Jump to content

Purple Reflection


Beyder28

Recommended Posts

Advertisement (gone after registration)

Thanks a lot for the postings on this subject. Really enjoy the insights brought forward.

 

Longitudinal and lateral chromatic aberration are caused by the same property: dispersion of wavelengths by the dispersive power of the types of glass used. The first correction (achromatic) brings red and blue together, leaving purple (blue + red) and green/cyaan still dispersed. Apo correction is the second order correction bringing three wafelengths together in the red, green and blue area. The severity of the chromatic aberration depends on the design of the lens, whether it is an achromat or apo lens. But the distribution ( more long or more lateral aberration) depends on the focal length of the lens: longitudinal aberration simply becomes lateral chromatic aberration when the rays become more oblique.

Link to post
Share on other sites

x
  • Replies 98
  • Created
  • Last Reply

'Chromatic aberration' points at a known lens error that is deeply understood by lens designers and can be reproduced using computer simulations of lenses. 'Purple fringing' is the name of something that occurs in images and is caused by a mixture of causes among which chromatic aberration. Sensor properties are also of importance.

You can show this by doing experiments: take an achromat and apo lens of the same focal length and expose with the same aperture and exposure time, taking an image of a problematic subject such as the necklace shown by the OP and inspect the purple fringing to see if chromatic aberration has anything to do with it.

Likewise you can leave the lens the same and compare two sensors or film and sensor. That way you can seperate the factos involved in purple fringing.

Link to post
Share on other sites

Absolutely amazing. I'm stunned. There is nothing mystical about the cause of purple fringing - it's longitudinal CA.

 

And no, an apochromat is in no way a guarantee that it will remove it. In fact the modern M mount APOs don't. APO technically means that it has been corrected for two wavelengths. Superapochromat for four etc The lens manufacturers, including Leica typically just mean that they've added some form of corrections. In the 90mm case, LoCA is unchanged.

 

Left 90 Cron V3 (non-APO), right 90 Cron AA (APO):

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!

 

I said that I'm stunned, and I mean it. Not that people don't know what CA and purple fringing is, but that they can't take in a very simple example that demonstrates the fact and perpetuate the myth that it's something mystical and sensor related. I don't know who started that myth, but it's as false as it is common. I suspect the wikipedia article on the subject may help spread it as it is vaguely misleading, but I doubt it started the whole thing.

 

You can easily look it up in any proper book on optics. The "Manual of Photography", Elizabeth Allen et al 10th edition for instance has a chapter on CA that explains it very well.

 

If you want something simple and online:

Chromatic aberration

 

Back to the example once more:

 

Again, this is at f/2, non-APO and APO side by side:

 

And this is at f/5.6:

 

The sensor is the same, the EV is the same. The difference is that in the second example the lens has been stopped down removing the LoCA and thus the purple fringing. The sensor is the same as is the exposure - the only thing that has changed is the aperture. You will never ever find purple fringing at the center of the frame when the lens has been stopped down enough, regardless of the exposure.

Link to post
Share on other sites

There is nothing mystical about the cause of purple fringing—it's longitudinal chromatic aberration.

No, it's not. And it won't become that way by repeating yourself a few times. If purple fringing was nothing but exaggerated (through over-exposure) longitudinal chromatic aberration then it would occur in green about as often as it does occur in purple. But it doesn't. It's alway purple and never green. So purple fringing is a different phenomenon.

Link to post
Share on other sites

No, it's not. And it won't become that way by repeating yourself a few times. If purple fringing was nothing but exaggerated (through over-exposure) longitudinal chromatic aberration then it would occur in green about as often as it does occur in purple. But it doesn't. It's alway purple and never green. So purple fringing is a different phenomenon.

 

No it wouldn't. Violet has a shorter wavelength than green and thus more energy and a sensor will pick it up more easily. Violet ("purple") is in the 400 nm range while green is around 550 nm. Green fringing does indeed occur, but less frequently because of its lower energy. Subsequently because of the shorter wavelength it's also more difficult to design a lens that corrects for it.

 

I have provided you with examples and external references. You have done nothing except made incorrect statements. Please provide a counter example, an alternative explanation of the images I've shown, or simply a credible scientific reference. Otherwise please stop posting misleading and false information.

