What happened to the Red Vignetting Thread

[marknorton]

Advertisement (gone after registration)

I was posting a reply and it had gone...

 

Earlier this morning, there was an interesting thread from Scott Kirkpatrick who had plotted red vignetting from two 28mm lenses against distance from the axis from white wall shots taken with an IR filter (presumably by Sean Reid).

 

Scott neatly side-stepped the issue of lens vignetting and overall in-camera vignetting correction by plotting 2*R/(B+G) which showed the residual red vignetting and the level of correction for it already built into the camera. The existing correction adjusted for the different characteristics of the two lenses to give essentially the same result and showed that the vignetting due to the filter is largely independent of aperture.

 

It was easy to see how a user selected "filter on" mode could apply additional correction to remove the effect of the filter. Scott's results showed the effect of red vignetting more clearly than a "Yup, there's cyan there" assessment.

 

An interesting thread, I'm wondering why it has gone

[andybarton]

It has been moderated (at the request of the graphs' owner) until the owner of the graphs has given his permission to the original poster for them to be published.

 

Apologies for the inconvenience.

 

My guess is that they will probably be back later today.

[scott kirkpatrick]

Mark, Sean and I have been exchanging ideas about how to analyze his pictures, even before I posted those curves. My profession is writing scientific papers, so that's what I did, posted a little scientific analysis. Sean's profession is showing things through pictures, but on his subscription site. He's not comfortable with my mode, and I think it puts us in a complicated position mixing public and proprietary information. What we are now thinking of doing is taking the analysis a bit further and posting it as an addendum on the RR site, which can then be discussed here.

 

I'll try to get this sorted out quickly. In the mean time you are welcome to pm or email me with any questions or to discuss. I do think that the conclusions -- what the Leica firmware does and how you might take advantage of it -- deserves public airing, just as the remarkable process in marking lenses to enable the firmware received a public airing here.

 

scott

[marknorton]

Well, I suppose we have to respect Sean's IP, but in the interests of us understanding what's going on, I like your analytical approach.

 

What was especially interesting from your results, Scott, was your suggestion that the existing cyan correction is there because of the red filtering of the sensor cover glass. It would be interesting to know if there is any correction applied with lens recognition turned off or whether the camera has to know the lens mounted to correct even for that.

 

When used with an IR filter, even the existing correction will "mop up" some of the additional cyan as Sean's results show.

 

It was interesting that the response to the magenta problem from Leica was so immediate - use filters and a firmware update. They obviously had all the necessary correction code in firmware already so it was "just" a case of re-calibrating the corrections when a filter is used.

 

The other open question is whether different makes of filter change the shape of those red vignetting curves...

[scott kirkpatrick]

Well, I suppose we have to respect Sean's IP, but in the interests of us understanding what's going on, I like your analytical approach.

 

Thanks

 

What was especially interesting from your results, Scott, was your suggestion that the existing cyan correction is there because of the red filtering of the sensor cover glass. It would be interesting to know if there is any correction applied with lens recognition turned off or whether the camera has to know the lens mounted to correct even for that.

 

We'll know that once Sean (or others, but his white wall is the perfect way to check this) shoots without an IR filter and with and without firmware enabled. Unfortunately, he has now coded up six or seven lenses and it's cloudy in Vermont, so I don't know when this will happen.

 

When used with an IR filter, even the existing correction will "mop up" some of the additional cyan as Sean's results show.

 

It was interesting that the response to the magenta problem from Leica was so immediate - use filters and a firmware update. They obviously had all the necessary correction code in firmware already so it was "just" a case of re-calibrating the corrections when a filter is used.

 

Even more important, they are writing a raw data transform that operates with the 14 bit data and produces the 8 bit sqrt information that goes into the .dng file. So they or their firmware partner or Capture One (or all three) have already written tools to extract all this behavior from shots of uniformly lighted white walls. And they have thought about what is the expected spectral distribution of red light in the scene so that they can choose a correction factor that will be good for most practical situations.

 

This is an interesting problem (as Joe Wisniewsky has already said). It requires things that all three partners have to contribute, before it is really understood. I can understand that it might have been properly appreciated only late in the product development cycle, and understood by executive management later still.

 

The other open question is whether different makes of filter change the shape of those red vignetting curves...

 

We'll see. I don't know, for example, if the B&W filters now available have an AR coating on their back sides, as Leica would probably insist in their versions to be available next year. The B&W catalog doesn't say. If you have one, could you look carefully at both sides to see?

