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

[LogicExtremist]

Both the RF 16 and RF 14-35 have substantial uncorrected geometric distortion. After correction, corner sharpness of the RF 14-35 is very good, while the RF 16's is only fair. This tells me that distortion correction alone does not substantially reduce sharpness.

Well, you can't draw that conclusion because its not an equal comparison. The better optics in the L-series lens would actually resolve a higher level of detail. There are reports that the images from software corrected lenses look worse on lower megapixel cameras, such as the R6 compared to the R5, because a higher megapixel sensor will capture much more data for the same part of an image, When you start with more accurate data, or more data, a fixed amount of image degradation due to software correction wont mess us the photo as badly, that's all we're seeing here.

The simple way to objectively test how much the software correction affects the image quality of the lens is to compare the RF 14-35mm f/4 lens to itself, by checking what happens to the periphery and corners when the lens is stopped down. Typically, lenses get sharper edge to edge when stopped down to a narrower aperture. Try the lens comparison below, which compares the sharpness of the lens at f/4 to the sharpness at f/8:

Canon RF 14-35mm F4 L IS USM Lens Image Quality

View the image quality delivered by the Canon RF 14-35mm F4 L IS USM Lens using ISO 12233 Resolution Chart lab test results. Compare the image quality of this lens with other lenses.

www.the-digital-picture.com

What is curious is that stopping down just reduces the vignetting, brightening up the image a bit, but does not improve sharpness at all!

Why? Like I said earlier, you can't stop down to sharpen pixels that were never captured in the first place but were added in after the fact.

 

[BBarn]

Well, you can't draw that conclusion because its not an equal comparison. The better optics in the L-series lens would actually resolve a higher level of detail.

The RF 14-35 is sharper in the corners after correction because it was sharper in the corners before correction. The RF 16 isn't as sharp in the corners after correction because it wasn't as sharp in the corners before correction. It's that simple. Distortion correction does not have a major effect on corner sharpness.

 

[LogicExtremist]

The RF 14-35 is sharper in the corners after correction because it was sharper in the corners before correction. The RF 16 isn't as sharp in the corners after correction because it wasn't as sharp in the corners before correction. It's that simple. Distortion correction does not have a major effect on corner sharpness.

Totally agree 100% with you on the first statement. Distortion correction doesn't have a major effect on corner sharpness? Those two lens test tools I posted for each of these lenses would suggest otherwise in my eyes, but if the image degradation doesn't look that bad to you, I can't argue with that...

For me, that level of software correction might be okay on the budget lenses when used for specific purposes such as the the 16mm f/2.8 STM used a a vlogging/video lens and the RF 24-240mm F4-6.3 IS USM as a lightweight do-everything travel lens, but not on a high-end L-series lens at that price.

As they say, YMMV (your mileage may vary), the satisfaction level might be different for every person.

 

[LogicExtremist]

Okay folks, found a decent review of this lens here -

Incidentally, one of the viewer comments on this video describes some testing they did with the RF 16mm f/2.8 STM and it's very telling, see below:

"In Lightroom, I used the EF 16-35 F2.8 L II's profile and manually slide the distortion to 3/4, then in the manual correction tab, distortion correction is set to 100. After cropping to closely match the composition of the same photo's in-camera processed JPG, I found out the crop results in a 29% loss of the sensor's real estate (21.42 mp vs 30.11 mp on the EOS R), which means the true resolution of the sensor is also lost. If you shoot full-sized jpgs in camera, it means 29% of the pixels are results of computation. It's hilarious when you think of it. Canon's definitely cheating in these newer lightweight ultra wide-angle lenses (the 14-35 F4 included) to digital corrections and fake resolutions. Video was fine as there's in camera correction. However, you are still giving up a lot of pixels for the same reason. I also think the lens will be unusable on other cameras that use RF mounts, like the RED v-raptor, as there will be no correction data applied to the footage. It's essentially an uncorrected 13mm or 14mm lens, faked into a 16mm by warping and cropping the image. I do not agree with this approach at all. I would rather see a proper but much bigger and more expensive design"

This is what I was suspecting...

The reviewer described it as a decent video lens, and suggested that it would make an excellent 24mm equivalent RF mount APSC lens, if Canon ever does release a crop sensor body, because it would only use the sharper centre portion of the lens.

