Review: Canon RF 14-35mm f/4L IS USM

[tron]

So does this lens has the FOV of a 14mm lens at the wide end when it is corrected or uncorrected?

(Since it has significant barrel distortion at the wide end)

 

[neuroanatomist]

There are very few lenses that don't need any distortion correction right? Even good EF lenses need a bit of pin-cushion or barrel distortion correction. So since people are going to have those corrections turned on anyways, what is the big deal if the trade off for more distortion in an uncorrected lenses is the fact that I can use the same filters on the 14-35, as I do on my 24-105 f4, and 100-500??

For me (as stated earlier), the 'big deal' is that geometric distortion correction can accentuate the effect of volume anamorphosis. Ok, so it'seo not really a big deal for me personally, because i don't use DPP so the geometric correction won't be forced, and DPP can't correct volume anamorphosis anyway. Given that, the trade-off of a 14mm wide end and 77mm filters is a good one (or would be, were I planning to buy the lens – I'm not, since I'm good with the adapted EF 16-35/4 and EF 11-24/4).

 

[neuroanatomist]

So does this lens has the FOV of a 14mm lens at the wide end when it is corrected or uncorrected?

(Since it has significant barrel distortion at the wide end)

14mm FoV after correction. That's clear from Bryan's statement that the the QA-77 chart (based on ISO 12233) he uses was captured at 14mm using the framing marks on the chart (i.e. framed using the EVF or LCD where the geometric corrections are applied live), but the RAW image viewed in C1 showed a wider framing with heavy distortion.

 

[tron]

14mm FoV after correction. That's clear from Bryan's statement that the the QA-77 chart (based on ISO 12233) he uses was captured at 14mm using the framing marks on the chart (i.e. framed using the EVF or LCD where the geometric corrections are applied live), but the RAW image viewed in C1 showed a wider framing with heavy distortion.

Thanks!

 

[Skyscraperfan]

Faking the image in the EVF is even worse. I always complained that the image of the EVF is not real and the distortion correction in the EVF is even worse than my fears. Shouldn't there be an option to see the same image you would get with an OVF? Instead the EVF shows us was Canon wants us to see.

And will the future be that Canon lenses will be corrected in the EVF, but not third party lenses?

 

[SHAMwow]

I think ya'll need to just skip this lens...

 

[neuroanatomist]

As for other ppls comments about the loss of mm after correction, we already know from other corrected canon lenses that after correction it is 14mm while the images without correction is in fact wider to allow for the correction.

Maybe I'm missing something, but isn't there another aspect to this 'correction' beyond the geometric distortion? The output image file size is fixed for a camera, e.g the R5 outputs a 45 MP image, 8192 x 5464 pixels. When I take the 14mm RAW file viewed in C1 and manually correct the barrel distortion then crop to the area that is a 14mm FoV (the 3:2 framing marks on the chart), the resulting image is ~12.5% smaller, e.g. an R5 image post-correction would only be ~39.4 MP. To output a 45 MP file, that would need to be upscaled.

If true, then not only is there a loss of IQ in the corners where the image is stretched to correct the distortion, there's also an overall loss of IQ when the resulting image is upsampled to the camera's normal output size.

 

[AJ]

So if the image after correction has a 14 mm FOV and it sharp, then who cares about any distortion of the "raw" image? It doesn't really matter in the end.
The barrel distortion is the compromise you end up with given a flat 77 mm front element. The alternative would be a big heavy bulging front element ala Tamron 15-30.
I'm actually surprised by how little peripheral shading this lens has, given its design. This is not something that's emphasized in the TDP review.
In all this looks like a nice lens. I would have preferred a few extra mm at the long end (e.g. 16-40) but that's not what's on offer. I think the only big thing to complain about is the hefty pricetag.

 

[tron]

Maybe peripheral shading is corrected too (at least up to a point). I do not know it's a hypothesis.

 

[dilbert]

The distortion is shocking. While I hate the idea of an adapter, no IQ improvement is worth the field curvature with this lens. All of that extra sharpness will get lost in post.


Is this lens Dead On Arrival?
btw, Cr Guy, this page:

https://www.canonrumors.com/recently-discontinued-ef-lenses/ (google finds this first)

NEEDS updates from this page:

Canon officially discontinues a lot more EF lenses

Canon has officially discontinued more EF lenses. Some of them are already discontinued, others will be discontinued sometime in 2021. Below are the addition

www.canonrumors.com

 

[Skyscraperfan]

So this EVF correction does not happen with every RF lens? I just checked "The Digital Picture" and the RF 70-200 f/4 and RF 15-35 f/2.8 both show visible distortion. So at least DPP does not force any correction on those two lenses. So I would suspect that it also is not corrected on the viewfinder.

