Geek Articles

Real World Testing of the New Canon 16-35mm f/2.8L III

Published December 15, 2016

Tags: Canon 11-24mm f/4L, Canon 16-35mm f/2.8L II, Canon 16-35mm f/2.8L III, Canon 16-35mm f/4L IS, Nikon 14-24mm f/2.8G ED AF-S, Nikon 16-35mm f/4G ED AF-S VR

Last month Roger did his standard series of optical bench tests on the new Canon 16-35mm f/2.8L III. But that doesn’t show you what it looks like in real pictures. I decided to take the new lens and a large selection of comparable lenses from Canon, as well as offerings from Nikon, Sony, and Tokina, and take them across the street to one of our wonderful Memphis parks. I captured the same scene with all the lenses using the Canon 5D mark IV for Canon mount, the Nikon D810 for Nikon mount, and the Sony a7RII for Sony mount (the LA-EA3 adapter was used for the A-mount lenses).

Note: I forgot the Tamron 15-30. Sorry, y’all.

Here’s the Canon 16-35mm f/2.8 mark III compared to the Canon 16-35mm f/2.8 mark II at 16mm f/2.8:

Canon 16-35mm Mark III at 16mm f/2.8

Canon 16-35mm Mark II at 16mm f/2.8

Here’s the III compared to the II at 35mm f/2.8:

Canon 16-35mm Mark III at 35mm f/2.8

Canon 16-35mm Mark II at 35mm f/2.8

You’ll see that the III is noticeably sharper everywhere in the frame. Here are center and edge crops at 16mm side by side:

And here are center and edge crops at 35mm side by side:

I used to always recommend the Canon 16-35mm f/4L IS version over the II for landscapes. I don’t really have to do that anymore since the III is better at f/2.8 than the IS version is wide open. Here’s the III compared to the IS at 16mm f/4 –

Canon 16-35mm Mark III at 16mm f/4

Canon 16-35mm f/4 at 16mm f/4

Here they are at 35mm f/4 –

Canon 16-35mm Mark III at 35mm f/4

Canon 16-35mm f/4 at 35mm f/4

I went ahead and also shot the same scene with the following lenses wide open:

Canon 11-24mm f/4 at 11mm, 16mm, and 24mm –

Canon 11-24mm at 11mm f/4

Canon 11-24mm at 16mm f/4

Canon 11-24mm at 24mm f/4

Tokina 16-28mm f/2.8 in EF mount at 16mm and 28mm

Tokina 16-28mm at 16mm

Tokina 16-28mm at 28mm

Nikon 14-24mm f/2.8 at 14mm, 16mm, and 24mm

Nikon 14-24mm at 14mm

Nikon 14-24mm at 16mm

Nikon 14-24mm at 24mm

Nikon 16-35mm f/2.8 at 16mm and 35mm

Nikon 16-35mm at 16mm

Nikon 16-35mm at 35mm

Nikon 17-35mm f/2.8 at 17mm and 35mm

Nikon 17-35mm at 17mm

Nikon 17-35mm at 35mm

Sony FE 16-35mm f/4 at 16mm and 35mm

Sony FE 16-35mm at 16mm

Sony FE 16-35mm at 35mm

Sony A-mount 16-35mm f/2.8 ZA SSM at 16mm and 35mm –

Sony A-Mount 16-35mm at 16mm

Sony A-Mount 16-35mm at 35mm

Sony A-mount 16-35mm f/2.8 ZA SSM II at 16mm and 35mm –

Sony A-Mount 16-35 ZA SSM II at 16mm

Sony A-Mount 16-35 ZA SSM II at 35mm

While it might be hard to see in these smaller jpegs, I think it’s pretty clear from these images that the new Canon 16-35mm f/2.8L III is a cut above the rest, like the MTF charts show in our previous post. Click the link here if you want to download full raw files of these images so you can compare them yourself.

