16-35 mm F4.0 IS first impressions - corner sharpness

[dcstep]

<p>Marcus, none of us are going to shoot film just because it's the only way to know the actual performance of our lenses. However, if it's possible, it'd be nice to see digital lenses tested on film bodies so that we could see what the native lens actually can do.</p>

<p>Take one Canon lens and three Canon bodies and take the same test image with each body. The three resulting images will not be the same when you examine the OOC RAWS in detail. Canon, DxO and, I assume Lightroom, take test images with lenses at every aperture and every focal length and every body that they support. The corrections a slightly different for every body. </p>

<p>This is not "gimmickry" but a necessary part of the digital capture process. You can pretend to test one part of a system by ignoring the two other parts, but then you're going to try to reach conclusions with incomplete data. </p>

[dcstep]

<p>Dan said:</p>

 

<blockquote>

<p>The point you are trying to make here is unclear to me.</p>

</blockquote>

<p>I was talking about the digital capture process introducing errors and the Raw conversion process removing errors. Unfortunately, Raw conversion done poorly also introduces errors. </p>

[Robin Smith]

<blockquote>

<p>This line of thinking assumes that the filtering and digital capture process doesn't do any damage</p>

</blockquote>

<p>No it doesn't, but one can minimize by not correcting for distortion or vignetting using software. Not sure what you mean by filtering. When one tests an optic it makes no sense to introduce extra variables. You are not going to test a lens with a circular polarizer on it and then claim it shows intense blues as if that is a feature of the lens. It's the same deal. It would be one thing if you had no alternative i.e on a fixed lens camera like a Sony RX1 or similar where the camera will always correct the RAW in camera and there's nothing you can do about it, but the whole point of testing a lens like this is to test it in isolation. G Dan's point and yours is that in practice using these corrections makes the raw optical performance much less important. That is a different point, it doesn't mean the lens has actually got any better optically. There's nothing to argue about really since there is no doubt that correcting for distortion will affect pixel level detail. Whether routine software correction is a good idea will depend on how bad the initial distortion was that needs to be corrected. The 50/1.4 has very little distortion in the scheme of things. As for perspective distortion, it will all depend, obviously, on the amount required to correct it: the 35/2 example is very mild so I'm not surprised you won't see much difference, but even then G Dan admits there is sharpness reduction. Personally, I use software correction when I need to, but I do not think that is the way to test a lens to show its optical prowess.</p>

[dcstep]

<p>So Robin, you seem to say that the in-camera Raw file is accurate representation of what the lens saw? If so, we disagree.</p>

[g dan mitchell]

The raw file supposedly contains the original luminosity data from the individual photo sites. The truth is slightly more

complicated.

 

Lenses don't "see" anything. Sensors sort of do "see" something transmitted to them through lenses, but rather than

"see," it is more accurate to say that they register or record luminosity values.

[Wouter Willemse]

<blockquote>

<p>This is not "gimmickry" but a necessary part of the digital capture process</p>

</blockquote>

<p>Here we disagree; it is not gimmickry, but it most certainly is not necessary. I shoot both digital and film, with the same lenses - how come on digital distortion correction would be necessary, while it would be perfectly fine to leave it as-is on film? Lens corrections are nice-to-have, but not the end-all-and-be-all of imaging. Let's not act like it makes the critical difference.</p>

<blockquote>

<p>However, if it's possible, it'd be nice to see digital lenses tested on film bodies so that we could see what the native lens actually can do.</p>

</blockquote>

<p>There is no such thing as a <em>digital lens</em>. Using it on film doesn't make it more native than it is on digital. Lenses are lenses, optical devices; they transmit light, and that's it. Some lenses have coatings optimised for the reflection of a sensor (instead of film), but that's as digital as lenses get. A sensor can render some optical defects more noticeable than film does, but the lens still does the very same thing - it is unaware of the recording medium that sits behind it (and yes, I use the same lenses on my digital bodies as I use on my film bodies).<br>

It is nonsens to try to define what is the most accurate representation of a scene. Anything in camera is just a bunch of data that needs interpretation (which raw processor is the accurate, which lens profile is most correct etc. etc.?). There are various ways, which all have their validity (with different advantages and disadvantages). Acting like there is one single correct way is underestimating seriously the complexity of the situation.</p>

