Review: Canon EF 16-35mm f/2.8L III

[Canon Rumors Guy]

<p>Roger at Lensrentals.com has completed his bench test of the Canon EF 16-35mm f/2.8L III wide angle zoom lens. It looks like the lens performance is the new pinnacle for wide angle zoom lenses and you should probably go ahead and buy one!</p>
<p>From Roger:</p>
<blockquote><p>This summary is quick and simple. From a resolution standpoint, the <a href="http://www.lensrentals.com/rent/canon/lenses/wide-angle/canon-16-35mm-f2.8l-iii" target="_blank">Canon 16-35mm f/2.8 Mk III</a> is the best f/2.8 wide-angle zoom available. You might be better served with the f/4 IS and some money in your pocket. There are also some very good wide-angle f/2.8 zooms available from third party manufacturers that are a lot less expensive and might offer more bang-for-the-buck. But if your style of photography needs the highest resolution you can get with a wide-angle lens, well this is it. I don’t use a wide-angle zoom all that often, but when I do, it will be this one. <a href="https://www.lensrentals.com/blog/2016/10/canon-16-35mm-f2-8l-mark-iii-optical-bench-tests/">Read the full review</a></p></blockquote>
<p>It looks like the EF 16-35mm f/2.8L III is in stock across most retailers.</p>
<p>Canon EF 16-35mm f/2.8L III: <a href="https://www.bhphotovideo.com/c/product/1274708-REG/canon_ef_16_35mm_f_2_8l_iii.html/BI/2466/KBID/3296">B&H Photo</a> | <a href="http://amzn.to/2bPL0jq">Amazon</a> | <a href="http://www.adorama.com/CA16353.html?KBID=64393">Adorama</a> | <a href="http://bit.ly/2bkKGfQ">Canon Store</a> | <a href="https://mpex.com/canon-ef-16-35mm-f-2-8l-iii-lens.html?acc=3">Midwest Photo</a></p>
<span id="pty_trigger"></span>

 

[Mt Spokane Photography]

I read Rogers article earlier today. It sounds like a stellar lens. I had the MK II version, but never really used it. I don't think I had a sharpness issue, just that its not the type of use I need.

I've stopped buying lenses just because they are the best of the best, and just purchase what gets used. But, if I used those focal lengths in low light, its a definite lens to get.

 

[pcho]

Thanks Roger for your comprehensive review. Confirms my own little brick wall test. I tested at F2.8 against the Ziess 15mm at 15mm and at f2.8 and I cannot separate the two in my brick wall test for sharpness. It's just that good

 

[rfdesigner]

I hate to rain on Canon parade, (this does look good) BUT.. and this is a rather big but...

We might be going down a blind ally with this "sharp in the corners" approach.

There's a link in the comments to Rogers excellent review showing:

4 stops of vignetting

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

Now if this is the way that manufacturers are going to be getting "sharp in the corners" (i.e. turn the lens into a pinhole lens by that point) then for things like astro it's fairly pointless.

i'll freely admit that this is not what you buy for astro work, ideally you get a manual focus wide aperture prime lens, you put all the money where it matters. But it still leaves the question.

Personally I'd like to see a new metric:

Sharpness vs Corner aperture.

in this lens case, it works out as F11, which is not what I'd call wide in the slightest.. (you're going to need the 5DIV higher DR if you want to correct that without visible noise in the corners.), my ancient and soft 28f1.8 has only 2 stops of vignetting, as does the 16-35f4IS so:

The 16-35 F4 is wider in the corners than this new F2.8

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=949
now that's not something I would have expected

 

[mb66energy]

[...]

in this lens case, it works out as F11, which is not what I'd call wide in the slightest.. (you're going to need the 5DIV higher DR if you want to correct that without visible noise in the corners.), my ancient and soft 28f1.8 has only 2 stops of vignetting, as does the 16-35f4IS so:

The 16-35 F4 is wider in the corners than this new F2.8 (1)

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=949
now that's not something I would have expected (2)

(1) Others neglect distortion and let it stay at 5% while correcting it in camera for the jpgs ... IMO vignetting is a little bit easier to correct while preserving contrast and sharpness - especially with modern high DR sensors where lifting the edges doesn't increase noise too much.
Dont't misunderstand me: A f/2.0 16-35 IS with 1 stop vignetting wide open, extraordinary sharpness and IS for 2000 EUR would be welcome ...