Author: Joey Miller

I’m Joey. I love cameras, especially old film cameras, and I can’t remember the last day I didn’t take a photo. Digital cameras are great, and they keep me employed, but I also still like processing my own film. I’m stuck somewhere in the middle. I shoot every single day, no matter what.

Posted in Geek Articles

How to Create a Photo Booth for an Event

How the Fuji X-Pro2 Resparked my Passion for Photography

Justin Croushore

I just got this lens and love it, albeit a few build quality issues. My Focus ring makes a weird wet sounding pop if turned from the infinity hard stop, and my outer barrel on the right side of the lens cracks and pops when its lightly touched. So far its functioning perfectly and is razor sharp. Just wondering if you guys had an experience with this at all? For $2200 that quality is disappointing. Canon wants to take a look at it since it sounds abnormal.
Patrick Chase

Edited my comment to reflect your correction. Thanks!
Brandon Dube

A lens designer will refer to a baffle as a baffle or a light trap. These are rarely used to introduce vignetting; reducing the semidiameter of an element can have an equal effect while reducing cost, instead of increasing cost by adding more parts.

You cannot look at defocused images of a point source to determine vignetting empirically, as you are unaware of the field curvature and astigmatism. Notably, astigmatism is almost always the limiting aberration in a wide-angle design. If a lens is astigmatic, you can produce the exact same “football” shape as if it were vignette.
Patrick Chase

EDITED to reflect Brandon’s feedback (he knows optical HW better than I do).

Lens designers use vignetting specifically to suppress aberrations on the periphery (by blocking peripheral rays). If it weren’t done then the wide-aperture corner performance would probably crater. Most wide-angle lenses vignette to some degree for that reason. I should be clear that when I say this I’m distinguishing aperture-dependent “vignetting” from aperture-independent “falloff”, which arises for different reasons.

Within digital workflows falloff and blurring are actually strongly coupled and exchangeable. They both cause information to be lost. Falloff does so in the intensity domain (signal is reduced, which means that the signal/noise ratio is reduced) while blurring does so in the spatial domain (neighboring pixels are correlated). The two forms of information loss can be exchanged in post, by sharpening and denoising. Sharpening (partially) removes spatial correlation but also amplifies noise and thereby degrades S/N. Denoising improves S/N but also increases spatial correlation.

I have the Canon 16-35 III, and based on a 15-year career in image-processing (including sharpening and denoising algorithm design), I think that Canon struck a reasonable balance with that lens given the current state of the post-processing art. Yes, the extreme corners are vignetted, but they don’t need anywhere near as much sharpening as with some other lenses, and the final result (after boosting the corners and sharpening to correct for aberrations) is as good as any other zoom in that class.
Michael Clark

When doing statistics, if you’re going to throw out the minimum value on one end you should also do the same with the maximum value on the other extreme. So kick out the 4.6 as well as the 1.7. Then average the rest. 3.5666666666667
Michael Clark

If you compare DxO’s results for the EF 70-200mm f/2.8 L IS II to everyone else’s it is abundantly clear either they are not getting blueprinted lenses direct from Canon like you suggest or the courier who delivered it that day was using Harambe to load the truck.
Ernest Green

Good points. In all reality, the 17-40 F4L is 98% of the performance *stopped down* for landscape photography with the exception being the EXTREME corners. The corners, 2/3 are very sharp at F8. Who peers into the corners anyway? I’ve printed maybe 5-6 of my ultra wide photos over my 13+ years of DSLR ultra wide photography. And between my 17-40 and my 16-35, i can’t tell the difference even when printed quite large. I think people obsess (myself included sometimes) over pixel peeping but most lenses are very sharp stopped down and as long as you have a good sensor behind it, you’ll be fine no matter what lens you use.
Brandon Dube

You cannot modify a given sample of a lens to reduce vignetting. In general, it will vary only on a difficult-to-measure level between samples.
Sator Photo

