Cv 15mm Versus Leica 18mm

[sandymc]

Advertisement (gone after registration)

Please take into account, that the "red shift" is not an issue of the M9 alone! It is as well visible - perhaps even more - with the 3.8/18 Super-Elmar on the M8. I think Leica has to tweak the in camera correction for this lens - perhaps for the WATE as well - in both cameras to avoid overcorrection of cyan shift.

 

That's right. So I think that with a bit of work, the issues at the moment can be resolved. So you probably shouldn't sell your wide-angle lens yet!

 

Sandy

[reven]

Well, yes if you have the lens which is perfect. Otherwise Leica can't optimize anything. Leica optimizes for the "perfect" 18, 21 or Wate. At that my last comment was directed. If you have one, you will have perfect in camera corrections. CornerFix is a great, start, but even better would be a real ColorCast correction alla MFDB. It is just amazing what it can do.

 

The important part is that everyone can make his OWN correction profile for HIS camera and lens combination. This is the ONLY way for perfect corrections.

[tashley]

Well, yes if you have the lens which is perfect. Otherwise Leica can't optimize anything. Leica optimizes for the "perfect" 18, 21 or Wate. At that my last comment was directed. If you have one, you will have perfect in camera corrections. CornerFix is a great, start, but even better would be a real ColorCast correction alla MFDB. It is just amazing what it can do.

 

The important part is that everyone can make his OWN correction profile for HIS camera and lens combination. This is the ONLY way for perfect corrections.

 

I agree and I use this method with my Phase One / Schneider setup BUT it is only just practical for slow-paced landscape work. For use with a handheld camera with a faster shooting rate it would require LCC frames to be shot for every aperture and every colour temperature on every lens and the user would then need a notebook to record the aperture used for every frame.

 

Trust me, it ain't gonna happen! But I have already asked Phase One to consider opening up their LCC feature to other cameras in C1 V5... just cos it would be useful sometimes.

 

T

[UliWer]

I hear the words, but I don't really understand them.

 

Which differences in individual lenses of the same type could cause things like the red shift we are talking about? I can imagine mechanical differences, which may cause misfocussing or even optical ones which might result in a variance of resolution (very theoretically), but I cannot see any correspondance between these variances and the red shift.

 

From what I understand very simplistically, the cyan-correction mainly depends on the amount of a lense's vignetting: the darker the edges become the more correction is applied. The Wate, 3.8/18 or 2.8/21 show different vignetting, so the individual in-camera-correction for one lens-type should not apply to another one. Therefore Leica introduced the individual coding. Though I cannot imagine that the vignetting of one and the same lens type differs so much with individual examples, that the in-camera-correction does work for one example but not for the other one.

[EricC]

That's right. So I think that with a bit of work, the issues at the moment can be resolved. So you probably shouldn't sell your wide-angle lens yet!

 

Sandy

 

 

Hi Sandy,

 

Apologies to all, for jumping in on the CV15mm Versus Leica 18mm thread.

 

I was thinking of selling my CV12mm until I received your help, and saw others on the board with similar problems working hard to get it fixed.

 

I still have not managed to photograph anything approaching 18% grey in normal sunlight to use as a reference file for several reasons.

 

One, when I do find something smooth enough and in sunlight it is not big enough for the 12mm FOV and Two, when I find something large enough for the FOV and in sunlight it is not smooth enough even when defocussed, with the DOF from the CV12mm it is rather difficult to get close enough to fill the frame and be sufficiently defocussed even at infinity.

 

I am still trying, however, having looked at Tim's photo's at the start of this thread the location of the red i am seeing matches his exactly, just a little more of it in mine.

[jml]

Have you already tried the CV15 on the M9 with an UV/IR filter? I did and it works!

By the way when I received (one of ) the (first) WATE the appropriate UV/IR filter was not yet available and on the M8 the red shift was also obvious. Once the UV/IR filter received and placed, it disappeared.

[UliWer]

Advertisement (gone after registration)

Have you already tried the CV15 on the M9 with an UV/IR filter? I did and it works!

By the way when I received (one of ) the (first) WATE the appropriate UV/IR filter was not yet available and on the M8 the red shift was also obvious. Once the UV/IR filter received and placed, it disappeared.

 

I only use the 3.8/18 with filter on the M8 and the red shift is - sometimes - there.

