The Tamron zoom 17-50mm f2.8 is one incredible lens for a crop sensor body. I tested a couple on a Canon Rebel xti, and they have truly impressive resolution and contrast out to the edges of the frames, as both the photozone and Husband's tests (I'm pretty sure you tested one of these lenses too, right?) also show.
Husband stated:
"However, if we take the adaptor out of the picture altogether, and look at Photozone's test – on a Nikon body – we see centre frame resolution scraping 2200 lwph, but the ZF25 doesn't even reach 1950 lwph on the DX frame edges until f8. Contrast this with the Tamron 17-50/2.8 zoom which resolves 1950 lwph at the frame edge . . . at f4 . . . at 17mm!"
Lotusm50 stated:
"I think the Tamron zoom lens comparison you pointed out only underlines the potential that this was a production problem. The lens, correctly executed as designed, simply could not be that bad."
I think the idea of testing on canon bodies using a third variable, the adaptors, are flawed from the get go. This is a disservice when to get a definitive answer to how these lenses perform is so easy - by testing them on the mount that they were designed for - Nikon.
Why are these NOT tested on nikon bodies? Or am I mistaken and they have been?
If they work as expected on nikon and then not on canon - then we have another mission to find out what is causing the poorer than hoped for performance there and if they do not perform well on the Nikon's then we can conclude either sample variations or simply poorer lenses than advertised. I can see the value in testing on canon bodies since that is the way so many may have to use them but it's only possible to draw final conclusions from using the platform they were designed for. The test methodology however seems to be the one thing we can rule out. It's simple, fair and real world.
If you compare the Photozone tests of the ZF25 (on D200) and the 24L, the ZF25 is consistently better.
This whole idea of comparing the ZF with a "properly sorted" 24L seems flawed, to me. I would say that a "properly sorted" ZF 25 aught to be compared to a "properly sorted" 24L, which seems to be the case with the Photozone test (although, I'm not quite sure about the 24L they tested).
In the end, the Photozone test tells me that the ZF25 is, indeed, a super lens. Maybe not the best option for EF mount due to possible adapter issues, but we've seen these issues before with various other alternative lenses and they are "fixable" issues, for the most part. On the other hand, the 24L is much faster and provides full AF and auto stop-down.
Time will tell whether Cosina can end up making more "good" samples of the ZF25 than Canon can make "good" samples of the 24L.
One other thing is certain. The ZF25 certainly does *not* blow away the competition quite as convincingly as does the CZ21.
If I had a 5D and the Imatest software, I'd love to do Photozone type testing of all sorts of alternative lenses adapted to EF mount. They seem to have something against adapters, and I guess their reasons are sort of obvious, but I'd like a similar resource devoted to adapted lenses.
Pavel wrote:
I think the idea of testing on canon bodies using a third variable, the adaptors, are flawed from the get go. This is a disservice when to get a definitive answer to how these lenses perform is so easy - by testing them on the mount that they were designed for - Nikon.
Why are these NOT tested on nikon bodies? Or am I mistaken and they have been?
If they work as expected on nikon and then not on canon - then we have another mission to find out what is causing the poorer than hoped for performance there and if they do not perform well on the Nikon's then we can conclude either sample variations or simply poorer lenses than advertised. I can see the value in testing on canon bodies since that is the way so many may have to use them but it's only possible to draw final conclusions from using the platform they were designed for. The test methodology however seems to be the one thing we can rule out. It's simple, fair and real world....Show more →
1) Hubsand refuses to test on Nikon, due to what we pay him.
Lotusm50 wrote:
Actually, that is not true in Zeiss' case. Zeiss' publish MTF's are based on tests from actual production lenses. Canon's MTF data is theoretical -- and there is a big difference between theoretical and actual MTF data.
If that's true for this lens, is the edge ff or a 1.5 crop?
The MTF charts have dimensions from center (in mm) listed on the horizontal axis of the graph, so we can see it is for the full image circle that the lenses are designed to produce, 44mm, which is full frame 35.
cogitech wrote:
Yes. Well, 1.5x on the Nikon side.
Maybe it is not 100% sound to compare the performance of the ZF25 and the 24L on crop bodies and then extrapolate, but I'm comfortable with it.
But the FF edges is the main issue. So if your a Nikon user, and sure that FF is a Canon marketing ploy, then there's no reason to consider most of the problems Hubsand found.
And as Hubsand said, the ZF25 looked darn good compared to most of what he looked at.
The 24L definitely does suffer from sample variation in the real world. The copy I have now seems to be unusually good. But the ZF25 – on a Canon body – has the finest centre frame performance of any lens at that focal length, and at f11 it is unsurpassed.
However, as used on either adaptor, the corners do not hold up at wide apertures, especially at close range. In all these areas, the 24L excels: it's designed to make the ZF25 look bad!
If we look at Photozone's tests – on a Nikon body – we see that the ZF25 only just scrapes a peak resolution at f8 of 1950 lwph at the DX frame edge (that's Zone B / mid-frame for the Canon user). Let's consider other optics DXO tested:
• The 17-55DX at 24mm exceeds 1950 lwph at f5.6.
• The Tamron 17-50/2.8 gets there at f4.
