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

gruhl28

Canon 70D
Jul 26, 2013
178
65
The lack of a mirror allows mirrorless designs to get as close as they want to the film or sensor. In SLR lenses, that is not possible, so where some lens designs would naturally have elements where the mirror is, those designs must be compromised to allow space for the mirror. It's why the EF 35/1.4 was something like ten times the volume of the Leica M 35/1.4 despite not being appreciably better optically.
This point made me wonder whether mirrorless lenses could now be made as small as typical Leica rangefinder lenses, at least if they were manual focus. And how much additional size would be required to add autofocus?

Hmm, come to think of it, I guess the RF 16mm and the RF 50 mm f/1.8 are about as small as Leica lenses. They don't seem to have the same optical quality as the Leicas, though. A line of small primes with size comparable to rangefinder lenses and very good optical quality would be quite attractive. Even if they were manual focus, with focus peaking I might be willing to give up autofocus for small size, and with IBIS you wouldn't need stabilization in the lenses. I doubt Canon will do this, probably wouldn't sell enough without autofocus, but with small elements they could probably add autofocus without making the lenses too much bigger.
 
Oct 31, 2020
192
232
I got the 16mm F2.8 two weeks ago and I'm going back and forth about whether I should keep or not. I have until tomorrow to decided whether I'll return it or not.

On the one hand, it is a cheap lense (at least for RF mount) and it is great value for what it is. Some of the pictures I took look actually pretty decent (some even better than "decent") but the vignetting (nightskies/ nightscapes) drives me nuts and the (in my opinion sometime horrible) IQ in the corners. Just doesn't work for landscapes...

Since I don't think I'll get any good astro out of it and so I'll needing a more capable UWA leise anyway, so why keep it? But every time I attempt to send it back, I wanna keep it so torn about what to do. Maybe I'll just keep it until an 15-35mm or 14-35mm because affordable.


Or I'll just wait until 2034 when Sigma finally announces their first RF UWA leise :)
 

Czardoom

EOS RP
Jan 27, 2020
359
773
Trouble is: these electronic corrections come at a cost, namely decrease of sharpness due to "corner stretching". But I guess this is acceptable for a $300 lens, designed mostly for street and vlogging.
....
Yes, corner sharpness is less than the uncorrected lens, but...

Better than the EF 17-40mm f/4L lens.
Very close to the 16-35mm f/4L lens.

It seems more like the reality is that trying to get corner sharpness optically is not so easy to do and was rarely done to the extent that it can be done with digital correction without a high cost and much more weight due to much more glass.
 
  • Like
Reactions: Del Paso

LogicExtremist

Lux pictor
Sep 26, 2021
189
117
Yes, corner sharpness is less than the uncorrected lens, but...

Better than the EF 17-40mm f/4L lens.
Very close to the 16-35mm f/4L lens.

It seems more like the reality is that trying to get corner sharpness optically is not so easy to do and was rarely done to the extent that it can be done with digital correction without a high cost and much more weight due to much more glass.
Hmmm, not really, there's no comparison between a budget lens and the EF 16-35mm f/4 L lens, which gets reviews like this:

"If you are looking for an extremely sharp ultra-wide angle zoom lens, the Canon EF 16-35mm f/4L IS USM Lens should be at the top of your consideration list. This lens delivers prime-grade image sharpness right into full frame corners and it has the overall performance to match, including AF speed and accuracy." (https://www.the-digital-picture.com/Reviews/Canon-EF-16-35mm-f-4-L-IS-USM-Lens.aspx)

There is no substitute for optical correction when it comes to corner sharpness on a wide lens, because the corrections must add in pixels that aren't really there. Of course, people will start arguing about accuracy of the software algorithms that guess what the pixels should be, but that's a moot point. Some post processing operations add data that was never there and also result in a loss of sharpness. Objectively, we can simply look at the images produced after software correction and clearly see that the corners are soft, end of story. :)

There is a bit of a strange rationalisation process going on with some comments. Not a criticism of any particular person here, I wish people can just let this little budget lens be what it is, and appreciate it for what it can do, rather than fuss over what it isn't and cant do. Denying or playing down its limitations isn't really helpful either in my opinion.

