The M9, lenses and the "red edge" problem

[Alnitak]

Advertisement (gone after registration)

The WATE is a retrofocus design as are the 24/1.4 , 21/1.4 and the super-elmar. This means that the angle of attack of the light rays on the sensor is not very steep. With the CV12 and CV15 (and a Zeiss mentioned in another thread) the light rays hit the sensor at a steep angle. How about the super-angulon that has similar design if I recall correctly?

 

Sofar I am convinced it is due to the steep angle of attack with some of the wide angle lenses. Recall that a few years ago Leica said that you could not make a digital M at all, for this very reason. The M9 sensor is pushing the boundaries of what is possible.

 

I think that the angle of incidence of the incoming light definitely plays a role here. The non-retrofocus designs of the CV lenses no doubt result in a much steeper angle of incidence relative to the individual pixels on the sensor, and while the microlenses can compensate for this to some extent, there are limits. As such, it may be that some lenses are just not suitable for a FF digital M. Like many other manufacturers, Leica may need to update some lens designs for a FF digital world (as will Zeiss and Voigtlander).

 

However, I also believe that any slight misalignments of a lens will exacerbate the problem and the steeper that angle of incidence (which would generally be related to the wider angle of view of a lens), the more sensitive it will be to any slight misalignment.

[Alnitak]

I tested a 21 Super Angulon on the M9 and the results where "brutal". Im not buying the problem is with the lens, it's the M9 sensor. I tried several different lens profiles in the M9 menu and none of them helped in any way. And it was not just the edge but the 1/3 of the left side. Im not buying that the lens is at fault because when used on film (negative/slide) the same 21 f4 Super Angulon is incredible.

Jeff: From your results, Im a little confused that you think the cyan drift is well corrected:confused:. Are we looking at the same images?. Because from your sample's they all look horrible. It's not just the weird color but the lack of uniform density that Im seeing. Anyways, thanks for posting your finding's. If you go to my Smug Mug site and go to the M9 section you can see some of my sample images with the M9 and the Leica 21 Super Angulon.

 

gregory

 

SmugMug Photo & Video Sharing. You look better here.

 

Gregory,

 

Your link to smugmug is not correct, but I managed to track down your gallery. Wow, the image from the 21 Super Angulon is terrible.

 

Now I am starting to think that maybe I was a bit wrong. The pattern is just too consistent across multiple lens types.

 

Today my thinking is that the its a combination of lens types and the camera. Clearly the lenses that have a very steep angle of incidence, are behind this problem. If what I am putting together is correct, all of the lenses that are exhibiting this problem are non-retrofocal designs, so they have fairly steep angles of incidence for the light in the corners and on the edges.

 

I suspect that the problem could be cropping up in a number of places, including the Bayer demosaicing algorithms, and the algorithms applied to deal with vignetting and cyan drift. The fact that the effect is always strongest on the left and bottom regardless of lens manufacturer seems to imply that the algorithms are not properly dealing with these types of lenses and the issues caused by the steep angle of incidence of the light rays.

 

It could be that certain lens designs will just not be suitable for a FF DRF. In other words, its not that the FF digital M was not possible, its just that its not possible for all lens designs, and newer designs will be required for certain types of lenses, namely the wider lenses. So, it's possible that this could be fixed in part in the camera, but not in whole without new lens designs.

 

As for the cyan drift, note I said "fairly well-corrected." There's a big difference. I do think that the 12mm coded as the 21mm 11134 looks pretty good, and in most cases will not present a problem used that way. The 15mm is still messy, but I do think that my copy is slightly decentered--and I still do believe that decentering has a role to play here, in that slight or subtle decentering can likely greatly exacerbate this situation, particularly with the ultra-wide lenses.

 

Jeff

[Nikkor AIS]

Jeff, Thanks for the heads up on my smugmug. When I click on it I get there at any rate it's rogaltacdesign. I think you spot on about the retro focus designs and the full frame sensor. I wish I still had my 21 F4 Nikkor which required mirror lock up to try on my D3. It would be an interesting test to see how the Nikon sensor would handle the same type of lens. The results would be interesting.

Getting back to the Leica M9 and the 21 Super Angulon yes it's a pretty bad combo. Kind of makes me sad as I love the lens on the M3/M6 However, as other's have mentioned there are solutions in post processing. When I get a M9 I'll just have to get the 21 1.4:p.

