Purple Reflection

[Lindolfi]

Advertisement (gone after registration)

OK, now concerning the theory I put forward in this post: click: I was wrong.

 

It has nothing to do with coating at all, which is shown by this experiment:

 

I just shot many overexposed backgrounds with many lenses and the only way I can reproduce anything near what was shown by ArtZ is this

 

 

And that was only possible at wide aperture with a non-apo lens, the 75/1.4 at f/1.4

 

Now to show that this heavy fringe is also caused by chromatic aberration, I focussed in front of the object, which generates green fringes:

 

 

(images at about 70% size and sections from the middle)

 

The halos Jaap is talking about in for instance images from a starry sky (like in this shot I made: click) is also caused by chromatic aberration (again shown by focussing closer by, turning the purple halos into green ones).

 

Don't forget that the blue and red may extend quite far into the dark area at a contrast point (around a star for instance), while this is not normally visible in images with gradual exposure, since the neighbouring light parts simply mask it.

[denoir]

Uh oh ... you are confusing longitudinal chromatic aberration and purple fringing again. The former occurs on film and on digital both. The latter on digital only.

 

Which part of the references that I posted do you have difficulty understanding? PF as Puts, Chambers as well as "The Manual of Photography", and all the others agree on is LoCA. Over and over again. Pick any book on optics.

 

This is the continued absurdity of this thread - that people completely ignore that this is basic optics and that there is nothing mystical about it. This could easily qualify for a Monty Python sketch. It doesn't matter how much we speculate back and forth - this is all trivial optics that stopped being controversial some 300 years ago. You might as well be arguing that the earth is flat.

[mjh]

This is the continued absurdity of this thread - that people completely ignore that this is basic optics and that there is nothing mystical about it.

That’s probably because people knew about longitudinal CA and were familiar with it, but still couldn’t make heads or tails of this new purple fringing phenomenon that started to rear its ugly head when photography went digital. So all kinds of explanations were offered like the (admittedly unlikely) birefringent-microlenses-hypothesis put forward by Canon’s Chuck Westfall.

[jaapv]

Here is an example of all colors of CA:

See how nicely it shows up my backfocus. I saved it just before it went into the bin. (DMR, VE 105-280)

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!

[Lindolfi]

Interesting example, Jaap. When you study the original you will most likely find that the area of optimal sharpness is in the purple area at the level of the top of the wing of the left bird in the water behind it. The significance of that is that in optimally sharp images you will find purple rather than cyan/green fringes around highlights and that is again related to the dispersion of the glass and the attempt to correct for it.

[Lindolfi]

This might also explain why a wide aperture and a fast shutter speed can lead to a different (higher) level of purple fringing than a smaller aperture and a slower shutter speed, even when the resulting exposure value remains constant. In the former case the drainage system briefly gets overloaded; in the latter, not.

.

 

That statement can also be tested by lowering the light level, keeping the aperture open and lengthening exposure time. If it is true what you say, it is the shutter speed which generates the problems, not the aperture. Have tested it and it is not true: the aperture is decisive, not the exposure time, as long as the highlights are overexposed. So the "briefly overloaded" is not the sensor factor that is responsible. Here the same scene exposed 1000 times (!) longer on the left (using a 1000x ND filter) than on the right: no difference in purple fringing. Lens: Summilux 75/1.4 at f/1.4. Left exposure 23 sec. right 1/45 sec.

 

[denoir]

Advertisement (gone after registration)

A couple of things - first the blue edge in the images in that other thread. I'm pretty sure it's simply a PP error - excessive CA correction where none should have been applied.

 

Take a look at this: a 28 Cron ASPH shot, stopped down a bit with no traces of purple fringing or any CA:

 

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!

 

100% crops. On the left side is a direct conversion from raw and on the right CA correction has been applied in Lightroom:

 

So by trying to correct for non-existent CA you can easily add a fringe.