[reven]

Well I don't think so. I still don't have my M9, but as far as I tried it with my M8 it should work out if you just shoot some generic corrections. So for example for the 18mm lens:

 

1. Wide Open (back light)

2. Wide Open (light coming from the front)

3. Medium aperture f8 (back light)

4. Medium aperture f8 (light coming from the front)

 

If you use the lens at 16 you perhaps should do the same thing there.

 

However as with my P65 and Rodenstock lenses, as long as I don't shift. I am fine with exactly these LLC files.

 

So One would not need to shoot a LLC file for every image. The approach above would already be MUCH better than any Leica correction can do, because it is specialized to one's lens camera combination.

[sandymc]

Well, the good news is that with a the help of images from Tim, Eric and Jono, the situation is starting to become at least a bit clearer, although not entirely clear. The evidence does seem to be pointing at lens de-centering (the optical center of the lens not being exactly aligned with the optical center of the sensor) being at least a significant contribution to what's going on. Hopefully I'll be able to post some results tomorrow.

 

Sandy

[sandymc]

I've done some further work on this question, and in summary:

 

1. I think the issue that we are seeing is caused by de-centering between the lens and sensor, leading to asymmetries in vignetting. It is not clear where exactly the de-centering is coming from, but the most likely explanation is manufacturing variations in the lens. This de-centering does also exist on the M8, but the impact is small enough to mostly be invisible.

 

2. It also seems likely that there is variation with aperture; the problem seems to get worse at larger apertures.

 

3. Leica's best course to fix the problem, or at least make it less visible, is probably just to reduce red correction for the 18mm. This will leave the right hand side of the images Tim showed under compensated, but that's a lot less visible that over compensation. This can easily be done in firmware.

 

4. CornerFix, in its current form, can't fix this problem. The reason is that CornerFix is designed only to deal with symmetrical vignetting. I'm currently working on a new version that removes this restriction.

 

For those interested in a whole bunch of charts, and a far more detailed explanation, I've posted one here: ChromaSoft: Vignetting Correction Issues on the Leica M9

 

BTW, I don't know why this happens mostly on the left hand side of images. I can only speculate that it has something to do with either the way Leica build and calibrate lenses, or the way the M9 is built.

 

Sandy

[Ario Arioldi]

I've done some test myself wit CV 15 (M version), Zeiss 18 and Leitz 21 (pre-aspherical). There is an evident asymmetry in color shift in the pics taken with CV 15 and to a lesser extent with Zeiss 18 and this asymetry basically disappears wit te Leitz 21.

It is a common observation that this asmmetry is localized in the left side of the pics.

Cheers,

Ario

[Shootist]

Pure speculation here.

But didn't Mark Norton post something about the battery being in the way for Leica to put a full frame sensor in the original M8 body?

 

If that is true then the question to asks is did Leica move the sensor ever so slightly to the right hand, shutter release, side of the body to be able to fit a FF sensor.

 

If they did this may have something to do with the vignetting on the left hand side of the image because the lens and image circle is now off center to the sensor and light rays nearer the edge of the image circle are darker then farther inside the image circle.

[EricC]

Wow Sandy, I have just read your detailed explanation.

 

I can only thank you once again for all of the very hard work and hours that you are putting into Cornerfix so that we can use almost any lens we wish on the M8 and M9.

[tashley]

Pure speculation here.

But didn't Mark Norton post something about the battery being in the way for Leica to put a full frame sensor in the original M8 body?

 

If that is true then the question to asks is did Leica move the sensor ever so slightly to the right hand, shutter release, side of the body to be able to fit a FF sensor.

 

If they did this may have something to do with the vignetting on the left hand side of the image because the lens and image circle is now off center to the sensor and light rays nearer the edge of the image circle are darker then farther inside the image circle.

 

Smart thinking Batman! Sandy, Mark, any thoughts? Do we see assymetrical vigging? Is the frame less sharp or more noisy on one side than the other?

 

T

[fotoism]

Pure speculation here.

But didn't Mark Norton post something about the battery being in the way for Leica to put a full frame sensor in the original M8 body?

 

If that is true then the question to asks is did Leica move the sensor ever so slightly to the right hand, shutter release, side of the body to be able to fit a FF sensor.

 

If they did this may have something to do with the vignetting on the left hand side of the image because the lens and image circle is now off center to the sensor and light rays nearer the edge of the image circle are darker then farther inside the image circle.

