Dear all,
I would like to get a wide/mid angle zoom for a full frame camra (nikon d750)
I'm looking to the following models:
- tamron AF 24-70 f/2.8 SP DI USD VC, or
-Sigma 24-105 f/4 DG OS HSM A
My dilemma is that the Tamron has a better image quality in term of sharpness, distortions, aberrations, speed etc..while the Sigma is more flexible due to the larger field of view (about 15% of my pictures has been shot within the 70-100mm range in the last 2 year).
Which lens do you suggest? Is there so big difference in term of perceived image quality to renounce to the higher flexibility provided by the Sigma lens?
Could you suggest any other valid alternative in the range of 24-70+ mm? (The nikon 24-70mm f/2.8 seems to be too heavy and expensive and provide similar performance, for such reason I'm not considering it)
Thank you in advance for any suggestion you may give me
Andrea
Hello and welcome Apedra,
at first I'd like to update your knowledge about field of view. I think, that's necessary since no one of these lenses has a bigger field of view (all start with 24 mm  ) but one has a larger range of focal length. I was curious how big the weight difference is:
See the "specs" on DxOMark (camera type I've set to D610 since that uses the same MP count as yours and the lenses only become slightly better on the standard D800)
825 - 900 - 885 grams (Tamron - Nikon - Sigma). I think 60...75 grams are no huge difference. Of those lenses, I own the Sigma and came from a different direction: I wanted a better one than the Nikon 24-85 which shows massive distortion. I didn't want the "best" zoom in that range, I already have 5 primes covering it but sometimes I want to have only one lens with me. For that purpose, the Sigma is great and the OS (= VC) is simply the best I've tried so far (better than Nikkor 70-200/4 at 105mm).
Downsides of the Sigma are (in my eyes)
- Massive weight (but somehow nomal in that range?)
- Nikon bodies put it on the same Focus-Adjustment bench as Micro Nikkor 105/2.8. Using both, I have to decide which correction value to use. Somehow stupid. Nikon's not interested in changing that, Sigma is not able to do so.
- filter thread is a bit exotic with M 82 (but that goes for Tamron as well).
Besides, not each lens tester consider the Tamron the better lens. At LenScore, the Sigma does look better. I was checking that site because of the minimal focus range:
0.38 - 0.38* - 0.45 grams (Tamron - Nikon - Sigma).
*Nikon says "at focal length 35 … 50 mm", which makes me a little bit suspicious. Anyway, with the Sigma I get a slightly bigger scale at close-ups.
In general, I can't see a heavy weight advantage of Sigma over Tamron. In particular, however, I do see an advantage: Using the Sigma USB-dock and investing some hours of testing, I can focus adjust the Sigma at 4 different focal lengths and 4 different distances. I know people saying a lens should be adjusted right out of the box, my experience about that is "wishful thinking" but good, the Sigma allows me to fulfill my wish 
Now, if you're in the situation your new 24-xx has to become your work-horse and you won't get better primes, I'd prefer recommending you the Tamron. Unfortunately, I can't: I made some bad experiences with Tamron service and a 17-50/2.8 lens. They might have become better in terms of service, but I'm not willing to find that out on my costs. Tamron needs to make a real fascinating lens to get me back on their boat, currently I look first on Sigma's portfolio, then on Nikon's and after that all the others.
Joju, thank you for the quick and very accurate reply and for the links provided.
You are right I used a wrong wording the two lens have the same FOV what is different is the focal length range.
I agree with most of your considerations but what is confusing me is the different results given by different testers. For example, optical characteristics such as distorsion, vignetting, resolution should be resonably the same for all, while photozone provide much better values for tamron (for this specific characteristics) while lenscore considers better the same optical characteristics of sigma.
In addition, regarding the photozone resolutions diagrams, please also consider that they use different scale for the two lens (i.e. from 1300 to 4000 lw/ph for the tamron and from 1200 to 3700 lw/ph for the sigma), If you put the numerical value in an excel sheet and compare the performance of the two lens in a single diagram using the same scale the difference is more evident.
Looking at the numbers measured by photozone do you believe that our eyes can normally perceive such difference?
Is there someone who has tested both lens and can provide his impression?
Thank you again
LenScore uses a 200MP SuperDuper lab-only sensor. For EVERY lens the same sensor. While PZ uses for each lens a normal body which in most cases are not the latest fashion in camera-technology. Their D3x from 2008/2009 still has the same MP as your D750 - but they haven't tested the Sigma 24-105/4 A on that body.
That's also the reason why DxO gets different results for the same lens (if it's a 3rd party product from Sigma, Tamron, Tokina etc.) The Tamron gets less points on a D610 than on a D810 - not to mention the even lower resolving Canon sensors. But while LenScore gives the best possible comparison (imo), that doesn't matter much if the AF module of the body and the lens' AF drive don't fit well together.
