Towards an explanation of the Italian Flag Phenomenon

[Xmas]

Advertisement (gone after registration)

My CV 15 was not fully corrected by 1.160. Not by a long shot. I doubtd that 1.162 will add much.

 

Is some one going to try the CV15 with diffeent Leica lens selections?

 

Noel

[otto.f]

All of a sudden this thread is about the new firmware, while all sensible tests in the quest for the explanation of the Italian Flag Phenomenon should be done without using the correction processing by the firmware (so with lens coding off). Anyway, it is of course interesting to see what the new firmware does, so I will eagerly run some tests.

 

What I like most about the file "Improvements in the Firmware Version 1_1.62_en.pdf" is the sentence

 

"Improvements in Italian translations"

 

Now that of course is right on topic in this Italian Flag Phenomenon thread!

 

Indeed is now that it is 'fixed', not at all proven that the phenomenon is understood

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[otto.f]

Is some one going to try the CV15 with diffeent Leica lens selections?

 

Noel

 

No we keep it to ourselves

[t024484]

Here is the result with the new software taken with the pinhole.

Lindolfi, you are right that new software cannot help in explaining IFP, but nevertheless I was curious to see if it makes a difference.

 

With exactly the same set up as with my last picture, I can definitely see a difference, because vignetting is less.

 

First is the full image and second the same image with only a circle in the middle and the rest blanked out.

 

Hans

 

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!

 

Edited June 15, 2011 by t024484

[Lindolfi]

Thanks Hans. What is crucial here is: did you have any manual coded lens switched on or not when taking the image with the pinhole?

 

Here is a comparison with and without coding with a 28/2.8 asph at f16 and uncoded there is a strong IFP as was to be expected since uncoded nothing is corrected.

 

Edited June 15, 2011 by Lindolfi

[Olsen]

I have just tested my CV15. Some improvement, but still the IFP is very visible at the left edge of the picture. The red/mangenta color does not seem to protrude so far into the picture as it previously did, though.

 

My WATE is - and was, perfect at 16 mm. What a lens, by the way!

[t024484]

Advertisement (gone after registration)

Thanks Hans. What is crucial here is: did you have any manual coded lens switched on or not when taking the image with the pinhole?

 

I took pictures with lens detection in auto and in off position.

In the automatic setting the camera reported the pinhole lens as uncoded.

Both settings gave me exactly the same result.

 

Hans

[t024484]

Lindolfi,

 

This is my 28mm/f2,8 at f16 uncoded.

Your image and mine are almost identical, after having raised the exposure of your picture short before clipping in Camera Raw.

 

Hans

 

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!

Edited June 15, 2011 by t024484

[t024484]

And this is an uncoded CV 15 / f4,5 at f16.

A real beauty as far as IFP concerns, the best so far.

 

Interestingly, there is hardly any difference between the uncoded 28 / f2,8 and the pinhole.

 

Hans

 

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!

Edited June 15, 2011 by t024484

[Lindolfi]

The pinhole exposures with the old and new firmware uncoded give me very little difference: top old firmware, bottom new. Now I have used your method of using the center not only for white-balance but also for scaling the levels to maximal values in the middle (producing that horrible little white crater in the middle).

 

[Lindolfi]

And this is an uncoded CV 15/f4,5 at f16.

A real beauty as far as IFP concerns, the best so far.

 

Hans

 

Yes Hans, I agree. Noticed that early on, so I put it back on my old Bessa body and not use it on the M9. The CV 12/5.6 is much more gentle with IFP.

[t024484]

The pinhole exposures with the old and new firmware uncoded give me very little difference: top old firmware, bottom new. Now I have used your method of using the center not only for white-balance but also for scaling the levels to maximal values in the middle (producing that horrible little white crater in the middle).

 

 

Lindolfi,

 

Obviously our pinholes cannot be used as a reliable tool to compare cameras.

But since our images with 28mm/f2.8 at f16 are almost identical, there seems to be little difference between sensors.

