Canon EOS EF-S 17-85 IS Lens Distortion

[david_laidler]

I've been using a Canon EOS 20D together with an EF-S 17-85 IS lens

for a few months now and everything has worked well. However, when

taking some pictures of buildings last weekend I was very surprised

by the degree of barrel distortion at the 17mm end. To me it makes

the lens unuseable at wide angle for this purpose. I've attached a

picture that I took at the 17mm end of a brick wall (which is

straight) to try and show the distortion. I'd be interested in what

any one thinks about the degree of distortion? It seems a lot worse

than any other Canon lens I've owned.<div></div>

[bobatkins]

It's quite unsuitable for taking pictures of brick walls from closeup, but how does it look on "normal" images?

 

All lenses distort. Wideangle zooms distort more than most. You can correct the distortion in many image editors if it bothers you.

[tommyinca]

I used Picture Window Pro to measure the distortion. It is about 4.5%. The higher number is properly due to the picture is taken close to the wall.<div></div>

[steve_dunn2]

<p>That's pretty much the amount of distortion I'd expect at the wide end of a lens like this. The 24-85, 28-105, and 28-135 lenses are all in the same ballpark (see the discussion of distortion with those lenses a few days ago in this very forum).</p>

[jim_larson1]

Geez. . .I really get tired of everyone saying that every lens fault, from barrel distortion to chromatic abberation, is easily and quickly corrected with software.

 

While we are at it. . .I am also tired of people expecting perfect exposure when photographing a white wall in tungsten light, and people expecting perfect flash exposure when photographing MIRRORS.

 

But drifting back on topic. . . .this distortion seems typical of the $200-$400 prosumer zoom lens group, of which the 17-85/IS is a member.

[chiswick_john]

"I really get tired of everyone saying that every lens fault, from barrel distortion to chromatic abberation, is easily and quickly corrected with software." - why exactly is that a problem?

[bobatkins]

I would not be totally surprised to one day see DLSRs that automatically corrected for distortion in the camera. The distortion data could be programmed into the lens' ROM chip and the body could apply the correction when it did its internal RAW -> JPEG conversion. It would be Custom Function #87.

 

Chromatic aberration could be corrected similarly, with Custom Function #88.

 

While some people might see these as "cheating", others would welcome them as an alternative to paying 2x-4x the price for a lens that didn't need such corrections. While optical correction is better, digital correction is probably 95% as good (some pixels have to be interpolated).

[marknagel]

Ease up on the lens! How do you know its not the mason's fault? He might of had a bad day and laid the bricks poorly that day:)

 

m

[paul hart]

<<I've attached a picture that I took at the 17mm end of a brick wall>>

 

Now let me ask you a question. Did you take this photograph in a public place where you could be seen? Or did you do it furtively in private?

 

Now another question: what does the answer to this question tell you?

[jim_larson1]

Bob;

 

Not to start a fight. . . but the 580EX flash has trouble zooming the head properly with a 20mm lens attached. Can you imagine how it would attempt to correct for chromatic aberation?

[eyelight]

Perhaps someone would explain how any very wide lens, close and perpendicular to a flat subject, would not show the sides smaller than the center ( some distorion ) at least a bit? While it can be corrected, isn't it true that the edges are more distant than the center?

 

That said, I think you can be happy that "everything has worked out well". I have several Canon lenses, mostly L series too. With a "crop factor" camera such as yours and my 1DMKII, life is pretty good. Unfortunately, when I try those same lenses on a 1DSMK-II, the distortion is worse, the Chromatic Aberation looks worse, and frankly, the edge sharpness is really poor. Be happy that the center is pretty good, that the lens was inexpensive, and well, that you can tote that camera to some nice shots!

 

Architectural and "Cityscape" shooting, including pictures of brick walls, does form a sort of acid test for equipment. If you want to see real edge to edge sharpness, and very low distortion, you need to have lenses that cover the format with more room to spare. If you want to see a very sharp wide angle, 4x5 cameras have much less trouble.

 

I have the 17-24L and the 16-35L, and both exhibit considerable distortion on less demanding subjects. While you didn't mention it, both leave a lot to be desired in edge sharpness too. Frankly I wouldn't be surprised if your 17-85 is as good or better than my 16-35L, which is very soft unless stopped down to f11.

 

I think it is reasonable to accept some limits in image quality in exchange for the portability and ease of use the DSLR cameras offer. In other words, you can make yourself unhappy in so many ways or be happy if you like.

 

I'm interested to hear comments from others about whether a very wide lens should be *completely* distortionless in this situation though.

[bobatkins]

<em>but the 580EX flash has trouble zooming the head properly with a 20mm lens attached. Can you imagine how it would attempt to correct for chromatic aberation?</em>

<p>

I don't see the connection. Unless you're implying that Canon are so incompetant that nothing they do will ever work, which would clearly be a ludicrous suggestion.

