M9 and Red Edge Redux--Part II, with 28mm Elmarit?

[hoppyman]

Advertisement (gone after registration)

So do I - I was just referring to one or two posts - not to the problem as experienced by a number of users.

 

Jaap, yes we both agree completely. I meant only to emphasise the positve parts of the discussion. I think that your comments were a reasonable part of the discussion and in no way patronising.

 

By way of further contribution I reviewed several thousand frames with wides on (two) M8's. I reduced my large sample to shots that have areas of white or grey or similar continuous tone on the left hand edge. I could detect some (small) effect with the new Summilux 21 wide open only (remember the Hessenpark forum meeting where we could try the new lenses).

 

On my M8 and one loaned to me by Customer Service during my upgrade, I cannot see any fault whatsoever in shots from the ZM 18, the Super Elmar18, the Elmarit 24, the Summicron 28 ASPH and Summicron 35 ASPH. I see a severe case of red edges in the first couple of shots with a borrowed Super Elmar that had no UV/IR on it but the camera was wrongly set for the UV/IR. The effect is completely gone from later shots with the filter and correct setting.Those persuade me personally that this is related to the camera firmware corrections (and hence can be further improved in the next firmware for the M9).

Snow is a great novelty for me so when I visit somewhere that has some I shoot it! Believe me that I have a LOT of shots containing snowfall at Yosemite, the Grand Canyon and Red Rock Canyon (Nevada). Nary a red edge in site.

 

 

uncropped 24 Elmarit M8

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!

[Advertisement]

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

[dwbell]

I think it's very valid for new users to postulate about this problem before dropping thousands on a new body.

 

I certainly would have loved to have know about it, rather than finding out after purchase that my brand new M9 with 35 cron - YES THIRTY FIVE, not 28, 21 etc. might suffer from frankly hideous red edges.

 

See my thread;

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/141888-purple-colouring-top-left-only.html

 

I've not sent it in to Leica though the problem is on their customer fix list internally. a) I don't want to be without my body, It has been suggested that a firmware fix may solve the problem.

 

It's by no means a "simple fix in post", it varies in intensity according to light, aperture, angel of main light etc. My custom profile in CornerFix does help alot, but knowing I would have to add an open source DNG re-write program to my workflow before importing to Lightroom would have made me ask the same questions in this forum before purchase. It's a Leica for goodness sake, it costs thousands and thousands of pounds and all the above applies!? I was annoyed, let down, and still am until it's fixed.

 

Kind regards,

DWBell

[parsod]

I discussed the red edge problem related to my Elmar 24 with the camera dealer who sold me the equipment and he put me in contact with the Leica agent for Sweden who is actually sitting in Denmark. Brian Felskov at Nordisk Fotoimport A/S brought my question with him when he visited Solms last week and today I got an email from him confirming that the problem of red edges with Elmar 24 on the M9 is a known problem at Leica and that they are working with a firmware updated. This means Leica acknowledge the problem and I guess a possible firmware fix would affect more wide-angles than just the Elmar 24 - so there is hope:)

 

/Paer

[hoppyman]

I wrote to Customer Service in Solms once again after these reports and received this information back. As when the broken cover glass reports emerged, my question was referred directly to Stefan Daniel who responded very promptly and openly. On this red edge question, technical staff provided this information to me. I have quoted the response verbatim. I would add my personal comment that I have never experienced any visible practical problem at all from about 2,000 frames with my M9. I posted my examples from all of my lenses from 35 and wider.

 

...............................................................................................................

"Thank you very much for your information.

We have had some complaints about the red edges, you are right. I will give you the full technical explanation for that.

On the one side, we do have a very short distance from the lens to the sensor (for this reason the m-system is so compact), on the other side the pixels of the ccd sensor are not symmetric. This combination produces a totally colored and inhomogenious RAW image. To provide a plain and neutral image, we do a lot of complex compensation, sensor corrections and lens vigentting corrections. What we achieve is a nearly neutral image, but unfortunately just "nearly" neutral. The tolerances for the M9 are even tighter as for the M8 with IR-filter, but unfortunately there are some very slight deviations towards red, the obviously are not acceptable for some special applications.