When aligment of the microlenses had been an issue, it would have become obvious here.

So alignment of microlenses can probably be deleted from the list as a cause of IFP.

 

The only thing remaining I see is the point you raised earlier: different pixel sensitivity depending on wavelength for light coming from different directions, which increases when the angle of incident becomes larger.

 

Hans

[otto.f]

And this is an uncoded CV 15 / f4,5 at f16.

A real beauty as far as IFP concerns, the best so far.

 

Interestingly, there is hardly any difference between the uncoded 28 / f2,8 and the pinhole.

 

Hans

 

[ATTACH]262833[/ATTACH]

 

At F8.0 though I can really live with it, setting it as Elmarit 28 non-asph, this is if you PP in LR3 and not in C1. BTW, I never used the CV 15 below F8.0 with the least acceptable quality, not on film , not on M8 either

[Lindolfi]

(Just a small correction Hans: I don't own a 28/2.8. I have tested the 21/2.8.)

 

Hans I agree with the fact that everything points at the colour wells (including their colour filters) being differently sensitive in different directions and that that difference differs for the three colours. The exact micro reason for this (presence of light reflecting electronics or reflectivity of the sides of the small square filters) is not so important now: we don't have enough data on the properties of these wells and my 60/0.85 achromat is not going to show it, even if I can squeeze sufficient light between the objective and the cover glass of the sensor. The shifted pinhole experiments show that it is not the place on the sensor but the direction and angle of rays which are conditions which evoking IFP.

 

My homewritten RAW converter in Matlab does show a difference with the RAW conversion from LightRoom: there is less IFP in my conversion. So I believe part of IFP is enhanced by the conversion. But the origin is most likely in the colour well layer and not in the microlens layer and not in the cover glass or their combination, nor in the optics themselves ( which is shown by our pinhole experiments)

 

Hans, Otto and others who contributed: thanks very much. I really enjoyed looking into this and have learned again in a group effort on this forum. This is what makes this forum shine: learning and having fun.

Edited June 15, 2011 by Lindolfi

[otto.f]

My homewritten RAW converter in Matlab does show a difference with the RAW conversion from LightRoom: there is less IFP in my conversion. So I believe part of IFP is enhanced by the conversion.

 

Yes that's also my impression, e.g. I see a better result with LR3 than with C1

[denoir]

First guys, let me just say: what a great thread! Fantastic contributions in the forms of the various tests and theories. I missed it completely although I've been very curious about the asymmetrical nature of the color cast for quite some while now. I've talked to a couple of people that work with CMOS sensor design and to my surprise they weren't really able to answer it. My own pet theory was that the micro lenses were diagonally offset from the pixel center because of some vias in the CCD preventing centered placement.

 

Yesterday however, I hit paydirt in the form of a guy over at the FM forums that also works with sensors. This is what he had to say:

 

In the Bayer pattern the red "clots" of absorption dye are slightly larger than the blue&green ones. They're still "square" as in equal width and height, but slightly larger.

 

The sensor surface underneath however, isn't rotationally symmetrical. The ccd vias are slightly rectangular, the Y dimension of the active pixel surface being larger than the X dimension. This is the same in all of the Kodak 6.8µm sensors, dating back to the "original" in the Oly E-1, 2002. They try to mitigate some of this effect with ITO gates and metal masks on top of the sensor (to stop some of the colour crosstalk via pixel>pixel light spread), but there's also the question of angle/wavelength imbalance.

 

If you balance the QE so that you have one "correct" whitebalance for perfectly perpendicular incident angles, the Kodak construction will by construction default unfortunately not have the same WB for incident light with lower angles. If you shine a white light source with perfect collimation (all light travels in perfect, parallel lines in space) on the sensor - first at 90º angle, and then at 30º angle - you wouldn't get the same WB result... Even though the light has the same CCT in both cases.

 

This defect is aggregated by the use of quite thick (not the best materials...) colour filters and assymetrical cell structures (the blue cell is even more assymetrical than the other cells, red is almost square) and you get angle-dependent colour faults.

