Corner sharpness: Four Thirds vs 35mm (bandwidth warning)

Started Oct 27, 2007 | Discussions thread
joe mama Forum Pro • Posts: 12,623
Good info!

Do you know a lens that has it's best center sharpness wide open and
gets worse stopped down one stop?

Really? When I look at the PZ tests of any lens, it doesn't show this. Please explain more, if you don't mind.

PZ says their 28/2.8 does this.

I took another look:

and it's sharper (very insignificantly so) at f / 4 compared to f / 2.8. However, interestingly, it's all downhill from there! I've never seen such a think on any of their other lenses. In fact, I took a look at a bunch of other lenses (including the 24 / 2.8 and 28 / 1.8) and this doesn't happen.

Also, look at the MTFs from Canon. 24-85 at 24mm takes a huge hit at
the edges.

28/1.8 does the double dip - gets worse moving out from the center,
then better again a little further out before taking the big dive at
the edge:

28/2.8 does a similar double dip though less pronounced. Based on
MTF, it would be expected to have the best f/8 corners of the three.

Yes, yes -- they all take huge dips. That's why I qualified in previous posts that I wouldn't be surprised if FF corners were softer in the last 100 pixels or so.

At f/11, my 28/2.8 had better extreme corners than my 24-105L.


Here's someone else's comparison against 13 other lenses at 28mm
tested on a 5D ->

It will take me a while to go through those.

In any event, I would certainly wecome a test of zoom vs zoom: the
14-42 on 4/3 vs the 24-85 on FF. I think that's an ideal

I think an even more ideal comparison would be the 14-42 on 4/3 vs
28-105 on 35mm or better yet, 28-90 II on 35mm.

All fair game in my book, but why? Price?

Remember, I have no agenda except the facts. If the facts invalidate
my theory, then I will change the theory to fit the facts. However,
as I see it, this test is exactly in line with what I would have
predicted. No surprises to me.

Do you agree that there is an appreciable sudden dropoff in the 5D
system image extreme corners relative to the rest of that photo
whereas the E-410 system shows worse global quality without the same
degree of dropoff in the corners relative to the rest of the photo?

Absolutely. I saw the same pics as you, right? : )

I thought that your predictions were at odds with that outcome.

Like I said above, I even agreed that in the extreme corners that smaller formats may hold an edge. Just in the same way I noted in my essay that for the same aperture (not f-ratio), the Sigma 30 / 1.4 on 1.6x does, in fact, vignette 1/8 stop less than the 50 / 1.4 on FF.

I know you always saw me as being overly dogmatic and absolute, but I never meant to say that every pixel of a FF image is a better pixel than a crop image. Nonetheless, I still maintain that the corners of 4/3 and FF are basically the same. Only in that last super-tiny patch does the 5D fall behind, and I certainly do not deny that.

Now that we have your sample (thanks again for the excellent comparison), anytime I make my claim, I'll link to your pics. If someone tells me that I'm full of it, I'll take it as a difference of opinion. I mean, I honestly believe that there is basically no difference in those corners, except that ultra small patch that I would never even notice, anyway.

Still, I totally grant you that my standards are not the same as others. As you may well know, I've been on the other side of the argument saying that my 100 / 2, as far as I was concerned, had "basically" the same IQ as the 135 / 2L, even though we both know that not to be true. It's just that the differences were so small to me as to be inconsequential, as with the corners of your pics.

I totally apologize for any misinterpretations on my position that I have given. Once again, I am not in disagreement with your tests, and think you've provided a valuable service to those considering different systems.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow