Cyan Vignetting Issue

[atufte]

Just come to think of this:

How on earth will Leica be able to fix the cyan vignetting issue, when
even with coded lenses the camera can't tell what aperture being used
on each shot...?

Because the cyan vignetting is much worse on larger apertures than it is on
small apertures, like my coded 35/2 shows slight cyan vignetting @ f/2 and it's
not evident at all from f/4 and up...

On wider lenses this problem will be very hard to correct, if aperture selected
can't be recognised by camera/software, any thoughts on this issue...?

(PS! Sorry if this topic have been discussed before....)

[wparsonsgisnet]

Alexander, it has been discussed and this particular point has been in the discussion, just not talked about much.

I agree that it is a problem for Leica, which is why I want cyan correction in software. In that form it would work for all lenses, not only for coded lenses.

[cme4brain]

Originally Posted by atufte

Just come to think of this:

How on earth will Leica be able to fix the cyan vignetting issue, when
even with coded lenses the camera can't tell what aperture being used
on each shot...?

Because the cyan vignetting is much worse on larger apertures than it is on
small apertures, like my coded 35/2 shows slight cyan vignetting @ f/2 and it's
not evident at all from f/4 and up...

On wider lenses this problem will be very hard to correct, if aperture selected
can't be recognised by camera/software, any thoughts on this issue...?

(PS! Sorry if this topic have been discussed before....)

I imagine Leica will deal with this problem like it has the others- forced compromises. The variation of cyan-vignetting with aperture is (relatively) minor compared to the overall correction of the cyan-vignetting depending on the lens focal length, it seems. Also, different brand lenses will vignette differently (see Sean Reid's evaluation of the CV 28mm- it hardly cyan-vignetts at all) as will older-design Leica lenses. It seems that the physical constraints of a digital rangefinder preclude some compromises, and if LeicaNuts weren't expecting perfection, it would be less of a problem. IR cut filters are necessary on the lens (read the current LFI magazine article- very descriptive of the difficulties with sensor-IR blocking technologies), even if in some instances they cause reflections/ghosting. The ability to use these wonderful Leica/Zeiss/Voigtlander lenses with such a great image chip is the focus here! SLR owners would give their left lens cap to have the sharpness of these lenses! The main reason I don't spend LESS money and buy a Canon 5D? The sensor is better than almost all canon lenses out there- it shows how crummy some SLR lenses actually are! The edge softness and reduced resolution of those expensive Canon L (and other lenses) compared to Leica mount lenses is so obvious, that I would really not think we should worry about aperture-related cyan-vignetting in the corners.

[scott kirkpatrick]

Originally Posted by atufte

Just come to think of this:

How on earth will Leica be able to fix the cyan vignetting issue, when
even with coded lenses the camera can't tell what aperture being used
on each shot...?

Because the cyan vignetting is much worse on larger apertures than it is on
small apertures, like my coded 35/2 shows slight cyan vignetting @ f/2 and it's
not evident at all from f/4 and up...

This is a pretty misunderstood topic.

The vignetting comes in two parts -- overall (less light reaches the corners), and red vignetting (the fraction of light that should be red is reduced in the corners, leaving a net cyan tinge). The first part depends on aperture, sometimes strongly. And the dependence is different for different lenses in funny ways -- some show excess overall vignetting only when wide open, say at 1.4 and 2, while the vignetting is unchanged above 2.8. Others show a steadily decreasing vignetting from wide open to f/8 or so. You can see this in Leica's pdf documents for each lens on their website.

But the red vignetting is independent of aperture. I obeserved that with Sean Reid's examples, and with a more recent post based on a test shot with the 21mm Elmarit.
So I believe the color effects can be handled in firmware, at least for coded lenses, but overall vignetting will require some additional work, unless Leica figures out how to estimate aperture.

scott

[carstenw]

The M8 *can* tell what aperture is being used. The little blue dot next to the large red Leica measures the light in the scene, and this is compared to the light entering the lens. The difference is attributed to the aperture setting, when compared to the theoretical max from the lens.

