Red edges and new firmware......

[UliWer]

Advertisement (gone after registration)

 

Lastly once in a while I keep hearing that even with the WATE, evidence of red edge can be seen. I thought it was coming from most who own this particular lens, that the WATE is one of the few ultra wide angle lenses that doesn't exhibit the red edge trait. Any conformation one way or another?

 

 

The examples given by viramati clearly show no red-edge issue at all.

 

Though here is an example with the WATE which does:

http://www.l-camera-forum.com/leica-forum/forum-zur-leica-m8/98843-problem-mit-dem-super-elmar-18mm.html#post1040557

 

This corresponds with my own experience with the 3.8/18 Super Elmar (and the M8!), that artificial light and white balance have a big influence on the phenomenon.

[vanhulsenbeek]

I have the WATE and have no problem with fringing.

 

 

That is my experience as well. The WATE is probably saved by its f/4 aperture: fine with me:)

[Bob Ross]

2) The effect can be seen faintly in the M8 as well as in the M9, if you take a lens with a really deep rear element and shoot a white wall, wide open, in warm room lighting, so it is not an M9 bug, just a problem that hadn't been noticeable before.

 

 

scott

HI Scott,

I was unaware that the problem showed at all with the M8. My "white wall" is an Expo/Disc, a consistant light source and manual WB to eliminate tints. My widest lens is a 28 Elmarit Asph. My original tests were to check the corners and I might have written off any faint gradient as lighting inconsistancy. I am curious, so I'll check again. The histogram should be a peak and any other lumps would indicate problems, to look for. I'll test with and without the IR filter, too.

Bob

[sandymc]

I think we should put to bed any ideas that this is caused by lens centring problems.

 

Well, we need to be a bit careful here - for the examples I've seen, the further off the measured centering of the lens is, the worse the red edge problem is. So I'm fairly sure that if we were to measure e.g., the WATEs and M9s that do have a red edge problem versus those that don't, we'd find that the lens centering as measured was good on the ones that don't have the problem, worse on the one that do.

 

But correlation isn't causality - there could be other things going on:

 

Firstly, whatever the root cause is could be itself causing the off-center reading - what I measure is the center of the vignetting, but that isn't necessarily the optical center of the lens.

 

Secondly, lens centering may make the problem worse, but its very likely that something else needs to be wrong for it to exist at all.

 

My personal theory (at the moment, and somewhat thanks to a discussion with Jono Slack) is that the microlens array on the sensor is slightly out of register, by less than a pixel's width. This results in light leakage between photosites which together with the RGGB pattern of the sensor itself gives the red-on-the-left result. But that's just my guess. At the moment. But as a theory, it also explains why de-centering contributes - the more de-centered a lens is, the more difference there is in what microlenses on the left and right of the sensor see, so making the problem worse, but not causing it. Of course, it may be possible that just having the microlens array somewhat out of register is enough to give the appearance of lens de-centering without it actually existing - that's the kind of test the Leica could do easily.

 

Sandy

[thrice]

Stefan Daniel says there will be a firmware to correct this issue in approximately 3-4 weeks.

 

As there is a lot of testing to do, obviously that date may stretch out a little.

 

We can all breathe a little easier and (at least those with Leica) don't have to sell all our wides

[Alnitak]

Well, we need to be a bit careful here - for the examples I've seen, the further off the measured centering of the lens is, the worse the red edge problem is. So I'm fairly sure that if we were to measure e.g., the WATEs and M9s that do have a red edge problem versus those that don't, we'd find that the lens centering as measured was good on the ones that don't have the problem, worse on the one that do.

 

But correlation isn't causality - there could be other things going on:

 

Firstly, whatever the root cause is could be itself causing the off-center reading - what I measure is the center of the vignetting, but that isn't necessarily the optical center of the lens.

 

Secondly, lens centering may make the problem worse, but its very likely that something else needs to be wrong for it to exist at all.

