Finally a lab test of Tamron AF 17-50mm f/2.8 XR Di II LD VC

>> The biggest weakness of the lens is the rather soft corner/border performance at large aperture settings. The quality increases when stopping down and it's certainly possible to achieve very high quality results from f/5.6 onwards (a bit of corner softness remains at 17mm). <<

Yeah, great - use at it 5.6! That beats the very purpose of buying a constant 2.8 aperture zoom lens. I bought this lens in December, exchanged it due to a F-- error, and returned the exchange after 5 days. I am not a pixel peeper, but the lens did not meet my expectations as a serious amateur :-(

Sadly Nikon does not have an equivalent lens (their 2.8 zoom does not have VR).

this review clearly states the VC model has worse IQ than the original model (which is fairly sharp at 2.8). the review is for canon mount, however.

I've been very interested in this lens, but since the review is for Canon, I'll tell ya' what I'm gonna do; go over to Samy's Camera (near me in Los Angeles) and test the lens for myself on my D70s, and on one of their D300 (the camera I intend to buy). I may even try a couple of the lenses just to check variations from lens to lens. This will be a real world test, of the sales people, the countertop, shelves, wide open in mixed light.

I don't find these strickly MTF-based evaluation too meaningful. Mind you that even the Nikon 17-50/2.8 only eared three star in their "optical quality" test. This emphasis on corner to corner sharpness is not always relevant to the way that we shoot. For example, most of the people would use f2.8 to achieve shallow DPF to pop the subject in the center of the frame; thus, what is the big deal if the corner is "soft?" If you compare the MTF number between the Nikon and this Tamron lens, you will find that even though the center performance of the Nikon is not too far off from that of the extreme corners, in terms of the absoulte MTF score, the center from the Tamron lens is in fact sharper than that from the Nikon:

Nikon at 2.8 from center to the extreme corners:
2000, 1800, 1500

Tamron:
2300, 1800, 1400

The dependence on MTF test in many ways may be misguided. It is easy to conduct the test by shooting a fixed target in the studio but how this lens performs in the field is often another story all together.

While I generally trust photozone reviews I fully agree with CC Chang here. While I do not own this lens I did PP for about 800 pics done by a friend of mine in available light at a conference, using only this lens mounted on a D90. Honestly I was pleased with results... I also borrowed it for a couple of days just to check the quality of VC... and it's impressing. Only when shooting in low light on noncontrasting surfaces I was dissapointed by it's merely hunting behavior.

Sure, I will not recommend this for a pro shooter... but for a person that is on a low budget this is a very decent lens.

good point, CC. photozone's MTF tests appears to confirm the 17-50 is sharper than the 17-55 at 2.8.

last night, i was shooting with the 17-50 in extreme low-light conditions (inside a dimly-lit nightclub, no flash). corner performance was essentially meaningless. the tamron+d300 combo did pretty good at 2.8, 1/200, iso 3200.

Pretty much every lens I have hunts in low light, low contrast situations, but I'm glad to hear that real world and center sharpness is good. I'm a professional and have no qualms what-so-ever using this lens, or any other quality third party lens for my work (Tokina 12-24, Sigma 50-150 in my bag).

i think sometimes people hunt for reasons not to like 3rd party glass other than its not OEM, and, at times, it's a bit of a reach. the fact is the third-party guys make lenses which plug gaps in the Canon/Nikon lineup, and some of them are gems. to me, its not about being a nikon fanboy, it s about having the right tools to get done what you want to get done.

I have the original (non-VR) version of the Tammy (sic). I have sold more prints from that lens than from any other I have owned. And I really can't see much reason for VR at a 75mm equivalent.

Funny, I greatly value my photography - and I can't find much fault in this lens.

It clearly can do things no other lens for Nikon can do, and apart from corner sharpness, I can't find much that would be negative about those lab tests, and of course corner sharpness at f2.8 is strongly overrated :)

Distortion at 17mm? Yup, but all others have it as well, at least some. Heck, even Nikon's primes have some. PTLens still does not support this new version of the Tamron, but its correction for the old version works reasonably well at 17mm. Just don't use it at other focal lengths, because there the new version is pretty nice anyway. And if you really shoot architecture professionally and can tolerate neither distortion nor software correction? Well, you can always get a Zeiss 21/2.8. Pretty expensive, not stabilized, no autofocus, but pretty much your only option. But then, for architecture you may be better off with FX and a 24mm T/S lens anyway.

