Report: Canon to launch 4 fast L prime lenses in the first half of 2024

[clearwaterphotog]

I have yet to see anyone post convincing evidence (or any evidence at all, for that matter) that digital correction is worse than optical correction for geometric distortion. But somehow, people are just sure it must be.

I've seen your comparisons and I don't disagree with them; the proof is clearly in the pudding. The 14-35, with all of its distortion, is still an impressively sharp lens in the corners, so I'm largely not as concerned with perfect optical correction (as much as I'd like it).

However, from a purely theoretical standpoint, optical correction allows infinite information to be retained. That is, glass is not converting the light to digital values and then doing some math on those values to produce a corrected image. A camera sensor does that. The ultimate result when correcting distortion digitally is going to be limited by the amount of resolution you have to "fill in the blanks" when squeezing/de-squeezing the image. That said, even if you have a 1MP image and do your de-squeeze, you're not trying to fill in extra MP worth of information; you're still only targeting a 1MP end result. So, any "missing" information that was not resolved, shouldn't matter/be visible, because the start and end size of the image are the same...? Maybe I'm confusing myself here, but it made sense in my head.

 

[clearwaterphotog]

I am part of the 28mm f/1.4 religion, ordered the EF Sigma the day it became available. When I think about photography as art - vs. journalism or documenting, when I need to carry 1 or more zooms to ensure I get a particular shot - 28mm on full-frame is the focal length that produces the closest reproduction of what I see with my own eyes. For others this is 35mm or 50mm, apparently, and so be it. It also matters how far away your subject is, and mine tend to be short-to-medium distances... I would probably choose 50mm if I was trying to capture things that were farther away on average.

I would love to replace my 15 - 35mm f/2.8 IS with a similar lens that gets a bit wider at the expense of the other end, like a 13 - 24mm or something, but only if it could take screw filters and those filters were 82mm or 77mm. That way I could justify carrying a big 28mm f/1.4 more frequently.... Then just toss a nifty 50 in the bag and I'm pretty much good to go for all the walkaround stuff I do. Or, if I'm doing family portraits at Christmas, replace the 28mm with the 24 - 105mm f/4 (for scampering children) and the EF 85mm f/1.4 IS or RF 135mm f/1.8 IS (for poses).

I have never owned a dedicated 28mm lens, but have used the FL a lot on phones and on my old Fuji 18-55 which started at 28mm equiv on the wide end. For landscapes, I find that it's a nice starting point, but I almost always want wider or tighter. For documentary/walkaround/travel type shooting, however, I really find 28mm to provide a natural field of view for me. It's wide enough to allow balance in the frame without having to take several steps back. I was trying out the RF50/1.8 in the city a couple weeks back and I found it too tight; though I've shot at 50mm a lot in the past on my Fuji system (35/1.4), I can't seem to get the hang of it now. I either want semi-wide that allows me to fill the frame with things I see within 10 feet of me, or I want semi-tele that allows me to isolate something interesting 50-300ft away (depending on its size ofc). This leads me to 28 and 85. I'm quite versed with 35mm and still find it just wide enough to capture what I see close to me, but sometimes it is just a smidge tight. 28mm also offers a slight amount of distortion which, from some photos I've perused, seems to add depth or pop, and that's also an effect I'd like to play with.

So, if a 28/1.4 actually does materialize, I could very well be interested. It definitely seems like a niche FL these days though.

 

[Del Paso]

I am part of the 28mm f/1.4 religion, ordered the EF Sigma the day it became available. When I think about photography as art - vs. journalism or documenting, when I need to carry 1 or more zooms to ensure I get a particular shot - 28mm on full-frame is the focal length that produces the closest reproduction of what I see with my own eyes. For others this is 35mm or 50mm, apparently, and so be it. It also matters how far away your subject is, and mine tend to be short-to-medium distances... I would probably choose 50mm if I was trying to capture things that were farther away on average.

I would love to replace my 15 - 35mm f/2.8 IS with a similar lens that gets a bit wider at the expense of the other end, like a 13 - 24mm or something, but only if it could take screw filters and those filters were 82mm or 77mm. That way I could justify carrying a big 28mm f/1.4 more frequently.... Then just toss a nifty 50 in the bag and I'm pretty much good to go for all the walkaround stuff I do. Or, if I'm doing family portraits at Christmas, replace the 28mm with the 24 - 105mm f/4 (for scampering children) and the EF 85mm f/1.4 IS or RF 135mm f/1.8 IS (for poses).