We basically have a compact, lightweight and cheap 16mm lens with a fast aperture, which is sharp in the centre and soft in the corners. That means it's not good for serious landscape photography, can't be used for astro unless stopped down to f/4, but makes an awesome carry around lens where a wide angle is needed but the corners don't matter. Casual landscape and group photos, vlogging and talking head videos come to mind. Any other applications?

One thing that baffles me is the questionable logic justifying software corrections on more expensive lenses. Like all engineering compromises, they come at a cost, and they can be made in Canon's favour, NOT the buyers! They can skimp on the lens formula, use software corrections to compensate, and sell the lens for the same price as an optically corrected one, but make a bigger profit. That's one possibility. Is that's what's happening? You guys tell me, are software corrected RF lenses selling at a cheaper price than optically corrected equivalents if such exist? Or is it not an apples for apples comparison because they're just trading a bit more sharpness for a lot more distortion?

Each lens has a specific use case, and where such compromises are fine, then so be it, if not, then it's necessary to use a lens with decent enough optics to captures all the data required for the necessary image quality. otherwise it's a pointless game of justification and self deception like this video:(!
 
Last edited:

HMC11

Travel
CR Pro
Sep 5, 2020
63
59
Hmmm, not really, there's no comparison between a budget lens and the EF 16-35mm f/4 L lens, which gets reviews like this:

"If you are looking for an extremely sharp ultra-wide angle zoom lens, the Canon EF 16-35mm f/4L IS USM Lens should be at the top of your consideration list. This lens delivers prime-grade image sharpness right into full frame corners and it has the overall performance to match, including AF speed and accuracy." (https://www.the-digital-picture.com/Reviews/Canon-EF-16-35mm-f-4-L-IS-USM-Lens.aspx)

There is no substitute for optical correction when it comes to corner sharpness on a wide lens, because the corrections must add in pixels that aren't really there. Of course, people will start arguing about accuracy of the software algorithms that guess what the pixels should be, but that's a moot point. Some post processing operations add data that was never there and also result in a loss of sharpness. Objectively, we can simply look at the images produced after software correction and clearly see that the corners are soft, end of story. :)

There is a bit of a strange rationalisation process going on with some comments. Not a criticism of any particular person here, I wish people can just let this little budget lens be what it is, and appreciate it for what it can do, rather than fuss over what it isn't and cant do. Denying or playing down its limitations isn't really helpful either in my opinion.

We basically have a compact, lightweight and cheap 16mm lens with a fast aperture, which is sharp in the centre and soft in the corners. That means it's not good for serious landscape photography, can't be used for astro unless stopped down to f/4, but makes an awesome carry around lens where a wide angle is needed but the corners don't matter. Casual landscape and group photos, vlogging and talking head videos come to mind. Any other applications?

One thing that baffles me is the questionable logic justifying software corrections on more expensive lenses. Like all engineering compromises, they come at a cost, and they can be made in Canon's favour, NOT the buyers! They can skimp on the lens formula, use software corrections to compensate, and sell the lens for the same price as an optically corrected one, but make a bigger profit. That's one possibility. Is that's what's happening? You guys tell me, are software corrected RF lenses selling at a cheaper price than optically corrected equivalents if such exist? Or is it not an apples for apples comparison because they're just trading a bit more sharpness for a lot more distortion?

Each lens has a specific use case, and where such compromises are fine, then so be it, if not, then it's necessary to use a lens with decent enough optics to captures all the data required for the necessary image quality. otherwise it's a pointless game of justification and self deception like this video:(!
Thank you for a clear-minded take. Dustin Abbott just released his review of the RF 14-35 f4. It seems that the lens follows a similar approach as the RF 16mm in that it relies on software to do some fairly major corrections, and that actual image is wider than 14mm but crops in to remove the 'physical' vignetting. Despite this, it seems that the 14-35 is still very sharp at almost the entire range (except at 35mm) when wide-open, which makes it a viable landscape lens. While psychologically it doesn't feel good to know at the back of the mind that the image is fairly heavily software corrected, the end results may well be more than satisfactory. Overall, had it been priced closer to the EF16-35, the 14-35 could be a fairly compelling purchase for me. Given the significant price difference, I am somewhat hesitant. I may just have to bite the bullet and get it if this is going to be the 'standard' approach in order to get a lighter and smaller lens with very good final image output.
 

