Equipment

A Quick Look at the New Sigma 24-35 f/2 Art

Published July 29, 2015

: Image Courtesy Sigma USA

I generally do MTF testing on multiple copies of a given lens so that I can present the averages, look at sample variation, and stuff like that. I’m always aware that looking at any single copy of a lens, especially a zoom lens, gives only a limited prediction of how other copies might look. But sometimes a limited prediction is still pretty exciting. As luck would have it, I was able to get a single copy of the Sigma 24-35mm f/2.0 DG HSM Art lens for testing today. It will be a week before we have enough copies to begin doing out multiple copy tests, so I thought I’d go ahead and post the results of the one copy.

Sigma is releasing this lens as something a bit different in the standard zoom arena. The 24-35mm f/2 is, as you might have noticed, an f/2.0 ZOOM for full-frame sensors. Hot damn, I’m excited about that, and after seeing what the Sigma 18-35mm f/1.8 performs like on a crop-sensor camera I’m pretty excited to test the performance of Sigma’s new full-frame lens. The Sigma has 18 elements in 13 groups with one low dispersion and 7 ultra-low dispersion elements. It is a very short zoom range, roughly half that of a 24-70mm lens, but is a full stop faster. But it is being marketed more as an alternative to prime lenses in this range, rather than simply a competitor to 24-70 f/2.8 zoom lenses.

MTF Curves

24mm Results

Since the Sigma is being considered, as an alternative to carrying several wide-aperture prime lenses, I thought we’d compare it to wide-aperture prime lenses. I’ll start by comparing it at 24mm to the two current Canon prime 24mm lenses.

: Roger Cicala and Brandon Dube, Lensrentals.com, 2015

The Sigma zoom is pretty impressive, particularly in the center where it actually has better resolution than the Canon 24mm f/2.8 IS lens, despite the Sigma’s stop wider aperture. It’s clearly better than the Canon 24mm f/1.4, but of course the Canon is being tested a full stop wider than the Sigma zoom. Off axis, though, the Sigma doesn’t do quite so well, with a lot of astigmatism in the outer 2/3 of the field. Towards the edges both of the Canon primes are doing better than the Sigma. But still, that’s pretty impressive for a zoom. Primes are expected to be better in the outer 1/3 of the field.

To make a bit more practical comparison, we compared the Sigma at 24mm with the Canon 24mm f/1.4 lens, this time shot at f/2.0 so the apertures were equal.

: Roger Cicala and Brandon Dube, Lensrentals.com, 2015

Stopped down to f/2.0, the Canon improves dramatically. It is now equal to the Sigma in the center and does better off center. But again, we’re comparing a $1,600 prime to a $1,000 zoom. The Sigma isn’t quite as good, but it’s certainly staying in the neighborhood, and that’s a pretty reasonable accomplishment, especially for a zoom lens.

28mm Results

Not too many people shoot a 28mm prime, but there are a couple of good comparisons, the Canon 28mm f/2.8 IS and Canon 28mm f/1.8 lenses.

The results are similar to those at 24mm. The Sigma is clearly better than the f/1.8 prime. Despite being a full stop of aperture faster, the Sigma zoom is nearly as good on-axis as the 28mm IS lens and not quite as good off-axis.

35mm Results

At 35mm we can make similar comparisons to what we did at 24mm, this time using the Canon 35mm f/2.0 IS and Canon 35mm f/1.4 L lens.

: Roger Cicala and Brandon Dube, Lensrentals.com, 2015

As at the other focal lengths, the Sigma is superb in the center, but has more astigmatism away from center. We can also directly compare the Sigma with the 35mm f/1.4 tested at f/2.0 to even up the apertures.