 

Gregory

 

http://www.rogaltacdesign@smugmug

[marknorton]

I do think Leica are dragging their feet in fixing the asymmetric red contamination problem or in coming clean as to why the problem exists. True, Cornerfix can deal with it but Leica should get it right in camera.

 

I still wonder whether they have fitted the sensor off-centre to avoid more fundamental redesign of the camera.

[Alnitak]

I do think Leica are dragging their feet in fixing the asymmetric red contamination problem or in coming clean as to why the problem exists. True, Cornerfix can deal with it but Leica should get it right in camera.

 

I still wonder whether they have fitted the sensor off-centre to avoid more fundamental redesign of the camera.

 

I was wondering the same thing; the pattern sure does seem to hint that the sensor might located off-center.

[Ario Arioldi]

I was wondering the same thing; the pattern sure does seem to hint that the sensor might located off-center.

I've seen asymmetrical color cast of the same kind also in Nikon D3x and Canon 1DsMKIII equipped with Distagon 21mm.

This suggests to me that the problem is not related to sensor decentering.

Cheers,

Ario

[Alnitak]

Advertisement (gone after registration)

Ario, that's very intriguing. Do you have any example images with either the D3x or 1DsIII?

[Ario Arioldi]

I do not have myself but same example can be seen in the Digilloyd review of the Zeiss lenses (by subscription).

The asymmetrical color cast is less visible (I guess because of the much larger distance from flange to sensor) but it is there, and is always showing more cyano on the right and more red on the left.

Cheers,

Ario

[jonoslack]

HI Ario

It used to happen on the Kodak 14n / SLR/n as well - where it was referred to as the 'italian flag'. I spent many happy hours doing spreadsheets of rgb values across the frame for different lenses and sending them off to Kodak . . . the things we do!

 

I don't totally understand it, but I think the reason it's red on the left and green on the right is to do with wavelengths rather than offset anything. Of course, the green on the right is less noticeable, but it is there.

[jaapv]

I do think Leica are dragging their feet in fixing the asymmetric red contamination problem or in coming clean as to why the problem exists. True, Cornerfix can deal with it but Leica should get it right in camera.

 

I still wonder whether they have fitted the sensor off-centre to avoid more fundamental redesign of the camera.

You mean that the battery is in the way as you pointed out before?

[UliWer]

...

I still wonder whether they have fitted the sensor off-centre to avoid more fundamental redesign of the camera.

 

The M8 shows the same issues with red shift on the edges, and you very often find the tendencies to the left as well - though sometimes it's equal on both sides.

 

You are right that Leica should give a solution. People who have been at a meeting in Wetzlar some weeks ago report that Leica is working on a solution which we may hope will be inculded in a new firmware. This firmware was announced for "soon".

 

So let's see.

[blatent liar]

Tangent on this topic.

 

Can the vignetting, and magenta be caused/exacerbated by a mildly torqued, or twisted sensor, as it is leveled and shimmed for focus accuracy in the factory.

 

In the Factory visit video, they were pointing out the small shims used to flush the sensor to the "film Plane", as most issues appear on the image left, Camera Sensor Right... I thought -- grasping at straws may not be so far off.

 

I recall a story about an aerospace helicopter chassis maker, who had a pregnant welder welding M2 Aluminum stays crooked, and the supplier could not figure out why the part steered leftward. The stay was crooked due to her modified posture. it adversely effected hundreds of sub-parts.

 

Just a thought; Likely just a silly thought.

 

-Max

[jaques]

If you watch the leica M9 factory video- when they are looking at those tiny shims- the Leica guy accidentally drops one back into the wrong draw in the box. That may explain at least one bad camera out there... ha ha. i doubt it.

[scott kirkpatrick]

I still wonder whether they have fitted the sensor off-centre to avoid more fundamental redesign of the camera.