[dcstep]

<p>Wouter said:</p>

 

<blockquote>

<p>Here we disagree; it is not gimmickry, but it most certainly is not necessary. I shoot both digital and film, with the same lenses - how come on digital distortion correction would be necessary, while it would be perfectly fine to leave it as-is on film? Lens corrections are nice-to-have, but not the end-all-and-be-all of imaging. Let's not act like it makes the critical difference.</p>

</blockquote>

<p>Exactly my point, there's no digital distortion on film, but there most likely is with digital capture. The degree of digital distortion may not be enough to worry with, but that's dependent on the body/lens combination. Zoom, lenses, particularly those including wide-angle, will have variable chromatic aberration, vignetting, geometric distortion, etc. which is more likely to be course enough at certain aperture/focal length combinations to warrant correction. Some digital bodies will respond to lens imperfections by adding purple fringing, moire, etc. that further aggravates the lens' imperfections.</p>

 

<blockquote>

<p>There is no such thing as a <em>digital lens</em>. Using it on film doesn't make it more native than it is on digital. Lenses are lenses, optical devices; they transmit light, and that's it. Some lenses have coatings optimised for the reflection of a sensor (instead of film), but that's as digital as lenses get. A sensor can render some optical defects more noticeable than film does, but the lens still does the very same thing - it is unaware of the recording medium that sits behind it (and yes, I use the same lenses on my digital bodies as I use on my film bodies).</p>

</blockquote>

<p>You said it yourself, "Some lenses have coatings optimized for the reflection of a sensor." Until digital capture came along, we didn't see these lenses. I believe that they're rightfully called "digital lenses", because they're optimized for digital capture. Still, with the right mount, you can put them on a film camera.</p>

 

<blockquote>

<p>It is nonsense to try to define what is the most accurate representation of a scene. Anything in camera is just a bunch of data that needs interpretation (which raw processor is the accurate, which lens profile is most correct etc. etc.?). There are various ways, which all have their validity (with different advantages and disadvantages). Acting like there is one single correct way is underestimating seriously the complexity of the situation.</p>

</blockquote>

<p>I believe that test patterns, models and certain static displays do allow a certain level of repeatable performance where the captured image can indeed be compared to reality and judged for accuracy. Despite their limitation, I find the test images at http://www.the-digital-picture.com/Reviews/Comparison-Tools.aspx useful. I also like the DPReview sample images using their little mock scenarios. Still, particularly with zooms, I think it'd be very useful to see those same images after applying the manufacturer's DLO correction in Raw conversion.</p>

[Marcus Ian]

<blockquote>

<p>Some lenses have coatings optimised for the reflection of a sensor (instead of film), but that's as digital as lenses get.</p>

</blockquote>

<p>Well, aside from the lenses that only mount to digital cameras (such as the EF-S lenses) ;) .</p>

<p>However, it is not a relevant distinction, because this lens is an EF (L) lens, and can therefore be mounted on any EOS camera ever made... awhile back, those included film cameras... Thank you for illustrating my point!</p>

<blockquote>

<p>The three resulting images will not be the same when you examine the OOC RAWS in detail.</p>

</blockquote>

<p>While I agree, it is not mere 'gimmickry', it<em> is</em> a software modification to an image file (aka image editing). It is the same as NR in that respect, and yes, while each sensor will have it's own 'opinion' about what it 'sees', the broad characteristics of the image will be exactly the same. ie. I can mount one to my 5D2s, then my old 5D, then a 5D3 or a 1Ds, and <em>all the images will share broadly similar distortions in the OOC Raws (or JPGs of course!)</em>. This is why we test lenses... to see how bad or good they are. -- In fact, if I shoot with it on my Elan, I would bet a fair amount of money that you would see very nearly exactly the <em>same distortion apparent in the image</em>.</p>

<p>I don't understand why some would argue that a lens's characteristics be defined by the results <em>after</em> image editing (in this case distortion correction). We don't typically define a sensor's noise characteristics after we've run NR on the image, specifically because every piece of software does it a bit differently. Same goes for this... and don't forget, while those who would use this lens on film are indeed few and far between, those who shoot JPGs are NOT.. Those who shoot JPGs may or may not ever use an image editor capable of distortion correction (or may not bother). Just because some do, doesn't mean ALL do. -- Which is exactly why understanding the UN-edited optical characteristics of a lens are so important.</p>

