35mm f/1.4 Shootout: Canon 35mm f/1.4L II VS Sigma 35mm Art VS Canon 35mm f/1.4L I

Posted by

Yes, indeed, I learned my lesson. I’ve had MTFs on a few copies of the Canon 35mm f/1.4 Mk II lens for a while now, but I wasn’t about to publish the results until I had a full 10-sample group done to get a thorough look at them.

If you don’t like pretty graphs, then I’ll give you the quick summary first: the new lens is significantly better than the original 35mm f/1.4, particularly in the edges and the corners. The difference is pretty significant and should be apparent in most photographs. That, of course, only tells you about the MTF. Others are already reviewing the lens and will comment on how it handles and focuses, what the bokeh and vignetting are like and whether that Blue Spectrum Refractive Optics makes all of the lateral color aberrations go away.

As an aside, for those of you who think Blue Spectrum Refractive Optics sounds like Canon has joined some of the other manufacturers in throwing out some marketing hyperbole, I agree with you. I’m not saying it doesn’t work, I’m just saying throwing magic words at me without telling me what it is and how it works, well, color me Cynical Blue. They do give some hints about what Blue Spectrum Optics are in a few places. If I was to guess, and I’m just guessing, it’s a new glass glue that refracts blue light and fills the space in a cemented group. I suspect it works really well. I suspect it would have worked just as well without the marketing department giving it a silly name and chanting spells over it.

But enough of my negativity. Let’s look at some positive results, beginning with MTF curves. I’ll put the Canon 35mm f/1.4 Mk II up against the original Canon 35mm f/1.4, the Zeiss 35mm f/1.4,  and the current 35mm focal length resolution champion, the Sigma 35mm f/1.4 Art.

 MTF Charts

As (almost) always, the MTF results we present are the average of 10 copies, each copy tested at 4 different rotations.

Roger Cicala and Aaron Closz, Olaf Optical Testing, 2015


The new Canon lens is clearly and definitely better than the very good lens it replaces, both on and off axis. It is also better, from an MTF standpoint, than the Zeiss 35mm f/1.4. When we look at the average MTF curves, it’s a tiny bit better, even, than the Sigma 35mm f/1.4 Art lens, although the difference is small. The two lenses are very close with the Canon being slightly better off axis (14 to 20mm) at 10 and 20 line pairs frequency. Could you tell the difference in a photograph? I doubt it, they’d be nearly identical. Other things like vignetting, focus accuracy, out of focus highlights, cost, etc., would be much more important to you than this slight MTF difference.

Copy Variation

Wide-angle, wide-aperture lenses tend to have some sample variation. That makes a difference, especially when we compare reviews of these lenses. The new Canon 35mm f/1.4 Mk II does pretty well compared to other wide angle lenses.

Roger Cicala and Aaron Closz, Olaf Optical Testing, 2015


Remember, the consistency number is a very blunt tool. Look at the graphs, too, they give you much more information. The one thing I think stands out about the graphs with all of these lenses is we see some variation all the way to the “0” point of the charts. That means there’s some variation in center sharpness, not just off-axis variation caused by a weak corner. But overall, both of the Canon lenses, as well as the Zeiss lens, have good consistency. The Sigma 35mm isn’t quite as good, but still acceptable. I’m still puzzled about why that is — the other Sigma Art series lenses tend to be extremely consistent.

My only real message here is that particularly when comparing MTF between the Canon 35mm f/1.4 Mk II to the Sigma 35mm f/1.4 Art, trying to split hairs and finding a difference between them is pretty silly. There’s probably as much difference between two copies of the Canon, or two copies of the Sigma, as there is between the Canon and Sigma overall. Either of those, from purely and MTF standpoint, is certainly better than the original Canon 35mm f/1.4 and even a bit better than the Zeiss 35mm f/1.4.

Again, that’s from an MTF standpoint. There’s a lot more to a lens than its MTF, but this is the one bit of information I can give that isn’t readily available elsewhere.

Everyone always asks me if I would buy this lens. Honestly, I’m not sure that I would just on the basis of the MTF charts. It’s the best 35mm by just a whisker over the Sigma 35mm f/1.4, but at nearly twice the price. As more reviewers weigh in there may well show other things that make it worth the price difference. A lot of people only consider Canon lenses, though, and I expect many of them will be upgrading from the Canon 35mm f/1.4 for the increased resolution the new lens gives them.


