I'd like to start a discussion with the goal of us identifying "the perfect CA target" for testing lenses.
Criteria:
- Must be able to indicate, with good accuracy, both types of CA (maybe this means two targets).
- Must be a readily available object, so that we can all use it.
Let's face it, we all own and use different lenses, and none of us have them all. It would be great if we could put our collective efforts together by shooting a common target (with a consistent method) and comparing the results. I'd love, for instance, to be able to compare the CA of my FD 35/2 to the Rokkor 35/1.8, EF 35/1.4 L, Nikkor 35/1.4, etc. It's just not realistic for any single one of us to do so (I don't think).
Is there a better target than the white ruler or white page of text in full sunlight?
Do we need two different targets for the two different types of CA?
I like a black/white ruler on white paper with random paperclips on the same paper troughout the lenght of the ruler. But I don't know if that is a good scenario?
I don't know about perfect but with some work one can make good things with common USAF charts and the infamous Nikon AF test chart.
CA
I suggest standard common printed USAF charts photographed at a distance of 55x the focal length, raw, ACR, everything default except for the white balance and exposure, curves set to a straight line, zero sharpening. Exposure set to 210 at the white paper background after the WB has been corrected.
Then we'll be able to see the CA in a way making it possible to compare it. If the, now advanced form of brick wall, shooter has focused correctly we'll also see how "sharp" the lens and camera combo is in at the edge and the corner (and the center if the photographer is generous).
LoCA
The AF test chart, about 30 degrees, a distance where the paper fills the frame at the the top border.
Too complicated perhaps. But I guess you mean something for comparisons.
BTW: A white page of text in full sunlight rules out fast lenses, no?
ulrikft wrote:
I like a black/white ruler on white paper with random paperclips on the same paper troughout the lenght of the ruler. But I don't know if that is a good scenario?
Jonas B wrote:
I don't know about perfect but with some work one can make good things with common USAF charts and the infamous Nikon AF test chart.
CA
I suggest standard common printed USAF charts photographed at a distance of 55x the focal length, raw, ACR, everything default except for the white balance and exposure, curves set to a straight line, zero sharpening. Exposure set to 210 at the white paper background after the WB has been corrected.
Then we'll be able to see the CA in a way making it possible to compare it. If the, now advanced form of brick wall, shooter has focused correctly we'll also see how "sharp" the lens and camera combo is in at the edge and the corner (and the center if the photographer is generous).
LoCA
The AF test chart, about 30 degrees, a distance where the paper fills the frame at the the top border.
Too complicated perhaps. But I guess you mean something for comparisons.
BTW: A white page of text in full sunlight rules out fast lenses, no?...Show more →
Sounds like a good method, but not everyone has immediate access to those test charts. Aren't they horribly expensive?
Yes, full sunlight does pose an issue with fast lenses. I suppose it doesn't really matter how bright it is, as long as there is a lot of contrast. Black text on white should show LoCA just as well with moderate light, right?
Black/white is the easiest way... But close focus is maybe not the best way to test for real scenarios. Both CA and LoCA is bound to be more pronounced (and more important maybe - by different amounts depending on lens configuration. An APO macro would do good, a telephoto lens may not be that good at MFD... But may be perfectly alright at medium/infinity distance)
My suggestion:
Two plain white papers with broad black markings (two stripes? Easy to replicate). Then you need a doorway with 2-4m clearance to the wall behind it, and same distance clearance in front of the doorway. Place one paper on the doorpost, and one on the wall you can see through the doorway. Now back off away from the doorway, the same distance as doorway to the wall behind it (3m doorway to wall means back off 3m).
Focus on close target (on the doorframe) and recompose so that the far target is centered in the picture. Take picture. Focus on far target, recompose so that the closer target is picture center, take picture.
This gives you medium distance half/double focal plane distance LoCA.
LCA can be done at the same time, just place the near target (in focus!) at the picture frame edge. Raw-convert without any type of CA correction or colour noise reduction, sharpening applied up to the sharpest you can get without ringing or "halos". Downsample to a common resolution (12MP seems high enough for almost anything you'd like to compare...).
How's that? I'd be glad to add in the few lenses that I have...
cogitech wrote:
Sounds like a good method, but not everyone has immediate access to those test charts. Aren't they horribly expensive?
Yes they are, but we don't need that extreme precision. It's perfectly OK to print an available PDF version using a common laser. That's nearly for free.
Paper clips sounds like fun though but I'm not really sure what one measures by that. Is it the lens or the sensor? And what light should be used? I don't say it is impossible. Maybe we can get a few samples and try to judge the method?
My issue is the changing weather from season to season. I'd like to be outside testing instead of some indoor set-up if possible. Lately I have discovered a new target here - I have a TV antenna on a pole about 25 feet above the house. I live out in the countryside. I do a shot of the sunlit metal antenna against the blue sky. Should work year round given the sun is out.
philber, yes that's a good test of CA, but not for LoCA. (we probably do really need two tests; one for CA and one for LoCA).
Yes, it would be difficult to standardize on that target, unless we all lived near that church. Even "similar" church columns under "similar" conditions wouldn't quite be accurate enough, I would think.
I guess the point would be to reduce the number of variable factors, in a way that would be easy for us all to replicate reliably.
problem is, light naturally refracts around hard edges; how much of captured result is due to lens chromatic aberration and how much has lens accurately captured?
Light is wonderful in surrounding obstacles to its path with rainbows.
For LoCA testing I like the white ruler with black markings test. It is what they use on photozone lens tests to show LoCA performance.
For CA I like the edge of a white house against a darker background and also chrome metal out in the sunlight. Making a truly standard test for CA is going to be hard.
Philbert and Cable:
The problem with that kind of scenario is that it is very hard to control exactly how much overexposure the whiteout is packing. That amount of overexposure is in part controlling how much of purple CA that is translated into the picture. The case of purple-fringing (fringeing?) is also dependent on how sensitive the camera you use is at the very ends of the visible-light spectrum. PF is mostly caused by extreme-ends (near-IR and near-UV) LoCA, which cause these "halos" around sharp contrast edges when you have full-spectrum light backlighting.
2Ev overexposure - THIS amount of PF is visible at -5Ev strength.
4Ev overexposure - almost certainly twice that amount of PF-width, since you're forcing 4x the light amount into the LoCA effect. If the effect was at -5Ev at two pixelwidths before, it will now be at -3Ev at the same pixelwidth, and at -5Ev two pixels further into the shaded area.
PF is a kind of LoCA, but one of the harder ones to measure/compare with any usable precision.
How about shooting straight objects in front of a softbox?
I think the trick is to come up with a lighting diagram with set positioning of lights, distances, camera settings etc. That way folks in different parts of the world (different weather seasons etc) could try to consistently replicate as many of the variables as possible.
Of course not everyone has lighting gear, but a controlled test procedure will yield the most consistent results that would "speak" to Alt Forum members...
