Review: Canon RF 16mm f/2.8 STM by Christopher Frost

[neuroanatomist]

Thanks! Better to take the time and get results you're happy with.

I don't really care if I'm happy with the results or disappointed by them, as long as I'm confident that they were generated in a robust and unbiased manner.

 

[LogicExtremist]

I'm not a lens tester, and have little interest in photographing flat images from a very short distance away with an ultra wide angle lens. I prefer to use a longer lenses for that sort of work. Nor do I have much interest in what an image looks like at 400% when viewed from a foot or so away on a large computer monitor.

With the above in mind, here's a brief summary of my limited still photography use of the RF 16mm f/2.8 on my EOS RP (so a budget lens on a low [relatively] cost body). I also should note that my comments are based on results from SOOC JPGs and RAW shots processed in Canon DPP (both of which perform lens corrections automatically).

In photos where the closest object to the lens is more than a few feet away, high magnifications appear to show pixelation becoming an issue by the time image quality becomes obviously soft. At lower magnifications like 100%, photos are reasonably sharp anywhere in the frame.

For very close-up images only inches away, differences between center and corner sharpness become more apparent. I find it very challenging to capture dynamic UWA shots that draw the viewer into the photo, images where some of the objects are often only inches away. Finding a good balance of sharpness/depth of field has always proven the most difficult task in capturing those images for me. And finding that good balance generally overshadows limitations of corner sharpness in an UWA lens (for me).

In photographing interior rooms in a house, most objects are more than a few feet away. So most things are rather small as captured with a 16mm lens. And unless the photos are going to be viewed at more than 100% (which is unlikely in most cases) everything will appear acceptably sharp from the RF 16. If you insist on zooming in to 400% and trying to read an open book on the coffee table 10' away, you are probably going to be disappointed.

If capturing outdoor landscapes where most everything of interest is 20+' away, the small size of those objects makes it nearly impossible to capture great detail of those objects, especially with a $300 lens on a 26MP body. So the inability to capture fine detail is going to be driven much more by the small size of the objects than a lack of lens sharpness. Even objects 20+' away in the center of the image where a lens is sharp are going to be too small to capture stunning detail.

I have no experience with astrophotography, so I won't comment on that. And I'm not into video, so I won't address that use either. As far as general use, or hiking/travel, I suppose the lens would work OK. But unless there is an object of interest that one can get real close to, or on top of, the picture from a UWA lens is likely to be boring. So for most of those occasions, I consider a UWA is a poor choice. Using a UWA lens one can surely "get it all in", but in most cases everything will probably be so small the viewer will be left wondering what the subject is (because everything is so small).

I'll also comment on the use of filters. The lens is threaded for 43mm filters. The standard 43mm UV filter I have on the lens does not appear to introduce any additional corner shading. I don't have a standard 43mm Grad or Polarizer to try, so I don't know if they will cut the corners or not.

If you are thinking about using step-up rings, be careful. There is a small lip on the lens front that extends forward beyond the retracted position of the lens. Any step-up ring 58mm or greater attached directly to the lens will hit that small lip when the lens retracts in the off position. Use of step-up rings in front of an attached UV filter negates that issue, but many photographers don't like to stack filters. A standard thickness 58mm polarizer or similar filter may shade the corners a bit when stacked with a standard 43mm UV filter. But a standard thickness 67mm polarizer or similar filter does not appear to shade the corners when stacked with a standard 43mm UV filter.

The use of "thin" filters obviously can impact the above in a positive way, but I don't know from experience how much of an improvement they might offer.

Overall, the lens is very handy and quite easy and enjoyable to use. Those wanting to photograph printed material a few inches away and/or enjoy searching images at 400% will probably be disappointed in the corner sharpness. Those taking pictures of most other things in typical ways will probably be happy with the lens.

Thanks, this is probably one of the most helpful and informative posts so far on the RF 16mm f/2.8.

I too have little interest in photographing flat images at any distance with any lens. More interested in real life subject matter viewed at 100% max and glad that's what you shared info on. The Canon RP at 26MP is close to the R and more than the R6, so that gives some idea of the expected resolving power of the lens at the centre.

