Purple corner when used Minolta G-Rokkor 28mm f3.5 L mount

[Hoppe]

Advertisement (gone after registration)

I recently got Minolta G-Rokkor 28mm f3.5 L mount. I tried it on M10 using a Kipon adapter and found that some photos had purple colour on both sides.

 

Could someone help me? What do I need to do?

[adan]

Wide-angle rangefinder lenses designed for film do not play well with full-color digital sensors. They need in-camera corrections applied to the picture files to remove the purple corners. This is why Leica has provided "6-bit" coding marks on their own lenses since 2006, and offers upgrades to add the coding to some older lenses, back to about 1980. So that the digital cameras will know what focal length you are using, and can then correct for the specific pattern of purple corners (which varies from lens type to lens type).

 

The problem is called "Italian Flag syndrome" because in many cases the color shifts are red/purple on one side, and greenish on the other side.

 

Note that SOME older wide-angles are very difficult to correct for. The Leica 21mm Super-Angulon f/3.4-f/4 lenses from the 1960s always produce strong purple corners, as does the Voigtlander 15mm from 1997 and the Zeiss 15mm Hologon from the 1960s. They are simply incompatible with color digital image sensors.

 

http://www.kenrockwell.com/leica/images/m9/examples/21mm-f4/L1001261-21mm-f4.jpg

 

Leica does not offer corrections designed specifically for lenses made by someone else: Minolta, Zeiss, Voigtlander, Canon or Nikon LTM, etc. But sometimes a correction for a Leica lens will produce "pretty good" results with those lenses.

 

As a first try at fixing this problem: with your Minolta lens mounted, open your M10's Main menu and go to the very first menu item: "Lens Detection."  Go into that menu item, and select "Manual M" and you will see a long list of LEICA M lenses. Scroll through the list to find "21mm f/2.8 11134" - this setting provides the strongest purple-removal in the corners, which your lens will likely need - even though it is just a 28mm lens. Press the "Info" button on the camera to set that as the active lens correction.

 

This is the M typ 240 lens menu - the M10 menu is identical except for the background color: http://www.waloszek.de/m240/lens_list_manual/DSC06080_600.jpg

 

Then take some pictures and see if the purple corners have been eliminated adequately.

 

If you find the corners are now too green, you can try a milder correction from the same menu - perhaps the setting for "28 f/2.8 11804." For solving this problem, the actual focal length chosen is not important - what you need is the correct pattern of purple removal, and since you lens is not from Leica originally, that may be a correction for a Leica 21 or 24 or 28, or even something else. Experimentation is free, and you will not damage anything by trying any or all of the settings.

 

If you want deeper technical details about why this occurs with small wideangle rangefinder lenses on Leica digital cameras, we can give you that. The very short explanation is, the shallow body of Leica M cameras compared to SLRs, and how close these small lenses like your Minolta come to the sensor, and the shallow angle of incidence with which their light rays hit the sensor in the far corners. Combined with the "checkerboard" of microscopic color filters (Bayer pattern) that allow digital sensors to see in color.

Edited December 24, 2017 by adan

[Advertisement]

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

[Hoppe]

adan,

 

I thank you so much for your kind advice. I really appreciate it. I’ll try it tomorrow. Hope it would correct the purple corners.

 

In your comment, you mentioned that should I am interested in a deeper technical detail abour why this occurs, you can give me that. If it isn’t too much trouble for you, I would be glad to read it. I thank you again.

[adan]

Let us start with film emulsion. It is a flat, thin, homogenous layer of light-sensitive silver compounds evenly distributed and suspended in translucent gelatin, smoothly coated on a base material (usually plastic) for strength. In modern (post-1935) color film there are three stacked layers for the three primary colors to be recorded (red, green, blue). It has a very simple and unorganized structure, relatively speaking. Any square micron of the emulsion is effectively identical to any other square micron. It is equally sensitive to light coming from any direction (except perhaps the back ), at any angle.

 

Therefore, in the 150 years or so that camera lenses were designed with film (or a coated glass plate) as the expected "target" for their projected light rays, little attention was paid to the imaging surface. It did not have any "special needs" to be considered ,and was assumed to simply respond as intended to any image projected on it. At least for general-purpose photography - special films and special lenses for technical or scientific work may have paid more precise attention to details.