Not sure why some people are a bit reluctant to dig a bit deeper to get a better understanding of what's going on.

 

[koenkooi]

Okay folks, found a decent review of this lens here -

Incidentally, one of the viewer comments on this video describes some testing they did with the RF 16mm f/2.8 STM and it's very telling, see below:

"In Lightroom, I used the EF 16-35 F2.8 L II's profile and manually slide the distortion to 3/4, then in the manual correction tab, distortion correction is set to 100. After cropping to closely match the composition of the same photo's in-camera processed JPG, I found out the crop results in a 29% loss of the sensor's real estate (21.42 mp vs 30.11 mp on the EOS R), which means the true resolution of the sensor is also lost. If you shoot full-sized jpgs in camera, it means 29% of the pixels are results of computation. It's hilarious when you think of it. Canon's definitely cheating in these newer lightweight ultra wide-angle lenses (the 14-35 F4 included) to digital corrections and fake resolutions. Video was fine as there's in camera correction. However, you are still giving up a lot of pixels for the same reason. I also think the lens will be unusable on other cameras that use RF mounts, like the RED v-raptor, as there will be no correction data applied to the footage. It's essentially an uncorrected 13mm or 14mm lens, faked into a 16mm by warping and cropping the image. I do not agree with this approach at all. I would rather see a proper but much bigger and more expensive design"

This is what I was suspecting...

The reviewer described it as a decent video lens, and suggested that it would make an excellent 24mm equivalent RF mount APSC lens, if Canon ever does release a crop sensor body, because it would only use the sharper centre portion of the lens.

Not sure why some people are a bit reluctant to dig a bit deeper to get a better understanding of what's going on.

So essentially: "I wish this cheap, light non-L lens was an expensive, heavy L lens." It's not like you can see from the outside how small and light it is.....

 

[LogicExtremist]

So essentially: "I wish this cheap, light non-L lens was an expensive, heavy L lens." It's not like you can see from the outside how small and light it is.....

That's not what the reviewer is saying! Did you even watch the review? He points out its pros and cons, and the small, compact pocketable size is virtually one of the things he mentions. He works as a photographer and he's pointing out the flaws which make it unsuitable for many of the traditional applications of a 16mm prime!

 

[gruhl28]

Well, you can't draw that conclusion because its not an equal comparison. The better optics in the L-series lens would actually resolve a higher level of detail. There are reports that the images from software corrected lenses look worse on lower megapixel cameras, such as the R6 compared to the R5, because a higher megapixel sensor will capture much more data for the same part of an image, When you start with more accurate data, or more data, a fixed amount of image degradation due to software correction wont mess us the photo as badly, that's all we're seeing here.

The simple way to objectively test how much the software correction affects the image quality of the lens is to compare the RF 14-35mm f/4 lens to itself, by checking what happens to the periphery and corners when the lens is stopped down. Typically, lenses get sharper edge to edge when stopped down to a narrower aperture. Try the lens comparison below, which compares the sharpness of the lens at f/4 to the sharpness at f/8:

Canon RF 14-35mm F4 L IS USM Lens Image Quality

View the image quality delivered by the Canon RF 14-35mm F4 L IS USM Lens using ISO 12233 Resolution Chart lab test results. Compare the image quality of this lens with other lenses.

www.the-digital-picture.com

What is curious is that stopping down just reduces the vignetting, brightening up the image a bit, but does not improve sharpness at all!

Why? Like I said earlier, you can't stop down to sharpen pixels that were never captured in the first place but were added in after the fact.

I agree that the way to objectively test how much the software correction affects the image quality of the lens is to compare the RF 14-35 to itself, or the 16mm to itself, but not by stopping down - by comparing the sharpness of the lens uncorrected with the sharpness after corrections. It's as simple as that; how does what ends up in the corners look before corrections vs. after corrections.

 

[gruhl28]

The RF 14-35 is sharper in the corners after correction because it was sharper in the corners before correction. The RF 16 isn't as sharp in the corners after correction because it wasn't as sharp in the corners before correction. It's that simple. Distortion correction does not have a major effect on corner sharpness.