The observations about lens filters you've provided are extremely handy, as many people like to use some sort of filter with landscape photography. Good to know standard filters don't cause mechanical vignetting and that using a filter under a step-up ring avoids mechanical inteference or the lens barrel when retracted. All UWA lenses can have issues with circular polarising filters, regardless of price, it's just an optical thing where dark patches can appear in certain positions.

You're right, any UWA lens is a specialised lens, and because of the wide field of view, subjects by necessity appear small, and cover less pixels on a sensor, so less details are captured. Most reviewers describe it as a great, affordable, carry-around 'just in case' UWA lens for photographers to add to their camera bag. Appreciate you sharing your experience!

For video, I've noticed from reviews that all UWA lenses can give wobbly "jello" corners in cameras with IBIS, so it's best switched off. Walk and talk style vlogging is way too bouncy without stabilisation though, so some reviewers recommend using the RP or R and their digital image stabilisation instead.

 

[LogicExtremist]

I don't really care if I'm happy with the results or disappointed by them, as long as I'm confident that they were generated in a robust and unbiased manner.

That's what I meant, that you're happy with the objectivity of the results, not the aesthetics lol!

 

[neuroanatomist]

A DPReview subscriber (crusliq) posted some images taken with the RF 16mm on an EOS R. Although not shot at infinity they give a sense of the degree of correction undertaken to remove barrel distortion. All @ f/2.8.
The first is the uncorrected image 6720 x 4480 pixels
In the second I applied +100 distortion correction in Lightroom.
The third is cropped such that the full vertical image is retained. This is 4994 x 3329 pixels or 16.6 MP, compared to 30.1 MP for the uncorrected image.
While this may be okay for an R5, the R6 is left with a little more than 11MP before the image is upscaled. I presume that this also applies for in-Camera corrections.

View attachment 201054View attachment 201055View attachment 201056

I'm in the midst of processing a series of shots taken with the RF 14-35/4, and it appears to me that the above test is flawed. When performing a manual correction for barrel distortion, dragging the ACR slider to +100 does not equate to correcting 100% of the distortion. Rather, the slider is merely an arbitrary scale of 100 units of correction in the (+) direction for correcting barrel distortion or in the (–) direction for correcting pincushion distortion. In other words, +100 is likely too much correction. In the case of the 14-35mm at 14mm, around +40 does a reasonable job of correcting the rather severe barrel distortion.

The other point is that it's a reasonable job. Manual distortion correction assumes the characteristic of the distortion is linear (along the image radial), and that's almost never going to be the case. So, if you drag the distortion correction slider to the point where horizontal lines in the corners of the scene appear horizontal in the image, then there is residual barrel distortion of lines in the mid-frame. Conversely, if you drag the distortion correction slider to the point where horizontal lines in the mid-frame of the scene appear horizontal in the image, pincushion distortion of lines in the corners is introduced. Basically, manual distortion correction is applying a linear process to a non-linear aberration.

That's where lens profiles come into play. Applying the lens profile for the 14-35/4 in ACR results in horizontal lines in both the corners and the mid-frame appearing horizontal in the image, i.e. the profile includes a non-linear correction for distortion. The same can be achieved in Photoshop using a warp transform, for example. Doing so does 'stretch' the corners of the image, so in that sense there are 'fake pixels' being created, but it allows geometric correction without the need to crop then upscale the image. In the case of an even more complex distortion (e.g. mustache) a warp-type transformation is really the only way it can be corrected.

More to come.....

 

[LogicExtremist]

I'm in the midst of processing a series of shots taken with the RF 14-35/4, and it appears to me that the above test is flawed. When performing a manual correction for barrel distortion, dragging the ACR slider to +100 does not equate to correcting 100% of the distortion. Rather, the slider is merely an arbitrary scale of 100 units of correction in the (+) direction for correcting barrel distortion or in the (–) direction for correcting pincushion distortion. In other words, +100 is likely too much correction. In the case of the 14-35mm at 14mm, around +40 does a reasonable job of correcting the rather severe barrel distortion.