I have just pulled out the DPR review results. Here is the running “meta-analysis” of currently available data. If you look carefully, the DPR results state there is -2.63 EV of falloff at 16 mm f/2.8:

https://www.dpreview.com/reviews/updating-a-classic-canon-16-35mm-f2-8l-iii-lens-review/2

(4 + 4.6 + 4.07 + 1.7 + 2.63 = ) ÷ 5 = 3.4

If you take out the outlier result from DXO Mark (could this be an average vignetting over multiple focal lengths rather than just maximum vignetting at 16mm f/2.8????) then here are the results:

(4 + 4.6 + 4.07 + 2.63 = ) ÷ 4 = 3.825

However you look at this, it is a poor result. DPR also fail to reveal how they their sourced their copy. Given how big this site is, I would guess it was submitted by the manufacturer unlike the other sites that source their own copies. That means that it could have been a specially tweaked copy of the lens to make it look its best, one unrepresentative of average production copies.

I would love it if Roger Cicala released data about average vignetting over ten independently sourced copies of this lens. That would settle it almost for good.
Panacea

And now, DPReview would differ greatly with your “numerous reviewers”. Citation of DxOMark’s hard data and subjective written review paints a vignetting picture of being almost the same as that of its predecessor (the EF 16-35mm f/2.8L II), and better than that of the EF 16-35mm f/4L IS USM, not just at matching apertures, but at wide-open as well (!).

https://www.dpreview.com/reviews/updating-a-classic-canon-16-35mm-f2-8-iii-lens-review

Chris M Williams, author (in Comments): “The DXO data reports a 2 and 1/3 stop decrease in the corners when shot at 16mm F2.8 and that matches what we saw in the field. Check out our real world image comparison widget and see for yourself, but our observations agree with the DXO data.”
Adam Sanford

Sator, I regret that I have only one like to give your post.
Sator Photo

I totally agree with this. Roger Cicala recently said that you should not rely on a single source for data about a lens, but that you should perform your own “meta-analysis” of pooled data from multiple independent sources:

https://www.lensrentals.com/blog/2016/08/pay-attention-to-the-man-behind-the-curtain/

The fact that the DXO vignetting result of -1.7 EV is an “outlier” reading means that you simply must ignore it, when several other independent tests suggest the value is worse than -4 EV. In this instance a rough meta-analysis of data means: (4 + 4.6 + 4.07 + 1.7) ÷ 4 = 3.6. If we take out the outlier result then (4 + 4.6 + 4.07) ÷ 3 = 4.2. We don’t have a precise value from TDP so I have just entered “4”, but the actual result was worse than -4 EV.

Unless DXO Mark provide sufficient explanation of the result, the best thing to do is to ignore this single reading. The claim that you should believe DXO Mark just because they said so, is credulous beyond belief. That they do not explain their testing methodology means that for all we know they test lenses after software correction of defects such as vignetting. Since they say nothing to contradict that, we cannot know what this or any of the other “magic numbers” that pop up on their website mean. As for their aggregate “DXOMark Score” and their single digit value for “sharpness”, I agree with Roger Cicala that without knowing how these were calculated such numbers are worse than useless. It reminds me of the Hitchhiker’s Guide to the Galaxy where at the end of universe you learn that the meaning of life is “42”. I quote Roger Cicala:

“We all know DxOMark likes to compress everything down to a single, and in my opinion less than useless, number. …Well it has some use since we immediately know that someone who says “DxO rated it as 83.2, the best lens ever!!!!” is a fanboy who doesn’t understand testing at all. Resist the temptation to try to reason with unreasonable people.”

I couldn’t agree more.