 

Next let us look at how the 28mm G-Rokkor or other compact short-focus lenses for Leica-type camera bodies work. One advantage of the Leica and other rangefinder cameras (and, of course, the recent "mirrorless" cameras) is that there is nothing inside the camera between the lens and the film except about 3mm of "shutter curtains." (Unlike SLRs, which need about 40-45mm of space between the lens and shutter for the reflex mirror to flip up and down). Leica cameras and the lenses for them standardize on about 28mm between the lens mount and the film, with the lenses often protruding even deeper into the camera than the mount (as your 28mm G-Rokkor does):

 

http://www.9days.hk/uploads/product/13F047_03.jpg

 

But what really matters is that your lens and similar 28mm lenses capture a view of the world that is 65° wide across the long dimension of a 24x36 negative or sensor. And in designs for the Leica, the lenses are relatively symmetrical, so that the light coming out of the lens at the back to hit the film also covers a total angle of about 65°. At each end of a horizontal picture, the light from the edges of your subject area is thus hitting the film (or sensor) at an angle of about 32.5° from the vertical - the "angle of incidence."

 

http://mathworld.wolfram.com/images/eps-gif/AngleOfIncidence_800.gif

 

Light passing straight through the lens from the center of the subject to the center of the film or sensor has an angle of incidence of zero.

 

As we mentioned, film is relatively insensitive to the incoming angle of light rays, so lens designers knew that almost any angle of incidence would work.

 

Unlike film, digital sensors have a very complex and non-homogenous 3D microscopic structure of silicon, glass, colored filters, etc. If one thinks of film as a smooth meadow, a digital sensor is like an inner city, divided up into blocks of tall buildings (on a microscopic scale). The net result is that digital sensors are far more "sensitive" than film to angle of incidence, and thus misbehave with lenses with a large angle of incidence at the picture edges/corners, that worked perfectly well on film. On the whole, current digital sensors work best when the angle of incidence is closer to zero.

 

Horizontally, sensors are divided up into squares - the pixels - each with different color filtering, the Bayer pattern or Bayer filter array. One key point suspected in the "Italian Flag" problem is that the Bayer filter pattern is not symmetrical - the blue pixels are always biased to one side while the red pixels are biased to the other side, no matter how you turn the pattern:

 

https://upload.wikimedia.org/wikipedia/commons/thumb/3/3c/Bayer_pattern.svg/2000px-Bayer_pattern.svg.png

 

Vertically, sensors are divided up into several layers, and most importantly, the imaging areas of the silicon (D and E) are not flat and homogenous, but deep "wells" divided by solid-state circuitry. Additionally important to the purple-corner problem is the existence of microlenses (F) capping each pixel, to refocus light down into the "wells".

 

http://keyassets.timeincuk.net/inspirewp/live/wp-content/uploads/sites/13/2010/08/Sensors1.jpg

 

This diagram demonstrates one current idea about what may cause the purple-corners or Italian-flag effects - leakage and/or reflection of light already filtered for one color (in this case green) into a neighboring pixel programmed to measure a different color (in this case blue). Due to the gaps between the vertical layers, the vertical structures, and the angle at which the incoming light hits the microlenses or the underlying silicon. And resulting in excess blue signal on one side of the picture, and (due to the bias of the Bayer pattern) excess red signal on the opposite side of the picture.

 

http://www.extremetech.com/wp-content/uploads/2013/03/Wiring-diagram-for-a-typical-front-illuminated-sensor.gif

 

Here are some real-world photomicrographs of an image sensor that may help clarify what the schematics above are showing, especially the 3D "wiring" etched into the silicon:

 

https://petapixel.com/2013/02/12/what-a-dslrs-cmos-sensor-looks-like-under-a-microscope/

 

You may ask why Sony and other mirrorless cameras and lenses do not have as much problem as Leica's lenses.

 

And the answer is, Sony and Fuji and Olympus have designed their mirrorless-system lenses specifically for digital sensors right from the beginning, with optics that reduce the angles of incidence. Your G-Rokkor and my 1981 21mm Elmarit-M were not designed for use with digital sensors (although fortunately Leica goes to some lengths to achieve significant backwards compatability through their camera corrections. At least for their own lenses, and coincidentally, sometimes for others')

Edited December 25, 2017 by adan

[anders]

Try this:

 

https://sourceforge.net/projects/cornerfix/

 

 

Merry Christmas

Anders

[adan]

A brief follow-up on "lenses designed for digital." Here is a diagram of a 8mm-30mm zoom lens designed for a (small) digital sensor, as in a Leica C or D-Lux P&S (or similar cameras from Nikon, Canon etc.).