Roger Cicala and Aaron Closz


September, 2015

Author: Roger Cicala

  • nice work. thanks for sharing

  • Doug

    Thanks Roger and Aaron for the report.

    How will Canon’s new BR element perform for astrophotography?

    Canon lenses, such as the 16-36 F2.8, always produced considerable chroma compared to the excellent Nikon 14-24 F2.8.

  • Mike

    @Lynn: I’m not sure if this analogy will help but typing is free so I will try:

    Imagine you are in charge of testing cars. What you are in charge of evaluating is how comfortable the car is along a road course of various objects. The rating scale for comfort is 0 to 1, but since no car is perfect, there is no 1 rating, ever. To make things interesting, you are going to sit in each test car 4 times at 4 different speeds, from reasonably low speed to medium, to medium fast, to high speed. As one might expect, the higher speed, the less the comfort. So while a car might get a .85 at 10mph, it won’t do anywhere near that at 50mph by nature of the test (50mph) being much tougher. So that’s the relationship between the % of contrast (the vertical) axis and the various colors of the lines on the graph. A 10 lp/mm test is “easier” (lower speed for the test car) than a 30 lp/mm test and so forth. Extending our analogy, if the course had a bunch of distinct sections of different road surfaces, with the first section being smooth pavement, the second being dirt road, the third being cobblestones, the fourth being bricks and rocks, you can see that as you increase the difficulty of the road surface, the comfort scale would drop. This corresponds to the distance from center to corners on the lens – easier to get real sharp in the center, much harder in the corners.

    The reason MTF is useful to us is that it gives us this three axis look at performance; we can see how the lens does at varying “speeds” (the increasing lp/mm) instead of just one – and thus we can somewhat gain an idea of whether the lens has good contrast and decent fine detail ability or (better) very good contrast and very good fine detail ability, AND we can see this across the frame (the different types of road surface). No lens is perfect, and some lens types (fast wide angles in particular) are much more difficult to design than others, so the lens designer has to make choices as to how they wish to trade one area of performance for another; it’s more about this set of choices as opposed to one lens being outright better, which of course isn’t what the brand fanboys want to hear. MTF graphs can give us a view into these choices. As an example, in the 35mm Canon vs 35mm Sigma charts, we can surmise (or perhaps the better word is place an educated guess) that the Sigma designer decided to put a little more emphasis on central area performance and a little less in corner performance a little bit compared to the Canon designer. You can see this by looking at the vertical scale differences at the varying resolutions at the center, mid frame, and corners. No one approach is necessarily better and it might depend on what you (the photographer) values more. And in the case of the two lenses, as Roger has said, in real life terms, the difference is quite subtle to essentially meaningless. (Based upon MTF, these two are quite good)

    The point I’m making though is that there is no way you can get this information out of a DPR widget, DXO rating, Lenstip chart, etc, as all of those are only telling you two parts of the puzzle – in our car analogy, they’re not giving you the various speeds, just one speed, so it’s a two axis test. Not as useful, although of course, easier to read. And a single number score is even worse, because we don’t even know what it is based upon, and thus, could be misleading. Lens design (and performance) isn’t simple enough to be able to be graded like a clasroom quiz.

  • > a very, very, very blunt instrument (just a single ‘figure of merit’ number)

    OK, makes sense (and thanks for the reply).

    I do wonder if the MTF values on the vertical axis could be read as a lp/mm or lp/ph value. For example, does 0.8 correspond to some resolution figure, like 1000? If so, would that be consistent across lenses, since no sensor is involved with the OLAF optical bench?

    Or not?

  • AndrewZ

    So regarding the blue Spectrum Optics imaging resources article on the canon expo seems to imply that it refracts blue light more than green or red (I would like to know how). Theoretically it can be used with other crown and flint glass to line up the focus depth of all three primary wavelengths. An mtf is not really going to show this improvement (except for lateral which is seen as less astigmatism because sagittal lines don’t srop as much) the improvement will be a lack of longitudal abberations (namely those horrible greeny colours in the background and purple ones in the foreground out of focus areas. That has me far more excited as that horrible green shift ruins more photos that lack of resolution ever would. Surprisingly though people don’t seem to notice it that much.

