Taking a different tack on corner smearing
Kamera Kat, my faithful wooden companion in my exploration of the corner-smearing conundrum. The Kat is almost exactly as high as my own face, from top of head to chin, so provides me with an infinitely patient headshot proxy.
I've been following this discussion with some trepidation. After
my initial experience with the G1, I actually had made up my mind to return it.
But then I started second-guessing myself: my test conditions were unusually awful even by my standards, and a surprising number of "civilians" who looked at the pictures liked them. I had the G1 all boxed up and ready to go back to the store, but kept putting it off, struggling to make a decision about it... and eventually the time limit for returning it had passed, and I still had it, and I had made up my mind to be pleased about that.
Then this corner-smearing thing came up, courtesy of Sean Reid and his pay-to-play website. This was a big deal to me, since my main justification for the G1 was as an alternative platform to my Epson R-D 1 for my odd collection of M-mount lenses.
I dropped my Reid subscription after the first year, but since he seems to have a more-than-just-good-friends relationship with Leica, I had to assume he has more experience with Leica optics than I do (having sold my last Leica lens about 20 years ago.) And some of the photos in this thread seemed to show issues very clearly.
So finally this evening I decided to do my own evaluation, but with my own methodology -- one a bit different from what Reid seems to have used and what others have used here. It's going to sound odd and unscientific to some of you, but it's based closely on how I, personally, take pictures and evaluate them. Here's what I did:
1) I almost never take pictures of brick walls or bookcases, but I do want the freedom to position my subject where I want it in the frame. Whether I decide to put the subject dead-center or off in some corner, I want to have it look reasonably detailed, not smeary. So instead of shooting single pictures and cropping out centers and corners, I shot pairs of pictures of the same subject: once in the center of the frame, and again recomposed so it's off in one corner.
2) I'm not interested in evaluating absolute image quality. What I want to find out is whether pictures taken with the G1 are going to be significantly worse than ones taken with my Epson R-D 1. The R-D 1 may not be the ultimate in image quality, but I know exactly what to expect from it and always use it with confidence -- so, I also shot pairs of pictures with the R-D 1 so I could use it as a baseline for comparison.
3) I shot all my test pictures from the same distance, because that's how I generally work: I'm in a fixed position, and choose different lenses to get different angles of view. The lenses I used were a 21mm f/2.8 Avenon and 28/1.9, 35/1.2, and 50/1.5 Cosina-Voigtlanders.
4) Most of my shooting is divided into two conditions: in poor light at wide apertures, or with studio flash at moderate apertures. So I shot each lens at its full aperture, whatever that was, and again at f/8.
5) I cropped the images down to the same subject area. This means different images have different numbers of pixels... but right or wrong, it's the way I evaluate images: I look at something, usually a person's face, and decide whether it has enough detail or not, whether it's filling most of the frame or is just a tiny section. Again, since all I was trying to do was compare the G1's performance to the R-D 1's, I felt this was a valid way to do it.
So now that you've waded through my rationalizations for my goofy-sounding but pragmatic test method, you may want to know: where are the pictures: Okay...
This set shows all of them scaled to a constant size, because I usually print to fixed sizes rather than enlarging images by a constant percentage. If you stretch your browser window wide enough, you'll be able to compare them four-across -- each row shows either a middle (M) or peripheral (P) version, shot with both cameras at full aperture and f/8.
This set shows the sections at their original sizes, for your pixel-peeping pleasure. You can click on any small image to see the full-frontal-pixel version.
And for extra controversial fun, I've deliberately made it slightly difficult to tell which camera shot which picture! The letter immediately after the aperture indicates it, but not obviously. (Hint: If you know the parent companies' names, it should be easy.)
Spoiler -- My takeaway: The full-aperture images are softer than the f/8 ones, and the corner images are smearier than the center ones, as you'd expect. But I couldn't find any clear-cut case where the R-D 1 produced a
usable image, while the G1 produced an
unusable one.
So, after all that, I've decided: at least with the particular cameras and lenses I have, and with my admittedly lax attitude toward Ultimate Image Quality, I'm not going to worry about corner smearing, no matter what Reid says. But remember:
Your mileage may vary!
Oops, nearly forgot... for reference, here's an uncropped view of the whole setting. This was with the G1 and the 21mm lens, which means it's almost exactly the same as the R-D 1 with the 28mm lens: