Canon EF 14-24 f/2.8L [CR2]
- Thread starter Canon Rumors
- Start date
- Status
Oh, absolutely. Controlled tests are only meaningful when done in duplicate, triplicate, etc. Multiple copies of lenses, for example. I've only presented data from one 16-35 here, for example, though I did present some aggregate data from 4 copies of the Nikon 14-24... showing you the best/worst performance out of a batch of 4 lenses.
But then, do you know my credentials? Do you trust my testing methodology? Etc., & so on & so forth. Which is why I tried to control as many variables as possible, & provide my testing conditions & RAW files if you'd like to see for yourself.
That being said, allow me to show you a quick visualization of photozone.de's MTF numbers I did of the Nikon 14-24 vs. the Canon 16-35.
As you can clearly see, Nikon extreme f/2.8 performance matches Canon's center f/11 performance (the left endpoint of the blue-dashed line matches the right endpoint of the red-solid line).
Wow, seems strangely reminiscent of the pictorial examples I showed!
Meanwhile, by f/5.6, the Nikon is as sharp at the extremes as the Canon lens ever is, anywhere.
Wow, again, seems strangely reminiscent of the pictorial examples I showed!
Now, let's take a look at extreme vs. center performance for both lenses. I'll plot extreme/center performance ratios, but this time with resolution (2-dimensional, meaning LPPH numbers have been squared) as a function of aperture), 1 being equivalent extreme vs. center performance:
Both lenses are so sharp at the center that the extreme performance only begins approximating center performance after center performance is depleted due to diffraction effects. But the Nikon lens almost approaches center/extreme equivalence by f/11, while the Canon still only shows ~79% extreme vs. center performance by f/11. At f/2.8, extreme resolution is at a staggeringly low 26% of center performance for Canon, while the Nikon extreme/center performance is at 62%.
Meanwhile, the Canon extremes never even come remotely close to the Nikon extreme performance, at best only achieving 61% resolution of the Nikon's best extreme resolution performance. Put another way, at best the Canon extreme performance is still only 82% the worst performance of the Nikon at its extremes.
Let me restate that: the best the Canon lens is capable of at the extremes is still only 61% of the best the Nikon lens is capable of at the extremes (2750^2/3500^2 LPPH).
61%. Sounds pretty close to 50%. Half the resolution, people.
Are my qualitative pictorial examples still so unbelievable? Science! It works!
But feel free to take my pictorial results that appear to, qualitatively, directly reflect photozone's quantitative findings with a grain of salt... or, better yet, do your own tests
Of course, if edge-to-edge sharpness has not bothered you in your own shots/printing workflow, then perhaps this is all entirely irrelevant to you!No, it's not. Not if you know what DXO is actually doing.
If you don't believe them/me, do the tests yourself (as I did). My numbers in my own lab setup with a 13.2 stop transmission step wedge agreed with DXO within 0.3EV. Not to mention real-world comparisons that showed that I could underexpose an image 8x (3 stops) on a D7000 & still get its shadows to look better, pushed 3 stops, than a 5D Mark III image exposed properly.
I still shoot Canon (for reasons I don't wish to enumerate here). But I'm not blind to the obvious, which is why I use a Nikon 14-24 on my Canon 5D Mark III (for landscapes).
Upvote
0
If you are closely examining the prints, then be sure that a 30" photo will be enough to tell the difference.briansquibb said:
@saringman thanks for the comparison man. The Nikon lens is clearly superior. I don't agree with jettatore's opinion that there is a flaw in your method, because there isn't. You did what was right. I think that you just got an "average" (or typical) 16-35 in your hands.
Upvote
0
J
Jettatore
Guest
IRT: akiskev -Both of the pictures in Saringman's tests that he kindly posted are of an un-usable quality at a 100% viewing, either on a monitor, or on a high end print, --if not viewing close up for fine detail either full images could work but it's not a nice picture so it's a moot point. Otherwise the 100% crop views on all of the above are unflattering at best and I would say downright unusable. I'm sure someone will respond to the contrary, but if you think either of those images, from the Nikon or the Canon, are usable for viewing at 100% without heavy editing, then I would suggest the visual arts are not for you. If I was going to print any such image I would make sure that the end print could never show such flaws, even if viewed close up, it would be better to see pixels or ink droplets before seeing visual flaws in the image. So akiskev, while you don't have to agree with me, I would have to not agree with you also. At the very least, I get my copy of the 16-35 to produce sharper results without even having to go up to f/11 and I would say from those samples, I'm not even impressed with Nikon's 14-24.
