Full frame vs APS-C: distortion & bad lens copies

-- "Reasons being severe distortion"

Can you supply a sample image showing your problem.

The point is, there are several different kinds of distortion, and not for all the lens is to blame.

- barrel distortion. This is caused by the lens.

- perspective distortion. This is caused by the user.

- Border distortion of rectilinear corrected lenses (usually only seen with UWA lenses). This kind of distortion is necessary to keep straight lines straight. To reach this goal, areas have to be streched the more you get to the corner. This is caused by design.

The latter two will not go away if you go FF.

For the barrel distortion ... photozone reports -1.25% for the 10-22 at 10mm and -2.47% for the 17-40 at 17mm. Both are measured on APS-C ... but since barrel distortion will not get better when using FF, you can expect at least the -2.47% on FF also. So the 10-22 on APS-C would be better here.

If distortion (barrel, moustache or pincushion) is a concern then you should process your images with PTLens software that will deal with it (and also correct for CA and vignetting). That might leave you a lot happier with your lenses, and be cheaper than trying to pursue an impossible Holy Grail.

http://epaperpress.com/ptlens/

Wide abd ultra-wide angle lenses produce a characteristic FOV which, requires a lot of care when framing. Tilting the camera and/or using off-center subjects, etc... can cause unwanted results. Some lenses have better performance than others with regards to correcting certain issues inherent to those lens designs.

If you want the best ultra wide primes and/or zooms, you have to spend like there is no tomorrow so, can't have it both ways. IF you want to save you have to be wiling to accept a level of compromise in performance.

But, unless you want to use an anamorphic system (!), that's the way ultra-wides work.

PT Lens works as do other techniques but, remember...you are stretching pixels so, even then there is a limit to what can be done without unwanted results.

I know this is a stupid question, and I will probably kick myself for not figuring out the acronym, but what is "CA"?

CA is Chromatic Aberration:

http://www.vanwalree.com/optics/chromatic.html

The 17-40 will show distortion on a 5D at the 17mm focal length, which means that you may not get completely straight horizons in a landscape image. DPP will fix that but you will loose a small amount of the image in the process. Distortion at 24mm is far less. (I know the difference between real distortion a prospective distortion)

While there are certainly bad equipment out there and I have found lenses that I did not like as much as others, but if you look hard enough for defects, you will likely find them in practically any piece of equipment.

Distortion of any type (barreling/pincushioning or perspective) is easy to correct with just about any photoediting package (e.g. Photoshop). However, if distortion is your primary concern, you might consider the Sigma 12-24 a an ultrawide. It won't give you the sharpness and contrast of, say, a 17-40L, nor will it control CA as well. There are also quality control issues, so you might not get a great copy. However, the lens does a phenomenal job of staying flat and true, with nearly zero barreling or pincushioning. It's also the widest angle rectilinear zoom ever engineered (wider than the 10-22 zooms because it can be used on a full frame). It's tied with the Voigtlander 12mm prime, which of course is only suited for a rangefinder. I rather like mine. It's a little engineering miracle, and you can't beat it for interior architectural work.

When people write about "distortion" on wide angle lenses they may be referring to a number of different issues, and the answer to your questions will be quite different depending upon what _you_ mean by this.

By their nature wide angle lenses create perspective distortion - any wide angle will do this, from the cheapest to the most expensive. This is the quality responsible for things like buildings tilting inwards, large noses on people you photograph close up, and so forth.

All lenses exhibit some barrel (outward bulging) and/or pincushion (inward bending) distortion and wide angle lenses can be especially prone to this - and it can be confused with the normal perspective distortion.

Wides tend to be perhaps more prone to "chromatic aberration" - color separation around high contrast areas most often in the corners of the frame. For example, the edge of a building or branches might have red/green lines around them.

Corner light fall-off (or "vignetting") is also found on virtually all lenses at their largest apertures, though it generally diminishes to insignificance as you stop down.

Corner softness - noticeable blurring in the far corners of the frame - afflicts many wide angle lenses, especially when shot at focal length extremes and wide open. This also diminishes as you stop down the aperture.

It is very important to understand what you mean by "distortion" for several reasons. First, it is quite possible that what you are seeing is simply the normal behavior of all lenses - e.g. corner fall-off and/or softness at large apertures. Trying to "fix" this by buying more lenses could be fruitless. Second, It could be a "problem" that simply requires a change in shooting technique rather than new gear. Third, some "distortion" can easily be fixed in post-processing, chromatic aberration for example. Fourth, if it is a "hardware problem" with the particular lenses you have used, it is critical to know what the issue is so that you can decide what lenses might improve this.

Dan

I'm refering to the "stretched out" look at the edges edges, the type that renders a normal car into a limo and makes cartoon heads of portraits..

>>I'm refering to the "stretched out" look at the edges edges, the type that renders a normal car into a limo and makes cartoon heads of portraits<<

Hence the "frame carefully" suggestion. ALL ultra-wide (and wide) angle lenses stretch image at the edges to some degree or another. Even PT Lens, the program suggest above. works by stretching the image (Pixels), except in the inverse. It's a matter of physics. There are limitation to 'capturing' a FOV of 110 degrees for example, and then, have it appear in our print the same as our brain perceives it when we observe the subject live.

