The image above should be viewed at "Original." (note: it's about one mB. Sorry for the large size, but I don't want jpg compression to muck up the study.)
If you've been studying photography and optical performance for any period of time, likely you've read that lenses perform best at their medium apertures. Of course, there are a few premium lenses that look rather amazing "wide open," but for the vast bulk of the lenses we use, the middle numbers provide the best performance.
So here's the proof. The example above is with the very common Nikon 50 f1.8 lens (AF-D), but you'll find similar performance with most lenses. Each image is a one-hundred-percent crop from the image's center, so we're looking at the sharpest area of the lens's field. All images received the same amount (+1) image sharpening in-camera (D70), with no post-processing. Camera was tripod-mounted with available light exposure.
Note that I did not include a shot at 1.8. It's no prettier than f2, just a touch softer (yikes).
So I'm a bit saddened by all this. I love my primes, and one of the reasons I buy prime is the performance one assumes to be had in terms of sharpness and range of available apertures.
Yes, and of course I know about the 1.4, and I know it performs better wide-open. My point here is not to slam any lens, or to seek recommendation. Rather, this post is to suggest that it's probably wise to run some tests on your lenses to find out where they really perform best. Then, you can make an informed decision about aperture versus shutter speed, and which lenses to put in your kit.
I've done similar tests on my 35 f2 AF (non-D, from my 8008 days!) and sadly, it also looks much better at f4 or smaller.
One nice surprise was an eighty dollar eBay purchase, the old Vivitar 135 f2.3 Series One. Very sharp from 2.8 and up.
My biggest surprise is the 18-70 kit lens. It's nearly as sharp wide open (granted, that being f3.5-4.5) as it is at any other aperture throughout its range. That lens is just amazing.
Of course, there is no denying that even if an image is soft wide-open, if that wide aperture allows you to capture an image, then it's worth it.
Back in the film days, few amateurs ever did resolution tests; it was too expensive and laborious. Now, we have no excuse. I don't regard any of this as pixel-peeping. It's learning to maximize the performance curve to take best advantage of what you have.
Now, what I said at the beginning, about this stuff being common knowledge, well, that's true. It's just that I always kind of thought it to be theoretical, the stuff of arcane bench tests, as meaningful as "line pairs."
And for you macro fans: I strongly suggest that you try a similar test on your 1:1 jewel to see if you really can get decent resolution at f22-32. I'm guessing you'll find that f13 to f16 blow f22 away, and you'll never stop fully down again once you see your results. Sure, maybe you'll give up a tad of DOF, but the area of sharp focus will be as sharp as it can be.
And for a lot of what we do, isn't that precisely the point?
What's the phrase? "F8 and be there." Ohhh, now I understand.