 

Now as for film, the shot below was taken with a Pentax 67 medium format camera with Ektar 100 film:

 

 

Crop:

 

As you can see there are definitely traces of PF here as well - although far less dramatic than you would have gotten with digital. The cause of the fringing is uncorrected LoCA but how strong the effect will be depends on other factors than just the lens.

[zlatkob]

For anyone who hasn't found it ... in Lightroom's Develop module, set Defringe to "Highlight Edges". The same adjustment is also available in Adobe Camera Raw under the Lens Corrections tab.

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!

[Lindolfi]

Thanks for the input, denoir. I don't think the purple fringing in the image by ArtZ is from incorrect CA compensation in postprocessing. That looks different with also cyan edges. The edges are quite broad compared to the ones I can generate under extreme conditions and they are quite blue compared to mine. Mystery can only be solved under controlled conditions, so that we can exclude a number of things.

 

Anyway if I read this whole thread again, it is time for a summing up of what we think about purple fringes.

 

There is experimental evidence put forward and there are also arguments which are hard to test, but presently I think the strongest case can be made for the primal cause of the purple fringe to be chromatic aberration of the optics (excluding the micro lenses of the sensor).

 

If we look at the example of denoir of the headlight of the car and other examples on the net: purple fringes are also found on film, but less enhanced. So that means there is also a role for the sensor there. How that works is still a matter of debate. One can put forward a story, but if there is no way of testing it, there are other stories to tell which stand next to it. Then it becomes a matter of opinion and that is not satisfying if we want to know the true cause.

 

An experiment I have thought of is to throw white light on the sensor with a hard shadow line, not coming from optics. This may be done by covering the sensor partly (without interfering with the action of the shutter) and using a point light source. Again if purple fringing is generated by the sensor, it should do so in this experiment and if not, it should not do so.

 

If I find the time, I'll do that experiment.

 

For now, the strongest case that can be made concerning purple fringing can be formulated as:

 

"purple fringing is caused by chromatic aberration and enhanced by some unknown sensor related process"

[jaapv]

For your high-contrast experiment I would advise testing an oblique 100% contrast line.

[Lindolfi]

Well, the experiment with covering the sensor partly while a point light source shines on it has been done.

 

The setup:

 

[1] M9 on tripod with a strip of black smooth and straight cut material partly covering the flange. No lens mounted.

 

[2] The front of a high power white cree led torch was covered with a cap in which a pinhole was made with a diameter of 0.1 mm

 

[3] The torch was placed at 3 meter distance. This gives a geometrical spread of the 0.1 mm light source over the sensor of 1 micron (0.001 mm), since the focal-flange distance is about 3 centimeter (1 % of the distance of the torch)

 

[4] The exposure of the M9 sensor was set to create mid grey of the uncovered part to check the shape of the schadow on the sensor (image [ A ])

 

[5] In a new image, the exposure was set to 6 stops longer to get heavy over exposure and to study the effect on the shadow area. (image [ B ])

 

These are the results (at 100%)

 

[ A ]

 

[ B ]

 

As you can see, the pinhole light produces wavefronts at the edge of the cover, which spread out over much more than 1 micron (the pixel pitch is about 7 micron) in image [ A ]

 

In image [ B ] you can see that the wavefronts start to eat into the shadow, producing all colours of the rainbow due to uneven excitation of colour wells of the Bayer matrix that are partly covered.

 

But although there is a huge overexposure, there is no purple fringing.

 

So also this experiment shows that the sensor can not produce purple fringing on its own, it needs a lens with chromatic aberration.

[Lindolfi]

The experiment that I did with the "knife-edge" in front of the sensor is pretty conclusive, but the wave diffraction at the distance of 30 mm is quite big, so you do not get a very sharp boundary on the sensor.

 

But, here is something for you: you already have such an experiment in dozens of images you have made yourself!

 

Select an image with a white background (like a cloudy sky) and a nice big flake of dirt on the sensor in that white area of the image. The flake is on the cover glass of the sensor, so less than a millimeter in front of the sensitive parts of the sensor.

