Daan,
I did a similar test series for you. My 16-35II shows similar behaviour: fringing @f/2.8, but not at the ends of the zoom range, like you described. As I guessed you would have used your 5DII, I used my 40D to achieve similar pixel density for the 100% crops. I could send you the files if you like. Hard to put them up here on FM, I'm travelling at this time and I have limited access to internet (and not much time to prepare/upload).
Anyway I think there's nothing wrong with your copy, AAMOF I never valued the 16-35II much for sharpness wide open and I think your samples look very good!
Let me know if I should send you anything.
Daan B wrote:
But why does it do this in the 20-28mm range, but not (so much) at 16mm and 35mm?
Daan, the 16-35 L II makes complex movements during zooming. The front group is furthest back somewhere between 24 and 28 mm focal length, for example. I don't know how this affects the light path, but it seems quite possible that that's the region that results in most sharply angled rays when wide open, even though the rear group makes a linear movement that puts it closest to the sensor at 16 mm FL. There could of course be other factors that affect lateral chromatic aberration vs FL, not just birefringent refraction. (I have no expertise, I'm just bringing up ideas that seem relevant. Hope that's ok.)
From a practical standpoint, I've encountered just this sort of fringing at very high contrast boundaries even with the 24/3.5 TS-E, which is generally highly resistant to chromatic aberration. I was able to minimize it in Lightroom 4.1 using the CA adjustment. I suspect that LR's built in profile for the 24-35 removes a good bit of CA. I'm certain Canon's DPP does too, I just don't happen to use it.
I'm not trying to discourage you from having Canon check the lens, but it may in fact be fine.
rabbitmountain wrote:
Daan,
I did a similar test series for you. My 16-35II shows similar behaviour: fringing @f/2.8, but not at the ends of the zoom range, like you described.
Wow. Thanks for doing your tests. I would love to see the results. PM sent
Since this happens on-axis and improves very rapidly upon stopping down, it's likely to be plain old spherical aberration (with traces of other aberrations too, no doubt).
It's not surprising that it's most visible at intermediate focal lengths. Modern zooms are often optimised for the extremes of their range. My Panasonic LX3's 24-60 mm-e lens is weakest around 28 mm-e, for example. There's no law that says the extremes must be weakest, though they often were in the past.
S Dilworth wrote:
Since this happens on-axis and improves very rapidly upon stopping down, it's likely to be plain old spherical aberration (with traces of other aberrations too, no doubt).
Nope. Spherical aberration does not show up as a magenta fringe or glow, nor is spherical aberration limited to high contrast edges. It's birefringence.
Photon wrote:
I suspect that LR's built in profile for the 24-35 removes a good bit of CA. I'm certain Canon's DPP does too, I just don't happen to use it.
I can confirm that LR4.1 doesn't take away any CA when Enable Profile Corrections is checked.
Monito wrote:
Spherical aberration does not show up as a magenta fringe or glow, nor is spherical aberration limited to high contrast edges. It's birefringence.
There's not much colour in Daan's example photo, Monito, and spherical aberration does indeed cause things to "glow." (Its effects are perhaps most often described as a glow.) Daan says "the whole picture itself lacks contrast and clarity, making it soft looking," which is consistent with spherical aberration. There's some chromatic aberration too, I'm sure. And other aberrations.
Spherical aberration isn't limited to high-contrast edges, but nor is any optical effect. It's just more visible there.
As soon as anyone starts blaming sensor micro-lenses for odd optical effects, I immediately become suspicious. But in this case, how do you suggest micro-lenses cause light to bleed across several pixels? Why would stopping down help so strongly, since the chief ray angle doesn't change? Why would it be so evident on-axis, where the light rays strike the sensor much more perpendicularly than at the periphery? Why would it happen at all in a recent, heavily retrofocus wide-angle, but not in any number of lenses with more severe chief ray angles?
The colour is magenta. Please look closely. It is not a red-green combination (not red on one side of a line or object, green on the other); it is magenta.