 

PREDICTABLE colour shifts, if you know the lens chacteristics, but even so.

 

To get the filter layer to work properly with light with high incident angles, the "normal" case you need to deposit them fairly centered over the active pixel areas. Unfortunately this means that they are DE-centered in the case of light with lower incident angles, due to the assymetrical nature of the CCD cell construction Kodak uses.

 

This decenterig means that you get one dominant colour fault in one direction from the image center, and the opposite colour fault in the opposite direction. And since the blue channel has the smallest active pixel surface, it has the lowest asymmetry. R-G shows a lot more deviation from corner>opposite corner.

 

The colour filters are not placed exactly over the center of the pixel (that's square), it's placed over the center of the active pixel surface (that's rectangular, and not equally transparent for all wavelengths).

 

So it's just as much a question of non-equal vignette as a question of colour "leakage". And the sensor is asymmetrical in construction, it isn't even mirror-symmetric in the horisontal direction if you look at a FEM side view. Red wavelengths have a much higher penetration capability, so it can go straight through a passivation layer, but not a metal. So one side of the "blocking" parts of the active pixel surface is transparent to red, the other isn't. The "red" transparent window in the pixel surface is pushed over to one side of the square containing the entire pixel, while the blue "window" is centered on the square.

 

So the answer is simply that the photosites have an asymmetric construction with materials with different absorption coefficients. Furthermore as different wavelengths have different energies there will be a difference in how well they can penetrate it.

[lars_bergquist]

Denoir, that was the most cogent and credible explanation I have seen yet. It does not answer all my questions – why red on one side but green on the other? – but it just smells right.

 

This seems to indicate that Leica should change over to a CMOS sensor for the M10. These sensors have been improving fast and any remaining differences in image quality between a M9 sensor and that in a FF Canon or Nikon, is probably already due mostly to the absence of the AA filter. The chance to overcome the oblique-ray colour shift is probably a more valid reason for such a changeover than the (to my mind unnecessary) wailing over low ISO. Leica's superiority in lenses will remain.

 

The old man from the Age of Kodachrome I

[mjh]

It does not answer all my questions – why red on one side but green on the other? – but it just smells right.

These are a lot of partial explanations rolled into one and frankly I’m not buying all of it just yet. However, if there is really an asymmetry in the light blocking behaviour between neighbouring pixels:

 

So it's just as much a question of non-equal vignette as a question of colour "leakage". And the sensor is asymmetrical in construction, it isn't even mirror-symmetric in the horisontal direction if you look at a FEM side view. Red wavelengths have a much higher penetration capability, so it can go straight through a passivation layer, but not a metal. So one side of the "blocking" parts of the active pixel surface is transparent to red, the other isn't.

 

then this alone would probably suffice to explain the Italian Flag Phenomenon. At least it’s a better explanation than any I have seen so far.

 

Furthermore as different wavelengths have different energies there will be a difference in how well they can penetrate it.

Just to prevent any possible misunderstanding: It is the lowest energy wavelengths (namely red) that penetrate deepest, not the other way round as one might presume. Just like the low energy red light from the sun penetrates deepest into the atmosphere whereas the high energy blue lights gets scattered, creating red sunsets and blue skies.

Edited August 7, 2011 by mjh

[jankap]

One interesting experiment has been left out.

Namely a pin hole with the hole out of the center.

Or with a lens offering perspective control, the similar effect of course.

Jan

[Advertisement]

Hallo,

Look here: Leica M

[sandymc]

then this alone would probably suffice to explain the Italian Flag Phenomenon. At least it’s a better explanation than any I have seen so far.

 

Indeed it could. You would have to assume the cells are "asymmetrically transparent" () in both dimensions to account for the left to right and top to bottom nature of Italian flag, but if it's possible in one dimension, then two is also quite possible. Interesting in the sense that the cure in design is obvious - mirror the cells. One wonders why that wasn't done anyway given the symmetrical nature of the sensor.

 

Sandy