: Roger Cicala and Brandon Dube, Lensrentals.com, 2015

Again, stopped down to f/2.0, the Canon 35mm is a bit better than the Sigma zoom. Some of you are going to note the really interesting tangential curves on the Sigma lens at 35mm. We spent some time looking at this effect and it’s a wicked tangential field curvature. If we did our MTF tests (as some do) by refocusing at each measured spot, that would be much less impressive. We see a similar MTF with a number of wide-aperture prime lenses, although not quite this impressively.

I’ll put up one last image which just repeats the Sigma results at 24, 28, and 35mm since we know most zooms have a weaker and stronger end.

: Roger Cicala and Brandon Dube, Lensrentals.com, 2015

In the case of the Sigma 24-35mm f/2.0 Art lens, things aren’t too dramatic. Center resolution is a bit better at 24mm but that’s about it. The lens is very close to equal at all zoom distances.

Summary

Some people are going to be a bit disappointed (I admit I am a bit myself) because they really wanted a zoom that was every bit as good as their prime lenses. What we actually have is a zoom that’s as good at f/2.0 in the center as their prime lenses, but like almost every other zoom isn’t quite as good as a prime off-axis. When even the best optical engineers (and lately Sigma seems to have some of the best optical engineers) design a wide-aperture zoom lens, some compromises have to be made. In this case there’s a lot of astigmatism away from center. There’s good news for you who are disappointed now: Computer image analysis (like Imatest and DxO) don’t show astigmatism directly so this lens will probably look better with those test results than it does on the optical bench.

But the glass here is definitely way more than half full. We have, for the first time, a modern zoom lens that reaches f/2.0 aperture and is wickedly sharp. That’s an amazing thing all by itself and this is a unique optic. What is really amazing to me is that it sells for $1,000. That’s way less than either of the f/1.4 primes we tested against it, less than two of the f/2.8 primes, and less than an f/2.8 24-70mm zoom (granted that zoom does go all the way out to 70mm).

Will everybody be interested in this lens? Nope. A lot of prime shooters want f/1.4 as an option. A lot of zoom shooters are happy with f/2.8 and a greater zoom range. And being interested in the lens doesn’t mean buying the lens. We’ll have to see how it performs in the field.

My first thought is given the off axis astigmatism (which Brandon described in fairly poetic terms of 5th and 7th order aberrations that I didn’t follow well enough to try to translate for you) people are either going to love or hate the bokeh. That will be more important to a lot of people than the absolute resolution. So will autofocus performance, flare resistance, and a lot of things beside absolute resolution.

But whether you love the lens, hate the lens, or don’t care about the lens, I think we all should give Sigma a thank you for having the guts to create the lens. The bar has been changed. An f/2.8 zoom lens isn’t the fastest zoom anymore. And all of us can appreciate the ripple effects of a game changing lens that’s priced as aggressively as this one is.

Roger Cicala and Brandon Dube

Lensrentals.com

July, 2015

Author: Roger Cicala

I’m Roger and I am the founder of Lensrentals.com. Hailed as one of the optic nerds here, I enjoy shooting collimated light through 30X microscope objectives in my spare time. When I do take real pictures I like using something different: a Medium format, or Pentax K1, or a Sony RX1R.

Posted in Equipment

24-70 f/2.8 Zoom MTF and Variation

Quick MTF of the Tokina 24-70 f/2.8 PRO Fx

John M

Came in here desperately hoping to see a comparison between the Sigma 35 1.4 ART – as I own one and am switching from Canon to Nikon, wondering if I could offload my 35 for the 24-35 zoom and forego the purchase of an ultra-wide (I haven’t shot a 24-70 in ages) without a drop in image quality. My impression is that yes, it’ll hurt me image quality wise given the off-axis comparisons to Canon first-party primes. Am I wrong?
Brandon

Jim,

Setting up two zoom and focus directions more than doubles the cost of the zoom and focus cam parts. They have to be cut/cast flipped in one axis for each make, doubling the cost (2x the tooling) and extra QC must be performed to ensure they are not accidentally mistaken for each other.