The other point is that it's a reasonable job. Manual distortion correction assumes the characteristic of the distortion is linear (along the image radial), and that's almost never going to be the case. So, if you drag the distortion correction slider to the point where horizontal lines in the corners of the scene appear horizontal in the image, then there is residual barrel distortion of lines in the mid-frame. Conversely, if you drag the distortion correction slider to the point where horizontal lines in the mid-frame of the scene appear horizontal in the image, pincushion distortion of lines in the corners is introduced. Basically, manual distortion correction is applying a linear process to a non-linear aberration.

That's where lens profiles come into play. Applying the lens profile for the 14-35/4 in ACR results in horizontal lines in both the corners and the mid-frame appearing horizontal in the image, i.e. the profile includes a non-linear correction for distortion. The same can be achieved in Photoshop using a warp transform, for example. Doing so does 'stretch' the corners of the image, so in that sense there are 'fake pixels' being created, but it allows geometric correction without the need to crop then upscale the image. In the case of an even more complex distortion (e.g. mustache) a warp-type transformation is really the only way it can be corrected.

More to come.....

I did bit of a search and fount that Dustin Abbott's review of the Canon RF 14-35mm F4L IS USM lens supports your findings:

"The distortion is extremely heavy; it required a +34 to achieve this manual correction, and it is also is not linear, so that results in a “mustache” pattern left behind. The standard correction profile obviously does a much cleaner job on the JPEG, and will on RAW images as well when that profile arrives."

Canon RF 14-35mm F4L IS USM Review - DustinAbbott.net

Photographer Dustin Abbott shares a thorough review of the Canon RF 14-35mm F4L IS USM wide angle zoom for Canon EOS R mirrorless.

dustinabbott.net

I've seen so many reviews and forum threads where people have justifiably complained about the lack of lens correction profiles in their favourite post-processing software package, only to be berated by some clueless know-it-all who claims they should just make their own lens profiles and it just took them five minutes to make one. They're missing the fact that the distortion is non-linear, and it wouldn't take Adobe one year to come up with a lens profile for the RF 50mm f/1.8 (which has much lower distortion than an UWA lens) if it was that easy!

 

[koenkooi]

[..] I've seen so many reviews and forum threads where people have justifiably complained about the lack of lens correction profiles in their favourite post-processing software package, only to be berated by some clueless know-it-all who claims they should just make their own lens profiles and it just took them five minutes to make one. They're missing the fact that the distortion is non-linear, and it wouldn't take Adobe one year to come up with a lens profile for the RF 50mm f/1.8 (which has much lower distortion than an UWA lens) if it was that easy!

You can make non-linear lens profiles your self, that's the whole point of using the pattern. I also don't believe you can conclude that it took Adobe a year because it's hard, it took a year because that's how Adobe prioritized it, regardless of difficulty.

 

[LogicExtremist]

For anyone who may be unfamiliar with what mustache distortion is, I found a good description here:

Lens Distorsions: Barrel, Pincushion & Mustache Distortion - Digital Photography Live

digitalphotographylive.com

"Mustache Distortion contains both barrel and pincushion. At the centre of an image a barrel distortion appears and gradually turns in a pincushion shape at the edge of the same image. Probably the worst kind of lens distortion. Both the barrel and pincushion distortions are quadratic. This means that their distortions are proportional to the square of their distance from the centre. But in mustache distortion in the centre it is quadratic due to the barrel being the dominant distortion, but at the edges it becomes quartic (to the power of 4) as both barrel and pincushion distortions are visible."

The B&H website also explains this quite well:

"Mustache distortion, which is most commonly found in less expensive wide-angle lenses, causes straight lines to curve both inward and outward as they crisscross the horizontal and vertical planes of the photograph. While barrel and pincushion distortions, which are also known as “radial distortions,” can usually be corrected post-capture in Photoshop, Lightroom, or other photo-editing applications, mustache distortions are more difficult to correct post-capture."