 

My personal theory (at the moment, and somewhat thanks to a discussion with Jono Slack) is that the microlens array on the sensor is slightly out of register, by less than a pixel's width. This results in light leakage between photosites which together with the RGGB pattern of the sensor itself gives the red-on-the-left result. But that's just my guess. At the moment. But as a theory, it also explains why de-centering contributes - the more de-centered a lens is, the more difference there is in what microlenses on the left and right of the sensor see, so making the problem worse, but not causing it. Of course, it may be possible that just having the microlens array somewhat out of register is enough to give the appearance of lens de-centering without it actually existing - that's the kind of test the Leica could do easily.

 

Sandy

 

I agree with Sandy's observations. I actually was able to test this with a pair of CV lenses that were visibly decentered (soft on one side) compared with copies that were optically fine, and the decentered copies had visibly stronger red edge effects. Initially, this led me to believe that the problem was basically a decentering problem, but it became clear that it was more than that.

 

I think that you are onto something with mis-registration of the RGGB Bayer array. This could explain why there is some inconsistency to this problem as lens decentering and sensor variation interact. However, if that's correct, a pure firmware fix will be a bit more complicated (although possible).

 

It will be interesting to see how this plays out. I would like to hear something directly from Leica rather than second-hand rumors.

 

Jeff

[Alnitak]

Advertisement (gone after registration)

Stefan Daniel says there will be a firmware to correct this issue in approximately 3-4 weeks.

 

As there is a lot of testing to do, obviously that date may stretch out a little.

 

We can all breathe a little easier and (at least those with Leica) don't have to sell all our wides

 

Has this been stated publicly somewhere?

 

Jeff

[scott kirkpatrick]

HI Scott,

I was unaware that the problem showed at all with the M8. My "white wall" is an Expo/Disc, a consistant light source and manual WB to eliminate tints. My widest lens is a 28 Elmarit Asph. My original tests were to check the corners and I might have written off any faint gradient as lighting inconsistancy. I am curious, so I'll check again. The histogram should be a peak and any other lumps would indicate problems, to look for. I'll test with and without the IR filter, too.

Bob

 

I posted some tests a month or two ago, using a 1960's Canon 19/3.5 in an LTM adapter. I've used it without noticing problems on my M8, but it would be even harder to use with the M9 than the CV15 has proven to be, since its rear element sticks even further back. So I shot a white wall, indoors, at a distance of about one foot with the lens set at infinity to throw things out of focus and maximize vignetting, both wide open and at f/8. With an M9 and then again with an M8. Sure enough the visibly awful pattern from the M9 would show some drifts that can easily be measured if you cropped the M9's image to its center 10 MPx. And that is what I saw on the M8. So I am not inclined to believe that there is anything wrong in the manufacture of sensors for either the M8 or the M9, although some variations are inevitable.

 

BTW, the ExpoDisc will vignette to a considerable extent, according to Sandy McG. When you use it for in-camera white balance, you also average over most of the image, so that's not a problem in practice. But it is not a perfect target to use for Cornerfix, just a good one.

 

Your 28/2.8 asph is a good test lens to use to see if the edge color problems are coming under control, since it has the rearmost exit pupil of all the lenses Leica currently supports.

 

scott

[sandymc]

FYI, blog post part 2 here: ChromaSoft: Will Leica fix the M9's red edges? Part 2

 

Sandy

[wlaidlaw]

Here are two images of a white wall. The first is taken with the WATE at 18mm/ f4. The lens detection was set to Manual - WATE 18mm. Red edges are quite apparent, with the usual phenomenon of more to the left. The second image is interesting and somewhat worrying. It is taken with a Zenitar SLR fish-eye lens on an M42 to Leica M adapter. Even though it is a considerably wider lens, red edges are less apparent, although what there is, is still more left. I am guessing this is due to its exit pupil being much further forward, since it is an SLR retro-focus lens. This would bear out Sandy and Jono's theory of light leakage/micro lens mis-alignment. I cannot see this being "magic-ed" away in firmware. If you remove it, surely you would get some image degradation as a result, given that the data from the sensor is in effect, corrupt. It is interesting that there is also less vignetting with the Zenitar, which indicates to me that the Kodak full frame sensor may be beyond its limits for an RF 18mm lens.

