart shows 2700 while the 5DII chart shows 3700 (I looked at the 85 F1.8 charts). This resolution difference is just full frame vs APS-C plus the 15 Mp vs 22 Mp sensor (the 7D / 60D will resolve more). What you may want to look at is the resolution shots on the digital picture as they tend to show you more what you can see. However, the choice of lens should not be confined to resolution. The factors to consider are edge vs center resolution, the difference in performance between a high contrast and a low contrast subject (good lenses perform well on both), colour and contrast, CA distortions and vignetting. You will also notice that stopped down the differences between a great lens and a reasonable lens (in resolution terms) start to disappear. You pay a lot for wide open edge performance - this is why some people shoot the Nikon 14-24 F2.8 on a Canon as it is a better lens than my 16-35 II. <br>

I have owned the 17-40 and it is a good lens but a bit soft at the edges. A quick look at the digital picture shows the 17-40 (at 17mm and F4) is sharper in the center but softer at the edges than the 17-85. Having owned the 17-40 I found the other factors (colour, contrast etc...) of this lens were very good and to me this "drawing" style is important. The 17-40 is also weatherproof and will work on full frame. The APS-C lens people rave about (I have not used it) is the 17-55 F2.8. As Scott says just take photos - over time you will decide what you like in a lens. Personally my favourite lenses are my Leica and Contax rangefinder lenses and my Fuji Medium Format lenses as they all have a "German style" which tends to be more balanced and less about high contrast center resolution. Bob Atkins has written a number of great articles on resolution - perhaps this may be of interest<br>

http://www.bobatkins.com/photography/digital/canon_eos_7D_review_4.html<br>

The Digital picture crops are at <br>

http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=251&Sample=0&FLI=0&API=0&LensComp=100&CameraComp=474&SampleComp=0&FLIComp=0&APIComp=0</p>

<p>Scott: Lines, from photozone: "line widths per picture height (LW/PH) which can be taken as a measure for sharpness"<br />I assumed photozone's testing criteria was something universally accepted but this might not be the case.<br>

JDM, I used the word 'cropped' only as it is used to denote APS-C sensor.<br>

I'm just trying to understand the technical side of how lenses are evaluated for sharpness. I know that absolute critical sharpness is often NOT the measure of a 'successful' photograph but I want at least to learn how to approach lens choices with more, not less knowledge.</p>

 

Rob,

From my 5D I have a photo of the Taj that can bear Nose to glass viewing on the 20x30 print on my wall.

I actually suggest firends to look that closely to see the beautiful detail on the Taj that they can see on the photo and then go back and enjoy the view as well.

20x30 to me is the ideal size for both details and far view at the same time. From a feet away it fills your field of vision transporting you into that location and at the same time the detail if there would allow you to see it well.

BTW for me anything that is around 12 inches is the nose to glass viewing distance as that is the distance at which our visual acuity is the most - human perceptual resolution is maximum.

<p>Bob - you need to understand resolution tests - they are performed across the lens using line pairs of varying contrast. In essence you have line pairs of different reflectance that will give a different contrast (e.g. a line of 75% reflectance then one of 25% reflectance is a 50% contrast). The test measures the ability of a lens to resolve at various contrast levels and various line spacings. Most tests now display an MTF charts - this plots contast on the vertical axis - 1 = 100% and (usually) distance across the bottom the numbers are distance in mm from the center of the frame - thus 20 is 20mm from the center of a 35mm frame. In general the plot shows a series of thick and thin lines thick is 10 line pairs per mm, thin is 30 lp/mm. The blue lines (in Canon's case) are at F8 and the black wide open. The solid and dashed lines are S and M (Sagital and Meridonial). From a layman's point of view you can regard these as evaluating the Bokeh of the lens - the closere the solid and dashed lines the more pleasing the Bokeh<br>

Here is a canon MTF chart<br>

http://usa.canon.com/cusa/consumer/products/cameras/ef_lens_lineup/ef_17_40mm_f_4l_usm</p>

<blockquote>

<p>I used the word 'cropped' only as it is used to denote APS-C sensor.</p>

</blockquote>

<p>Yes, I know. And I am saying what I did because this sort of usage too often leads to value judgments about "upgrading" to "full-frame". I am happy that <em>you</em> said "jump" not "upgrade."</p>

<p>APS-C and 35mm (Nikon does it better with "DX" and "FX") are just different formats.<br>

Although some medium-format fanboys will speak of "upgrading" to medium format, most of us recognize that this is not an "upgrade" but simply a <em>change</em>, like moving from medium format to the ultimate "upgrade" -- to wit, 8x10" view camera.</p>

My photography is about taking pictures of cameras and lenses. My best work is photos of MTF charts...