  • Roger,

    I’ve been impressed by the widget that DPR uses to present lens evaluations … such as:

    My impression is that LR and DPR are already working together in some areas. Perhaps something like an adaptation of the DPR lens evaluation widget could be jointly implemented.

    Or not?

  • Ben

    It is a substantial upgrade from MK1, but if all you do is sharing photo online when the outputs are heavily downsized, compressed, or went through heavy post processing then yea you probably can’t tell the worst MK2 to the best MK1, same thing can be said about all other MK2 lens too (24-70, 70-200, 24, 85,etc).

    My new 35L II is sharper from center to corner than my 100L IS Macro, 70-200 2.8 IS II, and 85L II. Infact, the 35L II @f/1.4 is sharper than the 85L II @f/2.0! Bokeh quality and Color reproduction looks similar to 85L II, which are not going to show on MTF graph.

  • adrian

    so, the conclusion would be that canon made a great lens… but if you take the worst mkII and compare it with the best mkI, you will most notice little to no difference at all…

  • I guess I’m more of a photographer than a gear geek, because for the life of me, I have no idea what “MTF” stands for, and unless I missed it, the article just assumes we all know. Here’s my non-scientific take: I’m a Nikon shooter, and I’ve owned the Sigma 35 and 50 Art lenses, and yes, they’re sharp, but I got rid of them and am sticking with the Nikon 35 and 58 f/1.4 lenses because, even though they may not be as good as the Sigma lenses when it comes to graphs, I just think my pictures look better with the Nikon lenses, especially as far as the out of focus areas of the photograph are rendered. Granted, this is all subjective and some people may prefer the look of the Sigma lenses, which I get, but for me, I’ll stick with Nikon.

  • Wayne S

    Thanks for all your work here for the 35mm shootout.
    Interesting, looks like the new Canon lens will be sharp and we will see about CA and bokeh. We will see how the new WA Otus will compare against these lenses although it will probably be wider than 35mm.

  • Ben

    Here’s quite a lot of test shots from the new 35L II, bokeh test with busy background and harsh lighting, flare test, corner sharpness, etc.


    CA and green/purple fringing is exceptionally well controlled. AF consistency and accuracy has been 100% so far. The image quality improvement over the MK1 is very obvious and substantial, same big leap as the 70-200 2.8 IS II.

  • Marc M

    I’ll be hugely interested in seeing how the new Tamron 35mm f/1.8 VC SP compares. And throw in the new 45mm with the 50mm lot, as well. Thus far there have been very good things being said but I’d love to see the charted differences as Tamron could easily have nailed, as Sigma did that great balance of quality and price point. The differences between light gathering and bokeh, when done right, are just not enough to convince me of the “advantages” of an f/1.4 lens over and f/1.8, especially if transmission on the 1.8 is excellent. Hope you will do some further testing, sir! And thank you, much, for always bringing that proper nerd rationale to the conversation. Working in the retail side of the industry as I do… We appreciate it a great deal, Roger. Thanks.

  • Alan

    Note the result will include the latest Tamron 35mm F1.8 VC

  • Brandon

    sorry, that should have been “the index is always nonlinear”

  • Brandon

    The index is always linear. If you take the index at the F,d, and C fraunhoffer lines you have three “samples” so to speak. You then define dispersion as Vd = (Nd – 1) / (NF – NC) for F=red, C=blue, d=yellow.

    You can also take the partial dispersion, Px,y which is just (Nx – Ny) / (NF – NC).

    If you plot different glasses you basically get a linear correlation between Vd and Px,y. Anomalous partial literally just means it departs from that relationship. It is a very desirable quality to have, as usually the color correction in the blue side of the spectrum is quite difficult.


  • CarVac


    When they say “anomalous partial dispersion” does that just mean “the refractive index is very nonlinear with wavelength”?

  • Brandon


    No optical grade plastic exhibits anomalous partial dispersion. BR is not a “plastic optic” so to speak.


  • Brian

    The “organic optical material” used to make “Blue Spectrum Refractive Optics” is presumably what is more commonly called plastic. It’s an amusing addition to all the other terms used to mean plastic, such as resin.

  • Someone

    Roger & Aaron, Thanks for my daily dose of humor. I’m now a more healthy person because of you two, and without any hocus pocus 😉 (or the treadmill). Keep up the good work!

  • Rob

    Roger it would be interesting to see howe they handle strong backlighting,I think the Canon will drop in price to a more reasonable street price and the weather sealing and lack of any firmware issues will make it very attractive to a lot of people.