 

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!

[ArtZ]

Wilson,

 

I understand now why you bought the M9... to match the colour of your Porsche :D:D

 

Be naughty

[wlaidlaw]

One further thought as to a possible cause (and I am fully girded to be shot down in flames on this) is could the left red edges be due to a polarisation phenomenon in the micro-lenses. If say, they are laid down with a slight diagonal polarisation component for higher light frequencies, this could reduce the blue/green component (all natural light is polarised to some degree) to a larger extent for light coming from the centre of the lens to the far right of the sensor than from the centre to the left. This would leave more red light to the right of the sensor, thereby giving red left edges.

 

Wilson

[jonoslack]

Here are two images of a white wall. The first is taken with the WATE at 18mm/ f4. The lens detection was set to Manual - WATE 18mm.

 

Just a point here Wilson, and one that other WATE users should know:

As I understand it there is absolutely no point in using manual lens selection with the WATE - it uses the same correction for each focal length. You will do just as well (and possibly better) by leaving it on auto - the only downside is the exif will always show 16mm

As others here with the WATE - although I have (very rarely) been able to detect the effect, out of thousands of shots with it I've yet to see one which even needs a fix.

 

all the best

[wlaidlaw]

Just a point here Wilson, and one that other WATE users should know:

As I understand it there is absolutely no point in using manual lens selection with the WATE - it uses the same correction for each focal length. You will do just as well (and possibly better) by leaving it on auto - the only downside is the exif will always show 16mm

As others here with the WATE - although I have (very rarely) been able to detect the effect, out of thousands of shots with it I've yet to see one which even needs a fix.

 

all the best

 

Jono,

 

Thanks for confirming that. I wondered if it might be the case. However, I did not want anybody to be able to point the finger and say "non-relevant result, as you did not have the WATE set to 18mm". At first I thought in real life shots, I was not getting the red left edge with the WATE but when I printed up to large size and looked carefully, in virtually every shot where there was a light coloured left edge - there it was. More apparent in some lights than others.

 

Wilson

[markowich]

very interesting. i have no red-issue with the 18mm super-elmar (though i got to dislike the lens for its hard to correct distorsion), neither with the LUX 24. it does seem to depend on tolerances and there is sample variation.

btw, i have tried the CV 15mm M mount and it gave WILD red edges.

peter

 

 

doubt it probably f5.6

here is one at f4 (LR 2.6, adobe standard calibration)

Mind you I'm not sure how accurate it is to do this on the site as it looks a little different here as to when it's viewed in LR2

 

 

 

[ATTACH]188391[/ATTACH]

[Mauribix]

very interesting. i have no red-issue with the 18mm super-elmar (though i got to dislike the lens for its hard to correct distorsion), neither with the LUX 24. it does seem to depend on tolerances and there is sample variation.

btw, i have tried the CV 15mm M mount and it gave WILD red edges.

peter

 

Markowich, I'm not aware of any of those lenses (24lux 18SE) who doesn't show the red edges problem. It will be nice to see 2 of them behaving differently with an M9.

Probably in most situation it's hardly noticeable, but may try and shoot upwards against the sky on a sunny (or well light) day, I bet you may find the red edges with both lenses @ full aperture.

I've seen it with my 28cron @ f2 as well, and that's supposed to vignette far less than the 24lux at FA (if the red edges corner was related to vignetting).

[jonoslack]

Markowich, I'm not aware of any of those lenses (24lux 18SE) who doesn't show the red edges problem. It will be nice to see 2 of them behaving differently with an M9.