 

Edit: The Digital Photography FAQ summarizes it quite well:

 

Purple fringing (PF) has been a subject of great debate in the digital photography community. A popular, but demonstrably incorrect, theory for some time was that PF was caused by sensor blooming. The following suggest strongly that blooming is not a factor:

  • PF decreases with smaller apertures and constant EV.
  • PF does not occur along readout lines in CCDs (as is typical with blooming) and CCDs are no more susceptible to PF than CMOS sensors, which have high blooming resistance.
  • PF remains the same with higher ISO and constant EV. (At high ISO only a fraction of the sensor's well capacity is used, so blooming is far less likely.)
  • PF varies with position in the frame
  • PF is typically radial.
  • PF is colored

Link to post
Share on other sites

Advertisement (gone after registration)

It's alway purple and never green. So purple fringing is a different phenomenon.

 

 

No, it is not always purple. Here in a picture of a necklace both purple and cyan (green) fringing. (100 procent section of an image with the 75/1.4 at f/1.4)

 

cyanpurplefringing.jpg

 

Explanation: the beads are either inside, in front or behind the plane of sharpest focus and the chromatic aberration creates the colours outside the on average white highlights. So why is purple more common: the human eye is most sensitve to green. So "good focus" focusses on green, while blue and red remain defocussed.

Link to post
Share on other sites

Here is another experiment, to show that the sensor indeed works differently with overexposed highlights than with well exposed ones in terms of purple fringes. The cause is the chromatic aberration and the sensor enhances it.

 

The first images is a composite of five exposures with the 75/1.4 on the M9 at f/1.4 of a small shiny disk on black velvet in the middle of the image exposed from left to right with 1/30, 1/90, 1/180, 1/350 and 1/750th of a second. Images at 100%. You can see that with the same lens producing the longitudinal colour shift, it can best be seen at the overexposed highlight and not in the correct exposed ones. If it only would be the lens producing the colour shift and the sensor would not enhance it, the purple would have to be visible in the other images as well.

 

purplefringatEV1.jpg

 

Now, you may say, "but perhaps the purple is hiding in the blacks in the short exposed images". Well that is partly true, but if you bring all images to the level of the first long exposed one, you do see that still the purple fringing is less in the short exposed ones.

 

 

purplefringatEV2.jpg

 

And so that means that the sensor does play a role in purple fringing, but the cause is chromatic aberration, as is shown by the highlights with cyan/green fringing I showed in my former posting.

Link to post
Share on other sites

No it wouldn't. Violet has a shorter wavelength than green and thus more energy and a sensor will pick it up more easily. Violet ("purple") is in the 400 nm range while green is around 550 nm. Green fringing does indeed occur, but less frequently because of its lower energy.

Your own image purportedly demonstrating purple fringing contradicts this: It clearly shows purple in front of the plane of sharpness and green behind that plane. Both green and purple are clearly visible (almost equally strong in fact). In this forum we have seen lots of examples for this phenomenon, i.e. a switch of fringing colour in front of and behind the plane of sharpness. That’s longitudinal chromatic aberration.

 

Purple fringing is different in that it is always purple and it often occurs even in the plane of sharpness where longitudinal CA is generally well corrected for.

Link to post
Share on other sites

Your own image purportedly demonstrating purple fringing contradicts this: It clearly shows purple in front of the plane of sharpness and green behind that plane. Both green and purple are clearly visible (almost equally strong in fact). In this forum we have seen lots of examples for this phenomenon, i.e. a switch of fringing colour in front of and behind the plane of sharpness. That’s longitudinal chromatic aberration.

 

Purple fringing is different in that it is always purple and it often occurs even in the plane of sharpness where longitudinal CA is generally well corrected for.

 

Where do you see a trace of green here (excluding the color of the car of course)?

 

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!

 

I guess you are referring to the focus chart shot. That's LoCA outside of the plane of focus and there you'll of course see both green an purple. Don't confuse LoCA in the OOF regions with the one you see at the focal plane. Purple fringing is the latter - LoCA still being the cause. Further more focusing distance matters as well as a variable.

 

And again, a simple demonstration, that I'm providing for the third time now is that it disappears when the lens is stopped down:

 

 

Why do you keep ignoring this very simple example that shows that it's all about LoCA that is corrected when the lens is stopped down?

Link to post
Share on other sites

And so that means that the sensor does play a role in purple fringing, but the cause is chromatic aberration, as is shown by the highlights with cyan/green fringing I showed in my former posting.

