IR and digital sensors - one for the engineers

It appears that a thin cover over a sensor such as Leica's .5mm cover is best

for good reasons, but when IR qualities are considered, then thicker is better?

 

That doesn't seem correct to me. Cannot a manufacturer put a thin hot filter

(IR mirror) coating on the sensor's cover in order to have the better of both

features - thin filter plus IR exclusion? Since it is IR that is being

mirrored-out, there would seem to be no issue with internal re-reflections...

unless there is some mysterious (to me) action within the lens coating.

 

Or is it that hot-filters must be thick for some reason?

 

For manufacturers of new lenses, would it be prudent, or even possible, to put

such coating on the front of the first lens element instead? (Surely, that

would obviate the virtue of the M8's using earlier lenses, but I digress.)

The IR filter is a dichroic filter. This works on the basis of interference, and is made up of layers of thin films.

 

A little looking with Google reveals that the exact passband of a dichroic filter depends on the angle of incidence. As you get skew, not only does the center of the passband move, it also breaks up into two peaks.

 

As Leica had said for a long time, making sensors work with the high angles of incidence found in the Leica M is challenging. They made the filter very thin to reduce purple fringing with high angles of indicence. They are judged by their wide angle performance, and due to the crop factor folks would be using lenses that were even closer to the sensor.

 

But it looks like they (and Kodak) went a bit too far. I suspect that if they change the filter on the sensor, the tradeoff is going to have to be more purple fringing. Or, slight color shift in the corners due to the IR stop-band being a little different at high angles, and cutting into the visible light part of the spectrum.

 

So I suspect Leica and Kodak are researching some hard tradeoffs, ones that they messed up a bit the first time. The separate IR filter approach might well result in the best photographic results. But so many Leica users are absolute snobs about not using filters, that it "degrades" their perfect lenses.

 

However, if Leica does settle on the filter approach, they should be polite and offer versions of the filter to fit ALL of their lenses, including oddball filter mounts like Summitar, Summarit, E34, and of course A36. Maybe even 40mm for the Canon RF lenses. They shouldn't limit themselves to the modern sizes.

"I suspect that if they change the filter on the sensor, the tradeoff is going to have to be more purple fringing."

 

It always seems to come down to the colour purple! Maybe thats why this was the colour of choice for Royalty for many centuries.

From the comments, on axis does not seem to raise the purple or fringing issues as described. Is it possible to make a IR gradient filter more effective from 1/3 off axis to the corners. Are both or either purple color shift and fringing issues on axis to 1/3 out?

No, the thickness of the subtrate does not have to do with

the transmission of the IR light.

As John says, a lens coating (anti-reflection or in this case IR-reflecting) is build up from several dielectric layers (sub micron thickness) with different indices of refraction. By interference some light is reflected and other is transmitted (might have lower losses than without coating).

The characteristics do depend on the angle, but not very strongly.

Esp. in the M8 I expect the angle to be between perpedicular (=0 degrees) and 30 deg. it should not be a problem.

In my work (laser laboratory) we used standard laser-mirrors that were HR (high reflectors R> 99%)width a bandwith of 690nm - 950 nm. The angle of reflection was between 0 and 45 deg.

Somthing like that is not difficult to make.

Probably 99% reflection will surpress the IR enough.

(I have to mention that I didn't read everything about M8 problem,

and don't know exactly what it is)

Changing the filter, or lack of, on the sensor would have major ramifications on the existing firmware, and there is also a physical reason for having such a thin filter now. Read more on my blog - tedmphoto.blogspot.com

<b>Ted Smith</b> <i>Changing the filter, or lack of, on the sensor would have major ramifications on the existing firmware</i><p>

It would seem that updating firmware is easy enough to do that it is a nonproblem.<p>

And thank you Joop Mes for your explanation. It helps me find a place to begin studying in the literature. I trust your expertise.<p>

<b>Joel Matherson</b> <i>

It always seems to come down to the colour purple! Maybe thats why this was the colour of choice for Royalty for many centuries.</i><p>

:) Purple coloring was so very hard to create that at one time the people who created the pigments, were guilded tradesmen and they and the current politic decided who got the colors of purple. So, if you see an old painting with a great deal of a color Purple (there are many variants), or a dominance of a certain blue, you know the artist had some clout! <p>

And then there is the color Violet... but I'm drifting.

