Canon Patent Application: Ultra-wide RF 14-20mm F2.0

SwissFrank

1N 3 1V 1Ds I II III R R5
Dec 9, 2018
526
361
I suspect that the lenses in the patent application will require a decent amount of software correction as well, which might not fit astro photography.
I keep asking for an example where the software correction is a visible let-down even in pixel-peeping 1:1 and haven't seen an example of it yet. Instead someone will say, something like "oh gosh, no, we don't want our stars to be stretched, I'll never buy this lens" but I haven't actually seen what this stretched star problem even is. It's not as they stretching them OUT of shape, but rather simply stretching them back INTO shape. Granted this kind of software correction will always cost you about a half-pixel of sharpness on average, but it also gives the designers far more freedom with the lens design, and with reduced constraints on geometry it's not clear to me that the image quality wouldn't be enough sharper that the result is better than an uncorrected lens.

To be clear I have no experience and would love to see a side-by-side where any uncorrected lens took better stars than a corrected one. It's certainly theoretically possible, but I just wonder if that's what actually happens with actual lenses in actual photos of actual subjects.

I was also mystified about whether the software-corrected vignetting was ever noticeable in actual photos and I can't remember how that conversation ended, but if someone happens to have some photos that show how the vignetting correction is causing an actual problem that is actually visible in an actual photo of an actual subject I'd be really interested in seeing that too (and apologies as I know one or two members--maybe yourself koenkooi--was helping me understand that but whether we got to a result or not is slipping my mind.)
 
  • Like
Reactions: 1 users
Upvote 0

Maximilian

The dark side - I've been there
CR Pro
Nov 7, 2013
5,715
8,669
Germany
... Have since switched to the RF 14-35/4 and 10-20/4 and am happy.

No need for an f/2 UWA. Now…an f/2 short tele, 70-150 or 70-200, to pair with the 28-70/2, that I’d buy.
I suppose, that'll be the way, I'd go, too, if I think that I'll need more or better IQ than the 24 mm my 24-105 delivers.

About the f/2 zooms: I prefer size/weight advantage of f/4 and f/2.8 over f/2. But great, that Canon offers both/all three lines.
 
Upvote 0

AJ

Sep 11, 2010
968
438
Canada
I keep asking for an example where the software correction is a visible let-down even in pixel-peeping 1:1 and haven't seen an example of it yet. Instead someone will say, something like "oh gosh, no, we don't want our stars to be stretched, I'll never buy this lens" but I haven't actually seen what this stretched star problem even is. It's not as they stretching them OUT of shape, but rather simply stretching them back INTO shape. Granted this kind of software correction will always cost you about a half-pixel of sharpness on average, but it also gives the designers far more freedom with the lens design, and with reduced constraints on geometry it's not clear to me that the image quality wouldn't be enough sharper that the result is better than an uncorrected lens.

To be clear I have no experience and would love to see a side-by-side where any uncorrected lens took better stars than a corrected one. It's certainly theoretically possible, but I just wonder if that's what actually happens with actual lenses in actual photos of actual subjects.

I was also mystified about whether the software-corrected vignetting was ever noticeable in actual photos and I can't remember how that conversation ended, but if someone happens to have some photos that show how the vignetting correction is causing an actual problem that is actually visible in an actual photo of an actual subject I'd be really interested in seeing that too (and apologies as I know one or two members--maybe yourself koenkooi--was helping me understand that but whether we got to a result or not is slipping my mind.)
Exactly. Look at any rectilinear ultrawide photo (optically corrected) and things do appear stretched in the corners. It's the nature of rectilinear projection. It's a similar problem to map projections which tend to distort by their very nature.
So when we take a barrel-distorted image and correct it to rectilinear, we trade fidelity in area for fidelity in direction, sacrificing fidelity in shape..
I for one was initially concerned that the digital correction would result in smearing due to stretch. Since the amount of stretch between adjacent pixels is quite small, the overall stretch effect is small. And because things look stretched anyways due to the rectilinear projection, any smear due to stretching is well hidden.
The bigger problem with lenses that barrel-distort to the point that corrections are not optional anymore (e.g. 14-35) is vignetting. Even after correction the vignetting is still noticeable.
 
Upvote 0

SwissFrank

1N 3 1V 1Ds I II III R R5
Dec 9, 2018
526
361
About the f/2 zooms: I prefer size/weight advantage of f/4 and f/2.8 over f/2. But great, that Canon offers both/all three lines.
Agreed.

In the 90s we NEEDED an f/2.8 trinity. Even 100 speed film was nowhere near as sharp and low noise/grain as ISO 5000 is now, so the option to halve your ISO was always attractive. Likewise, no IS or IBIS so you wanted a shutter of like 1/200 even for 50mm and again, the option to halve your shutter was always attractive. Meanwhile, due to unsharp lenses, bad AF or manual focus, camera movement unsharpness and film grain, you couldn't make your photos too big, so the bokeh of f/2.8 was great to give some pop at an image size of like 2.5x4cm/1x1.5".

Today we can almost always double shutter speed and have an equally attractive photo. IS lets you hand-hold to absurd lengths* so you don't need fast shutter. Lenses are sharp enough you can count eyelashes, and the sensor MP record it. Nearly every photo can be considered 25x40cm/10x15" for these reasons, and at that size, even f/4 gives a lot of pop to a subject. Indeed, opening up much wider risks making the bokeh the topic of the photo, not the putative subject. So, I've been thinking since I bought one of the first R's, that the trinity these days is f/4. I shoot the 14-35, 24-105, and the 100-500/4.5-7.1 which I basically consider a 70-200/2.8 + 1.4xTC with an optional 2.0xTC.

I've always loved aperture and owned the 50/1.0 and 85/1.2 from the day I got started in Canon and the 135/2 when it went on sale. I had the 2.8 trinity and probably did at least three upgrades as newer variations came out. I still have a 28/1.4 and 135/1.8, but for me these are special-use. (I have the 50/1.8 mounted in my backpack any time you see me out of the house, too.) But for just recording what's happening, f/4 I think is where it's at.
 
  • Like
Reactions: 1 user