[scott kirkpatrick]

Originally Posted by carstenw

The M8 *can* tell what aperture is being used. The little blue dot next to the large red Leica measures the light in the scene, and this is compared to the light entering the lens. The difference is attributed to the aperture setting, when compared to the theoretical max from the lens.

I've seen what is either shot to shot variation in the exposure, or the camera detecting the aperture change and making a slight correction, in the vignetting tests that have been presented so far. I can't tell if this is evidence that the camera is estimating the aperture, or just experimental error.

Sean Reid did an extensive set of tests (not all published), and only covered two apertures, showing one exposure for each, using 1.06 firmware, which does not attempt to correct for the IR-cut filter that is recommended. To do serious reverse engineering and really know what Leica's firmware is doing is about as much work as Leica's engineers will be doing to create the firmware in the first place. I think they have still considerable tuning to do, and I expect this to be an area in which the firmware steadily improves over the coming year.

scott

[marknorton]

Originally Posted by carstenw

The M8 *can* tell what aperture is being used. The little blue dot next to the large red Leica measures the light in the scene, and this is compared to the light entering the lens. The difference is attributed to the aperture setting, when compared to the theoretical max from the lens.

Carsten, I agree you could set up an M8 looking at an evenly illuminated surface and have the camera estimate the aperture in use in the way you describe.

Then, look at the real world factors which might spoil the party:

- different spectral responses for the two sensors
- different angles of view
- camera metering pattern
- effect of any filters on the lens
- different responses at extreme bright and dark

.. and so on. I know LFI said it could estimate the shooting aperture to within 1 stop but who knows if it's really doing this.

[atufte]

Originally Posted by wparsonsgisnet

Alexander, it has been discussed and this particular point has been in the discussion, just not talked about much.

I agree that it is a problem for Leica, which is why I want cyan correction in software. In that form it would work for all lenses, not only for coded lenses.

Sounds like a good idea to have the choice, in camera correction or post correction in raw conversion... But they probably already (and hopefully) have a solution for this coming up in fw 1.10 or later...what's seems odd to me is, they say they already have the fw fix for the vertical stripe/banding why not release that fw update now, and make the other fixes in later upcoming fw updates.. just to show some progress...

[chetccox]

They are still on vacation and not due to return to Solms until Monday the 8th. Maybe something will happen next week. Even just the corrected lens price list would be nice.

[wparsonsgisnet]

Not to start any noise, but what we need in the States is some European vacation standards!

What I said, about the day job getting in the way!

HNY to all.

[scott kirkpatrick]

Originally Posted by atufte

...what's seems odd to me is, they say they already have the fw fix for the vertical stripe/banding why not release that fw update now, and make the other fixes in later upcoming fw updates.. just to show some progress...

Testing is the reason why.

If they want to be sure that they have fixed the bugs they wanted to fix and haven't broken anything that was OK before, they have to test against previous test cases plus new examples that were to be fixed. This can get pretty time-consuming as the list gets longer. We want them to do it, and fix whatever turns up as a result, before releasing the next version of the firmware. So if they put out updated firmware every 8 weeks for a while, I'll be delighted. On that entirely hypothetical release schedule, we get 2-4 weeks to find what still hasn't been fixed, they get a month to work on further improvements, and there is two weeks left for testing.

scott

[mike prevette]

Again, the vignetting correction the M8 applies is simply based on the vignetting the chip creates due to the steap angle of the light hitting it. It does not correct for the vignetting that is already inherent to a lens. The aperture doesn't matter only the exit pupil of the lens matters.

_mike

[carstenw]

Originally Posted by chetccox

They are still on vacation and not due to return to Solms until Monday the 8th. Maybe something will happen next week. Even just the corrected lens price list would be nice.

Why do you think they are on holiday? I mean, maybe you are right, but here in Berlin there is no holiday after the 1st. In some parts of Germany, the 6th is a holiday, but this is on a Saturday this year anyway.