 

Well for this purpose I'm willing to hang my dirty sensor outside: After some digging I found a nice big piece of dirt in an almost white background and the image of the side of a building with some purple fringing. It was a 21/2.8 asph at f/11. Here part of the image at 100%. It is at the outer left boundary of the image.

 

 

 

Do you see purple fringing around the dirt flake image?

 

Well, again more proof that the sensor does not produce it. And the great thing is: you don't have to believe me, you can just look in your own collection.

[Beyder28]

For anyone who hasn't found it ... in Lightroom's Develop module, set Defringe to "Highlight Edges". The same adjustment is also available in Adobe Camera Raw under the Lens Corrections tab.

 

Does anyone know where in Aperture you can do this?

[Rick]

Lindolfi - pretty clever. I suppose the only thing that the dust example doesn't take into account is the angle of incident of the light hitting the sensor.

[Lindolfi]

That is true, RickLeica. Depends on the location of the dirt and the focal length. With my 0.1 mm point lightsource and an obstruction I place on the sensor, I would be able to make the light come from any direction, but that will not change the principle. The dust images I found all have a colourless boundary, independent of aperture, location on the sensor or focal length. So I am not ready to do that point source experiment (yet).

 

Until the advocates of a sensor related origin of purple fringes come up with better theories and experiments, I'll just leave the situation as it is for now.

[jan_brittenson]

Blooming of sensors is a real phenomenon, particularly known from CCDs used in astronomy, so the process of carry-over of light levels to nearby colour wells should not be readily dismissed.

How would this be purple? In fact, how would it be anything other than just a blurring effect? Basically, what you're describing is a gaussian blur [applied before demosaic].

[Lindolfi]

How would this be purple? In fact, how would it be anything other than just a blurring effect? Basically, what you're describing is a gaussian blur [applied before demosaic].

 

It is not a Gaussian blur in spatial sense: blooming usually has directionality. In this example http://www.l-camera-forum.com/leica-forum/leica-m9-forum/130720-m9-faqs-frequently-asked-questions-answers-3.html#post1742883 you can see that. And you are right that also all colours are involved, making the blooming be white, with the exception of the far borders, where it is a chance effect if certain colour wells happen not te be overflown. In this example you can see coloured speckles at the border, wich may be due to that chance effect.

 

The directionaility is known from the use of CCDs in astronomy.

 

As I wrote in my summing up post, the role of the sensor in enhancing the effect of chromatic aberration is yet to be explained. And presently I do not see how blooming can do that, given its directionality.

[jaapv]

This is a thread about the nature of purple fringing.

Posts about unrelated subjects will be (re)moved !

[Lindolfi]

Here is another consideration about chromatic aberration, purple fringing, sensors and film.

 

One of the questions we still have is why film is less prone to show purple fringing than sensors with the same lens. Now you could put forward that colour film has three colour layers (sometimes four) that are arranged in blue, green and red from top to bottom. Simple chromatic aberration consists of a spreading of the focal point from blue near the lens to red further away from the lens. And so you might argue that the colour layers of film are nicely arranged to compensate for the chromatic aberration: the blue focal point would fall on the blue layer and the red focal point on the red sensitive layer.

 

Now that would be the reason for film to be less prone to purple fringing. However, there are two problems with that story

 

[1] Usually lenses are achromats in which blue and red coincide, so the red and blue layer separated would make things worse than in a sensor (where they are in one layer).

 

[2] The colour layers in film do not take care of lateral chromatic aberration, only longitudinal aberration (were it not for objection [1]). The only way to get around that is to have a spherical film, like in the eye of a jumping spider in which indeed the colour sensitive cone layers are separated to compensate for the chromatic aberration of the lens of the eye of the jumping spider. In that respect the eyes of a jumping spider are millions of years ahead in design to that of our eyes. (the reference is here: click)

 

Anyway, we are still in need of an explanation of the sensitivity of digital sensors to the chromatic aberration of our lenses, compared to that of film.

[Advertisement]

Hallo,

Look here: Leica M