Changing the filter in the cover glass on the sensor should have minor to no software impact. The software doesn't talk to the sensor glass, it talks to the CCD chip under it.

 

I suspect a major issue with changing the sensor glass is whether the existing chips can be reworked. In any case, either Kodak or Leica has some inventory costs to eat if the chips that were already made have to be discarded. These are expensive chips, around $400 a pop.

 

I also cannot conceive of how adding an IR filter on the front of the lens can require the 6 bit coding. It's not like the coding will tell the software that there's an IR filter on there. I suspect that this is a communication muddle somewhere within Leica, aggravated by translation from German to English.

 

Keeping the IR light out of the CCD shouldn't require any firmware change. OK, maybe the reds would need to be slightly increased, but folks are used to fixing color balance in post-processing.

"The separate IR filter approach might well result in the best photographic results. But so many Leica users are absolute snobs about not using filters, that it "degrades" their perfect lenses."

 

From my observation there are at least as many people who keep a UV filter on all their lenses for protection. But even they acknowledge that there are specific instances where ghosting and flare are caused or increased by filters. In particular, where there are bright point sources of light just inside the shot, which the shade doesn't block. Such as shooting indoors where there are bright overhead spot or track lights, or outdoors at night where there are street lights. Two areas which, again coincidentally, seem to be popular subjects for "Leica Photography".

 

Additionally, even if we [hypothetically] agree to put IR blocking filters on all our lenses, the purple cast is only one of several unsettling faults in the M8's performance. There has been little talk about the others, the streaking/banding and ghosting. If those issues aren't solvable without a hardware rework of the sensor, then why wouldn't it be prudent to do over the sensor's IR coating at the same time?

"I also cannot conceive of how adding an IR filter on the front of the lens can require the 6 bit coding. It's not like the coding will tell the software that there's an IR filter on there."

 

No, and so far nobody has said they found the magenta cast different with different lenses. But if the "solution" is partly firmware it could be programmed such that the lens-coding-recognition function would have to be "ON" and sense a coded lens to apply its algorithms. That would definitely help sell more coded bayonets and thwart non-Leica lenses. But that's beyond the unthinkable because Leica would never do anything duplicitous...

Yes, the reflections inside the sensor are an interesting issue. Even mentioned in the Kodak press release about this sensor, how they tried to avoid them.

 

They obviously are a hardware artifact, and could only be addressed in hardware. They may just be a given for high angle of incidence. Unless the IR coating on the inside of the cover glass compromises the performance of the anti-reflection coatings?

 

I doubt that there's enough room between the sensor and the shutter to add a separate IR filter now, if that's the only solution.

 

Obviously these problems aren't inherent in the Kodak sensor cell design, since the same chip is used in the DMR. It's just the cover glass layers that's different on the M8 chip.

I don't think it is snobbish not to want to use filters. A 46mm Leica UV filter currently costs

around 100 bucks at B&H. The hot mirror filters are much more difficult to make, so they will

be more expensive. So if you have a few lenses, particularly ones of varying diameter, you are

going to have to pay a few hundred more dollars just to put filters on there. Add to that the

increase in flare and decrease in contrast associated with any filter, it just isn't an ideal

solution. If it is the only one, then people will have to make due, but I don't think you can

accuse people of being snobbish for not wanting to deal with it.

"If it is the only one, then people will have to make due"

 

More correctly, SOME people will ACCEPT them to make due. The question then becomes one of if there be enough of those people for Leica to sell enough M8s to at least break even on the M8 project, let alone make a profit.