LogicExtremist

Lux pictor
Sep 26, 2021
189
117
Thank you for a clear-minded take. Dustin Abbott just released his review of the RF 14-35 f4. It seems that the lens follows a similar approach as the RF 16mm in that it relies on software to do some fairly major corrections, and that actual image is wider than 14mm but crops in to remove the 'physical' vignetting. Despite this, it seems that the 14-35 is still very sharp at almost the entire range (except at 35mm) when wide-open, which makes it a viable landscape lens. While psychologically it doesn't feel good to know at the back of the mind that the image is fairly heavily software corrected, the end results may well be more than satisfactory. Overall, had it been priced closer to the EF16-35, the 14-35 could be a fairly compelling purchase for me. Given the significant price difference, I am somewhat hesitant. I may just have to bite the bullet and get it if this is going to be the 'standard' approach in order to get a lighter and smaller lens with very good final image output.
Good point, it depends on which way Canon is planning to go with their RF lenses in terms of design strategy. The new mirrorless platform provides a convenient cover for the use of software correction, which can be used as a way of giving less and charging more for it.

I looked at thedigitalpicture.com review, where the author of the article points out the issue of AI generated fake detail in the corners, but then plays that down:

"Stretching the image out to the as-framed composition requires AI. Although today's image correction AI is very good, AI does not really know what the subject details were in the stretched areas, and calling the result fake detail does not seem untrue.

Does the strong distortion correction matter? Psychologically it does, and an image captured from a non-distorted lens can similarly be up-sized to even higher resolution using AI, potentially giving it an advantage. That said, did you notice any corner issues until this point in the review? Likely not substantial ones.

I need to get over the psychological issue of the geometric distortion correction, but otherwise, this lens is a stellar performer."


I'm not sure how he can reduce the issue of AI generated fake detail in the corners down to a psychological issue, when it's actually an image reproduction issue that is visible. Surely, some subjects won't be affected, and what's good enough might be subjective, but here we're talking about a crop of the full image and a smearing of details on an L-series pro-grade lens that is not cheap. I can accept that as a compromise on the handy little RF 16mm f/2.8 budget lens, but that's an entry level budget lens! :)

The Rf 14-35mm Ff/4 L should be usable for landscape like the EF 16-35 f/4 L, and no amount of arguments can justify missing corner details because the optics can't deliver them. What the review ignores is that images most likely will be post-processed, and with a combination of heavy vignetting and AI generated fake detail, the details might degrade further and start to fall apart with certain post-processing tasks.

The correction also crops the image, so for those who fuss over details and maximum pixels over the subject, the whole sensor is not used, and this might not be much of a pixel loss percentage-wise, but it's nevertheless there.

So what do we get for the more expensive Rf lens compared to its EF counterpart? Doesn't look like much in terms of image quality, even when the corner issue is ignored.

"In the image quality comparison, the two lenses show rather similar image quality overall. Both lenses have slight advantages in specific comparisons. For example, the RF lens produces sharper periphery image quality at 28mm, and the EF lens is sharper in the center of the frame at 35mm f/4. Performing similarly in this comparison reflects positively on the RF lens — this EF lens is a great performer. The RF lens has dramatically stronger geometric distortion and has stronger lateral CA. The EF lens shows stronger peripheral shading at 35mm f/4 and slightly less at narrower apertures in the wider half of the focal length range."

Lens image comparison tool for Rf 14-35 f/4 L and EF 16-34 f/4 L here.

(https://www.the-digital-picture.com/Reviews/Canon-RF-14-35mm-F4-L-IS-USM-Lens.aspx)

I'm kind of glad I got the EF 16-35mm f/4 almost new a while back, I can't see much benefit for me in upgrading. For others, if having the native RF mount, 175g less weight and 2mm extra at the wide end is important, then there isn't much choice, other than paying not that much more, relatively speaking, and stepping up to the 15-35mm f/2.8 which is a much better engineered lens.