Of course, depending on your way of shooting, a weak AF system might not be a huge issue - but people who go for a zoom lens try to save time: No lens change, maybe no tripod, maybe phase detection AF only or most of there time - all that can make the optical second best your personal best purpose lens.
Also, an important part but not tested is reliability of the AF: How many shots out of 100 are sharp? In which conditions? Testing labs use a best shot method (focussing manually, with light variations and get the best resolving out of it). In daily use, things can be different. For me, reliable and well-performing shake reduction is a benefit - I've not always a tripod with me. Some samples of that you can find in this test shot gallery.
And we didn't talk about sample variation yet. Lensrentals.com made a couple of resolution (only) tests, just to see how big the difference between the worst and best lens of a lot (same manufacturer and type) is. Sometimes tiny sometimes you'll get the impression the manufacturers use some kind of random production 
How long will the AF work reliably? What about cold weather? How important is weather resistance to you? From the whole value of a lens, the optical bits are a major part, but what good is the best optics if the AF misses half the time? Or there's no AF and you have to rely on your focussing skills? Also, for Nikon owners, the Sigma zoom and focussing rings do the opposite what I'm used from Nikon zooms. So, if you have only one zoom, no problem, if you have a mixed collection, it can be a little bit confusing.
My advice would be: See if a dealer would borrow you the lens in question or you can rent it and get your own pictures. For my personal use the Sigma appears to be the best because my standard lens is either 35 or 50 when I walk around with two or three lenses with me (mostly primes). For travel the Sigma is pretty okay and performance wise on the level of the Nikon 24-120/4, if not better. (Another lens to consider, sorry)
re: lenscore - it's a good concept but a bit of flawed execution. There are optics in between the sensor and camera - wwhat is their purpose? Are they diffraction limited? It sounds like the optics are a focal reducer / beam expander / other image circle modifying component. Unles these optics are ultra fast and ultra well corrected (an impossible feat...) they degrade the performance of the lens in front of them. Depending what aberrations are present in those optics different lenses will suffer (or be apparently improved) on a case by case basis. For example, if those optics have overcorrected spherical aberration and a lens has undercorrected spherical, the sum is closer to perfect correction of spherical... Or if they have coma, that can actually improve off-axis resolution when a lens has imperfectly corrected spherical aberration...
The test also puts lenses outside design spec. What is the coverglass thickness on the sensor? Is there a coverglass? This will affect center performance some, and edge performance massively.
They also fail to disclose what they are doing. Are they imaging a test chart? If they are, its resolution must be extraordinary to avoid limiting the performance of the lens.
It would be easier to measure the lenses the way we do in my lab - through interferometry - than it is to do this. Of course such a test can only happen at infinity, but it also has the benefit of having "no" limit to the resolution of the test.
Well, other "lens test" labs are to me also black boxes. They can explain what they are doing and maybe also how they do it - but after all, what counts is the advantage of a comparison of incomparable parts of systems. Like you mentioned, there are sometimes thinner or thicker glass elements in front of a sensor. Extremely thick on mirrorless µ4/3 system, like I learnt from lensrentals.
LenScore just says
Based on these findings, we've created a lens testing system based on a custom-built camera using a 200MP super high resolution CCD sensor, an apochromatic high-precision lens group custom-made by one of the world's best industrial optics manufacturers and 5 exchangeable lens mount adapters. The sensor of our measuring camera is very slow and not designed for low light applications, but these restrictions are of no significance when measuring the optical performance of a lens.
and
The most expensive part of our test bench is not the 200MP digital back, it's the custom-made apochromatic lens group. We neither have the expertise to calculate such a lens group nor are we able to manufacture one. Both tasks have been commissioned to a highly regarded Swiss industrial optics manufacturer. They did a very good job, but they are very expensive. To measure APS-C lenses, or lenses for any other sensor format, we'll need a corresponding lens group. Another investment of this magnitude is currently not an option.
To me, it's the same device for all lenses, hence the same flaws and only for full frame lenses. Question is, what are the flaws and do those matter much? Will an excellent lens become crap and a kit lens shines unexpectedly? DxO Mark is testing different cameras (also very time consuming) and some of us do criticize them for giving too good (or bad) results. Using interferometry does say, how good a lens is on interferometry - and each aspect more is some kind of conclusion which might be right or wrong for our purposes. Like I said, a lens is a bit more than only test results. And I'm not talking of a "creamy rendering of colors" or whatever voodoobabble the lens poetry of some, mostly German, producers is trying to create. <_<
Optically a lens can be reduced to test results though, just not ones used by photographers. Interferometry is the only method I know of that reduces a lens to pure aberrations in a way that mimics design as well. In design you see a transverse ray plot. The focusing error of the ray is aberration. Different aberrations have different focusing errors throughout the pupil. Here is spherical aberration:
[sharedmedia=gallery:images:1315]
An interferometer will produce an interferogram - the fringes are wavefront error in a number of waves (multiple of ~630nm). Like a RIM plot (transverse ray plot), different aberrations manifest in different interferograms. Here is a series of interferograms across different field positions:
http://i.imgur.com/DEIaigC.png
What you see here is field constant coma and asymmetrical linear astigmatism. These are effects you won't see outside freeform optics (my area of research).