 

But if that's the case, would it not cause more than uneven vignetting on both sides, such as focusing...etc. due to the slight difference of lens to sensor distance?

[hoppyman]

I've done some further work on this question, and in summary:

 

1. I think the issue that we are seeing is caused by de-centering between the lens and sensor, leading to asymmetries in vignetting. It is not clear where exactly the de-centering is coming from, but the most likely explanation is manufacturing variations in the lens. This de-centering does also exist on the M8, but the impact is small enough to mostly be invisible.

 

2. It also seems likely that there is variation with aperture; the problem seems to get worse at larger apertures.

 

3. Leica's best course to fix the problem, or at least make it less visible, is probably just to reduce red correction for the 18mm. This will leave the right hand side of the images Tim showed under compensated, but that's a lot less visible that over compensation. This can easily be done in firmware.

 

4. CornerFix, in its current form, can't fix this problem. The reason is that CornerFix is designed only to deal with symmetrical vignetting. I'm currently working on a new version that removes this restriction.

 

For those interested in a whole bunch of charts, and a far more detailed explanation, I've posted one here: ChromaSoft: Vignetting Correction Issues on the Leica M9

 

BTW, I don't know why this happens mostly on the left hand side of images. I can only speculate that it has something to do with either the way Leica build and calibrate lenses, or the way the M9 is built.

 

Sandy

Hi Sandy. I am following this technical discussion with great interest. I have read your ChromaSoft article of course. Kudos to you and the contributors for good analysis and to you in particular for continuing to develop and make available for free, solutions for post processing.

Reasonable to say that the corner performance on 24x36 with the widest lenses is a worst case. (Actually I won't get the 18 now that the M9 is becoming available and I have my ZM 18 for sale. The 24 FoV is plenty for me for a 24x36 rangefinder).

 

Is it possible for one of the contributors to repeat the testing with the even diffuse light source on film? Could the natural vignetting then be subtracted to give the actual plotted error in your analysis?

When you mention that the effect is variable dependant on aperture, is this not just related to the natural vignetting of wide lenses? Leica quotes the vignetting for the 18 Super Elmar to be 2.3 stops at f/3.8. The decrease in vignetting can be clearly seen on their graph as the aperture is stopped down.

 

The widest Leica M lens (Summicron 28 ASPH.) that I have shot extensively on the M7 etc has always shown quite characteristic vignetting. That one at f/2.0 shows a very similar vignetting figure to the Super Elmar at f/3.8 and significantly less at comparable aperture.Can the contributors provide a comparison test with that or other wide modern Leica lenses?

I don't follow how the demonstrated left edge asymmetry can be related to lens build and/or calibration. The element shapes by definition must be symmetrical since that is the result of the (symmetrical) grinding/polishing methods employed on circular blanks. Seems reasonable to presume that also applies for the pressed blank mouldings prior to that point. Some lens elements must rotate during focusing movement too, whereas you are seeing a consistent bias. I would think that tolerances and error can be discounted in the case of Leica lenses.

 

Sensor positioning within the M8/9 bodies is known to now be extremely precise, measured and corrected to .01 mm increments at three points. I know that it has been speculated since that it could be displaced against the centre of the image circle for physical space reasons, however I think it would be practically impossible to allow for that in the micro lens design/pattern. That being so any speculated displacement would need to be addressed in the firmware functions, if it existed at all or was significant. Again the WATE and Super Elmar, each wide open would be the absolute worst cases.

 

On the correction, Stefan Daniel definitely states that it is not aperture dependant and (is) based on the best case, else the correction may be excessive for all but the worst case. I'll send a query to Leica to ask if they can confirm that comment in the interview.

[Shootist]

Sensor positioning within the M8/9 bodies is known to now be extremely precise, measured and corrected to .01 mm increments at three points. I know that it has been speculated since that it could be displaced against the centre of the image circle for physical space reasons, however I think it would be practically impossible to allow for that in the micro lens design/pattern. That being so any speculated displacement would need to be addressed in the firmware functions, if it existed at all or was significant. Again the WATE and Super Elmar, each wide open would be the absolute worst cases.

 

Yes but those 3 point mounting positions are for spacing the sensor from the lens mount and, as far as I know, there is no adjustment for up/down/left/right movement.