It may also depend on which WA you started with.
My very first WA prime was the Minolta Rokkor 2,5/28mm. Since then, I always had 28mm primes. The EF 28mm IS was not too convincing, the Sigma, optically superb, had, on the 5 D III & IV , an unreliable AF. Therefore, I'm happy with Canon "announcing" an RF L 28mm.

 

[neuroanatomist]

Therefore, I'm happy with Canon "announcing" an RF 28mm.

They already did! It's a great little lens...

 

[Quarkcharmed]

No, that's not what's happening.

Yes, there is 'stretching' going on...but before the stretching, there is a 'squishing'. The aberration being corrected with the stretching is barrel distortion. The purpose of the stretch is to algorithmically correct the squish. If the idea of stretching bothers you, think of it as it really is...unsquishing.

Here's a diagrammatic example of what's happening in the corner of an image from such a lens:

Barrel distortion makes straight lines into curves and circles into ovals, and correction restores them to straight lines and circles. In a lens that's optically well-corrected, lens elements do that job.

I have yet to see anyone post convincing evidence (or any evidence at all, for that matter) that digital correction is worse than optical correction for geometric distortion. But somehow, people are just sure it must be.

I remember when people didn't trust a calculator to perform arithmetic. Pencil and paper were better. Those people just knew it. The way you know about a good melon, or something.

Yeah. But you can't squish or stretch the grid of pixels. Almost any transformation on a raster image means loss of information. Distortion corrections are all lossy.

 

[neuroanatomist]

Distortion corrections are all lossy.

Yes, they are. That's my point – the distortion correction performed optically by the lens also costs resolution. Show me the corner of an ultrawide lens that's as sharp as the center. What you're saying is that it's somehow worse to lose the resolution during processing instead of losing it before the image is captured. Why?

Maybe it matters to you whether the detail is lost before or after the light hits the sensor. From my perspective, loss of resolution is loss of resolution. What I'm seeing is that the loss is no different between the corners of an optically corrected lens and the digitally corrected corners of a less optically corrected lens.

So, I'll ask again – what empirical evidence do you have that digital correction of geometric distortion after image capture is intrinsically worse than optical correction of that distortion before image capture? Or do you just have opinions and suspicions?

 

[neuroanatomist]

However, from a purely theoretical standpoint, optical correction allows infinite information to be retained.

Theoretical lenses are nice. I haven't found one that I can use to take pictures, have you? Regardless, optical correction allows capture of only the amount of information actually transmitted by the lens, whether that lens is theoretical or real.

Consider any lens MTF curve. For most manufacturers, those are theoretical curves calculated from the optical formulae of the lenses. Notice how the lines always drop as you go from left to right? That means there is less information available at the corners of the frame than at the center, even theoretically. That's exactly my point – distortion correction costs detail. That's true even if the distortion is corrected optically by the lens.

 

[Quarkcharmed]

You suspect. That’s nice. I’ve actually empirically tested it and found there was no difference. That was with the EOS R, which has among the lowest resolutions of current FF cameras (the 24 MP sensors in the R3 and R8 outresolve the 30 MP sensor in the 5DIV and R).

That was a good post, however, you found no difference between what and what exactly?

You'd need to compare an RF 16 or an RF 14-35 with digital corrections against the same lens with some imaginary redesigned (more expensive) optics.

 

[m4ndr4ke]

I'm totally ignorant on these matters so could people enlighten me: why a 28 as well as 24 and 35? Is that a gap that needs to be plugged?

It’s a different look. I don’t like the stretched corners on the 24mm, and the 35 doesn’t feel as useful as the 28, to me, in the sense of including “everything” in the frame.
The same way the 40mm is pretty much the middleground of the standard range, the 28mm is the middleground of the wide angle range.
The 28mm will give you a very neutral looking wide angle perspective, if that makes sense, with lower distortion than the 24. It’s “a more boring perspective” than the 24, which can be good.

Oh, and it pairs wonderfully with a 50mm. You can check out Larry Towell’s work, for instance, even though he uses one vertically and the other one horizontally.

