Conclusion on pixel peeping

What is the conclusion here? Did I find the same results as Stephan Kölliker?

Not quite. At 28mm, my CZ appears to be a bit better than his and his 28-75 appears to be a bit better than mine. The 24-105 are quite similar. At 70mm, my 28-70 is noticeably worse then his (at the center as well, so maybe I should indeed send this lens back for service), and the CZ and 24-105 appear quite similar.

As a side note: this test confirms what I already showed in this blog, namely that the 24-105 is indeed quite a good lens at f/8-f/11 (except for lateral chromatic aberration) …or rather that it is a good lens between 28mm and 70mm. At 24mm, the results between this lens and the CZ are vastly in favor of the CZ (even if the CZ corner sharpness is not very good at 24mm, the 24-105 is much worse).

But what about the CZ corners at 28mm? Are they as poor as Stephan thinks?

Let us examine the picture at f/5.6 again:

Isn't the right side sharper than the left side of this picture?

Let us look at the FTM again (from Sony Japan).

The scale at the bottom of this chart shows the distance to the center of a 24mmx36mm frame. So:
-12mm is the top or bottom edge of the frame (24 divided by 2)
-18mm is the left or right-hand edge (36 divided by 2)
-21.6mm is the distance to anyone of the 4 corners.
So what the FTM shows is that the picture stays sharp up to the extreme corners.


Let us see a bit more of the corner at f/2.8. The following picture is 800x1000 pixels, from the bottom left corner of the full picture. Here it is quite clear that the right-hand part of the picture is much sharper than the extreme corner. The top right corner is at 14mm from the center. I added figures to indicate the distance from the center in mm, so that you can compare the figure with the chart (but note that the chart is taken at 24mm and f/2.8 and this picture is at 28mm and f/5.6 so that the unsharp zone should be a bit smaller and further away from the center).

Note: that image is clickable and will open in its own window or tab if you want to examine it at the pixel level.


If we consider the full picture, the four corners will lack sharpness. Next, I tried to visualize the unsharp zone at f/5.6. Of course, this is an approximation: there is no clear limit to an "unsharp zone", but since the transition between "pretty sharp" and "pretty fuzzy" is rather steep on the FTM, we can draw some fuzzy limit, take or give some pixels. In addition, I drew the two crops: the small one as from Stephan and the bigger one from above (but without the numbers).

We find out that Stephan's crops are exactly from that steep transition zone.

Furthermore, we should realize that if the lens is not perfectly centered, that "transition zone" will move a bit (actually, the four corners will move and they won't be symmetrical). Could it be that Stephan's lens was not very well centered and that the crop ended in the worst corner? Only Stephan can tell, since he has access to the original pictures.


So what is the conclusion for us photographers?

Does the CZ have unsharp corners? Undoubtedly yes, even Sony acknowledges it.

Are the 28-75 or 24-105 better, then? Not really. I would say that the differences at f/8-f/11 are unlikely to be noticed on prints in practical use. At f/2.8, the CZ has better center sharpness than the Tamron (but not by much).

Does that mean that the price and weight of the CZ 24-70 are justified? Honestly, that is for you to decide. The CZ is better, but it is not perfect (which lens is?). However, in 90% of the pictures, you are not likely to see the difference in sharpness with the 28-75 unless you print huge enlargements or pixel peep. Then, the CZ is better built, has 24mm and SSM and the colors match better with the other CZ glass (16-35, 85 and 135mm). That and my Tamron does not work very well at 70mm, while I think it used to… YMMV, of course.

What if you want sharp corners at 28mm, then? Stephan suggests the Minolta 17-35mm f/2.8-4. The CZ 16-35mm f/2.8 is very good as well. What about 70mm? Stephan suggests the Minolta 28-135mm f/4-4.5. I don't own that lens, but the Sony 70-200 f/2.8 G is very sharp as well (see my test in November)...

70mm: Zeiss 2.8/24-70mm, Tamron 2.8/28-75mm, Minolta 3.5-4.5/24-105mm

These are the results of the "slanted horizon" at 70mm. The pictures are directly displayed at pixel level here, you don't need to click anything.

First the full aperture: f/2.8 on the CZ and Tamron, f/4.0 for the Minolta zoom.

lens	Corner	Center
CZ
Min.
Tam.

Then, the same pictures at f/5.6 for the three lenses.

lens	Corner	Center
CZ
Min.
Tam.

And the pictures at f/11:

lens	Corner	Center
CZ
Min.
Tam.

Once more, I'll let you jump to Stephan's page so that you can make your own opinion. I will give my interpretation of these tests in the next page.

28mm: Zeiss 2.8/24-70mm, Tamron 2.8/28-75mm, Minolta 3.5-4.5/24-105mm

These are the results of the "slanted horizon" at 28mm. The pictures are directly displayed at pixel level here, you don't need to click anything.