<p>Something to think about - and it is very important - when comparing cropped sensor and full frame formats, is the the lp/mm lens resolution values don't tell the whole story. If you put a lens with X lp/mm resolution on a cropped sensor camera and a lens with X lp/mm resolution on a full frame camera...</p>

<p>... the photograph made with the full frame camera has the potential to resolve more detail.</p>

<p>The key issue is what might be described as <em>line pairs per picture height</em> (or width) that result when the lens of X lp/mm is placed on a camera with, as I like to describe it, "more millimeters to hold line pairs." The short story is that the larger format always has the potential to resolve more detail. This was true with film and it continues to be true with digital, and the issue is most certainly not just how many photo sites are on the sensor.</p>

<p>If you are going to produce very high quality prints at the 20" x 30" size and you will work with detailed subjects in which high resolution is a concern and you will work with great care and from the tripod, then the FF format has some advantages for your photography that might make it a better choice.</p>

<p>Dan</p>

<p>Brad, how great for you. You never have to leave the house.</p>

 

 

<p><em>My photography is about taking pictures of cameras and lenses. My best work is photos of MTF charts...</em></p>

 

 

<p>What Brad said, ditto... Modern cameras and pro lenses are so good... yet, they are not sharp enough for your extremely hyper-critical work? Then go work for Canon or Nikkor and get that ''5 more lines per mm'' or whatever.</p>

<p>I sure don't get it. A 5D2 will take brilliant, sharp detailed 20x30 inch prints. A old 10D can do the same on smaller prints, say 19x13 inches. Sharp and detailed. Use good glass, good capture technique, and proper post processing to print techniques for the best output.</p>

<p>The old rule of thumb is that for close inspection you want 300 dots per inch. For 20x30, that comes to 54 megapixels. You need a very expensive medium format digital camera or you need to shoot 6x9 medium format film and scan it at 2400 dpi (which gets you close). I like to goof off with a crown graphic and 4x5 film, which is more than enough for a crisp 20x30 print that stands up to close inspection.</p>

<p>In reality, most DSLRs with a good lens and good technique can make a very good 20x30 print that will stand up to viewing at a few feet away. A t3i will be up to the task. A 7D will make getting the photo easier.</p>

<p>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.</p>

<p>I have found that a Tokina 11-16 is good enough for 20x30, as is the Tamron 17-50 and the Tokina 50-135. </p>

<p>Generally speaking, lenses made for APS-C sensors are sharper than lenses made for FF. So it is not surprising that 18-55 is sharper in the center than the 17-40 4L. But as G.Dan Mitchel explaned, the larger format of FF does not need as sharp lens to resolve a given total number of linepairs as the APS-C format.<br>

As pixel density increase, the difference in resolution between formats decrease but the larger format will almost always win.<br>

In the film days, it was a huge difference in resolution between FF (24x36 mm) and MF ( 55x68 mm) when using B&W Tri-X, but not so much when using the fine grain Kodak Tmax 100.</p>

 

<p>Let's get something straight here about my post and what I'm trying to get at:<br />I am not a pixel peeper or someone who is driven to care more about 'lines per inch' than in taking compelling photographic artwork.<br />But since I am looking into investing in a better sensor camera and better quality lenses, I want to understand what is the criteria on which their performance is based when a quality print is the ultimate goal. <br />I am the first to 'get' that resolution charts are not architecture, that megapixels are not portraits. <br />Ken, what you said, I have read before. That modern lenses are so good now that with proper capture, PP and printing, they are all pretty indistinguishable from one another and one is better off not obsessing over them and just go out and shoot.<br />And if that's 'one' of the truths learned from this post, that's all I'm looking for. And that same knowledge will be helpful to others potentially who happen upon this thread.<br />Thanks all.<br /><br /></p>

 

<p>Bob in the real world you tend to notice Bokeh, DOF, contrast, CA and distortion more than resolution in most shots with reasonable lenses. In big prints you will see softness, especially at the edge of the image if you have in focus objects. In general you can stop the lens down to solve the issue - indeed I try to never use my 35 F2 below F2.8 and my 50 F1.4 below F2 for this reason. Even when you do it is usually not fatal - my 16-35 F2.8 II gets pretty soft wide open and towards the 35mm end but people generally do not notice it. Without wishing to re-start a 7D vs 5DII debate (I own them both and they are fine cameras) you can tell the difference on a 20x30 print even at 100 ISO. That said you really have to look hard and unless you have the two side by side would never know. In the real world there is little to choose between them so long as they are at low ISO and correctly exposed - the 7D is a lot less tolerant of exposure errors than the 5DII (I almost always shoot RAW).<br>

If you plan to shoot at F5.6 then most lenses will work fine so long as the zoom range is not too long (long range zooms are very prone to distortion and CA - on a 35mm camera having a zoom that goes from wide angle to beyond about 70mm causes the lens to have to change from a retrofocus design to a telephoto design which lens designers tell me is where the problems start). The old rule of thumb (which I still go with) is that 3x zoom range is a good cut off.<br>

