comparing the 2 @ 2.8, yes the $100 is sharper, but at the same time, the 1.2 is sharper wide open then the 1.8 - to be expected for the price difference. really the 1.2 doesn't seem to be that great a lens, despite its price, across the range. if you compare the 1.4, its clearly the best of the 3 that canon makes

now if you want a comparison that makes my point, look at the 1.4 @ 2, and the 1.8 @ 2... which is sharper :) then compare both at 2.8.

to go further, if you look at the Nikon 50 1.4, its just plain nasty loaded with purple fringing, but at 2.8 looks very very good.

so lens design, glass & coating formulation all play roles. a lens builder can pick where they want their lens to be best, at the compromise of where it does less then great - wide open or stopped down a bit.

well if you look at the both wide open, 2.8 v 4 @ 16 and 17mm, the 17 F4 is sharper... but stop the 16 2.8 to 4, and now its better based purely on F stop. then at 5.6, you win because you may well be out of the faster lens's sweet spot for optical performance. I'm sure there are other lenses out there that would do better or worse, because in absolute terms the 17 F4 is better, but they are both generally close enough that no one would ever turn up either lens as bad.


ok, its been a long time since I sat thru classes on this stuff, and google failed to quickly turn up what I was looking for ... for now ;)

correct, wide open, stray light comes in from the edges of the lens reducing contrast and causing aberrations that reduce sharpness / contrast. at about 2 stops down from wide open, the iris blocks out most of this stray light, until you get the iris small enough to start causing diffraction on its own. this is what I was talking about in general terms of physics of the lens. I wasn't going to start putting up the math... its been way too long for that.

well in general you are right, the more glass, the more chance for reflections between elements to reduce contrast. thats where lens coatings come into play to reduce these. so in general a lens with fewer elements should be sharper then one with more. however, with coatings, and even painting the edges of the glass black, it can change this so that a lens with more elements can perform better then one with less. perhaps this is part of how some slower / cheaper lens can sometimes be sharper - they are simply made with fewer elements to reduce contrast, where as the wider more expensive super fast glass may require more glass, or more radical shapes to make it work.


pity me never having heard of such lens. I think you can always find an exception to just about anything. a faster lens will let more stray light in reducing contrast, so you stop it down towards 4 / 5.6 which is where more lenses do best ( very generalized rule, plenty of exceptions of lenses that do better or worse ).

back to my original statement edited " a larger front element will result in a sharper lens "

quoting -------

Picture sharpness also varies with f-number. The optimal f-stop varies with the lens characteristics. For modern standard lenses having 6 or 7 elements, the sharpest image is often obtained around f/5.6–f/8, while for older standard lenses having only 4 elements (Tessar formula) stopping to f/11 will give the sharpest image. The reason the sharpness is best at medium f-numbers is that the sharpness at high f-numbers is constrained by diffraction,[4] whereas at low f-numbers limitations of the lens design known as aberrations will dominate. The larger number of elements in modern lenses allow the designer to compensate for aberrations, allowing the lens to give better pictures at lower f-stops. Light falloff is also sensitive to f-stop. Many wide-angle lenses will show a significant light falloff (vignetting) at the edges for large apertures. To measure the actual resolution of the lens at the different f-numbers it is necessary to use a standardized measurement chart like the 1951 USAF resolution test chart.

------------
F-number - Wikipedia, the free encyclopedia

I'll add that coatings, APO / FL / exotic glass, and the refractive index of the particular glass used can all come into play here. even the internal construction of the lens can have play here - there are some lenses that have internal baffles to control stray light, or at the rear exit. this all means its easy to find exceptions to general rules.

also, because of how F stops are calculated, a larger diameter lens should be able to top down more before diffraction thru the iris takes place... well that would be a very interesting one to figure out.

Often, the advantage of a fast lens stopped down over a slow lens opened wide is that the fast lens will have less falloff. This is especially true of lenses from off-brand manufacturers who want to offer high-end specs for a low-end price. To achieve that, they simply let you open the aperture wider on a relatively small lens than where Canon, Nikon, Zeiss, or Leica would ever allow their designers to go.

Personally, for video on a DvSLR, I'm more sensitive to unwanted falloff than softness. You can always vignette in post. Un-vignetting isn't as successful. As long as you really nail focus on your subject, the sharpness results are usually acceptable with any quality lens.