[Wouter Willemse]

<p>David, what you call "digital distortions" are optical errors that for a large part are not due to digital at all, and occur similarly on film. Chromatic aberrations, distortion, vignetting - it all happens on film as well. It really has got zero to do with the recording medium in the camera. So you're whole point on how we should treat this differently on digital to level the playing field is, in my view, based on a complete misunderstanding. You make a distinction where there is none.</p>

<p>Marcus, I think we agree - though for me, in the end, images talk, so lens tests as such have a limited use (especially since they tend to compare two-dimensional things). </p>

[dcstep]

<p>Marcus said:</p>

 

<blockquote>

<p>However, it is not a relevant distinction, because this lens is an EF (L) lens, and can therefore be mounted on any EOS camera ever made... awhile back, those included film cameras... </p>

</blockquote>

<p>Great, now show me the test results on film. Thank you for illustrating my point!</p>

 

<blockquote>

<p>While I agree, it is not mere 'gimmickry', it<em> is</em> a software modification to an image file (aka image editing). It is the same as NR in that respect, and yes, while each sensor will have it's own 'opinion' about what it 'sees', the broad characteristics of the image will be exactly the same. ie. I can mount one to my 5D2s, then my old 5D, then a 5D3 or a 1Ds, and <em>all the images will share broadly similar distortions in the OOC Raws (or JPGs of course!)</em>. This is why we test lenses... to see how bad or good they are. -- In fact, if I shoot with it on my Elan, I would bet a fair amount of money that you would see very nearly exactly the <em>same distortion apparent in the image</em>.</p>

</blockquote>

<p>When testing, are we only looking at "broad characteristics" or are we interested in detail differences? If you're only interested in "broad characteristics" then you will not care about optimized performance, but some of us want to see optimized performance.</p>

 

<blockquote>

<p>I don't understand why some would argue that a lens's characteristics be defined by the results <em>after</em> image editing (in this case distortion correction). We don't typically define a sensor's noise characteristics after we've run NR on the image, specifically because every piece of software does it a bit differently. Same goes for this... and don't forget, while those who would use this lens on film are indeed few and far between, those who shoot JPGs are NOT.. Those who shoot JPGs may or may not ever use an image editor capable of distortion correction (or may not bother). Just because some do, doesn't mean ALL do. -- Which is exactly why understanding the UN-edited optical characteristics of a lens are so important.</p>

</blockquote>

<p>We're talking about lens testing, not body testing. Noise and in-camera JPGs are a representation of the body's ability to interpret the image, filtered and then captured on the sensor. When testing a body, we'll want to see in-camera JPGs to get an idea of color interpretation, avoidance of purple fringing, moire handling, dynamic range, noise production, etc., but that's only there for those that don't use external Raw conversion software. DPReview, for instance, will show Raw conversion with the manufacturer's external software and, sometimes, one or two other Raw conversion programs. I ignore the JPGs samples, but really get into the Raw conversion sample images. Many will only look at the JPG samples.</p>

<p>There's really no such thing as an un-edited digital image. Even the in-camera JPGs is based on the tastes of a Japanese committee. The in-camera processor applies pre-determined RGB, Contrast and, in some cases, Digital Lens Optimization, among other things. There is no "Raw image", it's a Raw data file.</p>

[dcstep]

<p>Wouter said:</p>

 

<blockquote>

<p>David, what you call "digital distortions" are optical errors that for a large part are not due to digital at all, and occur similarly on film. Chromatic aberrations, distortion, vignetting - it all happens on film as well. It really has got zero to do with the recording medium in the camera. So you're whole point on how we should treat this differently on digital to level the playing field is, in my view, based on a complete misunderstanding. You make a distinction where there is none.</p>

</blockquote>

<p>Wouter, sorry if I was unclear. I don't think I said that CA and vignetting were digital distortions. Those are optical distortions. Purple fringing and moire are a couple of digital distortions that I mentioned that can be excited by optical distortions. I meant to distinguish one from the other. Of course, DLO corrects for both optical and digital distortions, some more successfully than others.</p>

[Wouter Willemse]