FAQ: Wide-Angle Lenses

This is an FAQ post about wide-angle lenses.

www.bhphotovideo.com

An accurate lens corection profile made specifically for the lens can fix this kind of distortion.

 

[LogicExtremist]

You can make non-linear lens profiles your self, that's the whole point of using the pattern. I also don't believe you can conclude that it took Adobe a year because it's hard, it took a year because that's how Adobe prioritized it, regardless of difficulty.

I'm sure that Adobe aren't just doing what customers can do in a few minutes, there's no logical reason to assume that. To assume that they either use the same tool and do it the same way as a customer, and not any other way, would be an example of the logical fallacy of false dichotomy, because it's totally possible (and most likely) that they use something more precise, which takes far longer to produce a more accurate lens profile.

I'd say that with a leading software company's product, used by many professionals worldwide with extremely high standards, that company would put considerable effort to get things the best they can. They wouldn't keep world-class pro photogs needlessly waiting for lens profiles if the job could be completed by a junior staff member in an afternoon. Maybe the tool they provide for customers is so they can get a 'close-enough' profile for their needs, for people who think that is 'good enough', to tide customers over until the company develops proper lens profiles, could that be the case?

Even if the task took a whole afternoon, I'm sure Adobe could spare a few hours of an employee's time over a six month period to get that done. Something tells me there's probably much more involved...

I'm not implying that it takes 12 months to complete the task of creating a lens profile for a lens with only mild distortion such as the RF 50mm f/1.8, that would be absurd. What I'm saying is that it doesn't take 5 minutes!

 

[Johnw]

I would also suggest along the same lines that this RF 16mm is a worthy addition to the 24-240mm all-purpose lens.

I just got this lens and I like it for what it is. I find this lens to be better than the 24-240 at least considering the comparable wide ends. The 24-240 has so much distortion optically at 24 that it doesn't even fully illuminate a FF sensor. This lens is better than that, but the optical distortion is still pretty bad before correction. Still for $300, I like this lens for what it is, basically a cheap option to play around with UWA shots, or for when you want to get at least a decent image with an ultra compact form factor. I wasn't expect pristine rendering from a lens at this price point. My choice UWA lens that I really want is the Sigma 12-24 which actually has less distortion at 12 than this lens has at 16, but at more than 5x the price, that shouldn't come as much of a surprise. Similarly the 11-24 or the forthcoming RF 10-24 would obviously blow this lens away in terms of quality, the Samyang 14 RF prime does as well for a cheaper price point.

 

[neuroanatomist]

The 24-240 has so much distortion optically at 24 that it doesn't even fully illuminate a FF sensor.

I don’t believe that’s true. At the wide end of the zoom range, mechanical vignetting means the lens cannot cover the entire FF image circle, but the wide end is actually wider than 24mm.

The 14-35/4L is similar. Strong barrel distortion and mechanical vignetting at the wide end, but that wide end is actually ~13mm. Canon’s correction of the distortion results in a 14mm FoV at the wide end. DxO PhotoLab’s correction of the distortion results in a 13.5mm FoV, a nice bonus.

 

[Johnw]

I don’t believe that’s true. At the wide end of the zoom range, mechanical vignetting means the lens cannot cover the entire FF image circle, but the wide end is actually wider than 24mm.

That may indeed be the case. My main point was just that the 24-240 has more distortion at 24 than this lens does at 16, though that's not really surprising considering a 10x zoom is going to have some hard design compromises whereas its much easier in a prime to optimize the optics to do at least a halfway decent job at a fixed focal length.

The 14-35/4L is similar. Strong barrel distortion and mechanical vignetting at the wide end, but that wide end is actually ~13mm. Canon’s correction of the distortion results in a 14mm FoV at the wide end. DxO PhotoLab’s correction of the distortion results in a 13.5mm FoV, a nice bonus.

Another reason I don't like the 14-35 for UWA applications. It's clearly better suited to the long end but most people looking for a UWA lens don't care as much about 35. Mediocre performance can be forgiven in a cheap lens like this 16 given it's price, but if I were to spend as much on the 14-35 as it costs for UWA I would want better rendering at the lower end.