(2) Compare both lenses at f/4: same amount of vignetting! I think one has to compare both situations.

If someone needs the f/2.8 for overview portraits AND corner sharpness for landscape/architecture - maybe this is a "one lens for both"-approach at a premium price ...

 

[mb66energy]

I read Rogers article earlier today. It sounds like a stellar lens. I had the MK II version, but never really used it. I don't think I had a sharpness issue, just that its not the type of use I need.

I've stopped buying lenses just because they are the best of the best, and just purchase what gets used. But, if I used those focal lengths in low light, its a definite lens to get.

The same here: I am more on the "tele side", 40mm is wide and 24mm is WIIIIIIDE for me. So I will live with the EF-S 10-22 on the EOS M classic if I need an ultrawide. 16-35 isn't too interesting for me and I would choose the f/4.0 16-35 if I would go into ultrawide on a more or less regular basis.

About "buying lenses just because they are the best of the best": When I was in that mood I hadn't the money, when I had the money I hadn't the time to use them. For WHAT I DO I am really satisfied with two EOS5D classics and some unspectacular fix focals like 2.8 24 (old), 2.8 40, 2.0 100, 2.8 100 macro (USM, non-IS) + the great f/4.0 70-200 non-IS. All these lenses match the resolution of the 5D classic easily + operating the 5D classic is really simple and fun.

But Kudos to Canon for doing what is possible and creating instruments which may lead to exceptional photos in the right hands for the right purposes!
But please, Canon, think about creating a f/4.0 40-200 or 50-200 in the tradition of the great f/4.0 70-200 zooms - that would be a reason to invest again in Canon glass.

 

[MichaelG]

I shoot most with a wide angle lens and I was sure to replace the 16-35 II once the "III" is available. But 4-stop vignetting is absolutely not acceptable. The vignetting correction of the "II" now creates visible noise so that I usually don't correct. 4-stop is even worse. I think I wait for Version "IV".

 

[Random Orbits]

I hate to rain on Canon parade, (this does look good) BUT.. and this is a rather big but...

We might be going down a blind ally with this "sharp in the corners" approach.

There's a link in the comments to Rogers excellent review showing:

4 stops of vignetting

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

Now if this is the way that manufacturers are going to be getting "sharp in the corners" (i.e. turn the lens into a pinhole lens by that point) then for things like astro it's fairly pointless.

i'll freely admit that this is not what you buy for astro work, ideally you get a manual focus wide aperture prime lens, you put all the money where it matters. But it still leaves the question.

Personally I'd like to see a new metric:

Sharpness vs Corner aperture.

in this lens case, it works out as F11, which is not what I'd call wide in the slightest.. (you're going to need the 5DIV higher DR if you want to correct that without visible noise in the corners.), my ancient and soft 28f1.8 has only 2 stops of vignetting, as does the 16-35f4IS so:

The 16-35 F4 is wider in the corners than this new F2.8

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=949
now that's not something I would have expected

Vignetting at f/4 is about the same between the 16-35 f/2.8 III and 16-35 f/4 IS. You can use it at f/4 and get slightly better IQ than the f/4 without the IS and a higher price, or you can chose to use f/2.8 and live with the increased vignetting.

 

[TWI by Dustin Abbott]

I hate to rain on Canon parade, (this does look good) BUT.. and this is a rather big but...

We might be going down a blind ally with this "sharp in the corners" approach.

There's a link in the comments to Rogers excellent review showing:

4 stops of vignetting

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

Now if this is the way that manufacturers are going to be getting "sharp in the corners" (i.e. turn the lens into a pinhole lens by that point) then for things like astro it's fairly pointless.

i'll freely admit that this is not what you buy for astro work, ideally you get a manual focus wide aperture prime lens, you put all the money where it matters. But it still leaves the question.