 

[neuroanatomist]

You'd need to compare an RF 16 or an RF 14-35 with digital corrections against the same lens with some imaginary redesigned (more expensive) optics.

Obviously not possible. But, by using a lens that requires rather extreme digital correction (distortion sufficiently bad that the image doesn't fill the frame) with a substantially more expensive lens in the same class that at the same focal length requires no distortion correction, I would say I was stacking the deck in favor of finding that digital correction is worse. Except that it wasn't.

To be honest, I was quite surprised by the results because I went into it suspecting that optical correction was better. That demonstrates why empirical testing is better than suspicions.

 

[SwissFrank]

Optical corrections are applied before the image hits the sensor.
The entire sensor captures the corrected image.
In digital correction, the image has to be stretched digitally to correct for distortion which causes a loss in detail.

Does it cause a loss in detail vs. optical correction you can see even in 1:1 pixel-peeping or fat bits mode? If so do you have time to share an example?

 

[SwissFrank]

The same line of pixels lose some of the pixels, the rest will need to be stretched, but because they're discrete pixels, you'll have to fill the gaps with interpolated pixels. Loss of information in this case = loss of resolution.

OK, an R5 sensor is 8192 pixels wide and 36mm, right? So 227 pixels per mm. That's 110 line pairs per mm, or 110 lp/mm. Have you ever seen a lens that can render 110 lp/mm? What's the deepest test we've seen? I think I saw one site testing 60 lp/mm and I seem to recall the best lenses even in the center didn't have much contrast at 60 lp/mm.

In theory, sure, every source pixel's energy can end up on up to 4 destination pixels. That causes a blur of up to 1 pixel. That would harm 110 lp/mm details but I don't think lenses can really render those anyway. Can you find a photo of yours that has a strong jump from say white to black to white again in three neighboring pixels? That'd be the kind of detail that distortion correction would blur out. Do you have any photo that can show us an example of that? Maybe a perfectly-focused star? Or generally any pixel-peeping 1:1 or even fatbits that can show us how resolution is lost in practice rather than in theory?

To be clear I don't mean to be arguing with you, I'd legitimately like to learn something.

 

[Del Paso]

They already did! It's a great little lens...

View attachment 212950

I had stupidly forgotten that one.
According to all reviews, it must indeed be very good! And, since I prefer lighter lenses...next buy!

 

[EricN]

You suspect. That’s nice. I’ve actually empirically tested it and found there was no difference. That was with the EOS R, which has among the lowest resolutions of current FF cameras (the 24 MP sensors in the R3 and R8 outresolve the 30 MP sensor in the 5DIV and R).

I’m not suggesting that what I found will be true in all cases. But the observation that a digitally-corrected UWA zoom at the wide end delivers corners as sharp as a UWA zoom costing twice as much and used at the ‘sweet spot’ (from a geometric distortion standpoint) shows that optical corrections are not intrinsically superior, as you are claiming.

Have you tested your suspicions? What were your findings, and where can they be viewed?

Side note: for a fun exercise, compare the digitally corrected corners of the $300 RF 16/2.8 to the quality glass of the $2100 EF 14/2.8L II.

I bought the RF 15mm F/2.8 and RF 50mm F/1.8 for my niece because I wanted her to have more opportunity to play with depth of field more than the kit lenses offer. Like a good uncle, I tested the lenses on the R5 and the 50mm seemed about what I expected, but I was impressed with the 16mm's size, weight and cost enough to buy my own. It's amazing to me what they've accomplished in less than 10 years and I'm really looking forward to seeing what these new L primes will be capable of.

Edit: I meant, "RF 16mm"

 

[SwissFrank]

14, 24, 28, 35, 50, 85, 135 all make great sense for Canon to flex their optomechanical engineering muscle.

One sad note is that interchangeable-lens camera sales are down 90% or whatever from the peak. They never made a 28/1.4 in the past. I'm not sure how they could make one now, unless CAD/CAM makes it so much easier and cheaper to do so.

 

[neuroanatomist]

I had stupidly forgotten that one.
According to all reviews, it must indeed be very good! And, since I prefer lighter lenses...next buy!