Probably in most situation it's hardly noticeable, but may try and shoot upwards against the sky on a sunny (or well light) day, I bet you may find the red edges with both lenses @ full aperture.

I've seen it with my 28cron @ f2 as well, and that's supposed to vignette far less than the 24lux at FA (if the red edges corner was related to vignetting).

 

HI There Maurizio - Peter

I don't think it relates to the degree of vignetting (or lack of) but the angle of incidence.

The new 21 and 24 lux lenses are actually relatively telecentric, and thus not very susceptible to the red edge problem. I agree that you may actually find it with any lens . . although (IMHO) sunny skies make it more difficult to see than cloudy ones!.

 

As I understand it:

 

Most obvious:

18 SE

21 Elmarit

24 Elmarit

28 Elmarit Asph

 

Slight

WATE

28 'cron

 

Negligible

24 'lux

21 'lux

24 Elmar (I think)

 

N.B. I don't have all these lenses, just in case you were wondering!

 

But I still think that this needs to be kept in context, and I've seen so few shots where it's an issue, and then it was really easy to fix.

Much more of an issue with Zeiss and CV lenses . . . where Sandy's wonderful cornerfix is your best friend.

[wlaidlaw]

The new 21 and 24 lux lenses are actually relatively telecentric, and thus not very susceptible to the red edge problem.

 

Jono,

 

Considering how far back the innermost element is and its relatively small diameter, to have achieved relative image telecentricity with these lenses, is a real feat of optical design. If you think that to get to this degree of telecentricity, the virtual exit pupil has to lie at near infinity in front of the lens, you would feel that the back edges of the optical cell tube would prevent light rays hitting the outer part of the sensor.

 

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!

[D&A]

Jono wrote: >>>"Just a point here Wilson, and one that other WATE users should know:

As I understand it there is absolutely no point in using manual lens selection with the WATE - it uses the same correction for each focal length. You will do just as well (and possibly better) by leaving it on auto - the only downside is the exif will always show 16mm"<<<

 

Jono, Your explanation above explains why when I use hand coded Zeiss 18mm (coded as a WATE) with lens detection "on" with the M9, it shows up as a 16mm in the EXIF data...yet when the same lens is used on a M8.2, (with lens detection on)...it defaults to 18mm upon turning on the camera.

 

I'm surpirse Leica didn't realize (or maybe they did and there was no way around it)..that EXIF data regarding recorded focal length, would be incorrect if the WATE was used at 18 or 21mm (with lens detection on).

 

I also suspect that with certain lenses, although decentering of individual samples may make it more or less likely to see left side rear edege...I believe like any observed phenominon, the right set of conditions (lighting, subject etc.) often has to be present for some to observe it...especially if they weren't specifically aware or looking for it while shooting.

 

Dave (D&A)

[Advertisement]

Hallo,

Look here: Leica M

[adan]

WIlson, the Summilux 21/24 lenses don't have to be perfectly (or even strongly) telecentric to do better than the f/2.8 lenses - just enough so to meet the sensor requirements and camera corrections halfway.

 

The glass bits in the back half of the 21 'lux, as shown in your x-section, are virtually the same diameter all the way through - a larger and more constant-diameter "tunnel" that is closer to the size of the M9 sensor than the 2.8's. And more like, e.g. the new Nikkor 24 f/1.4.

 

There is also that nice wide negative rear element/group, which likely acts as a collimator. In addition a negative element on the back of the lens would act as a teleconverter, moving the effective exit pupil forward. In fact, minus that rear group, the rest of the 21 lux may amount to something like a retrofocus 18mm f/1.3 (with rear element ahead of the mount flange), with a built-in 1.2x TC.

 

If you think it weird to design a lens with a built-in TC - take a look at the 105 f/2.8 Nikkor macro - essentially a double-gauss 6-element lens, with a 3-element "TC" group behind: Nikon AF Micro-Nikkor 105mm f/2.8D Lense