New lens looks soft. Help!

[KevinBarrett]

[EDIT] I have removed the images from this post because of the poor quality of my test. The quality of my test as implemented is discussed further in this thread, but it would be irresponsible of me to allow anyone to come to a conclusion about this lens due to these images.

These are crops from the upper right-hand corner of large X-fine jpgs 4272 x 2878. Taken with my a700 and new Tamron 17-50/2.8

As you can see, there wasn't a very great increase in sharpness from 2.8 to 5.6, but at f11 it finally firmed up. I didn't take any between 5.6 and 11 because I thought I was going to be looking for that improvement between 2.8 and 5.6. F4 looks exactly like f2.8.

Should I send this lens back? Anybody else with a Tamron 17-50/2.8 have any input?

[01af]

As you can see, there wasn't a very great increase in sharpness from f/2.8 to f/5.6, but at f/11 it finally firmed up.

This is absolutely typical for the focus being slightly off. The apparent increase in sharpness at f/11 is due to depth-of-field. Re-do your test shots and pay careful attention to proper focus; use an eye-piece magnifier if you have one, or do focus bracketing. Shoot something that has fine detail before and behind the intended plane of focus so you can see where actual focus is falling at.

I returned two A700 bodies to the dealers right after checking them out, due to inaccurate auto-focus. Now I'll wait until after Photokina, then I'll probably purchase a Nikon D300. I am so fed up with inaccurate auto-focus ...

-- Olaf

[Dr. Harout]

... then I'll probably purchase a Nikon D300. I am so fed up with inaccurate auto-focus ...

-- Olaf

I'm quite positive that the same happens with other brands. OK, there might be a software compensation but that doesn't make it right.

[01af]

I'm quite positive that the same happens with other brands.

Sure. But in the Nikon D300, you can at least fine-tune the AF yourself, for up to 12 lenses individually. Furthermore, for some of my assignments I need a faster frame rate than 5 fps. The D300 can do up to 8 fps (with vertical grip) ... and I am afraid also the A900 won't be any faster than the A700. I am definitely going to miss the integrated image stabiliser ... but what the heck is an image stabiliser good for when AF is off most of the time?

Anyway, I'll wait until after Sep 23rd. It would be foolish to spend thousands on new photo gear three weeks prior to Photokina. Who knows---maybe the A900 is faster than anticipated!? However I'd be surprised ...

-- Olaf

[KevinBarrett]

Olaf, I hope you don't make any blanket assessments of Sony based on my amateur lens tests! Looking back through I noticed that the image shot at F11 was done with wide area autofocus with focus-hold while the others were done with spot-focus, then the corner moved into the corner of my frame. I'll redo the test again this afternoon when my battery is all charged up again.

[01af]

Olaf, I hope you don't make any blanket assessments of Sony based on my amateur lens tests!

Of course not! What a funny idea ...

-- Olaf

[KevinBarrett]

David, I know you're online, and I know you're in this thread...I need some help! What would be the best way to test this lens at this point?

[David Kilpatrick]

First, don't focus in the middle then swing the camera round to position the target in the corner, unless you are over 15 feet away. The closer you get, the worse the error will be. Assume the lens has a flat field, or is intended to - your test assumes it has a spherical plane of focus! That means using a brick wall, or a test target. The problem with test targets is that they must be used around 25-30X the focal length. At 17mm, that is just about 45cm - or the closest possible focus for many lenses. At this distance, the field is not flat, so you need to shoot a complex 3-D subject to be able to judge what is really happenning with full aperture.

Lens testing with close, flat field targets is a flawed procedure in terms of showing centre and edge sharpness, and also misreads distortion - that's why the best testing labs use five or more targets fixed to a large wall, often 3m x 2m or so in target area, to reduce the errors caused by using typical A1 to A0 size test charts. This is also why, although I review lenses for magazines, I do not make any resolution tests or shoot targets. I try to use the lens for a week or two on a wide variety of subjects before coming to ANY conclusions - and preferably on more than one camera body.