We did an improvement in a previous update, using the vignetting correction, this improved the effect visibly. But unfortunately we had to recognize that even this was not enough. There are still some complaints, therefore we decided to apply additional correction algorithms for the critical lenses. But developing these need unfortunately some time, because some sequences in the internal data management have to be changed, this is always a difficult and risky modification. Of course correcting this on a PC is an easy thing. But letting the camera do this needs a very simple algorithm, because we would not accept any delay in DNG mode.

We are planning to offer this update as soon as possible, but it will need at least until spring 2011.

I hope this answer helps you understand the confusion about the red edges."

...................................................................

[jaapv]

Thanks,Geoff. This provides much-needed clarity and hopefully some peace of mind to those of us that still have problems in this respect.

[ho_co]

Geoff, thanks for contacting Stefan and for posting his response. He is an excellent spokesman for the company, both knowledgeable and informative.

 

The unanswered question is why there is such variation among bodies and lenses in regard to the red left problem.

 

Stefan speaks of "critical lenses" and of a simple and implementable algorithm for the camera, and says delivery is at least three months away.

 

The fact that Leica still anticipates a firmware solution is encouraging.

 

 

Alan's link (in post 41 above) to the Phase One "lens cast" solution is interesting in this regard. If I understand correctly:

• Phase One furnishes a special piece of translucent glass (expensive to replace, IIRC).
  
• One sets the camera for a given shot, then exposes a frame through the white glass.
  
• One then removes the white glass and makes the intended shot.
  
• One uses Phase One's software to calculate and apply the needed correction based on those two exposures.
  

 

We're lucky to have Sandy's simple and direct Leica solution, CornerFix. We may need it for a while longer than we had hoped.

[hoppyman]

Advertisement (gone after registration)

Hi Howard. That information from Leica Camera technical staff who Stefan Daniel asked to respond to us, I think. I removed the personal details in line with Forum policy and of course anyone can contact Leica Camera Customer Service as I did.

 

I know nothing about the Phase One fix options personally and as shown I have not experienced any problems myself.

 

I was interested in the detail that Leica Camera provided on the "why". I posted it as a sticky in the M9 FAQ. I understand that this information has now been cross posted to the RFF. I shall cross post to the LUG for those interested there. Known facts are much better than speculation People will read what they want from the information naturally.

[Rick]

I wrote to Customer Service in Solms once again after these reports and received this information back. As when the broken cover glass reports emerged, my question was referred directly to Stefan Daniel who responded very promptly and openly. On this red edge question, technical staff provided this information to me. I have quoted the response verbatim. I would add my personal comment that I have never experienced any visible practical problem at all from about 2,000 frames with my M9. I posted my examples from all of my lenses from 35 and wider.

 

...............................................................................................................

"Thank you very much for your information.

We have had some complaints about the red edges, you are right. I will give you the full technical explanation for that.

On the one side, we do have a very short distance from the lens to the sensor (for this reason the m-system is so compact), on the other side the pixels of the ccd sensor are not symmetric. This combination produces a totally colored and inhomogenious RAW image. To provide a plain and neutral image, we do a lot of complex compensation, sensor corrections and lens vigentting corrections. What we achieve is a nearly neutral image, but unfortunately just "nearly" neutral. The tolerances for the M9 are even tighter as for the M8 with IR-filter, but unfortunately there are some very slight deviations towards red, the obviously are not acceptable for some special applications.

We did an improvement in a previous update, using the vignetting correction, this improved the effect visibly. But unfortunately we had to recognize that even this was not enough. There are still some complaints, therefore we decided to apply additional correction algorithms for the critical lenses. But developing these need unfortunately some time, because some sequences in the internal data management have to be changed, this is always a difficult and risky modification. Of course correcting this on a PC is an easy thing. But letting the camera do this needs a very simple algorithm, because we would not accept any delay in DNG mode.

We are planning to offer this update as soon as possible, but it will need at least until spring 2011.

I hope this answer helps you understand the confusion about the red edges."

...................................................................