We may be stepping into a whole new era of different compromises with the new RF range. We know they all have a heavy vignetting problem that's due to the RF mount design, and I can live with that. We've seen a molded plastic lens element included in the Rf 100-400, and some radical geometric distortion in a few lenses being rectified with software correction. Compromises imply that you give something to gain something else, and if we're not gaining much over the older EF lenses, but paying more, then we need to ask whether these are just cost cutting measures, or if we're gaining enough to offset what is being lost. Obviously, this will vary from lens to lens, and each needs to be considered against the buyers requirements. :oops:
 
Last edited:

neuroanatomist

I post too Much on Here!!
CR Pro
Jul 21, 2010
26,136
4,747
I may just have to bite the bullet and get it if this is going to be the 'standard' approach in order to get a lighter and smaller lens with very good final image output.
I think it is going to be the standard approach. Sony and Fuji have been doing this for years, even in high end lenses. Think about how many people on this forum have clamored for Canon to ‘keep up with the competition’. I guess they are.
 

LogicExtremist

Lux pictor
Sep 26, 2021
189
117
I think it is going to be the standard approach. Sony and Fuji have been doing this for years, even in high end lenses. Think about how many people on this forum have clamored for Canon to ‘keep up with the competition’. I guess they are.
Reminds me of the old saying "be careful what you wish for!" :confused:
 

Tom W

EOS R5
Sep 5, 2012
306
277
Interesting lens, and a good price, but I'm a little concerned at the excessive use of software corrections to make up for large lens weaknesses. I suppose it's a tradeoff for getting an ultra-wide that is this small and inexpensive, but I'd prefer less correction and better lens capability. I suppose that for $300, what I'm asking for is not necessarily possible.
 
  • Like
Reactions: SteveC

gruhl28

Canon 70D
Jul 26, 2013
178
65
I'm not sure this talk of "adding pixels" and using AI and creating things that weren't there is accurate. Does distortion correction really need or even use AI? Distortion correction has been around for years in LR, Photoshop, etc., long before anyone was talking about AI in post-processing. Distortion correction crops out some pixels and moves others, and maybe it averages and interpolates some to create new ones, I'm not sure. But it doesn't add things to the side of the frame, it actually crops out the sides, and it doesn't require AI to average two adjacent pixels to create a new one. Sure, this can magnify other aberrations and soften the image, but it's the final image that counts. If the corners are sharp even after distortion correction, as they are with the 14-35, then what difference does it make? The lens must be really sharp in the corners in order for the final pic to still be sharp after the stretching. Even with the corners of the 16mm not being as sharp as we would like, would even this level of quality have been possible in a lens this size and price without using distortion correction?

The statement made earlier, "The Rf 14-35mm Ff/4 L should be usable for landscape like the EF 16-35 f/4 L, and no amount of arguments can justify missing corner details because the optics can't deliver them" is wrong in my opinion. The RF 14-35 is not missing corner details - the corner detail is extremely good, just look at the pictures. This detail wasn't just "made up", it is because the lens is so sharp that even after stretching the result is still excellent. The designers sacrificed some optical distortion correction for excellent sharpness and other benefits, knowing that the distortion could be easily corrected in software, whereas a lack of resolution cannot be corrected - to make up for lack of resolution really would require AI. If AI were "creating" this detail then the16mm would have "sharp" corners also.
 
Last edited:

LSXPhotog

Automotive, Motorsports, Commerical, & Real Estate
CR Pro
Apr 2, 2015
554
581
www.diossiphotography.com
It's a shame that Canon has taken so long to release lens profiles for new RF glass to Adobe because the cat's out of the bag on this lens now...it's a lens that requires a truly hilariously amount of correction to create a presentable image. I'm finding it to be a brilliant lens for video and I purchased it exclusively as a gimbal ultra-wide for real estate video - which I'm confident it will do very well in. However...this lens is not a very good photography lens in regards to corner performance even after correction. It may be better to just crop the image even further to an 18-20mm to actually get good results. Things deteriorate rapidly at the corners of the frame - not a problem for video. It's just very sad that I was hoping for a better performance in the corners. I'll take it out a bunch next week and see if my opinion on it changed in regards to photography.
 

HMC11

Travel
CR Pro
Sep 5, 2020
63
59
Good point, it depends on which way Canon is planning to go with their RF lenses in terms of design strategy. The new mirrorless platform provides a convenient cover for the use of software correction, which can be used as a way of giving less and charging more for it.

