Optics

Given all the griping about Popeye's eyesight on DPReview, I was quite positively surprised by it -- at least on the 1.6x field of view crop. I'll put some film through my EOS-650 one of these days, just to take a peek at the corners. Even wide-open, Popeye sees enough detail to produce very nice 11 x 14 inch prints: this is quite a good achievement. The image holds together respectably well at all apertures and focal lengths. You can detect all kinds of interesting gremlins if you magnify to 200% and really look for them, but they're kept pretty nicely in their cages: nothing in particular jumps out at you. Some softening towards the corners is discernible at all focal lengths less than two stops from wide-open, but less than I had feared. There's very little CA for this type of lens, and extremely little distortion.

Stopped-down to f/11, the lens is impeccable: the optical quality should satisfy even the most critical of photographers.

Sharpness

It's interesting how deceptive "feelings" can be. I had already mentally framed a paragraph explaining that clearly the 12-24 was designed for maximal sharpness at the expense of contrast, not like the good old days, and so on. Then I started examining two frames I had shot within minutes of each other of the same subject, one with Popeye, the other with "Canon's best prime," the 50/1.4 USM. Both lenses were stopped down to their respective sweet spots. I was rather surprised -- there was very little difference between them. At its best, the Sigma clearly packs a quite a wallop in the sharpness department!

At top, a 100% crop from the Popeye at 14 mm, f/16. At bottom, the 50/1.4 USM at f/5.6. If you look closely, you will see that the 50 is just a tiny bit sharper and just maybe a little bit contrastier, but so little it could actually be in the eye of the beholder.

Center sharpness is impeccable at all apertures and focal lengths: the lens easily outresolves the sensor. Sharpness fall-off towards the corners is discernible until approximately two stops down from wide-open, but IMO the fall-off is pretty negligible: less than I had expected for a lens this wide. The fall-off affects contrast more than resolution, and is at least partly caused by vignetting. It responds to post-processing very well: a simple curves or levels tweak with a radial mask will improve the corners a great deal, and a similarly masked application of unsharp mask will almost erase the effect of the contrast fall-off (see Appendix below). An exceptional 20 mm prime on full-frame would do somewhat better, but not by much.

Wide-open at 12 mm, center of frame.

Wide-open at 12 mm, corner.

Two stops down, the lens is very sharp. I'm going to absolutely love shooting off-the-tripod landscapes with it. :-)

Stopped-down to f/11 at 12 mm, center of frame.

Stopped-down to f/11 at 12 mm, corner.

Interestingly, the lens maintains its characteristics very constant through the focal length range: the sharpness fall-off is very similar at 12 and at 24 mm.

Wide-open at 24 mm, center of frame.

Wide-open at 24 mm, corner.

For the rest of the sharpness test shots, see this page.

Chromatic aberration

Very positive surprise here: for CA, the lens is just about as good as any zoom I've seen, and better than some primes. I did catch one pretty clear instance of transverse CA (shiny object on a dark background near the edge of the frame), but I suspect it would take an 11 x 14 print to show it. In my CA test shots of branches against the sky, I had to blow up the picture to 200% actual pixels to be able to easily see the CA. This problem is kept well under control.

CA test shot: wide-open, corner of frame, slightly misfocused. Not bad.

Worst case of transverse CA I've seen in my RL shots.

Halation

The mystery of the "soft corners" explained: the lens is prone to a certain amount of halation wide-open near the edges. If you get a dark, sharp edge near the corner of the frame, you will get a noticeable "light spillover" that will decrease contrast and give an appearance of softness. Most lenses do this wide-open, and this one isn't particularly bad (e.g. the 50/1.4 does this noticeably more; of course, this is long gone by the time it hits f/4.5, though). I wonder what causes this, though? I'd guess comatic aberration...

Vignetting

Senate Square with tractor. 12 mm f/4.5.

Vignetting is noticeable if you have large, evenly lit areas in wide-open shots. However, it's nowhere near objectionable enough to make you want to use radial ND filters (good thing too, since there's no filter threads; I've no idea if you could make a gel filter like this either). In a more "chaotic" picture, the vignetting is completely unnoticeable, as it's overwhelmed by the lightness differences in the scene.

Distortion

I'm quite impressed by this characteristic: there's barely any distortion to be seen. Yeah, you can see it if you go through the trouble of shooting a rectangular object exactly at right angles and then using a straight reference to measure it, but at least I didn't notice it without jumping through these kinds of hoops. Distortion is well below my "noticing threshold" at all focal lengths.

There's the barest hint of barrel distortion visible at 12 mm -- a very good job, considering the field of view. I suspect you wouldn't spot it even on a picture like this if you weren't specifically looking for it.

Flare

My first flare test shots on slide were rather promising: Popeye appeared much better than I had expected, given the symbolic hood and bulgy front element. With the sun in the frame, one noticeable (and hard to clone out) flare spot showed up, but contrast loss was negligible. The flare spot was situated right on top of the sun, meaning that it would not have had a major impact on the "crop factor" area. With the sun just out of the frame, there were no distracting flare spots, and contrast loss was still very reasonable.

A flare "torture test." If anything, this situation should produce flare spots. There's only one really nasty one that I can see, and that's right on top of the sun. (Wide-open at 12 mm on Provia 100F.)

A second frame shot seconds afterwards, with the camera moved so the sun is just outside the frame. No distracting flare effects that I can see.

However, some further experimentation revealed that as I had originally expected, the lens does struggle with flare in strongly back-lit conditions. However, it is somewhat unpredictable, possibly because of the geometry of the pop-eyed front element: on some frames, flare is kept very well under control, while on others, it goes pretty much all over the place.

"Meditating," Helsinki, 2004. A tough situation for any lens to handle: directly into the sun. Not entirely unexpectedly, Popeye struggles: there are pretty obvious flare spots and other artifacts, plus significant contrast loss, although the contrast loss is largely restricted to the area around the sun.

It's pretty clear that I'm really going to miss the marvelous flare characteristics of my other wide-angle, the Tokina AT-X Pro 17/3.5. I haven't managed to get into a situation with it yet that would cause flare bad enough to seriously degrade the image. Looks like I'll have to learn to be more careful again, if the Popeye does become my primary wide-angle lens, as it's set out to be.

Phoenician seawall in Batroun, Lebanon. I shot this one with the Tokina AT-X Pro 17/3.5. I've done nothing in particular to get rid of flare effects, because there's very little that needs to be done: if you stare hard enough, you'll find a few faint, largish flare spots, hardly any contrast loss. I will miss the Tokina.

Color

I compared Popeye's color rendition against the 50/1.4 USM -- I undestand the same lens Canon uses for color comparisons. It did extremely well. There's the barest hint of a warm cast; even side-by-side, it's barely above the threshold of detection. In other words, the lens is as close to neutral as any I've seen -- no "Sigma yellow" problem here.

Fishing through ice. Popeye above, Canon below, with identical post-processing. There is a color difference, but it's tiny.