 

I do, too, since problems due to light reaching the chip at a sharp angle should be symmetric about the lens axis. But I am puzzled. The "Italian flag problem" -- red on the left, white in the middle, and green on the right -- which Jono, our "gentleman photographer," says dates back to the Kodak 14N, suggests that the chip and its firmware handlers overcompensate on the left hand side of the image and undercompensate at the right hand side of the image. Now the left hand side of the world is imaged on the right hand side of the chip and the right hand side of the world is imaged on the left hand side of the chip. (Right and left as seen standing behind the camera looking forward.) The battery on the left in the M8 was a bit too big to allow a full frame sensor to fit comfortably into the M8's casting, Mark Norton found. This would suggest that the sensor in the M9 is shifted a mm or so to the right, so that the rays of light coming from the left and striking the right pass through more green absorptive filter, while those coming from the right and striking the left side of the chip would pass through less green filter. If there is a basic correction applied to all raw data, even when the lens is not recognized, this gives a red white green pattern, but unfortunately it is reversed.

 

Here is a horrible example, using a lens which is not a good match to the M9. I shot a white wall under indoor artificial illumination using my Canon 19mm f/3.5 LTM mount lens. It's a nearly-symmetric design, which portrudes about 2 cm into the camera body (yes, the light metering seems to still work fine. At f/3.5 we see:

 

 

and even at f/8, when the luminance vignetting is reduced, but the color shifts should be unaffected:

 

 

the Italian flag pattern persists. The strongest corrections that I can find in the firmware tables, for the 21/2.8 pre-asph, doesn't improve this much. It is a job for Cornerfix, in its improved asymmetric mode.

 

I've sent Sandy vignetting series under daylight and artificial light, using the Summilux 21, the CV21/4.0 (nasty!), the Leica 28/2.8-asph (probably the most symmetric and least telecentric of Leica's wide angles, with color shifts that prove it), and others have contributed extreme examples from the CV12 and Cv15. Looking at these, it seems a good strategy is to correct all lenses at a medium aperture, say f/5.6 of f/8, since the color shifts appear to differ little with change of aperture. The aperture-sensitive effects which remain are luminance vignetting at the widest apertures. These can be addressed with additional profiles, or ignored, since they are a natural-looking effect.

 

scott

[sandymc]

For those interested, the two images below are what CornerFix can do with Scott's 19mm.

 

The first one is a default correction, the second tweaked (by careful selection of reference image for the profile among the various images that Scott sent me for even illumination, and using multiple equation mode to generate the profile)

 

Sandy

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!

[Ario Arioldi]

Hi Sandy,

while trying to build the best profile for my CV15 I found that the multiple equation option was producing inferior results compared with the default setting.

Just a curiosity, does this make any sense to you?

Cheers,

Ario

[sandymc]

Hi Sandy,

while trying to build the best profile for my CV15 I found that the multiple equation option was producing inferior results compared with the default setting.

Just a curiosity, does this make any sense to you?

Cheers,

Ario

 

Ario,

 

Yes - the problem with the multiple equation option is that it works well if the lighting, etc is entirely uniform, whatever the surface is, is entirely free of blemishes, etc. Anything less than a close to perfect reference image however, and CornerFix will start to try to correct for the imperfections in the image, rather than the imperfections in the lens/camera. The default "Single Equation" mode is a lot better at ignoring uneven lighting, etc.

 

Regards,

 

Sandy

[Ario Arioldi]

Sandy,

now is more clear to me.

And once again thanks for giving us such a useful tool.

Cheers,

Ario

[Alnitak]

Sandy,

 

How possible is it to generate a set of default profiles for lenses that could be distributed with the program? I find that its quite hard to get a really good flat frame to use for generating the profiles. If it were possible to have some standard .cpf files that could be used, it would be a great help.

 

Jeff

[Advertisement]

Hallo,

Look here: Leica M

[sandymc]

Sandy,

 

How possible is it to generate a set of default profiles for lenses that could be distributed with the program? I find that its quite hard to get a really good flat frame to use for generating the profiles. If it were possible to have some standard .cpf files that could be used, it would be a great help.

 

Jeff

 

Jeff,

 

Not by me

 

Although if someone else creates a library, I'll happily include it into the CornerFix distribution file.

 

A few people have tried to create libraries of profiles in the past; not sure how many of those are still active.

 

But a problem is that certainly for the M9, it appears that you really need a profile specifically for your combination of lens and camera. Various people have reported significant differences in vignetting with the same lens on different M9s, and different examples of the same model of lens on the same M9.

 

Regards,

 

Sandy