The Canon EOS and Photography Forums

Just a question.  I realize that different reviewers can get different results, but Bob's review of this lens and comparison to the cheapo EF 50/1.8 is completely at odds with the usually reliable Photozone reviews of the 50/1.2 (http://www.photozone.de/canon_eos_ff/472-canon_50_12_5d?start=1) and 50/1.8 (http://www.photozone.de/canon_eos_ff/415-canon_50_18_ff?start=1).  Bob's comparison shots of the air conditioner show the 1.2 to be far better than the 1.8 in the center at f/1.8, whereas the Photozone reviews show them to be about the same (though the Photozone 1.2 review gives figures for f/1.6 rather than for f/1.8).  Even more striking, Bob's shots show much clearer corners at f/2.8 with the 1.2, while Photozone shows terrible corners with the 1.2 lens at f/2.8 and quite good corner resolution on the 1.8 lens at f/2.8-- a dramatic difference, with 1874 lw/ph for the 1.2 lens and 2586 for the 1.8 lens--Photozone has the cheaper lens delivering 38% more resolution in the corners, yet Bob's sample shots show the more expensive lens giving dramatically more resolution there. 

So I've got to ask, what's going on?  A difference this huge can't be just a matter of opinion or minor technical differences.  I can't believe either Bob or Photozone would be lax enough to be comparing a properly focused image to an unfocused one.  The only thing I can think of is that you're essentially reviewing different lenses--it seems that Bob's got a defective 50/1.8 and Photozone's got a defective 50/1.2.  If that's what's going on, neither review is reliable--both are comparing a working lens to an essentially broken one.  So please tell me I'm wrong, and why!

I can't speak for photozone, but I can say that the results I presented were accurate for the two lenses I tested. There's always a possibility that any sample of any lens is either better or worse than an average sample. That's the problem with lens tests. As far as I know nobody ever tests 10 samples and presents the results of each one or even an average over the set. In fact it's very rare that even two samples are tested.

On the few occasions I've had two samples of a lens to look at, they've always been pretty close, but they've always been slightly different. There are sample to sample variations, but normally I'm assuming that they are small enough not to be significant.

I take lots and lots of shots when I'm testing lenses. Probably more than 100 and my general conclusions are based on the results of looking at all the images, not just the one or two examples I present. I would never base anything on a single shot or just two comparison shots. Uusually if I say that "A" is better than "B" under some set of conditions it means that I've reached that conclusion by looking at shots of resolution targets as well as real world images.

One point to be aware of is that shots of resolution test charts are normally done with a pretty close focusing distance. If the chart is 2ft x 3ft for example, a full frame shot with a 50mm lens would be taken at a distance of about 1.2m (4ft). Most of my real world shots are taken at much greater distances from the subject - conditions more representative of normal use. It's possible that lens characteristics may differ at different distances. If I notice a difference I'd comment on it (normally I don't see anything too significant in that regard). It's also more likely that real world outdoor shots will be affected by flare than controlled indoor studio shots.

As for focusing, I normally test using both AF and using manual focus with Live View and 10x magnification. If there's a difference I'll comment on it (as I will in an upcoming test but I don't want to give details now until I've carefully checked my results).

So I can't explain differences in lens test results other than sample to sample variations or some flaw (or at least some difference) in testing methodology. I'm pretty sure my results are good but I can't really comment on anyone else's data because I've no idea how it was obtained.

To add some more oranges to the apples, here is Ken Rockwell's review of the 50 mm macro:
http://www.kenrockwell.com/canon/lenses/50mm-macro.htm

He thinks that the macro is the best 50mm lens Canon makes.

I would not be in the slightest bit surprised of the EF 50/2.5 was the sharpest of all Canon's 50mm lenses at its optimum aperture. It's designed to be very sharp with a very flat field, as all macro lenses are.

If you want a 50mm lens that's as sharp as possible but no faster than f2.5, then it would be the one to get. However to say it's the "best" depends on what your use is. If your use is for low light work and/or portraits with a shallow depth of field, then of course it's the worst of all the 50mm lenses!

Thanks for the explanation.  That helps.  I can certainly understand why it's not really possible to check out several copies of a particular lens.  It also hadn't occurred to me that I was comparing near-distance test chart shots with normal-distance real world shots--thanks for pointing that out. 

(As a secondary note: www.slrgear.com has actually done some MULTI-SAMPLE reviews with very interesting results. Here is the one for the  Canon 50mm ƒ/1.4 USM. It can be found here: http://www.slrgear.com/articles/variation_canon50f14/canon50f14.htm ) Hope that helps as well.

This may be a late answer, but I did a comparison between two lenses and found very similar results to BobAtkins'. In fact, after I finished writing my blog I saw Bob's review and thought "someone took my text and translated into English"  :) .
http://halkboyleistiyor.blogspot.com/2010/06/50mm-f12l-ve-50mm-f18-mii-objektif.html
It's in Turkish but you may compare the sample pictures and crops.

Better late than never !

I don't know what the papers and charts & tests say...
The 50/1,2 is fantastic.
I had some difficulty realizing it with the first shots.
The dof is so small that shots get easily out of focus.
But the results are magical !
I absolutely adore it and couldn't do without it !

Even if overpriced, still worth every penny !
10 out of 10.
Can post results if anybody wants...
George