Bokeh is really fine for a zoom and worlds better than the Nikon 16-85 VR, CA is surprisingly good, and don't get me started on my pet feature, the short minimum focusing distance. And stabilization? A killer feature, even at 17mm. Once you have it, you never want to go back. All that in an affordable price range.

What's negative? A tendency to not focus at times. It does not happen often, but when it does, I need to switch the camera off and on again. This may be more a problem of the D300 than of the lens, but is is exhibited with no other lens that I have, and I have quite some.

The F-- error, well, I see that at times when it is cold, meaning sub-zero (Celsius). On the D300 it is reported as F1.2, but the source is the same, a problem in camera/lens communication. I have seen that with Tokina lenses as well, never with Nikon and Sigma. It also reliably goes away when I switch the camera off/on.

Finally there is some more ghosting than I'd like, when I have the sun in the frame, especially near the edge.

Nish, I'm sorry that it didn't work for you. For me it clearly does.

Understand me well, it does bother me that I sometimes have to reset the camera by switching it off/on, but what counts in the end are the pictures. This lens is far from perfect and due to that certain unreliability I would not recommend it for a sports shooter or a professional, but Nikon just has nothing that even gets near in usefulness. Up to now, 87 of my last blog entries were taken with that lens, that is more than with most of my lenses. That's almost three months of continuous, daily usage. Charts can tell you only part of the story, and it may come as a surprise, but it is not even the most important part.

Sadly Nikon does not have an equivalent lens (their 2.8 zoom does not have VR).

This test basically shows why Nikon does not and should not put VR into their f/2.8 standard zooms: the image quality will drop and since VR is not essential in this focal length range, it would be a net loss for quality conscious photographers. Although in event photography corner sharpness is typically not a priority (it can be in some group shots) in landscape photography it is, and people will use these lenses for landscapes too. The VR would only allow those blurred people in sharp surroundings handheld shots which I find appalling anyway. It's much better to use a tripod for those landscape and architectural shots and get the full image quality possible, corner to corner, and use high ISO for the people shots so that the shutter speed is high enough to stop movement. Thanks to Tamron we now have confirmation of the issues involved in putting VR into this type of optics.

Ilkka,

There are different kinds of people doing different types of photography :)

I for example photograph every day, be it summer or winter. I photograph landscapes, details, macros, street scenes, whatever I find that tickles my fancy that day. In winter there is almost always low light, and it's even worse during the week, when I can only take photographs before or after work. Some of my images are of people, the majority is not. Stabilization is incredibly valuable to me, yes it is the key feature of this lens for me. I suppose there are others in the same boat.

As to a standard zoom's image quality dropping because of stabilization, well, that's completely beyond my grasp. Every manufacturer on the planet stabilizes their standard zoom, be it 16,17-50+ on DX, be it the 24-70/2.8 on FX. Every manufacturer but Nikon that is. They do it by either stabilizing the lens or by stabilizing the sensor. Both work, and I have not the impression that the Nikon 17-55/2.8 or the Nikon 24-70/2.8 are vastly superior to the equivalents from Canon, Zeiss for Sony Alpha, Olympus, etc.

I know that the Nikon 24-70/2.8 is known as an extraordinarily good zoom (I don't have it and have not even tried it), I believe to remember that you own it and use it on a D700, and I happily admit that it may be the best of the bunch. But vastly superior? And what it is, it is for a price. Nikon just had to have something superior. This lens, along with the 14-24/2.8 and the D3, was their ticket back into professional photo journalism, and I admit as well, that stabilization makes no sense in photo journalism. That they left stabilization out may have been a simple matter of costs. Of course they can make the best lens in the world and of course they can make it stabilized, so could Canon and all the others, the question is only, where's the market? Who would buy it? They compromised, or they were not ready to stabilize and provide such high quality. Look at sensor cleaning: They compromised there as well, and in the end, with the D3s, they had made their homework.