The limits of an interferometric test is that you can't evaluate polychromatic aberrations this way, and the resutls are only for the wavelength of the laser. Typically interferometers contain an HeNe laser, but they can be made with other lasers as well. The machine can also only be run "at infinity", otherwise everything comes out wrong.
Other than that, interferometry is the method we use to evaluate produced optics for any factor. Different methods of interferometry can test wavefront error (aberrations), surface roughness (finer than any other method of physical measurement, might I add - SEMs barely scratch 1/1000th the precision), crystal structure, and perhaps other things I do not know about. Interferometry, done correctly, has *no* confounding variables. All you need is a reference optic with the same radius of curvature as the lens, which is simple enough to produce given the focal length of the lens, which can be physically measured to exacting precision.
Of course mechanical factors cannot be evaluated this way, but all optical factors can be analyzed independently through interferometry. You can even compute the coefficients of the aberrations in a merit function for the real lens through it, or the coefficients, normalization radius, etc, for aspheres.
My beef with lenscore's test is that any aberrations in their mysterious optic between the lens and sensor confound the results unless the optic is diffraction limited at an aperture faster than the lens being tested.
Quote:Optically a lens can be reduced to test results though, just not ones used by photographers. Interferometry is the only method I know of that reduces a lens to pure aberrations in a way that mimics design as well. In design you see a transverse ray plot. The focusing error of the ray is aberration. Different aberrations have different focusing errors throughout the pupil. Here is spherical aberration:
...
You can even compute the coefficients of the aberrations in a merit function for the real lens through it, or the coefficients, normalization radius, etc, for aspheres.
My beef with lenscore's test is that any aberrations in their mysterious optic between the lens and sensor confound the results unless the optic is diffraction limited at an aperture faster than the lens being tested.
One can compute an already computed lens, but what does that say about it's manufacturing quality? How does the analysis program gets it's formulas and coefficients? If not by originally long rows of experiments? Why don't you doubt them not a bit, too?
I have no idea nor experience with interferometry and the same goes for LenScore's testing method. I also can read the theory of an optical bench (also not to your taste, I guess  ) but I haven't seen one in the flesh nor tried to work with one, beware.
I guess (!) they use a big sensor and enlarge the picture given by the lens, sort of the way its' done by speed boosters on µ4/3 - and there the speed booster doesn't take away image quality. Of course you can doubt this optical adapter as you can doubt anything else. I just ask what's the point of spending lots of money in it and go public with the results? Only confusing amateurs or provoking discussion in their forums?
You realize, Scythels, we're moving very far away from the original question?
By searching for a link of how this speed boosting works (and I still guess, LenScore uses kind of the same device, just for FF lenses), I found also one about a comparison between the Sigma 24-105 and the Canon 24-105, together with a chart about both's sample variation: http://www.lensrentals.com/blog/2014/03/...comparison
The link for speedboosting by metabone magic from Roger Cigalla. And I see my memories were completely false, actually speed boosting is the opposite of what I think happening at LenScore. Anyway, as long as they use there "resolution and more" checking machine the same way on each lens (and they do it with kind of automated process) there are at least comparable results.
Thank you for your kind and competent comments.
As far as I can understand (sorry I'm not an expert in this field) I realized that each tester has its own methodology which may provide big difference in the results ( I was convinced there was a single methodolgy recognised by International authories (ISO, etc..).
Scythel, considering the methodology adopted in your laboratory based on interferometry (if I well understood it is the same methodology adopted in the photozone tests) how do you judge the three lenses (tamron 24-70 f2.8, sigma 24-105 f4 and nikon 24-120) in term of : sharpness, distortion, chromatics aberrations?
Could you also specify if the numerical differences raised from the test results are easily perceptible to the eye?
What I'm trying to understand if it is preferable to renounce a minimum of quality (perhaps highlighted by numerical tests but then not so much important to view) at the expense of greater flexibility
I'm also looking forward to Scythel's answer. Especially the bit with PZ's testing method. Not a single tester considers all factors, in my opinion, but a lot of them are reliable and excellent in delivering puzzle parts for the "bigger view". Unfortunately, they're not testing all the same lenses at around the same moment...
You're aware, Apedra, that with zooms there's no free lunch: High resolution in tele, not much distortions in wide angle and very small CA over a 4× range of focal length, as fast as possible and as light as possible - each of all those wishes is excluding others in some way. And it has to remain affordable...
|