First the full aperture: f/2.8 on the CZ and Tamron, f/4.0 for the Minolta zoom.

lens	Corner	Center
CZ
Min.
Tam.

Then, the same pictures at f/5.6 for the three lenses.

lens	Corner	Center
CZ
Min.
Tam.

And the pictures at f/11:

lens	Corner	Center
CZ
Min.
Tam.

I'll let you jump to Stephan's page so that you can make your own opinion. I will first present the pictures at 75mm in the following page and then give my interpretation of these tests.

A900: Zeiss 2.8/24-70mm, Tamron 2.8/28-75mm, Minolta 3.5-4.5/24-105mm

Not so long ago, Stephan Kölliker from artaphot published a lens comparison between the Sony - Carl Zeiss 24-70 f/2.8 zoom and 4 other lenses which provoked quite a stir on the dyxum forum (this thread) and on the dpreview Sony DSLR forum (this thread). Basically, the test criticised the corner sharpness of the Zeiss.

As you might imagine, people started to criticize the test, explained that Stephan must have used a bad copy and started various theories as to what the test was really showing. Photographers can be quite opinionated people and many do not like when expensive brand names are criticized. Such is life.

Quite pointedly, Stephan gave a link to the official FTM charts at Sony Japan and explained that if people doubted the results of test, they could redo it and post their own results.

Fair enough, here is my version of the test.

The lens I could gather for the test are:
-the Sony Carl Zeiss 24-70 f/2.8 (of course)
-the Minolta 24-105 f/3.5-4.5 (the same as Stephan used)
-the Tamron 28-75 f/2.8 (which should be of the same optical design as the Sony 28-75 f/2.8 that Stephan used).

I could not find the Minolta 28-135mm f/4-4.5 or Minolta 28-85mm f/3.5-4.5.

In all fairness, I should add that my copy of the Tamron behaves a bit strangely. I made particular efforts to get proper focus, but nevertheless what you see here may not represent the best that this lens can do.

The setup of the test is very simple, it is just a landscape image with the horizon diagonal through the corners. Here is the full picture, reduced for the web:

The tables in the following pages are crops at the pixel level, corner crop on the left and center crop on the right. The corner pictures are taken at the same place as Stephan (as far as I can tell): 200 pixels from the corner:

The center pictures are narrower for the table to fit on the page. The center pictures are just here so that we can see whether the lens was focused correctly (they come from the same file as the corner crops). We are not here to judge center sharpness here, everyone agrees that the CZ is quite sharp at the center.

Pictures were taken free-hand for lack of time (but we can check from the center pictures that there is no camera motion blur) and the raw files were processed in Apple's Aperture. No sharpening was added.

On the next page, the pictures at 28mm.

Conclusion or "why have lenses become so big?"

When I compared lenses, you may have noticed that I regularly made comments about size and weight. In particular, when comparing the Tamron 28-75 to its Carl Zeiss equivalent.

You will understand if you look at this picture (it's not the 24-70 but the similarly-sized 16-35):

On top of it, the Tamron is almost half the weight. I can insure you that you will notice when carrying the A900 with one lens or the other.

Maybe the Tamron is not the best example. After all, even if it can give very good pictures, I have had AF problems with it regularly.

I also commented about the 24-105 being an "ideal travel lens". Let us look at the difference:

Now, that is a lot of difference, and the lens has much bigger range as well. Of course, it is not as fast. In particular, if you want best results you should use f/8 (f/11 at 24mm). But for a day trek in the mountains, which one is the best choice?

Of course, you can even go smaller and lighter if you use a prime, here the 50mm:

Or look at the difference between the old 135mm f/2.8 prime and the huge 70-200 f/2.8 G zoom:

(and the prime has even a built-in hood... why don't they do that any more?)


Bottom line: there is more to a lens than test pictures. The lens with the worst picture quality is the one you left at home...


And it seem that every manufacturer is building their 35mm lenses bigger and heavier nowadays.

135mm, 3 lenses, 2 primes, 1 zoom, corner.

I am finally reaching the end of these comparison here (I won't do 200mm on this subject, it is not far enough). Corner performance of the:

Sony 70-200 f/2.8 G
Minolta 135mm f/2.8 (another 20 years old lens)
Sony 135mm T4.5 STF.

(Even if two lenses are called "Sony", all three are actually Minolta designs).

f-stop	Minolta 135mm f/2.8	Sony 70-200 f/2.8 G	Sony 135mm STF
f/2.8-T4.5
f/4-T5.6
f/5.6
f/8
f/11

Interestingly, the best corner performance comes from the tiny 135mm f/2.8 prime (look at the writing on the newspaper). The STF does not quite live to its reputation, since is the least sharp of the three. In all fairness however, I should point out that all three lenses are close to perfection at full aperture and that the differences are inconsequential.