And what does "AI" mean in regards to the correction anyway? Doesn't the image just have to be squeezed until lines are straight?

 

[dilbert]

So if the image after correction has a 14 mm FOV and it sharp, then who cares about any distortion of the "raw" image? It doesn't really matter in the end.
The barrel distortion is the compromise you end up with given a flat 77 mm front element. The alternative would be a big heavy bulging front element ala Tamron 15-30.
...
In all this looks like a nice lens. I would have preferred a few extra mm at the long end (e.g. 16-40) but that's not what's on offer.

Note that you aren't asking for 14-40, rather 16-40. Seems that you and many others would have preferred a narrower lens at the wide end. That would probably have moderated the price and distortion.

I wonder if Canon decided to one-up Tamron with their 15-30 zoom and go wider, without the bulbous front element thinking that everyone would just suck up their crazy distortion as something to pay extra for and fix in post, because everything gets fixed in post.

 

[Kit.]

Maybe I'm missing something, but isn't there another aspect to this 'correction' beyond the geometric distortion? The output image file size is fixed for a camera, e.g the R5 outputs a 45 MP image, 8192 x 5464 pixels. When I take the 14mm RAW file viewed in C1 and manually correct the barrel distortion then crop to the area that is a 14mm FoV (the 3:2 framing marks on the chart), the resulting image is ~12.5% smaller, e.g. an R5 image post-correction would only be ~39.4 MP. To output a 45 MP file, that would need to be upscaled.

If true, then not only is there a loss of IQ in the corners where the image is stretched to correct the distortion, there's also an overall loss of IQ when the resulting image is upsampled to the camera's normal output size.

If - for some weird reason - you want a 45MP JPEG, then yes.

The funny thing, though, is that the lens is the least sharp (wide open) at 35mm, where it needs no distortion correction at all.

 

[neuroanatomist]

If - for some weird reason - you want a 45MP JPEG, then yes.

There would be no if about it. The R5 and DPP are going to output images that are 8192 x 5464 (for the R5, obviously other camera have different output sizes), even though they need to be upscaled to get to that fixed output size. What algorithm is Canon using for upscaling? How good is it? We won't know, it's a black box, but regardless upscaling is adding information to the file that was not in the original scene.

Note that I'm not positive that's the case – maybe this lens on an R5 will yield a 39.5 MP image at 14mm, a 41 MP image at 16mm, etc. I doubt it, but I really don't know. Since Canon does a similar 'trick' with the RF 24-240, it would be occurring there as well and I haven't heard anyone mention that the 24mm images from the superzoom are lower resolution.

The 'workaround' is to use a different RAW converter, one that ignores or allows disabling of the default correction.

 

[Skyscraperfan]

If it is upscaled, it is either sharpened or filled with "AI" pixels. So if you sharpen it again later, it is sharpened twice, which you should avoid.

 

[SNJ Ops]

That distortion at 14mm is very very surprising and not in a good way..

 

[Skux]

I really thought the "we'll fix it in post" mentality would be limited to the cheaper RF lenses.

 

[aceflibble]

The distortion is shocking. While I hate the idea of an adapter, no IQ improvement is worth the field curvature with this lens. All of that extra sharpness will get lost in post.

Going by Bryan's results, there's no actual IQ improvement anyway. Sharpness is a dead-tie, each one varying slightly in different areas of the frame at different focal lengths, and fringing on the new lens is worse even aftrer corrections. Vignetting is also slightly worse, too, which usually corrects fine at low ISO but if you're shooting at medium or higher ISOs then that extra correction is going to bring out more noise; technically not a lens IQ problem, but an overall image IQ problem all the same.

Note that I'm not positive that's the case – maybe this lens on an R5 will yield a 39.5 MP image at 14mm, a 41 MP image at 16mm, etc. I doubt it, but I really don't know. Since Canon does a similar 'trick' with the RF 24-240, it would be occurring there as well and I haven't heard anyone mention that the 24mm images from the superzoom are lower resolution.

