Standard internet disclaimer applies: I have one copy of each of these. The Nikon 50/1.8G performs exactly as everyone says – very, very well. The Nikon 18-105, on the other hand, is possibly a bad copy; lots of people think theirs’ great, I think mine’s crap (not just versus a prime like the 50/1.8 – that’s never meant to be a fair fight – but against every other DSLR lens I’ve ever used, excepting a thirty-year-old 55/1.2).
All the following are focused at essentially infinity. The comparison was done from several typical photo spots in Yosemite, in case you care. It wasn’t as clear as I’d like on that day, but it was clear enough to show the differences between lenses and between apertures.
AF-S 18-105 @ 35mm (Why didn’t I compare at 50mm? Because 35mm was the better composition, and has relatively little geometric distortion)
- Very soft wide open (f/4.5).
- Substantially – surprisingly so – better at f/5.
- Optimal at f/5.6-f/6.3.
- Always soft in the corners.
For rough comparison (since I don’t own a 35mm prime):
AF-S 50/1.8 (same scene as above, just cropped in)
- Relatively soft ’til f/2.5 (compared to itself). Still better than the 18-105 @ 35 at any aperture. Yes, even wide open f/1.8 vs f/6.3 on the 18-105.
- Good at f/3.5-f/5. Strange weakness around f/5.6-f/6.3, particularly in the corners.
- Optimal around f/7.1 (though not significantly better than f/5). Much better than the 18-105 at any setting.
- Corners are a little soft wide open, but improve similarly to the centre from f/2.5 down. Quite good at f/3.5-f/5, soft again at f/5.6-f/6.3, then as good or better at f/7.1 as f/5 (it varies quite a bit, surprisingly, over time… it’s consistent across O(minutes), but O(hours) or greater and exactly how good f/7.1 is varies quite a lot… it’s not environmental because other lenses aren’t inconsistent like this).
- Declines rapidly from f/8 down. Yes, at f/8 it is softer than f/7.1 (consistently, though not always by much).
This is in typical conditions for me – hand-held, albeit braced, in Aperture-priority with auto-ISO on. Most of the photos came out at ISO 100 (it was a bright sunny day) but some were a tad higher. I took a couple of shots at each aperture, and kept the best. A few had motion blur in them, but not many; the keepers were almost all motion-blur-free. Certainly any residual motion blur was less significant than atmospheric distortion and the optics of the lenses.
One part of the results was new to me – the 50 being oddly weak from f/5.6 to f/6.3 – but the rest are things I’d already noticed from shooting with both these lenses for quite a while now.
Overall conclusions
At ~20″ display sizes (on screen), the 18-105 yielded a decent photo at f/6.3. But the photos from the 50 are visibly better, always. In fact you get roughly the same photo from both, but with a 3.5-stop advantage to the 50. (the 50 will have depth of field issues, perhaps, but the 18-105 will have soft corners anyway, so they’re actually more comparable than you’d think)
If I were printing these professionally, I honestly might try a print from the 18-105, but I wouldn’t set my hopes high (or print very large). It’s fine for a casual print; the difference is smaller than limits of basic paper anyway (but then almost anything is). Overall, it’s basically a good-light-only, casual-walkaround lens. Even on a tripod and with optimal settings it never really gets particularly sharp. Sharp enough, perhaps, but you can easily get much better results with other lenses (including lenses that are cheaper, if you’ll forgo things like focal length flexibility).
Also, I’d previously assumed the f/8 was a good default when I had enough light and didn’t need shallow depth of field. So says the internet. But I’d already tested that with my 80-400 and found it to be quite false. I assumed it was specific to that lens. But now I see these two lenses also are sub-optimal at f/8. Specifically, the edges & corners don’t get any sharper and the centre gets noticeably softer. There’s no apparent upside at all, other than the increased depth of field. f/7.1 is my new “optimal” default.
TODO: link to the original photos, once I upload them to Flickr.
