[Brightcolours]

I tried it with my 35mm f2 on my 450D.



In M, with the lens being "unknown" the images are a little bit darker than with the lens known, wide open with f2. Oddly enough, when trying with live view the "known" shot is a bit darker than the shot where the lens is unknown. Go figure?

[Guest]

Ok, I guess I have phrased my question incorrectly. Article shows that cross "in focus" had purple fringing and slightly defocused one had green fringing, after that it is stated that:



"The other colors of the spectrum are left defocused and add up to a magenta fringe. For this reason purple fringing is more common than green fringing".



Can it be that purple fringing was because the cross was slightly defocused as well, but it was on the other side of the focal plane? Like in popo's sample pictures a few posts above.

[genotypewriter]

[quote name='popo' timestamp='1291379798' post='4733']

Longitudinal CA is due to the different focus planes for different wavelengths of light right? That applies regardless of where in the frame you are hence can apply equally in the middle as well as the edges.[/quote]

Very good point!



GTW

[genotypewriter]

[quote name='Lomskij' timestamp='1291400276' post='4741']

Interesting. However I'm a bit confused by the following (longitudinal chromatic aberrations):



"As the human eye and autofocus systems are particularly sensitive to green light, both manual focus and autofocus tend to bring the green image in sharp focus. The other colors of the spectrum are left defocused and add up to a magenta fringe. For this reason purple fringing is more common than green fringing."



Is that so?

[/quote]

I can't see how this can be true for phase detect AF. Also if you're focusing on something black and white, the human eye can't differentiate between red, green and blue. But you still get fringing.



It's true that a Bayer sensor sees more contrast in green but I doubt it's the cause for PF because, like I said, PF doesn't exist with well corrected lenses.



So far the most suspicious thing to me is LoCA reacting with surrounding photosites.



GTW

[popo]

[quote name='Brightcolours' timestamp='1291420436' post='4748']

I tried it with my 35mm f2 on my 450D.



In M, with the lens being "unknown" the images are a little bit darker than with the lens known, wide open with f2. Oddly enough, when trying with live view the "known" shot is a bit darker than the shot where the lens is unknown. Go figure?

[/quote]

I just tried it myself with the same (450D, 35mm f/2) set to M mode (actual f/2 1/15s ISO100).



Put camera on tripod in front of my monitor, put up a grey background and focused to infinity. With lens connected normally the in camera histogram was slightly higher than with lens twisted. This was done twice to rule out any shot to shot variation. I repeated this in live view, same result as before. That is, with the lens connected normally it was slightly brighter histogram than twisted.



Just to rule it out, I did it again with fixed white balance as I think there might be a slight colour shift too. Didn't change results.







Example histograms from one of each.



Edit: Repeated the test at f/8 1s. Here the twisted shots were fractionally brighter than the normally connected ones, but by a much smaller degree than above. It was repeatable and not shot to shot variation. So this seems to confirm there is some aperture dependant effect going on.

[PuxaVida]

[quote name='popo' timestamp='1291456994' post='4753']

I just tried it myself with the same (450D, 35mm f/2) set to M mode (actual f/2 1/15s ISO100).



Put camera on tripod in front of my monitor, put up a grey background and focused to infinity. With lens connected normally the in camera histogram was slightly higher than with lens twisted. This was done twice to rule out any shot to shot variation. I repeated this in live view, same result as before. That is, with the lens connected normally it was slightly brighter histogram than twisted.



Just to rule it out, I did it again with fixed white balance as I think there might be a slight colour shift too. Didn't change results.







Example histograms from one of each.



Edit: Repeated the test at f/8 1s. Here the twisted shots were fractionally brighter than the normally connected ones, but by a much smaller degree than above. It was repeatable and not shot to shot variation. So this seems to confirm there is some aperture dependant effect going on.

[/quote]



Did you also try - both aperture known & unknown- @ ISO 200 or higher?

[Brightcolours]

[quote name='genotypewriter' timestamp='1291435398' post='4752']

I can't see how this can be true for phase detect AF. Also if you're focusing on something black and white, the human eye can't differentiate between red, green and blue. But you still get fringing.



It's true that a Bayer sensor sees more contrast in green but I doubt it's the cause for PF because, like I said, PF doesn't exist with well corrected lenses.