 

^^That.

 

The sensor will affect the rendering, but the cause is CA. No CA, no purple fringing.

Link to post
Share on other sites

Where do you see a trace of green here (excluding the color of the car of course)?

I was talking abour your first image, that of the focus chart. The pictures of the car window are showing purple fringing.

 

And again, a simple demonstration, that I'm providing for the third time now is that it disappears when the lens is stopped down

Yep, we all know that. But what does it prove? Purple fringing differs in several respects from longitudinal CA, one being its independence from distance (relative to the plane of sharpness) that was mentioned already. Another distinguishing characteristics is the fact that while longitudinal CA is visible even when the contrast is only moderate, as in your test shot of the focus chart, purple fringing crucially depends on overexposure (of which the specular highlights in your car window examples are a perfect example). However there is one feature that purple fringing and longitudinal CA have in common: Even when the total amount of light stays constant, an exposure with a smaller aperture (and correspondingly longer exposure time) will show both less purple fringing and less longitudinal CA. The image projected by the microlenses onto the sensor pixels is an image of the exit pupil and thus its size depends on the aperture. The larger the aperture, the larger the image projected on the sensor surface and the greater the variation of incident angles of rays hitting a given pixel. If purple fringing is caused by scattered light within the sensor (as I think it is) then this effect would be diminished when the pixel area actually receiving light was more narrow, as would be the case if the aperture was smaller. So the observation that stopping down reduces purple fringing proves nothing either way as it agrees with both explanations.

Link to post
Share on other sites

Yep, we all know that. But what does it prove?

 

It proves together with a bunch of other things that it's not the sensor. You have the same sensor and the image has the same EV but you get different results.

 

 

Purple fringing differs in several respects from longitudinal CA, one being its independence from distance (relative to the plane of sharpness) that was mentioned already. Another distinguishing characteristics is the fact that while longitudinal CA is visible even when the contrast is only moderate, as in your test shot of the focus chart, purple fringing crucially depends on overexposure (of which the specular highlights in your car window examples are a perfect example). However there is one feature that purple fringing and longitudinal CA have in common: Even when the total amount of light stays constant, an exposure with a smaller aperture (and correspondingly longer exposure time) will show both less purple fringing and less longitudinal CA. The image projected by the microlenses onto the sensor pixels is an image of the exit pupil and thus its size depends on the aperture. The larger the aperture, the larger the image projected on the sensor surface and the greater the variation of incident angles of rays hitting a given pixel. If purple fringing is caused by scattered light within the sensor (as I think it is) then this effect would be diminished when the pixel area actually receiving light was more narrow, as would be the case if the aperture was smaller. So the observation that stopping down reduces purple fringing proves nothing either way as it agrees with both explanations.

 

 

Ehm. I'm not sure how to respond to this as essentially there is not one correct sentence in this section. I'm hesitating because explaining the PF/CA issue is trivial compared to explaining some of the errors above - and I have apparently failed in the former. However, a few points:

 

* Purple fringing does not depend on overexposure but high contrast. In good lenses the contrast has to be extremely high, but not so with lenses with poorer correction. Here's an example with a Canon 85/1.8 lens (a notorious purple fringer), without any blown highlights:

 

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!

 

* The size of the projected image on the sensor does not in any way depend on the aperture. The image circle is constant regardless of the aperture.

 

* The image projected on the micro lenses are not an image of the exit pupil.

 

* Scattered light would appear as noise not CA.

Link to post
Share on other sites

Alright, one final attempt to demonstrate that it's lens dependent - i.e CA and not a question of the sensor.

 

First here are two crops, the first one from a 50 mm f/1.4 Canon, a notorious purple firnger and a Rokkor 58/1.2 that has exceptionally good LoCA corrections. Both are shot wide open

 

If it was a question of sensor then the f/1.2 lens should show more PF. This however is what we get:

 

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!

 

The Canon shows quite a bit of PF while the Rokkor is virtually free of it. You can make out a slight hint of purple, but nothing that comes close to the Canon lens.

 

As a second demonstration, we stop down the Canon and put it against a Zeiss 50/2 Makro Planar - also a lens with very good CA corrections.

 

 

The Canon shows a bit less PF than at f/1.4 as expected, but the Zeiss shows none - nor does it show any significant amount of green.