Oh yes, unlike Lens Rentals, DXO Mark rely on manufacturers to donate equipment for testing rather than sourcing their own. That means that only one copy of a manufacturer supplied lens is tested, and this may not be a representative sample of what is available out in the field. The only reason we know this is, unusually for them, DXO Mark admitted it themselves:

https://www.dxomark.com/Lenses/Sony/Sony-300mm-F28-G-II-mounted-on-Sony-SLT-Alpha-99__831

I quote: “Sony lenses we analyzed were provided by Sony itself”. We can guess that DXO Mark get to keep the donated lenses and retest them as new bodies are released for that mount.
DrJon

I have the f4 and looked at the III as there are some things about the edge performance of the f4 that annoy me (although the centre sharpness is amazing). I won’t be upgrading though and I don’t need to try a III to have some IMHO valid reasons, please explain if you feel I should use a III before saying the following:
(1) If I want to shoot video I’ll want the IS.
(2) The IS is surprisingly good, I was shooting in a Museum at 1/5 and 1/6 and getting a healthy proportion of keepers. This was at slightly higher ISO than would have been my first choice but mostly the f-stop was DoF controlled although some shots would have worked with f2.8 (I was after lots of sharpness over the frame though). An extra stop (but no IS) wouldn’t have got me to hand-holdability and a tripod wasn’t possible so I wouldn’t have got the shots with the III. (Examples: https://www.dpreview.com/forums/post/58453083 – just done as a quick test to help someone with a buying decision.)
(3) Outside I often want at least f5.6 of DoF and many times f8. Not that I wouldn’t use f2.8 occasionally if it was available.
Now if the III had a stabiliser…
(Oh, not saying the III wouldn’t be a good choice for a number of people, but indoors it’s going to be tricky if you can’t use a tripod, unless you need a fast shutter speed due to subject movement and just go for it with the ISO regardless of noise/noise-reduction-nasties and low-DR.)
Adam Sanford

DXO is a sensor site that rates lens’ overall score based on how many pixels are sitting behind them.
They use blackbox scoring methods that they have never published, and as result, no one takes them seriously (at least not with lenses).

However, here are some more reputable sources on this new Canon offering, who are almost in unison on this:

TDP: https://goo.gl/pO4uJi (Greater than 4 stops vignetting)
“At 16mm wide open at f/2.8, this lens has very strong vignetting with the corners darkened by over 4 stops, impacting the benefit that the f/2.8 aperture provides.”

PZ: https://goo.gl/6dwafV (4.6 stops vignetting)
“A major weakness is the high amount of light falloff in the image corners at 16mm @ f/2.8. A vignetting of 4.6 EV (f-stops) is WAY beyond our usual scale for full format lenses. Unless you have to go for f/2.8 or simply like the effect, you should stop down to at least f/5.6 in order to reduce the light falloff to an moderate degree.”

LensTip: https://goo.gl/e2sWwR (Greater than 4 stops vignetting)
“It would be difficult to call this situation other than dramatic. Such high vignetting we haven’t seen so far in our tests. At 16 mm and by f/2.8 in the frame corners disappears 75% of light (?4.07 EV)”

And this is not at all about the 16-35 f/4L IS — those are separate comments about landscape work. UWA f/2.8 zoom buyers need and expect a usable f/2.8 image for sports, events, astro, reportage, etc. Having to push corners 4+ stops on an ISO 3200 dark concert shot or long exposure + high ISO astro shot would probably not be ideal.

I’m raising this issue because Canon dramatically increased the wide-open resolution of this lens — it’s off the charts for its class — but it appears to have gone (rather dramatically) in the other direction for vignetting, and I’m asking the pros here to size up how important that might be.
Adam Sanford

That implies that I’m posting to validate my prior purchasing decisions, that I am irrational, that I’m trying to smear a product, etc. Could not be further from the truth.

I am saying that at apertures you will shoot for landscapes, the differences optically between the two lenses are imperceptible and you shouldn’t throw $1,000+ away for better f/4 performance. I found the article’s comment that ’16-35 f/2.8L III being sharper at f/2.8 than the f/4L IS @ f/4 makes it a better choice for landscapers’ as being off-target and I wanted to share some perspective on that.