I looked at thedigitalpicture.com review, where the author of the article points out the issue of AI generated fake detail in the corners, but then plays that down:

"Stretching the image out to the as-framed composition requires AI. Although today's image correction AI is very good, AI does not really know what the subject details were in the stretched areas, and calling the result fake detail does not seem untrue.

Does the strong distortion correction matter? Psychologically it does, and an image captured from a non-distorted lens can similarly be up-sized to even higher resolution using AI, potentially giving it an advantage. That said, did you notice any corner issues until this point in the review? Likely not substantial ones.

I need to get over the psychological issue of the geometric distortion correction, but otherwise, this lens is a stellar performer."


I'm not sure how he can reduce the issue of AI generated fake detail in the corners down to a psychological issue, when it's actually an image reproduction issue that is visible. Surely, some subjects won't be affected, and what's good enough might be subjective, but here we're talking about a crop of the full image and a smearing of details on an L-series pro-grade lens that is not cheap. I can accept that as a compromise on the handy little RF 16mm f/2.8 budget lens, but that's an entry level budget lens! :)

The Rf 14-35mm Ff/4 L should be usable for landscape like the EF 16-35 f/4 L, and no amount of arguments can justify missing corner details because the optics can't deliver them. What the review ignores is that images most likely will be post-processed, and with a combination of heavy vignetting and AI generated fake detail, the details might degrade further and start to fall apart with certain post-processing tasks.

The correction also crops the image, so for those who fuss over details and maximum pixels over the subject, the whole sensor is not used, and this might not be much of a pixel loss percentage-wise, but it's nevertheless there.

So what do we get for the more expensive Rf lens compared to its EF counterpart? Doesn't look like much in terms of image quality, even when the corner issue is ignored.

"In the image quality comparison, the two lenses show rather similar image quality overall. Both lenses have slight advantages in specific comparisons. For example, the RF lens produces sharper periphery image quality at 28mm, and the EF lens is sharper in the center of the frame at 35mm f/4. Performing similarly in this comparison reflects positively on the RF lens — this EF lens is a great performer. The RF lens has dramatically stronger geometric distortion and has stronger lateral CA. The EF lens shows stronger peripheral shading at 35mm f/4 and slightly less at narrower apertures in the wider half of the focal length range."

Lens image comparison tool for Rf 14-35 f/4 L and EF 16-34 f/4 L here.

(https://www.the-digital-picture.com/Reviews/Canon-RF-14-35mm-F4-L-IS-USM-Lens.aspx)

I'm kind of glad I got the EF 16-35mm f/4 almost new a while back, I can't see much benefit for me in upgrading. For others, if having the native RF mount, 175g less weight and 2mm extra at the wide end is important, then there isn't much choice, other than paying not that much more, relatively speaking, and stepping up to the 15-35mm f/2.8 which is a much better engineered lens.

We may be stepping into a whole new era of different compromises with the new RF range. We know they all have a heavy vignetting problem that's due to the RF mount design, and I can live with that. We've seen a molded plastic lens element included in the Rf 100-400, and some radical geometric distortion in a few lenses being rectified with software correction. Compromises imply that you give something to gain something else, and if we're not gaining much over the older EF lenses, but paying more, then we need to ask whether these are just cost cutting measures, or if we're gaining enough to offset what is being lost. Obviously, this will vary from lens to lens, and each needs to be considered against the buyers requirements. :oops:
As I don’t have the EF 16-35, I may need to do some mental gymnastics to persuade myself that this is worth going for . From ‘The digital picture’ review you shared, the distortion at 16mm seems not too bad. If so, I could take this as a RF 16-35 f4 lens, and that the premium over the EF version is for it being lighter, smaller, a tad sharper, better IS, and the control ring. Besides, with the EF launch price at $1199 in 2014 being about $1390 today, the ‘actual’ premium would be about $300. That may be easier to swallow .
 