Good lenses are corrected but, only to the extent that is possible, given their price range and focal length.

If you place a subject at the edges of any wide angle lens that subject will be distorted,

Oh, that's perspective distortion and is user controlled via technique. Sounds like you might enjoy a T/S lens.

"I'm refering to the "stretched out" look at the edges edges, the type that renders a normal car into a limo and makes cartoon heads of portraits.."

Exactly. That's why I asked my series of questions.

No different lens with the same focal length will change this in any substantial way. It is a normal characteristic of shooting ultra wide angle lenses.

You can minimize this by placing the horizon in the middle of the frame and being very careful to keep the camera level, however you cannot entirely eliminate such "distortions."

Dan

try demo at www.dxo.com - will fix the issues you have.

http://www.vanwalree.com/optics/distortion.html

It sounds like you're referring to the expansion of the image towards the margins by a rectilinear lens. That's a normal consequence of the design of the lens, with the objective being to keep straight lines straight and flat surfaces flat. However, a round object will appear oval in the margins, and a short hallway will look incredibly looooong. A fisheye lens doesn't generate this same type of distortion, but it also doesn't keep flat surfaces flat. That's why the image looks hemispherical. In short, you can't have it every which way. A camera will never see the world the way that we do, because our brains actively process what we see and piece it together actively into a coherent, three-dimensional whole. Oddly, there is not even a way to represent the whole of what we see on a flat piece of paper. (Trust me. I am/was a sensory physiologist. I and my colleagues have stuck many an electrode into the visual systems of many different animals to learn this.) The more the camera departs from what the eyeball would "see," the more alien the image becomes to us.

My most considered advice is to stopy buying wide angle lenses. Accept the phenomenon for what it is. Either embrace it, or abandon it.

Besides that, you mention the word "portraits." An ultrawide angle lens will never take a non-cartoonish portrait. You should be using something around 100mm on a full frame for head/shoulder shots and perhaps a 50 or more for full body shots, depending on how much room you have to move about. You will never be satisfied with a wide angle lens for portraiture, unless you're intentionally trying to make the subject look weird, distorted, or cartoonish.

One further note: If you have a lens that you feel looks soft or that exhibits more CA than you might expect, that could indeed be a lens defect afoot (but might just be the nature of the design). If you are concerned with the pattern of projection of the lens -- the distortion properties, if you will -- then that's the design of the lens and nothing else. It's not a matter of bad copy vs. good copy. It's also *approximately* what you're going to see on other lenses of similar design, give or take.

The better, and even some inexpensive wide angle lenses designed for 35mm film use will often test having lower distortion than their aps size DSLR counterparts, which are of much shorter FL. Also, FF DSLR sensors, even those in very expensive models, tend to have some light falloff problems, which exacerbates light falloff in even a good lens. Also, there tends to be more sensor reflective problems for edge performance at wide angle, according to reports I have read. It is therefore much more critical to choose the best possible quality in a wide angle for a FF DSLR than for a quality 35mm film SLR.

Wide angle continues to be a weaker area for digital, and my take is that film continues to be tops in that area.

Also, FF DSLR sensors, even those in very expensive models, tend to have some light falloff problems, which exacerbates light falloff in even a good lens.
Huh..? Could you support that statement with evidence, because I see much less light fall off on a 1Ds3 than on a 5D using the same lens (say, one with a known light fall-off issue wide open, such as EF 24/1.4L.) Both cameras sport FF sensors but there is a lot of magic going on on the 1Ds3...

I saw a writeup a couple of years back on this matter in POP Photo, and some experienced in both saying something like "ya'd better have the best lens possible for avoiding those problems with a FF DSLR, because any issue from the lens is compounded". But as you say, advancements may indeed have been made in the 1Ds3. Try going real wide, like 20mm or less, and compare edge results with using film.

I think Sarah is telling it like it is. I have stopped using any lens wider than 20mm (on full frame or 35mm film) because it's so difficult to frame images so as to control distortion. I don't have a ton of money to spend on bodies and lenses anyway, but I know you can't buy your way out of this situation. If you must have ultra-wide images, learn to live with distortion.

(By the way, fish-eye lenses are probably the least distorting ultra-wides, but people make the mistake of viewing the images from much too far away!)

But what Anesh described is not distortion.

I would ordinarily not use a lens wider than 28mm or so (on full frame) with people in the frame. For architecture ultrawides are fine as long as you align them just right (on a tripod, usually).

Sarah Fox, your explanation of Aneshﾴs "distortion" was excellent. I think the logical solution to the problem, which I consider real, is to carry a fisheye lens in addition to the rectilinear wideangle. My Sigma 10-20 is OK, btw.

You may find that no super wideangle lens creates the effect you want; whether on a FF or cropped sensor or; view camera etc. There is a reason animals have round eyeballs. You are mapping a 3D image to a flat plane; ie film or a sensor. Most prints are flat, most screens are flat. The distortion here is the thought that a mythical lens will please you.