You do not see birefringence like that on film, certainly not to that degree. I never heard of it or saw an example until DSLRs became prevalent.
From the 1D Mark II white paper:
"The ISO 3200 speed is one of several substantial improvements made possible by the Canon-developed, on-chip RGB primary color filter together with larger microlenses that have much smaller gaps between them than those on the 1Ds. These narrow gaps greatly increase the efficiency of light convergence while greatly reducing birefringence."
By the way, Daan, LensTip and Photozone both show relatively poor performance in the middle of the zoom range, for the on-axis image at f/2.8, with improvements at the extremes of the zoom range. These sites test at a fairly high spatial frequency, so spherical aberration isn't as evident in their tests as it is to your eyes, but their test results are compatible with your photos.
Spherical aberration causes focus shift when stopping down. So one way to check if there is significant spherical aberration, is to check if there is noticeable focus shift. To test, set your lens to the problematic focal length, put the camera on a tripod, and shoot a close flat subject (like a table surface) intercepting the optical axis at a shallow angle. The subject should have lots of fine detail (like wood grain). Shoot at f/2.8, f/4, f/5.6, and f/8, taking extreme care not to move the focus ring or disturb the camera. Wildly over-sharpen the resulting images, and check if the point of best focus appears to move away from the camera as the lens is stopped down.
A ruler or something similar would work instead of a table, but make sure it has lots of markings (such as mm markings rather than just cm ticks), and light it for high contrast.
One last thought: this lens has a replica (plastic) aspherical surface, which may be more susceptible to temperature variations than all-glass aspherical elements. It looks like a warm day in your photo. Had your camera been left in the sun? If so, the lens might have been at the edge of its performance envelope, for a worst-case result.
I thought you could take a mild insult on the chin, Monito. You dish them out often enough! Apologies.
My point is not that there is no false colour around the highlights (there is, as you note), but that the false colour is minor compared to the monochromatic aberrations present. On the face of it, I think the main issue here is spherical aberration. But ultra-wide zooms are incredibly complex, and there are other possibilities.
Effects caused by severe light angles at the sensor surface can be ruled out by considering the questions I posed.
The false colour surrounding the highlights (not actually in them) is magenta. It is not red-green spatially oriented, nor is it cyan-magenta spatially oriented (red on one side of a twig and green on the other would be a classic example).
There may be other defects, including perhaps improperly cleaned lens surfaces (it's a used lens), and possibly some aberration introduced by elements out of adjustment. But the glow referred to is coloured and it is associated with bright areas. It also seems to be the main defect.
There is a tiny amount of red-green chromatic aberration visible only at 200 % magnification (which is an extreme test) if you look very carefully at the vertical behind the coffee cup. However, that seems to be behind glass and thus it is not a proper test.
As to why birefringence will cover several pixels at an edge of a highlight area, ask Canon or an optical scientist. However, it seems to make sense to me that if light is being refracted by the sensel microlenses, and it is, that some would be refracted and splashed to neighbouring sensels. It takes strong light at contrasty edges and when you have a very good modern DSLR (not stated by the OP) and a very good modern lens it would be very much reduced and tamed.
What would be needed would be a test (by one person or entity) with identical conditions and processing that would compare four cases (think 2x2 matrix) with pre- and post- Mark II cameras in the 1D or 1Ds lines and the first 16-35 and the 16-35 Mark II.
The OP has not stated the processing of the example image.
Personally I am not convinced that the purple glow (seen most strongly it seems with the 85 f/1.8) should properly be called birefringence, but that is Canon's term for that kind of optical interaction that sensors have with lenses in some situations, so I call it that for lack of a better term.
Here are several examples of Zeiss birefringence on 5D2. Note the all-round purple glow.
As to the OP's lens and camera and Raw conversion (in-camera or off-camera), we are really only just speculating as to what the problem might be since the example is not a test. Was a lens hood used? Only the OP knows. Other info like what camera was used has been lost when the OP stripped off the EXIF info from the posted images.