Many harware issues can be fixed in software. Distortion for example. Chromatic aberrration for another example. This one is tougher but I can think of ways in which the effect could be mitigated by clever software.

 

Ther are also hardware fixes, though none of them simple and cheap at this point.

 

The bigger question is what Leica were thinking. They knew about these issues, they were reported to them by at least one "beta tester" and if they didn't find them themselves they're not very good at their job.

 

Did they think nobody would notice? Did they think nobody would care? Did they think that people would just buy the camera to put in a display case? Did they think they could make more money by selling a camera that most users would put up with even if it had a few problems than they could by holding up production for another 6 months and doing it right. I suspect the latter.

I'm guessing leica figured a few fussy people would notice, and even many of those would

make do.

 

I bet they had no clue, though, how quickly information is disseminated and propogates on

the web. There must

be more than a few leica managers who are syaing, "What the heck just happened..."

 

Bob, any chance you'll be testing one in the near future?

Angle of incidence for certain Leica lenses is much greater than 45 degrees. For example, the 21mm / f3.4 Super Angulon has a rear element that sits a few millimeters away from the film (or sensor). For this reason, it was unusable on the M5 (which had the metering stalk in front of the shutter) unless the rear element housing was modified (which it was for later production runs of this lens). I suspect that Leica wanted the M8 to be usable with as large a subset of existing lenses as possible, and to perform consistently with different lenses. Hence their choice of sensor IR filter. What I suspect is that the communication between engineering and marketing broke down in this instance: the (correct) engineering solution is to use the IR filter in front of the lens - but marketing didn't tell potential buyers about this. But it is still the correct solution...if your design brief is as given above.

Oops, modify the above: for the 21mm Super Angulon, assuming the nodal point in the center of the lens (symmetric design), the corners of full-frame film are at 45 degrees angle of incidence. So the M8 sensor corners are at <45 degrees, but still substantial.

<i>The hot mirror filters are much more difficult to make, so they will be more expensive.</i><p>

I think you should price them before saying that. They have been around for a very long time. They are not more expensive than the overpriced Leica filters on a size-by-size comparison. <p>

Proprietary fit might make them more expensive, however.

<i>Many harware issues can be fixed in software. Distortion for example. Chromatic aberrration for another example. This one is tougher but I can think of ways in which the effect could be mitigated by clever software. </i><p>

You would impress the heck out of all of us if you could find a way for the software to recognize objects which reflect an unacceptable amount of IR, then correct it. <p>

The RGB sensors can't tell what the object is in order to compensate the inappropriate luminosity (or even know what innapropriate means.)<p>

What kind of trick do you have in mind? AI? <p>

It's possible that the R, G and B sensors respond in a particular pattern to IR light. The trick would be to tell when that particular response pattern was due to IR. Not an easy task. We don't know if there's information in thr RAW data that might give a clue that excessive IR sensitivity was present. If it's due to the angle of incidence of the light on the dichroic filter, you'd expect the pattern to be dependant on the position in the frame, so you have another parameter to play with. Correlate certain patters in R, G, B and position in the frame and you might get some clues. You might mistakenly flag some real image with just the right(wrong) color and composition of course but it's possible you'd be right far more often than you'd be wrong - if you do it right.

 

I'm not employed by Leica as an optical imaging software engineer, so they'll have to figure it out for themselves. Maybe they can do it, maybe they can't.

 

However the proof that it can theoreticaly be done in software is self evident. You are software and you can tell when the problem exists and you can correct it to some extent in software (PhotoShop). All they need to do is make the software as smart as you are...

 

It's a kludge of course, and the right way to do it is in hardware. However hardware cost money and software kludges are cheap.

 

Of course they could always take the usual manufacturer's approach of kludging something for the M8 and fix it in the M9. There'd be even more incentive to upgrade and they'd have a year or two to come up with the solution if they get started now.