-Brandon
Jim Thomson

George,
Maybe they know that they sell more to Nikon users than Canon users.
Setting up two different rotation directions may not be as easy as you think.
George S.

@Edward Jenner:

Zoom ring direction has become a totally inconsistent mess at Sigma. Formerly, all their zoom rings turned the wrong/Nikon way – irrespective of lens mount variation. These days, their zoom lenses come with ONE direction irrepective of lens mount- but it might either be the Canon (+ Oly) or the Nikon (+Sony) way.

If possible, it would be great if Roger could possibly take a quick tally since he got so many lenses in house. And if confirmed, I’d love to see him pass the message to Sigma and Tamron (Tokina is better lately) that they are loosing sales to people like me – no matter how good and/or priceworthy their zoom lenses might be.
Blame

That’s sigma all right. Right on the edge between genius and insanity.

Perhaps more insanity this time because the Sigma 24mm f/1.4 wide open should deliver about as much light and more overall resolution to a cropped area of the sensor equivalent to 35mm.

Except in the central area (a circle of about 1/3 picture width in diameter) I can’t see a 36mp or 50mp camera affecting the outcome much. Of course stopped down things might be different but then why not a 24-70/2.8?

So there you have it. The Sigma 24/1.4. Better at 35mm and MUCH better at 24mm.
Edward Jenner

The zoom ring is weird. I was confused by George’s post since I have 3 sigma lenses (none ART) that all turn the ‘right’ way, including the recent 150-600C.

I though it was only Tamron and Nikon lenses that zoomed the ‘other’ way.

So different Sigma lenses’ zoom rings turn in different directions?
Andre Y

Thank you Brandon for the explanation!
George S.

Thanks for the test!

In theory i would be interested in this zoom – mainly for indoor/event use. Attractive combination of very decent optical performance and price. BUT as a Canon user I have zero tolerance for zoom rings turning in the wrong direction. As long as Sigma is not wiling to orient zoom rings aling Canon convention, i will not buy them, ART or not. i will miss shots in fast situations when i change lenses from Canon zoom to Sigma zoom by always zooming the wrong direction first. I want to be able to use my gear totally intuitively, especially when fast moving scenes demand my full visual attention.

It is beyond me, why Sigma refuses to equip even their flagship ART lenses with zoom rings that turn the right way in relation to the lens mount the lens is equipped with. Also, if the can only make them one way, why would theynot choose the standard of the clear market leader wuth the largest installed base of cameras in the market – i.e. Canon.

Same reason why i do not even consuder the very decent tamron 24-70/2.8 – despite attractive performance and price. Third party kens makers should not let such an easy to resolve technical feature limit their sales potential to potential buyers like me. I will not compromise on my gear selection. Lens makers will have to adhere to the user interface of my camera provider/s.
Brandon

Lee,

I chose to use the convention carried out in Code V, the design software I know best.
Lee Saxon

Why’d you guys reverse the usual convention on saggittal lines being solid and tangential lines being dashed? Kinda throws me off.
Roger Cicala

Wally, we had far more Canon lenses done so the comparison data was handy. And that’s for the simple reason that we have more Canon lenses so they were more available to do first. But I shoot too often with a D810 and 14-24 for anyone to call me a Nikon hater 🙂
Wally Kilburg

Why do you hate Nikon? You only seem to publish Canon stats. I guess I shall have to take my rental business to a Nikon friendly company when I rent.
Brandon

Andre,

It is not derived from lens data – just an example of the concept of balancing various orders of aberrations. Each aberration is mathematically represented by a power series expansion – each term is a different order, and the coefficient is the magnitude. For the first short while the first term rules, then it balances the second, then the third rules. The coefficients will generally reduce in size as the order increases.