I did communicate with Leica today, wrt. my M6 and M8, both of which need focus adjustments. The M6 got sent off, and the M8 will wait for now.

[chetccox]

Carsten

I spoke to the Leica dealer here in Warsaw today, and he told me they would not be back to work until Monday. He communicates with them daily, and I wondered when he might have some wide angle lenses. He said all the emails sent now only get automatic responses and there would be no one to talk to until they return on the 8th. I did not personally try to call them and only took his word for it. I do believe he should know.

[carstenw]

Well, I spoke to Jakob Eberhard of the service department today. He mailed me back about 15 minutes after I mailed him. There are no special German holidays at the moment, although many Germans extend their Christmas/New Year's holidays by a few days, and maybe Leica has something special for some people.

[jaapv]

What can I say. I spoke to three different people a Leica today, the whole place seemed open. One or two of the persons I wanted to speak to were ill, however. I suppose they are as hit by the flu as we are.

[rvaubel]

Originally Posted by carstenw

The M8 *can* tell what aperture is being used. The little blue dot next to the large red Leica measures the light in the scene, and this is compared to the light entering the lens. The difference is attributed to the aperture setting, when compared to the theoretical max from the lens.

Is this feature really there? Has any one been able to confirm this?

Rex

[ho_co]

Originally Posted by rvaubel

Is this feature really there? Has any one been able to confirm this?

Rex--
That's what I've been asking as well in another thread. As remarked above, LFI said it was there, but I haven't heard it mentioned since.

It wasn't mentioned according to an acquaintance who attended the post-photokina Solms M8 introduction for some LHSA members.

Even the EXIF data don't show even max aperture, let alone estimated shooting aperture, isn't that correct?

Seems a reasonable question, and one to keep in our (and Leica's) sights.

[marknorton]

HC, if you have a coded lens, the EXIF data shows both the focal length and the maximum aperture but not the (estimated) aperture.

As I've said elsewhere, I can see that the blue dot could help you estimate the working aperture in specific controlled situations. In real-world shooting - and isn't that where the M8 has fallen short of our expectations? - there are lots of things which mean the blue dot and the TTL sensor are not looking at the same thing or are not responding the same way to either colour or light level.

Take a 90mm lens, point it into a dark corner at set the lens to f2.0. Then shine a light at the camera outside the frame. the blue sensor will think it's bright, the TTL sensor will say, "not as bright as that here" and the camera will say, "lens must be stopped down - looks like f90 to me". Except of course the lens only goes down to f16 and is set to f2.0 anyway.

Then try to implement image correction based on that aperture estimation and you'll get the same variability we're seeing from AWB.

[scott kirkpatrick]

Originally Posted by rvaubel

Is this feature really there? Has any one been able to confirm this?

Rex

Hi, Rex --

I'm trying to see if there is a different overall vignetting correction applied as a function of aperture setting, but I don't have very many examples to work with. In cases, like Sean Reid's tests of 28mm lenses, where there are shots at f/2.8 and f/8, I do see a little more correction at 2.8 than at 8, but not enough to keep the wide open shots as even in illumination as the f/8 shots. And the differences are so small that it's more likely that this is experimental error, and the true behavior is a compromise setting, leaving the f/8 a bit overcorrected in the corners and f/2.8 undercorrected.

Sergiolov, is there room on the mount of your 21/4.5 to mark it as an Elmarit 21/2.8 and see what happens? I guess you would have to return to the white wall that you tested on in http://www.leica-camera-user.com/dig...ormers-m8.html
and redo the uncorrected shots for comparisons. At least it is easy to take two shots close together, one with firmware correction and one without, by holding the frame preview lever first in what Leica thinks is the "right" position for that lens and then in another position.

I've passed the analysis of the shots in the 28mm article back to Sean Reid to consider postiing on his site as a sort of technical appendix to that already very long article (whenever he digs out from under his yearend workload). There's a more extended discussion of vignetting included in that. Because of the new news of the IR filters, I expect the behavior of the firmware to keep changing in this area for a while, maybe several iterations, until they get it optimized.

scott