SwissFrank

from EOS 1N to R
Dec 9, 2018
660
371
I'm not sure this talk of "adding pixels" and using AI and creating things that weren't there is accurate. Does distortion correction really need or even use AI? Distortion correction has been around for years in LR, Photoshop, etc., long before anyone was talking about AI in post-processing. Distortion correction crops out some pixels and moves others, and maybe it averages and interpolates some to create new ones, I'm not sure. But it doesn't add things to the side of the frame, it actually crops out the sides, and it doesn't require AI to average two adjacent pixels to create a new one. Sure, this can magnify other aberrations and soften the image, but it's the final image that counts. If the corners are sharp even after distortion correction, as they are with the 14-35, then what difference does it make? The lens must be really sharp in the corners in order for the final pic to still be sharp after the stretching. Even with the corners of the 16mm not being as sharp as we would like, would even this level of quality have been possible in a lens this size and price without using distortion correction?

The statement made earlier, "The Rf 14-35mm Ff/4 L should be usable for landscape like the EF 16-35 f/4 L, and no amount of arguments can justify missing corner details because the optics can't deliver them" is totally wrong. The RF 14-35 is not missing corner details - the corner detail is extremely good, just look at the pictures. This detail wasn't just "made up", it is because the lens is so sharp that even after stretching the result is still excellent. The designers sacrificed some optical distortion correction for excellent sharpness and other benefits, knowing that the distortion could be easily corrected in software, whereas a lack of resolution cannot be corrected - to make up for lack of resolution really would require AI. If AI were "creating" this detail then the16mm would have "sharp" corners also.
You put it pretty well, Gruhl28. But I had a few minutes so I wrote a bit more in-depth summary.

Here is how distortion correction works.

Pincushion zoom simply means that as you get away from the center, magnification slowly increases just a bit. Barrel means magnification slowly drops. Lenses with just a few groups will have one or the other, but in today's 10-group monsters, you actually may have a curve that wiggles up and down a bit.

The lens type and the zoom and focus info are communicated to the camera via the mount. In an ideal world, the camera would then have all the tables of information to figure out, for every ideal radius from the center in a totally rectilinear ideal photo, where that data would have ended up in the raw data.

Then you just do the following:

1. Make a second image buffer to hold the corrected image. For every pixel in it, calculate the angle from the center, and distance. In other words convert to polar coordinates with the origin in the center of the image.

2. Figure out, from the data you have on this lens, what pixels in the raw image that data actually ended up on. This will be the same angle, in the polar coordinates, and simply be a different radius from the center.

3. Make an average of those pixels, weighted by how close they are to where the pixel should have been.


That's it, pretty much. My example describes each pixel as a point, but it's possible you'd get more accurate results (albeit probably not visible even to pixel peeping) by treating them as rectangles or even rectangle-like shapes with slightly curved sides. You'd also be able to do 8 pixels at once utilizing symmetries in trig functions.

So what is the downside? Every pixel is now the average of typically four raw pixels. However we think of resolution linearly, and those four raw pixels are in a 2x2 array, so in fact we'd call this "half" the resolution.

And yet what lenses can really render 1-pixel-wide features, even to the edges or corners? The 50/1.2 might come close but I'm not sure it manages this. I doubt any zooms or ultra-wide-angles do, and the lens in question is nowhere even near doing so.

Let's say it can only render something 3 pixels wide in the corner. On your black background you see dark grey, light grey, dark grey, in a sea of black. Thanks to this geometric correction, that blurry feature is now... 4 pixels wide. You'd see dark grey, medium light grey, medium light grey, dark grey, in a sea of black. Even pixel-peeping, I think you'd be hard-pressed to see the difference, maybe not even on a test chart. (And I note no-one's offering test charts to show this supposedly deleterious effect.)

OK, so in lens design, it seems nearly anything can trade off for nearly everything else. By loostening one up, you can improve some, many, or even all of the others.

  • price
  • size
  • weight
  • environmental concerns of lead or lanthanum in the glass
  • sharpness
  • distortion
  • vignetting
  • every variety of aberration

It's a big win to let the camera do the corrections that are trivial in software and that can be done at nearly no cost to image quantity. Distortion and vignetting* are two of these. You can then have the lens do better work at things the software CAN'T do, such as sharpness, light weight and so on.