 

Note the path of the edge or corner light rays (blue) at the wide-angle focal length (8mm, bottom). That path is NOT symmetrical - the light from the subject comes in at a wide angle (~35° from each side), but after passing through the lens, exits the lens and strikes the sensor at an angle of incidence of less than 10°, unlike our old film-era rangefinder wide-angles. This is called a "telecentric" lens design.

 

https://image.slidesharecdn.com/azoomlensdesignmethodjuly32013-131211180539-phpapp02/95/a-zoom-lens-design-method-july-3-2013-10-638.jpg?cb=1386786948

 

I can't find a ray-tracing diagram for the Leica Vario-Elmarit-SL 24-90 lens - but I am sure that that lens, designed in the digital era, behaves the same way, and thus even at 24mm, and on the same sensor size, does not produce the purple edges of old film 28/21 lenses on M cameras.

 

Could Leica design new "telecentric" lenses for the Leica M digitals? Certainly. But one characteristic of telecentric lenses is that they are generally very large relative to the size of the image area. The Vario-Elmarit 24-90 is larger and heavier than the SL camera itself. That would run contrary to the whole point of the M cameras.

 

It is probable that the Leica-M 21 and 24 Summilux f/1.4 lenses, designed in the digital era, and already large due to the aperture and angle of view, have some telecentricity designed into them.

 

SLR camera makers got lucky when it came to digital. Their wide-angle film lenses, even those designed decades before digital sensors existed, already come with a certain amount of telecentricity built in. Because, to clear the swinging mirror, those lenses needed asymmetrical retrofocus designs (back focus distance substantially longer than the nominal focal length). Note the difference in angle of incidence (b ) delivered by a retrofocus wide-angle (top) and a shorter, more symmetrical wide-angle (bottom) - I'm guessing this compares the Zeiss Distagon 21mm for Contax SLRs, and the Zeiss G-Biogon 21mm for the Contax G1/G2 "AF rangefinder" cameras:

 

http://pic.enorth.com.cn/0/03/48/96/3489683_304673.gif

 

Mount a Leica R (or other SLR) 15/19/21/24/28/35 on the M10, and even without coding or menu identification, they will likely avoid any purple corners. The one SLR exception is tilt-shift lenses, where by dialing the lens off-center for perspective or DOF control, the angle of incidence may get very large in one direction or another - that will sometimes cause purple corners or Italian Flag color distortions even with SLR cameras.

 

https://c4.staticflickr.com/3/2740/4278462691_ac24130c05_z.jpg

 

Phase One's Capture One Pro program has a module for LCC (lens cast correction) that builds profiles of purple-corner problems due to off-center T/S lenses, and corrects for them

 

https://i2.wp.com/blog.phaseone.com/wp-content/uploads/Tip902-Img7-full.jpg?ssl=1

Edited December 25, 2017 by adan

[jaapv]

Advertisement (gone after registration)

there is a flat-field correction plug-in for Lightroom and Photoshop as well.

[adan]

Yes. Leica's in-camera corrections for coded lenses, and Corner Fix, and Capture One LCC, and the Adobe flat-field plugin all do essentially the same thing - apply a "known" correction image or pattern of color casts on a white surface (a profile) onto other pictures, to subtract the unfortunate color cast.

 

On the whole, I'm happier just building a system that works "straight out of the camera" - Leica lenses with appropriate 6-bit coding (from the factory, or retrofitted, as on my 1980 21mm). The old KISS principle.

 

But I'm sure glad to see those modules out there, for those who want to "hot-rod" with legacy lenses.

[Hoppe]

adan, I thank you so much for taking your time to explain it. It’s a great reading. I really appreciate it.

 

You possess a great knowledge and understanding.

 

In addition, I also want to thank others for giving great advice.

[chattillion]

If you are an Adobe Bridge user you can use the optical correction for an MS Optics 17mm lens and it removes most of the vignetting.  

[Raid Amin]

My Rokkor-M 28/2.8 shows no such purple corners, but the Zeiss Hologon 16/8 shows a strong purple smearing in the corners for me.