S Dilworth wrote:
There's not much colour in Daan's example photo, Monito, and spherical aberration does indeed cause things to "glow." (Its effects are perhaps most often described as a glow.) Daan says "the whole picture itself lacks contrast and clarity, making it soft looking," which is consistent with spherical aberration.
Spherical aberration isn't limited to high-contrast edges, but nor is any optical effect. It's just more visible there.
Very true. If any color at all, it is more blue-ish than magenta. And the glow isn't limited to high contrast areas. It is all over the place, especially at 27mm 2.8. Sometimes much worse than other times. Shooting in the sun makes things worse.
S Dilworth wrote:
By the way, Daan, LensTip and Photozone both show relatively poor performance in the middle of the zoom range, for the on-axis image at f/2.8, with improvements at the extremes of the zoom range. These sites test at a fairly high spatial frequency, so spherical aberration isn't as evident in their tests as it is to your eyes, but their test results are compatible with your photos.
Ralph also sent me some crops where you can see that at 16 and 35mm there isn't any glowing, while in between those FL's there is. Exactly as my copy behaves.
Spherical aberration causes focus shift when stopping down. So one way to check if there is significant spherical aberration, is to check if there is noticeable focus shift. To test, set your lens to the problematic focal length, put the camera on a tripod, and shoot a close flat subject (like a table surface) intercepting the optical axis at a shallow angle. The subject should have lots of fine detail (like wood grain). Shoot at f/2.8, f/4, f/5.6, and f/8, taking extreme care not to move the focus ring or disturb the camera. Wildly over-sharpen the resulting images, and check if the point of best focus appears to move away from the camera as the lens is stopped down....Show more →
I will run some tests for focus shift.
One last thought: this lens has a replica (plastic) aspherical surface, which may be more susceptible to temperature variations than all-glass aspherical elements. It looks like a warm day in your photo. Had your camera been left in the sun? If so, the lens might have been at the edge of its performance envelope, for a worst-case result.
I never leave my camera in the sun
So if it is spherical aberration, it looks like a 'feature' of this lens. Since it isn't just my copy that shows it. Really disappointed in the wide open performance at the in-between FL's
Lars Johnsson wrote:
Every lens that I own (especially fast lenses) is like this more or less. Even my best Zeiss lenses or the most expensive Canon lenses. With a subject like yours. Black background and silver/white subject that even are a bit hot it's nearly impossible to avoid it
I'm not positive, but I don't think it corrects for this sort of thing - it mostly corrects pincushion/barrel and related distortion, CA (of the red-green type), and vignetting.
If you want to reduce this in the unusual case where it actually makes a difference and you can't stop down a bit, try the sharpening values I mentioned earlier.
Monito wrote:
The false colour surrounding the highlights (not actually in them) is magenta. It is not red-green spatially oriented, nor is it cyan-magenta spatially oriented (red on one side of a twig and green on the other would be a classic example).
It looks more like blue to me. But there isn't much color in the glowing.
There may be other defects, including perhaps improperly cleaned lens surfaces (it's a used lens), and possibly some aberration introduced by elements out of adjustment. But the glow referred to is coloured and it is associated with bright areas. It also seems to be the main defect.
The lens is brand new with a pristine front end. I don't know if any elements are out of alignment. How do you check for that? The glowing is in high contrast areas, but not limited to that. Shooting in the sun makes things worse.
It takes strong light at contrasty edges and when you have a very good modern DSLR (not stated by the OP) and a very good modern lens it would be very much reduced and tamed.
5D2
The OP has not stated the processing of the example image.[...]As to the OP's lens and camera and Raw conversion (in-camera or off-camera), we are really only just speculating as to what the problem might be since the example is not a test.
Fist image: RAW converted in LR3.6, exported with low sharpening for screen.
Images of bike: RAW converted with LR3.6 at default values. No sharpnening applied on export.