-Brandon
Andre Y

Thanks Brandon. That plot is also good information. Are those coefficients derived from real lens data? I’m just wondering if they’re indicative of actual lens performance.
Roger Cicala

Lee, I probably was a bit generous verbally — I think overall Sigma has become quite a game changer and I tend to be very supportive of new and different things. Plus, it’s not a lens I’d be interested in (I prefer small, not-quite-f/1.4 primes at these focal lengths) and I think sometimes I try so hard not to let my personal preferences come into play that I bend a bit too far in the other direction.
Lee Saxon

My reading of these MTF charts says your verbal description of the lens is somewhat generous. That is quite a rapid falloff.

I don’t understand Sigma’s design philosophy on this one. A range of focal lengths that narrow means this was never going to be a mass market lens. And what I believe is the fastest FF35 zoom ever was always going to make this an exotic. I think the target audience for such a lens would’ve accepted more size, weight, and cost for better performance.
Brandon

Joe,

Astigmatism is reduced linearly as the aperture is closed down. By comparison, spherical aberration is reduced by x^3 and coma is reduced by x^2.

I will measure this lens stopped down; eventually (nearly) every model will have stopped down data.

Regards,
Brandon
Joe

It appears that all tests of the Sigma are shot at maximum aperture. Were you able to test the Sigma closed down two or three stops to its optimum aperture. Would this affect astigmatism? Most people using an ultra-wide zoom will not be using it at maximum aperture and, if they’re trying to differentiate the foreground, then a longer lens would be more likely.
Brandon

If it is helpful, a plot of the astigmatism vs image height would have a shape something like this:

https://www.wolframalpha.com/input/?i=x^2+-+2x^4+%2B+0.33x^6+%2C+%28x+from+0+to+1%29
Brandon

Andre,

Lateral Color will tank the tangential measurement and “hold it down” so to speak without affecting the sagittal plane. I know the problem is field curvature by watching the target during testing – it grows uniformly in size when it moves out of focus, while lateral color is seen as two distinct images with some separation. In the case of e.g. the Canon 16-35 f4 IS, or the 24mm f1.4L II (in the absolute corner only) lateral color is largely the limiting aberration.

With this lens there is some lateral color but it is relatively minor and *much* smaller than the astigmatism.

A hint that it is astigmatism/field curvature is the shape of the MTF plot – points of inflection in lateral color are extremely rare while they are commonplace for astigmatism or petzval. The comment I made on the aberrations orders is due to the shape of the plot; where it holds mostly flat is the interaction of 3rd and balancing 5th order astigmatism, then 5th dominates and the MTF tanks. After that 7th order balances 5th and the MTF stays flat, and then the image height has grown to where 7th order dominates and the focal plane rockets back towards 0 (and would pass it, if the image extended far enough).

Regards,
Brandon
Roger Cicala

Peter, those graphs are 800 pixels as opposed to our usual 600 pixel limit. But at 600 pixels they just are unreadable. I wanted badly to keep them grouped so we made them larger realizing some people would have to scroll to see the right side. It’s an unfortunate limitation of our blog platform.
Andre Y

Brandon and Roger, thanks again for a timely test of an interesting lens.

You mention that this astigmatism is caused by field curvature, which is the root cause of astigmatism in general from what I’ve read? That is, the sagittal and tangential focusing surfaces are not coincident with each other.

Another cause is lateral CA, which causes divergent S&T curves, but is not astigmatism. Can you derive the cause of the divergent curves directly with your measurements, or do you need to do additional tests?

I’m wondering how many of the lenses in the other tests have divergent S&T curves caused by astigmatism and how many by lateral CA.

For anyone else who’s trying to decode all of this optics talk, I found this site to have a pretty good explanation (along with pictures!) of the various distortions and their causes:

http://toothwalker.org/optics.html
Peter C

Is it just me, or do others also frequently find some graphs cut off on the right hand margin? This is especially noticeable with this posting: all the three abreast graphs have the right hand one truncated.

All of the tests and posts are so informative, I hate to complain!
Jim Thomson

Sigma 24-25mm f/2.0 DG HSM Art lens :). I know it’s a narrow zoom range, but not quite that narrow.

