New owner, red edge problem with 18mm SE

[scott kirkpatrick]

Advertisement (gone after registration)

What appears curious to me is the non-symmetry of the phenomenon... Why is it so, while the system is symmetric?

 

Right. Look at Lars' pictures, made with a daylight flash. The red corner is strongest at the lower left, the green shift is strongest at the lower right. There are some color-dependent effects that might happen when light reaches the sensor at an angle. Red light penetrates much deeper into Silicon than does green or blue (Foveon's three-layer RGB cell depends on this effect). But the increasing angle is symmetric. The asymmetric elements in a CCD sensor are the cell itself (which is non-symmetric due to the placement of charge collection circuits and overflow drains), the Bayer array:

 

RGRGRGRG

GBGBGBGB

RGRGRGRG

GBGBGBGB

 

and perhaps the demosaicing software. What is surprising to me is that such small asymmetries could be magnified so strongly, leading to shifts seem over 100s of cells.

 

scott

[jlam]

My worry is that there may not be a single optimal correction for each lens, as the optimal corrections for different shots may be different depending on the subject distance and the characteristics of light rays coming from the subject (e.g. scattered vs directional).

[Advertisement]

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

[marknorton]

Has anyone received anything in the way of an official response from Leica to this problem?

[lars_bergquist]

What appears curious to me is the non-symmetry of the phenomenon... Why is it so, while the system is symmetric? Because the incident light (of the speedlight) is itself not captured in the axis of the lens?

Right, it's the 'Italian Flag Syndrome': red to the left, green to the right. I suspect that this has to do with dispersion in the microscopic lenses over the individual pixel cells.

 

Take the situation at the left side of the image, which means of course the right side of the sensor. As the light is refracted ('bent) it is also dispersed (split up in component colours) and some of that dispersion is lateral, not axial. So in this case, some of the violet/blue light 'spills over' and does not land in the pixel well, creating a violet/blue deficit, and thus a reddish cast. The situation is just the other way around on the other side, where some of the red light spills over, causing a green cast. Take this very amateurish hypothesis for what it is worth.

 

This is clearly a problem inherent in sensors with micro-lenses, but it should become more acute, the more acute the ray-bending is -- and therefore it pops up in the M9.

 

The old man from the Age of Redeye

[scott kirkpatrick]

Right, it's the 'Italian Flag Syndrome': red to the left, green to the right. I suspect that this has to do with dispersion in the microscopic lenses over the individual pixel cells.

 

Take the situation at the left side of the image, which means of course the right side of the sensor. As the light is refracted ('bent) it is also dispersed (split up in component colours) and some of that dispersion is lateral, not axial. So in this case, some of the violet/blue light 'spills over' and does not land in the pixel well, creating a violet/blue deficit, and thus a reddish cast. The situation is just the other way around on the other side, where some of the red light spills over, causing a green cast. Take this very amateurish hypothesis for what it is worth.

 

This is clearly a problem inherent in sensors with micro-lenses, but it should become more acute, the more acute the ray-bending is -- and therefore it pops up in the M9.

 

The old man from the Age of Redeye

 

I believe that the pattern of shifting the microlenses in towards the center is symmetric about the center of the sensor, so this wouldn't give rise to a left-right asymmetry, although it is another factor that could make colors get strange at the edges and corners of the image.

 

Also, the first camera to have the Italian flag syndrome (according to Jono Slack) was the Kodak 14N, which I don't believe used micro-sensors.

 

I think I found the champion worst-case lens for the M9 a little while ago, and showed a picture of its dramatic Italian flag. It is a Canon 19/3.5 LTM mount lens from the 1950's. The rear lens element comes within one cm of the sensor. The other extreme case I would like to see someday is the Zeiss 38mm lens that you find on the Hasselblad Superwide camera. Those cameras do take a digital back, but I don't have either camera or back, so the experiment is a bit expensive... This is a case where the rear lens element is within a cm of the film, and the image area is 57mm square.

 

scott

 

edit: Here's the permalink to the post I was thinking of. That thread has a pretty good discussion of this issue. No, there has been no information from Leica on what causes it, but I would hope that they and Kodak know pretty well what it is.

Edited January 28, 2010 by scott kirkpatrick
add link

[wattsy]

it's the 'Italian Flag Syndrome': red to the left, green to the right.

 

Italian flag.