(*We're talking distortion not vingetting but it's even simpler to correct for. The lens's ID, zoom and focus info should suffice to let the camera locate lens-specific data saying exactly how much darker it is than it should be at every radius from the center. So, for every pixel in your image, simply increase the value as dictated by that data. Now, on the Canon R and RP, this CAN hurt image quality because the corners may be so dark that low values are now zero values, and the information there's been lost. We may also find that thanks to vignetting we actually have image data less than pixel max values where something by rights should be a blown-out highlight, and we blow that out when correcting for vignetting. However, the R5 and newer cameras are starting to have VERY nice dynamic ranges that make both of these problems less and less an issue. So, on an R5, the 50/1.2 wide open is 3 stops darker in the corner as always, but where the R would end up with value 0 in the raw image, the R5 will have a low number but at least something that can be worked with.)
 
  • Like
Reactions: gruhl28 and HMC11

SwissFrank

from EOS 1N to R
Dec 9, 2018
660
371
This point made me wonder whether mirrorless lenses could now be made as small as typical Leica rangefinder lenses, at least if they were manual focus. And how much additional size would be required to add autofocus?

Hmm, come to think of it, I guess the RF 16mm and the RF 50 mm f/1.8 are about as small as Leica lenses. They don't seem to have the same optical quality as the Leicas, though. A line of small primes with size comparable to rangefinder lenses and very good optical quality would be quite attractive. Even if they were manual focus, with focus peaking I might be willing to give up autofocus for small size, and with IBIS you wouldn't need stabilization in the lenses. I doubt Canon will do this, probably wouldn't sell enough without autofocus, but with small elements they could probably add autofocus without making the lenses too much bigger.

Correct. Optics-wise there's no difference betwen the rangefinder lenses of yore (such as my 35/1.4 ASPH which is on my R as we speak) and today's "mirrorless" designs.

Not only COULD you make such lenses, but you can buy a $20 adapter that will literally mount any Leica M lens ever made onto your RF body.

The quality difference is in part cost, or maybe it'd be accurate to say the cost difference is one part quality. But even the old-style Leica lenses are utterly crushed, at least in resolution charts, by the new line of APO ASPH lenses that are quite large compared to the old ones. (See 50/2 for instance.)

Yes, I've suggested several times on this forum that Canon "should have" about the following lines of lenses:

  • a "street" line, emphasizing image quality, reliability and small size over everything else. Maybe 35/2, 135/2.8 These would be the "Leica M type".
  • an "amateur" line, with fancy features like macro, etc., emphasizing feature set and price over everything else. 35/1.4, 135/2.0
  • a "pro" line, emphasizing image quality, reliability, plus generous specs: 35/1.2, 135/1.4
  • a "halo" line, may not even be readily available, but to get people talking: 35/1.0, 135/1.0, similar to the EF1200/5.6 etc.
 

SwissFrank

from EOS 1N to R
Dec 9, 2018
660
371
I swear, you must have been beaten by a Capitalist as a young kid! You and Senator Warren should just go get a room!

> The new mirrorless platform provides a convenient cover for the use of software correction, which can be used as a way of giving less and charging more for it.

Retail price of the product often has little to do with the cost of manufacturing it, and everything to do with the perceived value.

If the result of using software correction is a smaller, sharper lens, and people choose to pay more, then Cannon makes money.



> I looked at thedigitalpicture.com review, where the author of the article points out the issue of AI generated fake detail in the corners

If it says anything about AI, it is lying to you, and you should take everything else they say with a grain of salt. BTW I wrote another post just now that explains exactly how distortion correction works. All you need to be able to do is convert to and from polar coordinates, multiple the radius by a correction factor provided from a table or function based on the lens type, zoom and focus, and read out the pixels nearest the resulting point to figure out what any given pixel in the distortion-corrected image should be.



> I'm not sure how he can reduce the issue of AI generated fake detail in the corners down to a psychological issue

He's lying about it. It's not AI.



> when it's actually an image reproduction issue that is visible. Surely, some subjects won't be affected

Basically, fine details will be blurred by up to a half-pixel's worth. Since most lenses get nowhere near resolving anything like that sharp, this is nearly invisible.