I’m really hard-pressed to know why a landscaper — a dedicated landscaper — would choose this f/2.8 lens. Unless you dabble in sports/events/astro, I really do believe the f/4L IS lens is the better call for the reasons I enumerated.
Darin

Well if it’s obvious how bad it is at 1400px like the Sony ZA 16-35 2.8 there is no need for larger 😉 But they do have a link for the raw images of them all: https://drive.google.com/drive/folders/0B2OzfD2A294ccW5mejhtdW1QUUk?usp=sharing
Panacea

DXOmark would differ greatly with your “numerous reviewers”. In fact, their shot-at-every-full-stop comprehensive data would paint the overall vignetting picture as while being a little worse than its Mk II predecessor, is actually *a little better* than that of the EF 16-35mm f/4L IS USM.

https://www.dxomark.com/Reviews/Canon-EF-16-35mm-f-2.8L-III-lens-review
Ivan Boden

I suppose Canon had to make lens improvements due to the higher megapixels of the the 5DIV and the 5DSR cameras. The higher resolution is less forgiving and the need for high resolving lenses is apparent.
Panacea

I do find it terribly amusing when people who own one product (the EF 16-35mm f/4L IS USM, in this case) take it upon themselves to attempt to nullify the opinions of experts who have access to many competing versions of the product to play with.

No, they couldn’t possibly know anything. Or know how a lens is used. If they like something you don’t own (like the EF 16-35mm f/2.8L III USM in question), it must be total ignorance on their part. Their first-hand experience taking pictures with it is nothing against the specs you can pull up. Your experience with the product you like is definitive… because you’ve used the product you own and you really like it. You are super-rational.
Turniphead

Interesting, but please could you either link the full resolution shots offsite, or post more 100% crops? How are we supposed to pixel peep when you only give us 1400px images 😉
Adam Sanford

And we lack any testing on how the f/2.8L III does ‘mechanical vignetting’-wise with a Lee setup — I’m referring to the Lee hardware in the field of view even after using UWA rings and such.

The 16-35 f/4L IS is very good at this (https://goo.gl/DRHw59) because it appears that Canon tucked the filter rings as close as humanly possible to the front element. I’m hopeful they have done the same with the f/2.8L III — I’m hard pressed to think of why they wouldn’t do this.
Sean Setters

Also, for some photographers, the EF 16-35mm f/4L IS USM’s 77mm front filter thread may be a blessing because of its increased compatibility with other lenses they may already own. The 82mm standard is certainly getting more popular, but… it’s still not as popular as 77mm and the larger filters are more expensive.
Adam Sanford

Also, two hard looks at this lens (this story and the optical bench results) and no comment at all about the vignetting at 16mm. I deeply value your take on lenses, and I found numerous reviewers pegging over *4 EV* darkening in the corners. Any thoughts on that for sports / astro / event people who would naturally be looking at an UWA f/2.8?
Adam Sanford

“I used to always recommend the Canon 16-35mm f/4L IS version over the II for landscapes. I don’t really have to do that anymore since the III is better at f/2.8 than the IS version is wide open.”

#Wince

1) Traditional non-astro landscape work is rarely shot at f/4, if at all.
2) The f/2.8L III costs twice as much for the extra aperture stop landscapers will never use.
3) The f/2.8L III lacks IS for handheld larger DOF work, like shooting the naves of churches, in museums, handheld nighttime cityscape walkabout shots, etc.
4) For those that have to lug their landscape gear on backcountry hikes and such, weight matters. The f/2.8L III is about a half pound heavier… again, for a stop you will never use.

The 16-35 f/4L IS is a far better call for landscapers despite whatever this real-world look or your prior OLAF testing might imply.

The only *landscapers* that should buy the 16-35 f/2.8L III are either generalists that shoot a little bit of everything or enthusiasts who shoot landscapes *now* but might want a future-proofed tool for other interests they may develop someday.

Geek Articles

Real World Testing of the New Canon 16-35mm f/2.8L III

Author: Joey Miller

Recent Posts