Olaf, I have done manual adjustment to my A700 AF using the accessible - unofficial - screws and the result is stunningly accurate. Lens choice still has some effect.

This shot on the Tamron 55-200mm at f7.1 was auto-focused using the centre sensor on the A700 with such fast movement it was hard to pick a single moment when the subjects were not tumbling about - and it's nailed the centre focus precisely on the open mouth:



See the original full size http://www.pbase.com/davidkilpatrick/image/102617303

Before I adjusted the Alpha 700, I was getting 'OK' shots but now they are just better all round. Before switching to another brand entirely I would be tempted to save money and try a focus point adjustment. Unlike the D7D, there is no goo on the screw adjusters - nothing to indicate, at all, that any adjustment has been made. So while it might invalidate a warranty there is absolutely no way of telling the label has been lifted and the settings tweaked. It only took a very slight turn - more like 1/12th than 1/6th - to fix the slightly poor focus I had been getting since January. The camera may always have been like that, but I did not notice any errors until January when I shot Cirque du Soleil.

David

[KevinBarrett]

I did as you suggested David and just took it around, using it at usual. I found that even at f/11 my copy still had half as much CA as my 18-70 in similar situations. However, it was fairly sharp toward the corners. I don't know how that stacks up against a CZ16-80, but it doesn't matter because I found that my Tamron had some specks on the front element, and its going back to bhphoto tomorrow.

I may go with a 16-80 next, based on David's reccomendation, as my budget suddenly changed in the past week as well.

[chriznatch]

I found that even at f/11 my copy still had half as much CA as my 18-70 in similar situations.

What does CA mean?

[KevinBarrett]

CA = chromatic abberation

It is a color-splitting phenomenon which increases towards the edges of the frame, most noticable at high contrast interfaces. It shows up as cyan and magenta (or blue and red, to put it more simply) fringing around details.

[chriznatch]

Thanks. Are you saying the Tamron did worse than the 18-70 with CA? I'm curious because I recently bought the Tamron 17-50 myself (mostly for low-light indoor work) but haven't had a chance to use it much yet. My initial impressions are that the lens is a touch softer than the much-maligned 18-200, although I took all the Tamron pictures at a higher ISO (400) than I usually use, and I managed to botch most of the exposures. So the long and short of it is that I haven't been able to form any real conclusions about the Tamron yet.

[KevinBarrett]

Oh no, not at all. What I meant was that whereas the Tamron 17-50's CA occupied a fringe about three or four pixels think in one scene at the very farthest edges, the Sony 18-70's CA occupied a fringe twice as thick, and more pronounced. The color splitting was stronger and the lens softer twoards the edges, combining to make an ugly rainbow smear around high contrast details. Most often, the Sony's CA was visible in an image sized for a desktop, whereas the Tamron had to be viewed at 100% to see the color-splitting. The Tamron 17-50 is a superior lens to the Sony 18-70, but I think my copy was rather off.

[EDIT] I should also add that the Tamron controlled barrel distortion very very well at the wide end, better than the Sony 16-105 as well as the 16-80 that I tried, though these two lenses do have a wider reach. The Tamron had very slight pincushioning at 50mm, though it wasn't really an issue. Surprisingly, I never found myself wanting a longer reach with the Tamron.

[chriznatch]

Good to hear, thanks for the info. Hopefully I'll have more time to experiment with this lens in the coming weeks!

[Vidgamer]

Wide open aperture at 17mm might be a worst-case test. I'll gladly accept that given the amazing performance overall. The lens is great, and it is affordable. You can find better, but you'll pay a lot more. And not necessarily all aspects will still be better. (For example, you may lose the f2.8 with the CZ16-80.) Even at f2.8 it's still pretty sharp, if imperfect. When I look back through photos and really like something, chances are, it was taken with this lens. I now rarely use the 50/1.7. And field curvature is not a problem, unless your subject is flat and you're close and you're wide open -- so maybe this isn't the best lens for photographing paintings at a museum.