> The Rf 14-35mm Ff/4 L should be usable for landscape like the EF 16-35 f/4 L, and no amount of arguments can justify missing corner details because the optics can't deliver them

The optics deliver the details fine. We don't know exactly what Canon traded off distortion for, but it could have been sharpness, for instance. It's totally believable that the distortion-corrected actual lens is sharper than a lens that was so tightly constrained to have zero distortion that they couldn't affordably give you a sharp image at all.



> What the review ignores is that images most likely will be post-processed, and with a combination of heavy vignetting and AI generated fake detail, the details might degrade further and start to fall apart with certain post-processing tasks.

You have to stop it with the goddamn AI fake detail BS. Go sit in a corner and breath deeply until you can talk rationally.

Meanwhile how does heavy vignetting degrade details? To answer my own question, a camera with limited dynamic range compared to the scene you need to capture may find some dark Zone II pixels going to Zone I, at which point they can't be brightened back up. But the R5 and future bodies will have lots of dynamic range so its not a problem. And even the R, with relatively modest dynamic range by today's standards, can lose three stops due to vignette correction and still have more dynamic range than a camera of say 2010 or so.



> The correction also crops the image, so for those who fuss over details and maximum pixels over the subject, the whole sensor is not used, and this might not be much of a pixel loss percentage-wise, but it's nevertheless there.

Red herring. You're not even losing 5% of your image here, and the fact you even ended up with a few extra pixels that will disappear in distortion correction is a side-effect of the fact that the lens designer chose to sacrifice easy-to-correct distortion for impossible-to-correct aspects such as size.

Think about it: if you had a rectilinearly perfect 14mm lens, you wouldn't have those pixels in the first place, and you'd be happy. So then a distorted lens captures the 14mm scene for you, and a few more pixels in the corners you didn't ask it for--and you're bawling that they're being snipped away??!!

Just shoot the initial image while using the center 95% of the finder. We lost more image area than that to slide mounts until 2002.



> So what do we get for the more expensive Rf lens compared to its EF counterpart? Doesn't look like much in terms of image quality, even when the corner issue is ignored.

Well first off, you do, if you're talking about the 14-35/4 vs. the 16-35/4. You get more image quality. And a substantial increase in zoom range. And far better autofocus, I'd say nice bokeh, lack of focus breathing, 2/3 the weight, and and and.

But the real reason I'd give you is that the EF version's value is going to plummet faster than the RF's. The price of a lens isn't the purchase price, it's the purchase price minus the sale price. Buy the RF used from a real human, and sell it when the follow-on model is announced, and it won't cost you much more than postage.

> We know they all have a heavy vignetting problem that's due to the RF mount design

Sigh, no, THINK FOR A SECOND. Every EF lens can be mounted on RF and none has an iota difference in vignetting. Ergo the mount has NOTHING to do with vignetting. However, lense designers can trade vignetting for practically anything else, and choose to. They can give you a far sharper lens, smaller, stronger, cheaper, or more features.

> Compromises imply that you give something to gain something else

Yes.
 

neuroanatomist

I post too Much on Here!!
CR Pro
Jul 21, 2010
26,136
4,747
The correction also crops the image, so for those who fuss over details and maximum pixels over the subject, the whole sensor is not used, and this might not be much of a pixel loss percentage-wise, but it's nevertheless there.
The issue there is that means it’s not just fake details in the corners after geometric corrections. The R3 outputs a 6000x4000 pixel RAW file. I presume that even though the wider-than-14mm image is cropped to a 14mm FoV without mechanical vignetting, the output will remain 6000x4000. That suggests that the cropped image must be up-scaled…fake pixels across the whole image.

I don’t know if that’s true, will be interesting to see if the RAW images at 14mm have larger pixel dimensions (i.e., use some of the non-effective pixels).

I'm kind of glad I got the EF 16-35mm f/4 almost new a while back, I can't see much benefit for me in upgrading. For others, if having the native RF mount, 175g less weight and 2mm extra at the wide end is important, then there isn't much more choice, other than paying not that much more relatively speaking and stepping up to the 15-35mm f/2.8 which is a much better engineered lens.
The RF is significantly shorter and lighter, for me that’s useful for travel. Although I sometimes travel with the 11-24/4 and the 17+24 TS-E, But sometimes I want a smaller and lighter kit and the 14-35 and 24-105 will make a nice kit for those occasions.
 
  • Like
Reactions: LogicExtremist