But then, we are not comparing the 24-70/2.8 and a Tamron 17-50/2.8. Wouldn't make much sense, would it? One is a high-price photojournalist tool that doubles as a fine landscape lens (even though it severely lacks stabilization and near focusing for so many other uses), and the other is a very much more versatile lens, that costs about a third and therefore may be excused for being worse at brick walls :)

CC Chang noted the MTF scores of the Nikkor and Tamron: Please note that the results are not comparable. Nikkor was tested on a 12.4Mpix body, and the Tamron on a EOS-50D, what is a 15.1Mpix body. The MTF values in this site should be compared only on a similar body. Interesting is their ratio though, and that is not flattering to the Tamron.
You may see the level of corner deterioration on this preliminary test:
(link)
also see on slrgear how some of the Tamron lenses deteriorate in corners. I am far from ignoring Tamron, as a matter of fact I use the 17-35 f/2.8-4 on my Nikon's for a long time, but that lens shines only on a DX size sensor. Corners on a full body are fatal, slrgear said even, the blurr has blown away they scale beyond any expectations.
This is probably the price for the miniaturization: Tamron really makes often a smaller, lighter product, with a great price advantage. Sadly, they neglect to uphold the standard of the optical properties a tad to much for my taste... I guess "F/2.8" sounds better than "F/4" in the specs, but the 17-35 be better not being used at f2.8 :-) I wish I could disable F/2.8 on this lens, but that's me. I like its low weight and compact size, though. Have you ever tried to carry around the 17-35/f2.8 Nikkor?

What about sample variation? Any lab test that doesn't pull in at least 3 to 5 production samples is utterly useless. And that's for a minimal degree of certainty. Ideally you would test around 50 of these. You have to be able to show out-liers in the extra-sharp range as well as extra-dull range, and demonstrate where the mean lies. This proves not only the lens design, but also the manufacturing precision.

This test, which only features 1 copy of the lens, could just as easily be the lowest quality out-liers that Tamron has produced. You can't know from this test. This lens already has a proven track record based on user response, and that's probably more reliable than a one-off test like this one.

essentially,it comes down to this: if you need corner sharpness but constant aperture isnt important, i.e. for landscapes, get the 16-85 VR.

if you need a fast constant aperture and want stabilization at the expense of losing some corner sharpness, get the 17-50 VC.

if the tamron sells well enough, it's quite possible we will see Nikon introduce a standard 2.8 zoom with VR, as thom hogan has been (not so subtly) hinting for a while now. if nikon can put VR in a plastic 18-55, they can do it in a pro-spec zoom, just maybe not at the size of the 17-50.

Eric,

I think this is wrong. The reason to get the 16-85 VR can only be the extended range.

Yes, it is sharper in the corners at f3.5 than the Tamron at f2.8, so what? Would you take landscape images at f2.8 or f3.5? Most of the time I wouldn't. And even if:

I've just tried the Tamron at f3.5, tried it with book shelfs (detail!), tried it with flash (it's still night here), and I can see a subtle sharpness falloff, only in the extreme corners, and I can only see it because I look for it. Even at f3.5, you would have a hard time seeing it, and for the 16-85 VR this is still wide open.

No, I suppose with the 16-85 you would shoot normally at at least f5.6, and by that the Tamron is stellar across the range. We're speaking of 50/1.8 sharpness here. And that's only at 17mm. Think of 24mm: the 16-85 just begins at f4, from 35mm at f4.5, and by 50mm it is at f5. At none of these focal lengths and at starting aperture it is a match for the Tamron.

Now take it the other way: Imagine a situation where you do want to take a scenic image at f2.8, for instance because it is night. Let it be architecture, for instance in a city, or let it be within a cathedral. It's quite a typical situation, and it's quite typical for situations where you either have no tripod or may not be allowed to use it.

In such situations the shot is frequently repeatable, thus I may go down from my normal 1/15s (auto ISO lower speed limit) to 1/8s or even 1/4s. With VR I have a sharp image, it may take me two or three attempts though, especially standing without support and shooting portrait format. Even in low light I may get away with ISO 200.

With the 17-55/2.8 at twice the price I may be lucky to get the shot at 1/15s, but I suppose 1/30s will be more likely, especially in portrait format. We're talking two to three stops, i.e. ISO 800-1600 here. Do you believe that the added corner sharpness of the 17-55/2.8 will still be there at these ISOs? And if were not talking extreme corners but center or off-center, for instance a typical "rule of thirds" composition? The Tamron will be much better than in the extreme corners. The Nikon may or may not still have a slight edge on the charts, but you would have a hard time seeing it, and, remember, that would be at the same ISO. But what with our fictual but not so unrealistic situation in the church or at night in the city? With an advantage of ISO 200 vs ISO 800-1600 across the frame and the main subject where main subjects typically are, don't you believe that any theoretical sharpness advantage of the Nikon, even if it were there at that point in the frame, would be hopelessly buried in noise?