  • Roger Cicala

    Lynn, a very, very, very blunt instrument. I understand that people want a single number to make things easy, but doing that basically eliminates all the measurements and brings it back to subjective.

    For example, let’s say I refuse to give a general number but just give you a sharpness score. What if you’re more interested in edge sharpness rather than center sharpness? But some lenses have good edge and great center sharpness, but are really pretty soft in the middle area. What if a lens has a lot of astigmatism? I could give it a very sharp rating based on the higher number of sagittal and tangential, a lower rating by using the worse of the two. I could average them and make it seem average, but a lens with no astigmatism and an average MTF looks very different than that other lens that has the same number but lots of astigmatism.

    Anyway, my own impression is there are plenty of places that will give you a single number. There aren’t any others that will give you the MTF data or multiple copy data. This is for the people who want more.

    I would, really, encourage you though to read about MTF. It’s not that hard to get a handle on and the amount of information you can take in at a glance is staggering: is a lens more contrasty, or does it resolve better, and in which parts. Even an idea of how the bokeh might look.


  • Roger Cicala


    It is odd, but it’s also real. I’m going to write a pretty long, technical article fairly soon about the myth that stopping down improves everything. It doesn’t always, at least for a stop or two. Some things, like field curvature, don’t improve at all. In a few situations with certain lenses even though things improve overall, certain measurements at certain locations actually worsen a bit. The good news is it’s more about measurements, in real life you’ll never (I don’t think) make an image worse by stopping down.

  • CarVac

    @Lynn Allan:

    You’re probably not the only one, but these charts are really for the gearheads like me who thrive on data: the more data the better.

    If the black and red lines are low, the lens is hazy. If the cyan and purple lines are low, the lens is soft. The green line is in the middle. If the dashed and solid lines diverge a lot, the lens is either smeary or has lateral CA.

    They don’t translate directly to resolution, but resolution is a very vague number; you can have a hazy high-resolution lens, or a contrasty low-resolution lens, each better for different purposes.

    That’s why they’re not boiled down to a single number here, and they really shouldn’t. There’s a lot more nuance to it that you can learn.

    For example, on Photozone, corner resolution is just a number. Is that tangential? Saggital? Somewhere between? Does it include CA? Is CA corrected? I don’t know, and even if I did know and they did it in a way that I disagree with then I can’t use the information anyway.

    If you want to know only one thing about these, look at the green lines and see how high they are. Higher is better, and the closer they are together, the better.

  • I can appreciate the difficulty in presenting the vast amount of data that OLAF measurements generate. However, I find it difficult to evaluate and compare lenses with the MTF charts. Am I the only one who mainly sees a bunch of squiggly lines?

    Would it be possible to provide overlapping images, like Photoshop layers to more directly compare lenses?

    Some numbers might help, like what photozone.de provides. The “Copy Variation” is a single number I find meaningful.

    Does a figure of, for example, 0.9 MTF-50 translate to a resolution value?

    I just encountered LenScore.org, and they boil down their measurements to a single number. Too much of a “blunt instrument”?

  • Tim

    There’s something odd about the Sigma 35mm Art measurements. Going from f1.4 to f2.0 the tangential MTF gets considerably worse, when you’d expect it to improve:


    And this effect isn’t obvious in the test chart images:


  • MayaTlab

    Oh well that might be coming sooner than I would have expected it then !

  • Just got mine today and I’m extremely happy. I haven’t used the sigma or the zeiss but this certainly improves on the original Canon in every way. Samples posted on CanonRumors.

  • Roger Cicala

    Maya, we have the mount made and it’s functioned well in several trials, but we’ve been too backed up to get any real data runs yet. I’m hopeful we’ll get some numbers next week.

  • MayaTlab

    Thanks a lot.

    I believe the Sigma’s variance chart is a little different from the one released in the 35mm variance article and in The-Digital-Picture comparison tool. Is it because you added more copies to the test ? Will it be possible in the future to update TDP’s comparison tool as you include more and more lenses in your variance charts or would that represent too much of a burden ?

    Also, do you have any update on your attempts to test Sony FE lenses with a special mount ?

  • CarVac

    You showed the Zeiss F/2 version in the picture at the top…

    Anyway, an impressive showing by Canon.

Follow on Feedly