Upvote
0
R
rushmore77
Guest
Hope it's smaller than the Nikon version.
For now my Samyang 14mm serves me very well. (Sometimes I can't believe how sharp it is.)
But canon 14-24 will be THE prefect lens to get, assuming the size is not too big.
I've been so disappointed with Canon lately but surely this is a great news.
For now my Samyang 14mm serves me very well. (Sometimes I can't believe how sharp it is.)
But canon 14-24 will be THE prefect lens to get, assuming the size is not too big.
I've been so disappointed with Canon lately but surely this is a great news.
Upvote
0
rushmore77 said:
Is this the same as the $400 Rokinon? I had one for two days and returned it because it wasn't sharp, even stopped down (I didn't expect much to start with, but even so)
Upvote
0
Respectfully, I really have no idea what you're talking about, & wonder if perhaps you're not used to viewing 22MP images at 1:1. As for it not being a 'nice picture', this was not a landscape shot, it was a test of a field-of-view that contained lots of buildings approximately equidistant from the camera for easy comparisons of sharpness.
The center 100% crops are tack sharp from both lenses for a default ACR setting of 25; you can make it much sharper if you'd like to. I don't think anyone here would disagree with me on this save for you. Are you following the links & viewing at 100%? Are you looking at the buildings/railings to judge sharpness? Don't look at trees blowing in the wind & cars moving... look at the static subjects, which are, again, tack sharp in the center, ruling out motion blur or whatever else you are invoking to attempt to invalidate these tests.
The side 100% crops are what they are, which was the entire point of this comparison.
And if you look at the numbers & graphs I posted earlier (e.g. 26% resolution at the extremes compared to the center at f/2.8 for the Canon, w/ the best the 16-35 can do at the extremes, i.e. f/8, still only 82% the resolution of the worst the Nikon does at its extremes at f/2.8 ), the pictorial examples I've provided should not surprise you at all, unless you also have a problem with photozone.de (who, by the way, do stellar tests).
If you're surprised by these results & would like to contribute to this discussion, I'd encourage you to also try shooting your own flat field tests, maybe even w/ a rental Nikon 14-24 to compare side-by-side. Try to walk in to the test unbiased, hard as that may be. If anything, I was initially biased toward Canon, b/c who'd want to go through all the extra trouble & cost to shoot w/ the Nikon lens on a Canon body, losing AF & aperture control?
As I said earlier, your examples do not provide any indication of how your 16-35 performs on a flat-field, though it's entirely plausible that it performs better than mine did (there are always tolerances in QC & manufacturing). Even if so, I doubt it would be by a wide margin, given the clear lack of decentering issues on my copy, my experience with multiple other copies of the 17-40, & photozone's numbers. Furthermore, your f/2.8 shot clearly shows incredible softness on a good portion of the left side (landscape orientation), even at web resolution. Again, no surprise there.
Upvote
0
Thanks Brian.
Here's an image I shot at f/18 on my 5D w/ a 17-40 a couple years ago:
Larger image here: http://f.cl.ly/items/1I120q0h1n1A1o432J3b/Seattle-RizalBridgeSunset.jpg
Looks plenty sharp at this size, of course, edge to edge. But in a 21"x34" print, the left side is clearly softer than the right. And this was f/18. Did my client that bought the image notice? Probably not
But then again, we are our own harshest critic, no?That was when I started my hunt for a better wide angle zoom... beginning w/ trying to get that lens fixed by Canon, buying another copy, buying a 16-35, etc. etc.
Upvote
0
I think it's always been acknowledged that the Nikon 14-24 is superior to any of Canon's wideangle zooms. You could probably increase the perception of sharpness in the 16-35 by correcting the CA, but it also shows how poor it is in comparison to the Nikon lens in that regard too. I know when I was researching a number of years ago, the original 16-35 wasn't really any better than the 17-40, that's why I went for the cheaper and lighter option of the 17-40. However, the MkII did also improve considerably over the MkI. As far as I'm aware, it still doesn't match the Nikon 14-24 though. On full frame, I find 17mm to be too wide most of the time, so I wouldn't get a great amount of use from the 14-24, otherwise I'd definitely consider it with an adaptor for landscapes, although filter use would be a pain. The 17-40 improves from around 24mm (at which point it's probably a little sharper than the 24-105 and definitely has less CA), so it would be interesting to see a comparison between your wideangle lenses at the longer end, to see if the differences are as marked.