[bobatkins]

Truely rectilinear wideangle lenses, by definition, show no barrel or pincushion distortion.

 

If you don't believe me, shoot the same brick wall using a camera with a pinhole lens. Pinhole lenses give zero distortion (i.e. are truely rectilinear). They may be slow and the results may not be very sharp, but the image won't show any barrel or pincushion distortion.

[david_laidler]

Thanks for all the replies. The reason I started photographing brick walls was because of the problems that I saw in my pictures and I wanted to test out the lens. I've attached an example to show you all. As well as the distortion, when taking pictures either horizontally or vertically I often see some tilt down towards the left, which I have never had a problem with when using my EOS film cameras. I saw another post on this issue, but everyone just seemed to think it was operator error and I can't see how this could be?

[david_laidler]

Another example image.

[chiswick_john]

"Frankly I wouldn't be surprised if your 17-85 is as good or better than my 16-35L, which is very soft unless stopped down to f11." - Ed you need to get your 16-35 serviced or adjusted - it beats all my primes at any aperture in this focal length on my 20D and a 1DS 2 that I have recently borrowed for evaluation - infact it gets softer at F11 and beyond due to diffraction.

[david_bedell]

"I often see some tilt down towards the left"

 

Are you subconciously looking at the distorted parts of straight lines near the edges of the frame when aligning your camera?

[jim_larson1]

I would like to ask a dumb question regarding software debarreling: What would happen if the photo was NOT taken perpendicular to the wall? Is the software capable of dealing with that problem?

 

I imagine the user is making these corrections by eye?

[steve_dunn2]

<p>I don't have any debarelling software so I haven't played with it and don't know what it can do. But I have Photoshop Elements 3, which has the ability to fix shots which have the "tall buildings falling backwards" look generated by a film plane which is not parallel to the subject. <a href="http://www.stevedunn.ca/photos/austria2004/r1f23-cathedral.jpg" target="_blank" title="Catholic cathedral in Schladming, Austria">This picture</a> had that look, and while it's not perfectly corrected, trust me when I say that this edited version looks much better than the original. (And yes, I know, this shot could do with some de-barreling, too.) Ditto for <a href="http://www.stevedunn.ca/photos/austria2004/r2f7-rathaus.jpg" target="_blank" title="Rathaus (town council hall) in Schladming, Austria">this picture</a>.</p>

[david_wright9]

I'm with David on this one: I've been taking photos of bookcases with my 17-85mm! These confirm the presence of severe distortion at both ends of the zoom range. The reason I took these pictures was the distortion is indeed bad enough to be visible in "normal" photos. I took some at a family occasion indoors using the wide angle to compensate for the lack of space. The result was pictures in which ceilings and door frames bent and swooped alarmingly.

 

I disagree with those who think the answer is post-processing with software. The problem is that these bend and stretch the whole image and cannot compensate for the relative position of objects in space. In my photos effect of straightening the lines in DxO or PT lens was to spread out the faces of people near the edge of the frames. The cure was worse than the disease.

 

The reason for trying the bookcases was that I had a chance to compare with the 17-40mm L to see if the distortion was an inevitable consequence of using such a wide angle. My conclusion was that distortion with the EF-S lens is strikingly worse. Enough to make oprdinary objects look strange to the non-technical eye.

 

I think this lens is something of an own-goal for Canon in the process of seeking to establish the EF-S standard.

[chiswick_john]

"was to spread out the faces of people near the edge of the frames" - this is a function of using a very wide lens - even a perfectly corrected wide lens will do this.

[david_laidler]

David, Thanks for the feedback. What is surprising to me is that the effect is clearly noticeable in "normal" photographs, as in the attached, where the Eifel Tower is clearly curved. I've never seen this kind of thing before in normal photographs. I guess I'm surprised that no one else really sees this as a problem. Or am I wrong?

[david_laidler]

One other point. Before I switched to the EOS 20D and 17-85 IS I had a Digital Rebel and the very cheap 18-55 kit lens. This camera and lens neither showed the distortion or tilt, when taking exactly the same kind of pictures. Unfortunately I sold them!

[dogbert]

I think I am with David too. My 28-135 shows a lot worse distortion than the 18-55 kit lens, Sigma 18-50 Ex DC, the Sigma 12-24 EX and even the 28-80 Mk III which is the crappiest lens I own. Maybe it is a function of the zoom range, which for Canon is fairly large.

[steve_dunn2]

<p>I think you're probably correct in pointing your finger at the zoom range, as the 28-135 has a significantly greater zoom range than any of the others in your list. Then again, clearly, zoom range is not the only thing determining distortion, else (going by photodo's measurements, as they're the only ones I have available) the 35-350 with its 10x range wouldn't have distortion numbers slightly better than the 5x 28-135. But look at the price difference :-)</p>