 

This is a fantastic post. I am amazed and impressed that Leica is capable of this kind of transparency. We all know that they are up against difficult design challenges to keep the M camera as compact as it is and to be able to use legacy lenses. I wish they would continue to update us on challenges like this in a transparent manner (like this). I don't have a problem at all with them having problems, but it is irritating to be left in the dark and not even know that they acknowledge that we are having problems. Just a post that they understand and are working on the issue would be so appreciated. A note about the technical challenge (like this one) would go a long way to helping us understand. Thanks Goeff.

[charles-k]

Thank you Geoff Fantastic reponse to an issue that has been aknowledged and being addressed.

 

Howard, I do have this issue for lenses 35 FL and less, but it is easily corrected in Capture One Pro or Corner Fix. This is following on from Alan's earlier post

 

I have four lenses that have difference responses regardless of the lens coding I use: 35 Summilux Asph II, 28 Cron, 28 Elmarit and ZM 25/2.8

 

Thanks to John Black, and his procedure, just take a shot with each lens +1 exp into the sky away from the sun with a Capture One LCC pocket calibrator or Expo Disc. Once saved this can then be quickly applied to your shots as part of the RAW conversion. I have had to only take one shot at f/2.8, but it can be taken different f stops (typically at f/8) settings should you require a more exact correction.

 

I have two M9 bodies, so this is important part of my workflow for now. Even when the firmware algorithm is released the procedure with a Capture One may prove to be more accurate, as it very specific and unique for each M9 and WA lens configuration.

 

I also use Sandy's Corner Fix which is excellent, when just working with DNG files.

[wlaidlaw]

Geoff,

 

Many thanks for your efforts with Leica. I have been trying the C1 fix as well as as Cornerfix. The issue with this is that any shoot may have shots with a number of different lenses, so it is a pain to check on EXIF to see what correction should be applied. In particular, when using the WATE, I do not even know from the EXIF, which focal length I am using. I see this as an issue also for a firmware upgrade, as again, the camera does not know the focal length that the WATE is set as. From my experience, both red edge and asymmetric (worst at lower left corner) vignetting is quite visible at 16mm but for real life shots as opposed to test shots of white walls, close to irrelevant at 21mm. I suppose Leica will have to correct for 18mm as a compromise. I don't see it happening but ideally there would be two firmware upgrades; one for cameras which do exhibit red edge and another for cameras which don't.

 

I said in a earlier post that I had not noticed red edge in real life pictures with my 25/2.8 Biogon. Just to check, I took an Expodisc shot today with it to set up a C1 correction and sure enough, there they are, together with the asymmetric vignetting. The posted shot is the out of camera JPEG and uncorrected apart from image size. My M9 was set to manual lens detection and 24/2.8 11878 Elmarit. Maybe I was better not knowing.

 

Wilson

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!

[parsod]

Thank you Geoff for sharing this information. It is good to hear it directly from the source which also confirms what the Nordic agent told me (even if it was in a much shorter format and in Danish:)). In the meantime I think it is nothing that prevents us from making great photos with the equipment we have even if it might require some more PP work from time to time.

 

/Paer

[thighslapper]

I wrote to Customer Service in Solms once again after these reports and received this information back. As when the broken cover glass reports emerged, my question was referred directly to Stefan Daniel who responded very promptly and openly. On this red edge question, technical staff provided this information to me. I have quoted the response verbatim. I would add my personal comment that I have never experienced any visible practical problem at all from about 2,000 frames with my M9. I posted my examples from all of my lenses from 35 and wider.

 

...............................................................................................................

"Thank you very much for your information.

We have had some complaints about the red edges, you are right. I will give you the full technical explanation for that.

On the one side, we do have a very short distance from the lens to the sensor (for this reason the m-system is so compact), on the other side the pixels of the ccd sensor are not symmetric. This combination produces a totally colored and inhomogenious RAW image. To provide a plain and neutral image, we do a lot of complex compensation, sensor corrections and lens vigentting corrections. What we achieve is a nearly neutral image, but unfortunately just "nearly" neutral. The tolerances for the M9 are even tighter as for the M8 with IR-filter, but unfortunately there are some very slight deviations towards red, the obviously are not acceptable for some special applications.