Upvote
0
sarangiman said:Thanks Brian.
Here's an image I shot at f/18 on my 5D w/ a 17-40 a couple years ago:
Larger image here: http://f.cl.ly/items/1I120q0h1n1A1o432J3b/Seattle-RizalBridgeSunset.jpg
Looks plenty sharp at this size, of course, edge to edge. But in a 21"x34" print, the left side is clearly softer than the right. And this was f/18. Did my client that bought the image notice? Probably not
That was when I started my hunt for a better wide angle zoom... beginning w/ trying to get that lens fixed by Canon, buying another copy, buying a 16-35, etc. etc.
thats a nicely done HDR, did you use photomatix? if so i'd be interested in what your settings were. I'm not normally a fan of anything out of photomatix these days but thats pretty decent
Upvote
0
No, that was a merge of 4 exposures all done using a combination of luminosity/hand-drawn masks. Thanks for the compliments.
Upvote
0
J
Jettatore
Guest
sarangiman said:Thanks Brian.
Here's an image I shot at f/18 on my 5D w/ a 17-40 a couple years ago:
Larger image here: http://f.cl.ly/items/1I120q0h1n1A1o432J3b/Seattle-RizalBridgeSunset.jpg
Looks plenty sharp at this size, of course, edge to edge. But in a 21"x34" print, the left side is clearly softer than the right. And this was f/18. Did my client that bought the image notice? Probably not
That was when I started my hunt for a better wide angle zoom... beginning w/ trying to get that lens fixed by Canon, buying another copy, buying a 16-35, etc. etc.
Visually this new post is a fairly nice image, but pretty sharp is not a quality that I would attribute to it. I think this is why we have been in disagreement the entire time, our eyes and ability to see and more importantly how we use language to describe things just aren't the same. Not better or worse, just, not the same. I would say in your new shot that the roads and buildings are approaching sharp, but not sharp, the sky 'is' close enough to sharp for a sky, and the trees and greenery are downright blurry. I would not call this image sharp, even casually.
And from your earlier post...
No, they aren't. They aren't even close to a definition of something I would call tack sharp, and 100% crops from 35mm DSLR's rarely are. Now, if you don't not understand the concept that it would be better to see visible pixels or ink droplets before noticing visual flaws in the image, like Chromatic abberation or low-contrast or blurred imagery that isn't purposeful, then i just don't know what to say to get on the same page with you so that we can actually communicate, as that was the central point of the entire previous post that you had difficulty in understanding. As to your other comment about not knowing 22+MP resolutions, I believe at least the center lens portion of an 18MP 7D crop would give me some idea as well as 5DIII, D800, 1DsMIII and 5DII RAW files and Phase 1 IQ80 files, so I hope that settles your question, as well, yes I know how to open the full images posted to this forum.sarangiman said:
This really is going no-where though, so if any further response isn't likely to clear up communication difficulties or lead to something new, please stop evoking a new response from me. It's a waste of everyone's time, including mine, because at this point, this overcomplicated, under-useful conversation is more of a communications error than anything else so far as I can tell.
I'll add one last thing to try and clarify what 'sharp/sharpness' defined -means to me. Sharpness, is closely related to, if not a viable synonym for 'contrast'. In order to perceive sharpness, there must be contrast. A solid black cube printed on solid black paper will result in zero contrast. A black cube printed on white paper will result in high or if you will 100% contrast at the intersections of black and white areas of the full image. If you keep this in mind, and relate that high contrast is a requirement of sharpness, and then go back and look at your 100% crops (maybe even viewing them in black and white to remove additional confusion, you will see that areas like the pavement grit detail, and even the railings, etc., have a lack of contrast and hence a lack of visual sharpness. Like I tried to explain earlier, when viewed at 100%, these crops are un-usable and any presentation of an image should not allow for such an examination, as it would be simply better to see ink droplets or pixels before being able to discern a lack of contrast and sharpness. For this reason, things like improving exposure of your images, can lead to much sharper pictures. The railing in your Nikon RAW/Crop samples is about the closest thing to sharp the image has going for it but isn't perfect and the Canon RAW/Crop sample overall is exposed less adequately by the photographer than your Nikon sample and hence portrays less sharpness overall, among other things... You will be able to see this readily if you convert the RAW files of both images to B&W using the same settings without any other RAW editing adjustments and then view them at the same 100% viewing level. You will see you did not match exposure in the images, and hence, measuring sharpness between the two of them at that point becomes rather a fruitless effort as the contrast will never be the same and hence the impression of sharpness will vary.