<p>David, read your own 10:47 am post: yes, you're being massively unclear. You do not distinguish at all between "digital distortions" and optical distortions (which you do not even mention), and you seem also to imply that only zooms have these optical issues show up in variable amounts depending on settings (while primes suffer the exact same). And your last post, probably you are unclear again, but purple fringing is an optical flaw and also happens on film....</p>

<p>Given that you do seem to be very convinced of your case for the greatness of "DLO", let's leave it at agreeing to disagree on its necessity for digital images.</p>

[dcstep]

<p>Yes Wouter, yes, I agree that we should agree to disagree.</p>

<p>For those interested in learning more about DLO, try it for yourselves, particularly with your zoom lenses, like the OP in this thread. Digital Photo Professional includes DLO modules for most EF and EF-L lens/body combinations, so it'll cost you only a little time to see if you want to include it in your workflow.</p>

[g dan mitchell]

<blockquote>

<p>Here we disagree; it is not gimmickry, but it most certainly is not necessary. I shoot both digital and film, with the same lenses - how come on digital distortion correction would be necessary, while it would be perfectly fine to leave it as-is on film? Lens corrections are nice-to-have, but not the end-all-and-be-all of imaging. Let's not act like it makes the critical difference.</p>

</blockquote>

<p>The "not necessary" argument is a slippery slope. You could argue that roll film is "not necessary," but rather is just "nice to have." You could say the same about reflex camera designs, filters, tilts and shifts, color filtering in film printing, use of papers with different contrast grades and surfaces, the quality of lighting used when prints are displayed, and just about anything else.</p>

<p>I would argue that if your photography benefits from the ability to correct or otherwise alter the captured image in post — as it traditionally and still almost always does — then it is "necessary" for your photography. The fact that the world will continue to rotate on its axis whether or not we can correct CA in post doesn't render the ability to correct CA unimportant.</p>

<p>I don't understand how this is such a big deal. If being able to minimize or eliminate barrel/pincushion distortion, fix CA, improve image sharpness, etc. are possible and can improve some photographs, why wouldn't we want these capabilities?</p>

<blockquote>

<p>"...but even then G Dan admits there is sharpness reduction"</p>

</blockquote>

<p>I think you may have accidentally left our adjectives such as insignificant or invisible in your paraphrase. My point was not that "there is a sharpness reduction," but that these corrections can be applied in the majority of cases with <strong>no visible deterioration</strong> of the image, even as a rather large print.</p>

<p>On a different topic and not directly related to the above quote: Anyone making a big deal out of the fact that digital capture "distorts" the image needs to slow down and think this through a lot more carefully. Film capture also "distorts" the original image in a number of ways, and the very act of making a photograph is far, far from what some imagine to be a "capture" of the objective reality of the subject of the photograph.</p>

<p>My motto is that "photographs always lie." And that is a Good Thing. And I'm not joking.</p>

[Robin Smith]

<blockquote>

<p>inspect this 100% crop <em>very</em> closely you <em>can</em> detect a small difference in the “sharpness” of the two photographs – but it is quite subtle even when viewed at 100%.</p>

</blockquote>

<p>That is just for a small perspective correction. You are arguing it doesn't matter, I tend to agree, but it does not mean that the performance of the tilted 35/2 with software correction is the same as say a tilt-shift lens equivalently shifted.</p>

[g dan mitchell]

<blockquote>

<p>That is just for a small perspective correction. You are arguing it doesn't matter, I tend to agree, but it does not mean that the performance of the tilted 35/2 with software correction is the same as say a tilt-shift lens equivalently shifted.</p>

</blockquote>

<p>Hmmm... I did not say it is "the same." What I did say (and provide some evidence for, I think) is that making such corrections in post does not degrade the image in a visible way, and that this approach is an entirely viable one, even for rather large prints.</p>

<p>A lot of forumtography writing is more about theory than reality. I've found that it is a good idea —and with digital it is often pretty easy, too — to actually check these things. I recall first "getting this" after I had spent a period of time avoiding apertures smaller than f/8 because "diffraction blur will soften you photographs." Then I made some photographs at various apertures and inspected the actual results and found out that the whole thing was overblown.</p>

<p>There are some things that TS lenses can do that you can't really do in post, but it is perfectly fine to make typical small corrections to barrel/pincushion distortion and some perspective distortion in software and the results are excellent.</p>

<p>Dan</p>