M9 + 50 Lux Asph - Too much purple fringing?

[jaapv]

Advertisement (gone after registration)

I find it fascinating, and I thought up a test to see what causes the fringing. If it is optical, there should be a green fringe on the opposite side , invisible because it is dark. But - it should be visible in the A channel of LAB. I looked - and there is none. So in this case it is indeed demosaicing fringing.

[tgray]

That would only be true if it were lateral chromatic aberration.

[mjh]

Here's another link to toothwalker's brilliant site with examples of longitudinal chromatic aberration causing purple and green fringing. His explanations are fully referenced to accepted textbooks, experts and manufacturers, you can use his reference list to correct those people too I assume You might struggle to correct Isaac Newton though.

I am well aware of the stuff on the toothwalker.org site; it is mostly very good, no doubt about that. Still I wonder why there is so much disagreement about the cause of purple fringing if all you would need to do was to look it up in a textbook? Longitudinal chromatic aberration is one of the potential cause of purple fringing, then there is Chuck Westfall’s birefringent-microlens-theory, while the Bibble Labs CEO (and developer of the Bibble Pro raw converter) once explained to me how it was all an artifact of the demosaicing algorithm and that his own algorithm prevented it.

 

Paul van Walree’s assumption that there is green as well as purple fringing (as one would suppose if it was caused by longitudinal CA) and that we always see purple rather than green fringing because the AF was more sensitive to blurriness in the green channel is a bit weak, I’m afraid. It isn’t even clear that AF sensors of DSLRs actually work this way (contrast-based AF systems of compact digicams might). As a matter of fact, his own comparison shots refute it as the example showing purple fringing is less sharp even in the green channel, so why would the AF prefer it over the still sharper green fringing version? Furthermore, the longitudinal CA theory doesn’t explain why purple fringing only happens at high contrast edges whereas if there is transverse CA, it will always be visible, just less so if the contrast is low. Purple fringing is quite a weak effect that needs a high contrast to become visible at all. If this had anything to do with the refractive index depending on the wavelength it would affect all light of the same wavelength in just the same way, so we would see the effect even when the contrast was low. Only we don’t.

 

That monochromatic red laser light will cause purple fringing when overexposing a sensor is easily shown; all you need is a red laser. But a green one should work just as well. That the fringing is purple rather than some other color is actually an artifact of the internal image processing: Of the RGB filters, the transmission of the green filters is generally the highest, so the raw image has a greenish tint that has to be corrected for when transforming the colors from the device-dependent colorspace of the sensor to a standard colorspace. This amplifies red and blue, resulting in a violet/purple tint whenever something affects red-, green-, and blue-sensitive pixels evenly. This is also the reason why smearing is usually purple/magenta (rather than white as one might expect), or why IR contamination causes colors to shift towards magenta, rather than some other color. Neither smearing (an electrical effect) nor IR contamination do care about what color a sensor pixel is supposed to be sensitive to. Now if there is some scattering of light within the sensor (after it has passed the RGB filters), this will affect all kinds of neighboring sensor pixels evenly as well, only the signals from the red- and blue-sensitive pixels will get amplified more – thus the purple color of the fringing. Naturally such an effect will only be visible once the contrast as very high, as with normal contrast the amount of stray light is negligible compared to the light reaching the photo diode directly. There may still be other causes of purple fringing (and of course longitudinal CA can in principle cause purple fringing; that isn’t the issue), but I am convinced that this is the most important one by far in digital photography.

[jaapv]

That would only be true if it were lateral chromatic aberration.

Not at all - in that case you get red-green, not purple. The point is that the green fringe woud be inside the black, not on the other side of the dark band as with lateral CA.

[pgk]

If anyone has their neighbourhood illuminated by sodium lighting (monochromatic) and can produce purple fringing under this illumination then this should deal with the chromatic aberration theory. Unfortunatey its somewhat mixed around my way or I'd pop out and try it.

[markowich]

this is a very weird discussion..newton would not be pleased at all----))). there is no way to isolate sensor effects from lens effects, from bayer interpolation effects and from RAW converter effects. the effect we see here is certainly caused by a combination of all three, but without lens imperfections it would simply not be there. this explains why certain lenses show more of it than others. and basically all sensors that i have ever seen do show it, CCD or CMOS, no difference. the idea that it is solely caused by the sensor seems to be leica-optics-division-in-house. peter

[tgray]

Advertisement (gone after registration)

Not at all - in that case you get red-green, not purple. The point is that the green fringe woud be inside the black, not on the other side of the dark band as with lateral CA.

 

I'm not following you. I'm pretty sure you can get red-green, or blue-yellow. And by red I mean magenta-ish, so it could be called purple, and by blue, I mean... blue. Toothwalker does have some good examples of this.

 

From the examples of longitudinal chromatic aberration I've seen, I typically only see one or the other fringe around a certain piece of the image, so I'm not sure if the LAB method would always pick up the difference as you suggest.

 

I did misinterpret your statement about what you meant by 'the other side', hence my statement. Still, just because you don't see a fringe in your LAB analysis, doesn't mean that it MUST be demosaicing.

[Guest Lotw]

try CaptureOne, I would be interested

[jaapv]

I'm not following you. I'm pretty sure you can get red-green, or blue-yellow. And by red I mean magenta-ish, so it could be called purple, and by blue, I mean... blue. Toothwalker does have some good examples of this.

 

From the examples of longitudinal chromatic aberration I've seen, I typically only see one or the other fringe around a certain piece of the image, so I'm not sure if the LAB method would always pick up the difference as you suggest.

 

I did misinterpret your statement about what you meant by 'the other side', hence my statement. Still, just because you don't see a fringe in your LAB analysis, doesn't mean that it MUST be demosaicing.