 

regards,

 

scott

[rvaubel]

..................We'll see. I don't know, for example, if the B&W filters now available have an AR coating on their back sides, as Leica would probably insist in their versions to be available next year. The B&W catalog doesn't say. If you have one, could you look carefully at both sides to see?

 

regards,

 

scott

 

Scott

 

The Heliopan is AR coated on both sides. The IR cut frequency is very similar on both the Heliopan and the B+W so I don't think there is a preference there.

 

Rex

[gravastar]

Advertisement (gone after registration)

[snip]

We'll see. I don't know, for example, if the B&W filters now available have an AR coating on their back sides, as Leica would probably insist in their versions to be available next year. The B&W catalog doesn't say. If you have one, could you look carefully at both sides to see?

 

I've looked at 5 different size B+W 486 filters and the IR interference coating appears to be on one side only with the other side AR coated. 4 out of 5 have the IR filter on their backside and one has it on the front.

 

The way to tell is to carefully place the corner of a white sheet of paper against the filter and look at the magenta reflection. If the corner of the paper and it's magenta reflection touch then the IR filter is on the side nearest to you. If there is a faint reflection touching the corner and the strong magenta one is separated by the thickness of the glass then you are viewing from the anti reflection side.

 

There's probably an optimum orientation for the coatings to minimise IR refections between the filter and sensor surfaces. Without knowing the reflectivities involved it's hard to say which is best. The one filter I have with the coatings reversed could just be chance, there's not enough samples to form a conclusion.

 

Bob.

[scott kirkpatrick]

I've looked at 5 different size B+W 486 filters and the IR interference coating appears to be on one side only with the other side AR coated. 4 out of 5 have the IR filter on their backside and one has it on the front.

 

The way to tell is to carefully place the corner of a white sheet of paper against the filter and look at the magenta reflection. If the corner of the paper and it's magenta reflection touch then the IR filter is on the side nearest to you. If there is a faint reflection touching the corner and the strong magenta one is separated by the thickness of the glass then you are viewing from the anti reflection side.

 

 

That's interesting that you can see which side the IR cut layers are on. How do you confirm that there is an AR coating on the other side? Naively, you would think the best thing is to have the IR cut filter on the front surface, where it keeps the IR out, and just use the AR coating on the back to keep ghosts from happening. But maybe these are such complex multilayer filters that the glass interface isn't that important. I have two 486's on order. I'll check and post what they seem to have.

 

scott

 

scott

[gravastar]

........ How do you confirm that there is an AR coating on the other side? .........

 

The faint reflection you see is tinged brownish green and not nearly as bright as you would get from a plain glass surface.

 

ETA. I forgot to say you do the paper reflection test on both sides to confirm.

 

Bob.

[Guest KennyS]

Hey Scott maybe someone else can supply you with some files so you can do what you set out to do, share your knowledge for free on a free forum. I applaud your and others' time and efforts to help the rest of us getting the M8 to work the way a camera should be expected to, without trying to make a buck off of it.

[Guest KennyS]

I've looked at 5 different size B+W 486 filters and the IR interference coating appears to be on one side only with the other side AR coated. 4 out of 5 have the IR filter on their backside and one has it on the front.

 

Interesting! That's really great to hear because the B+W protective coating is very strong, and this way the IR coating is protected by being not exposed. One of my worries was I would need to put a regular UV over the IR in some conditions to keep the IR coating from getting scratched up. Those IR filters aren't exactly cheap throwaways! If I got some that had the IR coatings on the outside I'd just take them apart and turn the glass around.

[sdai]

Since the original thread was deleted so my reply evaporated as well ... but I'll just repeat - with due respect to everyone involved - these tests have to be done in a strictly controlled lab environment, otherwise the potential error in measurement and data sampling is plain out of your control. The data charted in Scott's 28 cron graph was clearly flawed and there's no way the amount of light falloff can be 0% till 25% away from the image circle center.

[marknorton]

And as I was trying to explain to you Simon when the thread was deleted, Scott cleverly compared the red level to the average of the green and blue levels - that what what was plotted on the Y-axis (2*R/(B+G)) and this neatly removes any change in level caused by lens vignetting or general sensor vignetting correction, leaving just the vignetting due to the change in red level caused by the filter.

 

If no red was being filtered at all, those lines should have been straight lines across the top of the graph, irrespective of the lens vignetting characteristics.

 

You said in your original post that you didn't get it. Do you now?

[rvaubel]

And as I was trying to explain to you Simon when the thread was deleted, Scott cleverly compared the red level to the average of the green and blue levels - that what what was plotted on the Y-axis (2*R/(B+G)) and this neatly removes any change in level caused by lens vignetting or general sensor vignetting correction, leaving just the vignetting due to the change in red level caused by the filter.