I totally agree with the first part. As for how much distortion correction affects corner sharpness, the test is to simply compare the sharpness before vs. after correction. I'm not sure how much it affects sharpness, as I don't have 100% crops before and after correction, but someone who has the lenses can easily make this comparison.

 

[LogicExtremist]

I agree that the way to objectively test how much the software correction affects the image quality of the lens is to compare the RF 14-35 to itself, or the 16mm to itself, but not by stopping down - by comparing the sharpness of the lens uncorrected with the sharpness after corrections. It's as simple as that; how does what ends up in the corners look before corrections vs. after corrections.

Apologies, I don't quite understand because the original corners are cropped off the uncorrected image and therefore change position on the corrected image.

 

[gruhl28]

Totally agree 100% with you on the first statement. Distortion correction doesn't have a major effect on corner sharpness? Those two lens test tools I posted for each of these lenses would suggest otherwise in my eyes, but if the image degradation doesn't look that bad to you, I can't argue with that...

For me, that level of software correction might be okay on the budget lenses when used for specific purposes such as the the 16mm f/2.8 STM used a a vlogging/video lens and the RF 24-240mm F4-6.3 IS USM as a lightweight do-everything travel lens, but not on a high-end L-series lens at that price.

As they say, YMMV (your mileage may vary), the satisfaction level might be different for every person.

Not all aberrations are corrected by stopping down. You cannot assume that because corners are not improving when stopping down that that means that you are seeing constructed pixels that were never captured in the first place. I dont' know why so many people are missing the obvious - to see how much distortion correction affects sharpness, or resolution, or whatever, simply compare those in the uncorrected image with the corrected image.

 

[gruhl28]

Apologies, I don't quite understand because the original corners are cropped off the uncorrected image and therefore change position on the corrected image.

Look at the details that are in the corners after correction. Then look at that same part of the image before correction, the same details, which won't be all the way in the corners in this image. How do the sharpness, resolution, contrast compare?

 

[gruhl28]

Okay folks, found a decent review of this lens here -

Incidentally, one of the viewer comments on this video describes some testing they did with the RF 16mm f/2.8 STM and it's very telling, see below:

"In Lightroom, I used the EF 16-35 F2.8 L II's profile and manually slide the distortion to 3/4, then in the manual correction tab, distortion correction is set to 100. After cropping to closely match the composition of the same photo's in-camera processed JPG, I found out the crop results in a 29% loss of the sensor's real estate (21.42 mp vs 30.11 mp on the EOS R), which means the true resolution of the sensor is also lost. If you shoot full-sized jpgs in camera, it means 29% of the pixels are results of computation. It's hilarious when you think of it. Canon's definitely cheating in these newer lightweight ultra wide-angle lenses (the 14-35 F4 included) to digital corrections and fake resolutions. Video was fine as there's in camera correction. However, you are still giving up a lot of pixels for the same reason. I also think the lens will be unusable on other cameras that use RF mounts, like the RED v-raptor, as there will be no correction data applied to the footage. It's essentially an uncorrected 13mm or 14mm lens, faked into a 16mm by warping and cropping the image. I do not agree with this approach at all. I would rather see a proper but much bigger and more expensive design"

This is what I was suspecting...

The reviewer described it as a decent video lens, and suggested that it would make an excellent 24mm equivalent RF mount APSC lens, if Canon ever does release a crop sensor body, because it would only use the sharper centre portion of the lens.

Not sure why some people are a bit reluctant to dig a bit deeper to get a better understanding of what's going on.

If those numbers are accurate, and it is actually cropping out 29% of the image, that is a significant loss of resolution. That statement was not in the video, though, it was in a comment from someone else, so we don’t know if it is accurate or not.

 

[LogicExtremist]

Look at the details that are in the corners after correction. Then look at that same part of the image before correction, the same details, which won't be all the way in the corners in this image. How do the sharpness, resolution, contrast compare?

You'd probably need to crop the uncorrected to match the size of the corrected image in that case to make the comparison a bit easier visually, yeah that makes sense!

 

[LogicExtremist]

If those numbers are accurate, and it is actually cropping out 29% of the image, that is a significant loss of resolution.

That would definitely be a cause for concern, even more so for photos than video. It would be interesting to see how much those software-corrected images would hold up to extensive post processing like what is usually done with landscape photos, to see how the details hold up or whether the images begin to fall apart and start producing artifacts more readily.