Now say you accept some added noise because you need depth of field. You go to f5.6. That's two stops, we are at ISO 800 with the Tamron. We may need two or three attempts to hold the shot at 1/4s, but we would need the same with the Nikon at 1/15s or even 1/30s. Both lenses operate at maximum sharpness now. Under ideal light and in the lab, you may still be able to measure a slight sharpness advantage in the extreme corners for the Nikon. Our subject is not in the extreme corners though, and the light is low as it is. Where is the Nikon now? ISO 3200-6400, right? Forget about any theoretical advantage it may have. At that light it is severely hampered by sensor noise. ISO 800 vs ISO 3200-6400? This is an almost too easy win for the Tamron.

I may sound like being biased, I may even sound like being affiliated with Tamron, but that's not the case. I just own this lens and have used it for three months in the darkest time of the year. Really, I wouldn't so easily dismiss this lens :)

I think this is wrong. The reason to get the 16-85 VR can only be the extended range.

Yes, it is sharper in the corners at f3.5 than the Tamron at f2.8, so what? Would you take landscape images at f2.8 or f3.5?

LOL, andreas. you're preaching to the choir. the slow aperture is precisely the reason why the 16-85 doesnt really appeal to me. its specs and lab tests look great on paper, but in real world shooting it's crippled severely by that variable aperture. since you can't really get DoF isolation, especially at longer focal lengths, it's less useful as a portrait lens than 2.8 or faster lenses which cost a lot less. which kind of leaves landscape and walkaround. for landscape, VR is only somewhat useful since if you're really serious you're going to put it on a tripod and stop it down to at least 6.3 and dial in an appropriate shutter speed. and for walkaround, it's better outdoors than for urban street shooting or indoor use, where VR only compensates somewhat for a slow aperture.

No, I suppose with the 16-85 you would shoot normally at at least f5.6, and by that the Tamron is stellar across the range. We're speaking of 50/1.8 sharpness here. And that's only at 17mm. Think of 24mm: the 16-85 just begins at f4, from 35mm at f4.5, and by 50mm it is at f5. At none of these focal lengths and at starting aperture it is a match for the Tamron.

LOL again. after comparing the two lenses (16-85 VR and non-VC 17-50) on photozone, technically, the 16-85 is sharper than the tamron wide open (3.5 vs 2.8) at the widest focal setting, but not by much and only in the center. the two lenses actually have fairly similar characteristics at similar focal lengths and apertures, but surprisingly, the tamron has better corners at 50 mm and f/8!

so you'd have to conclude the only real advantage of the 16-85 for landscape is 1mm on the wide end, the extra reach on the long end, and VR for handholding in low light (mitigated by the VC model tamron, which photozone hasnt yet tested in nikon mount). however, the benefits of 2.8 outweigh those attributes for general photography, IMO. at 50mm the 16-85 has a max aperture of 5.6, which is really pitiful when you consider its cost.

as far as comparing the 17-55 and the 17-50 VC, i only own the non-VC tamron. IMO the 17-55's biggest plus is its build quality, second is its faster focus speed. optically, the tamron is as good or better, especially wide open. the drawbacks to this lens are its weight and its price--as photozone remarked, for that much, you expect awesome optical greatness unsurpassed and unequalled by a lens which costs 1/3 to 1/2 as much (i'm paraphrasing here). i've used my 17-50 since 2007 and its held up better than you'd expect from a plastic lens. and, it can go places where the 17-55 would be too heavy and obtrusive. with the tamron, i can usually get 1/15 handheld with good steadying technique because the lens is light. that's good enough for me.

Just another thing because I always read how light and small the Tamron is:

The new version with stabilization is really a totally different lens. It's bigger (72mm filters vs 67mm, 94mm length vs 83mm), it's heavier (570g vs 430g), has a different near focus distance (0.29m vs 0.27, one cm compensated by being longer though), has a different number of elements/groups (19/14 vs 16/13), and it even looks totally different.

Thus: don't expect a lens that feels small and light You'd be disappointed. Just for the record.

hopefully tamron will continue to make the non-VC model. it's kind of a good deal at $400-$450.