As always your article makes me feel like Oliver Twist.
“Please sir, I want some more.”

That first paragraph of Brandon’s reply to ginsbu sort of made my eyes glaze over. You really do need to get him a fancy lensrentals logo for his posts. I think it should include a warning. How about “lensrentals(logo)geek”?
I’m also a little concerned that I think I actually understood what he said, I may be spending to much time here, it’s bringing back memories of Calculus and BVP. Please don’t ever let him post about imaginary numbers.
ginsbu

Roger & Brandon,

Thanks for your quick and informative replies. I have wondered about their use of ‘horizonal’ and ‘vertical’, but had erroneously assumed that it was just a poor translation from polish. From your replies I gather that measuring horizontal and vertical MTF, rather than sagittal and tangential, is a limitation of Imatest. I wonder how much of the differences in results is due to measuring different planes, versus being due to a possible reduction in tangential field curvature at that closer focus distances at which Imatest is run.

Your point about different reviews is well taken, Roger. I’m certainly not trying to ignite hostilities among lens testers! 😉 This blog has prompted me to learn more about optics and your answers to my question help me to understand why testers might get quite different results.

I’m glad to hear some m4/3 tests are coming as part of this ambitious project!
Brandon

Kenny,

We will be measuring a full aperture series for each model at some point. Ten to fifteen or so models already have complete stop down data taken, wide open thru f/16. The field curvature means it won’t improve to the point of being ruler flat edge-to-edge, but it will be considerably better.
Kenny

This lens does interest me, but from a landscape perspective, and I’d rarely be using it at f/2. Do you guys have plan on testing any set of lenses at smaller apertures, like f/8? I ask this knowing full well that you guys are crazy busy 🙂

I’d imagine a lot of these lenses would perform pretty much exactly the same at f/8?
Brandon

ginsbu,

The vertical and horizontal orientations are an orthogonal coordinate set, but they make little sense for a rotationally symmetric system. In that case, it makes sense to have a radial component (the sagittal plane) and a component perpendicular to that (the tangential plane). Horizontal and vertical are each at about 45 degrees to these planes, so it is impossible to measure “true astigmatism” that way.

I have begun testing with a couple of micro four thirds lenses (Voigtlander 17.5, 42.5, and Panasonic 42.5 f/1.2) but they are carried in much lower volume by LensRentals, which makes testing 10 copies at the same time difficult.

Regards,
Brandon
Roger Cicala

ginsbu,

With all respect, the key here is “understood as the average difference between horizontal and vertical”. I’ve tested several thousand lenses on Imatest and I have noted what they note – but it’s not really astigmatism. Related to astigmatism, probably, but affected by some other factors. Horizontal and vertical aren’t the same as sagittal and tangential.

But one thing I emphasize all the time – taking the gestalt of several reviews is the most important thing for anyone considering a lens. Each of us has different methods and brings some slightly different information to the party. Everyone I know that does this kind of work does it conscientiously and carefully. We aren’t really reviewing, just laboratory testing. The reviewers give a lot more information about things we don’t even look at.

Roger
ginsbu

As it happens, LensTip just posted their review: http://www.lenstip.com/447.1-Lens_review-Sigma_A_24-35_mm_f_2.0_DG_HSM_Introduction.html
They have this to say about astigmatism: “When it comes to the astigmatism, understood as an average difference between horizontal and vertical MTF50 function values, it amounted to 4.6%. Any result lower than 5% we considered to be slight so the Sigma 24-35 mm deals with the correction of that off-axis aberration pretty well.”

The comparison with results here is interesting. I’d be interested to learn whether the difference is due to copy variation, testing method, focus distance, or something else.

Thanks to Roger and Brandon for this series of tests & posts. I primarily use m4/3 so I hope whatever technical challenges there may be to testing those lenses can be overcome. I’d also be interested to see how many of the tested lenses perform stopped down a bit.