If you look at the photozone tests of the 17-40 and the 18-55 kit lens you will see the kit lens has slightly more distortion (3.2% wide and 0.23% long vs 2.47% and 0.08% for the 17-40) but in both cases the distortion at the wide end is noticible although it can be corrected in post processing.<br>

Vignetting is quite severe on the kit lens (1.4 stops wide open at 18mm) compared to 2/3 of a stop for the 17-40. Again this can be fixed in post processing. In terms of resolution there is little to choose between the two lenses. CA performance from the two lenses is very similar. I have not used the 18-55 Kit lens but I suspect it's Bokeh is not as good as the 17-40 (although the 17-40 is not that great).<br>

So what are you paying the extra for?<br>

Well partly the kit lens is much better value due to much higher volumes but also:<br>

The 17-40 has a much larger image circle so you can use it on full frame<br>

the 17-40 is much better built and it has internal focusing and will almost certainly be more durable<br>

the 17-40 has USM focusing so it will have faster AF<br>

the 17-40 is weather sealed<br>

the 17-40 is USM so you get manual focusing in AF mode<br>

the 17-40 comes with a lens hood and soft pouch.<br>

Is this worth the difference in price - up to you. Buying good glass is a law of diminishing returns although lens hold their value very well. For example my 1988 FD series 85 F1.2 is worth more than I paid new. Similarly my EF 70-200 F2.8 non IS can be sold on ebay for more than I paid for it. This is generally not true of inferior glass which loses value quickly.</p>

<p>Bob,</p>

<p>I understand your quest, but technique will always have a far greater impact on your imaging than almost any lens.</p>

<p>Testers are almost always inconsistent, those that do have clinical testing procedures are subject to sample variation and technique and tend to be lab-rats. One problem is what are you testing? As you examine testers methodology you start to realise they are almost all testing different things and that testing has gotten so far away from actual real life use as to make the testing almost irrelevant. For instance, one grain of sand on a lens mount can make a huge difference in measured resolution, but that simply will not show in prints or electronic image reproduction. Manual focus is critical in lens testing, but how often are you going to use MF? Some always will but very few. If you AF then all resolution figures are out the window.</p>

<p>Even the cheapest lens will, technically, out resolve, any sensor, look at the sensor resolution in P&S's, they are over 200 MP if you extrapolate to a FF sensor, but they are not L lenses. But it is how the lens, sensor, AA filter etc all interact.</p>

<p>To really get to grips with comparisons, you need to think system resolution, not isolate lens resolution.</p>

<p><em>"I want to understand what is the criteria on which their performance is based when a quality print is the ultimate goal. " </em>The print is the ultimate measure. The size and viewing distance, the printing technique, the post processing and the capture technique all play major roles.</p>

<p>Thanks for further explanations.<br />Because I started with a medium format, first C220 then Bronica, I am very comfortable and almost always use a tripod with my 450D. I usually shoot at f8, use base ISO's and the 2 second timer. <br />I think this thread has given me the insight I need and can focus on the basic lens selections I need to capture the images that move me.<br>

Thanks everybody, Bob</p>

<p><em> 1. Is it true that the 17-40 does not provide more resolution of sharpness than the EF-S lenses, including my kit?</em></p>

<p>The 17-40L is a good lens, but not necessarily better than the best EF-S and crop lenses. As an example, my Tokina 11-16 f/2.8 is sharper than my Canon 17-40L. This is even more true with the 17-40L on FF.</p>

<p><em>2. Why are the resolution figures so different when tested on FF versus cropped?</em></p>

<p>Because of the way Photozone presents the results. They explicitly state that you cannot compare results between different sensors, even sensors with the same MP rating. You can only compare lenses on the same identical sensor model. They typically provide a scale next to the results (good, very good, etc.) which lets you roughly compare lenses between sensors.</p>

<p>Too many people try to read sensor and format differences into their tests. If you want to compare sensors or formats, you need to use tests designed to do that. DPReview and Imaging Resource provide these test results.</p>

<p>For the very disciplined and good technique users, there are a couple of lenses that really stand out as killer sharp (if you can call that resolution, probably not, but that is one of the many problems with this kind of thing :-) ). The 24 TS-E and the 300 f2.8 IS are both lenses I have used that made me take a breath at how sharp they are. Unbelievably the 50mm f1.4 at f8 is a very, very sharp lens too.</p>

<p><em>Detailed 20"x30" prints from a crop camera are, depending on your personal opinion of acceptable, unrealistic, unless you limit viewing distance.</em></p>

<p>They are the same from a 60D as from a 5D mkII. If there are viewing distance limits for one, they will be the same for the other.</p>

<p>30" is the upper bound for a landscape from either body. Less demanding subject matter or viewing conditions will allow you to enlarge more. A portrait, for example, will scale much larger. At 24-30" the best crop and FF bodies produce very good prints, though there is still room for improvement from future, higher resolution sensors.</p>

Right now my Keuffel and Esser slide rule is in Solms getting a CLA. I'll have something important to say on this subject when I get it

back and can do a few calculations...