For 21MP photos and large prints, sharpness comparisons mean a whole lot more than for video.

Thanks for the response, Kris. I should not have guessed that you guessed it. :) Sorry about that.

Well, it's easy to point out flaws in the work of someone else, but coming up with your own work is a lot more effort. :) The only comprehensive data sets I can find are for raw files, like the chart you linked to.

I'm sure there are some good ISO-vs-noise comparisons for 5D2 video out there somewhere, but the only ones I can find right now are the flawed ones, such as testing with the lens cap on, using the same exposure, and/or low-frequency flourescents, etc.

Compiling that sort of useful data for video is more difficult because it only applies to the specific dynamic range of the scene, color temperature of the light, in-camera settings (especially tone curve, contrast, and white balance), what tonal level of the image is analyzed, and how much detail it has.

For example, in a low dynamic range scene with magenta light, neutral white balance, and normal in-camera contrast settings looking at middle gray in a highly detailed section, ISO 800 looks great -- just a little bit of random noise, like grain -- and in fact a lot of people prefer the nice random "grain" of the higher ISO setting.

But in high dynamic range scene, with low-CRI tungsten light (e.g. 3000K), blue-ish white balance, low in-camera contrast settings, looking at deep shadows (e.g. 4 stops below middle gray), with HTP enabled, even ISO 160 will have the much disliked pattern noise (lines).

To make matters more complicated, in some cases, one setting may have less random noise, but more pattern noise. That's why most people prefer ISO 160 over ISO 100 -- the 160 actually has more random noise in the midtones and highlights, but it (sometimes) has less pattern noise in the deep shadows, making it preferred.

Sorry I can't be more helpful.

Just that chart.

I think it's because Canon put all their effort into creating fantastic bokeh -- sharpness wasn't a priority at all. I absolutely love that lens for environmental portraits, but I'd rather take the f/1.8 for landscapes.

You've got me there, but I would point out that there's still a four-fold difference in price. I maintain that if they quadrupled the cost of the 50mm f/1.8, they could make it even sharper than the 50mm f/1.4 when both are stopped down to the same f-stop.

I don't think it's the exception, though. If you have two similar lenses, one fast and one slow, and stop them down to the same f-stop, I think the slower lens will either be sharper or cheaper (or both). In further effort of demonstrating this as a general rule, here are a few more examples:

Nikon 17-35 f/2.8 ($1800) @ f/4 vs 16-35 f/4 @ f/4 ($1200) - the slow lens is much much sharper and much cheaper.

70-200 f/2.8 L IS (Mark I for $1900) @ f/4 vs 70-200 f/4 L IS ($1200) at f/4 - the slow lens is slightly sharper and much cheaper.

70-200 f/2.8 L ($1300) @ f/4 vs 70-200 f/4 L ($700) @ f/4 - the slow lens is softer but much cheaper.

24mm f/1.4 L Mark I ($1200 used?) @ f/2.8 vs 24mm f/2.8 ($300) @ f/2.8 - the slow lens is a little sharper and much cheaper.

I don't think that quote supports your position.

Let me illustrate it. f/10 on a 10mm optic is a very small aperture stop (1mm in diameter), while f/10 on a 2500mm optic like my telescope is much larger (250mm). The 10mm weighs 1 pound and the 2500mm weighs 50 pounds. But when they're both used on the same camera, the diffraction in image space (what photographers care about) is the exact same, because the f-number is the same.

I agree. Even if you had automatic vignetting correction, the corners would have more noise, which might be worse than just being darker.

let me add into the mix the Tamron 70-200 2.8. optically better then the original canon 70-200 2.8, and close enough with the V 2 of the canon that you couldn't tell them apart optically looking at images.

Tamron 70-200 2.8 vs Canon 70-200 2.8

and its a $700 lens, or 1/3 the price of the canon V2 which is something like $2300 or so ?. its internal zoom & focus as well, and I love shooting with this lens. However, I don't think price has really been much of a consideration here, unless maybe we were talking megabuck glass.

well the question is : is diffraction in a lens a function of the iris size ratio in relation to the overall size of the lens, or is it a function of the physical size of the light wavelengths passing thru the hole of an absolute size, like how a diffraction grating works.