If there is a magenta fringe caused by CA, the green part of the spectrum must be in the dark part of the image. In LAB one separates the luminosity channel from the colour information, so the green must show up in the A channel. If it isn't there, I think it cannot be optical CA.

[tgray]

If there is a magenta fringe caused by CA, the green part of the spectrum must be in the dark part of the image. In LAB one separates the luminosity channel from the colour information, so the green must show up in the A channel. If it isn't there, I think it cannot be optical CA.

 

I understand what you are saying. Still, try it on the examples of longitudinal CA on the toothwalker site (for lack of other 'reference' examples) and see you get. Even though what you say should be true, it might not play out like that in practice.

[thrice]

If that was the source of the purple fringing in this case, one should be able to construct a scenario where you can get only green fringing too. It's quite possible that the source of the fringing is from this, though I've read a quote from the designer of this lens that it is an APO design.

 

I own the 90mm APO-ASPH, which produces longitudinal CA, but not lateral CA. The term apochromatic as it applies to photographic lenses is used quite loosely. Toothwalker explains this on his website as well.

 

And you are correct, if the high contrast overexposed area was slightly in front of the plane of focus, one would get green fringing.

[tgray]

I own the 90mm APO-ASPH, which produces longitudinal CA, but not lateral CA. The term apochromatic as it applies to photographic lenses is used quite loosely. Toothwalker explains this on his website as well.

 

I know. I'm just throwing some facts out there. Or hearsay at least

[thrice]

I was going to try and look up some examples in my library to show some lenses will produce purple fringing quite strongly (35mm lux-R) whilst some are totally immune to it (Voigtlander 180mm APO-Lanthar) but I simply don't blow out large portions of my photographs. I will deliberately do so when I conduct the film/digital comparison, hopefully tomorrow.

[thrice]

I know. I'm just throwing some facts out there. Or hearsay at least

 

It is true that the 50/1.4 ASPH is based on the 75APO, and shares the same glass types. I think I read David Farkas' interview with Peter Karbe to which you're referring

[pgk]

Here's a detail of a photo taken on an 80mm f/1.4 R lens wide open (I've deliberately kept the file size low) on a Canon 1DS. The area with magenta fringing is in front of the point of focus whilst the (substantially more intrusive to my mind) green fringing is behind the point of focus. Whenever I try to figure out exactly what is happening I come to no final conclusion - other than that I have so far been unable to correct for this. FWIW some of my lenses seem quite prone (this 80/1.4 and the Canon 24/1.4 MkI for instance) whilst others are pretty immune (Canon 100mm macro and the 90/2.8 Elmarit M as examples). I do a substantial portion of my shooting with lenses wide open so this topic interests me.

 

I'd also say that looking at the MTF data on the 50/1.4 asph M does seem to indicate that the very corners may just show some degradation, so I presume that this is where we might notice it on FF?

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[jaapv]

It is true that the 50/1.4 ASPH is based on the 75APO, and shares the same glass types. I think I read David Farkas' interview with Peter Karbe to which you're referring

Leica does not claim APO correction for the Summiluxes, so there must be at least some CA visible. I just doubt that the rather strong effect we see is (only) CA. As soon as I get my M9 back ( should be this week) I'll start shooting the 90 AA in those conditions. With that lens we can be sure.

[pgk]

Another detail shot - this time from a 24/2.8 M asph lens on an M8 shooting below a tree canopy into bright sky. I am positive that this is not chromatic aberration which means that it must be sensor related - its not from the image corner and there are examples all over the frame. This whole issue of 'fringing' intrigues me and to date my web searching has produced nothing which I feel is at all definitive. Comments welcome. Perhaps another thread needs starting elsewhere on the topic of fringing so that examples can be posted?

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[jaapv]

Have a look here:

 

Fixing Blooming Artefacts in Photoshop (Blue and Red Halos around Lights in Nightshots) | Andre Gunther Photography

[ArtZ]

Another detail shot - this time from a 24/2.8 M asph lens on an M8 shooting below a tree canopy into bright sky. I am positive that this is not chromatic aberration which means that it must be sensor related - its not from the image corner and there are examples all over the frame. This whole issue of 'fringing' intrigues me and to date my web searching has produced nothing which I feel is at all definitive. Comments welcome. Perhaps another thread needs starting elsewhere on the topic of fringing so that examples can be posted?

 

Here're some more examples. All are crops at 100%.

This is not CA but a sensor issue... typical on CCD sensors.

 

1st picture: CV 75/2.5

2nd and 3rd picture: Summilux 50 ASPH

4th picture: Summilux 35 ASPH

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Advertisement]

Hallo,

Look here: Leica M

[tgray]

Here's a detail of a photo taken on an 80mm f/1.4 R lens wide open (I've deliberately kept the file size low) on a Canon 1DS.

 

...

 

I'd also say that looking at the MTF data on the 50/1.4 asph M does seem to indicate that the very corners may just show some degradation, so I presume that this is where we might notice it on FF?

 

That looks like CA to me. I don't know the exact set up of your shot, so I can't tell you what kind it is.

 

I don't think that is necessarily true about the MTF chart and the corners.

 

As far as the APO designation, the designer of the 50/1.4 ASPH did go on record saying it was an APO design. However, as thrice pointed out, calling a lens APO doesn't always mean much. I'm guessing some manufacturers use APO when their lens bring 3 wavelengths into focus, even if the secondary spectrum is large. I'm also guessing some manufacturers might do the opposite too; if the secondary spectrum is small, even if only 2 wavelengths are brought into focus, they still call it APO. Those last two are conjectures on my part. However, you must take APO on a lens with a grain of salt.