We did an improvement in a previous update, using the vignetting correction, this improved the effect visibly. But unfortunately we had to recognize that even this was not enough. There are still some complaints, therefore we decided to apply additional correction algorithms for the critical lenses. But developing these need unfortunately some time, because some sequences in the internal data management have to be changed, this is always a difficult and risky modification. Of course correcting this on a PC is an easy thing. But letting the camera do this needs a very simple algorithm, because we would not accept any delay in DNG mode.

We are planning to offer this update as soon as possible, but it will need at least until spring 2011.

I hope this answer helps you understand the confusion about the red edges."

...................................................................

 

So it is due to the assymetry of the array in the CCD sensor as I suggested and the acute angulation and the use of microlenses......

[wattsy]

the pixels of the ccd sensor are not symmetric

 

That's very interesting. Does anyone more technically minded (Michael?) know why this might be so.

[sandymc]

A Bayer pixel "quad" is:

 

RED GREEN

GREEN BLUE

 

So on the left side of the sensor, the light hitting the red cells is at a greater angle than the light hitting the blue pixel.

 

But on the right hand side, the light hitting the blue pixel is at the greater angle. And the greater the angle, the greater the light loss, because of micro-lens efficiency fall-off with angle, etc.

 

Essentially, the angle (which drives the light loss) is symmetrical round the center of the sensor, but the Bayer array is not perfectly symmetrical, so you get a left to right color drift.

 

However, I've always had a great deal of trouble believing that this was the only factor in red edge. The effect just seems much greater than the physics allow for.

 

Sandy

[hoppyman]

Wilson you never saw a problem in real world photographs with that Biogon lens. Then you shot an artificial target specifically intended to show the effect at its most visible. I can see a solution in there

 

Regarding your WATE, my understanding is.that the in-camera corrections are focal length specific if the separate manual options are selected. I may be wrong. I only ever shot a couple of frames with one. At least those settings (if done manually) are reported in EXIF but the camera has no way of knowing what length is used with auto detection as you say (unlike with the 28-35-50). I guess that manual option is a possible workaround for you in your particular situation?

 

Worth keeping in mind that the inter-related vignetting corrections in camera for any specific lens are always for the best case anyway since the aperture is not accurately determined and Leica Camera want to avoid possible over-correction (for smaller apertures). It is all a tricky and complex balancing act we are learning more about.

[bocaburger]

There are still some complaints, therefore we decided to apply additional correction algorithms for the critical lenses.

 

The worst red-edger I found is the non-ASPH 21mm Elmarit-M, which is after all a discontinued lens. I hope Leica considers it "critical", not just lenses they are currently selling.

[adan]

I also hope the 21 pre-ASPH Elmarit is among the lenses that will get new corrections. Half the time the red edge just fades into the scenery - but sometimes it is nasty. Since it is supported by 6-bit coding, there are already corrections for this lens in the firmware. They just need better values substituted.

 

Thighslapper - you win! I would not have thought an asymmetry of 13 microns (6.8 microns on each side in opposite directions) would create a red band 4mm wide. But I guess it is a cumulative effect.

 

Sandy, as I see it (trying not to duplicate everything SD said, and it is stuff you personally know inside and out, I'm sure) - Leica has multiple problems to solve.

 

1) overall monochromatic vignetting, part of which is simply natural to short-focus wide lenses (i.e. we saw it on film, too) and part of which is specific to digital sensor architecture.

 

2) cyan drift in the corners due to light going through the IR filter at an angle - at 45° for a 21, that is an effectively "greener" filter by approx 1.41x. Same as we saw in the M8 with external filters.

 

3) the assymmetric sensor problem

 

Leica fixes (1) by amplifying the signal from the corners (essentially an artificial localized increase in ISO). They fix (2) by progressively adding in red to the corners/edges (and it should be noted that "red edge" is therefore an overcorrection to one side due to (3), not an undercorrection, which would be cyan).

 

Kodak and Leica noted that the M9 sensor has a "different" red filter in the RGBG pattern than the M8 sensor - different in what way I don't know.

[C_R]

This might be interesting for Capture One users, before buying an expo disk, Sinar WB plate or anything expensive:

 

Product Comparison: White Balance Filters (Professional Photographer Magazine Web Exclusives)

 

Among those WB tools and cards, a cheap Melitta filter was fine for making a LCC profile, to be used in Capture One.