Personally I'd like to see a new metric:

Sharpness vs Corner aperture.

in this lens case, it works out as F11, which is not what I'd call wide in the slightest.. (you're going to need the 5DIV higher DR if you want to correct that without visible noise in the corners.), my ancient and soft 28f1.8 has only 2 stops of vignetting, as does the 16-35f4IS so:

The 16-35 F4 is wider in the corners than this new F2.8

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=949
now that's not something I would have expected

Even on the Mark IV there is still some noise that has been created in the corners after correction. Not bad, but very obvious when comparing the Tamron 15-30 after correction. It's only one copy of each, obviously, but while the Canon has a sharpness advantage in the middle at 16mm vs. 15mm, the Tamron (at least my copy) is clearly sharper in the corners.

 

[Alex_M]

According to TDP, Canon 11-24 F4 L equally dark in the extreme coners @11mm and F4 compared to the canon 16-35 F2.8 III wide open and @16mm. That does not stop the lens from being one of the most desirable ultra wide angle zoom lens. I suspect that Canon sacrificed light transmission in extreme corners in order to preserve corner sharpness and avoid bulging front element.

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?Lens=1073&Camera=979&FLIComp=0&APIComp=0&LensComp=977&CameraComp=453&FLI=0&API=0

 

[davidmurray]

Thanks for the review.

I am very tempted to get this lens, but it's a question of what to do first - to replace my car or buy this new lens.

I think I'll wait a little while to see the general reaction regarding the vignetting issue and then decide on 2.8 without IS or 4 with IS.

 

[fish_shooter]

Much of the light fall off is due to the cosine to the fourth power rule. There is probably some mechanical and optical vignetting as well - a result of keeping the front filter size as small as 82mm and having a large maximum aperture for this angle of view.

 

[rfdesigner]

Much of the light fall off is due to the cosine to the fourth power rule.

Go on.. I feel like I should know this but I can't seem to drag up why we're taking the fourth power of a cosine.

 

[ahsanford]

Much of the light fall off is due to the cosine to the fourth power rule. There is probably some mechanical and optical vignetting as well - a result of keeping the front filter size as small as 82mm and having a large maximum aperture for this angle of view.

The prior 16-35 f/2.8L II is also 82mm and it's a stop better when shot wide open:
http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

So that's not (entirely) it. Something else is at work here.

- A

 

[fish_shooter]

Much of the light fall off is due to the cosine to the fourth power rule.

Go on.. I feel like I should know this but I can't seem to drag up why we're taking the fourth power of a cosine.

It is an optical fundamental - the mathematical relationship describing light falloff of a lens.

 

[fish_shooter]

Much of the light fall off is due to the cosine to the fourth power rule. There is probably some mechanical and optical vignetting as well - a result of keeping the front filter size as small as 82mm and having a large maximum aperture for this angle of view.

The prior 16-35 f/2.8L II is also 82mm and it's a stop better when shot wide open:
http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

So that's not (entirely) it. Something else is at work here.

- A

This is something that Canon will have to tell us. You could mount your copy of the Mk III lens in a lathe and cut off the front parts that hold the lens shade and filter mounts and see if vignetting is reduced ;->>

For now you have a choice: 1 stop less light fall off or a sharper lens. Lens design is all about compromise.

 

[Larsskv]

Much of the light fall off is due to the cosine to the fourth power rule. There is probably some mechanical and optical vignetting as well - a result of keeping the front filter size as small as 82mm and having a large maximum aperture for this angle of view.

The prior 16-35 f/2.8L II is also 82mm and it's a stop better when shot wide open:
http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

So that's not (entirely) it. Something else is at work here.

- A

This is something that Canon will have to tell us. You could mount your copy of the Mk III lens in a lathe and cut off the front parts that hold the lens shade and filter mounts and see if vignetting is reduced ;->>

For now you have a choice: 1 stop less light fall off or a sharper lens. Lens design is all about compromise.