 

[neuroanatomist]

Another reason I don't like the 14-35 for UWA applications. It's clearly better suited to the long end but most people looking for a UWA lens don't care as much about 35. Mediocre performance can be forgiven in a cheap lens like this 16 given it's price, but if I were to spend as much on the 14-35 as it costs for UWA I would want better rendering at the lower end.

Have you tried it? I thought the same thing, before I did. I found that at 14mm, the heavily digitally corrected corners of the 14-35/4 were not significantly different from the lightly and mostly optically corrected corners of the far more expensive 11-24/4 (and the latter wasn’t at the end of its range).

RF 14-35mm f/4L IS – Distortion Correction Testing

Mirrorless cameras lack an optical viewfinder, so looking through the lens is always an electronic process. This fact enables lens manufacturers to include consideration of digital optical corrections in their lens designs and make trade-offs accordingly – more optical aberrations can be...

www.canonrumors.com

 

[Johnw]

Have you tried it? I thought the same thing, before I did. I found that at 14mm, the heavily digitally corrected corners of the 14-35/4 were not significantly different from the lightly and mostly optically corrected corners of the far more expensive 11-24/4 (and the latter wasn’t at the end of its range).

That's great if the digital correction really is that good. And I admit I have not used that lens personally. Astro is usually my main use case for UWA and my previous experience with lenses that require heavy digital correction in the corners is that they generally turn star fields into mush there. The 10-24 is also on my long term list already and I can't see myself purchasing more than one UWA zoom from Canon. For roughly the same price as the 14-35 the Sigma 12-24 seems like a better value to me for now. Weight is really the only downside I see on the Sigma and most UWA shooting I use a tripod for anyway. I've also owned the Sigma 14 1.8 and 14-24 2.8 in the past and found them both superb so I've been wanting to try that other Sigma as well until Canon releases something better (likely the 10-24).

 

[neuroanatomist]

That's great if the digital correction really is that good. And I admit I have not used that lens personally. Astro is usually my main use case for UWA and my previous experience with lenses that require heavy digital correction in the corners is that they generally turn star fields into mush there. The 10-24 is also on my long term list already and I can't see myself purchasing more than one UWA zoom from Canon. For roughly the same price as the 14-35 the Sigma 12-24 seems like a better value to me for now. Weight is really the only downside I see on the Sigma and most UWA shooting I use a tripod for anyway. I've also owned the Sigma 14 1.8 and 14-24 2.8 in the past and found them both superb so I've been wanting to try that other Sigma as well until Canon releases something better (likely the 10-24).

It surprised me. I don't shoot much astro, have a Samyang/Rokinon 14/2.8 that i have used for that. I doubt I'd buy an RF 10-24, there's a significant advantage in adapting the EF 11-24 with the drop-in filter (granted, not for astro).

 

[Johnw]

Filters are definitely relevant for astro. I've imaged the sun with a 20 stop ND in front of the EF-RF adapter variable ND.

Generally though filters are more relevant for astro at telephoto lengths than UWA. Still I felt the same as you did on the 11-24 about adapting the Sigmas partly for that reason.

 

[tron]

I have the lens too and I’m quite happy for my use cases.


For selfie I don’t care about corner sharpness, but all other aberrations are perfectly controlled.


For astro I won’t use the lens corrections but light frames to deal with vignetting .. no one cares about straight lines for Milky Way photography so I actually get sharpness and contrast where I need it. Coma is fairly okay, worst case I have to stop it down to f4 .. still better then the EF 16-35mm f4L for Milky Way I reckon, it had more aberrations traded for better corner sharpness.


Big plus for the smooth bokeh and close focusing distance!

most annoying is actually the lack of weather proofing, despite its small size I’d like to leave it out for Astro for a couple of hours and don’t have to take it down during a small drizzle.

Where do you base that? I strongly doubt it! Not only from my 16-35 4L IS experience in landscape and astro but also from lenstip,com full review. Sorry but I do not accept your statement. You can like the RF16 all you want or you can pursuade yourself - I do not care - but what you say about 16-35 does not hold.