This is Part Two of the rolling review of the 100mm macro lenses for Canon EF mount.
Studio Test. Ok folks, we are starting off with a fairly standard Imatest evaluation of our primary candidates. Just as a reminder, we will limit this test to only one focusing distance - 10m. Why 10m? For all 100mm lenses, 10m focusing distance roughly falls into medium range and closer to the infinity than to the MFD. I cannot test lenses at longer than 14m focusing distance due to the space constraints. And testing at very close focusing distances with Imatest does not really make sense due to the software's and imaging target's limitations - we will be testing the resolution of the target (printer that is) rather than the lens at anything closer than 2m (20x focal length of the lens). At 10m we get a 100x focal length ratio, which should be good enough to get us accurate results out of Imatest (as accurate as the software is capable of giving us, of course). Plus, we will do macro (MFD) and infinity tests separately, so, yea, 10m sounds like a good start.
Besides the usual MTF50 charts that we get out of Imatest, we will also take a look at the lateral CA that Imatest measures as well as vignetting and distortion figures. However, before we dig in into the results, let's review a couple of important things that we need to keep in perspective:
And now, without any further ado, here are the MTF50 charts for the tested four lenses
Ohh my, what do we have here? Canon is clearly the laggard at f/2.8, and not just by a hair. This is surprising to me - I tested (a different sample of this lens) back a few years ago and I don't remember it being that significantly off around borders at f/2.8. Granted though that I tested that lens on a lower resolution cameras - 10Mp Canon XTi and 12Mp Canon 5D. Actually, let me step back a bit and start by saying that if I were reviewing each of these lenses in isolation, I would have probably given them fairly high marks in the resolution department. Canon sticks out by a lot because the other lenses have just a darn high resolution, particularly Carl Zeiss. Even Canon's center resolution is pretty high (just not as high as Zeiss but high enough). It's just Canon's border resolution at f/2.8 that is showcasing some weakness - if we look at the crops from the imaging target below, we would cleealy see the delta that Imatest is pointing to.
The rest of the pack (lenses that is) have a fairly similar set of characteristics - center or borders. There are some differences here and there, most of them are within 10%-15% variance threshold, and even Canon shows very good numbers from f/4 and on. Carl Zeiss is the clear leader here and it bests all other lenses throughout the aperture range, in some cases just by a hairline, but still the leader. Leica and Voigtlander are pretty uniform across the frame and don't show significant variance from aperture to aperture. So it's probably safe to say that all lenses, except Canon, have superb resolving capacity. But what should we do with Canon though? The weakness it is showing is not due to the focusing error - I've tested and re-tested this lens a dozen times. The lens does have a very short focusing thrust, making it a huge PITA to focus manually since even a tini-tiny shift in the focusing ring results in a huge focusing changes. And trying to auto-focus it on a target - forget 'bout it! AF ends up all over the map, but practically never exactly on target. Yea, I don't want to rush with any conclusions yet - let's see how the lens performs in regular field testing, but I don't like this softness, it simply should not be there IMO. At least not that much. But let's move on for the moment.
Our next chart is lateral chromatic aberration. Here results are a little bit more in norm - Canon and Zeiss have slightly higher levels of CA around borders, while Leica and Voitlander, the two APO lenses, have lower CA levels. But neither of the lenses have any alarming levels of chromatic aberration - we will review the actual CA in subsequent parts when we do field testing, but if we are to believe Imatest, at these levels CA should not be visible in most cases.
As a general rule, short-tele lenses don't really exibit any significant distortion, and macro lenses in particular. The chart below confirms that - Canon shows the 'highest' barrel distortion at ~0.5%, while Voitlander shows 'highest' pincushion distortion at ~0.2%. Zeiss and Leica are essentially distortion free. But even Canon's relatively high distortion is fairly negligible and should not pose major problems - we will confirm that though in later parts of the review.
While distortion is quite low in all lenses, vignetting is a little bit more of an issue for these lenses, particularly for Zeiss, which is the fastest lens. At f/2 the lens shows ~1.8EV of falloff, which is certainly going to be visible in general photography and probably needs to be corrected either in camera or in post-processing. But at the same time, Zeiss has the lowest vignetting at f/2.8, where most other lenses average ~1.1EV of falloff. Beyond f/2.8 vignetting becomes much less of a problem and most likely would not visible in the virtual majority of cases.
Ok, so this is the end of part two of the review. The major takeaway here is Canon's surprising softness at f/2.8 around borders. It is not clear yet though, whether this softness is because the lens can't match the capabilities of a higher res camera or it's a problem with the sample itself. I will continue with the tests, using the same sample, but will also try to secure another to retest the lens under same conditions. Stay tuned for part three as well as an update on Canon 100/2.8.