With the TDP comparison tool, you can choose 16mm and a UV-filter attached. The filter ring (on the filter itself) has the potential to add to the vignetting, but in this case, the negative effect is small to my eye. Thus, the front parts doesn't seem to be the explanation.

http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?Lens=1073&Camera=979&FLI=0&API=0&LensComp=1073&CameraComp=979&FLIComp=5&APIComp=0

Maybe the coatings are causing much of the vignette?

 

[Alex_M]

In order to achieve the maximum corner sharpness as well as centre sharpness the focal plane has to be totally flat. Unfortunately flat focal plane results in extreme loss of light in the corners. Increased focal plane curvature results in brighter corners and decreased corner sharpness as a side effect... This is my understanding of why the older lens was brighter in the corners than the new one.

Much of the light fall off is due to the cosine to the fourth power rule. There is probably some mechanical and optical vignetting as well - a result of keeping the front filter size as small as 82mm and having a large maximum aperture for this angle of view.

The prior 16-35 f/2.8L II is also 82mm and it's a stop better when shot wide open:
http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?FLI=0&API=0&FLIComp=0&APIComp=0&Lens=1073&Camera=979&LensComp=412

So that's not (entirely) it. Something else is at work here.

- A

 

[rfdesigner]

Much of the light fall off is due to the cosine to the fourth power rule.

Go on.. I feel like I should know this but I can't seem to drag up why we're taking the fourth power of a cosine.

It is an optical fundamental - the mathematical relationship describing light falloff of a lens.

ok, I've got my head straight on this, and it isn't quite a fundamental. It's close because it's based on the principle of illuminating a circular aperture from an angle, i.e. at 90 degrees you get nought, straight on you get 100% and in between you get something that is described by the cos^4 rule.

Of course if you have "streering" lenses in front and behind the limiting element then the rules change, in theory you could mitigate the effect to some extent.

Calculating what you get from a 16mm focal length lens on a full frame camera, the "field of view" on the outside of the lens is greater than inside the camera body.. so using the wider FOV (to be fair) we get 2.7 stops of vignetting.

On an 11mm optic you get 4 stops of vignetting.

Well that's what I got and I'm still fumbling around this, if someone wants to correct please go ahead

 

[mb66energy]

Much of the light fall off is due to the cosine to the fourth power rule.

Go on.. I feel like I should know this but I can't seem to drag up why we're taking the fourth power of a cosine.

It is an optical fundamental - the mathematical relationship describing light falloff of a lens.

ok, I've got my head straight on this, and it isn't quite a fundamental. It's close because it's based on the principle of illuminating a circular aperture from an angle, i.e. at 90 degrees you get nought, straight on you get 100% and in between you get something that is described by the cos^4 rule.

Of course if you have "streering" lenses in front and behind the limiting element then the rules change, in theory you could mitigate the effect to some extent.

Calculating what you get from a 16mm focal length lens on a full frame camera, the "field of view" on the outside of the lens is greater than inside the camera body.. so using the wider FOV (to be fair) we get 2.7 stops of vignetting.

On an 11mm optic you get 4 stops of vignetting.

Well that's what I got and I'm still fumbling around this, if someone wants to correct please go ahead

Just stumbled over your remark about steering lenses (you ment that? I am a non-native writer ...): If you look at the aperture of an ultrawide and change the angle you see a strange behaviour: The aperture looks near circular from about 50° while a lens cap (as model for a simple circular hole) seems to be a flat ellipse ... just to back up your idea of steering lenses. Putting a light source on the back of the lens helps to see the effect. I just checked it during writing on my "poor mans ultra wide", the EF-S 10-22.

About light fall of observed on a circular hole: It should be the same like that of a square hole and I see only a simple cosine rule. I think the dramatic fall off is due to sensor reflectivity - but here too bending the light is helpful by other steering groups which try to reduce the angle of incidence.

More math is not possible at the moment after creating some examination questions in mathematics ...