 

I tried that out, and it worked perfectly for a critical snow/ fog scene.

Possibly also nice for making CornerFix profiles, better than a white wall for example

[Advertisement]

Hallo,

Look here: Leica M

[mjh]

That's very interesting. Does anyone more technically minded (Michael?) know why this might be so.

Sorry, I have no idea. I am wary of all these vague supposed “explanations” and “the pixels of the ccd sensor are not symmetric” is just more of the same.

 

Yes, the sensor pixels aren’t symmetric in that their microlenses are offset from the center by some amount that depends on the pixel’s distance from the optical axis. But is it really this kind of asymmetry that Leica refers to, and if so, how is this supposed to explain the red edges phenomenon? The effect of microlens shifting should be symmetric around the optical axis which doesn’t agree with the known facts of the red edges issue.

 

Or do they refer to the asymmetrical Bayer pattern of the RGB filters? As Sandy explained, the pattern is

 

RG

GB

 

Due to differences in the incident angle, the left column (and thus the red-sensitive pixel) might receive a little less light then the right column containing a blue-sensitive pixel – assuming that these 2 by 2 pixels are near the left edge of the sensor. But while the incident angle does indeed grow from the center of the sensor to its edges, the difference from one column of pixels to the next is much too small to make any difference that could account for a colour shift. And even if it the difference was greater, it still wouldn’t offer a plausible explanation. Let’s take a look at a slightly larger crop of sensor pixels:

 

RGRG

GBGB

 

If we still assume that pixels further to the left receive less light than pixels further to the right, the red pixel in the first row gets less light than the blue pixel in the second. But this blue-sensitive pixel receives less light than the red pixel in the third row … and so on. Even if the variance in the incident angle did affect neighbouring pixels to any noticeable amount, the demosaicing step interpolating between neighbouring pixels would suffice to compensate for this effect – more light hitting the pixel to the right would cancel out less light received by the pixel to the left.

 

But there is still another possible effect. Due to the incident angle growing towards the sensor’s edges, the light hitting a pixel near the edges might end up being absorbed by the wrong photo diode. Let’s look at the Bayer pattern again:

 

RGRG

GBGB

 

Light hitting the second column might get filtered by this column’s green or blue filters, but be registered by the photo diodes of the red or green pixels, respectively. Likewise, light hitting the third column (red/green) would end up being registered by the second row (green/blue).

 

If the sensor is lit by uniform white light, each pixel would get hit by the same amount of light, so one might assume that there would be no colour shift: A supposedly red-sensitive pixel would actually register green light and vice-versa, but as it was the same amount of light, it wouldn’t make a difference. Only it does. The transmission of the the green filters is generally greater than that of the red and blue filters so the light hitting the photo diodes is not uniform at all. What would be the effect? Supposedly red-sensitive pixels would register the stronger green light, and the same goes for the blue-sensitive pixels. Green-sensitive pixels would receive the somehwat weaker red or blue light. So while the camera expects less light to be registered in the red and blue pixels and gives these pixels an extra boost to compensate, the light actually registered by those photo diodes is the brighter green light. The net effect would a colour shift towards red and blue, i.e. magenta.

 

However, the effect on the right hand side of the sensor would be just the same. Despite the Bayer pattern’s asymmetry it would again be the red and blue pixels being favoured by getting hit by green light, so we would expect a symmetrical effect – an equal amount of a shift towards magenta on both sides. Nothing asymmetrical and certainly no Italian flag.

[ho_co]

... Regarding your WATE, my understanding is.that the in-camera corrections are focal length specific if the separate manual options are selected....

Interesting if that's so. When the lens was introduced with the M8, the explanation was that it had been designed so that all three focal lengths would need the same correction IIRC. Only when the IR-Cut filter was introduced did the three focal lengths need different corrections.

 

I'm not disagreeing, Geoff, just curious. Maybe someone with a WATE could shoot a neutral target without changing the focal length, but changing the lens ID information given the camera manually.

 

 

... Kodak and Leica noted that the M9 sensor has a "different" red filter in the RGBG pattern than the M8 sensor - different in what way I don't know.

I think Sandy and KammaGamma indicated that it is a different color.