 

If no red was being filtered at all, those lines should have been straight lines across the top of the graph, irrespective of the lens vignetting characteristics.

 

You said in your original post that you didn't get it. Do you now?

 

Mark

 

I didm't get it at first either. But after reading your explanation and studying the charts, I did.

 

Now, my question is, how is anyone else going to be able to follow this important discussion if we don't have the charts anymore? When and how are they going to be reposted?

 

Rex

[sdai]

Thanks a lot for your explanation, Mark ... I did understand (although not fully) what Scott was charting in his graph.

 

When I said "I don't get it" ... it was because I did't see any obvious change comparing Scott's charts to Leica's factory charts in terms of vignetting characteristics ... so what the filters, Leica's 6-bit coding, firmware, etc can do is very doubtful. And in this particular case with the 28 Elmarit ASPH, stopping down from f/2.8 to f/8, the change in the amount of light falloff is so small, I suspect any effort in vignetting correction would be meaningful.

[marknorton]

Sorry Simon, you are still not "on message".

 

Scott's charts and Leica's charts are not showing the same thing. Leica's charts show the overall lens vignetting which applies to all colours of the lens itself. Scott's charts show the additional vignetting of red due to the filter being on the front, expressed as a fraction of the blue-green average. The level of all three colours changes due to the overall vignetting, but the red suffers additional vignetting due to the filter and by plotting (2*R/(G+), the graph ignored all vignetting which applies to all three colours equally.

 

Those graphs showed, let's say, a drop from 1.0 (if there was no red filtering) to 0.95 and they also showed how the existing cyan correction - implemented according to Scott to correct for filtering in the sensor - increases the level to (let's say) 0.97. The additional drop in red level due to the filter can be corrected by changing the parameters the correction works with.

 

I made the point that the reason Leica were so quick to come up with the filters+firmware solution after the magenta problem was reported was because the code to do the correction was already in the camera so all they had to do was to revise it by measuring the filter on each coded lens. Not the work of a moment but I expect they have it down to fine art - take a few pictures and let a computer program analyse the resulting images.

[sdai]

I believe now I'm closer to be "on message", Mark ... so you were specifically talking about firmware correction to the added "benefit" of using filters (especially with regards to RED), right? Thanks again for the help in understanding.

[scott kirkpatrick]

Simon, thanks for reminding me to check the factory vignetting charts on these lenses. They are computed, not measured (Leica predicts, Zeiss measures, as I recall). And they talk about illumination. Even with film, how illumination at an angle translates into developed film density is tricky -- it depends on the type of film. With silicon photocells, whether ccd or Si, there's a different angle-dependence of quantum efficiency (ratio of photons in to electrons captured). Finally, the Kodak 10500 has offset microlenses, and that should reduce the effect of vignetting.

 

The interesting part of the factory charts is their aperture dependence. In the simplest approximation, there should be none, but a modern lens is very long and sometimes you can't see the whole aperture from the edges of the frame when it is wide open. See how the Summicron vignettes badly at f/2.0, and vignettes in the outer half of the frame at 2.8. Both lenses seem to vignette about the same at f/5.6 and above -40% at 15mm radius, which is about the edge of the frame in the M8. But the Summicron drops half again as much when it's wide open. Sean's shots at f/2.8 should show some of that. I'll take a look, now that we seem to be able to separate the overall vignetting from the effects on red.

 

Mark, thanks for explaining that I was plotting a ratio, not the actual fall-offs.

 

scott

[rvaubel]

This thread reminds me of a story bu Stanislous Lem "A Perfect Vacuum" . The charts are gone guys!!

 

Oh well. Carry one...at least I was one of the 4 people that say them.

 

Rex:rolleyes:

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[ho_co]

Simon--

Thanks for continuing your questiions till I began to get a handle on the matter!

 

Mark--

Thanks for explaining till I started to get it!

 

Scott--

Thanks for the posting.

 

It is awesome (in the true meaning of the word) to me how much data, knowledge and design go into an instrument like the M8!

 

And having you guys there to explain it to the rest of us is extremely helpful!

 

(I'm sure you're all also aware of the 32-page PDF of KAF-10500 specifications that Kodak makes available via their website, http://www.kodak.com/US/en/dpq/site/SENSORS/name/KAF-10500_product. Some on the forum have implied that Leica must have been unaware of the consequences of using this chip in regard to IR, but the data are all right there.)

 

Scott--good effort. Hope the graphs return soon!

 

--HC