 

[Random Orbits]

That would definitely be a cause for concern, even more so for photos than video. It would be interesting to see how much those software-corrected images would hold up to extensive post processing like what is usually done with landscape photos, to see how the details hold up or whether the images begin to fall apart and start producing artifacts more readily.

It depends on what you're comparing this to. It's a $300 lens. Yes, something like the RF 15-35 f/2.8L IS would be better, but that is over 2k. For absolute IQ, I'd choose the L lens, but this might come in handy if weight is a primary factor or for snapshots where the subject is in the center. If I'm trying to save weight, I might opt for the ultrawide zoom, the nifty fifty and a telephoto. Some my opt for this RF 16, a standard zoom and a telephoto. I do like that it's the same size as the RF 50 f/1.8.

 

[BBarn]

If geometric distortion correction caused a substantial reduction in sharpness, the corners of the RF 14-35 (which requires substantial geometric correction) wouldn't be sharp. But they are. Easy peasy. Geometric distortion correction can be accomplished with relatively little negative impact to image sharpness. If you want to know why the RF 16 isn't sharp in the corners, look elsewhere.

 

[gruhl28]

That would definitely be a cause for concern, even more so for photos than video. It would be interesting to see how much those software-corrected images would hold up to extensive post processing like what is usually done with landscape photos, to see how the details hold up or whether the images begin to fall apart and start producing artifacts more readily.

It would also be interesting to know whether jpg's and Raws processed in DPP lose 29% of their pixels, or whether Canon is actually doing something to "construct" additional pixels, whether that is simple interpolation or AI.

 

[BBarn]

EXIF data shows the corrected image sizes of both RAW and JPG (large) files taken with the RF 16mm f/2.8 match the full sensor pixel count.

 

[Aussie shooter]

For those doubting focus by wire with the 16/2.8, perhaps autofocus will work.

With my R5 I can autofocus on many stars with my 20/1.4 Sigma, 35/1.4 Tamron, 70-200/2.8 L IS, and my 31 year old 300/2.8 L. One needs to find a reasonably good star well above the horizon and it is a good idea to turn the lens to manual focusing once accurate auto focus has been achieved. I have accidentally pressed the focus button while astro photography was going on and had to start over again. Also, my M6 ii will autofocus the 22/2 on night skies.

I have the R6(which auto focuses at the same level as the R5) and absolutely get better focus by manually focusing while zoomed in ten times.

 

[neuroanatomist]

If I didn't have the EF lens, and I wanted the get a wide angle f/4 zoom, the 14-35mm is the only one available for the RF mount and I would consider it rather than invest in an old mount lens. The two considerations are value for money and image quality. If the IQ is suitable for your requirements, then it might be a matter of waiting till the Xmas or EOFY sales arrive!

I just sold my EF 16-35/4L IS.

The degree of cropping are reported for the 16/2.8 seems excessive, and given that it was an unsubstantiated comment on a review I take it with a large grain of salt. Regardless, to deliver the camera’s full resolution, the cropped images must be upscaled.

I also don’t find a comparison between two corrected profiles (Canon 14-35 vs Samyang 14) to be especially compelling. Distortion correction changes framing, and the degree to which distortion is corrected may differ between the profiles.

I am pretty sure that if Canon labels a lens 16mm or 14mm, the FoV delivered in the final images will be that. Keep in mind that focal length is specified at infinity focus. Since people testing for lens distortion aren’t using charts or walls large enough to fill the frame with the lenses focused at infinity, focus breathing must be considered. For example, the EF-S 18-200mm at the long end frames at ~150mm equivalent with a close subject, the EF 100/2.8L Macro frames like ~67mm at 1:1. The 16/2.8 probably has a fair bit of breathing, the 14-35 less so.

For me, an additional factor is the small size of the 14-35, compared to the 16-35/4 (with adapter). Also, the relevant comparison to me is both lenses at the wide end, so even if the 14-35 isn’t quite 14mm, it’s wider than 16mm, and smaller and lighter in my bag.

I ordered the RF 14-35, once it arrives I’ll run some comparisons, including vs my EF 11-24 for FoV.