That is all controlled by the placement of the mounting posts on the body and the placement of the holes in the sensor mounting board.

 

Wouldn't the slight displacement from center cause just this type thing to happen, especially with the widest of lenses and diminishing as the lenses get longer?

[sandymc]

When you mention that the effect is variable dependant on aperture, is this not just related to the natural vignetting of wide lenses? Leica quotes the vignetting for the 18 Super Elmar to be 2.3 stops at f/3.8. The decrease in vignetting can be clearly seen on their graph as the aperture is stopped down.

 

Not sure. That explanation makes sense, but I'm very aware that as yet our understanding of what's going on here is incomplete. I'm hoping that Tim can be persuaded to use his rig to make a series tests at different apertures. But doing long series of tests is actually quite time consuming.

 

I don't follow how the demonstrated left edge asymmetry can be related to lens build and/or calibration. The element shapes by definition must be symmetrical since that is the result of the (symmetrical) grinding/polishing methods employed on circular blanks. Seems reasonable to presume that also applies for the pressed blank mouldings prior to that point. Some lens elements must rotate during focusing movement too, whereas you are seeing a consistent bias. I would think that tolerances and error can be discounted in the case of Leica lenses.

 

I don't have a good understanding of how Leica builds and calibrates their lenses, but my simplistic understanding is that that its more important that the lens elements be aligned with each other than with mechanical center of the lens. So from an image quality perspective you can tolerate, e.g., the aperture blades not being quite aligned with the lens elements more easily than e.g., an internal lens element being misaligned with another element. But really, optical engineering isn't my area of expertise; I can only look at the sensor data and make inferences from that.

 

Any Leica optical engineers out there that want to provide some insight?????

 

Sensor positioning within the M8/9 bodies is known to now be extremely precise, measured and corrected to .01 mm increments at three points. I know that it has been speculated since that it could be displaced against the centre of the image circle for physical space reasons, however I think it would be practically impossible to allow for that in the micro lens design/pattern. That being so any speculated displacement would need to be addressed in the firmware functions, if it existed at all or was significant. Again the WATE and Super Elmar, each wide open would be the absolute worst cases.

 

Yes. I find it difficult to believe that what is happening is simply mechanical misalignment of the sensor; a back-of-the-envelope calculation suggests that the sensor would have to be more than 0.5 mm off to duplicate what I'm seeing. Admittedly in the xy axis, not the depth (focus) axis. But still, by optical standards, that's a mile. So I think something optical rather than mechanical is happening.

 

Sandy

[hoppyman]

Sandy, those of course are just my humble contribution. You guys are doing sterling work well beyond my ability!

The lens grinding and polishing detail is from personal observation at Solms.

As I understand it de-centring is undesirable for performance and is particularly tightly controlled as part of the manual assembly for Leica lenses. This is one of those things that you pay the extra money for.

However I am describing the lens elements relationships to each other within the lens body and the mount. That is different to optical centre displacement relative to the sensor xy axis?

The natural vignetting can be easily demonstrated from the graphs published for the lens (found at the technical data link from the page).

Leica Camera AG - Photography - NEW: LEICA SUPER-ELMAR-M 18 mm f/3.8 ASPH.

That does not prove it is the sole or primary cause for what you are describing, however it seems reasonable that there is a relationship.

Naturally the optical factors are much more complex than just that. Some vignetting too is part of the compromises in the designs but can actually make a positive contribution that can be observed in the MTF diagrams too (in the outer zone).

[Advertisement]

Hallo,

Look here: Leica M

[tashley]

I'm hoping that Tim can be persuaded to use his rig to make a series tests at different apertures. But doing long series of tests is actually quite time consuming.

Sandy

 

At this point a secondary Leica issue kicks in: as of yesterday I have returned both my M9 and my 18mm F3.8 for calibration. The RF on the M9 has been subtly out of whack since delivery. It won't take any of my existing lenses quite as far as infinity and though its focus is very accurate at short to mid-range, anything further out has been giving me a higher 'miss' rate. On the other hand the 18mm goes straight past infinity with abandon, galloping on into the even further reaches of the universe. If there are any.

 

Anyone else got an 18mm 3.8? I have now tried my own one and two others in two different stores and they all do this so maybe it's a design 'feature'?

 

Grr... so, sorry, but no more tests for a week or so at least!