This sort of correction isn't just cropping away the edges and presenting a lower res file, no. Some parts of the image are being squashed down while others are being stretched out. The final image resolution out of the camera is always the same.
What is being reduced is the optical resolution, since it is not physically possible for distortion to be corrected without lowering optical resolution, even if you keep the file resolution the same. So the file's pixel count may still be the full 45mp, but the optical resolution may only be 40mp, stretched out to 45mp. (And actually optical resolution is never that high to begin with, so really it's probably more like 30mp of optical resolution stretched out over a 45mp file.)

The 24-240mm is an interesting case because the non-linear distortion means the center of the image really doesn't need much correcting if you're focused more than a few feet form the lens. If you don't want a 3:2 file and you focused, say, six or seven feet away or more, then you can just crop the sides a little bit and you have a roughly 22mm shot which is much sharper than the corrected 24mm; Canon seem to over-correct the 24mm quite significantly and bring it to more like 25mm, possibly to account for the extra distortion the lens suffers from once you bring it close to minimum focus. It could be that this 14-35 is the same kind of case and perhaps a higher optical resolution could be retrieved from it if a squarer crop is acceptable and depending on the focus distance.


Still, I argue these are not things that anyone paying £1750 should have to be thinking about, anyway. The 24-240 gets away with it because it's a plastic superzoom made to be an hobbyist's jack-of-all-trades, and priced accordingly. This L lens is not advertised or sold as such and nobody dropping this kind of premium—again, 80% higher than the EF equivalent new, let alone third-parties—should be having to have these conversations.

 

[manwithafrotto]

Personally I can't wait to get this lens to chase my kids around with.. a lot of haters here I see. Might not be for you, and that's OK. For now I'll stick with the R6, 14-35f4, 50mm1.2, 100mm 2.8, and 100-500. I'd love to add a 24 or 28 1.2 then I think I'll be done buying for a while. Until the next 5 series body comes out I guess.. oof

I'm glad I unloaded all my old canon 5D3 and all EF Glass a few years ago and tried Sony for a while. Sold all of that and have switched back to Canon RF, and I don't see myself going anywhere else. I'm sticking with the amazing RF mount and lenses.

 

[aceflibble]

I've read the review and I have a lot of thoughts about this lens...and a lot of skin in the game, as I have TWO of them pre-ordered.

This is a very small lens. So small, in fact, that I am willing to forgive some of these perceived flaws because it will bring down the size/weight of Canon's ultra-wide lens options tremendously. If you don't value size and weight, then why are you even considering an f/4 lens? Go buy the f/2.8 and stop acting like anything slower than f/2.8 has no place in your bag at any price. LOL

I'm desperate for this lens. I am DONE adapting my EF ultra-wide and I'm eager to get to work with this thing. Is it perfect? No...and the RF 15-35 has some major flaws of its own - just like every other manufacturer's ultra-wide lenses. This, however, is the smallest and lightest I have ever seen, takes a 77mm filter and will be joining my camera bag by the end fo the month.

1) Good job reenforcing Canon's customer-unfriendly design practices and rewarding them for their marketing rather than their actual products. You are why we're having to have these conversations at all. Stop giving companies money until you know that they've actually delivered.
2) Tamron's 17-35mm f/2.8-4 is lighter and smaller than this, also uses a 77mm filter thread, and at least going by Bryan's results of this 14-35, the Tamron is optically better in the center and only very slightly worse in the furthest corners. The f/2.8 wide end is a bit soft but you can consider that aperture to just be a bonus, since the Canon lenses don't have it at all; stop it down to f/4 or beyond and the Tamron sharpens up significantly. (Or at least my copy did.) I've owned the Tamron, it wasn't my favourite wide-zoom by any means, but if size is important to you then it is your winner and optically it's very good. I did prefer the optics of the Canon 16-35mm f/4 and that's ultimately what I stuck with, but I appreciate not everyone wants to carry something that size around (and it is only size, not weight; the EF lens is only 80G heavier than this RF and only 150g heavier than the Tamron!) and/or pay that much.

However you spin it, however you want to try to justify it to yourself, you know that you've paid a gigantic, arbitrary premium, twice, for the sake of saving 80g (180g, if we're including the adapter), about ~1.5cm in length, and a clearly-flawed extra 2mm which is being completed via software rather than the glass you've ostensibly paid for. This is the problem—you are part of it—and this is what manufacturers (it's not Canon alone) need to not be rewarded for.