 

So these are two 50mm lenses shot at the same aperture, meaning the same amount of light at the same angles fall on the sensor. The only difference is in the lenses - one doesn't have good CA corrections while the other one does.

 

It's as simple as that - no need to invoke any mysterious sensor properties. And what's more, the whole sensor<->pf thing is just an internet forum myth. If you look up any credible sources either online or in books you'll see that there is no mystery about it. I've provided three or four references. I don't really know what more I can do.

Link to post
Share on other sites

Thanks denoir for taking the time to make and post these examples. They agree with what I found in my tests over the years. But If you look at the experiment with five exposures I posted, you probably agree that it points at the sensor being involved in enhancing the effect. That probably is the basis of the internet myth about the source of the purple fringes. With the advent of digital camera's people started to notice them.

 

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.

 

The lack of an AA filter has an influence on parts of the image with spatial frequencies just below the Nyquist frequency of the sensor, where they can also produce some wild colour artefacts, but usually they cover the whole gamut of the raiinbow, since it is a matter of chance whether color wells in the area of a black-white edge in the image are receiving part of the white light at that edge.

 

Anyway, I think it is quite clear what the cause is of the purple fringes that the OP showed us.

Link to post
Share on other sites

Thanks denoir for taking the time to make and post these examples. They agree with what I found in my tests over the years. But If you look at the experiment with five exposures I posted, you probably agree that it points at the sensor being involved in enhancing the effect. That probably is the basis of the internet myth about the source of the purple fringes. With the advent of digital camera's people started to notice them.

 

Absolutely. It is the sensor that has to deal with all that light in the end and very often in the PF case one is pushing its boundaries. I definitely agree that the sensor contributes to enhancing the effect. In the case of the M9, I'm also pretty sure that the lack of an AA filter helps accentuate the edges of the fringes, making them more apparent.

 

What I've just been saying is that CA is the cause of it.

 

Anyway in practice the solution to getting rid of PF is to a) stop down the lens B) use a lens with better corrections or c) remove in post processing.

Link to post
Share on other sites

The image circle is constant regardless of the aperture.

What else does it take do prove that you are having no idea what you are talking about?

 

By the way, did you notice that the Canon and Minolta images are focussed at different distances? So are the Canon and Zeiss images.

 

 

Alright, one final attempt to demonstrate that it's lens dependent - i. e. CA and not a question of the sensor.

Hey ... no-one denies it's lens- or aperture-dependent. Still, this little factoid alone does not prove it's the same as longitudinal chromatic aberration.

 

 

If it was a question of sensor then the f/1.2 lens should show more PF. This however is what we get ...

Purple fringing is a multi-factorial phenomenon—and thus way beyond anything you can grasp.

 

My problem is, I cannot fully and exhaustively explain what purple fringing is. But if I know one thing for certain then it's this: The real world is always more complex than you dare to believe. And purple fringing and longitudinal chromatic aberrations may share a few factors ... but they definitely are not the same thing.

 

And by the way—DxO is not a reliable online source ... after all, they invented DxO Mark which from the point of view of a photographer is the most stupid thing since Jackass :eek::p

Link to post
Share on other sites

What else does it take do prove that you are having no idea what you are talking about?

 

A bit of advice is to think a bit before posting.

 

The image circle always covers the full area of the sensor/film - regardless of aperture. Again, think about it just for one tiny second - do you think the projected image gets smaller when you use a smaller aperture? That at max aperture you are using the full size of the frame while at say f/16 you are just using a few pixels in the middle?

 

No actual knowledge is required, just a bit of logical thinking.

 

 

My problem is, I cannot fully and exhaustively explain what purple fringing is.

 

That part is perfectly clear. The question is why you then insist on spreading false information. There are people that can fully explain it and have done so.

 

Again, "The Manual of Photography" Elizabeth Allen et al covers it well as do a number of other books on optics. DxO mark is not a good benchmark, true. But that's because they weigh the different lens aspects in a weird way. That does not mean that they don't know optics. It is also not the only reference I've given.

 

Anyway, I've said my part, shown examples and given references. I can't do anything more so I'm done here.

Link to post
Share on other sites

A bit of advice is to think a bit before posting.

If only you would follow your own advice ... :rolleyes:

 

 

... do you think the projected image gets smaller when you use a smaller aperture?

No, I dont ... :rolleyes:

Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...