• The Tokina 12-24mm at 24mm gets there at f5.6
• The Nikon 18-200mm at 24mm gets there at f5.6
• Even the Nikon kit lens resolves 1950+ at the borders of a DX frame at f5.6
Lotusm50: are you suggesting that both tests were faulty? Or that all the adaptors have stopped working? Isn't it more plausible that there are teething problems with early samples?
What I mean is, the ZF25 edges (and centre) on 1.5x beat the 24L edges (and centre) on 1.6x in the Photozone tests (quite convincingly so), so I think it is not too far of a leap to assume that it would continue so to the edges of a FF.
As you can see, the ZF is about 250 to 300 points better than the 24L accross the board. Note that the ZF scale goes to 2250 and the 24L scale only goes to 2150.
Edit: I guess it is more like 150 to 250 points...
Edited by cogitech on Feb 06, 2007 at 01:03 PM GMT
Maybe hubsand's 24L would produce much better scores, and maybe his ZF25 would produce much worse scores, but my bet is there are (will be) more "good" ZF25s than 24Ls.
hubsand wrote:
The 24L definitely does suffer from sample variation in the real world. The copy I have now seems to be unusually good. But the ZF25 – on a Canon body – has the finest centre frame performance of any lens at that focal length, and at f11 it is unsurpassed.
However, as used on either adaptor, the corners do not hold up at wide apertures, especially at close range. In all these areas, the 24L excels: it's designed to make the ZF25 look bad!
If we look at Photozone's tests – on a Nikon body – we see that the ZF25 only just scrapes a peak resolution at f8 of 1950 lwph at the DX frame edge (that's Zone B / mid-frame for the Canon user). Let's consider other optics DXO tested:
• The 17-55DX at 24mm exceeds 1950 lwph at f5.6.
• The Tamron 17-50/2.8 gets there at f4.
• The Tokina 12-24mm at 24mm gets there at f5.6
• The Nikon 18-200mm at 24mm gets there at f5.6
• Even the Nikon kit lens resolves 1950+ at the borders of a DX frame at f5.6 ...Show more →
Forgive me if I am wrong, but I'd say the corners of all the lenses mentioned here would be pretty bad on FF
cogitech wrote:
What I mean is, the ZF25 edges (and centre) on 1.5x beat the 24L edges (and centre) on 1.6x in the Photozone tests (quite convincingly so), so I think it is not too far of a leap to assume that it would continue so to the edges of a FF.
Oh no. MTF charts often take a big dive after were a 1.5 or 1.6 crop camera stops.
But it does indirectly bring up the point that the 24L is a different purpose lens than the zf25. The fastest (and biggest) 24mm lenses is designed for people who need speed and/or want some dof control at 35mm.
A 2.8 24 is more of a general purpose lens. Often stopped down to maximum sharpness. but usable wide open when necessary.
dcmiller wrote:
Oh no. MTF charts often take a big dive after were a 1.5 or 1.6 crop camera stops.
But it does indirectly bring up the point that the 24L is a different purpose lens than the zf25. The fastest (and biggest) 24mm lenses is designed for people who need speed and/or want some dof control at 35mm.
A 2.8 24 is more of a general purpose lens. Often stopped down to maximum sharpness. but usable wide open when necessary.
Yes, I am aware of the drop-off that is so common, but is there reason to assume the drop-off of the Zeiss would be much worse than the Canon? (esp. considering that it bests the Canon accross the board?) I'd actually suspect that the ZF25 might actually curve back to sharper, like the CZ21 does.
You are absolutely correct about the different intended uses of these lenses. The ZF sucks at f1.4 and f2
cogitech wrote:
What I mean is, the ZF25 edges (and centre) on 1.5x beat the 24L edges (and centre) on 1.6x in the Photozone tests (quite convincingly so), so I think it is not too far of a leap to assume that it would continue so to the edges of a FF. http://www.photozone.de/8Reviews/lenses/zeiss_zf_25_28/mtf.gif http://www.photozone.de/8Reviews/lenses/canon_24_14/mtf.gif
As you can see, the ZF is about 250 to 300 points better than the 24L accross the board. Note that the ZF scale goes to 2250 and the 24L scale only goes to 2150.
hubsand wrote:
There's no doubt that his 24L was defective.
Given the wide disparity between the MTF's and what your test showed for the ZF 25mm, it wouldn't be possible that your ZF 25mm was similarly defective, would it?
Photozone's tests are very clearly stated to NOT be comparable between systems. Nikon tests on a 10MP D200 can not be compared to Canon tests done on an 8MP Rebel...
I revisited Hubsands page again and had a look at the MTFs: The ZF25 seems to be a good sample for the following reasons!
Speaking of the B Zone, the Canon is clearly better up to f4, they are on level at f5,6, probably the Zeiss is a bit better above.
In zone C the Zeiss' published MTF shows that wide open and at f5,6 the corners are not so sharp, see that dive at 20mm! Hubsands corner tests are from the last 250px, he had to find these more or less sharp corners up to f5,6
Zeiss may have sacrificed some qualtiy in the corners to get better performance in the center where crop cameras need the highest quality lenses to resolve up to 12mp and maybe more in the futur.
I guess we will not see better samples of the Zeiss but some may get a 24L which does not reach Hubsands lens due to Canons lousy CC.