I've been very happy with mine. Surprisingly good IQ and very convenient. Great on an R8 for street, but I also use it like I did the EF 40/2.8 as a small lens to put in my pocket when shooting an event with mainly with the RF 70-200. I can leave the telecom hanging on the BlackRapid strap for the few times I need something wider.

 

[EricN]

Just to add my opinion to the stack with the hopes that Canon Japan reads this. Much like many professionals I have been patiently awaiting wide high quality RF L-Series Primes. My personal take is these lenses must not rely heavily on lens correction. Meaning high quality optics, well corrected, ideally state of the art to fit in with the pedigree and performance of the current high performing L-Series Primes (50mm f/1.2, 85mm f/1.2, and 135mm f/1.8).

24mm and 35mm are bread and butter focal lengths for some and these are the most pressing needs. It's nice, if true, that Canon makes an L-Series 28mm. That would be "new" and possibly sway those who dance in the realm of Leica to look very hard at Canon if that glass sings.

The 14mm, whether it lands at f/1.4 or f/2 would be an accomplishment. Again, a highly corrected optic in the rectilinear sense would be very desired here.

Canon's inexpensive 35mm f/1.8, 50mm f/1.8, 85mm f/2 Macro straddle the line of very high value and well designed optics. The 24mm f/1.8 Macro however is borderline a fisheye lens without digital correction. The 16mm f/2.8 is a fisheye lens really with that much distortion. The most recent 28mm f/2.8 pancake I give a pass on this as it a bit of specialty lens and super compact. To that note the L-Series glass to command the price that they will launch at and to differentiate themselves further from the "close in maximum aperture" focal lengths ideally would be well designed glass first.

Though critical in my points here the current L-Series primes are some if not the highest quality glass on the market and I can say that safely owning various other contenders here. Which is why I want them to hit a homerun on these.

14, 24, 28, 35, 50, 85, 135 all make great sense for Canon to flex their optomechanical engineering muscle. Down the line, when possible, I'd love for them to tackle a new 200mm f/1.8 or f/2 with a new design, not just a new mount like the 400mm f/2.8. I suspect one of the reasons they've held off on a new 300mm is they are focusing on a new design from the ground up as well. Perhaps faster and/or smaller.

I also wouldn't mind something in the 18-21 range in the L-Series, but that would be a stretch for now. Zeiss has been MIA and a quality modern 21mm hasn't existed really outside of Sigma's ART 20mm in recent years. And Otus never finished the rest of the focal lengths. And to that point, though highly unlikely, a 40mm would be a pleasant surprise in either an L-Series or more likely a pancake to complement the 28mm. Wouldn't mind something in the 75-90mm range after that if they are looking to make a small trio of pancakes, which I think RF deserves.

To wrap up, I understand less expensive optics sell in higher volume and I'm okay relying even on digital lens correction in some cases. However, you have to play both sides as working professionals who are looking for "the best" still need tools as well without heavily relying extreme image manipulation to simply correct an image's distortion. Small distortion, much like the EF L-Series primes, that's fine. And the vastly improved optics of the RF 50mm f/1.2 versus the EF 50mm f/1.2 really show Canon at their best. I'd like to see them still pressing the gas and leading in that capacity.

I expect them to follow the other L prime's design philosophy, but I won't be too disappointed if there is more digital correction because it will help keep the price, weight and size lower.

 

[EricN]

Your landscape shots are done with f/1.4???
With f/8 there are 4 alternative full frame options on the market.

Depending on what the intent is, 1.4 could make a nice landscape photo.

 

[EricN]

When you do corrections for geometric distortions, consider a straight line (say a diagonal) that goes from an edge of the frame to the centre. Because you typically have to stretch the image (like with the RF 16mm), some pixels go out of the frame and you lose information. The same line of pixels lose some of the pixels, the rest will need to be stretched, but because they're discrete pixels, you'll have to fill the gaps with interpolated pixels. Loss of information in this case = loss of resolution.

This is going to happen optically, too. The correction technique is different but the outcome is nearly the same. In either case, most of your audience won't notice a difference.

 

[EricN]

I disagree... "coordinated" means 2 or more systems acting in synch. Yes there may be cases when the best outcome requires one of the systems to do nothing. But 1 system alone does not coordinate with itself

That was my understanding, but apparently it's not the correct way to discuss the topic. I'm happy to accept it either way.