This is my last post on this. I enjoyed some of this discussion and learned a bit, so thank you for that, but communication is often a very difficult thing, and as I mentioned before, this is getting way to complicated without being in person at a computer where we can easily learn what the other means and come to fruitful understandings and higher level conversations quickly. Best regards.
Upvote
0
I think what you're talking about is 'acutance', or local contrast... yes of course the buildings in this image will have higher acutance than the trees; the local contrast in the trees to begin with in this scene were very low. What is your point? That has absolutely no bearing on our conversation of lens MTF as a function of center vs. extreme.
You're right, we've had a total breakdown of communication. B/c when I said 'tack sharp', I thought I made it pretty clear that I meant 'about as sharp as you're going to get w/ any lens on this sensor'. Since this entire discussion was about comparing center vs. extreme sharpness on a 5D-ish camera, what would be the point of defining 'tack sharp' in any other way?
Actually, the exposures are exactly the same, barring T-stop differences between the lenses. 1/320s | f/2.8 @ISO 100 is equivalent to 1/40s | f/11 @ISO 200.
The flare decreases contrast in the Canon image, yes, & I mentioned this a while back. While I agree that contrast of course affects perceived acutance, I respectfully disagree that the decrease in contrast due to flare on the Canon shots prohibits one from seeing the clearly higher sharpness of the Nikon lens at the extremes.
Because if the decrease in contrast due to lens flare is confusing you about which of the following images is sharper:
Full-size image: http://cl.ly/HQaY/Nikon14-24_vs_Canon16-35_f2.8_Left.jpg
Or which of these images is sharper (remember: 100% crop taken from same portion of frame, so as to not give one shot an unfair advantage):
Full-size image: http://cl.ly/HTEK/Nikon14-24_vs_Canon16-35_Corner.jpg
... then you may have to get your eyes checked
Actually, here let me try to prove to you that the decreased contrast does not significantly affect your ability to judge which lens is sharper. I've done -100 contrast to the Nikon image. If you quantify the ratio of the brighter vs. darker siding on one of these buildings, the Nikon f/2.8 image (left) now has less contrast than the Canon f/11 image (right). Do you still have any trouble discerning which is sharper:
Full-size image: http://cl.ly/HTY9/Nikon14-24_-100Contrast_vs_Canon16-35_Corner.jpg
Seriously, I think I can rest my case now. If I were writing this formally as an article on a review site, yes, I would go out & redo the test on a cloudy day (would've done so already if I hadn't gotten rid of the 16-35) when there is absolutely no change in lighting between shots. But you can draw some pretty darn good conclusions from what I've posted here (barring copy variability, of course).
Unusable? Hardly. Especially in light of the -100 contrast (to the Nikon image) 'experiment' above.
Upvote
0
B
briansquibb
Guest
sarangiman said:
I have found that the 17-40 is best at f/8- f/11. After that the image gets softer, presumably defraction
Posting on this site further softens a jpeg
I compose my photos to be 10x8 or larger so reducing the image size to post here means that detail and impact is lost.
Below is a picture that I took (yesterday) where every hair can be seen on the face and the reflection in the eye, yet published here gives mush in comparison. Clicking on the hyperlink gives a better version, but still limited - yes for this picture I have 4 flash working on it (like a small flash gang)
When posting pictures here we have to take this into account
Attachments
Upvote
0
Yup, as I mentioned ages ago: different lighting due to sun sinking lower causing flare, & yet utterly irrelevant to the conclusions of this comparison, as I've already proven.
'Laughable' is more your denial of a study that shows exactly what most other studies have already shown... without doing your own tests. For example, the one psolberg linked to below...
Yup, thank you psolberg!
Upvote
0
- Status
Similar threads