 

It's not likley Leica will let me get my hands on an M8 to test, so I won't likely be writing a review. However I'm sure someone in this forum is capable of doing that and probably will before too long.

The infrared image would be in focus on a different plane than the visible light image. Is there any way that the camera could differentiate between the in focus visible light image and the out of focus infrared image?

I ran the problem past my old friend, Professor R. Goldberg, who suggested that Leica install a beamsplitter in the M8 and add a film back, thus allowing simultaneous digital and analog capture. Proprietary software could then compare the (developed) film image with the digital file, and in this manner the camera would "know" what was intended by the photographer. He refused to elaborate further but said he would contact Leica with his idea.

 

As a temporary fix until his hardware/software solution could be realized, he suggested shooting everything by flash with an IR interference filter on the flash unit, thus removing the problem at the source. "The problem is with the light," he observed, "so get rid of it!"

 

I tried to point out to him that IR isn't light but his services are in demand and he was rushing off to consult on a secret government project.

<i>I tried to point out to him that IR isn't light but his services are in demand and he was rushing off to consult on a secret government project.</i><p>

For which government? :( <p>

Of course flash produces a lot of IR, and IR is light... but I realize you were in a fantasy. <p>

<i>However the proof that it can theoreticaly be done in software is self evident. You are software and you can tell when the problem exists and you can correct it to some extent in software (PhotoShop). <u>All they need to do is make the software as smart as you are... </u></i><p>

Well, THAT shouldn't be hard. :) <p>

But you struck something of an idea, Bob A. What if the camera incorporated a tiny IR beam beside the lens which quickly scanned the subject, noting position of images with particularly great luminescence in the IR band, then applied adjustment to the area based upon the patttern? Or at least noted it. Hey! The M8 could have a voice module warning, something like "Red Rum, Red Rum", or maybe "Danger Will Robinson!" But seriously, there may be something in the IR probe up front.<p>

IF - Leica is going to replace the internal filtering over the sensor, there WILL be a need for

a firmware upgrade, because a thicker sensor, or a different coating, will affect the various

optical effects that the firmware corrects for with coded lenses.

 

The streaking/banding affects are already likely a firmware problem, not a sensor problem

per se. A likely suspect is the timing of the shift-register that transports the "packets" of

voltage from each pixel out along the chip to the edge - in this case gates are being left

open at the wrong microsecond and allowing excess voltage to flow into other pixels

along each row, causing the bands.

 

The timing is controlled by a programmable timing chip, so the fix (if this is the problem)

will simply be to reprogram the timing chip (via new firmware) so that the gates open and

close at the right instant.

 

The IR contamination "purples" cannot be fixed purely by firmware, or any software, as

already mentioned. Leica knows they can fix the problem - the only question is whether to

stick with the idea of external filters or do a recall for internal changes. I suspect, again as

already mentioned, that they are figuring out how much they would have to compromise

optical quality for any given amount of increased internal filtration - if at all.

 

I personally am happily taking delivery of my M8 "as is" - hopefully within a week. I got to

borrow a Leica rep's M8 over the weekend for 18 hours, took 150 pictures - and the

various "issues" only showed up in shots where I tried to make them happen. The rest of

the time I just got beautiful color with little vignetting and amazing sharpness right to the

corners.

 

I expect any firmware fix will be user-installable via a download. And I will wait and see

what the final suggested IR-filtration fix is before worrying about it.

"I suspect, again as already mentioned, that they are figuring out how much they would have to compromise optical quality for any given amount of increased internal filtration - if at all."

 

The techs might be pondering that issue. The corporate honchos are more likely figuring how much they would have to compromise their stock bonuses by doing a sensor redesign, as opposed to the windfall profits they could make selling an IR filter and a new coded bayonet for every lens out there to be used on an M8. The decision will be a no-brainer for them if enough people aren't willing to deny themselves a few months of immediate gratification.