So far the most suspicious thing to me is LoCA reacting with surrounding photosites.



GTW

[/quote]

If different colours have different focus planes, it may well be that AF systems are made to be more receptive to green in order to get the best results for us, the viewers of the results. Of course, black and white are not colours of light, so the same applies.



PF and LoCA are NOT the same thing. LoCA is green and magenta.

PF is purple to blue-ish (usually, it can also appear towards red oddly enough).

Nice PF example from the Sony 16mm f2.8 fisheye:



PF in lovely "orbs":



PF like we know it from compact digitals:





Longitudinal Chromatic Aberration:



LoCA from Canon EF 100mm f2.8mm macro:





Photosites do not "react" to eachother, LoCA is lens dependent. PF is lens dependent too, and seen mostly with cheaper compact digital cameras. Manufacturers have made changes in back element coating to combat PF, Sigma has introduced the DG(DC) line and Tazmron the Di(Di II) line specifically for PF (not LoCa). Nowadays you do not see PF very often anymore. You do see a lot of LoCA still, specifically with wide aperture lenses and macro lenses.

LaCA ("normal" CA) is mostly seen with standard zooms and wide lenses, but also with other lenses. Some brands show laCA more (often) than others (Nikon, Tokina, Pentax come to mind).

[popo]

[quote name='Brightcolours' timestamp='1291470806' post='4755']

PF and LoCA are NOT the same thing. LoCA is green and magenta.

[/quote]

I'm starting to wonder if "PF" is simply a sub-case of LoCA. See it as a parallel to the SLT ghosting effect. The relative intensity of the fringing in lower contrast scenes is different than the case you have very bright highlights.



I have an idea for a test I can try at home to show this...



Alternatively, there is a possible way to disprove it too. Anyone have a photo showing PF clearly in front or behind the focus region that isn't due to lateral CA, and also a photo from the same lens showing LoCA colours in the respective defocus region? If my suspicion is correct, then the PF can only happen where the LoCA is magenta-ish, and shouldn't elsewhere.

[Brightcolours]

[quote name='popo' timestamp='1291475496' post='4756']

I'm starting to wonder if "PF" is simply a sub-case of LoCA. See it as a parallel to the SLT ghosting effect. The relative intensity of the fringing in lower contrast scenes is different than the case you have very bright highlights.



I have an idea for a test I can try at home to show this...



Alternatively, there is a possible way to disprove it too. Anyone have a photo showing PF clearly in front or behind the focus region that isn't due to lateral CA, and also a photo from the same lens showing LoCA colours in the respective defocus region? If my suspicion is correct, then the PF can only happen where the LoCA is magenta-ish, and shouldn't elsewhere.

[/quote]

I have no idea why you wonder why PF is LoCA. It looks different, the colour usually is different, and where LoCA shows magenta and green, PF always is on its own in one colour (from blue/violet to red-ish).

[popo]

[quote name='Brightcolours' timestamp='1291476531' post='4757']

I have no idea why you wonder why PF is LoCA. It looks different, the colour usually is different, and where LoCA shows magenta and green, PF always is on its own in one colour (from blue/violet to red-ish).

[/quote]



Turning that around, why is there a need for a different mechanism when LoCA seems to me adequate to describe both. I haven't seen anything yet that suggests it isn't.





Full test output





Crop from bottom of top-middle.





Crop from top of bottom-middle.



Anyway, here's the result of my test. I created a test image with relative levels of 0% background, 25% for lower star backgrounds, and 100% for upper star backgrounds. Put the EF 85mm f/1.8 on the 450D, and manually focused on the vertical point of the middle stars at f/1.8. The lower contrast stars at the bottom show the expected regular LoCA colours, red/magenta-ish in front of the focal plane, and green-ish behind. Looking at the top 3 there's a lot of magenta "PF" on and in front for the focal plane. It's more interesting behind the focal plane, with a tiny amount of blue before going green mixing into yellow. The green/yellow thing seems dependant on the direction of contrast change with defocus.



This was obtained with one lens in one shot. It shows LoCA and PF happening, the only difference between them is the lightness of the region.



I'm open to discussion if there may be anything I've overlooked or misinterpreted here. Different lenses may exhibit different colours so could have the same test performed to see if that remains the case. I can repeat this for my other lenses on request.